IMP logo
IMP Manual  for IMP version 2.12.0
ChangeLog.md
1 ChangeLog {#changelog}
2 =========
3 
4 # 2.12.0 - 2019-12-06 # {#changelog_2_12_0}
5 - The Windows .exe installer now supports Python 3.8, and has dropped support
6  for Python 3.4.
7 - Packages are no longer provided for Ubuntu 14.04 (Trusty Tahr) since
8  it has reached end of life.
9 - The IMP distribution now includes the IMP::bayesianem module developed at
10  Institut Pasteur, Paris, for Bayesian handling of cryo-electron microscopy
11  density map data. See Bonomi et al. at https://doi.org/10.1101/113951 for
12  more information.
13 - The old IMP::pmi::representation::Representation class has been removed
14  from IMP.pmi. New applications should use IMP::pmi::topology::System instead.
15 - The IMP::pmi::restraints::crosslinking::ISDCrossLinkMS class for handling
16  crosslinking has been removed. Use
18  instead.
19 - The `rg` tool (part of the IMP::saxs module, used to compute radius of
20  gyration from a SAXS profile) is now called `compute_rg` for consistency
21  with other SAXS tools and to avoid conflicts with other packages.
22 - We now provide RPM packages for RedHat Enterprise Linux 8 (or compatible
23  operating systems such as CentOS 8).
24 - The RPM packages for RedHat Enterprise Linux 8 and for Fedora now use
25  Python 3 by default. If you need Python 2, install the IMP-python2 package
26  as well.
27 - IMP::algebra::Rotation3D::get_derivative(),
28  IMP::algebra::Rotation3D::get_gradient(),
29  IMP::algebra::get_gradient_of_composed_with_respect_to_first(), and
30  IMP::algebra::get_gradient_of_composed_with_respect_to_second() have been
31  deprecated and are superseded by
32  IMP::algebra::Rotation3D::get_gradient_of_rotated(),
33  IMP::algebra::Rotation3D::get_jacobian_of_rotated(),
34  IMP::algebra::get_jacobian_of_composed_wrt_first(), and
35  IMP::algebra::get_jacobian_of_composed_wrt_second(), respectively. By
36  default, the derivatives are now computed with respect to the unnormalized
37  quaternion and do not include the normalization operation.
38 - New methods are added to compute adjoint derivatives (reverse-mode
39  sensitivities) for compositions and actions of IMP::algebra::Rotation3D and
40  IMP::algebra::Transformation3D upon 3D vectors.
41 - Fixed a bug in nested rigid body derivative accumulation, where derivatives
42  with respect to quaternions were incorrectly projected to be orthogonal to
43  the quaternion.
44 - Reimplemented rigid body derivative accumulation to use the new adjoint
45  methods. The many-argument versions of
46  IMP::core::RigidBody::add_to_derivatives(),
47  IMP::core::RigidBody::add_to_rotational_derivatives(), and
48  IMP::core::NonRigidMember::add_to_internal_rotational_derivatives(), which
49  previously pulled adjoints from member global reference frame to member
50  local reference frame and parent global reference frame are now deprecated.
51  Pullback functionality is now handled by
52  IMP::core::RigidBody::pull_back_members_adjoints().
53 - IMP::isd::Weight is now constrained to the unit simplex, and methods were
54  added for adding to its derivatives. IMP::isd::Weight::add_weight() no longer
55  resets all the weights to the barycenter of the unit simplex and instead
56  initializes the new weight to 0. IMP::isd::Weight::get_number_of_states()
57  and IMP::isd::Weight::get_nstates_key() were deprecated and superseded by
58  IMP::isd::Weight::get_number_of_weights() and
59  IMP::isd::Weight::get_number_of_weights_key(), respectively.
60 
61 # 2.11.1 - 2019-07-18 # {#changelog_2_11_1}
62 - Bugfix: fix build system failures with CMake 3.12 and 3.13, and on Windows.
63 - Bugfix: IMP::atom::create_clone() now always copies mass, even of particles
64  that are not atoms.
65 - Bugfix: IMP::rmf::create_hierarchies() should no longer fail with
66  "The filename cannot be empty".
67 
68 # 2.11.0 - 2019-06-25 # {#changelog_2_11_0}
69 - A number of new tutorials are now available covering a variety of topics
70  from modeling complexes with IMP::pmi to writing new C++ code. See
71  the [tutorial index](https://integrativemodeling.org/tutorials/) for
72  more details.
73 - The build system has been updated to use new CMake functionality
74  (if available) to detect Python. If you have CMake 3.14 or later, this should
75  do a much better job of handling multiple Python installations (e.g. Macs
76  with Homebrew). It will default to using Python 3. To force use of Python 2,
77  pass `-DUSE_PYTHON2=on` to your CMake invocation.
78 - CMake files are now provided to make it straightforward to use IMP as
79  a library in other programs. See
80  [the tutorial](https://integrativemodeling.org/tutorials/using_cpp/)
81  for more details.
82 - CMake files are now provided to build new IMP modules outside of the IMP
83  build tree (i.e. a module can be built against an IMP binary).
84 - The IMP::OptionParser class has been removed. Use IMP::ArgumentParser instead.
85 - The IMP::test::TempDir and IMP::test::RunInTempDir utility classes have been
86  removed. Use the IMP::test::temporary_directory and
87  IMP::test::temporary_working_directory context managers instead.
88 
89 # 2.10.1 - 2019-02-26 # {#changelog_2_10_1}
90 - Add support for OpenCV 4.
91 - Fix IMP::isd `create_gmm.py` script to handle command line options correctly.
92 - Command line tools in the Mac and Ubuntu packages should now use system
93  Python (/usr/bin/python), not the first Python (e.g. Anaconda Python)
94  found in PATH (which might not be compatible with IMP's Python libraries).
95 
96 # 2.10.0 - 2019-01-08 # {#changelog_2_10_0}
97 - Applications using IMP::pmi should now use the IMP::pmi::topology::System
98  class to set up the system (sometimes referred to as "PMI2"), rather than
99  the older IMP::pmi::representation::Representation class ("PMI1").
100  This new class is faster and produces standard IMP Hierarchy objects. The
101  old PMI1-style classes are deprecated in this release and will be removed
102  in the next one. For support of published studies using "PMI1", a legacy
103  IMP::pmi1 module is provided (to use it, simply replace all imports of
104  "IMP.pmi" with "IMP.pmi1"). This module will receive only minor bug fixes,
105  and no new features.
106 - All IMP Python command line tools now use
107  [argparse](https://docs.python.org/2/library/argparse.html) rather than
108  [optparse](https://docs.python.org/2/library/optparse.html) for handling
109  options, as optparse is deprecated in modern Python versions. The
110  IMP::OptionParser class, which extends optparse to add standard IMP options,
111  is now deprecated. Use the similar IMP::ArgumentParser class instead.
112 - The bundled copy of [python-ihm](https://github.com/ihmwg/python-ihm) is
113  updated to the latest release. This supports faster (C-accelerated) reading
114  of mmCIF files, read and write support for BinaryCIF files, and support
115  for output of PMI2-style models. The new IMP::atom::read_mmcif function
116  allows reading a PDB file in mmCIF format, similarly to the existing
117  IMP::atom::read_pdb function.
118 - A new application of IMP is now available:
119  - [Architecture of Pol II(G) and molecular mechanism of transcription regulation by Gdown1](https://salilab.org/pol_ii_g/)
120 - The deprecated IMP::isd::FNormal::density() method has been removed.
121  Use IMP::isd::FNormal::get_density() instead.
122 - Handling of quaternions (for example in the calculation of rotational
123  derivatives) is now faster and more numerically stable.
124 - IMP::mpi now includes a workaround for an OpenMPI failure when used
125  with Python, so previous hacks (building OpenMPI with the
126  `--disable-dlopen` flag, or using `sys.setdlopenflags` in Python)
127  should no longer be necessary.
128 
129 # 2.9.0 - 2018-07-02 # {#changelog_2_9_0}
130 - [Eigen](https://eigen.tuxfamily.org/) is now required to build IMP.
131 - The Windows .exe IMP installer no longer supports Python 2.6. Python 2.7,
132  3.4, 3.5, and 3.6 are supported by this installer.
133 - A binary package is now provided for Ubuntu 18.04 LTS, Bionic Beaver.
134 - The IMP::atom::setup_as_rigid_body() function has been removed.
135  Use IMP::atom::create_rigid_body() instead.
136 - Constructors for IMP::container classes, IMP::core::BallMover,
137  IMP::core::RigidBodyMover, and IMP::core::IncrementalScoringFunction taking
138  a RigidBody or Particles have been removed. Use constructors taking
139  ParticleIndexes instead.
140 - The IMP::core::ConnectivityRestraint::get_connected_pairs() method has been
141  removed. Use get_connected_index_pairs() instead.
142 - The stats() profiling method has been removed from
143  IMP::isd::GaussianProcessInterpolationRestraint and
144  IMP::isd::MultivariateFNormalSufficient. Use a third party package (such as
145  gperftools) for profiling instead.
146 - Scores in IMP.saxs (and FoXS) now use the chi-square value rather than chi.
147 - New applications of IMP are now available:
148  - [Modeling of the entire yeast nuclear pore complex](https://salilab.org/npc2018), as well as
149  [modeling of the flexible FG regions](https://salilab.org/npc_fg_2018) and
150  modeling of the individual [Nup133](https://salilab.org/nup133),
151  [Nup82](https://salilab.org/nup82), and
152  [Pom152](https://salilab.org/pom152) subcomplexes.
153  - [Modeling of the Ecm29-proteasome complex](https://salilab.org/ecm29)
154 - The new modules IMP::npc and IMP::npctransport provide specialized
155  restraints and other functionality for modeling static nuclear pore complex
156  structures and transport through these structures, respectively, although
157  may also be useful for other modeling applications.
158 - IMP's mmCIF support for deposition of models in
159  [PDB-Dev](https://pdb-dev.wwpdb.org/), in the IMP::pmi and IMP::mmcif modules,
160  now utilizes the [python-ihm](https://github.com/ihmwg/python-ihm) library,
161  a copy of which is included with IMP.
162 - RMF files now store additional metadata, such as the primary sequence of
163  each chain, software used, information on each restraint (such as fit values),
164  and provenance (classes derived from IMP::core::Provenance) that describes
165  transformations of the system, such as initial input from a PDB file,
166  sampling, clustering, and filtering.
167 - The new experimental IMP::mmcif module allows generating basic mmCIF files
168  directly from an IMP::Model and/or RMF files.
169 
170 # 2.8.0 - 2017-08-16 # {#changelog_2_8_0}
171 - New applications of IMP are now available:
172  - [Modeling of the yeast Spindle Pole Body core](https://salilab.org/spb/)
173  - [Modeling of the SEA complex, a modulator of the TORC1 pathway](https://salilab.org/sea/)
174 - This will be the last release that includes packages for Ubuntu 12.04
175  (Precise Pangolin) since it has reached end of life. We will continue to
176  provide packages for newer Ubuntu LTS releases.
177 - The Anaconda and Windows .exe packages are now MPI-enabled. On Windows, in
178  order to use MPI, install [MS-MPI](https://msdn.microsoft.com/en-us/library/bb524831.aspx).
179 - IMP::em2d::PCAFitRestraint now uses binary PGM files as output, rather than
180  the less well supported text PGM format. (It can still read both text and
181  binary PGM files as input.)
182 - IMP::em::FitRestraint::cast() is deprecated. Use `get_from()` instead.
183 - The deprecated profiling (timing statistics) support has been removed.
184  A number of third party packages (such as gperftools) are available
185  for profiling and should be used instead.
186 - The new module IMP::spb provides specialized restraints, samplers, and
187  other objects to support modeling of the Spindle Pole Body (SPB). While
188  designed specifically for this system, they may be useful for modeling
189  similar systems.
190 
191 # 2.7.0 - 2017-04-13 # {#changelog_2_7_0}
192 - New applications of IMP are now available:
193  - [Modeling of the yeast Mediator complex](https://salilab.org/mediator/)
194  - [Modeling of the yeast and human TFIIH complex](https://salilab.org/tfiih/)
195  - [Modeling of complexes between the 26S proteasome and proteasome interacting proteins](https://salilab.org/26S-PIPs)
196  - [Human Complement modeling](https://salilab.org/Complement)
197  - [Modeling of a region of the fly genome](https://integrativemodeling.org/systems/?sys=22) using IMP and [TADbit](http://sgt.cnag.cat/3dg/tadbit/)
198 - Memory and CPU performance improvements across the code, particularly
199  in IMP::pmi.
200 - Preview of mmCIF support for deposition of integrative models in
201  [PDB-Dev](https://pdb-dev.wwpdb.org/). See the IMP::pmi::mmcif
202  module for more details.
203 - This will be the last release that includes packages for RedHat Enterprise
204  Linux 5 (and variants, such as CentOS 5), since RHEL 5 has reached end
205  of life.
206 - The IMP::em::SurfaceShellDensityMap::binaries() method has been replaced
207  with the (correctly spelled) IMP::em::SurfaceShellDensityMap::binarize()
208  method.
209 - IMP::multifit::WeightedExcludedVolumeRestraint no longer expects that the
210  IMP::core::RigidBody particles it acts on are also IMP::atom::Hierarchy
211  objects. Use an IMP::multifit::RigidLeavesRefiner in place of
212  IMP::core::LeavesRefiner to get the old behavior.
213 - IMP::modeller::MultipleBinormalRestraint has moved to the IMP::core
214  module.
215 - The IMP::core::setup_as_rigid_body() method has been removed.
216  Use IMP::core::create_rigid_body() instead.
217 - IMP::test:TempDir and IMP::test::RunInTempDir are deprecated in favor
218  of similar functions that work as Python context managers
219  (IMP::test::temporary_directory() and IMP::test::temporary_working_directory()
220  respectively).
221 
222 # 2.6.2 - 2016-05-25 # {#changelog_2_6_2}
223 - Add support for SWIG 3.0.8.
224 - Add support for Ubuntu 16.04, Xenial Xerus.
225 - Minor documentation fixes.
226 - Bugfix: fix detection of termini in IMP::atom::Selection.
227 
228 # 2.6.1 - 2016-04-14 # {#changelog_2_6_1}
229 - The 'cluster' binary is now called 'rmsd_cluster' to avoid conflicts with
230  other packages (e.g. graphviz also provides a 'cluster' binary).
231 - Minor bug fixes to PMI.
232 - Minor documentation fixes.
233 
234 # 2.6.0 - 2016-03-28 # {#changelog_2_6_0}
235 - This release includes most of the functionality of PMI2. This is an improved
236  version of PMI that is faster than PMI1, more modular, and easier to use.
237  See IMP::pmi for more information.
238 - A new application of IMP is now available:
239  [Modeling of exosome complexes from cross-link MS data](http://salilab.org/exosome).
240 - The IMP::base and IMP::kernel namespaces have been removed (use the top-level
241  IMP namespace instead).
242 - The IMP::display::ChimeraWriter class has been removed (use IMP::rmf instead).
243 - IMP::Model no longer has an implicit scoring function. This means that
244  Model::evaluate() and Model::add_restraint() have been removed (use an
245  IMP::ScoringFunction instead). IMP::Optimizer and IMP::Sampler now also
246  require an IMP::ScoringFunction to be explicitly set before they are used.
247 - IMP::atom::Hierarchy::get_as_xxx() methods are deprecated; use regular
248  decorators instead. For example, instead of h.get_as_residue() use
249  Residue(h). If you want to check for validity first,
250  use Residue::get_is_setup(h).
251 - IMP::atom::Selection can now select by element.
252 - IMP::modeller::load_restraints_file() and IMP::modeller::read_pdb() have
253  been removed. Use the IMP::modeller::ModelLoader class instead.
254 - IMP::atom::PDBSelector subclasses are more flexible, with a new
255  IMP::atom::XorPDBSelector class, and can be combined using standard
256  Python operators, in similar fashion to sets (&, |, ^, -, ~ operators).
257 - The following modules that use FFTW (which is licensed under the GPL)
258  are themselves now also licensed under the GPL: IMP::EMageFit,
259  IMP::cnmultifit, IMP::em2d, IMP::integrative_docking, and IMP::multifit.
260 - Bugfix: sampled EM density maps should now handle resolution correctly; maps
261  generated by the fixed code agree well with those generated by EMAN. The old
262  radius-dependent kernel parameter code is no longer used.
263 
264 # 2.5.0 - 2015-06-30 # {#changelog_2_5_0}
265 - Documentation has been revised and updated; the main manual now gives a
266  gentler introduction to IMP, with more technical details on the code
267  provided in a separate reference guide.
268 - The IMP::base and IMP::kernel namespaces are now deprecated. Functionality
269  in both base and kernel now lives in the top-level "IMP" namespace.
270 - IMP::atom::Selection::get_selected_particle_indexes() has a new
271  'with_representation' parameter. If turned on (the default), Selection
272  returns all representational (XYZ) particles that match the request -
273  for example, selecting a residue will typically return all of its atoms.
274  If turned off, Selection returns the highest matching particle instead -
275  so, in this case, the residue particle itself.
276 - All IMP::RAII objects (e.g. IMP::SetLogState) now support the context
277  manager protocol in Python (e.g. "with IMP.SetLogState() as s:")
278 - Several long-deprecated functions have been removed. These include:
279  - Several Restraint constructors that take Particles. Use equivalent
280  constructors that take an IMP::Model pointer and ParticleIndexes instead.
281  - IMP::core::MonteCarlo::get_number_of_forward_steps() (use
282  get_number_of_accepted_steps() instead).
283  - IMP::Model::remove_restraint(), IMP::Model::get_restraints(),
284  IMP::Model::get_number_of_restraints(), IMP::Model::get_restraints().
285  Store your restraints in an IMP::RestraintSet or use an
286  IMP::core::RestraintsScoringFunction. Related methods such as
287  IMP::Model::evaluate() will be removed in the next IMP release.
288  - IMP::Model::set_maximum_score() and related methods. Set the maximum
289  scores on the IMP::Restraint or IMP::RestraintSet directly, using
290  IMP::Restraint::set_maximum_score().
291  - IMP::Restraint::get_input_containers() and
292  IMP::Restraint::get_input_particles(). Use IMP::Restraint::get_inputs()
293  instead (similarly for get_outputs()).
294  - IMP::rmf::save_frame() no longer takes a frame index argument (it was
295  previously ignored anyway).
296 - Several functions have been deprecated in this release (and will be removed
297  in the next). These include several Restraint constructors that take Particle
298  or Particles objects (use constructors that take an IMP::Model and
299  ParticleIndex or ParticleIndexes instead); and add_particle*(),
300  set_particle*() and clear_particle*() in several Container objects (use add(),
301  set() and clear() with particle indexes instead).
302 - On platforms that support it, OpenMP is turned on by default. To speed up
303  some parts of IMP, call IMP::base::set_number_of_threads(). By default,
304  only a single thread is used.
305 - The RPM packages now include an -mpich subpackage that can be installed
306  to enable parallel computation, using the IMP.mpi module in concert
307  with the mpich library.
308 - We now provide packages for both of the currently-supported Ubuntu LTS
309  releases: Ubuntu 12.04 (Precise Pangolin) and Ubuntu 14.04 (Trusty Tahr).
310 - The default log level has changed from NONE to WARNING.
311 - A new command line tool, `multi_foxs`, allows for enumeration and scoring
312  of multi-state models against SAXS profiles.
313 - A new command line tool, `ligand_score_multiple`, acts similarly to the
314  existing `ligand_score` tool but scores each ligand pose generated by
315  docking software (such as PatchDock).
316 - New applications of IMP are now available:
317  - [Multi-state modeling of the histidine kinase, PhoQ](http://salilab.org/phoq).
318  - [Determination of the molecular architecture of the 40Sâ˘eIF1•eIF3 translation initiation complex](http://salilab.org/40S-eIF1-eIF3). - IMP::domino::DiscreteSampler::set_restraints() now allows the set of restraints used in DOMINO filtering to be controlled (previously, all restraints in the IMP::Model were used). - IMP::modeller::IMPRestraints now transfers an IMP::ScoringFunction into the MODELLER score, not the implicit (and deprecated) score of the entire IMP::Model. - Bugfix #775: coordinates of voxels in EM maps were previously inconsistently treated (they are supposed to be the centers of the voxels, but in some cases they were the lower left edge). The fix may result in maps and their bounding boxes appearing shifted by half the voxel size. # 2.4.0 - 2015-02-27 # {#changelog_2_4_0} - Python 3 is now supported. The Windows package now works out of the box with Python 3.3 or 3.4; a `-python3` subpackage is provided for Ubuntu and Fedora. (Apple does not yet provide Python 3 so the Mac package is not built with Python 3 support, but if you install Python 3 you can compile IMP from source to work with it.) - A new command line tool, `rrt_sample`, allows for running rapidly exploring random tree (RRT) based sampling, e.g. to sample conformations of multiple domains connected by flexible linkers. - Command line tools are now placed in the `bin` directory of modules (the old `bin` directory, used for utility programs that are not installed, is now called `utility`). The `applications` directory is removed. - Several Python modules that were only used by EMageFit have been moved from IMP::em2d to IMP::EMageFit. - A new tutorial that uses the IMP::pmi module is now available (modeling of the RNA Polymerase II stalk). # 2.3.1 - 2014-12-01 # {#changelog_2_3_1} - Minor bug fixes. - Update to IMP::pmi to support the Nup84 and rnapolii biological systems. # 2.3.0 - 2014-10-24 # {#changelog_2_3_0} - A new IMP::pmi module was added, containing high-level Python classes to simplify the representation, scoring, sampling and analyzing of models in IMP. This module is still under heavy development, and should be considered experimental. - Binary installers for 64-bit Windows are now available. - The `.py` extension has been removed from all Python applications, so that they are consistent with the C++ applications. - IMP::atom::Selection objects can now be combined using basic set operations (union, intersection, difference, symmetric difference). In Python the equivalent operators (|, &, -, ^ respectively) can be used. - The integrative docking (idock) application now employs SOAP to improve the scoring. - IMP::atom::CHARMMParameters::create_topology() now works with a wider range of hierarchies (previously it would only accept Residues that were direct children of Chains). For example, it should do the right thing when given a Hierarchy that contains no Chain particles, or one where Residues are grouped under Fragments. - Only Python 2.6 or later is now supported. We still provide IMP for RedHat Enterprise 5 (which ships with Python 2.4) but you will need to use the `python26` binary (part of EPEL) rather than the system default Python. # 2.2.1 - 2014-08-11 # {#changelog_2_2_1} - Bugfix: PoseScore (see the ligand_score application) should now match that used by the Pose&Rank web server (http://salilab.org/poseandrank/) # 2.2.0 - 2014-04-09 # {#changelog_2_2_0} - A module IMP.template was added to act as a template for creating new external modules. It includes setting up git hooks for format checks and things like that. - The Python support for grids was cleaned up. Hopefully this makes it more robust. If you were using a grid in Python that no longer exists, ping us, we can add it back. - The generic geometry functions were simplified slightly and renamed to IMP::algebra::get_vector_geometry(). - The IMP::base::Flag class was added to simplify adding flags in C++. - The `--log_level` and `--check_level` command line argument now take strings (eg VERBOSE) to specify their values, rather than cryptic (and a bit unstable) numbers. - Command line options are now divided into normal and advanced ones. You do `--help_advanced` to show the advanced ones. - Support for computing Connolly surfaces was added with IMP::algebra::get_connolly_surface(). In addition, there is now IMP::algebra::get_uniform_surface_cover(const IMP::algebra::Sphere3Ds&,double) to sample the surface of a collection of balls. - IMP::atom::reate_simplified_from_volume() was added to create a simplified representation of a hierarchy that preserves the surface are and occupied volume. It uses IMP::algebra::get_simplified_from_volume(). If you want to play with such representations, you can use the rmf_simply application. - IMP::atom::get_rmsd(IMP::atom::Selection, IMP::atom::Selection, const IMP::algebra::Transformation3D&) was deprecated. Use IMP::atom::get_rmsd_transforming_first() instead. - IMP::algebra::get_rmsd() and IMP::algebra::get_rmsd_transforming_first() were added. - The family of macros IMP_BASE_[SMALL,LARGE]_[UNORDERED,ORDERED]_[MAP,SET] have been added (eg IMP_BASE_SMALL_ORDERED_SET ()). They provide platform and boost-version independent ways to take advantage of the best set/map for an application. In particular, they use boost::container::flat_set and kin when available. - IMP::atom::State and IMP::atom::Representation decorators have been added to provide support for multiple states and multiple representation schemes. They are integrated with RMFs and IMP::atom::Selection. See atom/multiresolution.py and atom/multistate.py for examples. - The container implementations has been cleaned up and simplified a bit. This should not change most usage. - [RMF](http://salilab.github.io/rmf) has been updated. There are a few minor API changes, in addition to a new, faster format (`.rmf3`). - [Eigen](#Eigen) is now included with IMP. - The deprecated function IMP::kernel::Optimizer::set_restraints() was removed. Use IMP::kernel::Optimizer::set_scoring_function() instead. - The deprecated constructor for IMP::kernel::ModelObject that does not take an IMP::kernel::Model was removed. This propagates to constructors IMP::kernel::Restraint, IMP::kernel::ScoreState and kin. Make sure you pass an IMP::kernel::Model as the first constructor argument. Related to this, deprecated functions like get_is_part_of_model(), set_model() were removed as they don't serve any purpose. - The deprecated methods in IMP::kernel::Optimizer that pertain to attribute optimization were removed. Inherit from IMP::kernel::AttributeOptimizer instead if you want to use them in your optimizer. - Some deprecated functionality was removed. If you encounter problems, replace IMP::base::OwnerPointer with IMP::base::PointerMember, find logging stuff in IMP/base/log.h, object in IMP/base/Object.h and random in IMP/base/random.h. - IMP::atom::Chain now uses strings for the chain ids, to be consistent with changes in the PDB. PDB files are written using the first character of the chain string. - Added orientation-dependent SOAP scoring (IMP::score_functor::OrientedSoap and IMP::atom::OrientedSoapPairScore). Orientation-dependent SOAP scores include those that score loops (SOAP-Loop), protein-peptide interfaces (SOAP-Peptide) and proteins (SOAP-Protein). The library files themselves are rather large (~1.5GB) and so are not included with IMP. They can be downloaded separately from http://salilab.org/SOAP/. # 2.1.1 - 2013-11-04 # {#changelog_2_1_1} - Added support for Mac OS 10.9 # 2.1.0 - 2013-10-24 # {#changelog_2_1_0} - There is now some limited IMP-wide gathering of timing statistics. You can add `--statistics=filename` to the command line or use the IMP::base::show_timings() function to view a summary of where IMP has spent time. - Slack was being used inconsistently (and not always correctly) among the various containers and restraints. If you had heavily optimized the slack value for speed, you may want to revisit it. - Class name OwnerPointer was switched to PointerMember, to denote more clearly a ref-counting pointer that is a class member. - The dependency handling in IMP has been rewritten to simplify it and fix a bunch of bugs. There are now more checks of correctness, so behavior that might have silently worked (or not worked) before may now trigger errors. In particular, restraints/score states must have access to the IMP::kernel::Model to do much of anything, so either use constructors that take the model or call IMP::kernel::ModelObject::set_model(). Also, changing your set of inputs or outputs without calling IMP::kernel::ModelObject::set_has_dependencies(false) is now likely to trigger a check failure (it would just silently do the wrong thing before). - Decorators have been cleaned up to make them behave consistently regarding IMP::kernel::Particle vs IMP::kernel::ParticleIndexes. This change shouldn't break existing code, but decorators should move to the new macros IMP_DECORATOR_METHODS(), IMP_DECORATOR_SETUP_0() etc. - The IMP_GRAPH() and IMP_WEIGHTED_GRAPH() macros got an extra argument describing how to show the vertexes in the graph. They also now expose a method, eg IMP::kernel::show_as_graphviz() that writes the graph in graphviz format to an IMP::base::TextOutput. - Support and rules for deprecating code in IMP have been updated. There are new macros and instructions on how to use them that should result in much better visibility about changes. See the C++ wiki for IMP for more information. And there is now an encoded policy of keeping deprecated things around for 1 release after deprecation. - IMP is now much more selective about when it invalidates the dependency graph. See [Dependencies](http://github.com/salilab/imp/wiki/Dependencies) for more information. This change should make evaluation of isolated restraints much faster in certain situations where they were unexpectedly slow before. - IMP::atom::Selection has been completely rewritten as it was very brittle. The set of particles returned may change in some cases, in some cases because it was buggy before, in some cases because of new bugs. - Reminder: the IMP_PROTECTED macros didn't work out well, and are now deprecated. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading `_` on the method name, if you get `Swig director pure virtual method called IMP::kernel::PairScore::do_get_inputs` messages. - The generated `cmake` files are no longer stored in the repository. Instead `tools/build/setup_cmake.py` is run after checkouts and other git events that change the repository tree (assuming you have run `setup_git.py`). - Generation of documentation using `doxygen` has been refactored. It is now done on a per-module basis, via targets like `IMP.core-doc`. The main advantage is that changes to the doc for a module can be tested comparatively quickly, just by building that module's docs. On the down side, docs are no longer quite so heavily interconnected, so dependencies only work to things that the module actually depends on. - The restraint and score state statistics functions were removed from IMP::Model. They had been broken for quite a while, with no one complaining. Statistics can be added back easily if requested. - Added an IMP-specific fork of `git-flow`. It handles things like `README.md` files when you start and finish feature branches and provides a tool to nicely update IMP and display the change log. The main commands of use are `git imp feature start <feature_name>`, `git imp feature finish <feature_name>`, and `git imp update`. You may want to remove the `git-flow` lines from your `.git/config` file as you may accidentally type `git flow` instead of `git imp`. You need to run `setup_git.py` to set things up. - [RMF](http://salilab.github.com/rmf) is now included as a git submodule rather than copied into the IMP repository. You should rerun `setup_git.py` to make sure submodule stuff is initialized. # For IMP 2.0 and earlier # {#changelog_2_0} ## Spring 2013 - Modules and applications now have a _Maintainer_ in their `README.md`, giving the github id of the person to whom issues should be assigned. - `setup_git.py` supports the `--module` flag which will set up git hooks and things for a git repository that contains a module. - To clean up for the release, the module dependencies are now described in a file called `dependencies.py` and the module docs in a `README.md`, both in the main module directory. - IMP-wide support for progress bars was added with the IMP::base::set_progress_display() function and helpers. These are only displayed when the IMP::base::get_log_level() is IMP::base::PROGRESS. ## Winter 2012-2013 - Scons support was removed. Use cmake to build IMP, it is way better. - Examples are run with a `--test` argument when IMP is built with internal checks. Expensive examples should check this flag and shorten what they do in order to try to test all their code without taking too long. - The class IMP::core::NonRigidMember was added to allow particles that are not rigid but whose coordinates are specified via internal coordinates in some IMP::algebra::Reference3D that is part of an IMP::core::RigidBody. - The IMP_BIG_ENDIAN and IMP_LITTLE_ENDIAN preprocessor symbols are deprecated in favor of the BOOST_BIG_ENDIAN and BOOST_LITTLE_ENDIAN symbols. - Support for asymmetric moves was added to IMP::core::MonteCarlo. To do this, IMP::core::MonteCarloMover was added and IMP::core::Mover and IMP::core::MoverBase were deprecated and implemented in terms of the new class. This should require few or no changes in code that uses Monte Carlo support or implements movers (although movers should move to IMP::core::MonteCarlo). All IMP::core::MonteCarloMover objects keep track of statistics (number of moves and number of accepted moves), rather than that being done in the IMP::core::SerialMover. In addition, the poorly defined move probability was removed from IMP::core::MonteCarlo. - IMP::base::WarningContext now outputs warnings immediately. This change was needed as the previous behavior resulted in occasional core dumps when the libraries for log support were unloaded before the libraries where static IMP::base::WarningContexts were declared. And there was no easy way to detect such problems. - You can now (probably) build IMP with [CMake on IMP wiki](https://github.com/salilab/imp/wiki/Cmake). - The maximum logging level and check level can now be controlled independently of the build mode. To, for example, enable progress level logging in a fast build, add maxlog="PROGRESS" to your config. The maximum check level for `release` has been reduced to IMP::USAGE as that speeds things up a bit. - The way applications are structured and specified has been regularized and simplified. Python files are automatically copied/installed and binary executables are compiled from each found `.cpp` file linked with any `.cpp` files contained in a subdirectory called `lib`. Nothing needs to be put into the SConscript file. - The Python constants like IMP.algebra.has_cgal went away as swig autogenerates ones like IMP.algebra.IMP_ALGEBRA_HAS_IMP_CGAL. While the latter are a bit uglier, the constants aren't used much and having them be auto generated is nice. - There have been some major changes to the build system. The main one is that the SConscripts in the directories below the module directory aren't really used. They are being removed from the repository. Now everything really is off of file locations (eg all test_* in test is a test, each *.cpp in bin becomes a program etc). This greatly simplifies the build system, which has, historically been a disaster. You can now define dependencies by dropping files into a dependency subdirectory in your module, no need to mess with the SConscript. - Building IMP now always requires specifying a `repository` argument. This was always required for out of source builds, which is what one should be doing, but is now required for in source builds to, so as to make accidentally polluting your source with an in-source build harder. Use `repository="."` in that case. - The HDF5 functionality in RMF was reorganized. Specifically, the headers were moved into `RMF/HDF5`, the Python functionality into `RMF_HDF5` and the code into namespace `RMF::HDF5` and all mentions of `HDF5` or `hdf5` in the class and function names were removed. - Support for IMP-wide command line flags have been added to IMP::base in the base/flags.h file. They provide a simple command line parsing interface that can be used from C++ and Python. In addition, various functionality in IMP is automatically exposed on the command line when those functions are used. For example, one can control logging, checks or profiling. See IMP/base/flags.h for (some) more info. - OpenMP support in IMP is officially released. All tests pass. But there has been little specific testing so far. - Access to attributes using IMP::ParticleIndex and the IMP::Model now works from Python. See IMP::Model::add_attribute() and kin. - IMP::ParticleKey and IMP::ParticlesKey were renamed to IMP::ParticleIndexKey and IMP::ParticleIndexesKey respectively as that is what they store. Typedefs were provided to support old code. - Indexes have been added to the docs connecting classes with all the methods that return them, examples that use them and methods that take them. And all methods that are used in examples with the examples they are used in. Please check them out and report issues and ways to improve things. You can find links on the main page. - The docs for a modules/biological systems/applications should now be found in an `overview.dox` in the module `doc` folder. Modules in IMP svn have been updated. For external modules, you can just move a `overview.dox` from that was previously generated into your `doc` folder and (eventually) remove the lines from your `doc/SConscript`. This makes it easier to use the full range of doxygen markup, as well as removes the need to escape doxygen markup for Python. - Examples have now been simplified into full blown doxygen examples. This involves merging the `.py` and `.readme` files (putting `\#\#` in front of each line of the `.readme`) and adding a `## \\example module/examplename.py` line at the beginning. You can use `./tools/updaters/update_examples.py modules/mymodule` to update a module that was not already updated. - IMP::base::FailureHandler has been removed as it wasn't used. - Handling of data and includes has been simplified. All appropriate files in the corresponding directories are linked into build/installed. This is only likely to be a problem if you had broken headers sitting around `include`. If so, just comment them out, or move them some place nicer. - Modules now must be in a folder called `modules`, which must have a `description.py` file. - The build system should now be a little better about cleaning up after itself. - How headers/doc files/data/examples - The IMP_PROTECTED macros didn't work out well, and are now deprecated. Sorry. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading "_" from the method name. - Nice version numbers will now be generated if you use git. You need to rerun tools/git/developer_tools/init_git if you have an existing clone. ## Fall 2012 - 11/9: RMF is now an external library shipped with IMP. IMP will build it and install it automatically if you have cmake version 2.8 or higher installed. - 11/9: Updated support macros for IMP::Restraint and IMP::ScoreState classes have been added (IMP_RESTRAINT_2() and IMP_SCORE_STATE_2()) that make it easier to return dependencies on things other than IMP::Particle and IMP::ScoreState objects. In addition, the dependency functions on scores and modifiers have been updated similarly and standardized. In general this should not cause backward compatibility problems unless you were overriding the IMP::ModelObject::do_get_inputs() in IMP::Restraint and using the IMP_RESTRAINT() macro. In that case update to the IMP_RESTRAINT_2() macro. - 11/8: Support has been added for the C++11 keywords override (IMP_OVERRIDE) and final (IMP_FINAL). They allow you to tell the compiler to produce an error if the marked method does not override/implement a parent's method and if anything overrides a given method, respectively. - 11/8: IMP::SingletonModifier::apply_indexes() and IMP::SingletonScore::evaluate_indexes() now require an upper and lower bound index (you can just pass 0 and the size of the array). We can make them option if there is a good reason. - 11/8: IMP::SingletonDerivativeModifier and kin were merged into IMP::SingletonModifier and kin as the only difference, the presence of the IMP::DerivativeAccumulator (whose weight was always 1) wasn't very interested and the split added lots of complexity. If this has an impact on your code, let us know. - 11/8; Evaluation with containers was greatly simplified, moving many methods on containers. We can add backwards compatibility versions if that is interesting. - 11/7: Predicates, scores and modifiers written in Python need to update their methods for returning inputs and outputs. They should now have a method \c _do_get_outputs(self, m, particle_indexes) which should probably just return \c [m.get_particle(i) for in in particle_indexes] to return all the passed particles. - 11/7: It is now preferred that IMP::Restraint and IMP::ScoreState descendants report dependencies directly using IMP::ModelObject::do_get_inputs() and IMP::ModelObject::do_get_outputs(). To do this, you should use the IMP_RESTRAINT_2() and IMP_CONSTRAINT_2() macros instead. Methods like IMP::Restraint::get_input_particles() are now deprecated. This change was made as the old mechanism introduced a lot of extra complication and made it so that there was no way to directly depend on something that was not a Particle or a Container. - 11/3: IMP::container::InContainerPairFilter and friends now treat pairs as unordered. A flag was added to control this. - 9/5: IMP::base::Object classes now have a method IMP::base::Object::clear_caches() that clears all cache data in the object. This is not universally implemented, but now provides a correct path for doing that. The IMP_LIST() type macros call that method when their contents changes. ## Summer 2012 - 7/3: IMP_PERIODIC_OPTIMIZER_STATE() was superceded by the class IMP::core::PeriodicOptimizerState as it is rather cleaner than the macro. - 7/2: in order to make various base classes in IMP more Python friendly, macros IMP_PROTECTED_METHOD(), IMP_PROTECTED_CONSTRUCTOR() were added that properly expose such methods to Python. The standards checks now complain if \c protected is used an a class in the API. - 6/24: Executables from \c module/bin are now put in \b build/module_bin/module. This removes a recurring source of errors in the build script due to the origin and destination directory have the same name in scons. - 6/24: Benchmarks are now put int \c module/benchmark and build to \c build/benchmark/module to remove a recurring source of errors. - 6/18: The IMP::benchmark module supports the GPerfTools profilers when available. Benchmark executables (things in \c bin whose name starts with \c benchmark_) that use the IMP_BENCHMARK() macro have command line flags to turn on and cpu and memory profiling. - 6/12: The naming rules for .i files that are in module \c pyext directories changed. They should now be \c IMP_modulename.name.i instead of \c IMP_modulename_name.i. This is to better support underscores in names. ## Winter 2011/2012 - 20/4: A method, IMP::atom::read_pdb(IMP::base::TextInput, int, IMP::atom::Hierarchy), was added to load coordinates from a pdb file into an existing IMP::atom::Hierarchy. - 4/4: the filter types were superceded by the predicate types. You will only see a difference if you develop a filter class (eg IMP::PairFilter -derived class) or directly call methods on filters (specifically IMP::PairFilter::get_contains()). - 29/3: IMP::Particle::get_has_model() has been changed to IMP::ModelObject::get_is_part_of_model() to be consistent with other classes. - 28/3: IMP::core::Mover changed slightly, adding an extra method IMP::core::Mover::get_output_particles() and requiring the IMP::Model be passed to the constructor. All movers in IMP SVN were updated without any difficulty. This allows movers to use information contained in the dependency graph amongst particles so they can, for example, weight moves based on scores. - 26/3: IMP::SingletonContainerInput etc were added. Methods that take these as their arguments can be passed either containers or lists of particles (or tuples), as convenient. All methods that take IMP::SingletonContainer objects should eventually move over to taking this (changing just requires changing the argument from a pointer to an Input). - 22/3: Over the last week or so, which IMP headers include which other heads have been substantially simplified. This may result in compilation errors for code outside of \c svn that depended on IMP headers including other headers. In general, you want to explicitly include the header defining each thing you use. - 15/3: The vector used throughout IMP is now exposed at IMP::base::Vector and what it does it documented. - 10/3: IMP::domino::RestraintCache was introduced to provide a documented centralized place to handle restraint evaluation and caching in DOMINO. In the switch various bugs involving maximum scores and IMP::RestraintSet objects were fixed. - 10/3: Scoring functions in IMP are now implemented in IMP::ScoringFunction objects. All restraint evaluation is handled by them and they can cache various useful information to aid in accelerating restraint evaluation. Backward compatibility interfaces have mostly been provided, please complain if some important interface went away. This has greatly simplified the code as well as fixed various bugs in various types of restraint evaluation (particularly in the handling of maximum scores on IMP::Restraint and IMP::RestraintSet objects. - 29/2: The macros IMP_NAMED_TUPLE_1(),IMP_NAMED_TUPLE_2(),IMP_NAMED_TUPLE_3() were added for easy generation of simply struct for use in passing and returning values. The generated structs support comparison, output, hashing and have constructors. Use the macros instead of a boost::tuple or a a POD struct. - 16/2: Restraints now are evaluated with the correct weights when evaluated by themselves as opposed to always returning their weight as seen by the model as a whole. - 15/2: The IMP::algebra grid support was rearranged and simplified. Most of the changes should be transparent. However, the IMP::algebra::grids namespace got folded into IMP::algebra, which might break some code. And various (obscure) constructors of IMP::algebra::GridD were removed. All code in svn has been changed as needed. - 3/2: The IMP::domino HDF5 container has been revamped. This involves splitting it into a write version (IMP::domino::WriteHDF5AssignmentContainer) and a read version (IMP::domino::ReadHDF5AssignmentContainer). - 3/2: IMP::domino::WriteAssignmentContainer and IMP::domino::ReadAssignmentContainer were added to do direct binary file writing. This is faster than the HDF5 versions. - 3/2: The installation instructions have been revamped and the developer instruction, in particular, have been greatly simplified. - 24/1: \c tools/show-changes has been added that shows the changes in the history log since you last updated. - 24/1: \c tools/make-source has been added that adds a source/header pair to a module - 24/1: principal components analysis in IMP::algebra was generalized to work in any dimension. This required renaming the class to IMP::algebra::PrincipalComponentAnalysisD. - 24/1: the output of the IMP::algebra::PrincipalComponentAnalysisD::show() was changed to reflect the standards (values produce compact, single line output on show). The old show has been moved to IMP::multifit - 19/1: IMP::container::ConsecutivePairContainer had IMP::container::ExclusiveConsecutivePairContainer split off of it to make the two operating mode clearer and so that there could be a clear IMP::container::ExclusiveConsecutivePairFilter to use. - IMP::algebra::Grid::apply was added to provide faster access to all (or, later, part of a grid). It can be an order of magnitude or more faster than using the iterators when doing simple operations. - the scons config options \c pretty and \c color were added (defaulting to True). When they are true, build commands are mostly suppressed and a much briefer description of what is being done is printed out (in color). Note that \c color mode and compilation in emacs don't get along too well. - IMP::algebra::ParabolicFit was renamed to IMP::algebra::ParabolicFit2D (a backwards compatibility typedef is currently provided) - IMP::algebra::LinearFit was renamed to IMP::algebra::LinearFit2D. It now supports error bars on the y-values. ## Fall 2011 - IMP::compatibility::checked_vector was renamed to IMP::compatibility::vector and made available to IMP as IMP::vector. Typedefs should use this type (IMP::vector). - Various IMP::algebra plural typedefs which had been using std::vector (since they never were updated), have now been moved to IMP::vector. - IMP::statistics::HistogramD now supports display with \c matplotlib. - IMP::statistics::HistogramD supercedes IMP::statistics::Histogram - predicate base classes have been added: IMP::SingletonPredicate, IMP::PairPredicate etc. These take particles and return an integer. Current functionality using them includes IMP::core::PredicateSingletonScore that uses a predicate to choose which score to apply. - the core generic restraint and score state support was moved to the kernel to avoid breaking various scores outside of IMP svn. - the IMP::display module was refactored and moved up to be a dependency of IMP::core and IMP::atom rather than vice versa. The main observable results are that you should now use IMP::atom::HierarchyGeometry, IMP::atom::SelectionGeometry, IMP::atom::HierarchiesGeometry, IMP::core::XYZRGeometry, IMP::atom::BondGeometry etc instead of finding them in IMP::display. This change means that, display can be used internally when testing and debugging functionality in IMP::core and IMP::atom and that tests that use IMP::display don't need to build all of IMP::atom. - Decomposition of restraints now results in restraints that are faster to evaluate, using the generic restraint mechanism in IMP kernel. - IMP::algebra::get_unit_bounding_box_d() was changed to return a bounding box with radius 1 to be consistent with IMP::algebra::get_unit_sphere_d() - A general purpose scheme to associate data with tuples of hierarchy nodes was added to the RMF format. This deprecates the special purpose bond support. - the way various types of geometry is stored in RMF files was changed, breaking backwards compatibility. ## Summer 2011 - some functionality from the kernel has been spun off into a new module IMP.base. You probably won't notice anything. The only significant changes are that functions that accept or return IMP::base::TextInput and IMP::base::TextOutput must declare them with the \c base qualifier (eg base::TextInput) so that swig is happy. - the source for the kernel has been moved from \c kernel to \c modules/kernel. This is unlikely to effect anyone. - IMP::core::ExcludedVolumeRestraint now supports filters - IMP::domino::HDF5AssignmentTable now caches internally, which makes it several times faster than before. - you can now pick which Python to use with the \c python argument to scons. - the units for BD diffusion coefficients changed to be consistent with other IMP units. - IMP::domino::get_interaction_graph() now takes an IMP::RestraintsTemp, like other functions - The assert failure handling function was documented and renamed to IMP::handle_error() so it is easier to find when you want to use gdb. - to work around various issues with ref counting pointers and compiler eccentricities, we added emulation of the C++0x \c nullptr. You should prefer that to \c NULL in IMP code that you want to run cross platform. - IMP::IntsList, IMP::FloatsList and IMP::StringsList have been added to replace the various scattered solutions for passing such types. - IMP::Restraint::get_decomposition() and IMP::Restraint::get_current_decomposition() changed to return a single Restraint, instead of a list. This makes it easier to propagate the weight and maximum reliably. - IMP now supports usage of restraints without adding them to the IMP::Model scoring function. To do this, do IMP::Restraint::set_model(), passing the model. The restraint can then be evaluated. - IMP::domino::DominoSampler does not use the IMP::domino::OptimizeRestraints restraint optimization by default. Instead, if it is asked to create the IMP::domino::RestraintScoreSubsetFilterTable, it decomposes the restraints first (via IMP::create_decomposition()). This avoids several bugs and very fragile code. - the IMP::core::DiameterRestraint was ignoring the radii of the particles involved and so was giving artificially low scores. It now is on the actual diameter of this molecule. - the IMP::core::MonteCarlo optimizers now support incremental evaluation. This can be a lot faster when only a few particles are moved each step. - all the plural types in IMP have moved to bounds checked vectors when using debug builds with gcc. If you get a non-imp error about an invalid access, it is probably some code that walks off the end of an array. Please report it. - The IMP::core::MoverBase was changed to move it off of the IMP::SingletonContainer. This resulted in the IMP::core::BallMover and IMP::core::NormalMover having new constructors. Now create them from a list of particles, a list of float keys and a radius/standard deviation. This removes some complexity and inefficiencies as well is removes some scope for errors. - Temp and non-temp decorator lists have been merged in C++. This was done to simplify code since no one was using the non-temp variants. Now all decorator lists are temp (non-reference counted) variants. - restraints and restraint sets are now evaluated only using "canonical" weights. That is, each restraint has one weight for the model (the total over all the restraint sets it is contained in and will always be evaluated with that weight). This allows the evaluation of simple restraints to be faster. - The IMP_LIST() macros have been changed slightly. If your usage needs to take action upon addition, removal or changes to the list, use the IMP_LIST_ACTION() variant in your header. The IMP_LIST_IMPL() macro no longer takes the actions, and the last three (generally empty arguments) will have to be removed from the call to get it to work. This is needed to move towards more flexible and efficient passing of lists of values. - IMP::domino uses the new IMP::Model::evaluate_if_good() support to accelerate computations - IMP::core::MonteCarlo can use the IMP::Model::evaluate_if_good() support to more quickly reject very bad conformations. See IMP::core::MonteCarlo::set_maximum_difference(). - IMP::core::ExcludedVolumeRestraint has been rewritten to be faster. It is especially faster when used with IMP::Model::evaluate_if_good() - a new evaluation mode IMP::Model::evaluate_if_good() and IMP::Model::evaluate_if_below() has been added. It allows restraints to shortcut score computation when the score can be shown to be worse than a maximum value. - as part of moving towards more efficient use of evaluate_if_good() various deprecated methods have been removed from IMP::core::ExcludedVolumeRestraint and IMP::core::RigidClosePairsFinder. - incremental evaluation support was removed from IMP. The reason for doing this was that the code was complicated, created overhead even when incremental support was not used and was not commonly used. Further, decomposing the restraints and using the dependency graph yields similar results and is more transparent. - the RMF helper binaries have been moved to their own application so that they get installed. - IMP::display::Writer has been revised to better support writers that write multiple frames to a single file and ones that write binary data. The noticeable changes are that the existing IMP::display::Writer instances must be created with a file name. And the IMP::display::WriteOptimizerState just takes a writer rather than a write and a file name template. IMP::display::Writer::set_frame() has been added so that you can directly control the frame being written. ## Spring 2011 - An interactive mode has been added to IMP::domino::DominoSampler. The interactive mode is useful for understanding domino sampling runs and distributing domino computations across a cluster. - Functions IMP::domino::set_assignments() and IMP::domino::get_assignments() were added which support writing subset assignments to data fields in hdf5 files. These can be used to helping to distribute domino on a cluster as well as debugging the sampling process. - a Python function IMP.show_graphviz() has been added that attempts to render a graph with \c dot and then open the pdf. - The Python functions IMP.show_altgraph() and IMP.get_networkx_graph() have been added to provide alternatives for quick viewing of IMP graphs. Unfortunately, they still pale compared to graphviz. - IMP.atom.create_cover() was added to facilitate creating a summary particle for part of a hierarchy in order to accelerate restraints. - IMP.hdf5 was renamed to IMP.rmf, the proposed name of the file format that it supports. That also better separates the hdf5 support from the file format support. - The old I/O functions have been marked as deprecated in favor of IMP.hdf5. You may need to compile with deprecated="yes" if you use them. - The Python graph interface now has a remove_vertex method. - One can now access the model restraint evaluation statistics programmatically using IMP::Model::get_restaint_statistics(). Restraint statistics can be logged to a file during optimization using IMP::core::WriteRestraintScoreStatisticsOptimizerState. - IMP::domino::DominoSampler now allows you to specify the tree on which merges occur directly. Also, you now need to specify the merge tree rather than the JunctionTree if you want to do non-loopy inference and specify something about how it is done. Sorry, it makes the code a lot easier. - Preliminary symmetry support has been added via the IMP::core::TransformationSymmetry and IMP::core::TransformationAndReflectionSymmetry classes. ## Winter 2011 - There now is a common base class IMP::atom::Simulation for IMP::atom::BrownianDynamics and IMP::atom::MolecularDynamics. In merging them there were several minor changes made to both. The most noticeable may be that the function for setting the maximum time step length is now IMP::atom::Simulation::set_maximum_time_step(). The last time step length can be extracted by IMP::atom::Simulation::get_last_time_step() - IMP::core::MonteCarlo::set_temperature() was renamed to IMP::core::MonteCarlo::set_kt() as that is what it is doing. - A new module IMP.hdf5 has been added to provide more general support for saving and loading IMP::atom::Hierarchy structures to files. Multiple conformations of each hierarchy can be saved as well as information about geometry and restraints (although these can't be read at the moment, so they are not so useful). - IMP::algebra::grids::GridD now supports dense grids in arbitrary dimension through use of IMP::algebra::grids::DenseGridStorageD. - IMP::Optimizer now can use a list of restraint sets to compute the score rather than the complete Model score. - each IMP::Restraint now has its own weight. This is lightly tested. - A new log level IMP::PROGRESS has been added. The idea is that IMP::Sampler::get_sample() calls and other expensive loops can show a boost progress meter at this log level to give some idea when things are moving forward. - IMP::domino::TrivialParticleStates renamed to IMP::domino::IndexParticleStates to be less judgmental - IMP::domino::CompoundParticleStates and IMP::domino::RecursiveParticleStates added - IMP::domino::DominoSampler now supports maximum scores on IMP::RestraintSets (untested). - the name of the clustering methods have been changed to create_ from get_ since they create objects. - the names of the recently added metric based clustering support has been changed to Metric rather than Distance. - IMP now has the ability to have certain tests designated as expensive and to skip those tests when test-fast is run. The motivation for this is that the tests take a really long time making it impractical to run before every commit. And it is hard to tell which tests have not been run when running with the necessary multiple build processes. My thought is that tests that take about a second or less on a debug build should be left along (this is probably >95% of the tests), but the few that take longer should be marked as expensive. To do that, rename the test to expensive_text_XXX.py (from test_XXX.py) and add expensive_python_tests= env.IMPModuleGetExpensivePythonTests() to the IMPModuleTest call in the test SConscript. - the fuction IMP::display::create_restraint_geometry() was added to do what its name says. - runtime dimensional geometric objects have been added. These include IMP::algebra::VectorKD, IMP::algebra::BoundingBoxKD and IMP::algebra::SphereKD. The IMP::statistics::Embedding classes now use IMP::algebra::VectorKD instead of Floats. The sparse grids also support variable dimensions as to the nearest neighbor searches. - IMP::atom::DopePairScore has been added to IMP. - IMP::atom::CHARMMStereochemistryRestraint provides a high-level simplified interface to the bonded part of the CHARMM forcefield. - IMP modules are now built by first combining all the .cpp files into one, and then building that file. This greatly accelerates building of a module from scratch as build time is dominated by parsing headers. This does change the semantics of .cpp files as they see headers included by and functions declared by an arbitrary subset of other .cpp files. For now they are still required to be able to be compiled separately. To get the only behavior back either globally or on a per-module basis, see the percppcompilation parameter to scons. - Instructions and scripts have been added for building modules and things externally to the IMP build. See the installation guide. - IMP::atom::CHARMMParameters can now automatically map CHARMM-style atom and residue naming to PDB-style, so CHARMM topology files can be used unmodified. - IMP::atom::CHARMMTopology::add_sequence() provides a simple way to generate a topology from a string of amino acid one-letter codes. - IMP::atom::CHARMMTopology::add_coordinates() will generate atomic Cartesian coordinates from CHARMM internal coordinate information. - IMP::atom::CHARMMTopology::add_missing_atoms() and IMP::atom::remove_charmm_untyped_atoms() can now be used to ensure that atoms read from a PDB file match the CHARMM topology. - building of deprecated functionality is now false by default - the IMP_COMPARISONS(), IMP_COMPARISONS_1(), IMP_COMPARISONS_2() macros now take the name of the class as an additional (first) argument. This avoids the formerly needed, \c This typedef and makes them more consistent with the other macros in IMP. - modules can now be built externally to the IMP svn in addition to applications and biological systems. To do so, create a directory with your module in a subdir containing links to the IMP SConscript file and scons_tools and a config.py that has information needed to run IMP (eg an appropriate \c pythonpath, \c ldpath). - removed IMP::algebra::GridD::get_index(VectorD) as it wasn't very safe (in that vectors coordinates can get rounded between the IMP::algebra::GridD::get_has_index() and IMP::algebra::GridD::get_index() calls. Use IMP::algebra::GridD::get_nearest_index() or IMP::algebra::GridD::get_extended_index() or IMP::algebra::GridD::operator[]() instead. - renamed IMP::domino::ParticleStatesTable::get_particles() to IMP::domino::ParticleStatesTable::get_subset() as it is more useful to be able to get the subset containing all the particles. - rename IMP::restrainer::Restraint::print_all_restraints() to IMP::restrainer::Restraint::show_all_restraints() to be consistent with the rest of IMP. - renamed IMP::atom::get_conect_record_string() to IMP::atom::get_pdb_conect_record_string() to make clearer what it is. - class and method names are now spell checked. Add missing words to spelling_exceptions in the module test call. - the plural names (eg IMP.Particles()) have been removed from the Python side as they are just Python lists. - biological systems can now include internal modules. These provide a convenient place to put application specific C++ and Python code. See the example system for more info. - IMP_VALUES() now includes IMP_OUTPUT_OPERATOR() so you no longer need two namespace macros for each value. - the function IMP::core::Harmonic::k_from_standard_deviation was renamed to IMP::core::Harmonic::get_k_from_standard_deviation() to conform to the standards. - the IMP::atom::PDBSelector and IMP::atom::Mol2Selector became IMP::Object classes. All this means is that they need to be created with \c new in C++ code. It also means that one is no longer prohibited from storing them (not that there is much use in doing that). - the standards tests have been improved to test more useful aspects. Now class and function names are checked some as well as all modules have the checks for show and plural types. The error messages should explain how to fix things. - IMP::core::RigidBody::lazy_set_reference_frame() was renamed to IMP::core::RigidBody::set_reference_frame_lazy() to fit with the standards better. - the transform based methods on reference frame which have been deprecated for a while were removed - the hierarchy visiting and searching functions have been renamed to better fit the IMP naming scheme as we don't expect they were used outside of the IMP source. - IMP::atom::NonhydrogenMol2Selector was renamed IMP::atom::NonHydrogenMol2Selector to be consistent with the PDB version ## Fall 2010 - IMP::display::PlaneGeometry was added to display planes - the optimize methods on optimizers was moved to a "non-virtual interface". Usage has not changed, but optimizers should implement IMP::Optimizer::do_optimize() instead of IMP::Optimizer::optimize() to benefit from better checking. - the two-vector constructor for IMP::algebra::Plane3D now expects the second vector to be a unit vector as otherwise one could switch the order around without noticing. - the io functions in the kernel have been cleaned up a bit (the old ones are there for backwards compatibility). See IMP::write_particles(), IMP::write_particles_binary(), IMP::write_particles_to_buffer() and IMP::read_particles(), IMP::read_particles_binary(), IMP::read_particles_from_buffer() - IMP::core::WriteModelOptimizerState is going away as it duplicated IMP::WriteParticlesOptimizerState. - the IMP::algebra::Grid3D has been changed to make it support any (fixed) dimension when used with sparse storage. The template arguments have also been changed to remove a redundant one. See IMP::algebra::GridD. - the biological systems specification has been improved a little adding ability to run some rudimentary tests to make sure the scripts are ok and to not require that everything be run from the svn directory. - A new method IMP::atom::create_compatible_rigid_body() which creates a rigid body from a hierarchy which has the same internal coordinates as another rigid body. A way to make the rigid body construction more canonical would be nice (so we don't need this function), but I'm not sure how to do that. - Logging has been improved to better report the name of methods producing the log messages - a macro IMP_FUNCTION_LOG has been added which should be added to the start of non-member functions which produce significant logging output. - IMP::algebra::get_distance(const Rotation3D&,const Rotation3D&) was changed to use the Euclidean distance in R4 as that seems better behaved than the prior metric. - IMP.domino2 was renamed to IMP.domino and the old IMP.domino was moved - the documentation for classes and some methods now includes a list of all examples which use the class. Please report any incorrect links or phantom classes appearing in the docs. Unfortunately, functions have to be added individually. Please request any functions of particular interest. - applications and biological systems now have their own documentation pages - IMP::atom::Hierarchy requires radii on its leaves - IMP::atom::read_pdb add radii to the atoms - IMP::helper was merged into IMP::restrainer as all the functionality there was primarily to provide support for restrainer - IMP::domino2::SubsetEvaluatorTable was removed. Pass the model and IMP::domino2::ParticleStatesTable to the IMP::domino2::RestraintScoreSubsetFilterTable instead. - IMP::atom::Selection was added to allow specification of parts of IMP::atom::Hierarchy using biological sorts of names. A few functions to create restraints from them were added too. - a new example was added in the kernel to show IMP used on a biological system - IMP::display::PymolWriter now supports adding labels to things - IMP::display::PairRestraintGeometry and IMP::display::ConnectivityRestraintGeometry were added to aid in visualizing those restraints. - IMP::bullet was added with an optimizer to use hard body physics to resolve local collisions. - IMP::atom::setup_as_rigid_body() has been deprecated and replaced by IMP::atom::create_rigid_body(). The latter creates the rigid body as a separate particle so that the molecular hierarchy doesn't have to worry about having its bounding sphere large enough to enclose all the members. - IMP::core::KClosePairsPairScore was added to have count based cutoffs for close pair scoring. - IMP::algebra::get_surface_area_and_volume() was added to compute the two numbers for a set of balls ## Summer 2010 - The IMP::core::RigidDistancePairScore and IMP::core::ClosePairsPairScore have been revised to remove certain inconsistencies and ambiguities and to make them faster. Existing code using them will need to be slightly tweaked to work. - Javi added his 2D EM scoring functionality as IMP::em2D - some new ways to speed up scons have been added. See IMP::core::HarmonicDistancePairScore, IMP::core::SoftSpherePairScore and, for C++ users, the generic restraint support IMP::container::create_restraint() and IMP::core::create_restraint(). These each offer between 2x and 1.2x speedup when using simple scores (like harmonics on distances). In addition, a function IMP::container::get_slack_estimate() has been added which will estimate the correct slack to use with a IMP::container::ClosePairContainer. This will be extended to support IMP::core::ExcludedVolumeRestraint eventually. - support for dense and sparse grids have been added to IMP::algebra. See the IMP::algebra module page (I'm still working on how to make the docs clearer). - all write functions now have associated IMP::OptimizerState and IMP::FailureHandler classes. For examples, see IMP::WriteOptimizerState, IMP::WriteBinaryOptimizerState, IMP::WriteFailureHandler, IMP::atom::WritePDBOptimizerState etc. - functions IMP::write_binary_model(), IMP::read_binary_model() and IMP::read_configurations() have been added to support space and filesystem efficient storage of model configurations. Multiple configuration can be written to a single file. This functionality requires NetCDF be installed (NetCDF is a standard library for reading and writing binary data). - domino2 has been added. It provides a simpler interface on Keren's domino optimize as well as a number of new optimization and hooks for customizing the optimization process - a better explanation of how to use reference counted pointers has been added to IMP::Pointer. - all simple, planar polygons are now supported for display when CGAL is present ## Week of April 1, 2010 - All plural objects in Python are now simply Python lists rather than C++ recreations of Python lists. This makes compilation of the wrappers much, much, much faster. The visible changes are - types are not checked until the lists are passed to C++ - passing std::vector types by pointer and storing passed std::vector types by pointer will not work (but should never have been done anyway from code exposed to Python) - swig has issues with some overloads (not always for obvious reasons). These are being sorted out. None are commonly used. - get_particles() on a plural object doesn't work in Python (and is not needed in C++). Implicit conversions will always work. - All containers need to know the model at time of construction. If, after updating, you get an error on a container constructor, just add the model as the first argument. This also means that empty lists of particles in the constructors cause check failures. We can figure out a work around if this is too annoying. ## Week of March 13, 2010 - function get_rotation_in_radians_about_axis() was renamed to get_rotation_about_axis() as it didn't parse and we don't have the "in radians" elsewhere. - Added IMP::algebra::ReferenceFrame3D. As part of the cleanup, IMP::core::RigidBody::set_reference_frame() replaces set_transformation(). And the reference-frame related transformation construction functions were replaced. IMP::algebra::get_rotation_from_x_y_axes() was added to aid in the construction of reference frames (and replaces the algebra function taking three vectors). ## Week of March 6, 2010 - Publications now have a description field to give browsers an indication of what is in the publication to aid selection of what to read/cite. ## Week of Feb 27, 2010 - IMP::algebra::decompose_rotation_into_axis_angle() was renamed to IMP::algebra::get_angle_and_axis(). - renamed IMP::atom::bond() and IMP::atom::unbond() to IMP::atom::create_bond() and IMP::atom::destroy_bond() to be more consistent with things - renamed IMP::core::root() to IMP::core::get_root() ## Week of Feb 21, 2010 - IMP::Object::set_was_owned() was changed to IMP::Object::set_was_used() as that better reflects what the flag means. - various more interpolation and endian methods had their names fixed (by adding \c get_) - the function get_mean() on vectors was removed as it duplicated IMP::algebra::get_centroid() - build_Transformation3D_from_transformation3D was replaced by IMP.algebra.get_transformation_3d() - IMP::algebra::almost_equal() is now IMP::algebra::get_are_almost_equal() to fit the naming conventions. - the random vector generators have been cleaned up to remove redundant ones and make the names more consistent. They now have the form of either - IMP::algebra::get_random_vector_in() or IMP::algebra::get_random_vector_on() where there is one argument describing the volume or shape (such as an IMP::algebra::SphereD<4>). - or IMP::algebra::get_uniform_surface_cover() where the first argument is the primitive whose surface we are covering and the second is the number of points. Some good sed strings are - "s\#random_vector_in_unit_sphere()\#get_random_vector_in(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_in_unit_box()\#get_random_vector_in(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_on_unit_sphere()\#get_random_vector_on(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_on_unit_box()\#get_random_vector_on(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_in_box(\#get_random_vector_in(\#g" - "s\#random_vector_on_box(\#get_random_vector_on(\#g" - "s\#random_vector_in_sphere(\#get_random_vector_in(\#g" - "s\#random_vector_on_sphere(\#get_random_vector_on(\#g" - "s\#grid_cover\#get_grid_surface_cover\#g" - "s\#uniform_coverr\#get_uniform_surface_cover\#g" . - Spheres in IMP::algebra are now supported in general dimensions. See IMP::algebra::SphereD. - Many methods in IMP::algebra were renamed to be more consistent with other methods or general conventions. Most of these simply involved adding \c get_ as a prefix and/or add a \c _3d suffix when they did not take any arguments. - build_Rotation2D_from_Vector2D became IMP::algebra::get_rotation_to_x_axis() - most of the Rotation3D and Transformation3D functions (in addition, IMP::algebra::get_identity_rotation_3d() and IMP::algebra::get_identity_transformation_3d() added the \c _3d so we can add 2D versions at some point) - IMP::algebra::Transformation3D::get_transformed() and IMP::algebra::Rotation3D::get_rotated(). Replacing \c ".transform(" worked in the IMP code base. - IMP::algebra::get_distance(), this also carries over to the IMP::core::get_distance() The following global replacements are probably safe (in sed expressions) - "s\#core.distance(\#core.get_distance\#g" - "s\#algebra.distance(\#algebra.get_distance\#g" - "s\#\.transform(\#.get_transformed(\#g" Only in C++ - "s\#\.rotate(\#.get_rotated(\#g" Only in C++ - "s\#rotation_from\#get_rotation_from\#g" - "s\#rotation_in\#get_rotation_in\#g" - "s\#identity_rotation(\#get_identity_rotation_3d(" - "s\#::distance(\#::get_distance(\#g" . - IMP::algebra::Vector3D and IMP::algebra::Vector3Ds were replaced by the expanded (non-typedef) version in functions parsed by SWIG as the current situation was extremely brittle. For the moment, you should not use the typedefs in definitions that swig sees. Sorry. - IMP::algebra::Sphere3DPatch was renamed to IMP::algebra::SpherePatch3D to be consistent with the other geometry and allow us to support higher dimensions in the future. - rigid_align_first_to_second was renamed to IMP::algebra::get_transformation_aligning_first_to_second() and build_Transformation2D_from_point_sets was renamed to IMP::algebra::get_transformation_aligning_pair() ## Week of Feb 14, 2010 - The IMP::atom::Bond types were rearranged to reconcile the protein-centric classification with ligand support. - IMP::atom::Atom no longer has charge since we have IMP::atom::Charged. - IMP::atom::read_mol2() and IMP::atom::write_mol2() were added - IMP::atom::Selector and descendants renamed to IMP::atom::PDBSelector since we will soon have Mol2Selectors. In addition, the IMP::atom::IgnoreAlteriativesSelector is now IMP::atom::NonAlternativeSelector to keep the name positive, like the others. In the IMP body, all instances of the string Selector were for PDBSelectors and so sed was safe to use. - IMP::atom::read_pdb() lost the "ignore_alternatives" argument. Instead, all of the provided selectors, other than IMP::atom::AllSelector ignore alternatives internally. Code that uses the IMP::atom::AllSelector should probably change to the IMP::atom::NonAlternativeSelector. - IMP::example renamed (from examples) - The IMP_USAGE_CHECK() macro lost the (unused) third argument. ## Week of February 3, 2010 - The model io code as moved into the kernel so it is now IMP::write_model() and IMP::read_model() - Remove the version_info argument from the implementation macros. Instead, just pick up the module version info automatically. - Add a function Object::get_type_name() and Object::get_module() to get the respective bits of information about each object. - Change Object::show() to be a non-virtual wrapper which prints the object name and other info before passing control off to Object::do_show(). - The containers were moved to IMP.container. See that module page for a full list of what moved. In the process, the backwards compatibility functions were removed from IMP::container::PairsRestraint and kin. - IMP.atom.create_protein() now returns the created protein - the simplify functions have been moved from IMP.helper. They are now IMP.atom.create_simplified_along_backbone(). - IMP.helper.clone() has been renamed to IMP.atom.create_clone() - IMP.misc.ConnectingPairContainer has been moved to IMP.core.ConnectingPairContainer - The locations for pixels in IMP::em::DensityMap have been moved to the center of the corresponding voxel from the lower left corner. - IMP::em::get_transformed(), IMP::em::get_resampled() and IMP::em::get_density() have been added. - The documentation has been majorly cleaned up and simplified. Please look around and see if you like it. There are now two main pages, the introduction and the developers guide, rather than the previous plethora of pages. - IMP::algebra::Cone3D::get_bounding_sphere() has been removed since it was not of general use, not returning the bounding sphere you would expect and was trivial to implement. - IMP::algebra::projection() was removed since it didn't actually project onto the segment and the name didn't follow the normal conventions. And it wasn't used. - Most geometry now has namespace methods IMP::algebra::get_bounding_box(), IMP::algebra::get_surface_area(), IMP::algebra::get_volume(). ## Week of January 22, 2010 - A generic geometry layer has been added (see \ref genericgeometry). Using it, IMP::atom::rmsd(), IMP::algebra::alignment_taking_first_to_second() and the new IMP::atom::native_overlap() all can work on either collections of either IMP::core::XYZ particles or IMP::algebra::Vector3Ds. - Deprecated functionality has been removed from IMP.core. For those still using it, IMP::core::ClosePairsScoreState was replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairsScoreState was replaced by IMP::core::CloseBipartitePairContainer. - The macros for Scores changed slightly. Scores not in IMP svn which use the IMP_*_SCORE macros now have to implement the get_is_changed() function outside of the class declaration and scores which use the IMP_SIMPLE_*_SCORE now don't have to implement the functions to get interactions and used particles. - Rename IMP::core::MinimumPairScoreRestraint (and kin, Maximum, Singleton, Triplet and Quad) to IMP::core::MinimumPairRestraint. - Add IMP::core::MinumumPairScore (and kin). ## Week of January 1, 2010 - IMP::em is now using Particles to resample density maps and for fitting restraints. The previous ParticleAccessPoint mechanism is removed. - IMP::UsageException and IMP::InternalException are now reserved for their respective checks (and cannot be thrown by IMP_THROW()). This makes the distinction between API exceptions (which are thrown in any build mode and should be documented and tested) and helper exceptions which are only thrown when checks are turned on (and hence should not be documented) clearer. - IMP::core::RestraintSet moved to IMP::RestraintSet. The restraint sets now simply define a tree with weights which the model traverses as opposed to being an opaque IMP::Restraint themselves. This means that incremental evaluation of restraints in restraint sets now works and that restraints in restraint sets can be individually evaluated. Existing working usage should not have changed (so \c sed should manage updates just fine), but the internal implementation is completely new, so pay attention. - IMP now has a text i/o layer consisting of IMP::TextInput and IMP::TextOutput. These classes are convertible to C++ streams and implicitly constructable from both streams and file paths. As a result, functions that take them will work when called with a string for a file path, a Python file or a C++ stream. - The IMP.examples module now documents how to use the needed aspects of scons and SWIG. - The IMP.examples module is used as the basis for new modules created by the make-module script, so those include more documentation and are tested. - Module overview pages are now generated from the information passed in the SConscript file, rather than from the (now removed) modulename.dox file. See the examples file for an example of the things you can do there. - boost program options is now a checked for dependency. - The IMP::Model now uses dependency information from IMP::ScoreStates to automatically reorder the IMP::ScoreStates so that they are evaluated in the right order. Also, IMP::Model::evaluate() on a subset of the restraints now only updates the needed score states, not all of them. - Functions on rigid bodies no longer provide the IMP::core::RigidMemberRefiner as a default as it is, pretty much always, the wrong thing to use. That class may go away without notice in the future and is no longer documented. - the IMP::helper::create_simple_connectivity_on_rigid_bodies() and IMP::helper::create_simple_excluded_volume_on_rigid_bodies() now both take an IMP::Refiner as an (optional) final argument. This makes it so they can be used with the rigid bodies created from molecular hierarchies. This code still needs to be updated and the docs don't reflect what the code was or is doing, so use with caution (see notes in the header file). - The IMP::helper create* functions now all take non-Object inputs as const& to make clear what is input and that there are no memory management gotchas. - a new base class IMP::Constraint has been added for constraints. Various other classes have been renamed to be consistent. ## Week of December 11 - The interface of IMP::PairScore changed slightly to make the function signatures more uniform across various methods. Now all methods that take a pair of IMP::Particle objects take a const IMP::ParticlePair&. Previously some took that, some took an IMP::ParticlePair and some took two IMP::Particle*s. If you have an implementation of an IMP::PairScore outside of svn, you will have to update your function signatures (the compiler will tell you where). Backward compatible IMP::PairScore::evaluate() methods have been provided, so simple users of pair scores shouldn't see a difference. - IMP::core::PairRestraint takes a IMP::ParticlePair in the constructor instead of two IMP::Particle*s. A backwards compatible constructor can be added if needed. - IMP::ParticlePair::first and second went away since no one claimed to be using it and it complicated the implementation. - The generic restraints/score states/scores infrastructure has been extended to triplets and quads (for angles and torsion angles). ## Week of December 4 - IMP now makes use of Boost.Filesystem to manipulate paths. Things may not work well on systems where boost libs are not available which does not support posix paths. But libless boost installs generally don't exist outside of our test systems. - new functions like IMP::core::get_example_path(filename) and IMP::core::get_data_path(filename) allow one to easily find a data file named 'filename' for the examples and library internal data. - The IMP::display module has been revamped in order to support non-linear geometry. VRML support has been dropped. CMM support will be dropped unless someone requests it (as IMP::display::ChimeraWriter does all it does and more). ## Week of November 19 - imp.xcodeproj was moved into tools and updated accordingly - the bin directory finally went away - IMP::display::ChimeraWriter now supports the latest version of Chimera - IMP::display::CGOWriter was renamed to IMP::display::PymolWriter as it can do more than just CGO. - IMP::VectorOfRefCounted::operator[] can now be used for assignments just like a normal vector. No more need to use the IMP::VectorOfRefCounted::set() method. - the IMP::atom::Hierarchy constructor from an IMP::Particle* is now explicit (as were the decorator constructors) ## Week of November 12 - All exception types are now mapped to equivalent classes in Python. The class hierarchy is similar (e.g. IMP::ModelException derives from IMP::Exception in both C++ and Python) so all IMP exceptions can be caught in Python using "except IMP.Exception". For convenience, the IMP::IOException, IMP::ValueException and IMP::IndexException classes in Python also derive from the similarly-named standard Python classes (IOError, ValueError and IndexError respectively). ## week of October 29 Update - You can now add members to rigid bodies via IMP::core::RigidBody::add_member(). These members can be other rigid bodies (although this functionality is only lightly tested). - The code to get bounding volumes for IMP::atom::Hierarchy has been moved to the atom module, as as the clone and destroy code. They can be found at IMP::atom::clone(), IMP::atom::destroy(), IMP::atom::get_bounding_box(). - Added IMP::core::VolumeRestraint. - Several functions have been added to helper in creating restraints. They are create_simple_connectivity_on_rigid_bodies(), create_simple_connectivity_on_molecules(), create_simple_distance(), create_simple_diameter(), create_simple_em_fit(), and create_simple_excluded_volume_on_rigid_bodies(). Functions load_em_density_map() and set_rigid_bodies() are added as well. See IMP::helper for more information. - IMP::ScoreStates are further hidden. Namely, IMP::core::RigidBody constructs the needed score states internally, doing away with the need for the create functions (which have been done away with). In addition there are new decorators IMP::core::Cover and IMP::core::Centroid which manage their respective score states, further doing away with more create methods. The hierarchy-based create method for rigid bodies has been moved to the atom module and is called IMP::atom::rigid_body_setup_hierarchy(). - IMP::em::read_map() will figure out the file type to use automatically from the extension. Other such methods should follow suite. - A class IMP::core::StructureSet has been added to provide a convenient way to manage solutions of optimization runs. - IMP::core::RigidBodyMover now takes its angle argument in radians and is documented as such. - IMP::core::ClosePairsScoreState and IMP::core::CloseBipartitePairsScoreState have been replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairContainer. The score states have been marked as deprecated and are mostly still there, but are more limited than they were. - IMP::core::AllPairsPairContainer has been renamed to IMP::core::AllPairContainer and a bipartite version has been added. - All containers now keep track of what they looked like on the previous time step. - The IMP::core::ListPairContainer and IMP::core::ListSingletonContainer have had their interface reduced a bit in order to make implementing change tracking easier. - The set of exceptions has been simplified to go along with simplification of the checks. We now have IMP::InternalException and IMP::UsageException which are thrown by the respective checks (and the former is thrown by IMP_FAILURE()). When you want to to pass an exception back to Python use one of IMP::ValueException or IMP::IndexException or IMP::IOException. There is also a new macro, IMP_THROW() to be used when you want to throw an exception with a nice error message. Please look at code you have worked on and make sure that the exceptions make sense given the new clarifications. - An IMP::ModelException is now thrown when IMP::Particle attributes attain invalid values. This exception can be caught to recover from failures of IMP::core::ConjugateGradients or difficult restraints. ## Week of October 15 - The set of functions needed to get meta information about IMP::Restraint and IMP::ScoreState objects has hopefully stabilized. Sorry about the thrashing. If you have restraints or score states not in SVN, see the corresponding IMP_RESTRAINT type macros, the IMP::Interaction class and base classes for what exactly you need to implement. ## Week of October 8 - IMP::Interaction was modified to split IMP::interaction::get_used_particles() into IMP::interaction::get_read_particles() and IMP::interaction::get_write_particles(). This makes it possible to check for IMP::ScoreState objects being in an inconsistent order (so that the input to one is updated after it is read). Later, we will be able to automatically order the score states ensure that there are no problems. - IMP::atom::Hierarchy has been modifier to get rid of the IMP::atom::Hierarchy::Type. Instead, all types are contained in the decorators. IMP::atom::get_by_type() was also updated to reflect this. - IMP_check() and IMP_assert() have been replaced with IMP_USAGE_CHECK() and IMP_INTERNAL_CHECK() respectively. Hopefully this makes it clearer when to use which. IMP::set_check_level() has been revised accordingly (to use IMP::USAGE and IMP::USAGE_AND_INTERNAL). Also, IMP_failure() and IMP_not_implemented() have been renamed to IMP_FAILURE() and IMP_NOT_IMPLEMENTED() to make all macros be all caps. To update external code, - replace IMP.EXPENSIVE with IMP.USAGE_AND_INTERNAL - replace IMP.CHEAP with IMP.USAGE - replace IMP_assert and IMP_check as above. ## Week of October 1 - IMP::ScoreState and IMP::Restraint now better track the set of particles that they use and the dependencies between the particles. To this end, a new base class IMP::Interaction has been added with methods to get the inter-particle interactions and the set of particles used. - .cpp, .h, .py and SConscript files which contain the string "DO NOT COMMIT" will be rejected on checkin. Use this to make experimental or debugging changes you are making so you don't accidentally commit them. - The search module is no longer built by default and is not longer supported. It will get moved out of svn shortly. - IMP::ScoreState objects now also have IMP::Interacting::get_interacting_particles() methods. Soon IMP::PairScore and IMP::SingletonScore objects will too. ## Week of September 27 - Some support for gather statistics has been added to IMP::Model (see IMP::Model::set_gather_statistics()). If requested, an API can be provided so that the individual data points can be returned to programs. - A class IMP::core::RigidBodyDistancePairScore has been added which provides accelerated computations for the distance between two rigid bodies. - The SWIG input files were renamed from swig.i to swig.i-in to aid in dependency tracking. - rigid bodies now have two sets of particles associated with them in the various places they are used. One is the IMP::core::RigidBodyMember particles, as was previously the case. The other is defined through a refiner (and could be the RigidBodyMember particles if a IMP::core::RigidMembersRefiner is used). The idea is that you sometimes want some particles which are transformed along with the rigid body, but which are not really part of the representation of the rigid body. Examples would include marker points or multiresolution representations of proteins. ## Week of September 20 - all \c .i files have been renamed to start with \c IMP_ to help make dependency tracking easier. - static builds (builds with \c static=True) build both the static and dynamic libraries and just static executables - the \c profile build target has been removed as \c static=True does the important bit of what is needed. Just do \c linkflags=['-pg'] to add the needed link flag for \c gprof. - Restraints should now implement IMP::Restraint::unprotected_evaluate() instead of IMP::Restraint::evaluate(). The latter can now be called outside of IMP::Model::evaluate() and will return the right answer always. - \c bin/imppy.sh no longer works, as was previously warned. Use \c tools/imppy.sh instead. The bin directory will go away eventually, for now the script just tells you to use the tools version. ## Week of August 30 - A new score IMP::core::BoundingBox3DSingletonScore has been added to keep particles in a box. - str should now work better in Python when applied to IMP::Object objects and a few others. Objects should use the new IMP_SHOWABLE and IMP_SHOWABLE_INLINE macros instead of declaring show functions directory to make it so that they work in Python. It is now a bug to \c print or \c str to not return something useful in Python for any IMP object. ## Week of August 23 - each module now has a version, eg IMP::get_module_version_info() which is also returned by the various restraints of the module. The Python code checks that loaded libraries have the expected version numbers to try to provide better messages when runtime linking problems occur. - the various bounding_box methods have been renamed to ::get_bounding_box() to be more consistent with the other names. - Python documentation is now parsed by doxygen ## Week of August 15 - The Object::get_version_info() result now includes the SVN revision info when building from SVN. This means you can use it to more easily tell exactly what a built version of IMP was built from. ## Week of August 8 - Treatment of hetatoms in hierarchies (and hence pdbs) is now documented. See IMP::atom::Hierarchy. This involved broadening IMP::atom::Residue to handle ligands and having IMP::atom::Hierarchy::LIGAND, IMP::atom::Hierarchy::AMINOACID and IMP::atom::Hierarchy::NUCLEICACID be handled by IMP::atom::Residue. - IMP::atom::read_pdb() currently drops HOH ATOM records on the floor. We should add a IMP::atom::Water decorator if anyone wants to handle them (since they aren't atoms and don't have elements). - IMP::atom::Mass() has been added for decorating particles with mass. IMP::atom::Atom mass has been moved to that. - The scheme for generating atom types is now fully specified. See the IMP::atom::AtomType documentation. - IMP::atom::Atom and IMP::atom::Residue have been somewhat restructured. Some of the helper functions have been renamed to be more descriptive or removed (if they duplicated other easily accessible functionality). get_residue_index() has been removed (since you can just call get_index() on the return result of the IMP::atom::get_residue()). IMP::atom::get_chain() now returns an IMP::atom::Chain. root() was renamed to IMP::atom::get_root() for consistency. - SWIG files are now generated from a file called swig.i in the pyext directory. This file only has to contain module-specific commands such as lines to wrap headers and instantiate module templates. No more boilerplate. - The contents of bin are moving to tools since the two directories both contained assorted development tools (not all of which were executable). For now, bin/imppy.sh will still work, but that will go away soon. - IMP::Decorator::show() now just takes the stream (it used to take a prefix also, which was often ignored). Anyone who implemented a decorator will have to delete the second argument of their show function. ## Week of August 1 - Containers now have methods IMP::SingletonContainer::evaluate(), IMP::PairContainer::evaluate(), IMP::SingletonContainer::apply() and IMP::PairContainer::apply(). Using these can be significantly faster than looping over the particles externally. Similarly, there is IMP::SingletonScore::evaluate(), IMP::PairScore::evaluate(), IMP::SingletonModifier::apply(), IMP::PairModifier::apply(). - The decorator functions have been renamed to IMP::Decorator::setup_particle(), IMP::Decorator::decorate_particle() and IMP::Decorator::particle_is_instance(). A shell script \c tools/update-decorators is provided which updates the names in all code in subdirectories of the current directory. Please backup things first. - a function, IMP::Restraint::safe_evaluate() has been added to be used when requesting the value of restraints outside of the IMP::Model::evaluate(). ## Week of July 18 - The atom::BondPairContainer has been renamed to atom::BondPairFilter since it wasn't a real container. A real atom::BondPairContainer has been added. - Out of source builds mostly works. To do an out of source build, make a link to \c scons_tools and \c SConstruct from the source directory in the build directory and then add the line \command{repository="../path_to_src_dir"} in the \c config.py in the build directory. Then running \c scons in the build directory will put generated files there. This allows you to make both a \c release and \c fast build off of the same set of sources. Proper docs will be added soon. - The SConscript setup for a module has been greatly simplified so it should be easy enough for users to maintain their own module and updates to the build system shouldn't involve changes to user SConscripts any more. This does mean that any non-svn modules need to be updated. Look at IMP.core or another module and copy what is there. Again, proper docs will be added soon. ## Week of July 11 - Per Javi's suggestion, the conventions have been split into "Coding conventions" and "API conventions" and linked from the main page. - benchmarks got moved to their own module, IMP::benchmark and will now be tracked over time ## Week of July 4 - FilteredListSingletonContainer and FilteredPairSingletonContainer went away as there was too much hard to factor code shared with IMP::core::ListSingletonContainer and IMP::core::PairSingletonContainer. The IMP::core::ClosePairsScoreState is now handled by a list of filters on the class itself. That is, do IMP::core::ClosePairsScoreState::add_close_pair_filter() to add a filter. - Hooks have been added to allow actions to be taken on an assertion or check failure. See IMP::HandleFailure and its descendants for more details. - IMP::display::CGOWriter has been much improved and now allows provides lots of display-time control and easy building of animations. - Generic optimizer states have been added (IMP::core::SingletonsOptimizerState and IMP::core::PairsOptimizerState). ## Week of June 15 - box distributions (IMP::algebra::random_vector_on_box(), IMP::algebra::random_vector_in_box()) take bounding boxes - IMP::em::DensityMap are now IMP::Object objects. Some current allocations on the stack may need to be changed (since ref counted objects can't be stack allocated). - IMP::em::DensityMap::Read() is now deprecated (as is IMP::em::DensityMap::Write) as they violated a slew of conventions. Use IMP::em::read_map() and IMP::em::write_map() instead. ## Week of June 8 - new methods IMP::atom::add_radii() and IMP::atom::add_bonds(); - Hierarchys to IMP::atom::Hierarchies - IMP::Particles has been changed to a new type. It no longer plays quite so well with Python lists (you can't create one from an arbitrary Python list). - The containers for multiple decorators have been revised to make them more useful. See IMP::Decorators. You can now implicitly convert, for %example, a IMP::core::XYZs to an IMP::Particles (as well as use IMP::Decorators::get_particles()). All methods that take a collection of particles with a certain type now instead take a IMP::Decorators. - A method that has expectations about the particles it takes should take decorators rather than particles. That is, anything which expects IMP::core::XYZ particles, should take an IMP::core::XYZs instead of IMP::Particles. - IMP_LIST() now takes an extra argument (the name of the plural container) - In C++ IMP::Decorator objects are implicitly convertible to IMP::Particle* objects. Unfortunately, making this work in Python is problematic. - More methods take and return decorators rather than IMP::Particles. This means it is clearer what their expectations are and what they produce. ## Week of May 15 - The IMP::helper module was added and various functionality moved there. See its documentation page for a list of what is there. - IMP::algebra::SphericalCoords was renamed to IMP::algebra::SphericalVector3D to be consistent with other algebra classes. - IMP::algebra::Rotation2D is now IMP::InvalidDefault to be consistent with the other similar classes. - We are trying to be a bit more specific about what is expected from a IMP::atom::Hierarchy. Towards that end, there are two methods for extracting a representation from a hierarchy, IMP::atom::get_detailed_representation(), and IMP::atom::get_simplified_representation() - A method has been added to produce a simplified copy of an IMP::atom::Hierarchy. The method is still a bit crude and could benefit from other people inspecting the output and saying how it should be improved. - There is now an IMP::algebra::BoundingBoxD (and IMP::algebra::BoundingBox3D) to track bounding boxes. There are methods like IMP::algebra::bound_box(Sphere3D) which gets the bounding box of a sphere. - Each IMP::Object has its own log level. If IMP::Object::set_log_level() is called with a value other than IMP::DEFAULT, the value is used for code within and called by the object. To support this in your objects you need to add IMP_OBJECT_LOG to the start of each major method. It has been added to some of the core objects. ## Week of May 8 - IMP::atom::MolecularHierarchy has been renamed to IMP::atom::Hierarchy. Use \command{sed -i.old "s\#MolecularHierarchy\#Hierarchy\#g" my_files} - the long-deprecated Name(Decorator) was removed - The decorator suffix has been removed from decorator names. - Unit conventions have been established and documented for distance, energy, force and angle. - there is a function IMP::atom::create_protein() to create a coarse grained protein. The function is crude so far, but the idea is that we should have standard ways of generating such things so that everyone can benefit from best practices. - a class IMP::atom::BondGraph has been added so Boost.Graph algorithms can be used with bond graphs - the IMP-specific doxygen commands have been documented. - the doxygen commands unstable{name} and untested{name} have been added to doxygen to be used for marking methods and functions as likely to change or not well tested. Pass the name of the class or method as name. - bonds can now be added to molecular hierarchies automatically see IMP::atom::add_bonds() ## Week of May 1 - There is a IMP::Rotation2D - You can now extract fixed Euler ZYZ angles from a IMP::Rotation3D. This code can serve as a template for extracting other Euler angle conventions if they are needed. - IMP::algebra::rotation_between_two_vectors() changed to IMP::algebra::rotation_taking_first_to_second() to make it clear which direction the rotation goes. ## Week of April 27 - IMP::Decorator derived objects are now convertible to bool. So code like \code if (d) { blah } \endcode works. - IMP::Decorator::cast now returns a null decorator (\c DecoratorName()) rather than throwing an exception. This might make it actually useful. - New IMP::statistics module for statistical methods (e.g. clustering). ## Week of April 20 - Support for taking advantage of rigid bodies in collision detection has been added via the IMP::core::RigidClosePairsFinder. - All IMP::core::ClosePairsScoreState-related functionality has been optimized and should be noticeably faster. - The EMBED project has been merged into the IMP::em module. Thus, it is no longer necessary to obtain EMBED separately to build IMP::em. - All C++ methods which work with std::istream or std::ostream arguments should now be usable from Python: any Python file-like object can be used. - The SWIG interface now ties in to the reference counting mechanism used in the C++ code, for all classes derived from IMP::RefCounted. So, for %example, deleting a Python object should not free the underlying C++ object if another C++ object holds a reference to it, and unrefing a C++ object should not delete it if a Python object still holds a reference to it. There should be no further need to manually keep references to Python objects to prevent things from being prematurely cleaned up. This should be transparent to the user; please report any problems. - IMP::atom::AtomDecorator no longer inherits from IMP::core::XYZDecorator. ## Week of April 13 - The decorators IMP::atom::AtomDecorator, IMP::atom::ChainDecorator, IMP::atom::ResidueDecorator etc. now inherit from IMP::atom::MolecularHierarchyDecorator. This means they can set their IMP::atom::MolecularHierarchyDecorator::Type properly and saves a lot of typing in the common case. We can always make hierarchy decorators zero memory overhead if we find that the cost matters. - The base classes IMP::Comparable, IMP::NullDefault, IMP::ValidDefault and IMP::UninitializedDefault were added to document what the default constructors do for various objects in IMP. ## Week of April 6 - IMP::Object now defines pure virtual methods IMP::Object::get_version_info() and IMP::Object::show(). All non-abstract classes which inherit from IMP::Object must implement them. Doing this means they are more uniformly implemented and are documented in one place. - IMP::Object is now reference counted (the only objects this changes are IMP::Model and IMP::Optimizer). As a result optimizers and IMP::Model objects must be created on the heap. - IMP::RefCounted has had its name changed and no longer inherits from IMP::Object. This makes it easier to use independently. - the docs for reference counting (IMP::RefCounted) and and IMP::Decorator have been improved - running doxygen produces no warnings. Let's try to keep it that way. Run \command{scons doc > /dev/null} before you check in large blocks of new code to make sure that you didn't introduce warnings. - Helper classes have been added to make SingletonModifier and PairModifier objects look like functors for C++ and Python. This makes it so you can use \c map in Python and \c for_each in C++. See IMP::SingletonFunctor and IMP::PairFunctor for more information. - The recent changes page has moved to a doxygen page to make it easier to link changes back to the relevant source code. ## Week of March 20: - create_rigid_body(), create_rigid_bodies(), create_centroid() etc. now require you to add the score state to the model yourself. Whether the rigid body is enforced or not can be controlled by adding or removing the returned IMP::ScoreState. - All previously deprecated code has been removed. This includes indexes in particles, the no-argument constructor for particles and the functionality that was moved from core to atom. - The IMP::core::RigidBodyDecorator and IMP::core::RigidMemberDecorator have been simplified by removing the traits classes. There is now only one type of rigid body and it stores the list of member particles internally. - IMP::core::NameDecorator has been deprecated as IMP::Particles have built in names (use IMP::Particle::get_name(), IMP::Particle::set_name()). - IMP::atom::ChainDecorator has been added ## Week of March 13: - A new IMP::display::Writer, IMP::display::ChimeraWriter has been added which exports geometry to native Chimera constructs. It probably should be used instead of bild and cmm for most purposes as it combines the best of both types. - compose() for transformation is now a free function so compose(rotation, vector) will work via promotions. - the geometry classes have been cleaned up and are now more consistent with IMP standards (and more likely to do what their names/comments suggest). See IMP::algebra::Sphere3D, IMP::algebra::Cylinder3D, IMP::algebra::Segment3D, IMP::algebra::Plane3D etc. - ParticleRefiner has been renamed to IMP::Refiner since the Particle was long and unnecessary. €eIF1•eIF3 translation initiation complex](http://salilab.org/40S-eIF1-eIF3). - IMP::domino::DiscreteSampler::set_restraints() now allows the set of restraints used in DOMINO filtering to be controlled (previously, all restraints in the IMP::Model were used). - IMP::modeller::IMPRestraints now transfers an IMP::ScoringFunction into the MODELLER score, not the implicit (and deprecated) score of the entire IMP::Model. - Bugfix #775: coordinates of voxels in EM maps were previously inconsistently treated (they are supposed to be the centers of the voxels, but in some cases they were the lower left edge). The fix may result in maps and their bounding boxes appearing shifted by half the voxel size. # 2.4.0 - 2015-02-27 # {#changelog_2_4_0} - Python 3 is now supported. The Windows package now works out of the box with Python 3.3 or 3.4; a `-python3` subpackage is provided for Ubuntu and Fedora. (Apple does not yet provide Python 3 so the Mac package is not built with Python 3 support, but if you install Python 3 you can compile IMP from source to work with it.) - A new command line tool, `rrt_sample`, allows for running rapidly exploring random tree (RRT) based sampling, e.g. to sample conformations of multiple domains connected by flexible linkers. - Command line tools are now placed in the `bin` directory of modules (the old `bin` directory, used for utility programs that are not installed, is now called `utility`). The `applications` directory is removed. - Several Python modules that were only used by EMageFit have been moved from IMP::em2d to IMP::EMageFit. - A new tutorial that uses the IMP::pmi module is now available (modeling of the RNA Polymerase II stalk). # 2.3.1 - 2014-12-01 # {#changelog_2_3_1} - Minor bug fixes. - Update to IMP::pmi to support the Nup84 and rnapolii biological systems. # 2.3.0 - 2014-10-24 # {#changelog_2_3_0} - A new IMP::pmi module was added, containing high-level Python classes to simplify the representation, scoring, sampling and analyzing of models in IMP. This module is still under heavy development, and should be considered experimental. - Binary installers for 64-bit Windows are now available. - The `.py` extension has been removed from all Python applications, so that they are consistent with the C++ applications. - IMP::atom::Selection objects can now be combined using basic set operations (union, intersection, difference, symmetric difference). In Python the equivalent operators (|, &, -, ^ respectively) can be used. - The integrative docking (idock) application now employs SOAP to improve the scoring. - IMP::atom::CHARMMParameters::create_topology() now works with a wider range of hierarchies (previously it would only accept Residues that were direct children of Chains). For example, it should do the right thing when given a Hierarchy that contains no Chain particles, or one where Residues are grouped under Fragments. - Only Python 2.6 or later is now supported. We still provide IMP for RedHat Enterprise 5 (which ships with Python 2.4) but you will need to use the `python26` binary (part of EPEL) rather than the system default Python. # 2.2.1 - 2014-08-11 # {#changelog_2_2_1} - Bugfix: PoseScore (see the ligand_score application) should now match that used by the Pose&Rank web server (http://salilab.org/poseandrank/) # 2.2.0 - 2014-04-09 # {#changelog_2_2_0} - A module IMP.template was added to act as a template for creating new external modules. It includes setting up git hooks for format checks and things like that. - The Python support for grids was cleaned up. Hopefully this makes it more robust. If you were using a grid in Python that no longer exists, ping us, we can add it back. - The generic geometry functions were simplified slightly and renamed to IMP::algebra::get_vector_geometry(). - The IMP::base::Flag class was added to simplify adding flags in C++. - The `--log_level` and `--check_level` command line argument now take strings (eg VERBOSE) to specify their values, rather than cryptic (and a bit unstable) numbers. - Command line options are now divided into normal and advanced ones. You do `--help_advanced` to show the advanced ones. - Support for computing Connolly surfaces was added with IMP::algebra::get_connolly_surface(). In addition, there is now IMP::algebra::get_uniform_surface_cover(const IMP::algebra::Sphere3Ds&,double) to sample the surface of a collection of balls. - IMP::atom::reate_simplified_from_volume() was added to create a simplified representation of a hierarchy that preserves the surface are and occupied volume. It uses IMP::algebra::get_simplified_from_volume(). If you want to play with such representations, you can use the rmf_simply application. - IMP::atom::get_rmsd(IMP::atom::Selection, IMP::atom::Selection, const IMP::algebra::Transformation3D&) was deprecated. Use IMP::atom::get_rmsd_transforming_first() instead. - IMP::algebra::get_rmsd() and IMP::algebra::get_rmsd_transforming_first() were added. - The family of macros IMP_BASE_[SMALL,LARGE]_[UNORDERED,ORDERED]_[MAP,SET] have been added (eg IMP_BASE_SMALL_ORDERED_SET ()). They provide platform and boost-version independent ways to take advantage of the best set/map for an application. In particular, they use boost::container::flat_set and kin when available. - IMP::atom::State and IMP::atom::Representation decorators have been added to provide support for multiple states and multiple representation schemes. They are integrated with RMFs and IMP::atom::Selection. See atom/multiresolution.py and atom/multistate.py for examples. - The container implementations has been cleaned up and simplified a bit. This should not change most usage. - [RMF](http://salilab.github.io/rmf) has been updated. There are a few minor API changes, in addition to a new, faster format (`.rmf3`). - [Eigen](#Eigen) is now included with IMP. - The deprecated function IMP::kernel::Optimizer::set_restraints() was removed. Use IMP::kernel::Optimizer::set_scoring_function() instead. - The deprecated constructor for IMP::kernel::ModelObject that does not take an IMP::kernel::Model was removed. This propagates to constructors IMP::kernel::Restraint, IMP::kernel::ScoreState and kin. Make sure you pass an IMP::kernel::Model as the first constructor argument. Related to this, deprecated functions like get_is_part_of_model(), set_model() were removed as they don't serve any purpose. - The deprecated methods in IMP::kernel::Optimizer that pertain to attribute optimization were removed. Inherit from IMP::kernel::AttributeOptimizer instead if you want to use them in your optimizer. - Some deprecated functionality was removed. If you encounter problems, replace IMP::base::OwnerPointer with IMP::base::PointerMember, find logging stuff in IMP/base/log.h, object in IMP/base/Object.h and random in IMP/base/random.h. - IMP::atom::Chain now uses strings for the chain ids, to be consistent with changes in the PDB. PDB files are written using the first character of the chain string. - Added orientation-dependent SOAP scoring (IMP::score_functor::OrientedSoap and IMP::atom::OrientedSoapPairScore). Orientation-dependent SOAP scores include those that score loops (SOAP-Loop), protein-peptide interfaces (SOAP-Peptide) and proteins (SOAP-Protein). The library files themselves are rather large (~1.5GB) and so are not included with IMP. They can be downloaded separately from http://salilab.org/SOAP/. # 2.1.1 - 2013-11-04 # {#changelog_2_1_1} - Added support for Mac OS 10.9 # 2.1.0 - 2013-10-24 # {#changelog_2_1_0} - There is now some limited IMP-wide gathering of timing statistics. You can add `--statistics=filename` to the command line or use the IMP::base::show_timings() function to view a summary of where IMP has spent time. - Slack was being used inconsistently (and not always correctly) among the various containers and restraints. If you had heavily optimized the slack value for speed, you may want to revisit it. - Class name OwnerPointer was switched to PointerMember, to denote more clearly a ref-counting pointer that is a class member. - The dependency handling in IMP has been rewritten to simplify it and fix a bunch of bugs. There are now more checks of correctness, so behavior that might have silently worked (or not worked) before may now trigger errors. In particular, restraints/score states must have access to the IMP::kernel::Model to do much of anything, so either use constructors that take the model or call IMP::kernel::ModelObject::set_model(). Also, changing your set of inputs or outputs without calling IMP::kernel::ModelObject::set_has_dependencies(false) is now likely to trigger a check failure (it would just silently do the wrong thing before). - Decorators have been cleaned up to make them behave consistently regarding IMP::kernel::Particle vs IMP::kernel::ParticleIndexes. This change shouldn't break existing code, but decorators should move to the new macros IMP_DECORATOR_METHODS(), IMP_DECORATOR_SETUP_0() etc. - The IMP_GRAPH() and IMP_WEIGHTED_GRAPH() macros got an extra argument describing how to show the vertexes in the graph. They also now expose a method, eg IMP::kernel::show_as_graphviz() that writes the graph in graphviz format to an IMP::base::TextOutput. - Support and rules for deprecating code in IMP have been updated. There are new macros and instructions on how to use them that should result in much better visibility about changes. See the C++ wiki for IMP for more information. And there is now an encoded policy of keeping deprecated things around for 1 release after deprecation. - IMP is now much more selective about when it invalidates the dependency graph. See [Dependencies](http://github.com/salilab/imp/wiki/Dependencies) for more information. This change should make evaluation of isolated restraints much faster in certain situations where they were unexpectedly slow before. - IMP::atom::Selection has been completely rewritten as it was very brittle. The set of particles returned may change in some cases, in some cases because it was buggy before, in some cases because of new bugs. - Reminder: the IMP_PROTECTED macros didn't work out well, and are now deprecated. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading `_` on the method name, if you get `Swig director pure virtual method called IMP::kernel::PairScore::do_get_inputs` messages. - The generated `cmake` files are no longer stored in the repository. Instead `tools/build/setup_cmake.py` is run after checkouts and other git events that change the repository tree (assuming you have run `setup_git.py`). - Generation of documentation using `doxygen` has been refactored. It is now done on a per-module basis, via targets like `IMP.core-doc`. The main advantage is that changes to the doc for a module can be tested comparatively quickly, just by building that module's docs. On the down side, docs are no longer quite so heavily interconnected, so dependencies only work to things that the module actually depends on. - The restraint and score state statistics functions were removed from IMP::Model. They had been broken for quite a while, with no one complaining. Statistics can be added back easily if requested. - Added an IMP-specific fork of `git-flow`. It handles things like `README.md` files when you start and finish feature branches and provides a tool to nicely update IMP and display the change log. The main commands of use are `git imp feature start <feature_name>`, `git imp feature finish <feature_name>`, and `git imp update`. You may want to remove the `git-flow` lines from your `.git/config` file as you may accidentally type `git flow` instead of `git imp`. You need to run `setup_git.py` to set things up. - [RMF](http://salilab.github.com/rmf) is now included as a git submodule rather than copied into the IMP repository. You should rerun `setup_git.py` to make sure submodule stuff is initialized. # For IMP 2.0 and earlier # {#changelog_2_0} ## Spring 2013 - Modules and applications now have a _Maintainer_ in their `README.md`, giving the github id of the person to whom issues should be assigned. - `setup_git.py` supports the `--module` flag which will set up git hooks and things for a git repository that contains a module. - To clean up for the release, the module dependencies are now described in a file called `dependencies.py` and the module docs in a `README.md`, both in the main module directory. - IMP-wide support for progress bars was added with the IMP::base::set_progress_display() function and helpers. These are only displayed when the IMP::base::get_log_level() is IMP::base::PROGRESS. ## Winter 2012-2013 - Scons support was removed. Use cmake to build IMP, it is way better. - Examples are run with a `--test` argument when IMP is built with internal checks. Expensive examples should check this flag and shorten what they do in order to try to test all their code without taking too long. - The class IMP::core::NonRigidMember was added to allow particles that are not rigid but whose coordinates are specified via internal coordinates in some IMP::algebra::Reference3D that is part of an IMP::core::RigidBody. - The IMP_BIG_ENDIAN and IMP_LITTLE_ENDIAN preprocessor symbols are deprecated in favor of the BOOST_BIG_ENDIAN and BOOST_LITTLE_ENDIAN symbols. - Support for asymmetric moves was added to IMP::core::MonteCarlo. To do this, IMP::core::MonteCarloMover was added and IMP::core::Mover and IMP::core::MoverBase were deprecated and implemented in terms of the new class. This should require few or no changes in code that uses Monte Carlo support or implements movers (although movers should move to IMP::core::MonteCarlo). All IMP::core::MonteCarloMover objects keep track of statistics (number of moves and number of accepted moves), rather than that being done in the IMP::core::SerialMover. In addition, the poorly defined move probability was removed from IMP::core::MonteCarlo. - IMP::base::WarningContext now outputs warnings immediately. This change was needed as the previous behavior resulted in occasional core dumps when the libraries for log support were unloaded before the libraries where static IMP::base::WarningContexts were declared. And there was no easy way to detect such problems. - You can now (probably) build IMP with [CMake on IMP wiki](https://github.com/salilab/imp/wiki/Cmake). - The maximum logging level and check level can now be controlled independently of the build mode. To, for example, enable progress level logging in a fast build, add maxlog="PROGRESS" to your config. The maximum check level for `release` has been reduced to IMP::USAGE as that speeds things up a bit. - The way applications are structured and specified has been regularized and simplified. Python files are automatically copied/installed and binary executables are compiled from each found `.cpp` file linked with any `.cpp` files contained in a subdirectory called `lib`. Nothing needs to be put into the SConscript file. - The Python constants like IMP.algebra.has_cgal went away as swig autogenerates ones like IMP.algebra.IMP_ALGEBRA_HAS_IMP_CGAL. While the latter are a bit uglier, the constants aren't used much and having them be auto generated is nice. - There have been some major changes to the build system. The main one is that the SConscripts in the directories below the module directory aren't really used. They are being removed from the repository. Now everything really is off of file locations (eg all test_* in test is a test, each *.cpp in bin becomes a program etc). This greatly simplifies the build system, which has, historically been a disaster. You can now define dependencies by dropping files into a dependency subdirectory in your module, no need to mess with the SConscript. - Building IMP now always requires specifying a `repository` argument. This was always required for out of source builds, which is what one should be doing, but is now required for in source builds to, so as to make accidentally polluting your source with an in-source build harder. Use `repository="."` in that case. - The HDF5 functionality in RMF was reorganized. Specifically, the headers were moved into `RMF/HDF5`, the Python functionality into `RMF_HDF5` and the code into namespace `RMF::HDF5` and all mentions of `HDF5` or `hdf5` in the class and function names were removed. - Support for IMP-wide command line flags have been added to IMP::base in the base/flags.h file. They provide a simple command line parsing interface that can be used from C++ and Python. In addition, various functionality in IMP is automatically exposed on the command line when those functions are used. For example, one can control logging, checks or profiling. See IMP/base/flags.h for (some) more info. - OpenMP support in IMP is officially released. All tests pass. But there has been little specific testing so far. - Access to attributes using IMP::ParticleIndex and the IMP::Model now works from Python. See IMP::Model::add_attribute() and kin. - IMP::ParticleKey and IMP::ParticlesKey were renamed to IMP::ParticleIndexKey and IMP::ParticleIndexesKey respectively as that is what they store. Typedefs were provided to support old code. - Indexes have been added to the docs connecting classes with all the methods that return them, examples that use them and methods that take them. And all methods that are used in examples with the examples they are used in. Please check them out and report issues and ways to improve things. You can find links on the main page. - The docs for a modules/biological systems/applications should now be found in an `overview.dox` in the module `doc` folder. Modules in IMP svn have been updated. For external modules, you can just move a `overview.dox` from that was previously generated into your `doc` folder and (eventually) remove the lines from your `doc/SConscript`. This makes it easier to use the full range of doxygen markup, as well as removes the need to escape doxygen markup for Python. - Examples have now been simplified into full blown doxygen examples. This involves merging the `.py` and `.readme` files (putting `\#\#` in front of each line of the `.readme`) and adding a `## \\example module/examplename.py` line at the beginning. You can use `./tools/updaters/update_examples.py modules/mymodule` to update a module that was not already updated. - IMP::base::FailureHandler has been removed as it wasn't used. - Handling of data and includes has been simplified. All appropriate files in the corresponding directories are linked into build/installed. This is only likely to be a problem if you had broken headers sitting around `include`. If so, just comment them out, or move them some place nicer. - Modules now must be in a folder called `modules`, which must have a `description.py` file. - The build system should now be a little better about cleaning up after itself. - How headers/doc files/data/examples - The IMP_PROTECTED macros didn't work out well, and are now deprecated. Sorry. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading "_" from the method name. - Nice version numbers will now be generated if you use git. You need to rerun tools/git/developer_tools/init_git if you have an existing clone. ## Fall 2012 - 11/9: RMF is now an external library shipped with IMP. IMP will build it and install it automatically if you have cmake version 2.8 or higher installed. - 11/9: Updated support macros for IMP::Restraint and IMP::ScoreState classes have been added (IMP_RESTRAINT_2() and IMP_SCORE_STATE_2()) that make it easier to return dependencies on things other than IMP::Particle and IMP::ScoreState objects. In addition, the dependency functions on scores and modifiers have been updated similarly and standardized. In general this should not cause backward compatibility problems unless you were overriding the IMP::ModelObject::do_get_inputs() in IMP::Restraint and using the IMP_RESTRAINT() macro. In that case update to the IMP_RESTRAINT_2() macro. - 11/8: Support has been added for the C++11 keywords override (IMP_OVERRIDE) and final (IMP_FINAL). They allow you to tell the compiler to produce an error if the marked method does not override/implement a parent's method and if anything overrides a given method, respectively. - 11/8: IMP::SingletonModifier::apply_indexes() and IMP::SingletonScore::evaluate_indexes() now require an upper and lower bound index (you can just pass 0 and the size of the array). We can make them option if there is a good reason. - 11/8: IMP::SingletonDerivativeModifier and kin were merged into IMP::SingletonModifier and kin as the only difference, the presence of the IMP::DerivativeAccumulator (whose weight was always 1) wasn't very interested and the split added lots of complexity. If this has an impact on your code, let us know. - 11/8; Evaluation with containers was greatly simplified, moving many methods on containers. We can add backwards compatibility versions if that is interesting. - 11/7: Predicates, scores and modifiers written in Python need to update their methods for returning inputs and outputs. They should now have a method \c _do_get_outputs(self, m, particle_indexes) which should probably just return \c [m.get_particle(i) for in in particle_indexes] to return all the passed particles. - 11/7: It is now preferred that IMP::Restraint and IMP::ScoreState descendants report dependencies directly using IMP::ModelObject::do_get_inputs() and IMP::ModelObject::do_get_outputs(). To do this, you should use the IMP_RESTRAINT_2() and IMP_CONSTRAINT_2() macros instead. Methods like IMP::Restraint::get_input_particles() are now deprecated. This change was made as the old mechanism introduced a lot of extra complication and made it so that there was no way to directly depend on something that was not a Particle or a Container. - 11/3: IMP::container::InContainerPairFilter and friends now treat pairs as unordered. A flag was added to control this. - 9/5: IMP::base::Object classes now have a method IMP::base::Object::clear_caches() that clears all cache data in the object. This is not universally implemented, but now provides a correct path for doing that. The IMP_LIST() type macros call that method when their contents changes. ## Summer 2012 - 7/3: IMP_PERIODIC_OPTIMIZER_STATE() was superceded by the class IMP::core::PeriodicOptimizerState as it is rather cleaner than the macro. - 7/2: in order to make various base classes in IMP more Python friendly, macros IMP_PROTECTED_METHOD(), IMP_PROTECTED_CONSTRUCTOR() were added that properly expose such methods to Python. The standards checks now complain if \c protected is used an a class in the API. - 6/24: Executables from \c module/bin are now put in \b build/module_bin/module. This removes a recurring source of errors in the build script due to the origin and destination directory have the same name in scons. - 6/24: Benchmarks are now put int \c module/benchmark and build to \c build/benchmark/module to remove a recurring source of errors. - 6/18: The IMP::benchmark module supports the GPerfTools profilers when available. Benchmark executables (things in \c bin whose name starts with \c benchmark_) that use the IMP_BENCHMARK() macro have command line flags to turn on and cpu and memory profiling. - 6/12: The naming rules for .i files that are in module \c pyext directories changed. They should now be \c IMP_modulename.name.i instead of \c IMP_modulename_name.i. This is to better support underscores in names. ## Winter 2011/2012 - 20/4: A method, IMP::atom::read_pdb(IMP::base::TextInput, int, IMP::atom::Hierarchy), was added to load coordinates from a pdb file into an existing IMP::atom::Hierarchy. - 4/4: the filter types were superceded by the predicate types. You will only see a difference if you develop a filter class (eg IMP::PairFilter -derived class) or directly call methods on filters (specifically IMP::PairFilter::get_contains()). - 29/3: IMP::Particle::get_has_model() has been changed to IMP::ModelObject::get_is_part_of_model() to be consistent with other classes. - 28/3: IMP::core::Mover changed slightly, adding an extra method IMP::core::Mover::get_output_particles() and requiring the IMP::Model be passed to the constructor. All movers in IMP SVN were updated without any difficulty. This allows movers to use information contained in the dependency graph amongst particles so they can, for example, weight moves based on scores. - 26/3: IMP::SingletonContainerInput etc were added. Methods that take these as their arguments can be passed either containers or lists of particles (or tuples), as convenient. All methods that take IMP::SingletonContainer objects should eventually move over to taking this (changing just requires changing the argument from a pointer to an Input). - 22/3: Over the last week or so, which IMP headers include which other heads have been substantially simplified. This may result in compilation errors for code outside of \c svn that depended on IMP headers including other headers. In general, you want to explicitly include the header defining each thing you use. - 15/3: The vector used throughout IMP is now exposed at IMP::base::Vector and what it does it documented. - 10/3: IMP::domino::RestraintCache was introduced to provide a documented centralized place to handle restraint evaluation and caching in DOMINO. In the switch various bugs involving maximum scores and IMP::RestraintSet objects were fixed. - 10/3: Scoring functions in IMP are now implemented in IMP::ScoringFunction objects. All restraint evaluation is handled by them and they can cache various useful information to aid in accelerating restraint evaluation. Backward compatibility interfaces have mostly been provided, please complain if some important interface went away. This has greatly simplified the code as well as fixed various bugs in various types of restraint evaluation (particularly in the handling of maximum scores on IMP::Restraint and IMP::RestraintSet objects. - 29/2: The macros IMP_NAMED_TUPLE_1(),IMP_NAMED_TUPLE_2(),IMP_NAMED_TUPLE_3() were added for easy generation of simply struct for use in passing and returning values. The generated structs support comparison, output, hashing and have constructors. Use the macros instead of a boost::tuple or a a POD struct. - 16/2: Restraints now are evaluated with the correct weights when evaluated by themselves as opposed to always returning their weight as seen by the model as a whole. - 15/2: The IMP::algebra grid support was rearranged and simplified. Most of the changes should be transparent. However, the IMP::algebra::grids namespace got folded into IMP::algebra, which might break some code. And various (obscure) constructors of IMP::algebra::GridD were removed. All code in svn has been changed as needed. - 3/2: The IMP::domino HDF5 container has been revamped. This involves splitting it into a write version (IMP::domino::WriteHDF5AssignmentContainer) and a read version (IMP::domino::ReadHDF5AssignmentContainer). - 3/2: IMP::domino::WriteAssignmentContainer and IMP::domino::ReadAssignmentContainer were added to do direct binary file writing. This is faster than the HDF5 versions. - 3/2: The installation instructions have been revamped and the developer instruction, in particular, have been greatly simplified. - 24/1: \c tools/show-changes has been added that shows the changes in the history log since you last updated. - 24/1: \c tools/make-source has been added that adds a source/header pair to a module - 24/1: principal components analysis in IMP::algebra was generalized to work in any dimension. This required renaming the class to IMP::algebra::PrincipalComponentAnalysisD. - 24/1: the output of the IMP::algebra::PrincipalComponentAnalysisD::show() was changed to reflect the standards (values produce compact, single line output on show). The old show has been moved to IMP::multifit - 19/1: IMP::container::ConsecutivePairContainer had IMP::container::ExclusiveConsecutivePairContainer split off of it to make the two operating mode clearer and so that there could be a clear IMP::container::ExclusiveConsecutivePairFilter to use. - IMP::algebra::Grid::apply was added to provide faster access to all (or, later, part of a grid). It can be an order of magnitude or more faster than using the iterators when doing simple operations. - the scons config options \c pretty and \c color were added (defaulting to True). When they are true, build commands are mostly suppressed and a much briefer description of what is being done is printed out (in color). Note that \c color mode and compilation in emacs don't get along too well. - IMP::algebra::ParabolicFit was renamed to IMP::algebra::ParabolicFit2D (a backwards compatibility typedef is currently provided) - IMP::algebra::LinearFit was renamed to IMP::algebra::LinearFit2D. It now supports error bars on the y-values. ## Fall 2011 - IMP::compatibility::checked_vector was renamed to IMP::compatibility::vector and made available to IMP as IMP::vector. Typedefs should use this type (IMP::vector). - Various IMP::algebra plural typedefs which had been using std::vector (since they never were updated), have now been moved to IMP::vector. - IMP::statistics::HistogramD now supports display with \c matplotlib. - IMP::statistics::HistogramD supercedes IMP::statistics::Histogram - predicate base classes have been added: IMP::SingletonPredicate, IMP::PairPredicate etc. These take particles and return an integer. Current functionality using them includes IMP::core::PredicateSingletonScore that uses a predicate to choose which score to apply. - the core generic restraint and score state support was moved to the kernel to avoid breaking various scores outside of IMP svn. - the IMP::display module was refactored and moved up to be a dependency of IMP::core and IMP::atom rather than vice versa. The main observable results are that you should now use IMP::atom::HierarchyGeometry, IMP::atom::SelectionGeometry, IMP::atom::HierarchiesGeometry, IMP::core::XYZRGeometry, IMP::atom::BondGeometry etc instead of finding them in IMP::display. This change means that, display can be used internally when testing and debugging functionality in IMP::core and IMP::atom and that tests that use IMP::display don't need to build all of IMP::atom. - Decomposition of restraints now results in restraints that are faster to evaluate, using the generic restraint mechanism in IMP kernel. - IMP::algebra::get_unit_bounding_box_d() was changed to return a bounding box with radius 1 to be consistent with IMP::algebra::get_unit_sphere_d() - A general purpose scheme to associate data with tuples of hierarchy nodes was added to the RMF format. This deprecates the special purpose bond support. - the way various types of geometry is stored in RMF files was changed, breaking backwards compatibility. ## Summer 2011 - some functionality from the kernel has been spun off into a new module IMP.base. You probably won't notice anything. The only significant changes are that functions that accept or return IMP::base::TextInput and IMP::base::TextOutput must declare them with the \c base qualifier (eg base::TextInput) so that swig is happy. - the source for the kernel has been moved from \c kernel to \c modules/kernel. This is unlikely to effect anyone. - IMP::core::ExcludedVolumeRestraint now supports filters - IMP::domino::HDF5AssignmentTable now caches internally, which makes it several times faster than before. - you can now pick which Python to use with the \c python argument to scons. - the units for BD diffusion coefficients changed to be consistent with other IMP units. - IMP::domino::get_interaction_graph() now takes an IMP::RestraintsTemp, like other functions - The assert failure handling function was documented and renamed to IMP::handle_error() so it is easier to find when you want to use gdb. - to work around various issues with ref counting pointers and compiler eccentricities, we added emulation of the C++0x \c nullptr. You should prefer that to \c NULL in IMP code that you want to run cross platform. - IMP::IntsList, IMP::FloatsList and IMP::StringsList have been added to replace the various scattered solutions for passing such types. - IMP::Restraint::get_decomposition() and IMP::Restraint::get_current_decomposition() changed to return a single Restraint, instead of a list. This makes it easier to propagate the weight and maximum reliably. - IMP now supports usage of restraints without adding them to the IMP::Model scoring function. To do this, do IMP::Restraint::set_model(), passing the model. The restraint can then be evaluated. - IMP::domino::DominoSampler does not use the IMP::domino::OptimizeRestraints restraint optimization by default. Instead, if it is asked to create the IMP::domino::RestraintScoreSubsetFilterTable, it decomposes the restraints first (via IMP::create_decomposition()). This avoids several bugs and very fragile code. - the IMP::core::DiameterRestraint was ignoring the radii of the particles involved and so was giving artificially low scores. It now is on the actual diameter of this molecule. - the IMP::core::MonteCarlo optimizers now support incremental evaluation. This can be a lot faster when only a few particles are moved each step. - all the plural types in IMP have moved to bounds checked vectors when using debug builds with gcc. If you get a non-imp error about an invalid access, it is probably some code that walks off the end of an array. Please report it. - The IMP::core::MoverBase was changed to move it off of the IMP::SingletonContainer. This resulted in the IMP::core::BallMover and IMP::core::NormalMover having new constructors. Now create them from a list of particles, a list of float keys and a radius/standard deviation. This removes some complexity and inefficiencies as well is removes some scope for errors. - Temp and non-temp decorator lists have been merged in C++. This was done to simplify code since no one was using the non-temp variants. Now all decorator lists are temp (non-reference counted) variants. - restraints and restraint sets are now evaluated only using "canonical" weights. That is, each restraint has one weight for the model (the total over all the restraint sets it is contained in and will always be evaluated with that weight). This allows the evaluation of simple restraints to be faster. - The IMP_LIST() macros have been changed slightly. If your usage needs to take action upon addition, removal or changes to the list, use the IMP_LIST_ACTION() variant in your header. The IMP_LIST_IMPL() macro no longer takes the actions, and the last three (generally empty arguments) will have to be removed from the call to get it to work. This is needed to move towards more flexible and efficient passing of lists of values. - IMP::domino uses the new IMP::Model::evaluate_if_good() support to accelerate computations - IMP::core::MonteCarlo can use the IMP::Model::evaluate_if_good() support to more quickly reject very bad conformations. See IMP::core::MonteCarlo::set_maximum_difference(). - IMP::core::ExcludedVolumeRestraint has been rewritten to be faster. It is especially faster when used with IMP::Model::evaluate_if_good() - a new evaluation mode IMP::Model::evaluate_if_good() and IMP::Model::evaluate_if_below() has been added. It allows restraints to shortcut score computation when the score can be shown to be worse than a maximum value. - as part of moving towards more efficient use of evaluate_if_good() various deprecated methods have been removed from IMP::core::ExcludedVolumeRestraint and IMP::core::RigidClosePairsFinder. - incremental evaluation support was removed from IMP. The reason for doing this was that the code was complicated, created overhead even when incremental support was not used and was not commonly used. Further, decomposing the restraints and using the dependency graph yields similar results and is more transparent. - the RMF helper binaries have been moved to their own application so that they get installed. - IMP::display::Writer has been revised to better support writers that write multiple frames to a single file and ones that write binary data. The noticeable changes are that the existing IMP::display::Writer instances must be created with a file name. And the IMP::display::WriteOptimizerState just takes a writer rather than a write and a file name template. IMP::display::Writer::set_frame() has been added so that you can directly control the frame being written. ## Spring 2011 - An interactive mode has been added to IMP::domino::DominoSampler. The interactive mode is useful for understanding domino sampling runs and distributing domino computations across a cluster. - Functions IMP::domino::set_assignments() and IMP::domino::get_assignments() were added which support writing subset assignments to data fields in hdf5 files. These can be used to helping to distribute domino on a cluster as well as debugging the sampling process. - a Python function IMP.show_graphviz() has been added that attempts to render a graph with \c dot and then open the pdf. - The Python functions IMP.show_altgraph() and IMP.get_networkx_graph() have been added to provide alternatives for quick viewing of IMP graphs. Unfortunately, they still pale compared to graphviz. - IMP.atom.create_cover() was added to facilitate creating a summary particle for part of a hierarchy in order to accelerate restraints. - IMP.hdf5 was renamed to IMP.rmf, the proposed name of the file format that it supports. That also better separates the hdf5 support from the file format support. - The old I/O functions have been marked as deprecated in favor of IMP.hdf5. You may need to compile with deprecated="yes" if you use them. - The Python graph interface now has a remove_vertex method. - One can now access the model restraint evaluation statistics programmatically using IMP::Model::get_restaint_statistics(). Restraint statistics can be logged to a file during optimization using IMP::core::WriteRestraintScoreStatisticsOptimizerState. - IMP::domino::DominoSampler now allows you to specify the tree on which merges occur directly. Also, you now need to specify the merge tree rather than the JunctionTree if you want to do non-loopy inference and specify something about how it is done. Sorry, it makes the code a lot easier. - Preliminary symmetry support has been added via the IMP::core::TransformationSymmetry and IMP::core::TransformationAndReflectionSymmetry classes. ## Winter 2011 - There now is a common base class IMP::atom::Simulation for IMP::atom::BrownianDynamics and IMP::atom::MolecularDynamics. In merging them there were several minor changes made to both. The most noticeable may be that the function for setting the maximum time step length is now IMP::atom::Simulation::set_maximum_time_step(). The last time step length can be extracted by IMP::atom::Simulation::get_last_time_step() - IMP::core::MonteCarlo::set_temperature() was renamed to IMP::core::MonteCarlo::set_kt() as that is what it is doing. - A new module IMP.hdf5 has been added to provide more general support for saving and loading IMP::atom::Hierarchy structures to files. Multiple conformations of each hierarchy can be saved as well as information about geometry and restraints (although these can't be read at the moment, so they are not so useful). - IMP::algebra::grids::GridD now supports dense grids in arbitrary dimension through use of IMP::algebra::grids::DenseGridStorageD. - IMP::Optimizer now can use a list of restraint sets to compute the score rather than the complete Model score. - each IMP::Restraint now has its own weight. This is lightly tested. - A new log level IMP::PROGRESS has been added. The idea is that IMP::Sampler::get_sample() calls and other expensive loops can show a boost progress meter at this log level to give some idea when things are moving forward. - IMP::domino::TrivialParticleStates renamed to IMP::domino::IndexParticleStates to be less judgmental - IMP::domino::CompoundParticleStates and IMP::domino::RecursiveParticleStates added - IMP::domino::DominoSampler now supports maximum scores on IMP::RestraintSets (untested). - the name of the clustering methods have been changed to create_ from get_ since they create objects. - the names of the recently added metric based clustering support has been changed to Metric rather than Distance. - IMP now has the ability to have certain tests designated as expensive and to skip those tests when test-fast is run. The motivation for this is that the tests take a really long time making it impractical to run before every commit. And it is hard to tell which tests have not been run when running with the necessary multiple build processes. My thought is that tests that take about a second or less on a debug build should be left along (this is probably >95% of the tests), but the few that take longer should be marked as expensive. To do that, rename the test to expensive_text_XXX.py (from test_XXX.py) and add expensive_python_tests= env.IMPModuleGetExpensivePythonTests() to the IMPModuleTest call in the test SConscript. - the fuction IMP::display::create_restraint_geometry() was added to do what its name says. - runtime dimensional geometric objects have been added. These include IMP::algebra::VectorKD, IMP::algebra::BoundingBoxKD and IMP::algebra::SphereKD. The IMP::statistics::Embedding classes now use IMP::algebra::VectorKD instead of Floats. The sparse grids also support variable dimensions as to the nearest neighbor searches. - IMP::atom::DopePairScore has been added to IMP. - IMP::atom::CHARMMStereochemistryRestraint provides a high-level simplified interface to the bonded part of the CHARMM forcefield. - IMP modules are now built by first combining all the .cpp files into one, and then building that file. This greatly accelerates building of a module from scratch as build time is dominated by parsing headers. This does change the semantics of .cpp files as they see headers included by and functions declared by an arbitrary subset of other .cpp files. For now they are still required to be able to be compiled separately. To get the only behavior back either globally or on a per-module basis, see the percppcompilation parameter to scons. - Instructions and scripts have been added for building modules and things externally to the IMP build. See the installation guide. - IMP::atom::CHARMMParameters can now automatically map CHARMM-style atom and residue naming to PDB-style, so CHARMM topology files can be used unmodified. - IMP::atom::CHARMMTopology::add_sequence() provides a simple way to generate a topology from a string of amino acid one-letter codes. - IMP::atom::CHARMMTopology::add_coordinates() will generate atomic Cartesian coordinates from CHARMM internal coordinate information. - IMP::atom::CHARMMTopology::add_missing_atoms() and IMP::atom::remove_charmm_untyped_atoms() can now be used to ensure that atoms read from a PDB file match the CHARMM topology. - building of deprecated functionality is now false by default - the IMP_COMPARISONS(), IMP_COMPARISONS_1(), IMP_COMPARISONS_2() macros now take the name of the class as an additional (first) argument. This avoids the formerly needed, \c This typedef and makes them more consistent with the other macros in IMP. - modules can now be built externally to the IMP svn in addition to applications and biological systems. To do so, create a directory with your module in a subdir containing links to the IMP SConscript file and scons_tools and a config.py that has information needed to run IMP (eg an appropriate \c pythonpath, \c ldpath). - removed IMP::algebra::GridD::get_index(VectorD) as it wasn't very safe (in that vectors coordinates can get rounded between the IMP::algebra::GridD::get_has_index() and IMP::algebra::GridD::get_index() calls. Use IMP::algebra::GridD::get_nearest_index() or IMP::algebra::GridD::get_extended_index() or IMP::algebra::GridD::operator[]() instead. - renamed IMP::domino::ParticleStatesTable::get_particles() to IMP::domino::ParticleStatesTable::get_subset() as it is more useful to be able to get the subset containing all the particles. - rename IMP::restrainer::Restraint::print_all_restraints() to IMP::restrainer::Restraint::show_all_restraints() to be consistent with the rest of IMP. - renamed IMP::atom::get_conect_record_string() to IMP::atom::get_pdb_conect_record_string() to make clearer what it is. - class and method names are now spell checked. Add missing words to spelling_exceptions in the module test call. - the plural names (eg IMP.Particles()) have been removed from the Python side as they are just Python lists. - biological systems can now include internal modules. These provide a convenient place to put application specific C++ and Python code. See the example system for more info. - IMP_VALUES() now includes IMP_OUTPUT_OPERATOR() so you no longer need two namespace macros for each value. - the function IMP::core::Harmonic::k_from_standard_deviation was renamed to IMP::core::Harmonic::get_k_from_standard_deviation() to conform to the standards. - the IMP::atom::PDBSelector and IMP::atom::Mol2Selector became IMP::Object classes. All this means is that they need to be created with \c new in C++ code. It also means that one is no longer prohibited from storing them (not that there is much use in doing that). - the standards tests have been improved to test more useful aspects. Now class and function names are checked some as well as all modules have the checks for show and plural types. The error messages should explain how to fix things. - IMP::core::RigidBody::lazy_set_reference_frame() was renamed to IMP::core::RigidBody::set_reference_frame_lazy() to fit with the standards better. - the transform based methods on reference frame which have been deprecated for a while were removed - the hierarchy visiting and searching functions have been renamed to better fit the IMP naming scheme as we don't expect they were used outside of the IMP source. - IMP::atom::NonhydrogenMol2Selector was renamed IMP::atom::NonHydrogenMol2Selector to be consistent with the PDB version ## Fall 2010 - IMP::display::PlaneGeometry was added to display planes - the optimize methods on optimizers was moved to a "non-virtual interface". Usage has not changed, but optimizers should implement IMP::Optimizer::do_optimize() instead of IMP::Optimizer::optimize() to benefit from better checking. - the two-vector constructor for IMP::algebra::Plane3D now expects the second vector to be a unit vector as otherwise one could switch the order around without noticing. - the io functions in the kernel have been cleaned up a bit (the old ones are there for backwards compatibility). See IMP::write_particles(), IMP::write_particles_binary(), IMP::write_particles_to_buffer() and IMP::read_particles(), IMP::read_particles_binary(), IMP::read_particles_from_buffer() - IMP::core::WriteModelOptimizerState is going away as it duplicated IMP::WriteParticlesOptimizerState. - the IMP::algebra::Grid3D has been changed to make it support any (fixed) dimension when used with sparse storage. The template arguments have also been changed to remove a redundant one. See IMP::algebra::GridD. - the biological systems specification has been improved a little adding ability to run some rudimentary tests to make sure the scripts are ok and to not require that everything be run from the svn directory. - A new method IMP::atom::create_compatible_rigid_body() which creates a rigid body from a hierarchy which has the same internal coordinates as another rigid body. A way to make the rigid body construction more canonical would be nice (so we don't need this function), but I'm not sure how to do that. - Logging has been improved to better report the name of methods producing the log messages - a macro IMP_FUNCTION_LOG has been added which should be added to the start of non-member functions which produce significant logging output. - IMP::algebra::get_distance(const Rotation3D&,const Rotation3D&) was changed to use the Euclidean distance in R4 as that seems better behaved than the prior metric. - IMP.domino2 was renamed to IMP.domino and the old IMP.domino was moved - the documentation for classes and some methods now includes a list of all examples which use the class. Please report any incorrect links or phantom classes appearing in the docs. Unfortunately, functions have to be added individually. Please request any functions of particular interest. - applications and biological systems now have their own documentation pages - IMP::atom::Hierarchy requires radii on its leaves - IMP::atom::read_pdb add radii to the atoms - IMP::helper was merged into IMP::restrainer as all the functionality there was primarily to provide support for restrainer - IMP::domino2::SubsetEvaluatorTable was removed. Pass the model and IMP::domino2::ParticleStatesTable to the IMP::domino2::RestraintScoreSubsetFilterTable instead. - IMP::atom::Selection was added to allow specification of parts of IMP::atom::Hierarchy using biological sorts of names. A few functions to create restraints from them were added too. - a new example was added in the kernel to show IMP used on a biological system - IMP::display::PymolWriter now supports adding labels to things - IMP::display::PairRestraintGeometry and IMP::display::ConnectivityRestraintGeometry were added to aid in visualizing those restraints. - IMP::bullet was added with an optimizer to use hard body physics to resolve local collisions. - IMP::atom::setup_as_rigid_body() has been deprecated and replaced by IMP::atom::create_rigid_body(). The latter creates the rigid body as a separate particle so that the molecular hierarchy doesn't have to worry about having its bounding sphere large enough to enclose all the members. - IMP::core::KClosePairsPairScore was added to have count based cutoffs for close pair scoring. - IMP::algebra::get_surface_area_and_volume() was added to compute the two numbers for a set of balls ## Summer 2010 - The IMP::core::RigidDistancePairScore and IMP::core::ClosePairsPairScore have been revised to remove certain inconsistencies and ambiguities and to make them faster. Existing code using them will need to be slightly tweaked to work. - Javi added his 2D EM scoring functionality as IMP::em2D - some new ways to speed up scons have been added. See IMP::core::HarmonicDistancePairScore, IMP::core::SoftSpherePairScore and, for C++ users, the generic restraint support IMP::container::create_restraint() and IMP::core::create_restraint(). These each offer between 2x and 1.2x speedup when using simple scores (like harmonics on distances). In addition, a function IMP::container::get_slack_estimate() has been added which will estimate the correct slack to use with a IMP::container::ClosePairContainer. This will be extended to support IMP::core::ExcludedVolumeRestraint eventually. - support for dense and sparse grids have been added to IMP::algebra. See the IMP::algebra module page (I'm still working on how to make the docs clearer). - all write functions now have associated IMP::OptimizerState and IMP::FailureHandler classes. For examples, see IMP::WriteOptimizerState, IMP::WriteBinaryOptimizerState, IMP::WriteFailureHandler, IMP::atom::WritePDBOptimizerState etc. - functions IMP::write_binary_model(), IMP::read_binary_model() and IMP::read_configurations() have been added to support space and filesystem efficient storage of model configurations. Multiple configuration can be written to a single file. This functionality requires NetCDF be installed (NetCDF is a standard library for reading and writing binary data). - domino2 has been added. It provides a simpler interface on Keren's domino optimize as well as a number of new optimization and hooks for customizing the optimization process - a better explanation of how to use reference counted pointers has been added to IMP::Pointer. - all simple, planar polygons are now supported for display when CGAL is present ## Week of April 1, 2010 - All plural objects in Python are now simply Python lists rather than C++ recreations of Python lists. This makes compilation of the wrappers much, much, much faster. The visible changes are - types are not checked until the lists are passed to C++ - passing std::vector types by pointer and storing passed std::vector types by pointer will not work (but should never have been done anyway from code exposed to Python) - swig has issues with some overloads (not always for obvious reasons). These are being sorted out. None are commonly used. - get_particles() on a plural object doesn't work in Python (and is not needed in C++). Implicit conversions will always work. - All containers need to know the model at time of construction. If, after updating, you get an error on a container constructor, just add the model as the first argument. This also means that empty lists of particles in the constructors cause check failures. We can figure out a work around if this is too annoying. ## Week of March 13, 2010 - function get_rotation_in_radians_about_axis() was renamed to get_rotation_about_axis() as it didn't parse and we don't have the "in radians" elsewhere. - Added IMP::algebra::ReferenceFrame3D. As part of the cleanup, IMP::core::RigidBody::set_reference_frame() replaces set_transformation(). And the reference-frame related transformation construction functions were replaced. IMP::algebra::get_rotation_from_x_y_axes() was added to aid in the construction of reference frames (and replaces the algebra function taking three vectors). ## Week of March 6, 2010 - Publications now have a description field to give browsers an indication of what is in the publication to aid selection of what to read/cite. ## Week of Feb 27, 2010 - IMP::algebra::decompose_rotation_into_axis_angle() was renamed to IMP::algebra::get_angle_and_axis(). - renamed IMP::atom::bond() and IMP::atom::unbond() to IMP::atom::create_bond() and IMP::atom::destroy_bond() to be more consistent with things - renamed IMP::core::root() to IMP::core::get_root() ## Week of Feb 21, 2010 - IMP::Object::set_was_owned() was changed to IMP::Object::set_was_used() as that better reflects what the flag means. - various more interpolation and endian methods had their names fixed (by adding \c get_) - the function get_mean() on vectors was removed as it duplicated IMP::algebra::get_centroid() - build_Transformation3D_from_transformation3D was replaced by IMP.algebra.get_transformation_3d() - IMP::algebra::almost_equal() is now IMP::algebra::get_are_almost_equal() to fit the naming conventions. - the random vector generators have been cleaned up to remove redundant ones and make the names more consistent. They now have the form of either - IMP::algebra::get_random_vector_in() or IMP::algebra::get_random_vector_on() where there is one argument describing the volume or shape (such as an IMP::algebra::SphereD<4>). - or IMP::algebra::get_uniform_surface_cover() where the first argument is the primitive whose surface we are covering and the second is the number of points. Some good sed strings are - "s\#random_vector_in_unit_sphere()\#get_random_vector_in(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_in_unit_box()\#get_random_vector_in(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_on_unit_sphere()\#get_random_vector_on(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_on_unit_box()\#get_random_vector_on(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_in_box(\#get_random_vector_in(\#g" - "s\#random_vector_on_box(\#get_random_vector_on(\#g" - "s\#random_vector_in_sphere(\#get_random_vector_in(\#g" - "s\#random_vector_on_sphere(\#get_random_vector_on(\#g" - "s\#grid_cover\#get_grid_surface_cover\#g" - "s\#uniform_coverr\#get_uniform_surface_cover\#g" . - Spheres in IMP::algebra are now supported in general dimensions. See IMP::algebra::SphereD. - Many methods in IMP::algebra were renamed to be more consistent with other methods or general conventions. Most of these simply involved adding \c get_ as a prefix and/or add a \c _3d suffix when they did not take any arguments. - build_Rotation2D_from_Vector2D became IMP::algebra::get_rotation_to_x_axis() - most of the Rotation3D and Transformation3D functions (in addition, IMP::algebra::get_identity_rotation_3d() and IMP::algebra::get_identity_transformation_3d() added the \c _3d so we can add 2D versions at some point) - IMP::algebra::Transformation3D::get_transformed() and IMP::algebra::Rotation3D::get_rotated(). Replacing \c ".transform(" worked in the IMP code base. - IMP::algebra::get_distance(), this also carries over to the IMP::core::get_distance() The following global replacements are probably safe (in sed expressions) - "s\#core.distance(\#core.get_distance\#g" - "s\#algebra.distance(\#algebra.get_distance\#g" - "s\#\.transform(\#.get_transformed(\#g" Only in C++ - "s\#\.rotate(\#.get_rotated(\#g" Only in C++ - "s\#rotation_from\#get_rotation_from\#g" - "s\#rotation_in\#get_rotation_in\#g" - "s\#identity_rotation(\#get_identity_rotation_3d(" - "s\#::distance(\#::get_distance(\#g" . - IMP::algebra::Vector3D and IMP::algebra::Vector3Ds were replaced by the expanded (non-typedef) version in functions parsed by SWIG as the current situation was extremely brittle. For the moment, you should not use the typedefs in definitions that swig sees. Sorry. - IMP::algebra::Sphere3DPatch was renamed to IMP::algebra::SpherePatch3D to be consistent with the other geometry and allow us to support higher dimensions in the future. - rigid_align_first_to_second was renamed to IMP::algebra::get_transformation_aligning_first_to_second() and build_Transformation2D_from_point_sets was renamed to IMP::algebra::get_transformation_aligning_pair() ## Week of Feb 14, 2010 - The IMP::atom::Bond types were rearranged to reconcile the protein-centric classification with ligand support. - IMP::atom::Atom no longer has charge since we have IMP::atom::Charged. - IMP::atom::read_mol2() and IMP::atom::write_mol2() were added - IMP::atom::Selector and descendants renamed to IMP::atom::PDBSelector since we will soon have Mol2Selectors. In addition, the IMP::atom::IgnoreAlteriativesSelector is now IMP::atom::NonAlternativeSelector to keep the name positive, like the others. In the IMP body, all instances of the string Selector were for PDBSelectors and so sed was safe to use. - IMP::atom::read_pdb() lost the "ignore_alternatives" argument. Instead, all of the provided selectors, other than IMP::atom::AllSelector ignore alternatives internally. Code that uses the IMP::atom::AllSelector should probably change to the IMP::atom::NonAlternativeSelector. - IMP::example renamed (from examples) - The IMP_USAGE_CHECK() macro lost the (unused) third argument. ## Week of February 3, 2010 - The model io code as moved into the kernel so it is now IMP::write_model() and IMP::read_model() - Remove the version_info argument from the implementation macros. Instead, just pick up the module version info automatically. - Add a function Object::get_type_name() and Object::get_module() to get the respective bits of information about each object. - Change Object::show() to be a non-virtual wrapper which prints the object name and other info before passing control off to Object::do_show(). - The containers were moved to IMP.container. See that module page for a full list of what moved. In the process, the backwards compatibility functions were removed from IMP::container::PairsRestraint and kin. - IMP.atom.create_protein() now returns the created protein - the simplify functions have been moved from IMP.helper. They are now IMP.atom.create_simplified_along_backbone(). - IMP.helper.clone() has been renamed to IMP.atom.create_clone() - IMP.misc.ConnectingPairContainer has been moved to IMP.core.ConnectingPairContainer - The locations for pixels in IMP::em::DensityMap have been moved to the center of the corresponding voxel from the lower left corner. - IMP::em::get_transformed(), IMP::em::get_resampled() and IMP::em::get_density() have been added. - The documentation has been majorly cleaned up and simplified. Please look around and see if you like it. There are now two main pages, the introduction and the developers guide, rather than the previous plethora of pages. - IMP::algebra::Cone3D::get_bounding_sphere() has been removed since it was not of general use, not returning the bounding sphere you would expect and was trivial to implement. - IMP::algebra::projection() was removed since it didn't actually project onto the segment and the name didn't follow the normal conventions. And it wasn't used. - Most geometry now has namespace methods IMP::algebra::get_bounding_box(), IMP::algebra::get_surface_area(), IMP::algebra::get_volume(). ## Week of January 22, 2010 - A generic geometry layer has been added (see \ref genericgeometry). Using it, IMP::atom::rmsd(), IMP::algebra::alignment_taking_first_to_second() and the new IMP::atom::native_overlap() all can work on either collections of either IMP::core::XYZ particles or IMP::algebra::Vector3Ds. - Deprecated functionality has been removed from IMP.core. For those still using it, IMP::core::ClosePairsScoreState was replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairsScoreState was replaced by IMP::core::CloseBipartitePairContainer. - The macros for Scores changed slightly. Scores not in IMP svn which use the IMP_*_SCORE macros now have to implement the get_is_changed() function outside of the class declaration and scores which use the IMP_SIMPLE_*_SCORE now don't have to implement the functions to get interactions and used particles. - Rename IMP::core::MinimumPairScoreRestraint (and kin, Maximum, Singleton, Triplet and Quad) to IMP::core::MinimumPairRestraint. - Add IMP::core::MinumumPairScore (and kin). ## Week of January 1, 2010 - IMP::em is now using Particles to resample density maps and for fitting restraints. The previous ParticleAccessPoint mechanism is removed. - IMP::UsageException and IMP::InternalException are now reserved for their respective checks (and cannot be thrown by IMP_THROW()). This makes the distinction between API exceptions (which are thrown in any build mode and should be documented and tested) and helper exceptions which are only thrown when checks are turned on (and hence should not be documented) clearer. - IMP::core::RestraintSet moved to IMP::RestraintSet. The restraint sets now simply define a tree with weights which the model traverses as opposed to being an opaque IMP::Restraint themselves. This means that incremental evaluation of restraints in restraint sets now works and that restraints in restraint sets can be individually evaluated. Existing working usage should not have changed (so \c sed should manage updates just fine), but the internal implementation is completely new, so pay attention. - IMP now has a text i/o layer consisting of IMP::TextInput and IMP::TextOutput. These classes are convertible to C++ streams and implicitly constructable from both streams and file paths. As a result, functions that take them will work when called with a string for a file path, a Python file or a C++ stream. - The IMP.examples module now documents how to use the needed aspects of scons and SWIG. - The IMP.examples module is used as the basis for new modules created by the make-module script, so those include more documentation and are tested. - Module overview pages are now generated from the information passed in the SConscript file, rather than from the (now removed) modulename.dox file. See the examples file for an example of the things you can do there. - boost program options is now a checked for dependency. - The IMP::Model now uses dependency information from IMP::ScoreStates to automatically reorder the IMP::ScoreStates so that they are evaluated in the right order. Also, IMP::Model::evaluate() on a subset of the restraints now only updates the needed score states, not all of them. - Functions on rigid bodies no longer provide the IMP::core::RigidMemberRefiner as a default as it is, pretty much always, the wrong thing to use. That class may go away without notice in the future and is no longer documented. - the IMP::helper::create_simple_connectivity_on_rigid_bodies() and IMP::helper::create_simple_excluded_volume_on_rigid_bodies() now both take an IMP::Refiner as an (optional) final argument. This makes it so they can be used with the rigid bodies created from molecular hierarchies. This code still needs to be updated and the docs don't reflect what the code was or is doing, so use with caution (see notes in the header file). - The IMP::helper create* functions now all take non-Object inputs as const& to make clear what is input and that there are no memory management gotchas. - a new base class IMP::Constraint has been added for constraints. Various other classes have been renamed to be consistent. ## Week of December 11 - The interface of IMP::PairScore changed slightly to make the function signatures more uniform across various methods. Now all methods that take a pair of IMP::Particle objects take a const IMP::ParticlePair&. Previously some took that, some took an IMP::ParticlePair and some took two IMP::Particle*s. If you have an implementation of an IMP::PairScore outside of svn, you will have to update your function signatures (the compiler will tell you where). Backward compatible IMP::PairScore::evaluate() methods have been provided, so simple users of pair scores shouldn't see a difference. - IMP::core::PairRestraint takes a IMP::ParticlePair in the constructor instead of two IMP::Particle*s. A backwards compatible constructor can be added if needed. - IMP::ParticlePair::first and second went away since no one claimed to be using it and it complicated the implementation. - The generic restraints/score states/scores infrastructure has been extended to triplets and quads (for angles and torsion angles). ## Week of December 4 - IMP now makes use of Boost.Filesystem to manipulate paths. Things may not work well on systems where boost libs are not available which does not support posix paths. But libless boost installs generally don't exist outside of our test systems. - new functions like IMP::core::get_example_path(filename) and IMP::core::get_data_path(filename) allow one to easily find a data file named 'filename' for the examples and library internal data. - The IMP::display module has been revamped in order to support non-linear geometry. VRML support has been dropped. CMM support will be dropped unless someone requests it (as IMP::display::ChimeraWriter does all it does and more). ## Week of November 19 - imp.xcodeproj was moved into tools and updated accordingly - the bin directory finally went away - IMP::display::ChimeraWriter now supports the latest version of Chimera - IMP::display::CGOWriter was renamed to IMP::display::PymolWriter as it can do more than just CGO. - IMP::VectorOfRefCounted::operator[] can now be used for assignments just like a normal vector. No more need to use the IMP::VectorOfRefCounted::set() method. - the IMP::atom::Hierarchy constructor from an IMP::Particle* is now explicit (as were the decorator constructors) ## Week of November 12 - All exception types are now mapped to equivalent classes in Python. The class hierarchy is similar (e.g. IMP::ModelException derives from IMP::Exception in both C++ and Python) so all IMP exceptions can be caught in Python using "except IMP.Exception". For convenience, the IMP::IOException, IMP::ValueException and IMP::IndexException classes in Python also derive from the similarly-named standard Python classes (IOError, ValueError and IndexError respectively). ## week of October 29 Update - You can now add members to rigid bodies via IMP::core::RigidBody::add_member(). These members can be other rigid bodies (although this functionality is only lightly tested). - The code to get bounding volumes for IMP::atom::Hierarchy has been moved to the atom module, as as the clone and destroy code. They can be found at IMP::atom::clone(), IMP::atom::destroy(), IMP::atom::get_bounding_box(). - Added IMP::core::VolumeRestraint. - Several functions have been added to helper in creating restraints. They are create_simple_connectivity_on_rigid_bodies(), create_simple_connectivity_on_molecules(), create_simple_distance(), create_simple_diameter(), create_simple_em_fit(), and create_simple_excluded_volume_on_rigid_bodies(). Functions load_em_density_map() and set_rigid_bodies() are added as well. See IMP::helper for more information. - IMP::ScoreStates are further hidden. Namely, IMP::core::RigidBody constructs the needed score states internally, doing away with the need for the create functions (which have been done away with). In addition there are new decorators IMP::core::Cover and IMP::core::Centroid which manage their respective score states, further doing away with more create methods. The hierarchy-based create method for rigid bodies has been moved to the atom module and is called IMP::atom::rigid_body_setup_hierarchy(). - IMP::em::read_map() will figure out the file type to use automatically from the extension. Other such methods should follow suite. - A class IMP::core::StructureSet has been added to provide a convenient way to manage solutions of optimization runs. - IMP::core::RigidBodyMover now takes its angle argument in radians and is documented as such. - IMP::core::ClosePairsScoreState and IMP::core::CloseBipartitePairsScoreState have been replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairContainer. The score states have been marked as deprecated and are mostly still there, but are more limited than they were. - IMP::core::AllPairsPairContainer has been renamed to IMP::core::AllPairContainer and a bipartite version has been added. - All containers now keep track of what they looked like on the previous time step. - The IMP::core::ListPairContainer and IMP::core::ListSingletonContainer have had their interface reduced a bit in order to make implementing change tracking easier. - The set of exceptions has been simplified to go along with simplification of the checks. We now have IMP::InternalException and IMP::UsageException which are thrown by the respective checks (and the former is thrown by IMP_FAILURE()). When you want to to pass an exception back to Python use one of IMP::ValueException or IMP::IndexException or IMP::IOException. There is also a new macro, IMP_THROW() to be used when you want to throw an exception with a nice error message. Please look at code you have worked on and make sure that the exceptions make sense given the new clarifications. - An IMP::ModelException is now thrown when IMP::Particle attributes attain invalid values. This exception can be caught to recover from failures of IMP::core::ConjugateGradients or difficult restraints. ## Week of October 15 - The set of functions needed to get meta information about IMP::Restraint and IMP::ScoreState objects has hopefully stabilized. Sorry about the thrashing. If you have restraints or score states not in SVN, see the corresponding IMP_RESTRAINT type macros, the IMP::Interaction class and base classes for what exactly you need to implement. ## Week of October 8 - IMP::Interaction was modified to split IMP::interaction::get_used_particles() into IMP::interaction::get_read_particles() and IMP::interaction::get_write_particles(). This makes it possible to check for IMP::ScoreState objects being in an inconsistent order (so that the input to one is updated after it is read). Later, we will be able to automatically order the score states ensure that there are no problems. - IMP::atom::Hierarchy has been modifier to get rid of the IMP::atom::Hierarchy::Type. Instead, all types are contained in the decorators. IMP::atom::get_by_type() was also updated to reflect this. - IMP_check() and IMP_assert() have been replaced with IMP_USAGE_CHECK() and IMP_INTERNAL_CHECK() respectively. Hopefully this makes it clearer when to use which. IMP::set_check_level() has been revised accordingly (to use IMP::USAGE and IMP::USAGE_AND_INTERNAL). Also, IMP_failure() and IMP_not_implemented() have been renamed to IMP_FAILURE() and IMP_NOT_IMPLEMENTED() to make all macros be all caps. To update external code, - replace IMP.EXPENSIVE with IMP.USAGE_AND_INTERNAL - replace IMP.CHEAP with IMP.USAGE - replace IMP_assert and IMP_check as above. ## Week of October 1 - IMP::ScoreState and IMP::Restraint now better track the set of particles that they use and the dependencies between the particles. To this end, a new base class IMP::Interaction has been added with methods to get the inter-particle interactions and the set of particles used. - .cpp, .h, .py and SConscript files which contain the string "DO NOT COMMIT" will be rejected on checkin. Use this to make experimental or debugging changes you are making so you don't accidentally commit them. - The search module is no longer built by default and is not longer supported. It will get moved out of svn shortly. - IMP::ScoreState objects now also have IMP::Interacting::get_interacting_particles() methods. Soon IMP::PairScore and IMP::SingletonScore objects will too. ## Week of September 27 - Some support for gather statistics has been added to IMP::Model (see IMP::Model::set_gather_statistics()). If requested, an API can be provided so that the individual data points can be returned to programs. - A class IMP::core::RigidBodyDistancePairScore has been added which provides accelerated computations for the distance between two rigid bodies. - The SWIG input files were renamed from swig.i to swig.i-in to aid in dependency tracking. - rigid bodies now have two sets of particles associated with them in the various places they are used. One is the IMP::core::RigidBodyMember particles, as was previously the case. The other is defined through a refiner (and could be the RigidBodyMember particles if a IMP::core::RigidMembersRefiner is used). The idea is that you sometimes want some particles which are transformed along with the rigid body, but which are not really part of the representation of the rigid body. Examples would include marker points or multiresolution representations of proteins. ## Week of September 20 - all \c .i files have been renamed to start with \c IMP_ to help make dependency tracking easier. - static builds (builds with \c static=True) build both the static and dynamic libraries and just static executables - the \c profile build target has been removed as \c static=True does the important bit of what is needed. Just do \c linkflags=['-pg'] to add the needed link flag for \c gprof. - Restraints should now implement IMP::Restraint::unprotected_evaluate() instead of IMP::Restraint::evaluate(). The latter can now be called outside of IMP::Model::evaluate() and will return the right answer always. - \c bin/imppy.sh no longer works, as was previously warned. Use \c tools/imppy.sh instead. The bin directory will go away eventually, for now the script just tells you to use the tools version. ## Week of August 30 - A new score IMP::core::BoundingBox3DSingletonScore has been added to keep particles in a box. - str should now work better in Python when applied to IMP::Object objects and a few others. Objects should use the new IMP_SHOWABLE and IMP_SHOWABLE_INLINE macros instead of declaring show functions directory to make it so that they work in Python. It is now a bug to \c print or \c str to not return something useful in Python for any IMP object. ## Week of August 23 - each module now has a version, eg IMP::get_module_version_info() which is also returned by the various restraints of the module. The Python code checks that loaded libraries have the expected version numbers to try to provide better messages when runtime linking problems occur. - the various bounding_box methods have been renamed to ::get_bounding_box() to be more consistent with the other names. - Python documentation is now parsed by doxygen ## Week of August 15 - The Object::get_version_info() result now includes the SVN revision info when building from SVN. This means you can use it to more easily tell exactly what a built version of IMP was built from. ## Week of August 8 - Treatment of hetatoms in hierarchies (and hence pdbs) is now documented. See IMP::atom::Hierarchy. This involved broadening IMP::atom::Residue to handle ligands and having IMP::atom::Hierarchy::LIGAND, IMP::atom::Hierarchy::AMINOACID and IMP::atom::Hierarchy::NUCLEICACID be handled by IMP::atom::Residue. - IMP::atom::read_pdb() currently drops HOH ATOM records on the floor. We should add a IMP::atom::Water decorator if anyone wants to handle them (since they aren't atoms and don't have elements). - IMP::atom::Mass() has been added for decorating particles with mass. IMP::atom::Atom mass has been moved to that. - The scheme for generating atom types is now fully specified. See the IMP::atom::AtomType documentation. - IMP::atom::Atom and IMP::atom::Residue have been somewhat restructured. Some of the helper functions have been renamed to be more descriptive or removed (if they duplicated other easily accessible functionality). get_residue_index() has been removed (since you can just call get_index() on the return result of the IMP::atom::get_residue()). IMP::atom::get_chain() now returns an IMP::atom::Chain. root() was renamed to IMP::atom::get_root() for consistency. - SWIG files are now generated from a file called swig.i in the pyext directory. This file only has to contain module-specific commands such as lines to wrap headers and instantiate module templates. No more boilerplate. - The contents of bin are moving to tools since the two directories both contained assorted development tools (not all of which were executable). For now, bin/imppy.sh will still work, but that will go away soon. - IMP::Decorator::show() now just takes the stream (it used to take a prefix also, which was often ignored). Anyone who implemented a decorator will have to delete the second argument of their show function. ## Week of August 1 - Containers now have methods IMP::SingletonContainer::evaluate(), IMP::PairContainer::evaluate(), IMP::SingletonContainer::apply() and IMP::PairContainer::apply(). Using these can be significantly faster than looping over the particles externally. Similarly, there is IMP::SingletonScore::evaluate(), IMP::PairScore::evaluate(), IMP::SingletonModifier::apply(), IMP::PairModifier::apply(). - The decorator functions have been renamed to IMP::Decorator::setup_particle(), IMP::Decorator::decorate_particle() and IMP::Decorator::particle_is_instance(). A shell script \c tools/update-decorators is provided which updates the names in all code in subdirectories of the current directory. Please backup things first. - a function, IMP::Restraint::safe_evaluate() has been added to be used when requesting the value of restraints outside of the IMP::Model::evaluate(). ## Week of July 18 - The atom::BondPairContainer has been renamed to atom::BondPairFilter since it wasn't a real container. A real atom::BondPairContainer has been added. - Out of source builds mostly works. To do an out of source build, make a link to \c scons_tools and \c SConstruct from the source directory in the build directory and then add the line \command{repository="../path_to_src_dir"} in the \c config.py in the build directory. Then running \c scons in the build directory will put generated files there. This allows you to make both a \c release and \c fast build off of the same set of sources. Proper docs will be added soon. - The SConscript setup for a module has been greatly simplified so it should be easy enough for users to maintain their own module and updates to the build system shouldn't involve changes to user SConscripts any more. This does mean that any non-svn modules need to be updated. Look at IMP.core or another module and copy what is there. Again, proper docs will be added soon. ## Week of July 11 - Per Javi's suggestion, the conventions have been split into "Coding conventions" and "API conventions" and linked from the main page. - benchmarks got moved to their own module, IMP::benchmark and will now be tracked over time ## Week of July 4 - FilteredListSingletonContainer and FilteredPairSingletonContainer went away as there was too much hard to factor code shared with IMP::core::ListSingletonContainer and IMP::core::PairSingletonContainer. The IMP::core::ClosePairsScoreState is now handled by a list of filters on the class itself. That is, do IMP::core::ClosePairsScoreState::add_close_pair_filter() to add a filter. - Hooks have been added to allow actions to be taken on an assertion or check failure. See IMP::HandleFailure and its descendants for more details. - IMP::display::CGOWriter has been much improved and now allows provides lots of display-time control and easy building of animations. - Generic optimizer states have been added (IMP::core::SingletonsOptimizerState and IMP::core::PairsOptimizerState). ## Week of June 15 - box distributions (IMP::algebra::random_vector_on_box(), IMP::algebra::random_vector_in_box()) take bounding boxes - IMP::em::DensityMap are now IMP::Object objects. Some current allocations on the stack may need to be changed (since ref counted objects can't be stack allocated). - IMP::em::DensityMap::Read() is now deprecated (as is IMP::em::DensityMap::Write) as they violated a slew of conventions. Use IMP::em::read_map() and IMP::em::write_map() instead. ## Week of June 8 - new methods IMP::atom::add_radii() and IMP::atom::add_bonds(); - Hierarchys to IMP::atom::Hierarchies - IMP::Particles has been changed to a new type. It no longer plays quite so well with Python lists (you can't create one from an arbitrary Python list). - The containers for multiple decorators have been revised to make them more useful. See IMP::Decorators. You can now implicitly convert, for %example, a IMP::core::XYZs to an IMP::Particles (as well as use IMP::Decorators::get_particles()). All methods that take a collection of particles with a certain type now instead take a IMP::Decorators. - A method that has expectations about the particles it takes should take decorators rather than particles. That is, anything which expects IMP::core::XYZ particles, should take an IMP::core::XYZs instead of IMP::Particles. - IMP_LIST() now takes an extra argument (the name of the plural container) - In C++ IMP::Decorator objects are implicitly convertible to IMP::Particle* objects. Unfortunately, making this work in Python is problematic. - More methods take and return decorators rather than IMP::Particles. This means it is clearer what their expectations are and what they produce. ## Week of May 15 - The IMP::helper module was added and various functionality moved there. See its documentation page for a list of what is there. - IMP::algebra::SphericalCoords was renamed to IMP::algebra::SphericalVector3D to be consistent with other algebra classes. - IMP::algebra::Rotation2D is now IMP::InvalidDefault to be consistent with the other similar classes. - We are trying to be a bit more specific about what is expected from a IMP::atom::Hierarchy. Towards that end, there are two methods for extracting a representation from a hierarchy, IMP::atom::get_detailed_representation(), and IMP::atom::get_simplified_representation() - A method has been added to produce a simplified copy of an IMP::atom::Hierarchy. The method is still a bit crude and could benefit from other people inspecting the output and saying how it should be improved. - There is now an IMP::algebra::BoundingBoxD (and IMP::algebra::BoundingBox3D) to track bounding boxes. There are methods like IMP::algebra::bound_box(Sphere3D) which gets the bounding box of a sphere. - Each IMP::Object has its own log level. If IMP::Object::set_log_level() is called with a value other than IMP::DEFAULT, the value is used for code within and called by the object. To support this in your objects you need to add IMP_OBJECT_LOG to the start of each major method. It has been added to some of the core objects. ## Week of May 8 - IMP::atom::MolecularHierarchy has been renamed to IMP::atom::Hierarchy. Use \command{sed -i.old "s\#MolecularHierarchy\#Hierarchy\#g" my_files} - the long-deprecated Name(Decorator) was removed - The decorator suffix has been removed from decorator names. - Unit conventions have been established and documented for distance, energy, force and angle. - there is a function IMP::atom::create_protein() to create a coarse grained protein. The function is crude so far, but the idea is that we should have standard ways of generating such things so that everyone can benefit from best practices. - a class IMP::atom::BondGraph has been added so Boost.Graph algorithms can be used with bond graphs - the IMP-specific doxygen commands have been documented. - the doxygen commands unstable{name} and untested{name} have been added to doxygen to be used for marking methods and functions as likely to change or not well tested. Pass the name of the class or method as name. - bonds can now be added to molecular hierarchies automatically see IMP::atom::add_bonds() ## Week of May 1 - There is a IMP::Rotation2D - You can now extract fixed Euler ZYZ angles from a IMP::Rotation3D. This code can serve as a template for extracting other Euler angle conventions if they are needed. - IMP::algebra::rotation_between_two_vectors() changed to IMP::algebra::rotation_taking_first_to_second() to make it clear which direction the rotation goes. ## Week of April 27 - IMP::Decorator derived objects are now convertible to bool. So code like \code if (d) { blah } \endcode works. - IMP::Decorator::cast now returns a null decorator (\c DecoratorName()) rather than throwing an exception. This might make it actually useful. - New IMP::statistics module for statistical methods (e.g. clustering). ## Week of April 20 - Support for taking advantage of rigid bodies in collision detection has been added via the IMP::core::RigidClosePairsFinder. - All IMP::core::ClosePairsScoreState-related functionality has been optimized and should be noticeably faster. - The EMBED project has been merged into the IMP::em module. Thus, it is no longer necessary to obtain EMBED separately to build IMP::em. - All C++ methods which work with std::istream or std::ostream arguments should now be usable from Python: any Python file-like object can be used. - The SWIG interface now ties in to the reference counting mechanism used in the C++ code, for all classes derived from IMP::RefCounted. So, for %example, deleting a Python object should not free the underlying C++ object if another C++ object holds a reference to it, and unrefing a C++ object should not delete it if a Python object still holds a reference to it. There should be no further need to manually keep references to Python objects to prevent things from being prematurely cleaned up. This should be transparent to the user; please report any problems. - IMP::atom::AtomDecorator no longer inherits from IMP::core::XYZDecorator. ## Week of April 13 - The decorators IMP::atom::AtomDecorator, IMP::atom::ChainDecorator, IMP::atom::ResidueDecorator etc. now inherit from IMP::atom::MolecularHierarchyDecorator. This means they can set their IMP::atom::MolecularHierarchyDecorator::Type properly and saves a lot of typing in the common case. We can always make hierarchy decorators zero memory overhead if we find that the cost matters. - The base classes IMP::Comparable, IMP::NullDefault, IMP::ValidDefault and IMP::UninitializedDefault were added to document what the default constructors do for various objects in IMP. ## Week of April 6 - IMP::Object now defines pure virtual methods IMP::Object::get_version_info() and IMP::Object::show(). All non-abstract classes which inherit from IMP::Object must implement them. Doing this means they are more uniformly implemented and are documented in one place. - IMP::Object is now reference counted (the only objects this changes are IMP::Model and IMP::Optimizer). As a result optimizers and IMP::Model objects must be created on the heap. - IMP::RefCounted has had its name changed and no longer inherits from IMP::Object. This makes it easier to use independently. - the docs for reference counting (IMP::RefCounted) and and IMP::Decorator have been improved - running doxygen produces no warnings. Let's try to keep it that way. Run \command{scons doc > /dev/null} before you check in large blocks of new code to make sure that you didn't introduce warnings. - Helper classes have been added to make SingletonModifier and PairModifier objects look like functors for C++ and Python. This makes it so you can use \c map in Python and \c for_each in C++. See IMP::SingletonFunctor and IMP::PairFunctor for more information. - The recent changes page has moved to a doxygen page to make it easier to link changes back to the relevant source code. ## Week of March 20: - create_rigid_body(), create_rigid_bodies(), create_centroid() etc. now require you to add the score state to the model yourself. Whether the rigid body is enforced or not can be controlled by adding or removing the returned IMP::ScoreState. - All previously deprecated code has been removed. This includes indexes in particles, the no-argument constructor for particles and the functionality that was moved from core to atom. - The IMP::core::RigidBodyDecorator and IMP::core::RigidMemberDecorator have been simplified by removing the traits classes. There is now only one type of rigid body and it stores the list of member particles internally. - IMP::core::NameDecorator has been deprecated as IMP::Particles have built in names (use IMP::Particle::get_name(), IMP::Particle::set_name()). - IMP::atom::ChainDecorator has been added ## Week of March 13: - A new IMP::display::Writer, IMP::display::ChimeraWriter has been added which exports geometry to native Chimera constructs. It probably should be used instead of bild and cmm for most purposes as it combines the best of both types. - compose() for transformation is now a free function so compose(rotation, vector) will work via promotions. - the geometry classes have been cleaned up and are now more consistent with IMP standards (and more likely to do what their names/comments suggest). See IMP::algebra::Sphere3D, IMP::algebra::Cylinder3D, IMP::algebra::Segment3D, IMP::algebra::Plane3D etc. - ParticleRefiner has been renamed to IMP::Refiner since the Particle was long and unnecessary. ˘IF1•eIF3 translation initiation complex](http://salilab.org/40S-eIF1-eIF3). - IMP::domino::DiscreteSampler::set_restraints() now allows the set of restraints used in DOMINO filtering to be controlled (previously, all restraints in the IMP::Model were used). - IMP::modeller::IMPRestraints now transfers an IMP::ScoringFunction into the MODELLER score, not the implicit (and deprecated) score of the entire IMP::Model. - Bugfix #775: coordinates of voxels in EM maps were previously inconsistently treated (they are supposed to be the centers of the voxels, but in some cases they were the lower left edge). The fix may result in maps and their bounding boxes appearing shifted by half the voxel size. # 2.4.0 - 2015-02-27 # {#changelog_2_4_0} - Python 3 is now supported. The Windows package now works out of the box with Python 3.3 or 3.4; a `-python3` subpackage is provided for Ubuntu and Fedora. (Apple does not yet provide Python 3 so the Mac package is not built with Python 3 support, but if you install Python 3 you can compile IMP from source to work with it.) - A new command line tool, `rrt_sample`, allows for running rapidly exploring random tree (RRT) based sampling, e.g. to sample conformations of multiple domains connected by flexible linkers. - Command line tools are now placed in the `bin` directory of modules (the old `bin` directory, used for utility programs that are not installed, is now called `utility`). The `applications` directory is removed. - Several Python modules that were only used by EMageFit have been moved from IMP::em2d to IMP::EMageFit. - A new tutorial that uses the IMP::pmi module is now available (modeling of the RNA Polymerase II stalk). # 2.3.1 - 2014-12-01 # {#changelog_2_3_1} - Minor bug fixes. - Update to IMP::pmi to support the Nup84 and rnapolii biological systems. # 2.3.0 - 2014-10-24 # {#changelog_2_3_0} - A new IMP::pmi module was added, containing high-level Python classes to simplify the representation, scoring, sampling and analyzing of models in IMP. This module is still under heavy development, and should be considered experimental. - Binary installers for 64-bit Windows are now available. - The `.py` extension has been removed from all Python applications, so that they are consistent with the C++ applications. - IMP::atom::Selection objects can now be combined using basic set operations (union, intersection, difference, symmetric difference). In Python the equivalent operators (|, &, -, ^ respectively) can be used. - The integrative docking (idock) application now employs SOAP to improve the scoring. - IMP::atom::CHARMMParameters::create_topology() now works with a wider range of hierarchies (previously it would only accept Residues that were direct children of Chains). For example, it should do the right thing when given a Hierarchy that contains no Chain particles, or one where Residues are grouped under Fragments. - Only Python 2.6 or later is now supported. We still provide IMP for RedHat Enterprise 5 (which ships with Python 2.4) but you will need to use the `python26` binary (part of EPEL) rather than the system default Python. # 2.2.1 - 2014-08-11 # {#changelog_2_2_1} - Bugfix: PoseScore (see the ligand_score application) should now match that used by the Pose&Rank web server (http://salilab.org/poseandrank/) # 2.2.0 - 2014-04-09 # {#changelog_2_2_0} - A module IMP.template was added to act as a template for creating new external modules. It includes setting up git hooks for format checks and things like that. - The Python support for grids was cleaned up. Hopefully this makes it more robust. If you were using a grid in Python that no longer exists, ping us, we can add it back. - The generic geometry functions were simplified slightly and renamed to IMP::algebra::get_vector_geometry(). - The IMP::base::Flag class was added to simplify adding flags in C++. - The `--log_level` and `--check_level` command line argument now take strings (eg VERBOSE) to specify their values, rather than cryptic (and a bit unstable) numbers. - Command line options are now divided into normal and advanced ones. You do `--help_advanced` to show the advanced ones. - Support for computing Connolly surfaces was added with IMP::algebra::get_connolly_surface(). In addition, there is now IMP::algebra::get_uniform_surface_cover(const IMP::algebra::Sphere3Ds&,double) to sample the surface of a collection of balls. - IMP::atom::reate_simplified_from_volume() was added to create a simplified representation of a hierarchy that preserves the surface are and occupied volume. It uses IMP::algebra::get_simplified_from_volume(). If you want to play with such representations, you can use the rmf_simply application. - IMP::atom::get_rmsd(IMP::atom::Selection, IMP::atom::Selection, const IMP::algebra::Transformation3D&) was deprecated. Use IMP::atom::get_rmsd_transforming_first() instead. - IMP::algebra::get_rmsd() and IMP::algebra::get_rmsd_transforming_first() were added. - The family of macros IMP_BASE_[SMALL,LARGE]_[UNORDERED,ORDERED]_[MAP,SET] have been added (eg IMP_BASE_SMALL_ORDERED_SET ()). They provide platform and boost-version independent ways to take advantage of the best set/map for an application. In particular, they use boost::container::flat_set and kin when available. - IMP::atom::State and IMP::atom::Representation decorators have been added to provide support for multiple states and multiple representation schemes. They are integrated with RMFs and IMP::atom::Selection. See atom/multiresolution.py and atom/multistate.py for examples. - The container implementations has been cleaned up and simplified a bit. This should not change most usage. - [RMF](http://salilab.github.io/rmf) has been updated. There are a few minor API changes, in addition to a new, faster format (`.rmf3`). - [Eigen](#Eigen) is now included with IMP. - The deprecated function IMP::kernel::Optimizer::set_restraints() was removed. Use IMP::kernel::Optimizer::set_scoring_function() instead. - The deprecated constructor for IMP::kernel::ModelObject that does not take an IMP::kernel::Model was removed. This propagates to constructors IMP::kernel::Restraint, IMP::kernel::ScoreState and kin. Make sure you pass an IMP::kernel::Model as the first constructor argument. Related to this, deprecated functions like get_is_part_of_model(), set_model() were removed as they don't serve any purpose. - The deprecated methods in IMP::kernel::Optimizer that pertain to attribute optimization were removed. Inherit from IMP::kernel::AttributeOptimizer instead if you want to use them in your optimizer. - Some deprecated functionality was removed. If you encounter problems, replace IMP::base::OwnerPointer with IMP::base::PointerMember, find logging stuff in IMP/base/log.h, object in IMP/base/Object.h and random in IMP/base/random.h. - IMP::atom::Chain now uses strings for the chain ids, to be consistent with changes in the PDB. PDB files are written using the first character of the chain string. - Added orientation-dependent SOAP scoring (IMP::score_functor::OrientedSoap and IMP::atom::OrientedSoapPairScore). Orientation-dependent SOAP scores include those that score loops (SOAP-Loop), protein-peptide interfaces (SOAP-Peptide) and proteins (SOAP-Protein). The library files themselves are rather large (~1.5GB) and so are not included with IMP. They can be downloaded separately from http://salilab.org/SOAP/. # 2.1.1 - 2013-11-04 # {#changelog_2_1_1} - Added support for Mac OS 10.9 # 2.1.0 - 2013-10-24 # {#changelog_2_1_0} - There is now some limited IMP-wide gathering of timing statistics. You can add `--statistics=filename` to the command line or use the IMP::base::show_timings() function to view a summary of where IMP has spent time. - Slack was being used inconsistently (and not always correctly) among the various containers and restraints. If you had heavily optimized the slack value for speed, you may want to revisit it. - Class name OwnerPointer was switched to PointerMember, to denote more clearly a ref-counting pointer that is a class member. - The dependency handling in IMP has been rewritten to simplify it and fix a bunch of bugs. There are now more checks of correctness, so behavior that might have silently worked (or not worked) before may now trigger errors. In particular, restraints/score states must have access to the IMP::kernel::Model to do much of anything, so either use constructors that take the model or call IMP::kernel::ModelObject::set_model(). Also, changing your set of inputs or outputs without calling IMP::kernel::ModelObject::set_has_dependencies(false) is now likely to trigger a check failure (it would just silently do the wrong thing before). - Decorators have been cleaned up to make them behave consistently regarding IMP::kernel::Particle vs IMP::kernel::ParticleIndexes. This change shouldn't break existing code, but decorators should move to the new macros IMP_DECORATOR_METHODS(), IMP_DECORATOR_SETUP_0() etc. - The IMP_GRAPH() and IMP_WEIGHTED_GRAPH() macros got an extra argument describing how to show the vertexes in the graph. They also now expose a method, eg IMP::kernel::show_as_graphviz() that writes the graph in graphviz format to an IMP::base::TextOutput. - Support and rules for deprecating code in IMP have been updated. There are new macros and instructions on how to use them that should result in much better visibility about changes. See the C++ wiki for IMP for more information. And there is now an encoded policy of keeping deprecated things around for 1 release after deprecation. - IMP is now much more selective about when it invalidates the dependency graph. See [Dependencies](http://github.com/salilab/imp/wiki/Dependencies) for more information. This change should make evaluation of isolated restraints much faster in certain situations where they were unexpectedly slow before. - IMP::atom::Selection has been completely rewritten as it was very brittle. The set of particles returned may change in some cases, in some cases because it was buggy before, in some cases because of new bugs. - Reminder: the IMP_PROTECTED macros didn't work out well, and are now deprecated. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading `_` on the method name, if you get `Swig director pure virtual method called IMP::kernel::PairScore::do_get_inputs` messages. - The generated `cmake` files are no longer stored in the repository. Instead `tools/build/setup_cmake.py` is run after checkouts and other git events that change the repository tree (assuming you have run `setup_git.py`). - Generation of documentation using `doxygen` has been refactored. It is now done on a per-module basis, via targets like `IMP.core-doc`. The main advantage is that changes to the doc for a module can be tested comparatively quickly, just by building that module's docs. On the down side, docs are no longer quite so heavily interconnected, so dependencies only work to things that the module actually depends on. - The restraint and score state statistics functions were removed from IMP::Model. They had been broken for quite a while, with no one complaining. Statistics can be added back easily if requested. - Added an IMP-specific fork of `git-flow`. It handles things like `README.md` files when you start and finish feature branches and provides a tool to nicely update IMP and display the change log. The main commands of use are `git imp feature start <feature_name>`, `git imp feature finish <feature_name>`, and `git imp update`. You may want to remove the `git-flow` lines from your `.git/config` file as you may accidentally type `git flow` instead of `git imp`. You need to run `setup_git.py` to set things up. - [RMF](http://salilab.github.com/rmf) is now included as a git submodule rather than copied into the IMP repository. You should rerun `setup_git.py` to make sure submodule stuff is initialized. # For IMP 2.0 and earlier # {#changelog_2_0} ## Spring 2013 - Modules and applications now have a _Maintainer_ in their `README.md`, giving the github id of the person to whom issues should be assigned. - `setup_git.py` supports the `--module` flag which will set up git hooks and things for a git repository that contains a module. - To clean up for the release, the module dependencies are now described in a file called `dependencies.py` and the module docs in a `README.md`, both in the main module directory. - IMP-wide support for progress bars was added with the IMP::base::set_progress_display() function and helpers. These are only displayed when the IMP::base::get_log_level() is IMP::base::PROGRESS. ## Winter 2012-2013 - Scons support was removed. Use cmake to build IMP, it is way better. - Examples are run with a `--test` argument when IMP is built with internal checks. Expensive examples should check this flag and shorten what they do in order to try to test all their code without taking too long. - The class IMP::core::NonRigidMember was added to allow particles that are not rigid but whose coordinates are specified via internal coordinates in some IMP::algebra::Reference3D that is part of an IMP::core::RigidBody. - The IMP_BIG_ENDIAN and IMP_LITTLE_ENDIAN preprocessor symbols are deprecated in favor of the BOOST_BIG_ENDIAN and BOOST_LITTLE_ENDIAN symbols. - Support for asymmetric moves was added to IMP::core::MonteCarlo. To do this, IMP::core::MonteCarloMover was added and IMP::core::Mover and IMP::core::MoverBase were deprecated and implemented in terms of the new class. This should require few or no changes in code that uses Monte Carlo support or implements movers (although movers should move to IMP::core::MonteCarlo). All IMP::core::MonteCarloMover objects keep track of statistics (number of moves and number of accepted moves), rather than that being done in the IMP::core::SerialMover. In addition, the poorly defined move probability was removed from IMP::core::MonteCarlo. - IMP::base::WarningContext now outputs warnings immediately. This change was needed as the previous behavior resulted in occasional core dumps when the libraries for log support were unloaded before the libraries where static IMP::base::WarningContexts were declared. And there was no easy way to detect such problems. - You can now (probably) build IMP with [CMake on IMP wiki](https://github.com/salilab/imp/wiki/Cmake). - The maximum logging level and check level can now be controlled independently of the build mode. To, for example, enable progress level logging in a fast build, add maxlog="PROGRESS" to your config. The maximum check level for `release` has been reduced to IMP::USAGE as that speeds things up a bit. - The way applications are structured and specified has been regularized and simplified. Python files are automatically copied/installed and binary executables are compiled from each found `.cpp` file linked with any `.cpp` files contained in a subdirectory called `lib`. Nothing needs to be put into the SConscript file. - The Python constants like IMP.algebra.has_cgal went away as swig autogenerates ones like IMP.algebra.IMP_ALGEBRA_HAS_IMP_CGAL. While the latter are a bit uglier, the constants aren't used much and having them be auto generated is nice. - There have been some major changes to the build system. The main one is that the SConscripts in the directories below the module directory aren't really used. They are being removed from the repository. Now everything really is off of file locations (eg all test_* in test is a test, each *.cpp in bin becomes a program etc). This greatly simplifies the build system, which has, historically been a disaster. You can now define dependencies by dropping files into a dependency subdirectory in your module, no need to mess with the SConscript. - Building IMP now always requires specifying a `repository` argument. This was always required for out of source builds, which is what one should be doing, but is now required for in source builds to, so as to make accidentally polluting your source with an in-source build harder. Use `repository="."` in that case. - The HDF5 functionality in RMF was reorganized. Specifically, the headers were moved into `RMF/HDF5`, the Python functionality into `RMF_HDF5` and the code into namespace `RMF::HDF5` and all mentions of `HDF5` or `hdf5` in the class and function names were removed. - Support for IMP-wide command line flags have been added to IMP::base in the base/flags.h file. They provide a simple command line parsing interface that can be used from C++ and Python. In addition, various functionality in IMP is automatically exposed on the command line when those functions are used. For example, one can control logging, checks or profiling. See IMP/base/flags.h for (some) more info. - OpenMP support in IMP is officially released. All tests pass. But there has been little specific testing so far. - Access to attributes using IMP::ParticleIndex and the IMP::Model now works from Python. See IMP::Model::add_attribute() and kin. - IMP::ParticleKey and IMP::ParticlesKey were renamed to IMP::ParticleIndexKey and IMP::ParticleIndexesKey respectively as that is what they store. Typedefs were provided to support old code. - Indexes have been added to the docs connecting classes with all the methods that return them, examples that use them and methods that take them. And all methods that are used in examples with the examples they are used in. Please check them out and report issues and ways to improve things. You can find links on the main page. - The docs for a modules/biological systems/applications should now be found in an `overview.dox` in the module `doc` folder. Modules in IMP svn have been updated. For external modules, you can just move a `overview.dox` from that was previously generated into your `doc` folder and (eventually) remove the lines from your `doc/SConscript`. This makes it easier to use the full range of doxygen markup, as well as removes the need to escape doxygen markup for Python. - Examples have now been simplified into full blown doxygen examples. This involves merging the `.py` and `.readme` files (putting `\#\#` in front of each line of the `.readme`) and adding a `## \\example module/examplename.py` line at the beginning. You can use `./tools/updaters/update_examples.py modules/mymodule` to update a module that was not already updated. - IMP::base::FailureHandler has been removed as it wasn't used. - Handling of data and includes has been simplified. All appropriate files in the corresponding directories are linked into build/installed. This is only likely to be a problem if you had broken headers sitting around `include`. If so, just comment them out, or move them some place nicer. - Modules now must be in a folder called `modules`, which must have a `description.py` file. - The build system should now be a little better about cleaning up after itself. - How headers/doc files/data/examples - The IMP_PROTECTED macros didn't work out well, and are now deprecated. Sorry. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading "_" from the method name. - Nice version numbers will now be generated if you use git. You need to rerun tools/git/developer_tools/init_git if you have an existing clone. ## Fall 2012 - 11/9: RMF is now an external library shipped with IMP. IMP will build it and install it automatically if you have cmake version 2.8 or higher installed. - 11/9: Updated support macros for IMP::Restraint and IMP::ScoreState classes have been added (IMP_RESTRAINT_2() and IMP_SCORE_STATE_2()) that make it easier to return dependencies on things other than IMP::Particle and IMP::ScoreState objects. In addition, the dependency functions on scores and modifiers have been updated similarly and standardized. In general this should not cause backward compatibility problems unless you were overriding the IMP::ModelObject::do_get_inputs() in IMP::Restraint and using the IMP_RESTRAINT() macro. In that case update to the IMP_RESTRAINT_2() macro. - 11/8: Support has been added for the C++11 keywords override (IMP_OVERRIDE) and final (IMP_FINAL). They allow you to tell the compiler to produce an error if the marked method does not override/implement a parent's method and if anything overrides a given method, respectively. - 11/8: IMP::SingletonModifier::apply_indexes() and IMP::SingletonScore::evaluate_indexes() now require an upper and lower bound index (you can just pass 0 and the size of the array). We can make them option if there is a good reason. - 11/8: IMP::SingletonDerivativeModifier and kin were merged into IMP::SingletonModifier and kin as the only difference, the presence of the IMP::DerivativeAccumulator (whose weight was always 1) wasn't very interested and the split added lots of complexity. If this has an impact on your code, let us know. - 11/8; Evaluation with containers was greatly simplified, moving many methods on containers. We can add backwards compatibility versions if that is interesting. - 11/7: Predicates, scores and modifiers written in Python need to update their methods for returning inputs and outputs. They should now have a method \c _do_get_outputs(self, m, particle_indexes) which should probably just return \c [m.get_particle(i) for in in particle_indexes] to return all the passed particles. - 11/7: It is now preferred that IMP::Restraint and IMP::ScoreState descendants report dependencies directly using IMP::ModelObject::do_get_inputs() and IMP::ModelObject::do_get_outputs(). To do this, you should use the IMP_RESTRAINT_2() and IMP_CONSTRAINT_2() macros instead. Methods like IMP::Restraint::get_input_particles() are now deprecated. This change was made as the old mechanism introduced a lot of extra complication and made it so that there was no way to directly depend on something that was not a Particle or a Container. - 11/3: IMP::container::InContainerPairFilter and friends now treat pairs as unordered. A flag was added to control this. - 9/5: IMP::base::Object classes now have a method IMP::base::Object::clear_caches() that clears all cache data in the object. This is not universally implemented, but now provides a correct path for doing that. The IMP_LIST() type macros call that method when their contents changes. ## Summer 2012 - 7/3: IMP_PERIODIC_OPTIMIZER_STATE() was superceded by the class IMP::core::PeriodicOptimizerState as it is rather cleaner than the macro. - 7/2: in order to make various base classes in IMP more Python friendly, macros IMP_PROTECTED_METHOD(), IMP_PROTECTED_CONSTRUCTOR() were added that properly expose such methods to Python. The standards checks now complain if \c protected is used an a class in the API. - 6/24: Executables from \c module/bin are now put in \b build/module_bin/module. This removes a recurring source of errors in the build script due to the origin and destination directory have the same name in scons. - 6/24: Benchmarks are now put int \c module/benchmark and build to \c build/benchmark/module to remove a recurring source of errors. - 6/18: The IMP::benchmark module supports the GPerfTools profilers when available. Benchmark executables (things in \c bin whose name starts with \c benchmark_) that use the IMP_BENCHMARK() macro have command line flags to turn on and cpu and memory profiling. - 6/12: The naming rules for .i files that are in module \c pyext directories changed. They should now be \c IMP_modulename.name.i instead of \c IMP_modulename_name.i. This is to better support underscores in names. ## Winter 2011/2012 - 20/4: A method, IMP::atom::read_pdb(IMP::base::TextInput, int, IMP::atom::Hierarchy), was added to load coordinates from a pdb file into an existing IMP::atom::Hierarchy. - 4/4: the filter types were superceded by the predicate types. You will only see a difference if you develop a filter class (eg IMP::PairFilter -derived class) or directly call methods on filters (specifically IMP::PairFilter::get_contains()). - 29/3: IMP::Particle::get_has_model() has been changed to IMP::ModelObject::get_is_part_of_model() to be consistent with other classes. - 28/3: IMP::core::Mover changed slightly, adding an extra method IMP::core::Mover::get_output_particles() and requiring the IMP::Model be passed to the constructor. All movers in IMP SVN were updated without any difficulty. This allows movers to use information contained in the dependency graph amongst particles so they can, for example, weight moves based on scores. - 26/3: IMP::SingletonContainerInput etc were added. Methods that take these as their arguments can be passed either containers or lists of particles (or tuples), as convenient. All methods that take IMP::SingletonContainer objects should eventually move over to taking this (changing just requires changing the argument from a pointer to an Input). - 22/3: Over the last week or so, which IMP headers include which other heads have been substantially simplified. This may result in compilation errors for code outside of \c svn that depended on IMP headers including other headers. In general, you want to explicitly include the header defining each thing you use. - 15/3: The vector used throughout IMP is now exposed at IMP::base::Vector and what it does it documented. - 10/3: IMP::domino::RestraintCache was introduced to provide a documented centralized place to handle restraint evaluation and caching in DOMINO. In the switch various bugs involving maximum scores and IMP::RestraintSet objects were fixed. - 10/3: Scoring functions in IMP are now implemented in IMP::ScoringFunction objects. All restraint evaluation is handled by them and they can cache various useful information to aid in accelerating restraint evaluation. Backward compatibility interfaces have mostly been provided, please complain if some important interface went away. This has greatly simplified the code as well as fixed various bugs in various types of restraint evaluation (particularly in the handling of maximum scores on IMP::Restraint and IMP::RestraintSet objects. - 29/2: The macros IMP_NAMED_TUPLE_1(),IMP_NAMED_TUPLE_2(),IMP_NAMED_TUPLE_3() were added for easy generation of simply struct for use in passing and returning values. The generated structs support comparison, output, hashing and have constructors. Use the macros instead of a boost::tuple or a a POD struct. - 16/2: Restraints now are evaluated with the correct weights when evaluated by themselves as opposed to always returning their weight as seen by the model as a whole. - 15/2: The IMP::algebra grid support was rearranged and simplified. Most of the changes should be transparent. However, the IMP::algebra::grids namespace got folded into IMP::algebra, which might break some code. And various (obscure) constructors of IMP::algebra::GridD were removed. All code in svn has been changed as needed. - 3/2: The IMP::domino HDF5 container has been revamped. This involves splitting it into a write version (IMP::domino::WriteHDF5AssignmentContainer) and a read version (IMP::domino::ReadHDF5AssignmentContainer). - 3/2: IMP::domino::WriteAssignmentContainer and IMP::domino::ReadAssignmentContainer were added to do direct binary file writing. This is faster than the HDF5 versions. - 3/2: The installation instructions have been revamped and the developer instruction, in particular, have been greatly simplified. - 24/1: \c tools/show-changes has been added that shows the changes in the history log since you last updated. - 24/1: \c tools/make-source has been added that adds a source/header pair to a module - 24/1: principal components analysis in IMP::algebra was generalized to work in any dimension. This required renaming the class to IMP::algebra::PrincipalComponentAnalysisD. - 24/1: the output of the IMP::algebra::PrincipalComponentAnalysisD::show() was changed to reflect the standards (values produce compact, single line output on show). The old show has been moved to IMP::multifit - 19/1: IMP::container::ConsecutivePairContainer had IMP::container::ExclusiveConsecutivePairContainer split off of it to make the two operating mode clearer and so that there could be a clear IMP::container::ExclusiveConsecutivePairFilter to use. - IMP::algebra::Grid::apply was added to provide faster access to all (or, later, part of a grid). It can be an order of magnitude or more faster than using the iterators when doing simple operations. - the scons config options \c pretty and \c color were added (defaulting to True). When they are true, build commands are mostly suppressed and a much briefer description of what is being done is printed out (in color). Note that \c color mode and compilation in emacs don't get along too well. - IMP::algebra::ParabolicFit was renamed to IMP::algebra::ParabolicFit2D (a backwards compatibility typedef is currently provided) - IMP::algebra::LinearFit was renamed to IMP::algebra::LinearFit2D. It now supports error bars on the y-values. ## Fall 2011 - IMP::compatibility::checked_vector was renamed to IMP::compatibility::vector and made available to IMP as IMP::vector. Typedefs should use this type (IMP::vector). - Various IMP::algebra plural typedefs which had been using std::vector (since they never were updated), have now been moved to IMP::vector. - IMP::statistics::HistogramD now supports display with \c matplotlib. - IMP::statistics::HistogramD supercedes IMP::statistics::Histogram - predicate base classes have been added: IMP::SingletonPredicate, IMP::PairPredicate etc. These take particles and return an integer. Current functionality using them includes IMP::core::PredicateSingletonScore that uses a predicate to choose which score to apply. - the core generic restraint and score state support was moved to the kernel to avoid breaking various scores outside of IMP svn. - the IMP::display module was refactored and moved up to be a dependency of IMP::core and IMP::atom rather than vice versa. The main observable results are that you should now use IMP::atom::HierarchyGeometry, IMP::atom::SelectionGeometry, IMP::atom::HierarchiesGeometry, IMP::core::XYZRGeometry, IMP::atom::BondGeometry etc instead of finding them in IMP::display. This change means that, display can be used internally when testing and debugging functionality in IMP::core and IMP::atom and that tests that use IMP::display don't need to build all of IMP::atom. - Decomposition of restraints now results in restraints that are faster to evaluate, using the generic restraint mechanism in IMP kernel. - IMP::algebra::get_unit_bounding_box_d() was changed to return a bounding box with radius 1 to be consistent with IMP::algebra::get_unit_sphere_d() - A general purpose scheme to associate data with tuples of hierarchy nodes was added to the RMF format. This deprecates the special purpose bond support. - the way various types of geometry is stored in RMF files was changed, breaking backwards compatibility. ## Summer 2011 - some functionality from the kernel has been spun off into a new module IMP.base. You probably won't notice anything. The only significant changes are that functions that accept or return IMP::base::TextInput and IMP::base::TextOutput must declare them with the \c base qualifier (eg base::TextInput) so that swig is happy. - the source for the kernel has been moved from \c kernel to \c modules/kernel. This is unlikely to effect anyone. - IMP::core::ExcludedVolumeRestraint now supports filters - IMP::domino::HDF5AssignmentTable now caches internally, which makes it several times faster than before. - you can now pick which Python to use with the \c python argument to scons. - the units for BD diffusion coefficients changed to be consistent with other IMP units. - IMP::domino::get_interaction_graph() now takes an IMP::RestraintsTemp, like other functions - The assert failure handling function was documented and renamed to IMP::handle_error() so it is easier to find when you want to use gdb. - to work around various issues with ref counting pointers and compiler eccentricities, we added emulation of the C++0x \c nullptr. You should prefer that to \c NULL in IMP code that you want to run cross platform. - IMP::IntsList, IMP::FloatsList and IMP::StringsList have been added to replace the various scattered solutions for passing such types. - IMP::Restraint::get_decomposition() and IMP::Restraint::get_current_decomposition() changed to return a single Restraint, instead of a list. This makes it easier to propagate the weight and maximum reliably. - IMP now supports usage of restraints without adding them to the IMP::Model scoring function. To do this, do IMP::Restraint::set_model(), passing the model. The restraint can then be evaluated. - IMP::domino::DominoSampler does not use the IMP::domino::OptimizeRestraints restraint optimization by default. Instead, if it is asked to create the IMP::domino::RestraintScoreSubsetFilterTable, it decomposes the restraints first (via IMP::create_decomposition()). This avoids several bugs and very fragile code. - the IMP::core::DiameterRestraint was ignoring the radii of the particles involved and so was giving artificially low scores. It now is on the actual diameter of this molecule. - the IMP::core::MonteCarlo optimizers now support incremental evaluation. This can be a lot faster when only a few particles are moved each step. - all the plural types in IMP have moved to bounds checked vectors when using debug builds with gcc. If you get a non-imp error about an invalid access, it is probably some code that walks off the end of an array. Please report it. - The IMP::core::MoverBase was changed to move it off of the IMP::SingletonContainer. This resulted in the IMP::core::BallMover and IMP::core::NormalMover having new constructors. Now create them from a list of particles, a list of float keys and a radius/standard deviation. This removes some complexity and inefficiencies as well is removes some scope for errors. - Temp and non-temp decorator lists have been merged in C++. This was done to simplify code since no one was using the non-temp variants. Now all decorator lists are temp (non-reference counted) variants. - restraints and restraint sets are now evaluated only using "canonical" weights. That is, each restraint has one weight for the model (the total over all the restraint sets it is contained in and will always be evaluated with that weight). This allows the evaluation of simple restraints to be faster. - The IMP_LIST() macros have been changed slightly. If your usage needs to take action upon addition, removal or changes to the list, use the IMP_LIST_ACTION() variant in your header. The IMP_LIST_IMPL() macro no longer takes the actions, and the last three (generally empty arguments) will have to be removed from the call to get it to work. This is needed to move towards more flexible and efficient passing of lists of values. - IMP::domino uses the new IMP::Model::evaluate_if_good() support to accelerate computations - IMP::core::MonteCarlo can use the IMP::Model::evaluate_if_good() support to more quickly reject very bad conformations. See IMP::core::MonteCarlo::set_maximum_difference(). - IMP::core::ExcludedVolumeRestraint has been rewritten to be faster. It is especially faster when used with IMP::Model::evaluate_if_good() - a new evaluation mode IMP::Model::evaluate_if_good() and IMP::Model::evaluate_if_below() has been added. It allows restraints to shortcut score computation when the score can be shown to be worse than a maximum value. - as part of moving towards more efficient use of evaluate_if_good() various deprecated methods have been removed from IMP::core::ExcludedVolumeRestraint and IMP::core::RigidClosePairsFinder. - incremental evaluation support was removed from IMP. The reason for doing this was that the code was complicated, created overhead even when incremental support was not used and was not commonly used. Further, decomposing the restraints and using the dependency graph yields similar results and is more transparent. - the RMF helper binaries have been moved to their own application so that they get installed. - IMP::display::Writer has been revised to better support writers that write multiple frames to a single file and ones that write binary data. The noticeable changes are that the existing IMP::display::Writer instances must be created with a file name. And the IMP::display::WriteOptimizerState just takes a writer rather than a write and a file name template. IMP::display::Writer::set_frame() has been added so that you can directly control the frame being written. ## Spring 2011 - An interactive mode has been added to IMP::domino::DominoSampler. The interactive mode is useful for understanding domino sampling runs and distributing domino computations across a cluster. - Functions IMP::domino::set_assignments() and IMP::domino::get_assignments() were added which support writing subset assignments to data fields in hdf5 files. These can be used to helping to distribute domino on a cluster as well as debugging the sampling process. - a Python function IMP.show_graphviz() has been added that attempts to render a graph with \c dot and then open the pdf. - The Python functions IMP.show_altgraph() and IMP.get_networkx_graph() have been added to provide alternatives for quick viewing of IMP graphs. Unfortunately, they still pale compared to graphviz. - IMP.atom.create_cover() was added to facilitate creating a summary particle for part of a hierarchy in order to accelerate restraints. - IMP.hdf5 was renamed to IMP.rmf, the proposed name of the file format that it supports. That also better separates the hdf5 support from the file format support. - The old I/O functions have been marked as deprecated in favor of IMP.hdf5. You may need to compile with deprecated="yes" if you use them. - The Python graph interface now has a remove_vertex method. - One can now access the model restraint evaluation statistics programmatically using IMP::Model::get_restaint_statistics(). Restraint statistics can be logged to a file during optimization using IMP::core::WriteRestraintScoreStatisticsOptimizerState. - IMP::domino::DominoSampler now allows you to specify the tree on which merges occur directly. Also, you now need to specify the merge tree rather than the JunctionTree if you want to do non-loopy inference and specify something about how it is done. Sorry, it makes the code a lot easier. - Preliminary symmetry support has been added via the IMP::core::TransformationSymmetry and IMP::core::TransformationAndReflectionSymmetry classes. ## Winter 2011 - There now is a common base class IMP::atom::Simulation for IMP::atom::BrownianDynamics and IMP::atom::MolecularDynamics. In merging them there were several minor changes made to both. The most noticeable may be that the function for setting the maximum time step length is now IMP::atom::Simulation::set_maximum_time_step(). The last time step length can be extracted by IMP::atom::Simulation::get_last_time_step() - IMP::core::MonteCarlo::set_temperature() was renamed to IMP::core::MonteCarlo::set_kt() as that is what it is doing. - A new module IMP.hdf5 has been added to provide more general support for saving and loading IMP::atom::Hierarchy structures to files. Multiple conformations of each hierarchy can be saved as well as information about geometry and restraints (although these can't be read at the moment, so they are not so useful). - IMP::algebra::grids::GridD now supports dense grids in arbitrary dimension through use of IMP::algebra::grids::DenseGridStorageD. - IMP::Optimizer now can use a list of restraint sets to compute the score rather than the complete Model score. - each IMP::Restraint now has its own weight. This is lightly tested. - A new log level IMP::PROGRESS has been added. The idea is that IMP::Sampler::get_sample() calls and other expensive loops can show a boost progress meter at this log level to give some idea when things are moving forward. - IMP::domino::TrivialParticleStates renamed to IMP::domino::IndexParticleStates to be less judgmental - IMP::domino::CompoundParticleStates and IMP::domino::RecursiveParticleStates added - IMP::domino::DominoSampler now supports maximum scores on IMP::RestraintSets (untested). - the name of the clustering methods have been changed to create_ from get_ since they create objects. - the names of the recently added metric based clustering support has been changed to Metric rather than Distance. - IMP now has the ability to have certain tests designated as expensive and to skip those tests when test-fast is run. The motivation for this is that the tests take a really long time making it impractical to run before every commit. And it is hard to tell which tests have not been run when running with the necessary multiple build processes. My thought is that tests that take about a second or less on a debug build should be left along (this is probably >95% of the tests), but the few that take longer should be marked as expensive. To do that, rename the test to expensive_text_XXX.py (from test_XXX.py) and add expensive_python_tests= env.IMPModuleGetExpensivePythonTests() to the IMPModuleTest call in the test SConscript. - the fuction IMP::display::create_restraint_geometry() was added to do what its name says. - runtime dimensional geometric objects have been added. These include IMP::algebra::VectorKD, IMP::algebra::BoundingBoxKD and IMP::algebra::SphereKD. The IMP::statistics::Embedding classes now use IMP::algebra::VectorKD instead of Floats. The sparse grids also support variable dimensions as to the nearest neighbor searches. - IMP::atom::DopePairScore has been added to IMP. - IMP::atom::CHARMMStereochemistryRestraint provides a high-level simplified interface to the bonded part of the CHARMM forcefield. - IMP modules are now built by first combining all the .cpp files into one, and then building that file. This greatly accelerates building of a module from scratch as build time is dominated by parsing headers. This does change the semantics of .cpp files as they see headers included by and functions declared by an arbitrary subset of other .cpp files. For now they are still required to be able to be compiled separately. To get the only behavior back either globally or on a per-module basis, see the percppcompilation parameter to scons. - Instructions and scripts have been added for building modules and things externally to the IMP build. See the installation guide. - IMP::atom::CHARMMParameters can now automatically map CHARMM-style atom and residue naming to PDB-style, so CHARMM topology files can be used unmodified. - IMP::atom::CHARMMTopology::add_sequence() provides a simple way to generate a topology from a string of amino acid one-letter codes. - IMP::atom::CHARMMTopology::add_coordinates() will generate atomic Cartesian coordinates from CHARMM internal coordinate information. - IMP::atom::CHARMMTopology::add_missing_atoms() and IMP::atom::remove_charmm_untyped_atoms() can now be used to ensure that atoms read from a PDB file match the CHARMM topology. - building of deprecated functionality is now false by default - the IMP_COMPARISONS(), IMP_COMPARISONS_1(), IMP_COMPARISONS_2() macros now take the name of the class as an additional (first) argument. This avoids the formerly needed, \c This typedef and makes them more consistent with the other macros in IMP. - modules can now be built externally to the IMP svn in addition to applications and biological systems. To do so, create a directory with your module in a subdir containing links to the IMP SConscript file and scons_tools and a config.py that has information needed to run IMP (eg an appropriate \c pythonpath, \c ldpath). - removed IMP::algebra::GridD::get_index(VectorD) as it wasn't very safe (in that vectors coordinates can get rounded between the IMP::algebra::GridD::get_has_index() and IMP::algebra::GridD::get_index() calls. Use IMP::algebra::GridD::get_nearest_index() or IMP::algebra::GridD::get_extended_index() or IMP::algebra::GridD::operator[]() instead. - renamed IMP::domino::ParticleStatesTable::get_particles() to IMP::domino::ParticleStatesTable::get_subset() as it is more useful to be able to get the subset containing all the particles. - rename IMP::restrainer::Restraint::print_all_restraints() to IMP::restrainer::Restraint::show_all_restraints() to be consistent with the rest of IMP. - renamed IMP::atom::get_conect_record_string() to IMP::atom::get_pdb_conect_record_string() to make clearer what it is. - class and method names are now spell checked. Add missing words to spelling_exceptions in the module test call. - the plural names (eg IMP.Particles()) have been removed from the Python side as they are just Python lists. - biological systems can now include internal modules. These provide a convenient place to put application specific C++ and Python code. See the example system for more info. - IMP_VALUES() now includes IMP_OUTPUT_OPERATOR() so you no longer need two namespace macros for each value. - the function IMP::core::Harmonic::k_from_standard_deviation was renamed to IMP::core::Harmonic::get_k_from_standard_deviation() to conform to the standards. - the IMP::atom::PDBSelector and IMP::atom::Mol2Selector became IMP::Object classes. All this means is that they need to be created with \c new in C++ code. It also means that one is no longer prohibited from storing them (not that there is much use in doing that). - the standards tests have been improved to test more useful aspects. Now class and function names are checked some as well as all modules have the checks for show and plural types. The error messages should explain how to fix things. - IMP::core::RigidBody::lazy_set_reference_frame() was renamed to IMP::core::RigidBody::set_reference_frame_lazy() to fit with the standards better. - the transform based methods on reference frame which have been deprecated for a while were removed - the hierarchy visiting and searching functions have been renamed to better fit the IMP naming scheme as we don't expect they were used outside of the IMP source. - IMP::atom::NonhydrogenMol2Selector was renamed IMP::atom::NonHydrogenMol2Selector to be consistent with the PDB version ## Fall 2010 - IMP::display::PlaneGeometry was added to display planes - the optimize methods on optimizers was moved to a "non-virtual interface". Usage has not changed, but optimizers should implement IMP::Optimizer::do_optimize() instead of IMP::Optimizer::optimize() to benefit from better checking. - the two-vector constructor for IMP::algebra::Plane3D now expects the second vector to be a unit vector as otherwise one could switch the order around without noticing. - the io functions in the kernel have been cleaned up a bit (the old ones are there for backwards compatibility). See IMP::write_particles(), IMP::write_particles_binary(), IMP::write_particles_to_buffer() and IMP::read_particles(), IMP::read_particles_binary(), IMP::read_particles_from_buffer() - IMP::core::WriteModelOptimizerState is going away as it duplicated IMP::WriteParticlesOptimizerState. - the IMP::algebra::Grid3D has been changed to make it support any (fixed) dimension when used with sparse storage. The template arguments have also been changed to remove a redundant one. See IMP::algebra::GridD. - the biological systems specification has been improved a little adding ability to run some rudimentary tests to make sure the scripts are ok and to not require that everything be run from the svn directory. - A new method IMP::atom::create_compatible_rigid_body() which creates a rigid body from a hierarchy which has the same internal coordinates as another rigid body. A way to make the rigid body construction more canonical would be nice (so we don't need this function), but I'm not sure how to do that. - Logging has been improved to better report the name of methods producing the log messages - a macro IMP_FUNCTION_LOG has been added which should be added to the start of non-member functions which produce significant logging output. - IMP::algebra::get_distance(const Rotation3D&,const Rotation3D&) was changed to use the Euclidean distance in R4 as that seems better behaved than the prior metric. - IMP.domino2 was renamed to IMP.domino and the old IMP.domino was moved - the documentation for classes and some methods now includes a list of all examples which use the class. Please report any incorrect links or phantom classes appearing in the docs. Unfortunately, functions have to be added individually. Please request any functions of particular interest. - applications and biological systems now have their own documentation pages - IMP::atom::Hierarchy requires radii on its leaves - IMP::atom::read_pdb add radii to the atoms - IMP::helper was merged into IMP::restrainer as all the functionality there was primarily to provide support for restrainer - IMP::domino2::SubsetEvaluatorTable was removed. Pass the model and IMP::domino2::ParticleStatesTable to the IMP::domino2::RestraintScoreSubsetFilterTable instead. - IMP::atom::Selection was added to allow specification of parts of IMP::atom::Hierarchy using biological sorts of names. A few functions to create restraints from them were added too. - a new example was added in the kernel to show IMP used on a biological system - IMP::display::PymolWriter now supports adding labels to things - IMP::display::PairRestraintGeometry and IMP::display::ConnectivityRestraintGeometry were added to aid in visualizing those restraints. - IMP::bullet was added with an optimizer to use hard body physics to resolve local collisions. - IMP::atom::setup_as_rigid_body() has been deprecated and replaced by IMP::atom::create_rigid_body(). The latter creates the rigid body as a separate particle so that the molecular hierarchy doesn't have to worry about having its bounding sphere large enough to enclose all the members. - IMP::core::KClosePairsPairScore was added to have count based cutoffs for close pair scoring. - IMP::algebra::get_surface_area_and_volume() was added to compute the two numbers for a set of balls ## Summer 2010 - The IMP::core::RigidDistancePairScore and IMP::core::ClosePairsPairScore have been revised to remove certain inconsistencies and ambiguities and to make them faster. Existing code using them will need to be slightly tweaked to work. - Javi added his 2D EM scoring functionality as IMP::em2D - some new ways to speed up scons have been added. See IMP::core::HarmonicDistancePairScore, IMP::core::SoftSpherePairScore and, for C++ users, the generic restraint support IMP::container::create_restraint() and IMP::core::create_restraint(). These each offer between 2x and 1.2x speedup when using simple scores (like harmonics on distances). In addition, a function IMP::container::get_slack_estimate() has been added which will estimate the correct slack to use with a IMP::container::ClosePairContainer. This will be extended to support IMP::core::ExcludedVolumeRestraint eventually. - support for dense and sparse grids have been added to IMP::algebra. See the IMP::algebra module page (I'm still working on how to make the docs clearer). - all write functions now have associated IMP::OptimizerState and IMP::FailureHandler classes. For examples, see IMP::WriteOptimizerState, IMP::WriteBinaryOptimizerState, IMP::WriteFailureHandler, IMP::atom::WritePDBOptimizerState etc. - functions IMP::write_binary_model(), IMP::read_binary_model() and IMP::read_configurations() have been added to support space and filesystem efficient storage of model configurations. Multiple configuration can be written to a single file. This functionality requires NetCDF be installed (NetCDF is a standard library for reading and writing binary data). - domino2 has been added. It provides a simpler interface on Keren's domino optimize as well as a number of new optimization and hooks for customizing the optimization process - a better explanation of how to use reference counted pointers has been added to IMP::Pointer. - all simple, planar polygons are now supported for display when CGAL is present ## Week of April 1, 2010 - All plural objects in Python are now simply Python lists rather than C++ recreations of Python lists. This makes compilation of the wrappers much, much, much faster. The visible changes are - types are not checked until the lists are passed to C++ - passing std::vector types by pointer and storing passed std::vector types by pointer will not work (but should never have been done anyway from code exposed to Python) - swig has issues with some overloads (not always for obvious reasons). These are being sorted out. None are commonly used. - get_particles() on a plural object doesn't work in Python (and is not needed in C++). Implicit conversions will always work. - All containers need to know the model at time of construction. If, after updating, you get an error on a container constructor, just add the model as the first argument. This also means that empty lists of particles in the constructors cause check failures. We can figure out a work around if this is too annoying. ## Week of March 13, 2010 - function get_rotation_in_radians_about_axis() was renamed to get_rotation_about_axis() as it didn't parse and we don't have the "in radians" elsewhere. - Added IMP::algebra::ReferenceFrame3D. As part of the cleanup, IMP::core::RigidBody::set_reference_frame() replaces set_transformation(). And the reference-frame related transformation construction functions were replaced. IMP::algebra::get_rotation_from_x_y_axes() was added to aid in the construction of reference frames (and replaces the algebra function taking three vectors). ## Week of March 6, 2010 - Publications now have a description field to give browsers an indication of what is in the publication to aid selection of what to read/cite. ## Week of Feb 27, 2010 - IMP::algebra::decompose_rotation_into_axis_angle() was renamed to IMP::algebra::get_angle_and_axis(). - renamed IMP::atom::bond() and IMP::atom::unbond() to IMP::atom::create_bond() and IMP::atom::destroy_bond() to be more consistent with things - renamed IMP::core::root() to IMP::core::get_root() ## Week of Feb 21, 2010 - IMP::Object::set_was_owned() was changed to IMP::Object::set_was_used() as that better reflects what the flag means. - various more interpolation and endian methods had their names fixed (by adding \c get_) - the function get_mean() on vectors was removed as it duplicated IMP::algebra::get_centroid() - build_Transformation3D_from_transformation3D was replaced by IMP.algebra.get_transformation_3d() - IMP::algebra::almost_equal() is now IMP::algebra::get_are_almost_equal() to fit the naming conventions. - the random vector generators have been cleaned up to remove redundant ones and make the names more consistent. They now have the form of either - IMP::algebra::get_random_vector_in() or IMP::algebra::get_random_vector_on() where there is one argument describing the volume or shape (such as an IMP::algebra::SphereD<4>). - or IMP::algebra::get_uniform_surface_cover() where the first argument is the primitive whose surface we are covering and the second is the number of points. Some good sed strings are - "s\#random_vector_in_unit_sphere()\#get_random_vector_in(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_in_unit_box()\#get_random_vector_in(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_on_unit_sphere()\#get_random_vector_on(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_on_unit_box()\#get_random_vector_on(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_in_box(\#get_random_vector_in(\#g" - "s\#random_vector_on_box(\#get_random_vector_on(\#g" - "s\#random_vector_in_sphere(\#get_random_vector_in(\#g" - "s\#random_vector_on_sphere(\#get_random_vector_on(\#g" - "s\#grid_cover\#get_grid_surface_cover\#g" - "s\#uniform_coverr\#get_uniform_surface_cover\#g" . - Spheres in IMP::algebra are now supported in general dimensions. See IMP::algebra::SphereD. - Many methods in IMP::algebra were renamed to be more consistent with other methods or general conventions. Most of these simply involved adding \c get_ as a prefix and/or add a \c _3d suffix when they did not take any arguments. - build_Rotation2D_from_Vector2D became IMP::algebra::get_rotation_to_x_axis() - most of the Rotation3D and Transformation3D functions (in addition, IMP::algebra::get_identity_rotation_3d() and IMP::algebra::get_identity_transformation_3d() added the \c _3d so we can add 2D versions at some point) - IMP::algebra::Transformation3D::get_transformed() and IMP::algebra::Rotation3D::get_rotated(). Replacing \c ".transform(" worked in the IMP code base. - IMP::algebra::get_distance(), this also carries over to the IMP::core::get_distance() The following global replacements are probably safe (in sed expressions) - "s\#core.distance(\#core.get_distance\#g" - "s\#algebra.distance(\#algebra.get_distance\#g" - "s\#\.transform(\#.get_transformed(\#g" Only in C++ - "s\#\.rotate(\#.get_rotated(\#g" Only in C++ - "s\#rotation_from\#get_rotation_from\#g" - "s\#rotation_in\#get_rotation_in\#g" - "s\#identity_rotation(\#get_identity_rotation_3d(" - "s\#::distance(\#::get_distance(\#g" . - IMP::algebra::Vector3D and IMP::algebra::Vector3Ds were replaced by the expanded (non-typedef) version in functions parsed by SWIG as the current situation was extremely brittle. For the moment, you should not use the typedefs in definitions that swig sees. Sorry. - IMP::algebra::Sphere3DPatch was renamed to IMP::algebra::SpherePatch3D to be consistent with the other geometry and allow us to support higher dimensions in the future. - rigid_align_first_to_second was renamed to IMP::algebra::get_transformation_aligning_first_to_second() and build_Transformation2D_from_point_sets was renamed to IMP::algebra::get_transformation_aligning_pair() ## Week of Feb 14, 2010 - The IMP::atom::Bond types were rearranged to reconcile the protein-centric classification with ligand support. - IMP::atom::Atom no longer has charge since we have IMP::atom::Charged. - IMP::atom::read_mol2() and IMP::atom::write_mol2() were added - IMP::atom::Selector and descendants renamed to IMP::atom::PDBSelector since we will soon have Mol2Selectors. In addition, the IMP::atom::IgnoreAlteriativesSelector is now IMP::atom::NonAlternativeSelector to keep the name positive, like the others. In the IMP body, all instances of the string Selector were for PDBSelectors and so sed was safe to use. - IMP::atom::read_pdb() lost the "ignore_alternatives" argument. Instead, all of the provided selectors, other than IMP::atom::AllSelector ignore alternatives internally. Code that uses the IMP::atom::AllSelector should probably change to the IMP::atom::NonAlternativeSelector. - IMP::example renamed (from examples) - The IMP_USAGE_CHECK() macro lost the (unused) third argument. ## Week of February 3, 2010 - The model io code as moved into the kernel so it is now IMP::write_model() and IMP::read_model() - Remove the version_info argument from the implementation macros. Instead, just pick up the module version info automatically. - Add a function Object::get_type_name() and Object::get_module() to get the respective bits of information about each object. - Change Object::show() to be a non-virtual wrapper which prints the object name and other info before passing control off to Object::do_show(). - The containers were moved to IMP.container. See that module page for a full list of what moved. In the process, the backwards compatibility functions were removed from IMP::container::PairsRestraint and kin. - IMP.atom.create_protein() now returns the created protein - the simplify functions have been moved from IMP.helper. They are now IMP.atom.create_simplified_along_backbone(). - IMP.helper.clone() has been renamed to IMP.atom.create_clone() - IMP.misc.ConnectingPairContainer has been moved to IMP.core.ConnectingPairContainer - The locations for pixels in IMP::em::DensityMap have been moved to the center of the corresponding voxel from the lower left corner. - IMP::em::get_transformed(), IMP::em::get_resampled() and IMP::em::get_density() have been added. - The documentation has been majorly cleaned up and simplified. Please look around and see if you like it. There are now two main pages, the introduction and the developers guide, rather than the previous plethora of pages. - IMP::algebra::Cone3D::get_bounding_sphere() has been removed since it was not of general use, not returning the bounding sphere you would expect and was trivial to implement. - IMP::algebra::projection() was removed since it didn't actually project onto the segment and the name didn't follow the normal conventions. And it wasn't used. - Most geometry now has namespace methods IMP::algebra::get_bounding_box(), IMP::algebra::get_surface_area(), IMP::algebra::get_volume(). ## Week of January 22, 2010 - A generic geometry layer has been added (see \ref genericgeometry). Using it, IMP::atom::rmsd(), IMP::algebra::alignment_taking_first_to_second() and the new IMP::atom::native_overlap() all can work on either collections of either IMP::core::XYZ particles or IMP::algebra::Vector3Ds. - Deprecated functionality has been removed from IMP.core. For those still using it, IMP::core::ClosePairsScoreState was replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairsScoreState was replaced by IMP::core::CloseBipartitePairContainer. - The macros for Scores changed slightly. Scores not in IMP svn which use the IMP_*_SCORE macros now have to implement the get_is_changed() function outside of the class declaration and scores which use the IMP_SIMPLE_*_SCORE now don't have to implement the functions to get interactions and used particles. - Rename IMP::core::MinimumPairScoreRestraint (and kin, Maximum, Singleton, Triplet and Quad) to IMP::core::MinimumPairRestraint. - Add IMP::core::MinumumPairScore (and kin). ## Week of January 1, 2010 - IMP::em is now using Particles to resample density maps and for fitting restraints. The previous ParticleAccessPoint mechanism is removed. - IMP::UsageException and IMP::InternalException are now reserved for their respective checks (and cannot be thrown by IMP_THROW()). This makes the distinction between API exceptions (which are thrown in any build mode and should be documented and tested) and helper exceptions which are only thrown when checks are turned on (and hence should not be documented) clearer. - IMP::core::RestraintSet moved to IMP::RestraintSet. The restraint sets now simply define a tree with weights which the model traverses as opposed to being an opaque IMP::Restraint themselves. This means that incremental evaluation of restraints in restraint sets now works and that restraints in restraint sets can be individually evaluated. Existing working usage should not have changed (so \c sed should manage updates just fine), but the internal implementation is completely new, so pay attention. - IMP now has a text i/o layer consisting of IMP::TextInput and IMP::TextOutput. These classes are convertible to C++ streams and implicitly constructable from both streams and file paths. As a result, functions that take them will work when called with a string for a file path, a Python file or a C++ stream. - The IMP.examples module now documents how to use the needed aspects of scons and SWIG. - The IMP.examples module is used as the basis for new modules created by the make-module script, so those include more documentation and are tested. - Module overview pages are now generated from the information passed in the SConscript file, rather than from the (now removed) modulename.dox file. See the examples file for an example of the things you can do there. - boost program options is now a checked for dependency. - The IMP::Model now uses dependency information from IMP::ScoreStates to automatically reorder the IMP::ScoreStates so that they are evaluated in the right order. Also, IMP::Model::evaluate() on a subset of the restraints now only updates the needed score states, not all of them. - Functions on rigid bodies no longer provide the IMP::core::RigidMemberRefiner as a default as it is, pretty much always, the wrong thing to use. That class may go away without notice in the future and is no longer documented. - the IMP::helper::create_simple_connectivity_on_rigid_bodies() and IMP::helper::create_simple_excluded_volume_on_rigid_bodies() now both take an IMP::Refiner as an (optional) final argument. This makes it so they can be used with the rigid bodies created from molecular hierarchies. This code still needs to be updated and the docs don't reflect what the code was or is doing, so use with caution (see notes in the header file). - The IMP::helper create* functions now all take non-Object inputs as const& to make clear what is input and that there are no memory management gotchas. - a new base class IMP::Constraint has been added for constraints. Various other classes have been renamed to be consistent. ## Week of December 11 - The interface of IMP::PairScore changed slightly to make the function signatures more uniform across various methods. Now all methods that take a pair of IMP::Particle objects take a const IMP::ParticlePair&. Previously some took that, some took an IMP::ParticlePair and some took two IMP::Particle*s. If you have an implementation of an IMP::PairScore outside of svn, you will have to update your function signatures (the compiler will tell you where). Backward compatible IMP::PairScore::evaluate() methods have been provided, so simple users of pair scores shouldn't see a difference. - IMP::core::PairRestraint takes a IMP::ParticlePair in the constructor instead of two IMP::Particle*s. A backwards compatible constructor can be added if needed. - IMP::ParticlePair::first and second went away since no one claimed to be using it and it complicated the implementation. - The generic restraints/score states/scores infrastructure has been extended to triplets and quads (for angles and torsion angles). ## Week of December 4 - IMP now makes use of Boost.Filesystem to manipulate paths. Things may not work well on systems where boost libs are not available which does not support posix paths. But libless boost installs generally don't exist outside of our test systems. - new functions like IMP::core::get_example_path(filename) and IMP::core::get_data_path(filename) allow one to easily find a data file named 'filename' for the examples and library internal data. - The IMP::display module has been revamped in order to support non-linear geometry. VRML support has been dropped. CMM support will be dropped unless someone requests it (as IMP::display::ChimeraWriter does all it does and more). ## Week of November 19 - imp.xcodeproj was moved into tools and updated accordingly - the bin directory finally went away - IMP::display::ChimeraWriter now supports the latest version of Chimera - IMP::display::CGOWriter was renamed to IMP::display::PymolWriter as it can do more than just CGO. - IMP::VectorOfRefCounted::operator[] can now be used for assignments just like a normal vector. No more need to use the IMP::VectorOfRefCounted::set() method. - the IMP::atom::Hierarchy constructor from an IMP::Particle* is now explicit (as were the decorator constructors) ## Week of November 12 - All exception types are now mapped to equivalent classes in Python. The class hierarchy is similar (e.g. IMP::ModelException derives from IMP::Exception in both C++ and Python) so all IMP exceptions can be caught in Python using "except IMP.Exception". For convenience, the IMP::IOException, IMP::ValueException and IMP::IndexException classes in Python also derive from the similarly-named standard Python classes (IOError, ValueError and IndexError respectively). ## week of October 29 Update - You can now add members to rigid bodies via IMP::core::RigidBody::add_member(). These members can be other rigid bodies (although this functionality is only lightly tested). - The code to get bounding volumes for IMP::atom::Hierarchy has been moved to the atom module, as as the clone and destroy code. They can be found at IMP::atom::clone(), IMP::atom::destroy(), IMP::atom::get_bounding_box(). - Added IMP::core::VolumeRestraint. - Several functions have been added to helper in creating restraints. They are create_simple_connectivity_on_rigid_bodies(), create_simple_connectivity_on_molecules(), create_simple_distance(), create_simple_diameter(), create_simple_em_fit(), and create_simple_excluded_volume_on_rigid_bodies(). Functions load_em_density_map() and set_rigid_bodies() are added as well. See IMP::helper for more information. - IMP::ScoreStates are further hidden. Namely, IMP::core::RigidBody constructs the needed score states internally, doing away with the need for the create functions (which have been done away with). In addition there are new decorators IMP::core::Cover and IMP::core::Centroid which manage their respective score states, further doing away with more create methods. The hierarchy-based create method for rigid bodies has been moved to the atom module and is called IMP::atom::rigid_body_setup_hierarchy(). - IMP::em::read_map() will figure out the file type to use automatically from the extension. Other such methods should follow suite. - A class IMP::core::StructureSet has been added to provide a convenient way to manage solutions of optimization runs. - IMP::core::RigidBodyMover now takes its angle argument in radians and is documented as such. - IMP::core::ClosePairsScoreState and IMP::core::CloseBipartitePairsScoreState have been replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairContainer. The score states have been marked as deprecated and are mostly still there, but are more limited than they were. - IMP::core::AllPairsPairContainer has been renamed to IMP::core::AllPairContainer and a bipartite version has been added. - All containers now keep track of what they looked like on the previous time step. - The IMP::core::ListPairContainer and IMP::core::ListSingletonContainer have had their interface reduced a bit in order to make implementing change tracking easier. - The set of exceptions has been simplified to go along with simplification of the checks. We now have IMP::InternalException and IMP::UsageException which are thrown by the respective checks (and the former is thrown by IMP_FAILURE()). When you want to to pass an exception back to Python use one of IMP::ValueException or IMP::IndexException or IMP::IOException. There is also a new macro, IMP_THROW() to be used when you want to throw an exception with a nice error message. Please look at code you have worked on and make sure that the exceptions make sense given the new clarifications. - An IMP::ModelException is now thrown when IMP::Particle attributes attain invalid values. This exception can be caught to recover from failures of IMP::core::ConjugateGradients or difficult restraints. ## Week of October 15 - The set of functions needed to get meta information about IMP::Restraint and IMP::ScoreState objects has hopefully stabilized. Sorry about the thrashing. If you have restraints or score states not in SVN, see the corresponding IMP_RESTRAINT type macros, the IMP::Interaction class and base classes for what exactly you need to implement. ## Week of October 8 - IMP::Interaction was modified to split IMP::interaction::get_used_particles() into IMP::interaction::get_read_particles() and IMP::interaction::get_write_particles(). This makes it possible to check for IMP::ScoreState objects being in an inconsistent order (so that the input to one is updated after it is read). Later, we will be able to automatically order the score states ensure that there are no problems. - IMP::atom::Hierarchy has been modifier to get rid of the IMP::atom::Hierarchy::Type. Instead, all types are contained in the decorators. IMP::atom::get_by_type() was also updated to reflect this. - IMP_check() and IMP_assert() have been replaced with IMP_USAGE_CHECK() and IMP_INTERNAL_CHECK() respectively. Hopefully this makes it clearer when to use which. IMP::set_check_level() has been revised accordingly (to use IMP::USAGE and IMP::USAGE_AND_INTERNAL). Also, IMP_failure() and IMP_not_implemented() have been renamed to IMP_FAILURE() and IMP_NOT_IMPLEMENTED() to make all macros be all caps. To update external code, - replace IMP.EXPENSIVE with IMP.USAGE_AND_INTERNAL - replace IMP.CHEAP with IMP.USAGE - replace IMP_assert and IMP_check as above. ## Week of October 1 - IMP::ScoreState and IMP::Restraint now better track the set of particles that they use and the dependencies between the particles. To this end, a new base class IMP::Interaction has been added with methods to get the inter-particle interactions and the set of particles used. - .cpp, .h, .py and SConscript files which contain the string "DO NOT COMMIT" will be rejected on checkin. Use this to make experimental or debugging changes you are making so you don't accidentally commit them. - The search module is no longer built by default and is not longer supported. It will get moved out of svn shortly. - IMP::ScoreState objects now also have IMP::Interacting::get_interacting_particles() methods. Soon IMP::PairScore and IMP::SingletonScore objects will too. ## Week of September 27 - Some support for gather statistics has been added to IMP::Model (see IMP::Model::set_gather_statistics()). If requested, an API can be provided so that the individual data points can be returned to programs. - A class IMP::core::RigidBodyDistancePairScore has been added which provides accelerated computations for the distance between two rigid bodies. - The SWIG input files were renamed from swig.i to swig.i-in to aid in dependency tracking. - rigid bodies now have two sets of particles associated with them in the various places they are used. One is the IMP::core::RigidBodyMember particles, as was previously the case. The other is defined through a refiner (and could be the RigidBodyMember particles if a IMP::core::RigidMembersRefiner is used). The idea is that you sometimes want some particles which are transformed along with the rigid body, but which are not really part of the representation of the rigid body. Examples would include marker points or multiresolution representations of proteins. ## Week of September 20 - all \c .i files have been renamed to start with \c IMP_ to help make dependency tracking easier. - static builds (builds with \c static=True) build both the static and dynamic libraries and just static executables - the \c profile build target has been removed as \c static=True does the important bit of what is needed. Just do \c linkflags=['-pg'] to add the needed link flag for \c gprof. - Restraints should now implement IMP::Restraint::unprotected_evaluate() instead of IMP::Restraint::evaluate(). The latter can now be called outside of IMP::Model::evaluate() and will return the right answer always. - \c bin/imppy.sh no longer works, as was previously warned. Use \c tools/imppy.sh instead. The bin directory will go away eventually, for now the script just tells you to use the tools version. ## Week of August 30 - A new score IMP::core::BoundingBox3DSingletonScore has been added to keep particles in a box. - str should now work better in Python when applied to IMP::Object objects and a few others. Objects should use the new IMP_SHOWABLE and IMP_SHOWABLE_INLINE macros instead of declaring show functions directory to make it so that they work in Python. It is now a bug to \c print or \c str to not return something useful in Python for any IMP object. ## Week of August 23 - each module now has a version, eg IMP::get_module_version_info() which is also returned by the various restraints of the module. The Python code checks that loaded libraries have the expected version numbers to try to provide better messages when runtime linking problems occur. - the various bounding_box methods have been renamed to ::get_bounding_box() to be more consistent with the other names. - Python documentation is now parsed by doxygen ## Week of August 15 - The Object::get_version_info() result now includes the SVN revision info when building from SVN. This means you can use it to more easily tell exactly what a built version of IMP was built from. ## Week of August 8 - Treatment of hetatoms in hierarchies (and hence pdbs) is now documented. See IMP::atom::Hierarchy. This involved broadening IMP::atom::Residue to handle ligands and having IMP::atom::Hierarchy::LIGAND, IMP::atom::Hierarchy::AMINOACID and IMP::atom::Hierarchy::NUCLEICACID be handled by IMP::atom::Residue. - IMP::atom::read_pdb() currently drops HOH ATOM records on the floor. We should add a IMP::atom::Water decorator if anyone wants to handle them (since they aren't atoms and don't have elements). - IMP::atom::Mass() has been added for decorating particles with mass. IMP::atom::Atom mass has been moved to that. - The scheme for generating atom types is now fully specified. See the IMP::atom::AtomType documentation. - IMP::atom::Atom and IMP::atom::Residue have been somewhat restructured. Some of the helper functions have been renamed to be more descriptive or removed (if they duplicated other easily accessible functionality). get_residue_index() has been removed (since you can just call get_index() on the return result of the IMP::atom::get_residue()). IMP::atom::get_chain() now returns an IMP::atom::Chain. root() was renamed to IMP::atom::get_root() for consistency. - SWIG files are now generated from a file called swig.i in the pyext directory. This file only has to contain module-specific commands such as lines to wrap headers and instantiate module templates. No more boilerplate. - The contents of bin are moving to tools since the two directories both contained assorted development tools (not all of which were executable). For now, bin/imppy.sh will still work, but that will go away soon. - IMP::Decorator::show() now just takes the stream (it used to take a prefix also, which was often ignored). Anyone who implemented a decorator will have to delete the second argument of their show function. ## Week of August 1 - Containers now have methods IMP::SingletonContainer::evaluate(), IMP::PairContainer::evaluate(), IMP::SingletonContainer::apply() and IMP::PairContainer::apply(). Using these can be significantly faster than looping over the particles externally. Similarly, there is IMP::SingletonScore::evaluate(), IMP::PairScore::evaluate(), IMP::SingletonModifier::apply(), IMP::PairModifier::apply(). - The decorator functions have been renamed to IMP::Decorator::setup_particle(), IMP::Decorator::decorate_particle() and IMP::Decorator::particle_is_instance(). A shell script \c tools/update-decorators is provided which updates the names in all code in subdirectories of the current directory. Please backup things first. - a function, IMP::Restraint::safe_evaluate() has been added to be used when requesting the value of restraints outside of the IMP::Model::evaluate(). ## Week of July 18 - The atom::BondPairContainer has been renamed to atom::BondPairFilter since it wasn't a real container. A real atom::BondPairContainer has been added. - Out of source builds mostly works. To do an out of source build, make a link to \c scons_tools and \c SConstruct from the source directory in the build directory and then add the line \command{repository="../path_to_src_dir"} in the \c config.py in the build directory. Then running \c scons in the build directory will put generated files there. This allows you to make both a \c release and \c fast build off of the same set of sources. Proper docs will be added soon. - The SConscript setup for a module has been greatly simplified so it should be easy enough for users to maintain their own module and updates to the build system shouldn't involve changes to user SConscripts any more. This does mean that any non-svn modules need to be updated. Look at IMP.core or another module and copy what is there. Again, proper docs will be added soon. ## Week of July 11 - Per Javi's suggestion, the conventions have been split into "Coding conventions" and "API conventions" and linked from the main page. - benchmarks got moved to their own module, IMP::benchmark and will now be tracked over time ## Week of July 4 - FilteredListSingletonContainer and FilteredPairSingletonContainer went away as there was too much hard to factor code shared with IMP::core::ListSingletonContainer and IMP::core::PairSingletonContainer. The IMP::core::ClosePairsScoreState is now handled by a list of filters on the class itself. That is, do IMP::core::ClosePairsScoreState::add_close_pair_filter() to add a filter. - Hooks have been added to allow actions to be taken on an assertion or check failure. See IMP::HandleFailure and its descendants for more details. - IMP::display::CGOWriter has been much improved and now allows provides lots of display-time control and easy building of animations. - Generic optimizer states have been added (IMP::core::SingletonsOptimizerState and IMP::core::PairsOptimizerState). ## Week of June 15 - box distributions (IMP::algebra::random_vector_on_box(), IMP::algebra::random_vector_in_box()) take bounding boxes - IMP::em::DensityMap are now IMP::Object objects. Some current allocations on the stack may need to be changed (since ref counted objects can't be stack allocated). - IMP::em::DensityMap::Read() is now deprecated (as is IMP::em::DensityMap::Write) as they violated a slew of conventions. Use IMP::em::read_map() and IMP::em::write_map() instead. ## Week of June 8 - new methods IMP::atom::add_radii() and IMP::atom::add_bonds(); - Hierarchys to IMP::atom::Hierarchies - IMP::Particles has been changed to a new type. It no longer plays quite so well with Python lists (you can't create one from an arbitrary Python list). - The containers for multiple decorators have been revised to make them more useful. See IMP::Decorators. You can now implicitly convert, for %example, a IMP::core::XYZs to an IMP::Particles (as well as use IMP::Decorators::get_particles()). All methods that take a collection of particles with a certain type now instead take a IMP::Decorators. - A method that has expectations about the particles it takes should take decorators rather than particles. That is, anything which expects IMP::core::XYZ particles, should take an IMP::core::XYZs instead of IMP::Particles. - IMP_LIST() now takes an extra argument (the name of the plural container) - In C++ IMP::Decorator objects are implicitly convertible to IMP::Particle* objects. Unfortunately, making this work in Python is problematic. - More methods take and return decorators rather than IMP::Particles. This means it is clearer what their expectations are and what they produce. ## Week of May 15 - The IMP::helper module was added and various functionality moved there. See its documentation page for a list of what is there. - IMP::algebra::SphericalCoords was renamed to IMP::algebra::SphericalVector3D to be consistent with other algebra classes. - IMP::algebra::Rotation2D is now IMP::InvalidDefault to be consistent with the other similar classes. - We are trying to be a bit more specific about what is expected from a IMP::atom::Hierarchy. Towards that end, there are two methods for extracting a representation from a hierarchy, IMP::atom::get_detailed_representation(), and IMP::atom::get_simplified_representation() - A method has been added to produce a simplified copy of an IMP::atom::Hierarchy. The method is still a bit crude and could benefit from other people inspecting the output and saying how it should be improved. - There is now an IMP::algebra::BoundingBoxD (and IMP::algebra::BoundingBox3D) to track bounding boxes. There are methods like IMP::algebra::bound_box(Sphere3D) which gets the bounding box of a sphere. - Each IMP::Object has its own log level. If IMP::Object::set_log_level() is called with a value other than IMP::DEFAULT, the value is used for code within and called by the object. To support this in your objects you need to add IMP_OBJECT_LOG to the start of each major method. It has been added to some of the core objects. ## Week of May 8 - IMP::atom::MolecularHierarchy has been renamed to IMP::atom::Hierarchy. Use \command{sed -i.old "s\#MolecularHierarchy\#Hierarchy\#g" my_files} - the long-deprecated Name(Decorator) was removed - The decorator suffix has been removed from decorator names. - Unit conventions have been established and documented for distance, energy, force and angle. - there is a function IMP::atom::create_protein() to create a coarse grained protein. The function is crude so far, but the idea is that we should have standard ways of generating such things so that everyone can benefit from best practices. - a class IMP::atom::BondGraph has been added so Boost.Graph algorithms can be used with bond graphs - the IMP-specific doxygen commands have been documented. - the doxygen commands unstable{name} and untested{name} have been added to doxygen to be used for marking methods and functions as likely to change or not well tested. Pass the name of the class or method as name. - bonds can now be added to molecular hierarchies automatically see IMP::atom::add_bonds() ## Week of May 1 - There is a IMP::Rotation2D - You can now extract fixed Euler ZYZ angles from a IMP::Rotation3D. This code can serve as a template for extracting other Euler angle conventions if they are needed. - IMP::algebra::rotation_between_two_vectors() changed to IMP::algebra::rotation_taking_first_to_second() to make it clear which direction the rotation goes. ## Week of April 27 - IMP::Decorator derived objects are now convertible to bool. So code like \code if (d) { blah } \endcode works. - IMP::Decorator::cast now returns a null decorator (\c DecoratorName()) rather than throwing an exception. This might make it actually useful. - New IMP::statistics module for statistical methods (e.g. clustering). ## Week of April 20 - Support for taking advantage of rigid bodies in collision detection has been added via the IMP::core::RigidClosePairsFinder. - All IMP::core::ClosePairsScoreState-related functionality has been optimized and should be noticeably faster. - The EMBED project has been merged into the IMP::em module. Thus, it is no longer necessary to obtain EMBED separately to build IMP::em. - All C++ methods which work with std::istream or std::ostream arguments should now be usable from Python: any Python file-like object can be used. - The SWIG interface now ties in to the reference counting mechanism used in the C++ code, for all classes derived from IMP::RefCounted. So, for %example, deleting a Python object should not free the underlying C++ object if another C++ object holds a reference to it, and unrefing a C++ object should not delete it if a Python object still holds a reference to it. There should be no further need to manually keep references to Python objects to prevent things from being prematurely cleaned up. This should be transparent to the user; please report any problems. - IMP::atom::AtomDecorator no longer inherits from IMP::core::XYZDecorator. ## Week of April 13 - The decorators IMP::atom::AtomDecorator, IMP::atom::ChainDecorator, IMP::atom::ResidueDecorator etc. now inherit from IMP::atom::MolecularHierarchyDecorator. This means they can set their IMP::atom::MolecularHierarchyDecorator::Type properly and saves a lot of typing in the common case. We can always make hierarchy decorators zero memory overhead if we find that the cost matters. - The base classes IMP::Comparable, IMP::NullDefault, IMP::ValidDefault and IMP::UninitializedDefault were added to document what the default constructors do for various objects in IMP. ## Week of April 6 - IMP::Object now defines pure virtual methods IMP::Object::get_version_info() and IMP::Object::show(). All non-abstract classes which inherit from IMP::Object must implement them. Doing this means they are more uniformly implemented and are documented in one place. - IMP::Object is now reference counted (the only objects this changes are IMP::Model and IMP::Optimizer). As a result optimizers and IMP::Model objects must be created on the heap. - IMP::RefCounted has had its name changed and no longer inherits from IMP::Object. This makes it easier to use independently. - the docs for reference counting (IMP::RefCounted) and and IMP::Decorator have been improved - running doxygen produces no warnings. Let's try to keep it that way. Run \command{scons doc > /dev/null} before you check in large blocks of new code to make sure that you didn't introduce warnings. - Helper classes have been added to make SingletonModifier and PairModifier objects look like functors for C++ and Python. This makes it so you can use \c map in Python and \c for_each in C++. See IMP::SingletonFunctor and IMP::PairFunctor for more information. - The recent changes page has moved to a doxygen page to make it easier to link changes back to the relevant source code. ## Week of March 20: - create_rigid_body(), create_rigid_bodies(), create_centroid() etc. now require you to add the score state to the model yourself. Whether the rigid body is enforced or not can be controlled by adding or removing the returned IMP::ScoreState. - All previously deprecated code has been removed. This includes indexes in particles, the no-argument constructor for particles and the functionality that was moved from core to atom. - The IMP::core::RigidBodyDecorator and IMP::core::RigidMemberDecorator have been simplified by removing the traits classes. There is now only one type of rigid body and it stores the list of member particles internally. - IMP::core::NameDecorator has been deprecated as IMP::Particles have built in names (use IMP::Particle::get_name(), IMP::Particle::set_name()). - IMP::atom::ChainDecorator has been added ## Week of March 13: - A new IMP::display::Writer, IMP::display::ChimeraWriter has been added which exports geometry to native Chimera constructs. It probably should be used instead of bild and cmm for most purposes as it combines the best of both types. - compose() for transformation is now a free function so compose(rotation, vector) will work via promotions. - the geometry classes have been cleaned up and are now more consistent with IMP standards (and more likely to do what their names/comments suggest). See IMP::algebra::Sphere3D, IMP::algebra::Cylinder3D, IMP::algebra::Segment3D, IMP::algebra::Plane3D etc. - ParticleRefiner has been renamed to IMP::Refiner since the Particle was long and unnecessary. eIF1â˘eIF3 translation initiation complex](http://salilab.org/40S-eIF1-eIF3). - IMP::domino::DiscreteSampler::set_restraints() now allows the set of restraints used in DOMINO filtering to be controlled (previously, all restraints in the IMP::Model were used). - IMP::modeller::IMPRestraints now transfers an IMP::ScoringFunction into the MODELLER score, not the implicit (and deprecated) score of the entire IMP::Model. - Bugfix #775: coordinates of voxels in EM maps were previously inconsistently treated (they are supposed to be the centers of the voxels, but in some cases they were the lower left edge). The fix may result in maps and their bounding boxes appearing shifted by half the voxel size. # 2.4.0 - 2015-02-27 # {#changelog_2_4_0} - Python 3 is now supported. The Windows package now works out of the box with Python 3.3 or 3.4; a `-python3` subpackage is provided for Ubuntu and Fedora. (Apple does not yet provide Python 3 so the Mac package is not built with Python 3 support, but if you install Python 3 you can compile IMP from source to work with it.) - A new command line tool, `rrt_sample`, allows for running rapidly exploring random tree (RRT) based sampling, e.g. to sample conformations of multiple domains connected by flexible linkers. - Command line tools are now placed in the `bin` directory of modules (the old `bin` directory, used for utility programs that are not installed, is now called `utility`). The `applications` directory is removed. - Several Python modules that were only used by EMageFit have been moved from IMP::em2d to IMP::EMageFit. - A new tutorial that uses the IMP::pmi module is now available (modeling of the RNA Polymerase II stalk). # 2.3.1 - 2014-12-01 # {#changelog_2_3_1} - Minor bug fixes. - Update to IMP::pmi to support the Nup84 and rnapolii biological systems. # 2.3.0 - 2014-10-24 # {#changelog_2_3_0} - A new IMP::pmi module was added, containing high-level Python classes to simplify the representation, scoring, sampling and analyzing of models in IMP. This module is still under heavy development, and should be considered experimental. - Binary installers for 64-bit Windows are now available. - The `.py` extension has been removed from all Python applications, so that they are consistent with the C++ applications. - IMP::atom::Selection objects can now be combined using basic set operations (union, intersection, difference, symmetric difference). In Python the equivalent operators (|, &, -, ^ respectively) can be used. - The integrative docking (idock) application now employs SOAP to improve the scoring. - IMP::atom::CHARMMParameters::create_topology() now works with a wider range of hierarchies (previously it would only accept Residues that were direct children of Chains). For example, it should do the right thing when given a Hierarchy that contains no Chain particles, or one where Residues are grouped under Fragments. - Only Python 2.6 or later is now supported. We still provide IMP for RedHat Enterprise 5 (which ships with Python 2.4) but you will need to use the `python26` binary (part of EPEL) rather than the system default Python. # 2.2.1 - 2014-08-11 # {#changelog_2_2_1} - Bugfix: PoseScore (see the ligand_score application) should now match that used by the Pose&Rank web server (http://salilab.org/poseandrank/) # 2.2.0 - 2014-04-09 # {#changelog_2_2_0} - A module IMP.template was added to act as a template for creating new external modules. It includes setting up git hooks for format checks and things like that. - The Python support for grids was cleaned up. Hopefully this makes it more robust. If you were using a grid in Python that no longer exists, ping us, we can add it back. - The generic geometry functions were simplified slightly and renamed to IMP::algebra::get_vector_geometry(). - The IMP::base::Flag class was added to simplify adding flags in C++. - The `--log_level` and `--check_level` command line argument now take strings (eg VERBOSE) to specify their values, rather than cryptic (and a bit unstable) numbers. - Command line options are now divided into normal and advanced ones. You do `--help_advanced` to show the advanced ones. - Support for computing Connolly surfaces was added with IMP::algebra::get_connolly_surface(). In addition, there is now IMP::algebra::get_uniform_surface_cover(const IMP::algebra::Sphere3Ds&,double) to sample the surface of a collection of balls. - IMP::atom::reate_simplified_from_volume() was added to create a simplified representation of a hierarchy that preserves the surface are and occupied volume. It uses IMP::algebra::get_simplified_from_volume(). If you want to play with such representations, you can use the rmf_simply application. - IMP::atom::get_rmsd(IMP::atom::Selection, IMP::atom::Selection, const IMP::algebra::Transformation3D&) was deprecated. Use IMP::atom::get_rmsd_transforming_first() instead. - IMP::algebra::get_rmsd() and IMP::algebra::get_rmsd_transforming_first() were added. - The family of macros IMP_BASE_[SMALL,LARGE]_[UNORDERED,ORDERED]_[MAP,SET] have been added (eg IMP_BASE_SMALL_ORDERED_SET ()). They provide platform and boost-version independent ways to take advantage of the best set/map for an application. In particular, they use boost::container::flat_set and kin when available. - IMP::atom::State and IMP::atom::Representation decorators have been added to provide support for multiple states and multiple representation schemes. They are integrated with RMFs and IMP::atom::Selection. See atom/multiresolution.py and atom/multistate.py for examples. - The container implementations has been cleaned up and simplified a bit. This should not change most usage. - [RMF](http://salilab.github.io/rmf) has been updated. There are a few minor API changes, in addition to a new, faster format (`.rmf3`). - [Eigen](#Eigen) is now included with IMP. - The deprecated function IMP::kernel::Optimizer::set_restraints() was removed. Use IMP::kernel::Optimizer::set_scoring_function() instead. - The deprecated constructor for IMP::kernel::ModelObject that does not take an IMP::kernel::Model was removed. This propagates to constructors IMP::kernel::Restraint, IMP::kernel::ScoreState and kin. Make sure you pass an IMP::kernel::Model as the first constructor argument. Related to this, deprecated functions like get_is_part_of_model(), set_model() were removed as they don't serve any purpose. - The deprecated methods in IMP::kernel::Optimizer that pertain to attribute optimization were removed. Inherit from IMP::kernel::AttributeOptimizer instead if you want to use them in your optimizer. - Some deprecated functionality was removed. If you encounter problems, replace IMP::base::OwnerPointer with IMP::base::PointerMember, find logging stuff in IMP/base/log.h, object in IMP/base/Object.h and random in IMP/base/random.h. - IMP::atom::Chain now uses strings for the chain ids, to be consistent with changes in the PDB. PDB files are written using the first character of the chain string. - Added orientation-dependent SOAP scoring (IMP::score_functor::OrientedSoap and IMP::atom::OrientedSoapPairScore). Orientation-dependent SOAP scores include those that score loops (SOAP-Loop), protein-peptide interfaces (SOAP-Peptide) and proteins (SOAP-Protein). The library files themselves are rather large (~1.5GB) and so are not included with IMP. They can be downloaded separately from http://salilab.org/SOAP/. # 2.1.1 - 2013-11-04 # {#changelog_2_1_1} - Added support for Mac OS 10.9 # 2.1.0 - 2013-10-24 # {#changelog_2_1_0} - There is now some limited IMP-wide gathering of timing statistics. You can add `--statistics=filename` to the command line or use the IMP::base::show_timings() function to view a summary of where IMP has spent time. - Slack was being used inconsistently (and not always correctly) among the various containers and restraints. If you had heavily optimized the slack value for speed, you may want to revisit it. - Class name OwnerPointer was switched to PointerMember, to denote more clearly a ref-counting pointer that is a class member. - The dependency handling in IMP has been rewritten to simplify it and fix a bunch of bugs. There are now more checks of correctness, so behavior that might have silently worked (or not worked) before may now trigger errors. In particular, restraints/score states must have access to the IMP::kernel::Model to do much of anything, so either use constructors that take the model or call IMP::kernel::ModelObject::set_model(). Also, changing your set of inputs or outputs without calling IMP::kernel::ModelObject::set_has_dependencies(false) is now likely to trigger a check failure (it would just silently do the wrong thing before). - Decorators have been cleaned up to make them behave consistently regarding IMP::kernel::Particle vs IMP::kernel::ParticleIndexes. This change shouldn't break existing code, but decorators should move to the new macros IMP_DECORATOR_METHODS(), IMP_DECORATOR_SETUP_0() etc. - The IMP_GRAPH() and IMP_WEIGHTED_GRAPH() macros got an extra argument describing how to show the vertexes in the graph. They also now expose a method, eg IMP::kernel::show_as_graphviz() that writes the graph in graphviz format to an IMP::base::TextOutput. - Support and rules for deprecating code in IMP have been updated. There are new macros and instructions on how to use them that should result in much better visibility about changes. See the C++ wiki for IMP for more information. And there is now an encoded policy of keeping deprecated things around for 1 release after deprecation. - IMP is now much more selective about when it invalidates the dependency graph. See [Dependencies](http://github.com/salilab/imp/wiki/Dependencies) for more information. This change should make evaluation of isolated restraints much faster in certain situations where they were unexpectedly slow before. - IMP::atom::Selection has been completely rewritten as it was very brittle. The set of particles returned may change in some cases, in some cases because it was buggy before, in some cases because of new bugs. - Reminder: the IMP_PROTECTED macros didn't work out well, and are now deprecated. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading `_` on the method name, if you get `Swig director pure virtual method called IMP::kernel::PairScore::do_get_inputs` messages. - The generated `cmake` files are no longer stored in the repository. Instead `tools/build/setup_cmake.py` is run after checkouts and other git events that change the repository tree (assuming you have run `setup_git.py`). - Generation of documentation using `doxygen` has been refactored. It is now done on a per-module basis, via targets like `IMP.core-doc`. The main advantage is that changes to the doc for a module can be tested comparatively quickly, just by building that module's docs. On the down side, docs are no longer quite so heavily interconnected, so dependencies only work to things that the module actually depends on. - The restraint and score state statistics functions were removed from IMP::Model. They had been broken for quite a while, with no one complaining. Statistics can be added back easily if requested. - Added an IMP-specific fork of `git-flow`. It handles things like `README.md` files when you start and finish feature branches and provides a tool to nicely update IMP and display the change log. The main commands of use are `git imp feature start <feature_name>`, `git imp feature finish <feature_name>`, and `git imp update`. You may want to remove the `git-flow` lines from your `.git/config` file as you may accidentally type `git flow` instead of `git imp`. You need to run `setup_git.py` to set things up. - [RMF](http://salilab.github.com/rmf) is now included as a git submodule rather than copied into the IMP repository. You should rerun `setup_git.py` to make sure submodule stuff is initialized. # For IMP 2.0 and earlier # {#changelog_2_0} ## Spring 2013 - Modules and applications now have a _Maintainer_ in their `README.md`, giving the github id of the person to whom issues should be assigned. - `setup_git.py` supports the `--module` flag which will set up git hooks and things for a git repository that contains a module. - To clean up for the release, the module dependencies are now described in a file called `dependencies.py` and the module docs in a `README.md`, both in the main module directory. - IMP-wide support for progress bars was added with the IMP::base::set_progress_display() function and helpers. These are only displayed when the IMP::base::get_log_level() is IMP::base::PROGRESS. ## Winter 2012-2013 - Scons support was removed. Use cmake to build IMP, it is way better. - Examples are run with a `--test` argument when IMP is built with internal checks. Expensive examples should check this flag and shorten what they do in order to try to test all their code without taking too long. - The class IMP::core::NonRigidMember was added to allow particles that are not rigid but whose coordinates are specified via internal coordinates in some IMP::algebra::Reference3D that is part of an IMP::core::RigidBody. - The IMP_BIG_ENDIAN and IMP_LITTLE_ENDIAN preprocessor symbols are deprecated in favor of the BOOST_BIG_ENDIAN and BOOST_LITTLE_ENDIAN symbols. - Support for asymmetric moves was added to IMP::core::MonteCarlo. To do this, IMP::core::MonteCarloMover was added and IMP::core::Mover and IMP::core::MoverBase were deprecated and implemented in terms of the new class. This should require few or no changes in code that uses Monte Carlo support or implements movers (although movers should move to IMP::core::MonteCarlo). All IMP::core::MonteCarloMover objects keep track of statistics (number of moves and number of accepted moves), rather than that being done in the IMP::core::SerialMover. In addition, the poorly defined move probability was removed from IMP::core::MonteCarlo. - IMP::base::WarningContext now outputs warnings immediately. This change was needed as the previous behavior resulted in occasional core dumps when the libraries for log support were unloaded before the libraries where static IMP::base::WarningContexts were declared. And there was no easy way to detect such problems. - You can now (probably) build IMP with [CMake on IMP wiki](https://github.com/salilab/imp/wiki/Cmake). - The maximum logging level and check level can now be controlled independently of the build mode. To, for example, enable progress level logging in a fast build, add maxlog="PROGRESS" to your config. The maximum check level for `release` has been reduced to IMP::USAGE as that speeds things up a bit. - The way applications are structured and specified has been regularized and simplified. Python files are automatically copied/installed and binary executables are compiled from each found `.cpp` file linked with any `.cpp` files contained in a subdirectory called `lib`. Nothing needs to be put into the SConscript file. - The Python constants like IMP.algebra.has_cgal went away as swig autogenerates ones like IMP.algebra.IMP_ALGEBRA_HAS_IMP_CGAL. While the latter are a bit uglier, the constants aren't used much and having them be auto generated is nice. - There have been some major changes to the build system. The main one is that the SConscripts in the directories below the module directory aren't really used. They are being removed from the repository. Now everything really is off of file locations (eg all test_* in test is a test, each *.cpp in bin becomes a program etc). This greatly simplifies the build system, which has, historically been a disaster. You can now define dependencies by dropping files into a dependency subdirectory in your module, no need to mess with the SConscript. - Building IMP now always requires specifying a `repository` argument. This was always required for out of source builds, which is what one should be doing, but is now required for in source builds to, so as to make accidentally polluting your source with an in-source build harder. Use `repository="."` in that case. - The HDF5 functionality in RMF was reorganized. Specifically, the headers were moved into `RMF/HDF5`, the Python functionality into `RMF_HDF5` and the code into namespace `RMF::HDF5` and all mentions of `HDF5` or `hdf5` in the class and function names were removed. - Support for IMP-wide command line flags have been added to IMP::base in the base/flags.h file. They provide a simple command line parsing interface that can be used from C++ and Python. In addition, various functionality in IMP is automatically exposed on the command line when those functions are used. For example, one can control logging, checks or profiling. See IMP/base/flags.h for (some) more info. - OpenMP support in IMP is officially released. All tests pass. But there has been little specific testing so far. - Access to attributes using IMP::ParticleIndex and the IMP::Model now works from Python. See IMP::Model::add_attribute() and kin. - IMP::ParticleKey and IMP::ParticlesKey were renamed to IMP::ParticleIndexKey and IMP::ParticleIndexesKey respectively as that is what they store. Typedefs were provided to support old code. - Indexes have been added to the docs connecting classes with all the methods that return them, examples that use them and methods that take them. And all methods that are used in examples with the examples they are used in. Please check them out and report issues and ways to improve things. You can find links on the main page. - The docs for a modules/biological systems/applications should now be found in an `overview.dox` in the module `doc` folder. Modules in IMP svn have been updated. For external modules, you can just move a `overview.dox` from that was previously generated into your `doc` folder and (eventually) remove the lines from your `doc/SConscript`. This makes it easier to use the full range of doxygen markup, as well as removes the need to escape doxygen markup for Python. - Examples have now been simplified into full blown doxygen examples. This involves merging the `.py` and `.readme` files (putting `\#\#` in front of each line of the `.readme`) and adding a `## \\example module/examplename.py` line at the beginning. You can use `./tools/updaters/update_examples.py modules/mymodule` to update a module that was not already updated. - IMP::base::FailureHandler has been removed as it wasn't used. - Handling of data and includes has been simplified. All appropriate files in the corresponding directories are linked into build/installed. This is only likely to be a problem if you had broken headers sitting around `include`. If so, just comment them out, or move them some place nicer. - Modules now must be in a folder called `modules`, which must have a `description.py` file. - The build system should now be a little better about cleaning up after itself. - How headers/doc files/data/examples - The IMP_PROTECTED macros didn't work out well, and are now deprecated. Sorry. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading "_" from the method name. - Nice version numbers will now be generated if you use git. You need to rerun tools/git/developer_tools/init_git if you have an existing clone. ## Fall 2012 - 11/9: RMF is now an external library shipped with IMP. IMP will build it and install it automatically if you have cmake version 2.8 or higher installed. - 11/9: Updated support macros for IMP::Restraint and IMP::ScoreState classes have been added (IMP_RESTRAINT_2() and IMP_SCORE_STATE_2()) that make it easier to return dependencies on things other than IMP::Particle and IMP::ScoreState objects. In addition, the dependency functions on scores and modifiers have been updated similarly and standardized. In general this should not cause backward compatibility problems unless you were overriding the IMP::ModelObject::do_get_inputs() in IMP::Restraint and using the IMP_RESTRAINT() macro. In that case update to the IMP_RESTRAINT_2() macro. - 11/8: Support has been added for the C++11 keywords override (IMP_OVERRIDE) and final (IMP_FINAL). They allow you to tell the compiler to produce an error if the marked method does not override/implement a parent's method and if anything overrides a given method, respectively. - 11/8: IMP::SingletonModifier::apply_indexes() and IMP::SingletonScore::evaluate_indexes() now require an upper and lower bound index (you can just pass 0 and the size of the array). We can make them option if there is a good reason. - 11/8: IMP::SingletonDerivativeModifier and kin were merged into IMP::SingletonModifier and kin as the only difference, the presence of the IMP::DerivativeAccumulator (whose weight was always 1) wasn't very interested and the split added lots of complexity. If this has an impact on your code, let us know. - 11/8; Evaluation with containers was greatly simplified, moving many methods on containers. We can add backwards compatibility versions if that is interesting. - 11/7: Predicates, scores and modifiers written in Python need to update their methods for returning inputs and outputs. They should now have a method \c _do_get_outputs(self, m, particle_indexes) which should probably just return \c [m.get_particle(i) for in in particle_indexes] to return all the passed particles. - 11/7: It is now preferred that IMP::Restraint and IMP::ScoreState descendants report dependencies directly using IMP::ModelObject::do_get_inputs() and IMP::ModelObject::do_get_outputs(). To do this, you should use the IMP_RESTRAINT_2() and IMP_CONSTRAINT_2() macros instead. Methods like IMP::Restraint::get_input_particles() are now deprecated. This change was made as the old mechanism introduced a lot of extra complication and made it so that there was no way to directly depend on something that was not a Particle or a Container. - 11/3: IMP::container::InContainerPairFilter and friends now treat pairs as unordered. A flag was added to control this. - 9/5: IMP::base::Object classes now have a method IMP::base::Object::clear_caches() that clears all cache data in the object. This is not universally implemented, but now provides a correct path for doing that. The IMP_LIST() type macros call that method when their contents changes. ## Summer 2012 - 7/3: IMP_PERIODIC_OPTIMIZER_STATE() was superceded by the class IMP::core::PeriodicOptimizerState as it is rather cleaner than the macro. - 7/2: in order to make various base classes in IMP more Python friendly, macros IMP_PROTECTED_METHOD(), IMP_PROTECTED_CONSTRUCTOR() were added that properly expose such methods to Python. The standards checks now complain if \c protected is used an a class in the API. - 6/24: Executables from \c module/bin are now put in \b build/module_bin/module. This removes a recurring source of errors in the build script due to the origin and destination directory have the same name in scons. - 6/24: Benchmarks are now put int \c module/benchmark and build to \c build/benchmark/module to remove a recurring source of errors. - 6/18: The IMP::benchmark module supports the GPerfTools profilers when available. Benchmark executables (things in \c bin whose name starts with \c benchmark_) that use the IMP_BENCHMARK() macro have command line flags to turn on and cpu and memory profiling. - 6/12: The naming rules for .i files that are in module \c pyext directories changed. They should now be \c IMP_modulename.name.i instead of \c IMP_modulename_name.i. This is to better support underscores in names. ## Winter 2011/2012 - 20/4: A method, IMP::atom::read_pdb(IMP::base::TextInput, int, IMP::atom::Hierarchy), was added to load coordinates from a pdb file into an existing IMP::atom::Hierarchy. - 4/4: the filter types were superceded by the predicate types. You will only see a difference if you develop a filter class (eg IMP::PairFilter -derived class) or directly call methods on filters (specifically IMP::PairFilter::get_contains()). - 29/3: IMP::Particle::get_has_model() has been changed to IMP::ModelObject::get_is_part_of_model() to be consistent with other classes. - 28/3: IMP::core::Mover changed slightly, adding an extra method IMP::core::Mover::get_output_particles() and requiring the IMP::Model be passed to the constructor. All movers in IMP SVN were updated without any difficulty. This allows movers to use information contained in the dependency graph amongst particles so they can, for example, weight moves based on scores. - 26/3: IMP::SingletonContainerInput etc were added. Methods that take these as their arguments can be passed either containers or lists of particles (or tuples), as convenient. All methods that take IMP::SingletonContainer objects should eventually move over to taking this (changing just requires changing the argument from a pointer to an Input). - 22/3: Over the last week or so, which IMP headers include which other heads have been substantially simplified. This may result in compilation errors for code outside of \c svn that depended on IMP headers including other headers. In general, you want to explicitly include the header defining each thing you use. - 15/3: The vector used throughout IMP is now exposed at IMP::base::Vector and what it does it documented. - 10/3: IMP::domino::RestraintCache was introduced to provide a documented centralized place to handle restraint evaluation and caching in DOMINO. In the switch various bugs involving maximum scores and IMP::RestraintSet objects were fixed. - 10/3: Scoring functions in IMP are now implemented in IMP::ScoringFunction objects. All restraint evaluation is handled by them and they can cache various useful information to aid in accelerating restraint evaluation. Backward compatibility interfaces have mostly been provided, please complain if some important interface went away. This has greatly simplified the code as well as fixed various bugs in various types of restraint evaluation (particularly in the handling of maximum scores on IMP::Restraint and IMP::RestraintSet objects. - 29/2: The macros IMP_NAMED_TUPLE_1(),IMP_NAMED_TUPLE_2(),IMP_NAMED_TUPLE_3() were added for easy generation of simply struct for use in passing and returning values. The generated structs support comparison, output, hashing and have constructors. Use the macros instead of a boost::tuple or a a POD struct. - 16/2: Restraints now are evaluated with the correct weights when evaluated by themselves as opposed to always returning their weight as seen by the model as a whole. - 15/2: The IMP::algebra grid support was rearranged and simplified. Most of the changes should be transparent. However, the IMP::algebra::grids namespace got folded into IMP::algebra, which might break some code. And various (obscure) constructors of IMP::algebra::GridD were removed. All code in svn has been changed as needed. - 3/2: The IMP::domino HDF5 container has been revamped. This involves splitting it into a write version (IMP::domino::WriteHDF5AssignmentContainer) and a read version (IMP::domino::ReadHDF5AssignmentContainer). - 3/2: IMP::domino::WriteAssignmentContainer and IMP::domino::ReadAssignmentContainer were added to do direct binary file writing. This is faster than the HDF5 versions. - 3/2: The installation instructions have been revamped and the developer instruction, in particular, have been greatly simplified. - 24/1: \c tools/show-changes has been added that shows the changes in the history log since you last updated. - 24/1: \c tools/make-source has been added that adds a source/header pair to a module - 24/1: principal components analysis in IMP::algebra was generalized to work in any dimension. This required renaming the class to IMP::algebra::PrincipalComponentAnalysisD. - 24/1: the output of the IMP::algebra::PrincipalComponentAnalysisD::show() was changed to reflect the standards (values produce compact, single line output on show). The old show has been moved to IMP::multifit - 19/1: IMP::container::ConsecutivePairContainer had IMP::container::ExclusiveConsecutivePairContainer split off of it to make the two operating mode clearer and so that there could be a clear IMP::container::ExclusiveConsecutivePairFilter to use. - IMP::algebra::Grid::apply was added to provide faster access to all (or, later, part of a grid). It can be an order of magnitude or more faster than using the iterators when doing simple operations. - the scons config options \c pretty and \c color were added (defaulting to True). When they are true, build commands are mostly suppressed and a much briefer description of what is being done is printed out (in color). Note that \c color mode and compilation in emacs don't get along too well. - IMP::algebra::ParabolicFit was renamed to IMP::algebra::ParabolicFit2D (a backwards compatibility typedef is currently provided) - IMP::algebra::LinearFit was renamed to IMP::algebra::LinearFit2D. It now supports error bars on the y-values. ## Fall 2011 - IMP::compatibility::checked_vector was renamed to IMP::compatibility::vector and made available to IMP as IMP::vector. Typedefs should use this type (IMP::vector). - Various IMP::algebra plural typedefs which had been using std::vector (since they never were updated), have now been moved to IMP::vector. - IMP::statistics::HistogramD now supports display with \c matplotlib. - IMP::statistics::HistogramD supercedes IMP::statistics::Histogram - predicate base classes have been added: IMP::SingletonPredicate, IMP::PairPredicate etc. These take particles and return an integer. Current functionality using them includes IMP::core::PredicateSingletonScore that uses a predicate to choose which score to apply. - the core generic restraint and score state support was moved to the kernel to avoid breaking various scores outside of IMP svn. - the IMP::display module was refactored and moved up to be a dependency of IMP::core and IMP::atom rather than vice versa. The main observable results are that you should now use IMP::atom::HierarchyGeometry, IMP::atom::SelectionGeometry, IMP::atom::HierarchiesGeometry, IMP::core::XYZRGeometry, IMP::atom::BondGeometry etc instead of finding them in IMP::display. This change means that, display can be used internally when testing and debugging functionality in IMP::core and IMP::atom and that tests that use IMP::display don't need to build all of IMP::atom. - Decomposition of restraints now results in restraints that are faster to evaluate, using the generic restraint mechanism in IMP kernel. - IMP::algebra::get_unit_bounding_box_d() was changed to return a bounding box with radius 1 to be consistent with IMP::algebra::get_unit_sphere_d() - A general purpose scheme to associate data with tuples of hierarchy nodes was added to the RMF format. This deprecates the special purpose bond support. - the way various types of geometry is stored in RMF files was changed, breaking backwards compatibility. ## Summer 2011 - some functionality from the kernel has been spun off into a new module IMP.base. You probably won't notice anything. The only significant changes are that functions that accept or return IMP::base::TextInput and IMP::base::TextOutput must declare them with the \c base qualifier (eg base::TextInput) so that swig is happy. - the source for the kernel has been moved from \c kernel to \c modules/kernel. This is unlikely to effect anyone. - IMP::core::ExcludedVolumeRestraint now supports filters - IMP::domino::HDF5AssignmentTable now caches internally, which makes it several times faster than before. - you can now pick which Python to use with the \c python argument to scons. - the units for BD diffusion coefficients changed to be consistent with other IMP units. - IMP::domino::get_interaction_graph() now takes an IMP::RestraintsTemp, like other functions - The assert failure handling function was documented and renamed to IMP::handle_error() so it is easier to find when you want to use gdb. - to work around various issues with ref counting pointers and compiler eccentricities, we added emulation of the C++0x \c nullptr. You should prefer that to \c NULL in IMP code that you want to run cross platform. - IMP::IntsList, IMP::FloatsList and IMP::StringsList have been added to replace the various scattered solutions for passing such types. - IMP::Restraint::get_decomposition() and IMP::Restraint::get_current_decomposition() changed to return a single Restraint, instead of a list. This makes it easier to propagate the weight and maximum reliably. - IMP now supports usage of restraints without adding them to the IMP::Model scoring function. To do this, do IMP::Restraint::set_model(), passing the model. The restraint can then be evaluated. - IMP::domino::DominoSampler does not use the IMP::domino::OptimizeRestraints restraint optimization by default. Instead, if it is asked to create the IMP::domino::RestraintScoreSubsetFilterTable, it decomposes the restraints first (via IMP::create_decomposition()). This avoids several bugs and very fragile code. - the IMP::core::DiameterRestraint was ignoring the radii of the particles involved and so was giving artificially low scores. It now is on the actual diameter of this molecule. - the IMP::core::MonteCarlo optimizers now support incremental evaluation. This can be a lot faster when only a few particles are moved each step. - all the plural types in IMP have moved to bounds checked vectors when using debug builds with gcc. If you get a non-imp error about an invalid access, it is probably some code that walks off the end of an array. Please report it. - The IMP::core::MoverBase was changed to move it off of the IMP::SingletonContainer. This resulted in the IMP::core::BallMover and IMP::core::NormalMover having new constructors. Now create them from a list of particles, a list of float keys and a radius/standard deviation. This removes some complexity and inefficiencies as well is removes some scope for errors. - Temp and non-temp decorator lists have been merged in C++. This was done to simplify code since no one was using the non-temp variants. Now all decorator lists are temp (non-reference counted) variants. - restraints and restraint sets are now evaluated only using "canonical" weights. That is, each restraint has one weight for the model (the total over all the restraint sets it is contained in and will always be evaluated with that weight). This allows the evaluation of simple restraints to be faster. - The IMP_LIST() macros have been changed slightly. If your usage needs to take action upon addition, removal or changes to the list, use the IMP_LIST_ACTION() variant in your header. The IMP_LIST_IMPL() macro no longer takes the actions, and the last three (generally empty arguments) will have to be removed from the call to get it to work. This is needed to move towards more flexible and efficient passing of lists of values. - IMP::domino uses the new IMP::Model::evaluate_if_good() support to accelerate computations - IMP::core::MonteCarlo can use the IMP::Model::evaluate_if_good() support to more quickly reject very bad conformations. See IMP::core::MonteCarlo::set_maximum_difference(). - IMP::core::ExcludedVolumeRestraint has been rewritten to be faster. It is especially faster when used with IMP::Model::evaluate_if_good() - a new evaluation mode IMP::Model::evaluate_if_good() and IMP::Model::evaluate_if_below() has been added. It allows restraints to shortcut score computation when the score can be shown to be worse than a maximum value. - as part of moving towards more efficient use of evaluate_if_good() various deprecated methods have been removed from IMP::core::ExcludedVolumeRestraint and IMP::core::RigidClosePairsFinder. - incremental evaluation support was removed from IMP. The reason for doing this was that the code was complicated, created overhead even when incremental support was not used and was not commonly used. Further, decomposing the restraints and using the dependency graph yields similar results and is more transparent. - the RMF helper binaries have been moved to their own application so that they get installed. - IMP::display::Writer has been revised to better support writers that write multiple frames to a single file and ones that write binary data. The noticeable changes are that the existing IMP::display::Writer instances must be created with a file name. And the IMP::display::WriteOptimizerState just takes a writer rather than a write and a file name template. IMP::display::Writer::set_frame() has been added so that you can directly control the frame being written. ## Spring 2011 - An interactive mode has been added to IMP::domino::DominoSampler. The interactive mode is useful for understanding domino sampling runs and distributing domino computations across a cluster. - Functions IMP::domino::set_assignments() and IMP::domino::get_assignments() were added which support writing subset assignments to data fields in hdf5 files. These can be used to helping to distribute domino on a cluster as well as debugging the sampling process. - a Python function IMP.show_graphviz() has been added that attempts to render a graph with \c dot and then open the pdf. - The Python functions IMP.show_altgraph() and IMP.get_networkx_graph() have been added to provide alternatives for quick viewing of IMP graphs. Unfortunately, they still pale compared to graphviz. - IMP.atom.create_cover() was added to facilitate creating a summary particle for part of a hierarchy in order to accelerate restraints. - IMP.hdf5 was renamed to IMP.rmf, the proposed name of the file format that it supports. That also better separates the hdf5 support from the file format support. - The old I/O functions have been marked as deprecated in favor of IMP.hdf5. You may need to compile with deprecated="yes" if you use them. - The Python graph interface now has a remove_vertex method. - One can now access the model restraint evaluation statistics programmatically using IMP::Model::get_restaint_statistics(). Restraint statistics can be logged to a file during optimization using IMP::core::WriteRestraintScoreStatisticsOptimizerState. - IMP::domino::DominoSampler now allows you to specify the tree on which merges occur directly. Also, you now need to specify the merge tree rather than the JunctionTree if you want to do non-loopy inference and specify something about how it is done. Sorry, it makes the code a lot easier. - Preliminary symmetry support has been added via the IMP::core::TransformationSymmetry and IMP::core::TransformationAndReflectionSymmetry classes. ## Winter 2011 - There now is a common base class IMP::atom::Simulation for IMP::atom::BrownianDynamics and IMP::atom::MolecularDynamics. In merging them there were several minor changes made to both. The most noticeable may be that the function for setting the maximum time step length is now IMP::atom::Simulation::set_maximum_time_step(). The last time step length can be extracted by IMP::atom::Simulation::get_last_time_step() - IMP::core::MonteCarlo::set_temperature() was renamed to IMP::core::MonteCarlo::set_kt() as that is what it is doing. - A new module IMP.hdf5 has been added to provide more general support for saving and loading IMP::atom::Hierarchy structures to files. Multiple conformations of each hierarchy can be saved as well as information about geometry and restraints (although these can't be read at the moment, so they are not so useful). - IMP::algebra::grids::GridD now supports dense grids in arbitrary dimension through use of IMP::algebra::grids::DenseGridStorageD. - IMP::Optimizer now can use a list of restraint sets to compute the score rather than the complete Model score. - each IMP::Restraint now has its own weight. This is lightly tested. - A new log level IMP::PROGRESS has been added. The idea is that IMP::Sampler::get_sample() calls and other expensive loops can show a boost progress meter at this log level to give some idea when things are moving forward. - IMP::domino::TrivialParticleStates renamed to IMP::domino::IndexParticleStates to be less judgmental - IMP::domino::CompoundParticleStates and IMP::domino::RecursiveParticleStates added - IMP::domino::DominoSampler now supports maximum scores on IMP::RestraintSets (untested). - the name of the clustering methods have been changed to create_ from get_ since they create objects. - the names of the recently added metric based clustering support has been changed to Metric rather than Distance. - IMP now has the ability to have certain tests designated as expensive and to skip those tests when test-fast is run. The motivation for this is that the tests take a really long time making it impractical to run before every commit. And it is hard to tell which tests have not been run when running with the necessary multiple build processes. My thought is that tests that take about a second or less on a debug build should be left along (this is probably >95% of the tests), but the few that take longer should be marked as expensive. To do that, rename the test to expensive_text_XXX.py (from test_XXX.py) and add expensive_python_tests= env.IMPModuleGetExpensivePythonTests() to the IMPModuleTest call in the test SConscript. - the fuction IMP::display::create_restraint_geometry() was added to do what its name says. - runtime dimensional geometric objects have been added. These include IMP::algebra::VectorKD, IMP::algebra::BoundingBoxKD and IMP::algebra::SphereKD. The IMP::statistics::Embedding classes now use IMP::algebra::VectorKD instead of Floats. The sparse grids also support variable dimensions as to the nearest neighbor searches. - IMP::atom::DopePairScore has been added to IMP. - IMP::atom::CHARMMStereochemistryRestraint provides a high-level simplified interface to the bonded part of the CHARMM forcefield. - IMP modules are now built by first combining all the .cpp files into one, and then building that file. This greatly accelerates building of a module from scratch as build time is dominated by parsing headers. This does change the semantics of .cpp files as they see headers included by and functions declared by an arbitrary subset of other .cpp files. For now they are still required to be able to be compiled separately. To get the only behavior back either globally or on a per-module basis, see the percppcompilation parameter to scons. - Instructions and scripts have been added for building modules and things externally to the IMP build. See the installation guide. - IMP::atom::CHARMMParameters can now automatically map CHARMM-style atom and residue naming to PDB-style, so CHARMM topology files can be used unmodified. - IMP::atom::CHARMMTopology::add_sequence() provides a simple way to generate a topology from a string of amino acid one-letter codes. - IMP::atom::CHARMMTopology::add_coordinates() will generate atomic Cartesian coordinates from CHARMM internal coordinate information. - IMP::atom::CHARMMTopology::add_missing_atoms() and IMP::atom::remove_charmm_untyped_atoms() can now be used to ensure that atoms read from a PDB file match the CHARMM topology. - building of deprecated functionality is now false by default - the IMP_COMPARISONS(), IMP_COMPARISONS_1(), IMP_COMPARISONS_2() macros now take the name of the class as an additional (first) argument. This avoids the formerly needed, \c This typedef and makes them more consistent with the other macros in IMP. - modules can now be built externally to the IMP svn in addition to applications and biological systems. To do so, create a directory with your module in a subdir containing links to the IMP SConscript file and scons_tools and a config.py that has information needed to run IMP (eg an appropriate \c pythonpath, \c ldpath). - removed IMP::algebra::GridD::get_index(VectorD) as it wasn't very safe (in that vectors coordinates can get rounded between the IMP::algebra::GridD::get_has_index() and IMP::algebra::GridD::get_index() calls. Use IMP::algebra::GridD::get_nearest_index() or IMP::algebra::GridD::get_extended_index() or IMP::algebra::GridD::operator[]() instead. - renamed IMP::domino::ParticleStatesTable::get_particles() to IMP::domino::ParticleStatesTable::get_subset() as it is more useful to be able to get the subset containing all the particles. - rename IMP::restrainer::Restraint::print_all_restraints() to IMP::restrainer::Restraint::show_all_restraints() to be consistent with the rest of IMP. - renamed IMP::atom::get_conect_record_string() to IMP::atom::get_pdb_conect_record_string() to make clearer what it is. - class and method names are now spell checked. Add missing words to spelling_exceptions in the module test call. - the plural names (eg IMP.Particles()) have been removed from the Python side as they are just Python lists. - biological systems can now include internal modules. These provide a convenient place to put application specific C++ and Python code. See the example system for more info. - IMP_VALUES() now includes IMP_OUTPUT_OPERATOR() so you no longer need two namespace macros for each value. - the function IMP::core::Harmonic::k_from_standard_deviation was renamed to IMP::core::Harmonic::get_k_from_standard_deviation() to conform to the standards. - the IMP::atom::PDBSelector and IMP::atom::Mol2Selector became IMP::Object classes. All this means is that they need to be created with \c new in C++ code. It also means that one is no longer prohibited from storing them (not that there is much use in doing that). - the standards tests have been improved to test more useful aspects. Now class and function names are checked some as well as all modules have the checks for show and plural types. The error messages should explain how to fix things. - IMP::core::RigidBody::lazy_set_reference_frame() was renamed to IMP::core::RigidBody::set_reference_frame_lazy() to fit with the standards better. - the transform based methods on reference frame which have been deprecated for a while were removed - the hierarchy visiting and searching functions have been renamed to better fit the IMP naming scheme as we don't expect they were used outside of the IMP source. - IMP::atom::NonhydrogenMol2Selector was renamed IMP::atom::NonHydrogenMol2Selector to be consistent with the PDB version ## Fall 2010 - IMP::display::PlaneGeometry was added to display planes - the optimize methods on optimizers was moved to a "non-virtual interface". Usage has not changed, but optimizers should implement IMP::Optimizer::do_optimize() instead of IMP::Optimizer::optimize() to benefit from better checking. - the two-vector constructor for IMP::algebra::Plane3D now expects the second vector to be a unit vector as otherwise one could switch the order around without noticing. - the io functions in the kernel have been cleaned up a bit (the old ones are there for backwards compatibility). See IMP::write_particles(), IMP::write_particles_binary(), IMP::write_particles_to_buffer() and IMP::read_particles(), IMP::read_particles_binary(), IMP::read_particles_from_buffer() - IMP::core::WriteModelOptimizerState is going away as it duplicated IMP::WriteParticlesOptimizerState. - the IMP::algebra::Grid3D has been changed to make it support any (fixed) dimension when used with sparse storage. The template arguments have also been changed to remove a redundant one. See IMP::algebra::GridD. - the biological systems specification has been improved a little adding ability to run some rudimentary tests to make sure the scripts are ok and to not require that everything be run from the svn directory. - A new method IMP::atom::create_compatible_rigid_body() which creates a rigid body from a hierarchy which has the same internal coordinates as another rigid body. A way to make the rigid body construction more canonical would be nice (so we don't need this function), but I'm not sure how to do that. - Logging has been improved to better report the name of methods producing the log messages - a macro IMP_FUNCTION_LOG has been added which should be added to the start of non-member functions which produce significant logging output. - IMP::algebra::get_distance(const Rotation3D&,const Rotation3D&) was changed to use the Euclidean distance in R4 as that seems better behaved than the prior metric. - IMP.domino2 was renamed to IMP.domino and the old IMP.domino was moved - the documentation for classes and some methods now includes a list of all examples which use the class. Please report any incorrect links or phantom classes appearing in the docs. Unfortunately, functions have to be added individually. Please request any functions of particular interest. - applications and biological systems now have their own documentation pages - IMP::atom::Hierarchy requires radii on its leaves - IMP::atom::read_pdb add radii to the atoms - IMP::helper was merged into IMP::restrainer as all the functionality there was primarily to provide support for restrainer - IMP::domino2::SubsetEvaluatorTable was removed. Pass the model and IMP::domino2::ParticleStatesTable to the IMP::domino2::RestraintScoreSubsetFilterTable instead. - IMP::atom::Selection was added to allow specification of parts of IMP::atom::Hierarchy using biological sorts of names. A few functions to create restraints from them were added too. - a new example was added in the kernel to show IMP used on a biological system - IMP::display::PymolWriter now supports adding labels to things - IMP::display::PairRestraintGeometry and IMP::display::ConnectivityRestraintGeometry were added to aid in visualizing those restraints. - IMP::bullet was added with an optimizer to use hard body physics to resolve local collisions. - IMP::atom::setup_as_rigid_body() has been deprecated and replaced by IMP::atom::create_rigid_body(). The latter creates the rigid body as a separate particle so that the molecular hierarchy doesn't have to worry about having its bounding sphere large enough to enclose all the members. - IMP::core::KClosePairsPairScore was added to have count based cutoffs for close pair scoring. - IMP::algebra::get_surface_area_and_volume() was added to compute the two numbers for a set of balls ## Summer 2010 - The IMP::core::RigidDistancePairScore and IMP::core::ClosePairsPairScore have been revised to remove certain inconsistencies and ambiguities and to make them faster. Existing code using them will need to be slightly tweaked to work. - Javi added his 2D EM scoring functionality as IMP::em2D - some new ways to speed up scons have been added. See IMP::core::HarmonicDistancePairScore, IMP::core::SoftSpherePairScore and, for C++ users, the generic restraint support IMP::container::create_restraint() and IMP::core::create_restraint(). These each offer between 2x and 1.2x speedup when using simple scores (like harmonics on distances). In addition, a function IMP::container::get_slack_estimate() has been added which will estimate the correct slack to use with a IMP::container::ClosePairContainer. This will be extended to support IMP::core::ExcludedVolumeRestraint eventually. - support for dense and sparse grids have been added to IMP::algebra. See the IMP::algebra module page (I'm still working on how to make the docs clearer). - all write functions now have associated IMP::OptimizerState and IMP::FailureHandler classes. For examples, see IMP::WriteOptimizerState, IMP::WriteBinaryOptimizerState, IMP::WriteFailureHandler, IMP::atom::WritePDBOptimizerState etc. - functions IMP::write_binary_model(), IMP::read_binary_model() and IMP::read_configurations() have been added to support space and filesystem efficient storage of model configurations. Multiple configuration can be written to a single file. This functionality requires NetCDF be installed (NetCDF is a standard library for reading and writing binary data). - domino2 has been added. It provides a simpler interface on Keren's domino optimize as well as a number of new optimization and hooks for customizing the optimization process - a better explanation of how to use reference counted pointers has been added to IMP::Pointer. - all simple, planar polygons are now supported for display when CGAL is present ## Week of April 1, 2010 - All plural objects in Python are now simply Python lists rather than C++ recreations of Python lists. This makes compilation of the wrappers much, much, much faster. The visible changes are - types are not checked until the lists are passed to C++ - passing std::vector types by pointer and storing passed std::vector types by pointer will not work (but should never have been done anyway from code exposed to Python) - swig has issues with some overloads (not always for obvious reasons). These are being sorted out. None are commonly used. - get_particles() on a plural object doesn't work in Python (and is not needed in C++). Implicit conversions will always work. - All containers need to know the model at time of construction. If, after updating, you get an error on a container constructor, just add the model as the first argument. This also means that empty lists of particles in the constructors cause check failures. We can figure out a work around if this is too annoying. ## Week of March 13, 2010 - function get_rotation_in_radians_about_axis() was renamed to get_rotation_about_axis() as it didn't parse and we don't have the "in radians" elsewhere. - Added IMP::algebra::ReferenceFrame3D. As part of the cleanup, IMP::core::RigidBody::set_reference_frame() replaces set_transformation(). And the reference-frame related transformation construction functions were replaced. IMP::algebra::get_rotation_from_x_y_axes() was added to aid in the construction of reference frames (and replaces the algebra function taking three vectors). ## Week of March 6, 2010 - Publications now have a description field to give browsers an indication of what is in the publication to aid selection of what to read/cite. ## Week of Feb 27, 2010 - IMP::algebra::decompose_rotation_into_axis_angle() was renamed to IMP::algebra::get_angle_and_axis(). - renamed IMP::atom::bond() and IMP::atom::unbond() to IMP::atom::create_bond() and IMP::atom::destroy_bond() to be more consistent with things - renamed IMP::core::root() to IMP::core::get_root() ## Week of Feb 21, 2010 - IMP::Object::set_was_owned() was changed to IMP::Object::set_was_used() as that better reflects what the flag means. - various more interpolation and endian methods had their names fixed (by adding \c get_) - the function get_mean() on vectors was removed as it duplicated IMP::algebra::get_centroid() - build_Transformation3D_from_transformation3D was replaced by IMP.algebra.get_transformation_3d() - IMP::algebra::almost_equal() is now IMP::algebra::get_are_almost_equal() to fit the naming conventions. - the random vector generators have been cleaned up to remove redundant ones and make the names more consistent. They now have the form of either - IMP::algebra::get_random_vector_in() or IMP::algebra::get_random_vector_on() where there is one argument describing the volume or shape (such as an IMP::algebra::SphereD<4>). - or IMP::algebra::get_uniform_surface_cover() where the first argument is the primitive whose surface we are covering and the second is the number of points. Some good sed strings are - "s\#random_vector_in_unit_sphere()\#get_random_vector_in(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_in_unit_box()\#get_random_vector_in(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_on_unit_sphere()\#get_random_vector_on(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_on_unit_box()\#get_random_vector_on(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_in_box(\#get_random_vector_in(\#g" - "s\#random_vector_on_box(\#get_random_vector_on(\#g" - "s\#random_vector_in_sphere(\#get_random_vector_in(\#g" - "s\#random_vector_on_sphere(\#get_random_vector_on(\#g" - "s\#grid_cover\#get_grid_surface_cover\#g" - "s\#uniform_coverr\#get_uniform_surface_cover\#g" . - Spheres in IMP::algebra are now supported in general dimensions. See IMP::algebra::SphereD. - Many methods in IMP::algebra were renamed to be more consistent with other methods or general conventions. Most of these simply involved adding \c get_ as a prefix and/or add a \c _3d suffix when they did not take any arguments. - build_Rotation2D_from_Vector2D became IMP::algebra::get_rotation_to_x_axis() - most of the Rotation3D and Transformation3D functions (in addition, IMP::algebra::get_identity_rotation_3d() and IMP::algebra::get_identity_transformation_3d() added the \c _3d so we can add 2D versions at some point) - IMP::algebra::Transformation3D::get_transformed() and IMP::algebra::Rotation3D::get_rotated(). Replacing \c ".transform(" worked in the IMP code base. - IMP::algebra::get_distance(), this also carries over to the IMP::core::get_distance() The following global replacements are probably safe (in sed expressions) - "s\#core.distance(\#core.get_distance\#g" - "s\#algebra.distance(\#algebra.get_distance\#g" - "s\#\.transform(\#.get_transformed(\#g" Only in C++ - "s\#\.rotate(\#.get_rotated(\#g" Only in C++ - "s\#rotation_from\#get_rotation_from\#g" - "s\#rotation_in\#get_rotation_in\#g" - "s\#identity_rotation(\#get_identity_rotation_3d(" - "s\#::distance(\#::get_distance(\#g" . - IMP::algebra::Vector3D and IMP::algebra::Vector3Ds were replaced by the expanded (non-typedef) version in functions parsed by SWIG as the current situation was extremely brittle. For the moment, you should not use the typedefs in definitions that swig sees. Sorry. - IMP::algebra::Sphere3DPatch was renamed to IMP::algebra::SpherePatch3D to be consistent with the other geometry and allow us to support higher dimensions in the future. - rigid_align_first_to_second was renamed to IMP::algebra::get_transformation_aligning_first_to_second() and build_Transformation2D_from_point_sets was renamed to IMP::algebra::get_transformation_aligning_pair() ## Week of Feb 14, 2010 - The IMP::atom::Bond types were rearranged to reconcile the protein-centric classification with ligand support. - IMP::atom::Atom no longer has charge since we have IMP::atom::Charged. - IMP::atom::read_mol2() and IMP::atom::write_mol2() were added - IMP::atom::Selector and descendants renamed to IMP::atom::PDBSelector since we will soon have Mol2Selectors. In addition, the IMP::atom::IgnoreAlteriativesSelector is now IMP::atom::NonAlternativeSelector to keep the name positive, like the others. In the IMP body, all instances of the string Selector were for PDBSelectors and so sed was safe to use. - IMP::atom::read_pdb() lost the "ignore_alternatives" argument. Instead, all of the provided selectors, other than IMP::atom::AllSelector ignore alternatives internally. Code that uses the IMP::atom::AllSelector should probably change to the IMP::atom::NonAlternativeSelector. - IMP::example renamed (from examples) - The IMP_USAGE_CHECK() macro lost the (unused) third argument. ## Week of February 3, 2010 - The model io code as moved into the kernel so it is now IMP::write_model() and IMP::read_model() - Remove the version_info argument from the implementation macros. Instead, just pick up the module version info automatically. - Add a function Object::get_type_name() and Object::get_module() to get the respective bits of information about each object. - Change Object::show() to be a non-virtual wrapper which prints the object name and other info before passing control off to Object::do_show(). - The containers were moved to IMP.container. See that module page for a full list of what moved. In the process, the backwards compatibility functions were removed from IMP::container::PairsRestraint and kin. - IMP.atom.create_protein() now returns the created protein - the simplify functions have been moved from IMP.helper. They are now IMP.atom.create_simplified_along_backbone(). - IMP.helper.clone() has been renamed to IMP.atom.create_clone() - IMP.misc.ConnectingPairContainer has been moved to IMP.core.ConnectingPairContainer - The locations for pixels in IMP::em::DensityMap have been moved to the center of the corresponding voxel from the lower left corner. - IMP::em::get_transformed(), IMP::em::get_resampled() and IMP::em::get_density() have been added. - The documentation has been majorly cleaned up and simplified. Please look around and see if you like it. There are now two main pages, the introduction and the developers guide, rather than the previous plethora of pages. - IMP::algebra::Cone3D::get_bounding_sphere() has been removed since it was not of general use, not returning the bounding sphere you would expect and was trivial to implement. - IMP::algebra::projection() was removed since it didn't actually project onto the segment and the name didn't follow the normal conventions. And it wasn't used. - Most geometry now has namespace methods IMP::algebra::get_bounding_box(), IMP::algebra::get_surface_area(), IMP::algebra::get_volume(). ## Week of January 22, 2010 - A generic geometry layer has been added (see \ref genericgeometry). Using it, IMP::atom::rmsd(), IMP::algebra::alignment_taking_first_to_second() and the new IMP::atom::native_overlap() all can work on either collections of either IMP::core::XYZ particles or IMP::algebra::Vector3Ds. - Deprecated functionality has been removed from IMP.core. For those still using it, IMP::core::ClosePairsScoreState was replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairsScoreState was replaced by IMP::core::CloseBipartitePairContainer. - The macros for Scores changed slightly. Scores not in IMP svn which use the IMP_*_SCORE macros now have to implement the get_is_changed() function outside of the class declaration and scores which use the IMP_SIMPLE_*_SCORE now don't have to implement the functions to get interactions and used particles. - Rename IMP::core::MinimumPairScoreRestraint (and kin, Maximum, Singleton, Triplet and Quad) to IMP::core::MinimumPairRestraint. - Add IMP::core::MinumumPairScore (and kin). ## Week of January 1, 2010 - IMP::em is now using Particles to resample density maps and for fitting restraints. The previous ParticleAccessPoint mechanism is removed. - IMP::UsageException and IMP::InternalException are now reserved for their respective checks (and cannot be thrown by IMP_THROW()). This makes the distinction between API exceptions (which are thrown in any build mode and should be documented and tested) and helper exceptions which are only thrown when checks are turned on (and hence should not be documented) clearer. - IMP::core::RestraintSet moved to IMP::RestraintSet. The restraint sets now simply define a tree with weights which the model traverses as opposed to being an opaque IMP::Restraint themselves. This means that incremental evaluation of restraints in restraint sets now works and that restraints in restraint sets can be individually evaluated. Existing working usage should not have changed (so \c sed should manage updates just fine), but the internal implementation is completely new, so pay attention. - IMP now has a text i/o layer consisting of IMP::TextInput and IMP::TextOutput. These classes are convertible to C++ streams and implicitly constructable from both streams and file paths. As a result, functions that take them will work when called with a string for a file path, a Python file or a C++ stream. - The IMP.examples module now documents how to use the needed aspects of scons and SWIG. - The IMP.examples module is used as the basis for new modules created by the make-module script, so those include more documentation and are tested. - Module overview pages are now generated from the information passed in the SConscript file, rather than from the (now removed) modulename.dox file. See the examples file for an example of the things you can do there. - boost program options is now a checked for dependency. - The IMP::Model now uses dependency information from IMP::ScoreStates to automatically reorder the IMP::ScoreStates so that they are evaluated in the right order. Also, IMP::Model::evaluate() on a subset of the restraints now only updates the needed score states, not all of them. - Functions on rigid bodies no longer provide the IMP::core::RigidMemberRefiner as a default as it is, pretty much always, the wrong thing to use. That class may go away without notice in the future and is no longer documented. - the IMP::helper::create_simple_connectivity_on_rigid_bodies() and IMP::helper::create_simple_excluded_volume_on_rigid_bodies() now both take an IMP::Refiner as an (optional) final argument. This makes it so they can be used with the rigid bodies created from molecular hierarchies. This code still needs to be updated and the docs don't reflect what the code was or is doing, so use with caution (see notes in the header file). - The IMP::helper create* functions now all take non-Object inputs as const& to make clear what is input and that there are no memory management gotchas. - a new base class IMP::Constraint has been added for constraints. Various other classes have been renamed to be consistent. ## Week of December 11 - The interface of IMP::PairScore changed slightly to make the function signatures more uniform across various methods. Now all methods that take a pair of IMP::Particle objects take a const IMP::ParticlePair&. Previously some took that, some took an IMP::ParticlePair and some took two IMP::Particle*s. If you have an implementation of an IMP::PairScore outside of svn, you will have to update your function signatures (the compiler will tell you where). Backward compatible IMP::PairScore::evaluate() methods have been provided, so simple users of pair scores shouldn't see a difference. - IMP::core::PairRestraint takes a IMP::ParticlePair in the constructor instead of two IMP::Particle*s. A backwards compatible constructor can be added if needed. - IMP::ParticlePair::first and second went away since no one claimed to be using it and it complicated the implementation. - The generic restraints/score states/scores infrastructure has been extended to triplets and quads (for angles and torsion angles). ## Week of December 4 - IMP now makes use of Boost.Filesystem to manipulate paths. Things may not work well on systems where boost libs are not available which does not support posix paths. But libless boost installs generally don't exist outside of our test systems. - new functions like IMP::core::get_example_path(filename) and IMP::core::get_data_path(filename) allow one to easily find a data file named 'filename' for the examples and library internal data. - The IMP::display module has been revamped in order to support non-linear geometry. VRML support has been dropped. CMM support will be dropped unless someone requests it (as IMP::display::ChimeraWriter does all it does and more). ## Week of November 19 - imp.xcodeproj was moved into tools and updated accordingly - the bin directory finally went away - IMP::display::ChimeraWriter now supports the latest version of Chimera - IMP::display::CGOWriter was renamed to IMP::display::PymolWriter as it can do more than just CGO. - IMP::VectorOfRefCounted::operator[] can now be used for assignments just like a normal vector. No more need to use the IMP::VectorOfRefCounted::set() method. - the IMP::atom::Hierarchy constructor from an IMP::Particle* is now explicit (as were the decorator constructors) ## Week of November 12 - All exception types are now mapped to equivalent classes in Python. The class hierarchy is similar (e.g. IMP::ModelException derives from IMP::Exception in both C++ and Python) so all IMP exceptions can be caught in Python using "except IMP.Exception". For convenience, the IMP::IOException, IMP::ValueException and IMP::IndexException classes in Python also derive from the similarly-named standard Python classes (IOError, ValueError and IndexError respectively). ## week of October 29 Update - You can now add members to rigid bodies via IMP::core::RigidBody::add_member(). These members can be other rigid bodies (although this functionality is only lightly tested). - The code to get bounding volumes for IMP::atom::Hierarchy has been moved to the atom module, as as the clone and destroy code. They can be found at IMP::atom::clone(), IMP::atom::destroy(), IMP::atom::get_bounding_box(). - Added IMP::core::VolumeRestraint. - Several functions have been added to helper in creating restraints. They are create_simple_connectivity_on_rigid_bodies(), create_simple_connectivity_on_molecules(), create_simple_distance(), create_simple_diameter(), create_simple_em_fit(), and create_simple_excluded_volume_on_rigid_bodies(). Functions load_em_density_map() and set_rigid_bodies() are added as well. See IMP::helper for more information. - IMP::ScoreStates are further hidden. Namely, IMP::core::RigidBody constructs the needed score states internally, doing away with the need for the create functions (which have been done away with). In addition there are new decorators IMP::core::Cover and IMP::core::Centroid which manage their respective score states, further doing away with more create methods. The hierarchy-based create method for rigid bodies has been moved to the atom module and is called IMP::atom::rigid_body_setup_hierarchy(). - IMP::em::read_map() will figure out the file type to use automatically from the extension. Other such methods should follow suite. - A class IMP::core::StructureSet has been added to provide a convenient way to manage solutions of optimization runs. - IMP::core::RigidBodyMover now takes its angle argument in radians and is documented as such. - IMP::core::ClosePairsScoreState and IMP::core::CloseBipartitePairsScoreState have been replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairContainer. The score states have been marked as deprecated and are mostly still there, but are more limited than they were. - IMP::core::AllPairsPairContainer has been renamed to IMP::core::AllPairContainer and a bipartite version has been added. - All containers now keep track of what they looked like on the previous time step. - The IMP::core::ListPairContainer and IMP::core::ListSingletonContainer have had their interface reduced a bit in order to make implementing change tracking easier. - The set of exceptions has been simplified to go along with simplification of the checks. We now have IMP::InternalException and IMP::UsageException which are thrown by the respective checks (and the former is thrown by IMP_FAILURE()). When you want to to pass an exception back to Python use one of IMP::ValueException or IMP::IndexException or IMP::IOException. There is also a new macro, IMP_THROW() to be used when you want to throw an exception with a nice error message. Please look at code you have worked on and make sure that the exceptions make sense given the new clarifications. - An IMP::ModelException is now thrown when IMP::Particle attributes attain invalid values. This exception can be caught to recover from failures of IMP::core::ConjugateGradients or difficult restraints. ## Week of October 15 - The set of functions needed to get meta information about IMP::Restraint and IMP::ScoreState objects has hopefully stabilized. Sorry about the thrashing. If you have restraints or score states not in SVN, see the corresponding IMP_RESTRAINT type macros, the IMP::Interaction class and base classes for what exactly you need to implement. ## Week of October 8 - IMP::Interaction was modified to split IMP::interaction::get_used_particles() into IMP::interaction::get_read_particles() and IMP::interaction::get_write_particles(). This makes it possible to check for IMP::ScoreState objects being in an inconsistent order (so that the input to one is updated after it is read). Later, we will be able to automatically order the score states ensure that there are no problems. - IMP::atom::Hierarchy has been modifier to get rid of the IMP::atom::Hierarchy::Type. Instead, all types are contained in the decorators. IMP::atom::get_by_type() was also updated to reflect this. - IMP_check() and IMP_assert() have been replaced with IMP_USAGE_CHECK() and IMP_INTERNAL_CHECK() respectively. Hopefully this makes it clearer when to use which. IMP::set_check_level() has been revised accordingly (to use IMP::USAGE and IMP::USAGE_AND_INTERNAL). Also, IMP_failure() and IMP_not_implemented() have been renamed to IMP_FAILURE() and IMP_NOT_IMPLEMENTED() to make all macros be all caps. To update external code, - replace IMP.EXPENSIVE with IMP.USAGE_AND_INTERNAL - replace IMP.CHEAP with IMP.USAGE - replace IMP_assert and IMP_check as above. ## Week of October 1 - IMP::ScoreState and IMP::Restraint now better track the set of particles that they use and the dependencies between the particles. To this end, a new base class IMP::Interaction has been added with methods to get the inter-particle interactions and the set of particles used. - .cpp, .h, .py and SConscript files which contain the string "DO NOT COMMIT" will be rejected on checkin. Use this to make experimental or debugging changes you are making so you don't accidentally commit them. - The search module is no longer built by default and is not longer supported. It will get moved out of svn shortly. - IMP::ScoreState objects now also have IMP::Interacting::get_interacting_particles() methods. Soon IMP::PairScore and IMP::SingletonScore objects will too. ## Week of September 27 - Some support for gather statistics has been added to IMP::Model (see IMP::Model::set_gather_statistics()). If requested, an API can be provided so that the individual data points can be returned to programs. - A class IMP::core::RigidBodyDistancePairScore has been added which provides accelerated computations for the distance between two rigid bodies. - The SWIG input files were renamed from swig.i to swig.i-in to aid in dependency tracking. - rigid bodies now have two sets of particles associated with them in the various places they are used. One is the IMP::core::RigidBodyMember particles, as was previously the case. The other is defined through a refiner (and could be the RigidBodyMember particles if a IMP::core::RigidMembersRefiner is used). The idea is that you sometimes want some particles which are transformed along with the rigid body, but which are not really part of the representation of the rigid body. Examples would include marker points or multiresolution representations of proteins. ## Week of September 20 - all \c .i files have been renamed to start with \c IMP_ to help make dependency tracking easier. - static builds (builds with \c static=True) build both the static and dynamic libraries and just static executables - the \c profile build target has been removed as \c static=True does the important bit of what is needed. Just do \c linkflags=['-pg'] to add the needed link flag for \c gprof. - Restraints should now implement IMP::Restraint::unprotected_evaluate() instead of IMP::Restraint::evaluate(). The latter can now be called outside of IMP::Model::evaluate() and will return the right answer always. - \c bin/imppy.sh no longer works, as was previously warned. Use \c tools/imppy.sh instead. The bin directory will go away eventually, for now the script just tells you to use the tools version. ## Week of August 30 - A new score IMP::core::BoundingBox3DSingletonScore has been added to keep particles in a box. - str should now work better in Python when applied to IMP::Object objects and a few others. Objects should use the new IMP_SHOWABLE and IMP_SHOWABLE_INLINE macros instead of declaring show functions directory to make it so that they work in Python. It is now a bug to \c print or \c str to not return something useful in Python for any IMP object. ## Week of August 23 - each module now has a version, eg IMP::get_module_version_info() which is also returned by the various restraints of the module. The Python code checks that loaded libraries have the expected version numbers to try to provide better messages when runtime linking problems occur. - the various bounding_box methods have been renamed to ::get_bounding_box() to be more consistent with the other names. - Python documentation is now parsed by doxygen ## Week of August 15 - The Object::get_version_info() result now includes the SVN revision info when building from SVN. This means you can use it to more easily tell exactly what a built version of IMP was built from. ## Week of August 8 - Treatment of hetatoms in hierarchies (and hence pdbs) is now documented. See IMP::atom::Hierarchy. This involved broadening IMP::atom::Residue to handle ligands and having IMP::atom::Hierarchy::LIGAND, IMP::atom::Hierarchy::AMINOACID and IMP::atom::Hierarchy::NUCLEICACID be handled by IMP::atom::Residue. - IMP::atom::read_pdb() currently drops HOH ATOM records on the floor. We should add a IMP::atom::Water decorator if anyone wants to handle them (since they aren't atoms and don't have elements). - IMP::atom::Mass() has been added for decorating particles with mass. IMP::atom::Atom mass has been moved to that. - The scheme for generating atom types is now fully specified. See the IMP::atom::AtomType documentation. - IMP::atom::Atom and IMP::atom::Residue have been somewhat restructured. Some of the helper functions have been renamed to be more descriptive or removed (if they duplicated other easily accessible functionality). get_residue_index() has been removed (since you can just call get_index() on the return result of the IMP::atom::get_residue()). IMP::atom::get_chain() now returns an IMP::atom::Chain. root() was renamed to IMP::atom::get_root() for consistency. - SWIG files are now generated from a file called swig.i in the pyext directory. This file only has to contain module-specific commands such as lines to wrap headers and instantiate module templates. No more boilerplate. - The contents of bin are moving to tools since the two directories both contained assorted development tools (not all of which were executable). For now, bin/imppy.sh will still work, but that will go away soon. - IMP::Decorator::show() now just takes the stream (it used to take a prefix also, which was often ignored). Anyone who implemented a decorator will have to delete the second argument of their show function. ## Week of August 1 - Containers now have methods IMP::SingletonContainer::evaluate(), IMP::PairContainer::evaluate(), IMP::SingletonContainer::apply() and IMP::PairContainer::apply(). Using these can be significantly faster than looping over the particles externally. Similarly, there is IMP::SingletonScore::evaluate(), IMP::PairScore::evaluate(), IMP::SingletonModifier::apply(), IMP::PairModifier::apply(). - The decorator functions have been renamed to IMP::Decorator::setup_particle(), IMP::Decorator::decorate_particle() and IMP::Decorator::particle_is_instance(). A shell script \c tools/update-decorators is provided which updates the names in all code in subdirectories of the current directory. Please backup things first. - a function, IMP::Restraint::safe_evaluate() has been added to be used when requesting the value of restraints outside of the IMP::Model::evaluate(). ## Week of July 18 - The atom::BondPairContainer has been renamed to atom::BondPairFilter since it wasn't a real container. A real atom::BondPairContainer has been added. - Out of source builds mostly works. To do an out of source build, make a link to \c scons_tools and \c SConstruct from the source directory in the build directory and then add the line \command{repository="../path_to_src_dir"} in the \c config.py in the build directory. Then running \c scons in the build directory will put generated files there. This allows you to make both a \c release and \c fast build off of the same set of sources. Proper docs will be added soon. - The SConscript setup for a module has been greatly simplified so it should be easy enough for users to maintain their own module and updates to the build system shouldn't involve changes to user SConscripts any more. This does mean that any non-svn modules need to be updated. Look at IMP.core or another module and copy what is there. Again, proper docs will be added soon. ## Week of July 11 - Per Javi's suggestion, the conventions have been split into "Coding conventions" and "API conventions" and linked from the main page. - benchmarks got moved to their own module, IMP::benchmark and will now be tracked over time ## Week of July 4 - FilteredListSingletonContainer and FilteredPairSingletonContainer went away as there was too much hard to factor code shared with IMP::core::ListSingletonContainer and IMP::core::PairSingletonContainer. The IMP::core::ClosePairsScoreState is now handled by a list of filters on the class itself. That is, do IMP::core::ClosePairsScoreState::add_close_pair_filter() to add a filter. - Hooks have been added to allow actions to be taken on an assertion or check failure. See IMP::HandleFailure and its descendants for more details. - IMP::display::CGOWriter has been much improved and now allows provides lots of display-time control and easy building of animations. - Generic optimizer states have been added (IMP::core::SingletonsOptimizerState and IMP::core::PairsOptimizerState). ## Week of June 15 - box distributions (IMP::algebra::random_vector_on_box(), IMP::algebra::random_vector_in_box()) take bounding boxes - IMP::em::DensityMap are now IMP::Object objects. Some current allocations on the stack may need to be changed (since ref counted objects can't be stack allocated). - IMP::em::DensityMap::Read() is now deprecated (as is IMP::em::DensityMap::Write) as they violated a slew of conventions. Use IMP::em::read_map() and IMP::em::write_map() instead. ## Week of June 8 - new methods IMP::atom::add_radii() and IMP::atom::add_bonds(); - Hierarchys to IMP::atom::Hierarchies - IMP::Particles has been changed to a new type. It no longer plays quite so well with Python lists (you can't create one from an arbitrary Python list). - The containers for multiple decorators have been revised to make them more useful. See IMP::Decorators. You can now implicitly convert, for %example, a IMP::core::XYZs to an IMP::Particles (as well as use IMP::Decorators::get_particles()). All methods that take a collection of particles with a certain type now instead take a IMP::Decorators. - A method that has expectations about the particles it takes should take decorators rather than particles. That is, anything which expects IMP::core::XYZ particles, should take an IMP::core::XYZs instead of IMP::Particles. - IMP_LIST() now takes an extra argument (the name of the plural container) - In C++ IMP::Decorator objects are implicitly convertible to IMP::Particle* objects. Unfortunately, making this work in Python is problematic. - More methods take and return decorators rather than IMP::Particles. This means it is clearer what their expectations are and what they produce. ## Week of May 15 - The IMP::helper module was added and various functionality moved there. See its documentation page for a list of what is there. - IMP::algebra::SphericalCoords was renamed to IMP::algebra::SphericalVector3D to be consistent with other algebra classes. - IMP::algebra::Rotation2D is now IMP::InvalidDefault to be consistent with the other similar classes. - We are trying to be a bit more specific about what is expected from a IMP::atom::Hierarchy. Towards that end, there are two methods for extracting a representation from a hierarchy, IMP::atom::get_detailed_representation(), and IMP::atom::get_simplified_representation() - A method has been added to produce a simplified copy of an IMP::atom::Hierarchy. The method is still a bit crude and could benefit from other people inspecting the output and saying how it should be improved. - There is now an IMP::algebra::BoundingBoxD (and IMP::algebra::BoundingBox3D) to track bounding boxes. There are methods like IMP::algebra::bound_box(Sphere3D) which gets the bounding box of a sphere. - Each IMP::Object has its own log level. If IMP::Object::set_log_level() is called with a value other than IMP::DEFAULT, the value is used for code within and called by the object. To support this in your objects you need to add IMP_OBJECT_LOG to the start of each major method. It has been added to some of the core objects. ## Week of May 8 - IMP::atom::MolecularHierarchy has been renamed to IMP::atom::Hierarchy. Use \command{sed -i.old "s\#MolecularHierarchy\#Hierarchy\#g" my_files} - the long-deprecated Name(Decorator) was removed - The decorator suffix has been removed from decorator names. - Unit conventions have been established and documented for distance, energy, force and angle. - there is a function IMP::atom::create_protein() to create a coarse grained protein. The function is crude so far, but the idea is that we should have standard ways of generating such things so that everyone can benefit from best practices. - a class IMP::atom::BondGraph has been added so Boost.Graph algorithms can be used with bond graphs - the IMP-specific doxygen commands have been documented. - the doxygen commands unstable{name} and untested{name} have been added to doxygen to be used for marking methods and functions as likely to change or not well tested. Pass the name of the class or method as name. - bonds can now be added to molecular hierarchies automatically see IMP::atom::add_bonds() ## Week of May 1 - There is a IMP::Rotation2D - You can now extract fixed Euler ZYZ angles from a IMP::Rotation3D. This code can serve as a template for extracting other Euler angle conventions if they are needed. - IMP::algebra::rotation_between_two_vectors() changed to IMP::algebra::rotation_taking_first_to_second() to make it clear which direction the rotation goes. ## Week of April 27 - IMP::Decorator derived objects are now convertible to bool. So code like \code if (d) { blah } \endcode works. - IMP::Decorator::cast now returns a null decorator (\c DecoratorName()) rather than throwing an exception. This might make it actually useful. - New IMP::statistics module for statistical methods (e.g. clustering). ## Week of April 20 - Support for taking advantage of rigid bodies in collision detection has been added via the IMP::core::RigidClosePairsFinder. - All IMP::core::ClosePairsScoreState-related functionality has been optimized and should be noticeably faster. - The EMBED project has been merged into the IMP::em module. Thus, it is no longer necessary to obtain EMBED separately to build IMP::em. - All C++ methods which work with std::istream or std::ostream arguments should now be usable from Python: any Python file-like object can be used. - The SWIG interface now ties in to the reference counting mechanism used in the C++ code, for all classes derived from IMP::RefCounted. So, for %example, deleting a Python object should not free the underlying C++ object if another C++ object holds a reference to it, and unrefing a C++ object should not delete it if a Python object still holds a reference to it. There should be no further need to manually keep references to Python objects to prevent things from being prematurely cleaned up. This should be transparent to the user; please report any problems. - IMP::atom::AtomDecorator no longer inherits from IMP::core::XYZDecorator. ## Week of April 13 - The decorators IMP::atom::AtomDecorator, IMP::atom::ChainDecorator, IMP::atom::ResidueDecorator etc. now inherit from IMP::atom::MolecularHierarchyDecorator. This means they can set their IMP::atom::MolecularHierarchyDecorator::Type properly and saves a lot of typing in the common case. We can always make hierarchy decorators zero memory overhead if we find that the cost matters. - The base classes IMP::Comparable, IMP::NullDefault, IMP::ValidDefault and IMP::UninitializedDefault were added to document what the default constructors do for various objects in IMP. ## Week of April 6 - IMP::Object now defines pure virtual methods IMP::Object::get_version_info() and IMP::Object::show(). All non-abstract classes which inherit from IMP::Object must implement them. Doing this means they are more uniformly implemented and are documented in one place. - IMP::Object is now reference counted (the only objects this changes are IMP::Model and IMP::Optimizer). As a result optimizers and IMP::Model objects must be created on the heap. - IMP::RefCounted has had its name changed and no longer inherits from IMP::Object. This makes it easier to use independently. - the docs for reference counting (IMP::RefCounted) and and IMP::Decorator have been improved - running doxygen produces no warnings. Let's try to keep it that way. Run \command{scons doc > /dev/null} before you check in large blocks of new code to make sure that you didn't introduce warnings. - Helper classes have been added to make SingletonModifier and PairModifier objects look like functors for C++ and Python. This makes it so you can use \c map in Python and \c for_each in C++. See IMP::SingletonFunctor and IMP::PairFunctor for more information. - The recent changes page has moved to a doxygen page to make it easier to link changes back to the relevant source code. ## Week of March 20: - create_rigid_body(), create_rigid_bodies(), create_centroid() etc. now require you to add the score state to the model yourself. Whether the rigid body is enforced or not can be controlled by adding or removing the returned IMP::ScoreState. - All previously deprecated code has been removed. This includes indexes in particles, the no-argument constructor for particles and the functionality that was moved from core to atom. - The IMP::core::RigidBodyDecorator and IMP::core::RigidMemberDecorator have been simplified by removing the traits classes. There is now only one type of rigid body and it stores the list of member particles internally. - IMP::core::NameDecorator has been deprecated as IMP::Particles have built in names (use IMP::Particle::get_name(), IMP::Particle::set_name()). - IMP::atom::ChainDecorator has been added ## Week of March 13: - A new IMP::display::Writer, IMP::display::ChimeraWriter has been added which exports geometry to native Chimera constructs. It probably should be used instead of bild and cmm for most purposes as it combines the best of both types. - compose() for transformation is now a free function so compose(rotation, vector) will work via promotions. - the geometry classes have been cleaned up and are now more consistent with IMP standards (and more likely to do what their names/comments suggest). See IMP::algebra::Sphere3D, IMP::algebra::Cylinder3D, IMP::algebra::Segment3D, IMP::algebra::Plane3D etc. - ParticleRefiner has been renamed to IMP::Refiner since the Particle was long and unnecessary. €eIF3 translation initiation complex](http://salilab.org/40S-eIF1-eIF3). - IMP::domino::DiscreteSampler::set_restraints() now allows the set of restraints used in DOMINO filtering to be controlled (previously, all restraints in the IMP::Model were used). - IMP::modeller::IMPRestraints now transfers an IMP::ScoringFunction into the MODELLER score, not the implicit (and deprecated) score of the entire IMP::Model. - Bugfix #775: coordinates of voxels in EM maps were previously inconsistently treated (they are supposed to be the centers of the voxels, but in some cases they were the lower left edge). The fix may result in maps and their bounding boxes appearing shifted by half the voxel size. # 2.4.0 - 2015-02-27 # {#changelog_2_4_0} - Python 3 is now supported. The Windows package now works out of the box with Python 3.3 or 3.4; a `-python3` subpackage is provided for Ubuntu and Fedora. (Apple does not yet provide Python 3 so the Mac package is not built with Python 3 support, but if you install Python 3 you can compile IMP from source to work with it.) - A new command line tool, `rrt_sample`, allows for running rapidly exploring random tree (RRT) based sampling, e.g. to sample conformations of multiple domains connected by flexible linkers. - Command line tools are now placed in the `bin` directory of modules (the old `bin` directory, used for utility programs that are not installed, is now called `utility`). The `applications` directory is removed. - Several Python modules that were only used by EMageFit have been moved from IMP::em2d to IMP::EMageFit. - A new tutorial that uses the IMP::pmi module is now available (modeling of the RNA Polymerase II stalk). # 2.3.1 - 2014-12-01 # {#changelog_2_3_1} - Minor bug fixes. - Update to IMP::pmi to support the Nup84 and rnapolii biological systems. # 2.3.0 - 2014-10-24 # {#changelog_2_3_0} - A new IMP::pmi module was added, containing high-level Python classes to simplify the representation, scoring, sampling and analyzing of models in IMP. This module is still under heavy development, and should be considered experimental. - Binary installers for 64-bit Windows are now available. - The `.py` extension has been removed from all Python applications, so that they are consistent with the C++ applications. - IMP::atom::Selection objects can now be combined using basic set operations (union, intersection, difference, symmetric difference). In Python the equivalent operators (|, &, -, ^ respectively) can be used. - The integrative docking (idock) application now employs SOAP to improve the scoring. - IMP::atom::CHARMMParameters::create_topology() now works with a wider range of hierarchies (previously it would only accept Residues that were direct children of Chains). For example, it should do the right thing when given a Hierarchy that contains no Chain particles, or one where Residues are grouped under Fragments. - Only Python 2.6 or later is now supported. We still provide IMP for RedHat Enterprise 5 (which ships with Python 2.4) but you will need to use the `python26` binary (part of EPEL) rather than the system default Python. # 2.2.1 - 2014-08-11 # {#changelog_2_2_1} - Bugfix: PoseScore (see the ligand_score application) should now match that used by the Pose&Rank web server (http://salilab.org/poseandrank/) # 2.2.0 - 2014-04-09 # {#changelog_2_2_0} - A module IMP.template was added to act as a template for creating new external modules. It includes setting up git hooks for format checks and things like that. - The Python support for grids was cleaned up. Hopefully this makes it more robust. If you were using a grid in Python that no longer exists, ping us, we can add it back. - The generic geometry functions were simplified slightly and renamed to IMP::algebra::get_vector_geometry(). - The IMP::base::Flag class was added to simplify adding flags in C++. - The `--log_level` and `--check_level` command line argument now take strings (eg VERBOSE) to specify their values, rather than cryptic (and a bit unstable) numbers. - Command line options are now divided into normal and advanced ones. You do `--help_advanced` to show the advanced ones. - Support for computing Connolly surfaces was added with IMP::algebra::get_connolly_surface(). In addition, there is now IMP::algebra::get_uniform_surface_cover(const IMP::algebra::Sphere3Ds&,double) to sample the surface of a collection of balls. - IMP::atom::reate_simplified_from_volume() was added to create a simplified representation of a hierarchy that preserves the surface are and occupied volume. It uses IMP::algebra::get_simplified_from_volume(). If you want to play with such representations, you can use the rmf_simply application. - IMP::atom::get_rmsd(IMP::atom::Selection, IMP::atom::Selection, const IMP::algebra::Transformation3D&) was deprecated. Use IMP::atom::get_rmsd_transforming_first() instead. - IMP::algebra::get_rmsd() and IMP::algebra::get_rmsd_transforming_first() were added. - The family of macros IMP_BASE_[SMALL,LARGE]_[UNORDERED,ORDERED]_[MAP,SET] have been added (eg IMP_BASE_SMALL_ORDERED_SET ()). They provide platform and boost-version independent ways to take advantage of the best set/map for an application. In particular, they use boost::container::flat_set and kin when available. - IMP::atom::State and IMP::atom::Representation decorators have been added to provide support for multiple states and multiple representation schemes. They are integrated with RMFs and IMP::atom::Selection. See atom/multiresolution.py and atom/multistate.py for examples. - The container implementations has been cleaned up and simplified a bit. This should not change most usage. - [RMF](http://salilab.github.io/rmf) has been updated. There are a few minor API changes, in addition to a new, faster format (`.rmf3`). - [Eigen](#Eigen) is now included with IMP. - The deprecated function IMP::kernel::Optimizer::set_restraints() was removed. Use IMP::kernel::Optimizer::set_scoring_function() instead. - The deprecated constructor for IMP::kernel::ModelObject that does not take an IMP::kernel::Model was removed. This propagates to constructors IMP::kernel::Restraint, IMP::kernel::ScoreState and kin. Make sure you pass an IMP::kernel::Model as the first constructor argument. Related to this, deprecated functions like get_is_part_of_model(), set_model() were removed as they don't serve any purpose. - The deprecated methods in IMP::kernel::Optimizer that pertain to attribute optimization were removed. Inherit from IMP::kernel::AttributeOptimizer instead if you want to use them in your optimizer. - Some deprecated functionality was removed. If you encounter problems, replace IMP::base::OwnerPointer with IMP::base::PointerMember, find logging stuff in IMP/base/log.h, object in IMP/base/Object.h and random in IMP/base/random.h. - IMP::atom::Chain now uses strings for the chain ids, to be consistent with changes in the PDB. PDB files are written using the first character of the chain string. - Added orientation-dependent SOAP scoring (IMP::score_functor::OrientedSoap and IMP::atom::OrientedSoapPairScore). Orientation-dependent SOAP scores include those that score loops (SOAP-Loop), protein-peptide interfaces (SOAP-Peptide) and proteins (SOAP-Protein). The library files themselves are rather large (~1.5GB) and so are not included with IMP. They can be downloaded separately from http://salilab.org/SOAP/. # 2.1.1 - 2013-11-04 # {#changelog_2_1_1} - Added support for Mac OS 10.9 # 2.1.0 - 2013-10-24 # {#changelog_2_1_0} - There is now some limited IMP-wide gathering of timing statistics. You can add `--statistics=filename` to the command line or use the IMP::base::show_timings() function to view a summary of where IMP has spent time. - Slack was being used inconsistently (and not always correctly) among the various containers and restraints. If you had heavily optimized the slack value for speed, you may want to revisit it. - Class name OwnerPointer was switched to PointerMember, to denote more clearly a ref-counting pointer that is a class member. - The dependency handling in IMP has been rewritten to simplify it and fix a bunch of bugs. There are now more checks of correctness, so behavior that might have silently worked (or not worked) before may now trigger errors. In particular, restraints/score states must have access to the IMP::kernel::Model to do much of anything, so either use constructors that take the model or call IMP::kernel::ModelObject::set_model(). Also, changing your set of inputs or outputs without calling IMP::kernel::ModelObject::set_has_dependencies(false) is now likely to trigger a check failure (it would just silently do the wrong thing before). - Decorators have been cleaned up to make them behave consistently regarding IMP::kernel::Particle vs IMP::kernel::ParticleIndexes. This change shouldn't break existing code, but decorators should move to the new macros IMP_DECORATOR_METHODS(), IMP_DECORATOR_SETUP_0() etc. - The IMP_GRAPH() and IMP_WEIGHTED_GRAPH() macros got an extra argument describing how to show the vertexes in the graph. They also now expose a method, eg IMP::kernel::show_as_graphviz() that writes the graph in graphviz format to an IMP::base::TextOutput. - Support and rules for deprecating code in IMP have been updated. There are new macros and instructions on how to use them that should result in much better visibility about changes. See the C++ wiki for IMP for more information. And there is now an encoded policy of keeping deprecated things around for 1 release after deprecation. - IMP is now much more selective about when it invalidates the dependency graph. See [Dependencies](http://github.com/salilab/imp/wiki/Dependencies) for more information. This change should make evaluation of isolated restraints much faster in certain situations where they were unexpectedly slow before. - IMP::atom::Selection has been completely rewritten as it was very brittle. The set of particles returned may change in some cases, in some cases because it was buggy before, in some cases because of new bugs. - Reminder: the IMP_PROTECTED macros didn't work out well, and are now deprecated. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading `_` on the method name, if you get `Swig director pure virtual method called IMP::kernel::PairScore::do_get_inputs` messages. - The generated `cmake` files are no longer stored in the repository. Instead `tools/build/setup_cmake.py` is run after checkouts and other git events that change the repository tree (assuming you have run `setup_git.py`). - Generation of documentation using `doxygen` has been refactored. It is now done on a per-module basis, via targets like `IMP.core-doc`. The main advantage is that changes to the doc for a module can be tested comparatively quickly, just by building that module's docs. On the down side, docs are no longer quite so heavily interconnected, so dependencies only work to things that the module actually depends on. - The restraint and score state statistics functions were removed from IMP::Model. They had been broken for quite a while, with no one complaining. Statistics can be added back easily if requested. - Added an IMP-specific fork of `git-flow`. It handles things like `README.md` files when you start and finish feature branches and provides a tool to nicely update IMP and display the change log. The main commands of use are `git imp feature start <feature_name>`, `git imp feature finish <feature_name>`, and `git imp update`. You may want to remove the `git-flow` lines from your `.git/config` file as you may accidentally type `git flow` instead of `git imp`. You need to run `setup_git.py` to set things up. - [RMF](http://salilab.github.com/rmf) is now included as a git submodule rather than copied into the IMP repository. You should rerun `setup_git.py` to make sure submodule stuff is initialized. # For IMP 2.0 and earlier # {#changelog_2_0} ## Spring 2013 - Modules and applications now have a _Maintainer_ in their `README.md`, giving the github id of the person to whom issues should be assigned. - `setup_git.py` supports the `--module` flag which will set up git hooks and things for a git repository that contains a module. - To clean up for the release, the module dependencies are now described in a file called `dependencies.py` and the module docs in a `README.md`, both in the main module directory. - IMP-wide support for progress bars was added with the IMP::base::set_progress_display() function and helpers. These are only displayed when the IMP::base::get_log_level() is IMP::base::PROGRESS. ## Winter 2012-2013 - Scons support was removed. Use cmake to build IMP, it is way better. - Examples are run with a `--test` argument when IMP is built with internal checks. Expensive examples should check this flag and shorten what they do in order to try to test all their code without taking too long. - The class IMP::core::NonRigidMember was added to allow particles that are not rigid but whose coordinates are specified via internal coordinates in some IMP::algebra::Reference3D that is part of an IMP::core::RigidBody. - The IMP_BIG_ENDIAN and IMP_LITTLE_ENDIAN preprocessor symbols are deprecated in favor of the BOOST_BIG_ENDIAN and BOOST_LITTLE_ENDIAN symbols. - Support for asymmetric moves was added to IMP::core::MonteCarlo. To do this, IMP::core::MonteCarloMover was added and IMP::core::Mover and IMP::core::MoverBase were deprecated and implemented in terms of the new class. This should require few or no changes in code that uses Monte Carlo support or implements movers (although movers should move to IMP::core::MonteCarlo). All IMP::core::MonteCarloMover objects keep track of statistics (number of moves and number of accepted moves), rather than that being done in the IMP::core::SerialMover. In addition, the poorly defined move probability was removed from IMP::core::MonteCarlo. - IMP::base::WarningContext now outputs warnings immediately. This change was needed as the previous behavior resulted in occasional core dumps when the libraries for log support were unloaded before the libraries where static IMP::base::WarningContexts were declared. And there was no easy way to detect such problems. - You can now (probably) build IMP with [CMake on IMP wiki](https://github.com/salilab/imp/wiki/Cmake). - The maximum logging level and check level can now be controlled independently of the build mode. To, for example, enable progress level logging in a fast build, add maxlog="PROGRESS" to your config. The maximum check level for `release` has been reduced to IMP::USAGE as that speeds things up a bit. - The way applications are structured and specified has been regularized and simplified. Python files are automatically copied/installed and binary executables are compiled from each found `.cpp` file linked with any `.cpp` files contained in a subdirectory called `lib`. Nothing needs to be put into the SConscript file. - The Python constants like IMP.algebra.has_cgal went away as swig autogenerates ones like IMP.algebra.IMP_ALGEBRA_HAS_IMP_CGAL. While the latter are a bit uglier, the constants aren't used much and having them be auto generated is nice. - There have been some major changes to the build system. The main one is that the SConscripts in the directories below the module directory aren't really used. They are being removed from the repository. Now everything really is off of file locations (eg all test_* in test is a test, each *.cpp in bin becomes a program etc). This greatly simplifies the build system, which has, historically been a disaster. You can now define dependencies by dropping files into a dependency subdirectory in your module, no need to mess with the SConscript. - Building IMP now always requires specifying a `repository` argument. This was always required for out of source builds, which is what one should be doing, but is now required for in source builds to, so as to make accidentally polluting your source with an in-source build harder. Use `repository="."` in that case. - The HDF5 functionality in RMF was reorganized. Specifically, the headers were moved into `RMF/HDF5`, the Python functionality into `RMF_HDF5` and the code into namespace `RMF::HDF5` and all mentions of `HDF5` or `hdf5` in the class and function names were removed. - Support for IMP-wide command line flags have been added to IMP::base in the base/flags.h file. They provide a simple command line parsing interface that can be used from C++ and Python. In addition, various functionality in IMP is automatically exposed on the command line when those functions are used. For example, one can control logging, checks or profiling. See IMP/base/flags.h for (some) more info. - OpenMP support in IMP is officially released. All tests pass. But there has been little specific testing so far. - Access to attributes using IMP::ParticleIndex and the IMP::Model now works from Python. See IMP::Model::add_attribute() and kin. - IMP::ParticleKey and IMP::ParticlesKey were renamed to IMP::ParticleIndexKey and IMP::ParticleIndexesKey respectively as that is what they store. Typedefs were provided to support old code. - Indexes have been added to the docs connecting classes with all the methods that return them, examples that use them and methods that take them. And all methods that are used in examples with the examples they are used in. Please check them out and report issues and ways to improve things. You can find links on the main page. - The docs for a modules/biological systems/applications should now be found in an `overview.dox` in the module `doc` folder. Modules in IMP svn have been updated. For external modules, you can just move a `overview.dox` from that was previously generated into your `doc` folder and (eventually) remove the lines from your `doc/SConscript`. This makes it easier to use the full range of doxygen markup, as well as removes the need to escape doxygen markup for Python. - Examples have now been simplified into full blown doxygen examples. This involves merging the `.py` and `.readme` files (putting `\#\#` in front of each line of the `.readme`) and adding a `## \\example module/examplename.py` line at the beginning. You can use `./tools/updaters/update_examples.py modules/mymodule` to update a module that was not already updated. - IMP::base::FailureHandler has been removed as it wasn't used. - Handling of data and includes has been simplified. All appropriate files in the corresponding directories are linked into build/installed. This is only likely to be a problem if you had broken headers sitting around `include`. If so, just comment them out, or move them some place nicer. - Modules now must be in a folder called `modules`, which must have a `description.py` file. - The build system should now be a little better about cleaning up after itself. - How headers/doc files/data/examples - The IMP_PROTECTED macros didn't work out well, and are now deprecated. Sorry. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading "_" from the method name. - Nice version numbers will now be generated if you use git. You need to rerun tools/git/developer_tools/init_git if you have an existing clone. ## Fall 2012 - 11/9: RMF is now an external library shipped with IMP. IMP will build it and install it automatically if you have cmake version 2.8 or higher installed. - 11/9: Updated support macros for IMP::Restraint and IMP::ScoreState classes have been added (IMP_RESTRAINT_2() and IMP_SCORE_STATE_2()) that make it easier to return dependencies on things other than IMP::Particle and IMP::ScoreState objects. In addition, the dependency functions on scores and modifiers have been updated similarly and standardized. In general this should not cause backward compatibility problems unless you were overriding the IMP::ModelObject::do_get_inputs() in IMP::Restraint and using the IMP_RESTRAINT() macro. In that case update to the IMP_RESTRAINT_2() macro. - 11/8: Support has been added for the C++11 keywords override (IMP_OVERRIDE) and final (IMP_FINAL). They allow you to tell the compiler to produce an error if the marked method does not override/implement a parent's method and if anything overrides a given method, respectively. - 11/8: IMP::SingletonModifier::apply_indexes() and IMP::SingletonScore::evaluate_indexes() now require an upper and lower bound index (you can just pass 0 and the size of the array). We can make them option if there is a good reason. - 11/8: IMP::SingletonDerivativeModifier and kin were merged into IMP::SingletonModifier and kin as the only difference, the presence of the IMP::DerivativeAccumulator (whose weight was always 1) wasn't very interested and the split added lots of complexity. If this has an impact on your code, let us know. - 11/8; Evaluation with containers was greatly simplified, moving many methods on containers. We can add backwards compatibility versions if that is interesting. - 11/7: Predicates, scores and modifiers written in Python need to update their methods for returning inputs and outputs. They should now have a method \c _do_get_outputs(self, m, particle_indexes) which should probably just return \c [m.get_particle(i) for in in particle_indexes] to return all the passed particles. - 11/7: It is now preferred that IMP::Restraint and IMP::ScoreState descendants report dependencies directly using IMP::ModelObject::do_get_inputs() and IMP::ModelObject::do_get_outputs(). To do this, you should use the IMP_RESTRAINT_2() and IMP_CONSTRAINT_2() macros instead. Methods like IMP::Restraint::get_input_particles() are now deprecated. This change was made as the old mechanism introduced a lot of extra complication and made it so that there was no way to directly depend on something that was not a Particle or a Container. - 11/3: IMP::container::InContainerPairFilter and friends now treat pairs as unordered. A flag was added to control this. - 9/5: IMP::base::Object classes now have a method IMP::base::Object::clear_caches() that clears all cache data in the object. This is not universally implemented, but now provides a correct path for doing that. The IMP_LIST() type macros call that method when their contents changes. ## Summer 2012 - 7/3: IMP_PERIODIC_OPTIMIZER_STATE() was superceded by the class IMP::core::PeriodicOptimizerState as it is rather cleaner than the macro. - 7/2: in order to make various base classes in IMP more Python friendly, macros IMP_PROTECTED_METHOD(), IMP_PROTECTED_CONSTRUCTOR() were added that properly expose such methods to Python. The standards checks now complain if \c protected is used an a class in the API. - 6/24: Executables from \c module/bin are now put in \b build/module_bin/module. This removes a recurring source of errors in the build script due to the origin and destination directory have the same name in scons. - 6/24: Benchmarks are now put int \c module/benchmark and build to \c build/benchmark/module to remove a recurring source of errors. - 6/18: The IMP::benchmark module supports the GPerfTools profilers when available. Benchmark executables (things in \c bin whose name starts with \c benchmark_) that use the IMP_BENCHMARK() macro have command line flags to turn on and cpu and memory profiling. - 6/12: The naming rules for .i files that are in module \c pyext directories changed. They should now be \c IMP_modulename.name.i instead of \c IMP_modulename_name.i. This is to better support underscores in names. ## Winter 2011/2012 - 20/4: A method, IMP::atom::read_pdb(IMP::base::TextInput, int, IMP::atom::Hierarchy), was added to load coordinates from a pdb file into an existing IMP::atom::Hierarchy. - 4/4: the filter types were superceded by the predicate types. You will only see a difference if you develop a filter class (eg IMP::PairFilter -derived class) or directly call methods on filters (specifically IMP::PairFilter::get_contains()). - 29/3: IMP::Particle::get_has_model() has been changed to IMP::ModelObject::get_is_part_of_model() to be consistent with other classes. - 28/3: IMP::core::Mover changed slightly, adding an extra method IMP::core::Mover::get_output_particles() and requiring the IMP::Model be passed to the constructor. All movers in IMP SVN were updated without any difficulty. This allows movers to use information contained in the dependency graph amongst particles so they can, for example, weight moves based on scores. - 26/3: IMP::SingletonContainerInput etc were added. Methods that take these as their arguments can be passed either containers or lists of particles (or tuples), as convenient. All methods that take IMP::SingletonContainer objects should eventually move over to taking this (changing just requires changing the argument from a pointer to an Input). - 22/3: Over the last week or so, which IMP headers include which other heads have been substantially simplified. This may result in compilation errors for code outside of \c svn that depended on IMP headers including other headers. In general, you want to explicitly include the header defining each thing you use. - 15/3: The vector used throughout IMP is now exposed at IMP::base::Vector and what it does it documented. - 10/3: IMP::domino::RestraintCache was introduced to provide a documented centralized place to handle restraint evaluation and caching in DOMINO. In the switch various bugs involving maximum scores and IMP::RestraintSet objects were fixed. - 10/3: Scoring functions in IMP are now implemented in IMP::ScoringFunction objects. All restraint evaluation is handled by them and they can cache various useful information to aid in accelerating restraint evaluation. Backward compatibility interfaces have mostly been provided, please complain if some important interface went away. This has greatly simplified the code as well as fixed various bugs in various types of restraint evaluation (particularly in the handling of maximum scores on IMP::Restraint and IMP::RestraintSet objects. - 29/2: The macros IMP_NAMED_TUPLE_1(),IMP_NAMED_TUPLE_2(),IMP_NAMED_TUPLE_3() were added for easy generation of simply struct for use in passing and returning values. The generated structs support comparison, output, hashing and have constructors. Use the macros instead of a boost::tuple or a a POD struct. - 16/2: Restraints now are evaluated with the correct weights when evaluated by themselves as opposed to always returning their weight as seen by the model as a whole. - 15/2: The IMP::algebra grid support was rearranged and simplified. Most of the changes should be transparent. However, the IMP::algebra::grids namespace got folded into IMP::algebra, which might break some code. And various (obscure) constructors of IMP::algebra::GridD were removed. All code in svn has been changed as needed. - 3/2: The IMP::domino HDF5 container has been revamped. This involves splitting it into a write version (IMP::domino::WriteHDF5AssignmentContainer) and a read version (IMP::domino::ReadHDF5AssignmentContainer). - 3/2: IMP::domino::WriteAssignmentContainer and IMP::domino::ReadAssignmentContainer were added to do direct binary file writing. This is faster than the HDF5 versions. - 3/2: The installation instructions have been revamped and the developer instruction, in particular, have been greatly simplified. - 24/1: \c tools/show-changes has been added that shows the changes in the history log since you last updated. - 24/1: \c tools/make-source has been added that adds a source/header pair to a module - 24/1: principal components analysis in IMP::algebra was generalized to work in any dimension. This required renaming the class to IMP::algebra::PrincipalComponentAnalysisD. - 24/1: the output of the IMP::algebra::PrincipalComponentAnalysisD::show() was changed to reflect the standards (values produce compact, single line output on show). The old show has been moved to IMP::multifit - 19/1: IMP::container::ConsecutivePairContainer had IMP::container::ExclusiveConsecutivePairContainer split off of it to make the two operating mode clearer and so that there could be a clear IMP::container::ExclusiveConsecutivePairFilter to use. - IMP::algebra::Grid::apply was added to provide faster access to all (or, later, part of a grid). It can be an order of magnitude or more faster than using the iterators when doing simple operations. - the scons config options \c pretty and \c color were added (defaulting to True). When they are true, build commands are mostly suppressed and a much briefer description of what is being done is printed out (in color). Note that \c color mode and compilation in emacs don't get along too well. - IMP::algebra::ParabolicFit was renamed to IMP::algebra::ParabolicFit2D (a backwards compatibility typedef is currently provided) - IMP::algebra::LinearFit was renamed to IMP::algebra::LinearFit2D. It now supports error bars on the y-values. ## Fall 2011 - IMP::compatibility::checked_vector was renamed to IMP::compatibility::vector and made available to IMP as IMP::vector. Typedefs should use this type (IMP::vector). - Various IMP::algebra plural typedefs which had been using std::vector (since they never were updated), have now been moved to IMP::vector. - IMP::statistics::HistogramD now supports display with \c matplotlib. - IMP::statistics::HistogramD supercedes IMP::statistics::Histogram - predicate base classes have been added: IMP::SingletonPredicate, IMP::PairPredicate etc. These take particles and return an integer. Current functionality using them includes IMP::core::PredicateSingletonScore that uses a predicate to choose which score to apply. - the core generic restraint and score state support was moved to the kernel to avoid breaking various scores outside of IMP svn. - the IMP::display module was refactored and moved up to be a dependency of IMP::core and IMP::atom rather than vice versa. The main observable results are that you should now use IMP::atom::HierarchyGeometry, IMP::atom::SelectionGeometry, IMP::atom::HierarchiesGeometry, IMP::core::XYZRGeometry, IMP::atom::BondGeometry etc instead of finding them in IMP::display. This change means that, display can be used internally when testing and debugging functionality in IMP::core and IMP::atom and that tests that use IMP::display don't need to build all of IMP::atom. - Decomposition of restraints now results in restraints that are faster to evaluate, using the generic restraint mechanism in IMP kernel. - IMP::algebra::get_unit_bounding_box_d() was changed to return a bounding box with radius 1 to be consistent with IMP::algebra::get_unit_sphere_d() - A general purpose scheme to associate data with tuples of hierarchy nodes was added to the RMF format. This deprecates the special purpose bond support. - the way various types of geometry is stored in RMF files was changed, breaking backwards compatibility. ## Summer 2011 - some functionality from the kernel has been spun off into a new module IMP.base. You probably won't notice anything. The only significant changes are that functions that accept or return IMP::base::TextInput and IMP::base::TextOutput must declare them with the \c base qualifier (eg base::TextInput) so that swig is happy. - the source for the kernel has been moved from \c kernel to \c modules/kernel. This is unlikely to effect anyone. - IMP::core::ExcludedVolumeRestraint now supports filters - IMP::domino::HDF5AssignmentTable now caches internally, which makes it several times faster than before. - you can now pick which Python to use with the \c python argument to scons. - the units for BD diffusion coefficients changed to be consistent with other IMP units. - IMP::domino::get_interaction_graph() now takes an IMP::RestraintsTemp, like other functions - The assert failure handling function was documented and renamed to IMP::handle_error() so it is easier to find when you want to use gdb. - to work around various issues with ref counting pointers and compiler eccentricities, we added emulation of the C++0x \c nullptr. You should prefer that to \c NULL in IMP code that you want to run cross platform. - IMP::IntsList, IMP::FloatsList and IMP::StringsList have been added to replace the various scattered solutions for passing such types. - IMP::Restraint::get_decomposition() and IMP::Restraint::get_current_decomposition() changed to return a single Restraint, instead of a list. This makes it easier to propagate the weight and maximum reliably. - IMP now supports usage of restraints without adding them to the IMP::Model scoring function. To do this, do IMP::Restraint::set_model(), passing the model. The restraint can then be evaluated. - IMP::domino::DominoSampler does not use the IMP::domino::OptimizeRestraints restraint optimization by default. Instead, if it is asked to create the IMP::domino::RestraintScoreSubsetFilterTable, it decomposes the restraints first (via IMP::create_decomposition()). This avoids several bugs and very fragile code. - the IMP::core::DiameterRestraint was ignoring the radii of the particles involved and so was giving artificially low scores. It now is on the actual diameter of this molecule. - the IMP::core::MonteCarlo optimizers now support incremental evaluation. This can be a lot faster when only a few particles are moved each step. - all the plural types in IMP have moved to bounds checked vectors when using debug builds with gcc. If you get a non-imp error about an invalid access, it is probably some code that walks off the end of an array. Please report it. - The IMP::core::MoverBase was changed to move it off of the IMP::SingletonContainer. This resulted in the IMP::core::BallMover and IMP::core::NormalMover having new constructors. Now create them from a list of particles, a list of float keys and a radius/standard deviation. This removes some complexity and inefficiencies as well is removes some scope for errors. - Temp and non-temp decorator lists have been merged in C++. This was done to simplify code since no one was using the non-temp variants. Now all decorator lists are temp (non-reference counted) variants. - restraints and restraint sets are now evaluated only using "canonical" weights. That is, each restraint has one weight for the model (the total over all the restraint sets it is contained in and will always be evaluated with that weight). This allows the evaluation of simple restraints to be faster. - The IMP_LIST() macros have been changed slightly. If your usage needs to take action upon addition, removal or changes to the list, use the IMP_LIST_ACTION() variant in your header. The IMP_LIST_IMPL() macro no longer takes the actions, and the last three (generally empty arguments) will have to be removed from the call to get it to work. This is needed to move towards more flexible and efficient passing of lists of values. - IMP::domino uses the new IMP::Model::evaluate_if_good() support to accelerate computations - IMP::core::MonteCarlo can use the IMP::Model::evaluate_if_good() support to more quickly reject very bad conformations. See IMP::core::MonteCarlo::set_maximum_difference(). - IMP::core::ExcludedVolumeRestraint has been rewritten to be faster. It is especially faster when used with IMP::Model::evaluate_if_good() - a new evaluation mode IMP::Model::evaluate_if_good() and IMP::Model::evaluate_if_below() has been added. It allows restraints to shortcut score computation when the score can be shown to be worse than a maximum value. - as part of moving towards more efficient use of evaluate_if_good() various deprecated methods have been removed from IMP::core::ExcludedVolumeRestraint and IMP::core::RigidClosePairsFinder. - incremental evaluation support was removed from IMP. The reason for doing this was that the code was complicated, created overhead even when incremental support was not used and was not commonly used. Further, decomposing the restraints and using the dependency graph yields similar results and is more transparent. - the RMF helper binaries have been moved to their own application so that they get installed. - IMP::display::Writer has been revised to better support writers that write multiple frames to a single file and ones that write binary data. The noticeable changes are that the existing IMP::display::Writer instances must be created with a file name. And the IMP::display::WriteOptimizerState just takes a writer rather than a write and a file name template. IMP::display::Writer::set_frame() has been added so that you can directly control the frame being written. ## Spring 2011 - An interactive mode has been added to IMP::domino::DominoSampler. The interactive mode is useful for understanding domino sampling runs and distributing domino computations across a cluster. - Functions IMP::domino::set_assignments() and IMP::domino::get_assignments() were added which support writing subset assignments to data fields in hdf5 files. These can be used to helping to distribute domino on a cluster as well as debugging the sampling process. - a Python function IMP.show_graphviz() has been added that attempts to render a graph with \c dot and then open the pdf. - The Python functions IMP.show_altgraph() and IMP.get_networkx_graph() have been added to provide alternatives for quick viewing of IMP graphs. Unfortunately, they still pale compared to graphviz. - IMP.atom.create_cover() was added to facilitate creating a summary particle for part of a hierarchy in order to accelerate restraints. - IMP.hdf5 was renamed to IMP.rmf, the proposed name of the file format that it supports. That also better separates the hdf5 support from the file format support. - The old I/O functions have been marked as deprecated in favor of IMP.hdf5. You may need to compile with deprecated="yes" if you use them. - The Python graph interface now has a remove_vertex method. - One can now access the model restraint evaluation statistics programmatically using IMP::Model::get_restaint_statistics(). Restraint statistics can be logged to a file during optimization using IMP::core::WriteRestraintScoreStatisticsOptimizerState. - IMP::domino::DominoSampler now allows you to specify the tree on which merges occur directly. Also, you now need to specify the merge tree rather than the JunctionTree if you want to do non-loopy inference and specify something about how it is done. Sorry, it makes the code a lot easier. - Preliminary symmetry support has been added via the IMP::core::TransformationSymmetry and IMP::core::TransformationAndReflectionSymmetry classes. ## Winter 2011 - There now is a common base class IMP::atom::Simulation for IMP::atom::BrownianDynamics and IMP::atom::MolecularDynamics. In merging them there were several minor changes made to both. The most noticeable may be that the function for setting the maximum time step length is now IMP::atom::Simulation::set_maximum_time_step(). The last time step length can be extracted by IMP::atom::Simulation::get_last_time_step() - IMP::core::MonteCarlo::set_temperature() was renamed to IMP::core::MonteCarlo::set_kt() as that is what it is doing. - A new module IMP.hdf5 has been added to provide more general support for saving and loading IMP::atom::Hierarchy structures to files. Multiple conformations of each hierarchy can be saved as well as information about geometry and restraints (although these can't be read at the moment, so they are not so useful). - IMP::algebra::grids::GridD now supports dense grids in arbitrary dimension through use of IMP::algebra::grids::DenseGridStorageD. - IMP::Optimizer now can use a list of restraint sets to compute the score rather than the complete Model score. - each IMP::Restraint now has its own weight. This is lightly tested. - A new log level IMP::PROGRESS has been added. The idea is that IMP::Sampler::get_sample() calls and other expensive loops can show a boost progress meter at this log level to give some idea when things are moving forward. - IMP::domino::TrivialParticleStates renamed to IMP::domino::IndexParticleStates to be less judgmental - IMP::domino::CompoundParticleStates and IMP::domino::RecursiveParticleStates added - IMP::domino::DominoSampler now supports maximum scores on IMP::RestraintSets (untested). - the name of the clustering methods have been changed to create_ from get_ since they create objects. - the names of the recently added metric based clustering support has been changed to Metric rather than Distance. - IMP now has the ability to have certain tests designated as expensive and to skip those tests when test-fast is run. The motivation for this is that the tests take a really long time making it impractical to run before every commit. And it is hard to tell which tests have not been run when running with the necessary multiple build processes. My thought is that tests that take about a second or less on a debug build should be left along (this is probably >95% of the tests), but the few that take longer should be marked as expensive. To do that, rename the test to expensive_text_XXX.py (from test_XXX.py) and add expensive_python_tests= env.IMPModuleGetExpensivePythonTests() to the IMPModuleTest call in the test SConscript. - the fuction IMP::display::create_restraint_geometry() was added to do what its name says. - runtime dimensional geometric objects have been added. These include IMP::algebra::VectorKD, IMP::algebra::BoundingBoxKD and IMP::algebra::SphereKD. The IMP::statistics::Embedding classes now use IMP::algebra::VectorKD instead of Floats. The sparse grids also support variable dimensions as to the nearest neighbor searches. - IMP::atom::DopePairScore has been added to IMP. - IMP::atom::CHARMMStereochemistryRestraint provides a high-level simplified interface to the bonded part of the CHARMM forcefield. - IMP modules are now built by first combining all the .cpp files into one, and then building that file. This greatly accelerates building of a module from scratch as build time is dominated by parsing headers. This does change the semantics of .cpp files as they see headers included by and functions declared by an arbitrary subset of other .cpp files. For now they are still required to be able to be compiled separately. To get the only behavior back either globally or on a per-module basis, see the percppcompilation parameter to scons. - Instructions and scripts have been added for building modules and things externally to the IMP build. See the installation guide. - IMP::atom::CHARMMParameters can now automatically map CHARMM-style atom and residue naming to PDB-style, so CHARMM topology files can be used unmodified. - IMP::atom::CHARMMTopology::add_sequence() provides a simple way to generate a topology from a string of amino acid one-letter codes. - IMP::atom::CHARMMTopology::add_coordinates() will generate atomic Cartesian coordinates from CHARMM internal coordinate information. - IMP::atom::CHARMMTopology::add_missing_atoms() and IMP::atom::remove_charmm_untyped_atoms() can now be used to ensure that atoms read from a PDB file match the CHARMM topology. - building of deprecated functionality is now false by default - the IMP_COMPARISONS(), IMP_COMPARISONS_1(), IMP_COMPARISONS_2() macros now take the name of the class as an additional (first) argument. This avoids the formerly needed, \c This typedef and makes them more consistent with the other macros in IMP. - modules can now be built externally to the IMP svn in addition to applications and biological systems. To do so, create a directory with your module in a subdir containing links to the IMP SConscript file and scons_tools and a config.py that has information needed to run IMP (eg an appropriate \c pythonpath, \c ldpath). - removed IMP::algebra::GridD::get_index(VectorD) as it wasn't very safe (in that vectors coordinates can get rounded between the IMP::algebra::GridD::get_has_index() and IMP::algebra::GridD::get_index() calls. Use IMP::algebra::GridD::get_nearest_index() or IMP::algebra::GridD::get_extended_index() or IMP::algebra::GridD::operator[]() instead. - renamed IMP::domino::ParticleStatesTable::get_particles() to IMP::domino::ParticleStatesTable::get_subset() as it is more useful to be able to get the subset containing all the particles. - rename IMP::restrainer::Restraint::print_all_restraints() to IMP::restrainer::Restraint::show_all_restraints() to be consistent with the rest of IMP. - renamed IMP::atom::get_conect_record_string() to IMP::atom::get_pdb_conect_record_string() to make clearer what it is. - class and method names are now spell checked. Add missing words to spelling_exceptions in the module test call. - the plural names (eg IMP.Particles()) have been removed from the Python side as they are just Python lists. - biological systems can now include internal modules. These provide a convenient place to put application specific C++ and Python code. See the example system for more info. - IMP_VALUES() now includes IMP_OUTPUT_OPERATOR() so you no longer need two namespace macros for each value. - the function IMP::core::Harmonic::k_from_standard_deviation was renamed to IMP::core::Harmonic::get_k_from_standard_deviation() to conform to the standards. - the IMP::atom::PDBSelector and IMP::atom::Mol2Selector became IMP::Object classes. All this means is that they need to be created with \c new in C++ code. It also means that one is no longer prohibited from storing them (not that there is much use in doing that). - the standards tests have been improved to test more useful aspects. Now class and function names are checked some as well as all modules have the checks for show and plural types. The error messages should explain how to fix things. - IMP::core::RigidBody::lazy_set_reference_frame() was renamed to IMP::core::RigidBody::set_reference_frame_lazy() to fit with the standards better. - the transform based methods on reference frame which have been deprecated for a while were removed - the hierarchy visiting and searching functions have been renamed to better fit the IMP naming scheme as we don't expect they were used outside of the IMP source. - IMP::atom::NonhydrogenMol2Selector was renamed IMP::atom::NonHydrogenMol2Selector to be consistent with the PDB version ## Fall 2010 - IMP::display::PlaneGeometry was added to display planes - the optimize methods on optimizers was moved to a "non-virtual interface". Usage has not changed, but optimizers should implement IMP::Optimizer::do_optimize() instead of IMP::Optimizer::optimize() to benefit from better checking. - the two-vector constructor for IMP::algebra::Plane3D now expects the second vector to be a unit vector as otherwise one could switch the order around without noticing. - the io functions in the kernel have been cleaned up a bit (the old ones are there for backwards compatibility). See IMP::write_particles(), IMP::write_particles_binary(), IMP::write_particles_to_buffer() and IMP::read_particles(), IMP::read_particles_binary(), IMP::read_particles_from_buffer() - IMP::core::WriteModelOptimizerState is going away as it duplicated IMP::WriteParticlesOptimizerState. - the IMP::algebra::Grid3D has been changed to make it support any (fixed) dimension when used with sparse storage. The template arguments have also been changed to remove a redundant one. See IMP::algebra::GridD. - the biological systems specification has been improved a little adding ability to run some rudimentary tests to make sure the scripts are ok and to not require that everything be run from the svn directory. - A new method IMP::atom::create_compatible_rigid_body() which creates a rigid body from a hierarchy which has the same internal coordinates as another rigid body. A way to make the rigid body construction more canonical would be nice (so we don't need this function), but I'm not sure how to do that. - Logging has been improved to better report the name of methods producing the log messages - a macro IMP_FUNCTION_LOG has been added which should be added to the start of non-member functions which produce significant logging output. - IMP::algebra::get_distance(const Rotation3D&,const Rotation3D&) was changed to use the Euclidean distance in R4 as that seems better behaved than the prior metric. - IMP.domino2 was renamed to IMP.domino and the old IMP.domino was moved - the documentation for classes and some methods now includes a list of all examples which use the class. Please report any incorrect links or phantom classes appearing in the docs. Unfortunately, functions have to be added individually. Please request any functions of particular interest. - applications and biological systems now have their own documentation pages - IMP::atom::Hierarchy requires radii on its leaves - IMP::atom::read_pdb add radii to the atoms - IMP::helper was merged into IMP::restrainer as all the functionality there was primarily to provide support for restrainer - IMP::domino2::SubsetEvaluatorTable was removed. Pass the model and IMP::domino2::ParticleStatesTable to the IMP::domino2::RestraintScoreSubsetFilterTable instead. - IMP::atom::Selection was added to allow specification of parts of IMP::atom::Hierarchy using biological sorts of names. A few functions to create restraints from them were added too. - a new example was added in the kernel to show IMP used on a biological system - IMP::display::PymolWriter now supports adding labels to things - IMP::display::PairRestraintGeometry and IMP::display::ConnectivityRestraintGeometry were added to aid in visualizing those restraints. - IMP::bullet was added with an optimizer to use hard body physics to resolve local collisions. - IMP::atom::setup_as_rigid_body() has been deprecated and replaced by IMP::atom::create_rigid_body(). The latter creates the rigid body as a separate particle so that the molecular hierarchy doesn't have to worry about having its bounding sphere large enough to enclose all the members. - IMP::core::KClosePairsPairScore was added to have count based cutoffs for close pair scoring. - IMP::algebra::get_surface_area_and_volume() was added to compute the two numbers for a set of balls ## Summer 2010 - The IMP::core::RigidDistancePairScore and IMP::core::ClosePairsPairScore have been revised to remove certain inconsistencies and ambiguities and to make them faster. Existing code using them will need to be slightly tweaked to work. - Javi added his 2D EM scoring functionality as IMP::em2D - some new ways to speed up scons have been added. See IMP::core::HarmonicDistancePairScore, IMP::core::SoftSpherePairScore and, for C++ users, the generic restraint support IMP::container::create_restraint() and IMP::core::create_restraint(). These each offer between 2x and 1.2x speedup when using simple scores (like harmonics on distances). In addition, a function IMP::container::get_slack_estimate() has been added which will estimate the correct slack to use with a IMP::container::ClosePairContainer. This will be extended to support IMP::core::ExcludedVolumeRestraint eventually. - support for dense and sparse grids have been added to IMP::algebra. See the IMP::algebra module page (I'm still working on how to make the docs clearer). - all write functions now have associated IMP::OptimizerState and IMP::FailureHandler classes. For examples, see IMP::WriteOptimizerState, IMP::WriteBinaryOptimizerState, IMP::WriteFailureHandler, IMP::atom::WritePDBOptimizerState etc. - functions IMP::write_binary_model(), IMP::read_binary_model() and IMP::read_configurations() have been added to support space and filesystem efficient storage of model configurations. Multiple configuration can be written to a single file. This functionality requires NetCDF be installed (NetCDF is a standard library for reading and writing binary data). - domino2 has been added. It provides a simpler interface on Keren's domino optimize as well as a number of new optimization and hooks for customizing the optimization process - a better explanation of how to use reference counted pointers has been added to IMP::Pointer. - all simple, planar polygons are now supported for display when CGAL is present ## Week of April 1, 2010 - All plural objects in Python are now simply Python lists rather than C++ recreations of Python lists. This makes compilation of the wrappers much, much, much faster. The visible changes are - types are not checked until the lists are passed to C++ - passing std::vector types by pointer and storing passed std::vector types by pointer will not work (but should never have been done anyway from code exposed to Python) - swig has issues with some overloads (not always for obvious reasons). These are being sorted out. None are commonly used. - get_particles() on a plural object doesn't work in Python (and is not needed in C++). Implicit conversions will always work. - All containers need to know the model at time of construction. If, after updating, you get an error on a container constructor, just add the model as the first argument. This also means that empty lists of particles in the constructors cause check failures. We can figure out a work around if this is too annoying. ## Week of March 13, 2010 - function get_rotation_in_radians_about_axis() was renamed to get_rotation_about_axis() as it didn't parse and we don't have the "in radians" elsewhere. - Added IMP::algebra::ReferenceFrame3D. As part of the cleanup, IMP::core::RigidBody::set_reference_frame() replaces set_transformation(). And the reference-frame related transformation construction functions were replaced. IMP::algebra::get_rotation_from_x_y_axes() was added to aid in the construction of reference frames (and replaces the algebra function taking three vectors). ## Week of March 6, 2010 - Publications now have a description field to give browsers an indication of what is in the publication to aid selection of what to read/cite. ## Week of Feb 27, 2010 - IMP::algebra::decompose_rotation_into_axis_angle() was renamed to IMP::algebra::get_angle_and_axis(). - renamed IMP::atom::bond() and IMP::atom::unbond() to IMP::atom::create_bond() and IMP::atom::destroy_bond() to be more consistent with things - renamed IMP::core::root() to IMP::core::get_root() ## Week of Feb 21, 2010 - IMP::Object::set_was_owned() was changed to IMP::Object::set_was_used() as that better reflects what the flag means. - various more interpolation and endian methods had their names fixed (by adding \c get_) - the function get_mean() on vectors was removed as it duplicated IMP::algebra::get_centroid() - build_Transformation3D_from_transformation3D was replaced by IMP.algebra.get_transformation_3d() - IMP::algebra::almost_equal() is now IMP::algebra::get_are_almost_equal() to fit the naming conventions. - the random vector generators have been cleaned up to remove redundant ones and make the names more consistent. They now have the form of either - IMP::algebra::get_random_vector_in() or IMP::algebra::get_random_vector_on() where there is one argument describing the volume or shape (such as an IMP::algebra::SphereD<4>). - or IMP::algebra::get_uniform_surface_cover() where the first argument is the primitive whose surface we are covering and the second is the number of points. Some good sed strings are - "s\#random_vector_in_unit_sphere()\#get_random_vector_in(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_in_unit_box()\#get_random_vector_in(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_on_unit_sphere()\#get_random_vector_on(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_on_unit_box()\#get_random_vector_on(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_in_box(\#get_random_vector_in(\#g" - "s\#random_vector_on_box(\#get_random_vector_on(\#g" - "s\#random_vector_in_sphere(\#get_random_vector_in(\#g" - "s\#random_vector_on_sphere(\#get_random_vector_on(\#g" - "s\#grid_cover\#get_grid_surface_cover\#g" - "s\#uniform_coverr\#get_uniform_surface_cover\#g" . - Spheres in IMP::algebra are now supported in general dimensions. See IMP::algebra::SphereD. - Many methods in IMP::algebra were renamed to be more consistent with other methods or general conventions. Most of these simply involved adding \c get_ as a prefix and/or add a \c _3d suffix when they did not take any arguments. - build_Rotation2D_from_Vector2D became IMP::algebra::get_rotation_to_x_axis() - most of the Rotation3D and Transformation3D functions (in addition, IMP::algebra::get_identity_rotation_3d() and IMP::algebra::get_identity_transformation_3d() added the \c _3d so we can add 2D versions at some point) - IMP::algebra::Transformation3D::get_transformed() and IMP::algebra::Rotation3D::get_rotated(). Replacing \c ".transform(" worked in the IMP code base. - IMP::algebra::get_distance(), this also carries over to the IMP::core::get_distance() The following global replacements are probably safe (in sed expressions) - "s\#core.distance(\#core.get_distance\#g" - "s\#algebra.distance(\#algebra.get_distance\#g" - "s\#\.transform(\#.get_transformed(\#g" Only in C++ - "s\#\.rotate(\#.get_rotated(\#g" Only in C++ - "s\#rotation_from\#get_rotation_from\#g" - "s\#rotation_in\#get_rotation_in\#g" - "s\#identity_rotation(\#get_identity_rotation_3d(" - "s\#::distance(\#::get_distance(\#g" . - IMP::algebra::Vector3D and IMP::algebra::Vector3Ds were replaced by the expanded (non-typedef) version in functions parsed by SWIG as the current situation was extremely brittle. For the moment, you should not use the typedefs in definitions that swig sees. Sorry. - IMP::algebra::Sphere3DPatch was renamed to IMP::algebra::SpherePatch3D to be consistent with the other geometry and allow us to support higher dimensions in the future. - rigid_align_first_to_second was renamed to IMP::algebra::get_transformation_aligning_first_to_second() and build_Transformation2D_from_point_sets was renamed to IMP::algebra::get_transformation_aligning_pair() ## Week of Feb 14, 2010 - The IMP::atom::Bond types were rearranged to reconcile the protein-centric classification with ligand support. - IMP::atom::Atom no longer has charge since we have IMP::atom::Charged. - IMP::atom::read_mol2() and IMP::atom::write_mol2() were added - IMP::atom::Selector and descendants renamed to IMP::atom::PDBSelector since we will soon have Mol2Selectors. In addition, the IMP::atom::IgnoreAlteriativesSelector is now IMP::atom::NonAlternativeSelector to keep the name positive, like the others. In the IMP body, all instances of the string Selector were for PDBSelectors and so sed was safe to use. - IMP::atom::read_pdb() lost the "ignore_alternatives" argument. Instead, all of the provided selectors, other than IMP::atom::AllSelector ignore alternatives internally. Code that uses the IMP::atom::AllSelector should probably change to the IMP::atom::NonAlternativeSelector. - IMP::example renamed (from examples) - The IMP_USAGE_CHECK() macro lost the (unused) third argument. ## Week of February 3, 2010 - The model io code as moved into the kernel so it is now IMP::write_model() and IMP::read_model() - Remove the version_info argument from the implementation macros. Instead, just pick up the module version info automatically. - Add a function Object::get_type_name() and Object::get_module() to get the respective bits of information about each object. - Change Object::show() to be a non-virtual wrapper which prints the object name and other info before passing control off to Object::do_show(). - The containers were moved to IMP.container. See that module page for a full list of what moved. In the process, the backwards compatibility functions were removed from IMP::container::PairsRestraint and kin. - IMP.atom.create_protein() now returns the created protein - the simplify functions have been moved from IMP.helper. They are now IMP.atom.create_simplified_along_backbone(). - IMP.helper.clone() has been renamed to IMP.atom.create_clone() - IMP.misc.ConnectingPairContainer has been moved to IMP.core.ConnectingPairContainer - The locations for pixels in IMP::em::DensityMap have been moved to the center of the corresponding voxel from the lower left corner. - IMP::em::get_transformed(), IMP::em::get_resampled() and IMP::em::get_density() have been added. - The documentation has been majorly cleaned up and simplified. Please look around and see if you like it. There are now two main pages, the introduction and the developers guide, rather than the previous plethora of pages. - IMP::algebra::Cone3D::get_bounding_sphere() has been removed since it was not of general use, not returning the bounding sphere you would expect and was trivial to implement. - IMP::algebra::projection() was removed since it didn't actually project onto the segment and the name didn't follow the normal conventions. And it wasn't used. - Most geometry now has namespace methods IMP::algebra::get_bounding_box(), IMP::algebra::get_surface_area(), IMP::algebra::get_volume(). ## Week of January 22, 2010 - A generic geometry layer has been added (see \ref genericgeometry). Using it, IMP::atom::rmsd(), IMP::algebra::alignment_taking_first_to_second() and the new IMP::atom::native_overlap() all can work on either collections of either IMP::core::XYZ particles or IMP::algebra::Vector3Ds. - Deprecated functionality has been removed from IMP.core. For those still using it, IMP::core::ClosePairsScoreState was replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairsScoreState was replaced by IMP::core::CloseBipartitePairContainer. - The macros for Scores changed slightly. Scores not in IMP svn which use the IMP_*_SCORE macros now have to implement the get_is_changed() function outside of the class declaration and scores which use the IMP_SIMPLE_*_SCORE now don't have to implement the functions to get interactions and used particles. - Rename IMP::core::MinimumPairScoreRestraint (and kin, Maximum, Singleton, Triplet and Quad) to IMP::core::MinimumPairRestraint. - Add IMP::core::MinumumPairScore (and kin). ## Week of January 1, 2010 - IMP::em is now using Particles to resample density maps and for fitting restraints. The previous ParticleAccessPoint mechanism is removed. - IMP::UsageException and IMP::InternalException are now reserved for their respective checks (and cannot be thrown by IMP_THROW()). This makes the distinction between API exceptions (which are thrown in any build mode and should be documented and tested) and helper exceptions which are only thrown when checks are turned on (and hence should not be documented) clearer. - IMP::core::RestraintSet moved to IMP::RestraintSet. The restraint sets now simply define a tree with weights which the model traverses as opposed to being an opaque IMP::Restraint themselves. This means that incremental evaluation of restraints in restraint sets now works and that restraints in restraint sets can be individually evaluated. Existing working usage should not have changed (so \c sed should manage updates just fine), but the internal implementation is completely new, so pay attention. - IMP now has a text i/o layer consisting of IMP::TextInput and IMP::TextOutput. These classes are convertible to C++ streams and implicitly constructable from both streams and file paths. As a result, functions that take them will work when called with a string for a file path, a Python file or a C++ stream. - The IMP.examples module now documents how to use the needed aspects of scons and SWIG. - The IMP.examples module is used as the basis for new modules created by the make-module script, so those include more documentation and are tested. - Module overview pages are now generated from the information passed in the SConscript file, rather than from the (now removed) modulename.dox file. See the examples file for an example of the things you can do there. - boost program options is now a checked for dependency. - The IMP::Model now uses dependency information from IMP::ScoreStates to automatically reorder the IMP::ScoreStates so that they are evaluated in the right order. Also, IMP::Model::evaluate() on a subset of the restraints now only updates the needed score states, not all of them. - Functions on rigid bodies no longer provide the IMP::core::RigidMemberRefiner as a default as it is, pretty much always, the wrong thing to use. That class may go away without notice in the future and is no longer documented. - the IMP::helper::create_simple_connectivity_on_rigid_bodies() and IMP::helper::create_simple_excluded_volume_on_rigid_bodies() now both take an IMP::Refiner as an (optional) final argument. This makes it so they can be used with the rigid bodies created from molecular hierarchies. This code still needs to be updated and the docs don't reflect what the code was or is doing, so use with caution (see notes in the header file). - The IMP::helper create* functions now all take non-Object inputs as const& to make clear what is input and that there are no memory management gotchas. - a new base class IMP::Constraint has been added for constraints. Various other classes have been renamed to be consistent. ## Week of December 11 - The interface of IMP::PairScore changed slightly to make the function signatures more uniform across various methods. Now all methods that take a pair of IMP::Particle objects take a const IMP::ParticlePair&. Previously some took that, some took an IMP::ParticlePair and some took two IMP::Particle*s. If you have an implementation of an IMP::PairScore outside of svn, you will have to update your function signatures (the compiler will tell you where). Backward compatible IMP::PairScore::evaluate() methods have been provided, so simple users of pair scores shouldn't see a difference. - IMP::core::PairRestraint takes a IMP::ParticlePair in the constructor instead of two IMP::Particle*s. A backwards compatible constructor can be added if needed. - IMP::ParticlePair::first and second went away since no one claimed to be using it and it complicated the implementation. - The generic restraints/score states/scores infrastructure has been extended to triplets and quads (for angles and torsion angles). ## Week of December 4 - IMP now makes use of Boost.Filesystem to manipulate paths. Things may not work well on systems where boost libs are not available which does not support posix paths. But libless boost installs generally don't exist outside of our test systems. - new functions like IMP::core::get_example_path(filename) and IMP::core::get_data_path(filename) allow one to easily find a data file named 'filename' for the examples and library internal data. - The IMP::display module has been revamped in order to support non-linear geometry. VRML support has been dropped. CMM support will be dropped unless someone requests it (as IMP::display::ChimeraWriter does all it does and more). ## Week of November 19 - imp.xcodeproj was moved into tools and updated accordingly - the bin directory finally went away - IMP::display::ChimeraWriter now supports the latest version of Chimera - IMP::display::CGOWriter was renamed to IMP::display::PymolWriter as it can do more than just CGO. - IMP::VectorOfRefCounted::operator[] can now be used for assignments just like a normal vector. No more need to use the IMP::VectorOfRefCounted::set() method. - the IMP::atom::Hierarchy constructor from an IMP::Particle* is now explicit (as were the decorator constructors) ## Week of November 12 - All exception types are now mapped to equivalent classes in Python. The class hierarchy is similar (e.g. IMP::ModelException derives from IMP::Exception in both C++ and Python) so all IMP exceptions can be caught in Python using "except IMP.Exception". For convenience, the IMP::IOException, IMP::ValueException and IMP::IndexException classes in Python also derive from the similarly-named standard Python classes (IOError, ValueError and IndexError respectively). ## week of October 29 Update - You can now add members to rigid bodies via IMP::core::RigidBody::add_member(). These members can be other rigid bodies (although this functionality is only lightly tested). - The code to get bounding volumes for IMP::atom::Hierarchy has been moved to the atom module, as as the clone and destroy code. They can be found at IMP::atom::clone(), IMP::atom::destroy(), IMP::atom::get_bounding_box(). - Added IMP::core::VolumeRestraint. - Several functions have been added to helper in creating restraints. They are create_simple_connectivity_on_rigid_bodies(), create_simple_connectivity_on_molecules(), create_simple_distance(), create_simple_diameter(), create_simple_em_fit(), and create_simple_excluded_volume_on_rigid_bodies(). Functions load_em_density_map() and set_rigid_bodies() are added as well. See IMP::helper for more information. - IMP::ScoreStates are further hidden. Namely, IMP::core::RigidBody constructs the needed score states internally, doing away with the need for the create functions (which have been done away with). In addition there are new decorators IMP::core::Cover and IMP::core::Centroid which manage their respective score states, further doing away with more create methods. The hierarchy-based create method for rigid bodies has been moved to the atom module and is called IMP::atom::rigid_body_setup_hierarchy(). - IMP::em::read_map() will figure out the file type to use automatically from the extension. Other such methods should follow suite. - A class IMP::core::StructureSet has been added to provide a convenient way to manage solutions of optimization runs. - IMP::core::RigidBodyMover now takes its angle argument in radians and is documented as such. - IMP::core::ClosePairsScoreState and IMP::core::CloseBipartitePairsScoreState have been replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairContainer. The score states have been marked as deprecated and are mostly still there, but are more limited than they were. - IMP::core::AllPairsPairContainer has been renamed to IMP::core::AllPairContainer and a bipartite version has been added. - All containers now keep track of what they looked like on the previous time step. - The IMP::core::ListPairContainer and IMP::core::ListSingletonContainer have had their interface reduced a bit in order to make implementing change tracking easier. - The set of exceptions has been simplified to go along with simplification of the checks. We now have IMP::InternalException and IMP::UsageException which are thrown by the respective checks (and the former is thrown by IMP_FAILURE()). When you want to to pass an exception back to Python use one of IMP::ValueException or IMP::IndexException or IMP::IOException. There is also a new macro, IMP_THROW() to be used when you want to throw an exception with a nice error message. Please look at code you have worked on and make sure that the exceptions make sense given the new clarifications. - An IMP::ModelException is now thrown when IMP::Particle attributes attain invalid values. This exception can be caught to recover from failures of IMP::core::ConjugateGradients or difficult restraints. ## Week of October 15 - The set of functions needed to get meta information about IMP::Restraint and IMP::ScoreState objects has hopefully stabilized. Sorry about the thrashing. If you have restraints or score states not in SVN, see the corresponding IMP_RESTRAINT type macros, the IMP::Interaction class and base classes for what exactly you need to implement. ## Week of October 8 - IMP::Interaction was modified to split IMP::interaction::get_used_particles() into IMP::interaction::get_read_particles() and IMP::interaction::get_write_particles(). This makes it possible to check for IMP::ScoreState objects being in an inconsistent order (so that the input to one is updated after it is read). Later, we will be able to automatically order the score states ensure that there are no problems. - IMP::atom::Hierarchy has been modifier to get rid of the IMP::atom::Hierarchy::Type. Instead, all types are contained in the decorators. IMP::atom::get_by_type() was also updated to reflect this. - IMP_check() and IMP_assert() have been replaced with IMP_USAGE_CHECK() and IMP_INTERNAL_CHECK() respectively. Hopefully this makes it clearer when to use which. IMP::set_check_level() has been revised accordingly (to use IMP::USAGE and IMP::USAGE_AND_INTERNAL). Also, IMP_failure() and IMP_not_implemented() have been renamed to IMP_FAILURE() and IMP_NOT_IMPLEMENTED() to make all macros be all caps. To update external code, - replace IMP.EXPENSIVE with IMP.USAGE_AND_INTERNAL - replace IMP.CHEAP with IMP.USAGE - replace IMP_assert and IMP_check as above. ## Week of October 1 - IMP::ScoreState and IMP::Restraint now better track the set of particles that they use and the dependencies between the particles. To this end, a new base class IMP::Interaction has been added with methods to get the inter-particle interactions and the set of particles used. - .cpp, .h, .py and SConscript files which contain the string "DO NOT COMMIT" will be rejected on checkin. Use this to make experimental or debugging changes you are making so you don't accidentally commit them. - The search module is no longer built by default and is not longer supported. It will get moved out of svn shortly. - IMP::ScoreState objects now also have IMP::Interacting::get_interacting_particles() methods. Soon IMP::PairScore and IMP::SingletonScore objects will too. ## Week of September 27 - Some support for gather statistics has been added to IMP::Model (see IMP::Model::set_gather_statistics()). If requested, an API can be provided so that the individual data points can be returned to programs. - A class IMP::core::RigidBodyDistancePairScore has been added which provides accelerated computations for the distance between two rigid bodies. - The SWIG input files were renamed from swig.i to swig.i-in to aid in dependency tracking. - rigid bodies now have two sets of particles associated with them in the various places they are used. One is the IMP::core::RigidBodyMember particles, as was previously the case. The other is defined through a refiner (and could be the RigidBodyMember particles if a IMP::core::RigidMembersRefiner is used). The idea is that you sometimes want some particles which are transformed along with the rigid body, but which are not really part of the representation of the rigid body. Examples would include marker points or multiresolution representations of proteins. ## Week of September 20 - all \c .i files have been renamed to start with \c IMP_ to help make dependency tracking easier. - static builds (builds with \c static=True) build both the static and dynamic libraries and just static executables - the \c profile build target has been removed as \c static=True does the important bit of what is needed. Just do \c linkflags=['-pg'] to add the needed link flag for \c gprof. - Restraints should now implement IMP::Restraint::unprotected_evaluate() instead of IMP::Restraint::evaluate(). The latter can now be called outside of IMP::Model::evaluate() and will return the right answer always. - \c bin/imppy.sh no longer works, as was previously warned. Use \c tools/imppy.sh instead. The bin directory will go away eventually, for now the script just tells you to use the tools version. ## Week of August 30 - A new score IMP::core::BoundingBox3DSingletonScore has been added to keep particles in a box. - str should now work better in Python when applied to IMP::Object objects and a few others. Objects should use the new IMP_SHOWABLE and IMP_SHOWABLE_INLINE macros instead of declaring show functions directory to make it so that they work in Python. It is now a bug to \c print or \c str to not return something useful in Python for any IMP object. ## Week of August 23 - each module now has a version, eg IMP::get_module_version_info() which is also returned by the various restraints of the module. The Python code checks that loaded libraries have the expected version numbers to try to provide better messages when runtime linking problems occur. - the various bounding_box methods have been renamed to ::get_bounding_box() to be more consistent with the other names. - Python documentation is now parsed by doxygen ## Week of August 15 - The Object::get_version_info() result now includes the SVN revision info when building from SVN. This means you can use it to more easily tell exactly what a built version of IMP was built from. ## Week of August 8 - Treatment of hetatoms in hierarchies (and hence pdbs) is now documented. See IMP::atom::Hierarchy. This involved broadening IMP::atom::Residue to handle ligands and having IMP::atom::Hierarchy::LIGAND, IMP::atom::Hierarchy::AMINOACID and IMP::atom::Hierarchy::NUCLEICACID be handled by IMP::atom::Residue. - IMP::atom::read_pdb() currently drops HOH ATOM records on the floor. We should add a IMP::atom::Water decorator if anyone wants to handle them (since they aren't atoms and don't have elements). - IMP::atom::Mass() has been added for decorating particles with mass. IMP::atom::Atom mass has been moved to that. - The scheme for generating atom types is now fully specified. See the IMP::atom::AtomType documentation. - IMP::atom::Atom and IMP::atom::Residue have been somewhat restructured. Some of the helper functions have been renamed to be more descriptive or removed (if they duplicated other easily accessible functionality). get_residue_index() has been removed (since you can just call get_index() on the return result of the IMP::atom::get_residue()). IMP::atom::get_chain() now returns an IMP::atom::Chain. root() was renamed to IMP::atom::get_root() for consistency. - SWIG files are now generated from a file called swig.i in the pyext directory. This file only has to contain module-specific commands such as lines to wrap headers and instantiate module templates. No more boilerplate. - The contents of bin are moving to tools since the two directories both contained assorted development tools (not all of which were executable). For now, bin/imppy.sh will still work, but that will go away soon. - IMP::Decorator::show() now just takes the stream (it used to take a prefix also, which was often ignored). Anyone who implemented a decorator will have to delete the second argument of their show function. ## Week of August 1 - Containers now have methods IMP::SingletonContainer::evaluate(), IMP::PairContainer::evaluate(), IMP::SingletonContainer::apply() and IMP::PairContainer::apply(). Using these can be significantly faster than looping over the particles externally. Similarly, there is IMP::SingletonScore::evaluate(), IMP::PairScore::evaluate(), IMP::SingletonModifier::apply(), IMP::PairModifier::apply(). - The decorator functions have been renamed to IMP::Decorator::setup_particle(), IMP::Decorator::decorate_particle() and IMP::Decorator::particle_is_instance(). A shell script \c tools/update-decorators is provided which updates the names in all code in subdirectories of the current directory. Please backup things first. - a function, IMP::Restraint::safe_evaluate() has been added to be used when requesting the value of restraints outside of the IMP::Model::evaluate(). ## Week of July 18 - The atom::BondPairContainer has been renamed to atom::BondPairFilter since it wasn't a real container. A real atom::BondPairContainer has been added. - Out of source builds mostly works. To do an out of source build, make a link to \c scons_tools and \c SConstruct from the source directory in the build directory and then add the line \command{repository="../path_to_src_dir"} in the \c config.py in the build directory. Then running \c scons in the build directory will put generated files there. This allows you to make both a \c release and \c fast build off of the same set of sources. Proper docs will be added soon. - The SConscript setup for a module has been greatly simplified so it should be easy enough for users to maintain their own module and updates to the build system shouldn't involve changes to user SConscripts any more. This does mean that any non-svn modules need to be updated. Look at IMP.core or another module and copy what is there. Again, proper docs will be added soon. ## Week of July 11 - Per Javi's suggestion, the conventions have been split into "Coding conventions" and "API conventions" and linked from the main page. - benchmarks got moved to their own module, IMP::benchmark and will now be tracked over time ## Week of July 4 - FilteredListSingletonContainer and FilteredPairSingletonContainer went away as there was too much hard to factor code shared with IMP::core::ListSingletonContainer and IMP::core::PairSingletonContainer. The IMP::core::ClosePairsScoreState is now handled by a list of filters on the class itself. That is, do IMP::core::ClosePairsScoreState::add_close_pair_filter() to add a filter. - Hooks have been added to allow actions to be taken on an assertion or check failure. See IMP::HandleFailure and its descendants for more details. - IMP::display::CGOWriter has been much improved and now allows provides lots of display-time control and easy building of animations. - Generic optimizer states have been added (IMP::core::SingletonsOptimizerState and IMP::core::PairsOptimizerState). ## Week of June 15 - box distributions (IMP::algebra::random_vector_on_box(), IMP::algebra::random_vector_in_box()) take bounding boxes - IMP::em::DensityMap are now IMP::Object objects. Some current allocations on the stack may need to be changed (since ref counted objects can't be stack allocated). - IMP::em::DensityMap::Read() is now deprecated (as is IMP::em::DensityMap::Write) as they violated a slew of conventions. Use IMP::em::read_map() and IMP::em::write_map() instead. ## Week of June 8 - new methods IMP::atom::add_radii() and IMP::atom::add_bonds(); - Hierarchys to IMP::atom::Hierarchies - IMP::Particles has been changed to a new type. It no longer plays quite so well with Python lists (you can't create one from an arbitrary Python list). - The containers for multiple decorators have been revised to make them more useful. See IMP::Decorators. You can now implicitly convert, for %example, a IMP::core::XYZs to an IMP::Particles (as well as use IMP::Decorators::get_particles()). All methods that take a collection of particles with a certain type now instead take a IMP::Decorators. - A method that has expectations about the particles it takes should take decorators rather than particles. That is, anything which expects IMP::core::XYZ particles, should take an IMP::core::XYZs instead of IMP::Particles. - IMP_LIST() now takes an extra argument (the name of the plural container) - In C++ IMP::Decorator objects are implicitly convertible to IMP::Particle* objects. Unfortunately, making this work in Python is problematic. - More methods take and return decorators rather than IMP::Particles. This means it is clearer what their expectations are and what they produce. ## Week of May 15 - The IMP::helper module was added and various functionality moved there. See its documentation page for a list of what is there. - IMP::algebra::SphericalCoords was renamed to IMP::algebra::SphericalVector3D to be consistent with other algebra classes. - IMP::algebra::Rotation2D is now IMP::InvalidDefault to be consistent with the other similar classes. - We are trying to be a bit more specific about what is expected from a IMP::atom::Hierarchy. Towards that end, there are two methods for extracting a representation from a hierarchy, IMP::atom::get_detailed_representation(), and IMP::atom::get_simplified_representation() - A method has been added to produce a simplified copy of an IMP::atom::Hierarchy. The method is still a bit crude and could benefit from other people inspecting the output and saying how it should be improved. - There is now an IMP::algebra::BoundingBoxD (and IMP::algebra::BoundingBox3D) to track bounding boxes. There are methods like IMP::algebra::bound_box(Sphere3D) which gets the bounding box of a sphere. - Each IMP::Object has its own log level. If IMP::Object::set_log_level() is called with a value other than IMP::DEFAULT, the value is used for code within and called by the object. To support this in your objects you need to add IMP_OBJECT_LOG to the start of each major method. It has been added to some of the core objects. ## Week of May 8 - IMP::atom::MolecularHierarchy has been renamed to IMP::atom::Hierarchy. Use \command{sed -i.old "s\#MolecularHierarchy\#Hierarchy\#g" my_files} - the long-deprecated Name(Decorator) was removed - The decorator suffix has been removed from decorator names. - Unit conventions have been established and documented for distance, energy, force and angle. - there is a function IMP::atom::create_protein() to create a coarse grained protein. The function is crude so far, but the idea is that we should have standard ways of generating such things so that everyone can benefit from best practices. - a class IMP::atom::BondGraph has been added so Boost.Graph algorithms can be used with bond graphs - the IMP-specific doxygen commands have been documented. - the doxygen commands unstable{name} and untested{name} have been added to doxygen to be used for marking methods and functions as likely to change or not well tested. Pass the name of the class or method as name. - bonds can now be added to molecular hierarchies automatically see IMP::atom::add_bonds() ## Week of May 1 - There is a IMP::Rotation2D - You can now extract fixed Euler ZYZ angles from a IMP::Rotation3D. This code can serve as a template for extracting other Euler angle conventions if they are needed. - IMP::algebra::rotation_between_two_vectors() changed to IMP::algebra::rotation_taking_first_to_second() to make it clear which direction the rotation goes. ## Week of April 27 - IMP::Decorator derived objects are now convertible to bool. So code like \code if (d) { blah } \endcode works. - IMP::Decorator::cast now returns a null decorator (\c DecoratorName()) rather than throwing an exception. This might make it actually useful. - New IMP::statistics module for statistical methods (e.g. clustering). ## Week of April 20 - Support for taking advantage of rigid bodies in collision detection has been added via the IMP::core::RigidClosePairsFinder. - All IMP::core::ClosePairsScoreState-related functionality has been optimized and should be noticeably faster. - The EMBED project has been merged into the IMP::em module. Thus, it is no longer necessary to obtain EMBED separately to build IMP::em. - All C++ methods which work with std::istream or std::ostream arguments should now be usable from Python: any Python file-like object can be used. - The SWIG interface now ties in to the reference counting mechanism used in the C++ code, for all classes derived from IMP::RefCounted. So, for %example, deleting a Python object should not free the underlying C++ object if another C++ object holds a reference to it, and unrefing a C++ object should not delete it if a Python object still holds a reference to it. There should be no further need to manually keep references to Python objects to prevent things from being prematurely cleaned up. This should be transparent to the user; please report any problems. - IMP::atom::AtomDecorator no longer inherits from IMP::core::XYZDecorator. ## Week of April 13 - The decorators IMP::atom::AtomDecorator, IMP::atom::ChainDecorator, IMP::atom::ResidueDecorator etc. now inherit from IMP::atom::MolecularHierarchyDecorator. This means they can set their IMP::atom::MolecularHierarchyDecorator::Type properly and saves a lot of typing in the common case. We can always make hierarchy decorators zero memory overhead if we find that the cost matters. - The base classes IMP::Comparable, IMP::NullDefault, IMP::ValidDefault and IMP::UninitializedDefault were added to document what the default constructors do for various objects in IMP. ## Week of April 6 - IMP::Object now defines pure virtual methods IMP::Object::get_version_info() and IMP::Object::show(). All non-abstract classes which inherit from IMP::Object must implement them. Doing this means they are more uniformly implemented and are documented in one place. - IMP::Object is now reference counted (the only objects this changes are IMP::Model and IMP::Optimizer). As a result optimizers and IMP::Model objects must be created on the heap. - IMP::RefCounted has had its name changed and no longer inherits from IMP::Object. This makes it easier to use independently. - the docs for reference counting (IMP::RefCounted) and and IMP::Decorator have been improved - running doxygen produces no warnings. Let's try to keep it that way. Run \command{scons doc > /dev/null} before you check in large blocks of new code to make sure that you didn't introduce warnings. - Helper classes have been added to make SingletonModifier and PairModifier objects look like functors for C++ and Python. This makes it so you can use \c map in Python and \c for_each in C++. See IMP::SingletonFunctor and IMP::PairFunctor for more information. - The recent changes page has moved to a doxygen page to make it easier to link changes back to the relevant source code. ## Week of March 20: - create_rigid_body(), create_rigid_bodies(), create_centroid() etc. now require you to add the score state to the model yourself. Whether the rigid body is enforced or not can be controlled by adding or removing the returned IMP::ScoreState. - All previously deprecated code has been removed. This includes indexes in particles, the no-argument constructor for particles and the functionality that was moved from core to atom. - The IMP::core::RigidBodyDecorator and IMP::core::RigidMemberDecorator have been simplified by removing the traits classes. There is now only one type of rigid body and it stores the list of member particles internally. - IMP::core::NameDecorator has been deprecated as IMP::Particles have built in names (use IMP::Particle::get_name(), IMP::Particle::set_name()). - IMP::atom::ChainDecorator has been added ## Week of March 13: - A new IMP::display::Writer, IMP::display::ChimeraWriter has been added which exports geometry to native Chimera constructs. It probably should be used instead of bild and cmm for most purposes as it combines the best of both types. - compose() for transformation is now a free function so compose(rotation, vector) will work via promotions. - the geometry classes have been cleaned up and are now more consistent with IMP standards (and more likely to do what their names/comments suggest). See IMP::algebra::Sphere3D, IMP::algebra::Cylinder3D, IMP::algebra::Segment3D, IMP::algebra::Plane3D etc. - ParticleRefiner has been renamed to IMP::Refiner since the Particle was long and unnecessary. ˘IF3 translation initiation complex](http://salilab.org/40S-eIF1-eIF3). - IMP::domino::DiscreteSampler::set_restraints() now allows the set of restraints used in DOMINO filtering to be controlled (previously, all restraints in the IMP::Model were used). - IMP::modeller::IMPRestraints now transfers an IMP::ScoringFunction into the MODELLER score, not the implicit (and deprecated) score of the entire IMP::Model. - Bugfix #775: coordinates of voxels in EM maps were previously inconsistently treated (they are supposed to be the centers of the voxels, but in some cases they were the lower left edge). The fix may result in maps and their bounding boxes appearing shifted by half the voxel size. # 2.4.0 - 2015-02-27 # {#changelog_2_4_0} - Python 3 is now supported. The Windows package now works out of the box with Python 3.3 or 3.4; a `-python3` subpackage is provided for Ubuntu and Fedora. (Apple does not yet provide Python 3 so the Mac package is not built with Python 3 support, but if you install Python 3 you can compile IMP from source to work with it.) - A new command line tool, `rrt_sample`, allows for running rapidly exploring random tree (RRT) based sampling, e.g. to sample conformations of multiple domains connected by flexible linkers. - Command line tools are now placed in the `bin` directory of modules (the old `bin` directory, used for utility programs that are not installed, is now called `utility`). The `applications` directory is removed. - Several Python modules that were only used by EMageFit have been moved from IMP::em2d to IMP::EMageFit. - A new tutorial that uses the IMP::pmi module is now available (modeling of the RNA Polymerase II stalk). # 2.3.1 - 2014-12-01 # {#changelog_2_3_1} - Minor bug fixes. - Update to IMP::pmi to support the Nup84 and rnapolii biological systems. # 2.3.0 - 2014-10-24 # {#changelog_2_3_0} - A new IMP::pmi module was added, containing high-level Python classes to simplify the representation, scoring, sampling and analyzing of models in IMP. This module is still under heavy development, and should be considered experimental. - Binary installers for 64-bit Windows are now available. - The `.py` extension has been removed from all Python applications, so that they are consistent with the C++ applications. - IMP::atom::Selection objects can now be combined using basic set operations (union, intersection, difference, symmetric difference). In Python the equivalent operators (|, &, -, ^ respectively) can be used. - The integrative docking (idock) application now employs SOAP to improve the scoring. - IMP::atom::CHARMMParameters::create_topology() now works with a wider range of hierarchies (previously it would only accept Residues that were direct children of Chains). For example, it should do the right thing when given a Hierarchy that contains no Chain particles, or one where Residues are grouped under Fragments. - Only Python 2.6 or later is now supported. We still provide IMP for RedHat Enterprise 5 (which ships with Python 2.4) but you will need to use the `python26` binary (part of EPEL) rather than the system default Python. # 2.2.1 - 2014-08-11 # {#changelog_2_2_1} - Bugfix: PoseScore (see the ligand_score application) should now match that used by the Pose&Rank web server (http://salilab.org/poseandrank/) # 2.2.0 - 2014-04-09 # {#changelog_2_2_0} - A module IMP.template was added to act as a template for creating new external modules. It includes setting up git hooks for format checks and things like that. - The Python support for grids was cleaned up. Hopefully this makes it more robust. If you were using a grid in Python that no longer exists, ping us, we can add it back. - The generic geometry functions were simplified slightly and renamed to IMP::algebra::get_vector_geometry(). - The IMP::base::Flag class was added to simplify adding flags in C++. - The `--log_level` and `--check_level` command line argument now take strings (eg VERBOSE) to specify their values, rather than cryptic (and a bit unstable) numbers. - Command line options are now divided into normal and advanced ones. You do `--help_advanced` to show the advanced ones. - Support for computing Connolly surfaces was added with IMP::algebra::get_connolly_surface(). In addition, there is now IMP::algebra::get_uniform_surface_cover(const IMP::algebra::Sphere3Ds&,double) to sample the surface of a collection of balls. - IMP::atom::reate_simplified_from_volume() was added to create a simplified representation of a hierarchy that preserves the surface are and occupied volume. It uses IMP::algebra::get_simplified_from_volume(). If you want to play with such representations, you can use the rmf_simply application. - IMP::atom::get_rmsd(IMP::atom::Selection, IMP::atom::Selection, const IMP::algebra::Transformation3D&) was deprecated. Use IMP::atom::get_rmsd_transforming_first() instead. - IMP::algebra::get_rmsd() and IMP::algebra::get_rmsd_transforming_first() were added. - The family of macros IMP_BASE_[SMALL,LARGE]_[UNORDERED,ORDERED]_[MAP,SET] have been added (eg IMP_BASE_SMALL_ORDERED_SET ()). They provide platform and boost-version independent ways to take advantage of the best set/map for an application. In particular, they use boost::container::flat_set and kin when available. - IMP::atom::State and IMP::atom::Representation decorators have been added to provide support for multiple states and multiple representation schemes. They are integrated with RMFs and IMP::atom::Selection. See atom/multiresolution.py and atom/multistate.py for examples. - The container implementations has been cleaned up and simplified a bit. This should not change most usage. - [RMF](http://salilab.github.io/rmf) has been updated. There are a few minor API changes, in addition to a new, faster format (`.rmf3`). - [Eigen](#Eigen) is now included with IMP. - The deprecated function IMP::kernel::Optimizer::set_restraints() was removed. Use IMP::kernel::Optimizer::set_scoring_function() instead. - The deprecated constructor for IMP::kernel::ModelObject that does not take an IMP::kernel::Model was removed. This propagates to constructors IMP::kernel::Restraint, IMP::kernel::ScoreState and kin. Make sure you pass an IMP::kernel::Model as the first constructor argument. Related to this, deprecated functions like get_is_part_of_model(), set_model() were removed as they don't serve any purpose. - The deprecated methods in IMP::kernel::Optimizer that pertain to attribute optimization were removed. Inherit from IMP::kernel::AttributeOptimizer instead if you want to use them in your optimizer. - Some deprecated functionality was removed. If you encounter problems, replace IMP::base::OwnerPointer with IMP::base::PointerMember, find logging stuff in IMP/base/log.h, object in IMP/base/Object.h and random in IMP/base/random.h. - IMP::atom::Chain now uses strings for the chain ids, to be consistent with changes in the PDB. PDB files are written using the first character of the chain string. - Added orientation-dependent SOAP scoring (IMP::score_functor::OrientedSoap and IMP::atom::OrientedSoapPairScore). Orientation-dependent SOAP scores include those that score loops (SOAP-Loop), protein-peptide interfaces (SOAP-Peptide) and proteins (SOAP-Protein). The library files themselves are rather large (~1.5GB) and so are not included with IMP. They can be downloaded separately from http://salilab.org/SOAP/. # 2.1.1 - 2013-11-04 # {#changelog_2_1_1} - Added support for Mac OS 10.9 # 2.1.0 - 2013-10-24 # {#changelog_2_1_0} - There is now some limited IMP-wide gathering of timing statistics. You can add `--statistics=filename` to the command line or use the IMP::base::show_timings() function to view a summary of where IMP has spent time. - Slack was being used inconsistently (and not always correctly) among the various containers and restraints. If you had heavily optimized the slack value for speed, you may want to revisit it. - Class name OwnerPointer was switched to PointerMember, to denote more clearly a ref-counting pointer that is a class member. - The dependency handling in IMP has been rewritten to simplify it and fix a bunch of bugs. There are now more checks of correctness, so behavior that might have silently worked (or not worked) before may now trigger errors. In particular, restraints/score states must have access to the IMP::kernel::Model to do much of anything, so either use constructors that take the model or call IMP::kernel::ModelObject::set_model(). Also, changing your set of inputs or outputs without calling IMP::kernel::ModelObject::set_has_dependencies(false) is now likely to trigger a check failure (it would just silently do the wrong thing before). - Decorators have been cleaned up to make them behave consistently regarding IMP::kernel::Particle vs IMP::kernel::ParticleIndexes. This change shouldn't break existing code, but decorators should move to the new macros IMP_DECORATOR_METHODS(), IMP_DECORATOR_SETUP_0() etc. - The IMP_GRAPH() and IMP_WEIGHTED_GRAPH() macros got an extra argument describing how to show the vertexes in the graph. They also now expose a method, eg IMP::kernel::show_as_graphviz() that writes the graph in graphviz format to an IMP::base::TextOutput. - Support and rules for deprecating code in IMP have been updated. There are new macros and instructions on how to use them that should result in much better visibility about changes. See the C++ wiki for IMP for more information. And there is now an encoded policy of keeping deprecated things around for 1 release after deprecation. - IMP is now much more selective about when it invalidates the dependency graph. See [Dependencies](http://github.com/salilab/imp/wiki/Dependencies) for more information. This change should make evaluation of isolated restraints much faster in certain situations where they were unexpectedly slow before. - IMP::atom::Selection has been completely rewritten as it was very brittle. The set of particles returned may change in some cases, in some cases because it was buggy before, in some cases because of new bugs. - Reminder: the IMP_PROTECTED macros didn't work out well, and are now deprecated. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading `_` on the method name, if you get `Swig director pure virtual method called IMP::kernel::PairScore::do_get_inputs` messages. - The generated `cmake` files are no longer stored in the repository. Instead `tools/build/setup_cmake.py` is run after checkouts and other git events that change the repository tree (assuming you have run `setup_git.py`). - Generation of documentation using `doxygen` has been refactored. It is now done on a per-module basis, via targets like `IMP.core-doc`. The main advantage is that changes to the doc for a module can be tested comparatively quickly, just by building that module's docs. On the down side, docs are no longer quite so heavily interconnected, so dependencies only work to things that the module actually depends on. - The restraint and score state statistics functions were removed from IMP::Model. They had been broken for quite a while, with no one complaining. Statistics can be added back easily if requested. - Added an IMP-specific fork of `git-flow`. It handles things like `README.md` files when you start and finish feature branches and provides a tool to nicely update IMP and display the change log. The main commands of use are `git imp feature start <feature_name>`, `git imp feature finish <feature_name>`, and `git imp update`. You may want to remove the `git-flow` lines from your `.git/config` file as you may accidentally type `git flow` instead of `git imp`. You need to run `setup_git.py` to set things up. - [RMF](http://salilab.github.com/rmf) is now included as a git submodule rather than copied into the IMP repository. You should rerun `setup_git.py` to make sure submodule stuff is initialized. # For IMP 2.0 and earlier # {#changelog_2_0} ## Spring 2013 - Modules and applications now have a _Maintainer_ in their `README.md`, giving the github id of the person to whom issues should be assigned. - `setup_git.py` supports the `--module` flag which will set up git hooks and things for a git repository that contains a module. - To clean up for the release, the module dependencies are now described in a file called `dependencies.py` and the module docs in a `README.md`, both in the main module directory. - IMP-wide support for progress bars was added with the IMP::base::set_progress_display() function and helpers. These are only displayed when the IMP::base::get_log_level() is IMP::base::PROGRESS. ## Winter 2012-2013 - Scons support was removed. Use cmake to build IMP, it is way better. - Examples are run with a `--test` argument when IMP is built with internal checks. Expensive examples should check this flag and shorten what they do in order to try to test all their code without taking too long. - The class IMP::core::NonRigidMember was added to allow particles that are not rigid but whose coordinates are specified via internal coordinates in some IMP::algebra::Reference3D that is part of an IMP::core::RigidBody. - The IMP_BIG_ENDIAN and IMP_LITTLE_ENDIAN preprocessor symbols are deprecated in favor of the BOOST_BIG_ENDIAN and BOOST_LITTLE_ENDIAN symbols. - Support for asymmetric moves was added to IMP::core::MonteCarlo. To do this, IMP::core::MonteCarloMover was added and IMP::core::Mover and IMP::core::MoverBase were deprecated and implemented in terms of the new class. This should require few or no changes in code that uses Monte Carlo support or implements movers (although movers should move to IMP::core::MonteCarlo). All IMP::core::MonteCarloMover objects keep track of statistics (number of moves and number of accepted moves), rather than that being done in the IMP::core::SerialMover. In addition, the poorly defined move probability was removed from IMP::core::MonteCarlo. - IMP::base::WarningContext now outputs warnings immediately. This change was needed as the previous behavior resulted in occasional core dumps when the libraries for log support were unloaded before the libraries where static IMP::base::WarningContexts were declared. And there was no easy way to detect such problems. - You can now (probably) build IMP with [CMake on IMP wiki](https://github.com/salilab/imp/wiki/Cmake). - The maximum logging level and check level can now be controlled independently of the build mode. To, for example, enable progress level logging in a fast build, add maxlog="PROGRESS" to your config. The maximum check level for `release` has been reduced to IMP::USAGE as that speeds things up a bit. - The way applications are structured and specified has been regularized and simplified. Python files are automatically copied/installed and binary executables are compiled from each found `.cpp` file linked with any `.cpp` files contained in a subdirectory called `lib`. Nothing needs to be put into the SConscript file. - The Python constants like IMP.algebra.has_cgal went away as swig autogenerates ones like IMP.algebra.IMP_ALGEBRA_HAS_IMP_CGAL. While the latter are a bit uglier, the constants aren't used much and having them be auto generated is nice. - There have been some major changes to the build system. The main one is that the SConscripts in the directories below the module directory aren't really used. They are being removed from the repository. Now everything really is off of file locations (eg all test_* in test is a test, each *.cpp in bin becomes a program etc). This greatly simplifies the build system, which has, historically been a disaster. You can now define dependencies by dropping files into a dependency subdirectory in your module, no need to mess with the SConscript. - Building IMP now always requires specifying a `repository` argument. This was always required for out of source builds, which is what one should be doing, but is now required for in source builds to, so as to make accidentally polluting your source with an in-source build harder. Use `repository="."` in that case. - The HDF5 functionality in RMF was reorganized. Specifically, the headers were moved into `RMF/HDF5`, the Python functionality into `RMF_HDF5` and the code into namespace `RMF::HDF5` and all mentions of `HDF5` or `hdf5` in the class and function names were removed. - Support for IMP-wide command line flags have been added to IMP::base in the base/flags.h file. They provide a simple command line parsing interface that can be used from C++ and Python. In addition, various functionality in IMP is automatically exposed on the command line when those functions are used. For example, one can control logging, checks or profiling. See IMP/base/flags.h for (some) more info. - OpenMP support in IMP is officially released. All tests pass. But there has been little specific testing so far. - Access to attributes using IMP::ParticleIndex and the IMP::Model now works from Python. See IMP::Model::add_attribute() and kin. - IMP::ParticleKey and IMP::ParticlesKey were renamed to IMP::ParticleIndexKey and IMP::ParticleIndexesKey respectively as that is what they store. Typedefs were provided to support old code. - Indexes have been added to the docs connecting classes with all the methods that return them, examples that use them and methods that take them. And all methods that are used in examples with the examples they are used in. Please check them out and report issues and ways to improve things. You can find links on the main page. - The docs for a modules/biological systems/applications should now be found in an `overview.dox` in the module `doc` folder. Modules in IMP svn have been updated. For external modules, you can just move a `overview.dox` from that was previously generated into your `doc` folder and (eventually) remove the lines from your `doc/SConscript`. This makes it easier to use the full range of doxygen markup, as well as removes the need to escape doxygen markup for Python. - Examples have now been simplified into full blown doxygen examples. This involves merging the `.py` and `.readme` files (putting `\#\#` in front of each line of the `.readme`) and adding a `## \\example module/examplename.py` line at the beginning. You can use `./tools/updaters/update_examples.py modules/mymodule` to update a module that was not already updated. - IMP::base::FailureHandler has been removed as it wasn't used. - Handling of data and includes has been simplified. All appropriate files in the corresponding directories are linked into build/installed. This is only likely to be a problem if you had broken headers sitting around `include`. If so, just comment them out, or move them some place nicer. - Modules now must be in a folder called `modules`, which must have a `description.py` file. - The build system should now be a little better about cleaning up after itself. - How headers/doc files/data/examples - The IMP_PROTECTED macros didn't work out well, and are now deprecated. Sorry. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading "_" from the method name. - Nice version numbers will now be generated if you use git. You need to rerun tools/git/developer_tools/init_git if you have an existing clone. ## Fall 2012 - 11/9: RMF is now an external library shipped with IMP. IMP will build it and install it automatically if you have cmake version 2.8 or higher installed. - 11/9: Updated support macros for IMP::Restraint and IMP::ScoreState classes have been added (IMP_RESTRAINT_2() and IMP_SCORE_STATE_2()) that make it easier to return dependencies on things other than IMP::Particle and IMP::ScoreState objects. In addition, the dependency functions on scores and modifiers have been updated similarly and standardized. In general this should not cause backward compatibility problems unless you were overriding the IMP::ModelObject::do_get_inputs() in IMP::Restraint and using the IMP_RESTRAINT() macro. In that case update to the IMP_RESTRAINT_2() macro. - 11/8: Support has been added for the C++11 keywords override (IMP_OVERRIDE) and final (IMP_FINAL). They allow you to tell the compiler to produce an error if the marked method does not override/implement a parent's method and if anything overrides a given method, respectively. - 11/8: IMP::SingletonModifier::apply_indexes() and IMP::SingletonScore::evaluate_indexes() now require an upper and lower bound index (you can just pass 0 and the size of the array). We can make them option if there is a good reason. - 11/8: IMP::SingletonDerivativeModifier and kin were merged into IMP::SingletonModifier and kin as the only difference, the presence of the IMP::DerivativeAccumulator (whose weight was always 1) wasn't very interested and the split added lots of complexity. If this has an impact on your code, let us know. - 11/8; Evaluation with containers was greatly simplified, moving many methods on containers. We can add backwards compatibility versions if that is interesting. - 11/7: Predicates, scores and modifiers written in Python need to update their methods for returning inputs and outputs. They should now have a method \c _do_get_outputs(self, m, particle_indexes) which should probably just return \c [m.get_particle(i) for in in particle_indexes] to return all the passed particles. - 11/7: It is now preferred that IMP::Restraint and IMP::ScoreState descendants report dependencies directly using IMP::ModelObject::do_get_inputs() and IMP::ModelObject::do_get_outputs(). To do this, you should use the IMP_RESTRAINT_2() and IMP_CONSTRAINT_2() macros instead. Methods like IMP::Restraint::get_input_particles() are now deprecated. This change was made as the old mechanism introduced a lot of extra complication and made it so that there was no way to directly depend on something that was not a Particle or a Container. - 11/3: IMP::container::InContainerPairFilter and friends now treat pairs as unordered. A flag was added to control this. - 9/5: IMP::base::Object classes now have a method IMP::base::Object::clear_caches() that clears all cache data in the object. This is not universally implemented, but now provides a correct path for doing that. The IMP_LIST() type macros call that method when their contents changes. ## Summer 2012 - 7/3: IMP_PERIODIC_OPTIMIZER_STATE() was superceded by the class IMP::core::PeriodicOptimizerState as it is rather cleaner than the macro. - 7/2: in order to make various base classes in IMP more Python friendly, macros IMP_PROTECTED_METHOD(), IMP_PROTECTED_CONSTRUCTOR() were added that properly expose such methods to Python. The standards checks now complain if \c protected is used an a class in the API. - 6/24: Executables from \c module/bin are now put in \b build/module_bin/module. This removes a recurring source of errors in the build script due to the origin and destination directory have the same name in scons. - 6/24: Benchmarks are now put int \c module/benchmark and build to \c build/benchmark/module to remove a recurring source of errors. - 6/18: The IMP::benchmark module supports the GPerfTools profilers when available. Benchmark executables (things in \c bin whose name starts with \c benchmark_) that use the IMP_BENCHMARK() macro have command line flags to turn on and cpu and memory profiling. - 6/12: The naming rules for .i files that are in module \c pyext directories changed. They should now be \c IMP_modulename.name.i instead of \c IMP_modulename_name.i. This is to better support underscores in names. ## Winter 2011/2012 - 20/4: A method, IMP::atom::read_pdb(IMP::base::TextInput, int, IMP::atom::Hierarchy), was added to load coordinates from a pdb file into an existing IMP::atom::Hierarchy. - 4/4: the filter types were superceded by the predicate types. You will only see a difference if you develop a filter class (eg IMP::PairFilter -derived class) or directly call methods on filters (specifically IMP::PairFilter::get_contains()). - 29/3: IMP::Particle::get_has_model() has been changed to IMP::ModelObject::get_is_part_of_model() to be consistent with other classes. - 28/3: IMP::core::Mover changed slightly, adding an extra method IMP::core::Mover::get_output_particles() and requiring the IMP::Model be passed to the constructor. All movers in IMP SVN were updated without any difficulty. This allows movers to use information contained in the dependency graph amongst particles so they can, for example, weight moves based on scores. - 26/3: IMP::SingletonContainerInput etc were added. Methods that take these as their arguments can be passed either containers or lists of particles (or tuples), as convenient. All methods that take IMP::SingletonContainer objects should eventually move over to taking this (changing just requires changing the argument from a pointer to an Input). - 22/3: Over the last week or so, which IMP headers include which other heads have been substantially simplified. This may result in compilation errors for code outside of \c svn that depended on IMP headers including other headers. In general, you want to explicitly include the header defining each thing you use. - 15/3: The vector used throughout IMP is now exposed at IMP::base::Vector and what it does it documented. - 10/3: IMP::domino::RestraintCache was introduced to provide a documented centralized place to handle restraint evaluation and caching in DOMINO. In the switch various bugs involving maximum scores and IMP::RestraintSet objects were fixed. - 10/3: Scoring functions in IMP are now implemented in IMP::ScoringFunction objects. All restraint evaluation is handled by them and they can cache various useful information to aid in accelerating restraint evaluation. Backward compatibility interfaces have mostly been provided, please complain if some important interface went away. This has greatly simplified the code as well as fixed various bugs in various types of restraint evaluation (particularly in the handling of maximum scores on IMP::Restraint and IMP::RestraintSet objects. - 29/2: The macros IMP_NAMED_TUPLE_1(),IMP_NAMED_TUPLE_2(),IMP_NAMED_TUPLE_3() were added for easy generation of simply struct for use in passing and returning values. The generated structs support comparison, output, hashing and have constructors. Use the macros instead of a boost::tuple or a a POD struct. - 16/2: Restraints now are evaluated with the correct weights when evaluated by themselves as opposed to always returning their weight as seen by the model as a whole. - 15/2: The IMP::algebra grid support was rearranged and simplified. Most of the changes should be transparent. However, the IMP::algebra::grids namespace got folded into IMP::algebra, which might break some code. And various (obscure) constructors of IMP::algebra::GridD were removed. All code in svn has been changed as needed. - 3/2: The IMP::domino HDF5 container has been revamped. This involves splitting it into a write version (IMP::domino::WriteHDF5AssignmentContainer) and a read version (IMP::domino::ReadHDF5AssignmentContainer). - 3/2: IMP::domino::WriteAssignmentContainer and IMP::domino::ReadAssignmentContainer were added to do direct binary file writing. This is faster than the HDF5 versions. - 3/2: The installation instructions have been revamped and the developer instruction, in particular, have been greatly simplified. - 24/1: \c tools/show-changes has been added that shows the changes in the history log since you last updated. - 24/1: \c tools/make-source has been added that adds a source/header pair to a module - 24/1: principal components analysis in IMP::algebra was generalized to work in any dimension. This required renaming the class to IMP::algebra::PrincipalComponentAnalysisD. - 24/1: the output of the IMP::algebra::PrincipalComponentAnalysisD::show() was changed to reflect the standards (values produce compact, single line output on show). The old show has been moved to IMP::multifit - 19/1: IMP::container::ConsecutivePairContainer had IMP::container::ExclusiveConsecutivePairContainer split off of it to make the two operating mode clearer and so that there could be a clear IMP::container::ExclusiveConsecutivePairFilter to use. - IMP::algebra::Grid::apply was added to provide faster access to all (or, later, part of a grid). It can be an order of magnitude or more faster than using the iterators when doing simple operations. - the scons config options \c pretty and \c color were added (defaulting to True). When they are true, build commands are mostly suppressed and a much briefer description of what is being done is printed out (in color). Note that \c color mode and compilation in emacs don't get along too well. - IMP::algebra::ParabolicFit was renamed to IMP::algebra::ParabolicFit2D (a backwards compatibility typedef is currently provided) - IMP::algebra::LinearFit was renamed to IMP::algebra::LinearFit2D. It now supports error bars on the y-values. ## Fall 2011 - IMP::compatibility::checked_vector was renamed to IMP::compatibility::vector and made available to IMP as IMP::vector. Typedefs should use this type (IMP::vector). - Various IMP::algebra plural typedefs which had been using std::vector (since they never were updated), have now been moved to IMP::vector. - IMP::statistics::HistogramD now supports display with \c matplotlib. - IMP::statistics::HistogramD supercedes IMP::statistics::Histogram - predicate base classes have been added: IMP::SingletonPredicate, IMP::PairPredicate etc. These take particles and return an integer. Current functionality using them includes IMP::core::PredicateSingletonScore that uses a predicate to choose which score to apply. - the core generic restraint and score state support was moved to the kernel to avoid breaking various scores outside of IMP svn. - the IMP::display module was refactored and moved up to be a dependency of IMP::core and IMP::atom rather than vice versa. The main observable results are that you should now use IMP::atom::HierarchyGeometry, IMP::atom::SelectionGeometry, IMP::atom::HierarchiesGeometry, IMP::core::XYZRGeometry, IMP::atom::BondGeometry etc instead of finding them in IMP::display. This change means that, display can be used internally when testing and debugging functionality in IMP::core and IMP::atom and that tests that use IMP::display don't need to build all of IMP::atom. - Decomposition of restraints now results in restraints that are faster to evaluate, using the generic restraint mechanism in IMP kernel. - IMP::algebra::get_unit_bounding_box_d() was changed to return a bounding box with radius 1 to be consistent with IMP::algebra::get_unit_sphere_d() - A general purpose scheme to associate data with tuples of hierarchy nodes was added to the RMF format. This deprecates the special purpose bond support. - the way various types of geometry is stored in RMF files was changed, breaking backwards compatibility. ## Summer 2011 - some functionality from the kernel has been spun off into a new module IMP.base. You probably won't notice anything. The only significant changes are that functions that accept or return IMP::base::TextInput and IMP::base::TextOutput must declare them with the \c base qualifier (eg base::TextInput) so that swig is happy. - the source for the kernel has been moved from \c kernel to \c modules/kernel. This is unlikely to effect anyone. - IMP::core::ExcludedVolumeRestraint now supports filters - IMP::domino::HDF5AssignmentTable now caches internally, which makes it several times faster than before. - you can now pick which Python to use with the \c python argument to scons. - the units for BD diffusion coefficients changed to be consistent with other IMP units. - IMP::domino::get_interaction_graph() now takes an IMP::RestraintsTemp, like other functions - The assert failure handling function was documented and renamed to IMP::handle_error() so it is easier to find when you want to use gdb. - to work around various issues with ref counting pointers and compiler eccentricities, we added emulation of the C++0x \c nullptr. You should prefer that to \c NULL in IMP code that you want to run cross platform. - IMP::IntsList, IMP::FloatsList and IMP::StringsList have been added to replace the various scattered solutions for passing such types. - IMP::Restraint::get_decomposition() and IMP::Restraint::get_current_decomposition() changed to return a single Restraint, instead of a list. This makes it easier to propagate the weight and maximum reliably. - IMP now supports usage of restraints without adding them to the IMP::Model scoring function. To do this, do IMP::Restraint::set_model(), passing the model. The restraint can then be evaluated. - IMP::domino::DominoSampler does not use the IMP::domino::OptimizeRestraints restraint optimization by default. Instead, if it is asked to create the IMP::domino::RestraintScoreSubsetFilterTable, it decomposes the restraints first (via IMP::create_decomposition()). This avoids several bugs and very fragile code. - the IMP::core::DiameterRestraint was ignoring the radii of the particles involved and so was giving artificially low scores. It now is on the actual diameter of this molecule. - the IMP::core::MonteCarlo optimizers now support incremental evaluation. This can be a lot faster when only a few particles are moved each step. - all the plural types in IMP have moved to bounds checked vectors when using debug builds with gcc. If you get a non-imp error about an invalid access, it is probably some code that walks off the end of an array. Please report it. - The IMP::core::MoverBase was changed to move it off of the IMP::SingletonContainer. This resulted in the IMP::core::BallMover and IMP::core::NormalMover having new constructors. Now create them from a list of particles, a list of float keys and a radius/standard deviation. This removes some complexity and inefficiencies as well is removes some scope for errors. - Temp and non-temp decorator lists have been merged in C++. This was done to simplify code since no one was using the non-temp variants. Now all decorator lists are temp (non-reference counted) variants. - restraints and restraint sets are now evaluated only using "canonical" weights. That is, each restraint has one weight for the model (the total over all the restraint sets it is contained in and will always be evaluated with that weight). This allows the evaluation of simple restraints to be faster. - The IMP_LIST() macros have been changed slightly. If your usage needs to take action upon addition, removal or changes to the list, use the IMP_LIST_ACTION() variant in your header. The IMP_LIST_IMPL() macro no longer takes the actions, and the last three (generally empty arguments) will have to be removed from the call to get it to work. This is needed to move towards more flexible and efficient passing of lists of values. - IMP::domino uses the new IMP::Model::evaluate_if_good() support to accelerate computations - IMP::core::MonteCarlo can use the IMP::Model::evaluate_if_good() support to more quickly reject very bad conformations. See IMP::core::MonteCarlo::set_maximum_difference(). - IMP::core::ExcludedVolumeRestraint has been rewritten to be faster. It is especially faster when used with IMP::Model::evaluate_if_good() - a new evaluation mode IMP::Model::evaluate_if_good() and IMP::Model::evaluate_if_below() has been added. It allows restraints to shortcut score computation when the score can be shown to be worse than a maximum value. - as part of moving towards more efficient use of evaluate_if_good() various deprecated methods have been removed from IMP::core::ExcludedVolumeRestraint and IMP::core::RigidClosePairsFinder. - incremental evaluation support was removed from IMP. The reason for doing this was that the code was complicated, created overhead even when incremental support was not used and was not commonly used. Further, decomposing the restraints and using the dependency graph yields similar results and is more transparent. - the RMF helper binaries have been moved to their own application so that they get installed. - IMP::display::Writer has been revised to better support writers that write multiple frames to a single file and ones that write binary data. The noticeable changes are that the existing IMP::display::Writer instances must be created with a file name. And the IMP::display::WriteOptimizerState just takes a writer rather than a write and a file name template. IMP::display::Writer::set_frame() has been added so that you can directly control the frame being written. ## Spring 2011 - An interactive mode has been added to IMP::domino::DominoSampler. The interactive mode is useful for understanding domino sampling runs and distributing domino computations across a cluster. - Functions IMP::domino::set_assignments() and IMP::domino::get_assignments() were added which support writing subset assignments to data fields in hdf5 files. These can be used to helping to distribute domino on a cluster as well as debugging the sampling process. - a Python function IMP.show_graphviz() has been added that attempts to render a graph with \c dot and then open the pdf. - The Python functions IMP.show_altgraph() and IMP.get_networkx_graph() have been added to provide alternatives for quick viewing of IMP graphs. Unfortunately, they still pale compared to graphviz. - IMP.atom.create_cover() was added to facilitate creating a summary particle for part of a hierarchy in order to accelerate restraints. - IMP.hdf5 was renamed to IMP.rmf, the proposed name of the file format that it supports. That also better separates the hdf5 support from the file format support. - The old I/O functions have been marked as deprecated in favor of IMP.hdf5. You may need to compile with deprecated="yes" if you use them. - The Python graph interface now has a remove_vertex method. - One can now access the model restraint evaluation statistics programmatically using IMP::Model::get_restaint_statistics(). Restraint statistics can be logged to a file during optimization using IMP::core::WriteRestraintScoreStatisticsOptimizerState. - IMP::domino::DominoSampler now allows you to specify the tree on which merges occur directly. Also, you now need to specify the merge tree rather than the JunctionTree if you want to do non-loopy inference and specify something about how it is done. Sorry, it makes the code a lot easier. - Preliminary symmetry support has been added via the IMP::core::TransformationSymmetry and IMP::core::TransformationAndReflectionSymmetry classes. ## Winter 2011 - There now is a common base class IMP::atom::Simulation for IMP::atom::BrownianDynamics and IMP::atom::MolecularDynamics. In merging them there were several minor changes made to both. The most noticeable may be that the function for setting the maximum time step length is now IMP::atom::Simulation::set_maximum_time_step(). The last time step length can be extracted by IMP::atom::Simulation::get_last_time_step() - IMP::core::MonteCarlo::set_temperature() was renamed to IMP::core::MonteCarlo::set_kt() as that is what it is doing. - A new module IMP.hdf5 has been added to provide more general support for saving and loading IMP::atom::Hierarchy structures to files. Multiple conformations of each hierarchy can be saved as well as information about geometry and restraints (although these can't be read at the moment, so they are not so useful). - IMP::algebra::grids::GridD now supports dense grids in arbitrary dimension through use of IMP::algebra::grids::DenseGridStorageD. - IMP::Optimizer now can use a list of restraint sets to compute the score rather than the complete Model score. - each IMP::Restraint now has its own weight. This is lightly tested. - A new log level IMP::PROGRESS has been added. The idea is that IMP::Sampler::get_sample() calls and other expensive loops can show a boost progress meter at this log level to give some idea when things are moving forward. - IMP::domino::TrivialParticleStates renamed to IMP::domino::IndexParticleStates to be less judgmental - IMP::domino::CompoundParticleStates and IMP::domino::RecursiveParticleStates added - IMP::domino::DominoSampler now supports maximum scores on IMP::RestraintSets (untested). - the name of the clustering methods have been changed to create_ from get_ since they create objects. - the names of the recently added metric based clustering support has been changed to Metric rather than Distance. - IMP now has the ability to have certain tests designated as expensive and to skip those tests when test-fast is run. The motivation for this is that the tests take a really long time making it impractical to run before every commit. And it is hard to tell which tests have not been run when running with the necessary multiple build processes. My thought is that tests that take about a second or less on a debug build should be left along (this is probably >95% of the tests), but the few that take longer should be marked as expensive. To do that, rename the test to expensive_text_XXX.py (from test_XXX.py) and add expensive_python_tests= env.IMPModuleGetExpensivePythonTests() to the IMPModuleTest call in the test SConscript. - the fuction IMP::display::create_restraint_geometry() was added to do what its name says. - runtime dimensional geometric objects have been added. These include IMP::algebra::VectorKD, IMP::algebra::BoundingBoxKD and IMP::algebra::SphereKD. The IMP::statistics::Embedding classes now use IMP::algebra::VectorKD instead of Floats. The sparse grids also support variable dimensions as to the nearest neighbor searches. - IMP::atom::DopePairScore has been added to IMP. - IMP::atom::CHARMMStereochemistryRestraint provides a high-level simplified interface to the bonded part of the CHARMM forcefield. - IMP modules are now built by first combining all the .cpp files into one, and then building that file. This greatly accelerates building of a module from scratch as build time is dominated by parsing headers. This does change the semantics of .cpp files as they see headers included by and functions declared by an arbitrary subset of other .cpp files. For now they are still required to be able to be compiled separately. To get the only behavior back either globally or on a per-module basis, see the percppcompilation parameter to scons. - Instructions and scripts have been added for building modules and things externally to the IMP build. See the installation guide. - IMP::atom::CHARMMParameters can now automatically map CHARMM-style atom and residue naming to PDB-style, so CHARMM topology files can be used unmodified. - IMP::atom::CHARMMTopology::add_sequence() provides a simple way to generate a topology from a string of amino acid one-letter codes. - IMP::atom::CHARMMTopology::add_coordinates() will generate atomic Cartesian coordinates from CHARMM internal coordinate information. - IMP::atom::CHARMMTopology::add_missing_atoms() and IMP::atom::remove_charmm_untyped_atoms() can now be used to ensure that atoms read from a PDB file match the CHARMM topology. - building of deprecated functionality is now false by default - the IMP_COMPARISONS(), IMP_COMPARISONS_1(), IMP_COMPARISONS_2() macros now take the name of the class as an additional (first) argument. This avoids the formerly needed, \c This typedef and makes them more consistent with the other macros in IMP. - modules can now be built externally to the IMP svn in addition to applications and biological systems. To do so, create a directory with your module in a subdir containing links to the IMP SConscript file and scons_tools and a config.py that has information needed to run IMP (eg an appropriate \c pythonpath, \c ldpath). - removed IMP::algebra::GridD::get_index(VectorD) as it wasn't very safe (in that vectors coordinates can get rounded between the IMP::algebra::GridD::get_has_index() and IMP::algebra::GridD::get_index() calls. Use IMP::algebra::GridD::get_nearest_index() or IMP::algebra::GridD::get_extended_index() or IMP::algebra::GridD::operator[]() instead. - renamed IMP::domino::ParticleStatesTable::get_particles() to IMP::domino::ParticleStatesTable::get_subset() as it is more useful to be able to get the subset containing all the particles. - rename IMP::restrainer::Restraint::print_all_restraints() to IMP::restrainer::Restraint::show_all_restraints() to be consistent with the rest of IMP. - renamed IMP::atom::get_conect_record_string() to IMP::atom::get_pdb_conect_record_string() to make clearer what it is. - class and method names are now spell checked. Add missing words to spelling_exceptions in the module test call. - the plural names (eg IMP.Particles()) have been removed from the Python side as they are just Python lists. - biological systems can now include internal modules. These provide a convenient place to put application specific C++ and Python code. See the example system for more info. - IMP_VALUES() now includes IMP_OUTPUT_OPERATOR() so you no longer need two namespace macros for each value. - the function IMP::core::Harmonic::k_from_standard_deviation was renamed to IMP::core::Harmonic::get_k_from_standard_deviation() to conform to the standards. - the IMP::atom::PDBSelector and IMP::atom::Mol2Selector became IMP::Object classes. All this means is that they need to be created with \c new in C++ code. It also means that one is no longer prohibited from storing them (not that there is much use in doing that). - the standards tests have been improved to test more useful aspects. Now class and function names are checked some as well as all modules have the checks for show and plural types. The error messages should explain how to fix things. - IMP::core::RigidBody::lazy_set_reference_frame() was renamed to IMP::core::RigidBody::set_reference_frame_lazy() to fit with the standards better. - the transform based methods on reference frame which have been deprecated for a while were removed - the hierarchy visiting and searching functions have been renamed to better fit the IMP naming scheme as we don't expect they were used outside of the IMP source. - IMP::atom::NonhydrogenMol2Selector was renamed IMP::atom::NonHydrogenMol2Selector to be consistent with the PDB version ## Fall 2010 - IMP::display::PlaneGeometry was added to display planes - the optimize methods on optimizers was moved to a "non-virtual interface". Usage has not changed, but optimizers should implement IMP::Optimizer::do_optimize() instead of IMP::Optimizer::optimize() to benefit from better checking. - the two-vector constructor for IMP::algebra::Plane3D now expects the second vector to be a unit vector as otherwise one could switch the order around without noticing. - the io functions in the kernel have been cleaned up a bit (the old ones are there for backwards compatibility). See IMP::write_particles(), IMP::write_particles_binary(), IMP::write_particles_to_buffer() and IMP::read_particles(), IMP::read_particles_binary(), IMP::read_particles_from_buffer() - IMP::core::WriteModelOptimizerState is going away as it duplicated IMP::WriteParticlesOptimizerState. - the IMP::algebra::Grid3D has been changed to make it support any (fixed) dimension when used with sparse storage. The template arguments have also been changed to remove a redundant one. See IMP::algebra::GridD. - the biological systems specification has been improved a little adding ability to run some rudimentary tests to make sure the scripts are ok and to not require that everything be run from the svn directory. - A new method IMP::atom::create_compatible_rigid_body() which creates a rigid body from a hierarchy which has the same internal coordinates as another rigid body. A way to make the rigid body construction more canonical would be nice (so we don't need this function), but I'm not sure how to do that. - Logging has been improved to better report the name of methods producing the log messages - a macro IMP_FUNCTION_LOG has been added which should be added to the start of non-member functions which produce significant logging output. - IMP::algebra::get_distance(const Rotation3D&,const Rotation3D&) was changed to use the Euclidean distance in R4 as that seems better behaved than the prior metric. - IMP.domino2 was renamed to IMP.domino and the old IMP.domino was moved - the documentation for classes and some methods now includes a list of all examples which use the class. Please report any incorrect links or phantom classes appearing in the docs. Unfortunately, functions have to be added individually. Please request any functions of particular interest. - applications and biological systems now have their own documentation pages - IMP::atom::Hierarchy requires radii on its leaves - IMP::atom::read_pdb add radii to the atoms - IMP::helper was merged into IMP::restrainer as all the functionality there was primarily to provide support for restrainer - IMP::domino2::SubsetEvaluatorTable was removed. Pass the model and IMP::domino2::ParticleStatesTable to the IMP::domino2::RestraintScoreSubsetFilterTable instead. - IMP::atom::Selection was added to allow specification of parts of IMP::atom::Hierarchy using biological sorts of names. A few functions to create restraints from them were added too. - a new example was added in the kernel to show IMP used on a biological system - IMP::display::PymolWriter now supports adding labels to things - IMP::display::PairRestraintGeometry and IMP::display::ConnectivityRestraintGeometry were added to aid in visualizing those restraints. - IMP::bullet was added with an optimizer to use hard body physics to resolve local collisions. - IMP::atom::setup_as_rigid_body() has been deprecated and replaced by IMP::atom::create_rigid_body(). The latter creates the rigid body as a separate particle so that the molecular hierarchy doesn't have to worry about having its bounding sphere large enough to enclose all the members. - IMP::core::KClosePairsPairScore was added to have count based cutoffs for close pair scoring. - IMP::algebra::get_surface_area_and_volume() was added to compute the two numbers for a set of balls ## Summer 2010 - The IMP::core::RigidDistancePairScore and IMP::core::ClosePairsPairScore have been revised to remove certain inconsistencies and ambiguities and to make them faster. Existing code using them will need to be slightly tweaked to work. - Javi added his 2D EM scoring functionality as IMP::em2D - some new ways to speed up scons have been added. See IMP::core::HarmonicDistancePairScore, IMP::core::SoftSpherePairScore and, for C++ users, the generic restraint support IMP::container::create_restraint() and IMP::core::create_restraint(). These each offer between 2x and 1.2x speedup when using simple scores (like harmonics on distances). In addition, a function IMP::container::get_slack_estimate() has been added which will estimate the correct slack to use with a IMP::container::ClosePairContainer. This will be extended to support IMP::core::ExcludedVolumeRestraint eventually. - support for dense and sparse grids have been added to IMP::algebra. See the IMP::algebra module page (I'm still working on how to make the docs clearer). - all write functions now have associated IMP::OptimizerState and IMP::FailureHandler classes. For examples, see IMP::WriteOptimizerState, IMP::WriteBinaryOptimizerState, IMP::WriteFailureHandler, IMP::atom::WritePDBOptimizerState etc. - functions IMP::write_binary_model(), IMP::read_binary_model() and IMP::read_configurations() have been added to support space and filesystem efficient storage of model configurations. Multiple configuration can be written to a single file. This functionality requires NetCDF be installed (NetCDF is a standard library for reading and writing binary data). - domino2 has been added. It provides a simpler interface on Keren's domino optimize as well as a number of new optimization and hooks for customizing the optimization process - a better explanation of how to use reference counted pointers has been added to IMP::Pointer. - all simple, planar polygons are now supported for display when CGAL is present ## Week of April 1, 2010 - All plural objects in Python are now simply Python lists rather than C++ recreations of Python lists. This makes compilation of the wrappers much, much, much faster. The visible changes are - types are not checked until the lists are passed to C++ - passing std::vector types by pointer and storing passed std::vector types by pointer will not work (but should never have been done anyway from code exposed to Python) - swig has issues with some overloads (not always for obvious reasons). These are being sorted out. None are commonly used. - get_particles() on a plural object doesn't work in Python (and is not needed in C++). Implicit conversions will always work. - All containers need to know the model at time of construction. If, after updating, you get an error on a container constructor, just add the model as the first argument. This also means that empty lists of particles in the constructors cause check failures. We can figure out a work around if this is too annoying. ## Week of March 13, 2010 - function get_rotation_in_radians_about_axis() was renamed to get_rotation_about_axis() as it didn't parse and we don't have the "in radians" elsewhere. - Added IMP::algebra::ReferenceFrame3D. As part of the cleanup, IMP::core::RigidBody::set_reference_frame() replaces set_transformation(). And the reference-frame related transformation construction functions were replaced. IMP::algebra::get_rotation_from_x_y_axes() was added to aid in the construction of reference frames (and replaces the algebra function taking three vectors). ## Week of March 6, 2010 - Publications now have a description field to give browsers an indication of what is in the publication to aid selection of what to read/cite. ## Week of Feb 27, 2010 - IMP::algebra::decompose_rotation_into_axis_angle() was renamed to IMP::algebra::get_angle_and_axis(). - renamed IMP::atom::bond() and IMP::atom::unbond() to IMP::atom::create_bond() and IMP::atom::destroy_bond() to be more consistent with things - renamed IMP::core::root() to IMP::core::get_root() ## Week of Feb 21, 2010 - IMP::Object::set_was_owned() was changed to IMP::Object::set_was_used() as that better reflects what the flag means. - various more interpolation and endian methods had their names fixed (by adding \c get_) - the function get_mean() on vectors was removed as it duplicated IMP::algebra::get_centroid() - build_Transformation3D_from_transformation3D was replaced by IMP.algebra.get_transformation_3d() - IMP::algebra::almost_equal() is now IMP::algebra::get_are_almost_equal() to fit the naming conventions. - the random vector generators have been cleaned up to remove redundant ones and make the names more consistent. They now have the form of either - IMP::algebra::get_random_vector_in() or IMP::algebra::get_random_vector_on() where there is one argument describing the volume or shape (such as an IMP::algebra::SphereD<4>). - or IMP::algebra::get_uniform_surface_cover() where the first argument is the primitive whose surface we are covering and the second is the number of points. Some good sed strings are - "s\#random_vector_in_unit_sphere()\#get_random_vector_in(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_in_unit_box()\#get_random_vector_in(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_on_unit_sphere()\#get_random_vector_on(IMP.algebra.get_unit_sphere_3d())\#g" - "s\#random_vector_on_unit_box()\#get_random_vector_on(IMP.algebra.get_unit_bounding_box_3d())\#g" - "s\#random_vector_in_box(\#get_random_vector_in(\#g" - "s\#random_vector_on_box(\#get_random_vector_on(\#g" - "s\#random_vector_in_sphere(\#get_random_vector_in(\#g" - "s\#random_vector_on_sphere(\#get_random_vector_on(\#g" - "s\#grid_cover\#get_grid_surface_cover\#g" - "s\#uniform_coverr\#get_uniform_surface_cover\#g" . - Spheres in IMP::algebra are now supported in general dimensions. See IMP::algebra::SphereD. - Many methods in IMP::algebra were renamed to be more consistent with other methods or general conventions. Most of these simply involved adding \c get_ as a prefix and/or add a \c _3d suffix when they did not take any arguments. - build_Rotation2D_from_Vector2D became IMP::algebra::get_rotation_to_x_axis() - most of the Rotation3D and Transformation3D functions (in addition, IMP::algebra::get_identity_rotation_3d() and IMP::algebra::get_identity_transformation_3d() added the \c _3d so we can add 2D versions at some point) - IMP::algebra::Transformation3D::get_transformed() and IMP::algebra::Rotation3D::get_rotated(). Replacing \c ".transform(" worked in the IMP code base. - IMP::algebra::get_distance(), this also carries over to the IMP::core::get_distance() The following global replacements are probably safe (in sed expressions) - "s\#core.distance(\#core.get_distance\#g" - "s\#algebra.distance(\#algebra.get_distance\#g" - "s\#\.transform(\#.get_transformed(\#g" Only in C++ - "s\#\.rotate(\#.get_rotated(\#g" Only in C++ - "s\#rotation_from\#get_rotation_from\#g" - "s\#rotation_in\#get_rotation_in\#g" - "s\#identity_rotation(\#get_identity_rotation_3d(" - "s\#::distance(\#::get_distance(\#g" . - IMP::algebra::Vector3D and IMP::algebra::Vector3Ds were replaced by the expanded (non-typedef) version in functions parsed by SWIG as the current situation was extremely brittle. For the moment, you should not use the typedefs in definitions that swig sees. Sorry. - IMP::algebra::Sphere3DPatch was renamed to IMP::algebra::SpherePatch3D to be consistent with the other geometry and allow us to support higher dimensions in the future. - rigid_align_first_to_second was renamed to IMP::algebra::get_transformation_aligning_first_to_second() and build_Transformation2D_from_point_sets was renamed to IMP::algebra::get_transformation_aligning_pair() ## Week of Feb 14, 2010 - The IMP::atom::Bond types were rearranged to reconcile the protein-centric classification with ligand support. - IMP::atom::Atom no longer has charge since we have IMP::atom::Charged. - IMP::atom::read_mol2() and IMP::atom::write_mol2() were added - IMP::atom::Selector and descendants renamed to IMP::atom::PDBSelector since we will soon have Mol2Selectors. In addition, the IMP::atom::IgnoreAlteriativesSelector is now IMP::atom::NonAlternativeSelector to keep the name positive, like the others. In the IMP body, all instances of the string Selector were for PDBSelectors and so sed was safe to use. - IMP::atom::read_pdb() lost the "ignore_alternatives" argument. Instead, all of the provided selectors, other than IMP::atom::AllSelector ignore alternatives internally. Code that uses the IMP::atom::AllSelector should probably change to the IMP::atom::NonAlternativeSelector. - IMP::example renamed (from examples) - The IMP_USAGE_CHECK() macro lost the (unused) third argument. ## Week of February 3, 2010 - The model io code as moved into the kernel so it is now IMP::write_model() and IMP::read_model() - Remove the version_info argument from the implementation macros. Instead, just pick up the module version info automatically. - Add a function Object::get_type_name() and Object::get_module() to get the respective bits of information about each object. - Change Object::show() to be a non-virtual wrapper which prints the object name and other info before passing control off to Object::do_show(). - The containers were moved to IMP.container. See that module page for a full list of what moved. In the process, the backwards compatibility functions were removed from IMP::container::PairsRestraint and kin. - IMP.atom.create_protein() now returns the created protein - the simplify functions have been moved from IMP.helper. They are now IMP.atom.create_simplified_along_backbone(). - IMP.helper.clone() has been renamed to IMP.atom.create_clone() - IMP.misc.ConnectingPairContainer has been moved to IMP.core.ConnectingPairContainer - The locations for pixels in IMP::em::DensityMap have been moved to the center of the corresponding voxel from the lower left corner. - IMP::em::get_transformed(), IMP::em::get_resampled() and IMP::em::get_density() have been added. - The documentation has been majorly cleaned up and simplified. Please look around and see if you like it. There are now two main pages, the introduction and the developers guide, rather than the previous plethora of pages. - IMP::algebra::Cone3D::get_bounding_sphere() has been removed since it was not of general use, not returning the bounding sphere you would expect and was trivial to implement. - IMP::algebra::projection() was removed since it didn't actually project onto the segment and the name didn't follow the normal conventions. And it wasn't used. - Most geometry now has namespace methods IMP::algebra::get_bounding_box(), IMP::algebra::get_surface_area(), IMP::algebra::get_volume(). ## Week of January 22, 2010 - A generic geometry layer has been added (see \ref genericgeometry). Using it, IMP::atom::rmsd(), IMP::algebra::alignment_taking_first_to_second() and the new IMP::atom::native_overlap() all can work on either collections of either IMP::core::XYZ particles or IMP::algebra::Vector3Ds. - Deprecated functionality has been removed from IMP.core. For those still using it, IMP::core::ClosePairsScoreState was replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairsScoreState was replaced by IMP::core::CloseBipartitePairContainer. - The macros for Scores changed slightly. Scores not in IMP svn which use the IMP_*_SCORE macros now have to implement the get_is_changed() function outside of the class declaration and scores which use the IMP_SIMPLE_*_SCORE now don't have to implement the functions to get interactions and used particles. - Rename IMP::core::MinimumPairScoreRestraint (and kin, Maximum, Singleton, Triplet and Quad) to IMP::core::MinimumPairRestraint. - Add IMP::core::MinumumPairScore (and kin). ## Week of January 1, 2010 - IMP::em is now using Particles to resample density maps and for fitting restraints. The previous ParticleAccessPoint mechanism is removed. - IMP::UsageException and IMP::InternalException are now reserved for their respective checks (and cannot be thrown by IMP_THROW()). This makes the distinction between API exceptions (which are thrown in any build mode and should be documented and tested) and helper exceptions which are only thrown when checks are turned on (and hence should not be documented) clearer. - IMP::core::RestraintSet moved to IMP::RestraintSet. The restraint sets now simply define a tree with weights which the model traverses as opposed to being an opaque IMP::Restraint themselves. This means that incremental evaluation of restraints in restraint sets now works and that restraints in restraint sets can be individually evaluated. Existing working usage should not have changed (so \c sed should manage updates just fine), but the internal implementation is completely new, so pay attention. - IMP now has a text i/o layer consisting of IMP::TextInput and IMP::TextOutput. These classes are convertible to C++ streams and implicitly constructable from both streams and file paths. As a result, functions that take them will work when called with a string for a file path, a Python file or a C++ stream. - The IMP.examples module now documents how to use the needed aspects of scons and SWIG. - The IMP.examples module is used as the basis for new modules created by the make-module script, so those include more documentation and are tested. - Module overview pages are now generated from the information passed in the SConscript file, rather than from the (now removed) modulename.dox file. See the examples file for an example of the things you can do there. - boost program options is now a checked for dependency. - The IMP::Model now uses dependency information from IMP::ScoreStates to automatically reorder the IMP::ScoreStates so that they are evaluated in the right order. Also, IMP::Model::evaluate() on a subset of the restraints now only updates the needed score states, not all of them. - Functions on rigid bodies no longer provide the IMP::core::RigidMemberRefiner as a default as it is, pretty much always, the wrong thing to use. That class may go away without notice in the future and is no longer documented. - the IMP::helper::create_simple_connectivity_on_rigid_bodies() and IMP::helper::create_simple_excluded_volume_on_rigid_bodies() now both take an IMP::Refiner as an (optional) final argument. This makes it so they can be used with the rigid bodies created from molecular hierarchies. This code still needs to be updated and the docs don't reflect what the code was or is doing, so use with caution (see notes in the header file). - The IMP::helper create* functions now all take non-Object inputs as const& to make clear what is input and that there are no memory management gotchas. - a new base class IMP::Constraint has been added for constraints. Various other classes have been renamed to be consistent. ## Week of December 11 - The interface of IMP::PairScore changed slightly to make the function signatures more uniform across various methods. Now all methods that take a pair of IMP::Particle objects take a const IMP::ParticlePair&. Previously some took that, some took an IMP::ParticlePair and some took two IMP::Particle*s. If you have an implementation of an IMP::PairScore outside of svn, you will have to update your function signatures (the compiler will tell you where). Backward compatible IMP::PairScore::evaluate() methods have been provided, so simple users of pair scores shouldn't see a difference. - IMP::core::PairRestraint takes a IMP::ParticlePair in the constructor instead of two IMP::Particle*s. A backwards compatible constructor can be added if needed. - IMP::ParticlePair::first and second went away since no one claimed to be using it and it complicated the implementation. - The generic restraints/score states/scores infrastructure has been extended to triplets and quads (for angles and torsion angles). ## Week of December 4 - IMP now makes use of Boost.Filesystem to manipulate paths. Things may not work well on systems where boost libs are not available which does not support posix paths. But libless boost installs generally don't exist outside of our test systems. - new functions like IMP::core::get_example_path(filename) and IMP::core::get_data_path(filename) allow one to easily find a data file named 'filename' for the examples and library internal data. - The IMP::display module has been revamped in order to support non-linear geometry. VRML support has been dropped. CMM support will be dropped unless someone requests it (as IMP::display::ChimeraWriter does all it does and more). ## Week of November 19 - imp.xcodeproj was moved into tools and updated accordingly - the bin directory finally went away - IMP::display::ChimeraWriter now supports the latest version of Chimera - IMP::display::CGOWriter was renamed to IMP::display::PymolWriter as it can do more than just CGO. - IMP::VectorOfRefCounted::operator[] can now be used for assignments just like a normal vector. No more need to use the IMP::VectorOfRefCounted::set() method. - the IMP::atom::Hierarchy constructor from an IMP::Particle* is now explicit (as were the decorator constructors) ## Week of November 12 - All exception types are now mapped to equivalent classes in Python. The class hierarchy is similar (e.g. IMP::ModelException derives from IMP::Exception in both C++ and Python) so all IMP exceptions can be caught in Python using "except IMP.Exception". For convenience, the IMP::IOException, IMP::ValueException and IMP::IndexException classes in Python also derive from the similarly-named standard Python classes (IOError, ValueError and IndexError respectively). ## week of October 29 Update - You can now add members to rigid bodies via IMP::core::RigidBody::add_member(). These members can be other rigid bodies (although this functionality is only lightly tested). - The code to get bounding volumes for IMP::atom::Hierarchy has been moved to the atom module, as as the clone and destroy code. They can be found at IMP::atom::clone(), IMP::atom::destroy(), IMP::atom::get_bounding_box(). - Added IMP::core::VolumeRestraint. - Several functions have been added to helper in creating restraints. They are create_simple_connectivity_on_rigid_bodies(), create_simple_connectivity_on_molecules(), create_simple_distance(), create_simple_diameter(), create_simple_em_fit(), and create_simple_excluded_volume_on_rigid_bodies(). Functions load_em_density_map() and set_rigid_bodies() are added as well. See IMP::helper for more information. - IMP::ScoreStates are further hidden. Namely, IMP::core::RigidBody constructs the needed score states internally, doing away with the need for the create functions (which have been done away with). In addition there are new decorators IMP::core::Cover and IMP::core::Centroid which manage their respective score states, further doing away with more create methods. The hierarchy-based create method for rigid bodies has been moved to the atom module and is called IMP::atom::rigid_body_setup_hierarchy(). - IMP::em::read_map() will figure out the file type to use automatically from the extension. Other such methods should follow suite. - A class IMP::core::StructureSet has been added to provide a convenient way to manage solutions of optimization runs. - IMP::core::RigidBodyMover now takes its angle argument in radians and is documented as such. - IMP::core::ClosePairsScoreState and IMP::core::CloseBipartitePairsScoreState have been replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairContainer. The score states have been marked as deprecated and are mostly still there, but are more limited than they were. - IMP::core::AllPairsPairContainer has been renamed to IMP::core::AllPairContainer and a bipartite version has been added. - All containers now keep track of what they looked like on the previous time step. - The IMP::core::ListPairContainer and IMP::core::ListSingletonContainer have had their interface reduced a bit in order to make implementing change tracking easier. - The set of exceptions has been simplified to go along with simplification of the checks. We now have IMP::InternalException and IMP::UsageException which are thrown by the respective checks (and the former is thrown by IMP_FAILURE()). When you want to to pass an exception back to Python use one of IMP::ValueException or IMP::IndexException or IMP::IOException. There is also a new macro, IMP_THROW() to be used when you want to throw an exception with a nice error message. Please look at code you have worked on and make sure that the exceptions make sense given the new clarifications. - An IMP::ModelException is now thrown when IMP::Particle attributes attain invalid values. This exception can be caught to recover from failures of IMP::core::ConjugateGradients or difficult restraints. ## Week of October 15 - The set of functions needed to get meta information about IMP::Restraint and IMP::ScoreState objects has hopefully stabilized. Sorry about the thrashing. If you have restraints or score states not in SVN, see the corresponding IMP_RESTRAINT type macros, the IMP::Interaction class and base classes for what exactly you need to implement. ## Week of October 8 - IMP::Interaction was modified to split IMP::interaction::get_used_particles() into IMP::interaction::get_read_particles() and IMP::interaction::get_write_particles(). This makes it possible to check for IMP::ScoreState objects being in an inconsistent order (so that the input to one is updated after it is read). Later, we will be able to automatically order the score states ensure that there are no problems. - IMP::atom::Hierarchy has been modifier to get rid of the IMP::atom::Hierarchy::Type. Instead, all types are contained in the decorators. IMP::atom::get_by_type() was also updated to reflect this. - IMP_check() and IMP_assert() have been replaced with IMP_USAGE_CHECK() and IMP_INTERNAL_CHECK() respectively. Hopefully this makes it clearer when to use which. IMP::set_check_level() has been revised accordingly (to use IMP::USAGE and IMP::USAGE_AND_INTERNAL). Also, IMP_failure() and IMP_not_implemented() have been renamed to IMP_FAILURE() and IMP_NOT_IMPLEMENTED() to make all macros be all caps. To update external code, - replace IMP.EXPENSIVE with IMP.USAGE_AND_INTERNAL - replace IMP.CHEAP with IMP.USAGE - replace IMP_assert and IMP_check as above. ## Week of October 1 - IMP::ScoreState and IMP::Restraint now better track the set of particles that they use and the dependencies between the particles. To this end, a new base class IMP::Interaction has been added with methods to get the inter-particle interactions and the set of particles used. - .cpp, .h, .py and SConscript files which contain the string "DO NOT COMMIT" will be rejected on checkin. Use this to make experimental or debugging changes you are making so you don't accidentally commit them. - The search module is no longer built by default and is not longer supported. It will get moved out of svn shortly. - IMP::ScoreState objects now also have IMP::Interacting::get_interacting_particles() methods. Soon IMP::PairScore and IMP::SingletonScore objects will too. ## Week of September 27 - Some support for gather statistics has been added to IMP::Model (see IMP::Model::set_gather_statistics()). If requested, an API can be provided so that the individual data points can be returned to programs. - A class IMP::core::RigidBodyDistancePairScore has been added which provides accelerated computations for the distance between two rigid bodies. - The SWIG input files were renamed from swig.i to swig.i-in to aid in dependency tracking. - rigid bodies now have two sets of particles associated with them in the various places they are used. One is the IMP::core::RigidBodyMember particles, as was previously the case. The other is defined through a refiner (and could be the RigidBodyMember particles if a IMP::core::RigidMembersRefiner is used). The idea is that you sometimes want some particles which are transformed along with the rigid body, but which are not really part of the representation of the rigid body. Examples would include marker points or multiresolution representations of proteins. ## Week of September 20 - all \c .i files have been renamed to start with \c IMP_ to help make dependency tracking easier. - static builds (builds with \c static=True) build both the static and dynamic libraries and just static executables - the \c profile build target has been removed as \c static=True does the important bit of what is needed. Just do \c linkflags=['-pg'] to add the needed link flag for \c gprof. - Restraints should now implement IMP::Restraint::unprotected_evaluate() instead of IMP::Restraint::evaluate(). The latter can now be called outside of IMP::Model::evaluate() and will return the right answer always. - \c bin/imppy.sh no longer works, as was previously warned. Use \c tools/imppy.sh instead. The bin directory will go away eventually, for now the script just tells you to use the tools version. ## Week of August 30 - A new score IMP::core::BoundingBox3DSingletonScore has been added to keep particles in a box. - str should now work better in Python when applied to IMP::Object objects and a few others. Objects should use the new IMP_SHOWABLE and IMP_SHOWABLE_INLINE macros instead of declaring show functions directory to make it so that they work in Python. It is now a bug to \c print or \c str to not return something useful in Python for any IMP object. ## Week of August 23 - each module now has a version, eg IMP::get_module_version_info() which is also returned by the various restraints of the module. The Python code checks that loaded libraries have the expected version numbers to try to provide better messages when runtime linking problems occur. - the various bounding_box methods have been renamed to ::get_bounding_box() to be more consistent with the other names. - Python documentation is now parsed by doxygen ## Week of August 15 - The Object::get_version_info() result now includes the SVN revision info when building from SVN. This means you can use it to more easily tell exactly what a built version of IMP was built from. ## Week of August 8 - Treatment of hetatoms in hierarchies (and hence pdbs) is now documented. See IMP::atom::Hierarchy. This involved broadening IMP::atom::Residue to handle ligands and having IMP::atom::Hierarchy::LIGAND, IMP::atom::Hierarchy::AMINOACID and IMP::atom::Hierarchy::NUCLEICACID be handled by IMP::atom::Residue. - IMP::atom::read_pdb() currently drops HOH ATOM records on the floor. We should add a IMP::atom::Water decorator if anyone wants to handle them (since they aren't atoms and don't have elements). - IMP::atom::Mass() has been added for decorating particles with mass. IMP::atom::Atom mass has been moved to that. - The scheme for generating atom types is now fully specified. See the IMP::atom::AtomType documentation. - IMP::atom::Atom and IMP::atom::Residue have been somewhat restructured. Some of the helper functions have been renamed to be more descriptive or removed (if they duplicated other easily accessible functionality). get_residue_index() has been removed (since you can just call get_index() on the return result of the IMP::atom::get_residue()). IMP::atom::get_chain() now returns an IMP::atom::Chain. root() was renamed to IMP::atom::get_root() for consistency. - SWIG files are now generated from a file called swig.i in the pyext directory. This file only has to contain module-specific commands such as lines to wrap headers and instantiate module templates. No more boilerplate. - The contents of bin are moving to tools since the two directories both contained assorted development tools (not all of which were executable). For now, bin/imppy.sh will still work, but that will go away soon. - IMP::Decorator::show() now just takes the stream (it used to take a prefix also, which was often ignored). Anyone who implemented a decorator will have to delete the second argument of their show function. ## Week of August 1 - Containers now have methods IMP::SingletonContainer::evaluate(), IMP::PairContainer::evaluate(), IMP::SingletonContainer::apply() and IMP::PairContainer::apply(). Using these can be significantly faster than looping over the particles externally. Similarly, there is IMP::SingletonScore::evaluate(), IMP::PairScore::evaluate(), IMP::SingletonModifier::apply(), IMP::PairModifier::apply(). - The decorator functions have been renamed to IMP::Decorator::setup_particle(), IMP::Decorator::decorate_particle() and IMP::Decorator::particle_is_instance(). A shell script \c tools/update-decorators is provided which updates the names in all code in subdirectories of the current directory. Please backup things first. - a function, IMP::Restraint::safe_evaluate() has been added to be used when requesting the value of restraints outside of the IMP::Model::evaluate(). ## Week of July 18 - The atom::BondPairContainer has been renamed to atom::BondPairFilter since it wasn't a real container. A real atom::BondPairContainer has been added. - Out of source builds mostly works. To do an out of source build, make a link to \c scons_tools and \c SConstruct from the source directory in the build directory and then add the line \command{repository="../path_to_src_dir"} in the \c config.py in the build directory. Then running \c scons in the build directory will put generated files there. This allows you to make both a \c release and \c fast build off of the same set of sources. Proper docs will be added soon. - The SConscript setup for a module has been greatly simplified so it should be easy enough for users to maintain their own module and updates to the build system shouldn't involve changes to user SConscripts any more. This does mean that any non-svn modules need to be updated. Look at IMP.core or another module and copy what is there. Again, proper docs will be added soon. ## Week of July 11 - Per Javi's suggestion, the conventions have been split into "Coding conventions" and "API conventions" and linked from the main page. - benchmarks got moved to their own module, IMP::benchmark and will now be tracked over time ## Week of July 4 - FilteredListSingletonContainer and FilteredPairSingletonContainer went away as there was too much hard to factor code shared with IMP::core::ListSingletonContainer and IMP::core::PairSingletonContainer. The IMP::core::ClosePairsScoreState is now handled by a list of filters on the class itself. That is, do IMP::core::ClosePairsScoreState::add_close_pair_filter() to add a filter. - Hooks have been added to allow actions to be taken on an assertion or check failure. See IMP::HandleFailure and its descendants for more details. - IMP::display::CGOWriter has been much improved and now allows provides lots of display-time control and easy building of animations. - Generic optimizer states have been added (IMP::core::SingletonsOptimizerState and IMP::core::PairsOptimizerState). ## Week of June 15 - box distributions (IMP::algebra::random_vector_on_box(), IMP::algebra::random_vector_in_box()) take bounding boxes - IMP::em::DensityMap are now IMP::Object objects. Some current allocations on the stack may need to be changed (since ref counted objects can't be stack allocated). - IMP::em::DensityMap::Read() is now deprecated (as is IMP::em::DensityMap::Write) as they violated a slew of conventions. Use IMP::em::read_map() and IMP::em::write_map() instead. ## Week of June 8 - new methods IMP::atom::add_radii() and IMP::atom::add_bonds(); - Hierarchys to IMP::atom::Hierarchies - IMP::Particles has been changed to a new type. It no longer plays quite so well with Python lists (you can't create one from an arbitrary Python list). - The containers for multiple decorators have been revised to make them more useful. See IMP::Decorators. You can now implicitly convert, for %example, a IMP::core::XYZs to an IMP::Particles (as well as use IMP::Decorators::get_particles()). All methods that take a collection of particles with a certain type now instead take a IMP::Decorators. - A method that has expectations about the particles it takes should take decorators rather than particles. That is, anything which expects IMP::core::XYZ particles, should take an IMP::core::XYZs instead of IMP::Particles. - IMP_LIST() now takes an extra argument (the name of the plural container) - In C++ IMP::Decorator objects are implicitly convertible to IMP::Particle* objects. Unfortunately, making this work in Python is problematic. - More methods take and return decorators rather than IMP::Particles. This means it is clearer what their expectations are and what they produce. ## Week of May 15 - The IMP::helper module was added and various functionality moved there. See its documentation page for a list of what is there. - IMP::algebra::SphericalCoords was renamed to IMP::algebra::SphericalVector3D to be consistent with other algebra classes. - IMP::algebra::Rotation2D is now IMP::InvalidDefault to be consistent with the other similar classes. - We are trying to be a bit more specific about what is expected from a IMP::atom::Hierarchy. Towards that end, there are two methods for extracting a representation from a hierarchy, IMP::atom::get_detailed_representation(), and IMP::atom::get_simplified_representation() - A method has been added to produce a simplified copy of an IMP::atom::Hierarchy. The method is still a bit crude and could benefit from other people inspecting the output and saying how it should be improved. - There is now an IMP::algebra::BoundingBoxD (and IMP::algebra::BoundingBox3D) to track bounding boxes. There are methods like IMP::algebra::bound_box(Sphere3D) which gets the bounding box of a sphere. - Each IMP::Object has its own log level. If IMP::Object::set_log_level() is called with a value other than IMP::DEFAULT, the value is used for code within and called by the object. To support this in your objects you need to add IMP_OBJECT_LOG to the start of each major method. It has been added to some of the core objects. ## Week of May 8 - IMP::atom::MolecularHierarchy has been renamed to IMP::atom::Hierarchy. Use \command{sed -i.old "s\#MolecularHierarchy\#Hierarchy\#g" my_files} - the long-deprecated Name(Decorator) was removed - The decorator suffix has been removed from decorator names. - Unit conventions have been established and documented for distance, energy, force and angle. - there is a function IMP::atom::create_protein() to create a coarse grained protein. The function is crude so far, but the idea is that we should have standard ways of generating such things so that everyone can benefit from best practices. - a class IMP::atom::BondGraph has been added so Boost.Graph algorithms can be used with bond graphs - the IMP-specific doxygen commands have been documented. - the doxygen commands unstable{name} and untested{name} have been added to doxygen to be used for marking methods and functions as likely to change or not well tested. Pass the name of the class or method as name. - bonds can now be added to molecular hierarchies automatically see IMP::atom::add_bonds() ## Week of May 1 - There is a IMP::Rotation2D - You can now extract fixed Euler ZYZ angles from a IMP::Rotation3D. This code can serve as a template for extracting other Euler angle conventions if they are needed. - IMP::algebra::rotation_between_two_vectors() changed to IMP::algebra::rotation_taking_first_to_second() to make it clear which direction the rotation goes. ## Week of April 27 - IMP::Decorator derived objects are now convertible to bool. So code like \code if (d) { blah } \endcode works. - IMP::Decorator::cast now returns a null decorator (\c DecoratorName()) rather than throwing an exception. This might make it actually useful. - New IMP::statistics module for statistical methods (e.g. clustering). ## Week of April 20 - Support for taking advantage of rigid bodies in collision detection has been added via the IMP::core::RigidClosePairsFinder. - All IMP::core::ClosePairsScoreState-related functionality has been optimized and should be noticeably faster. - The EMBED project has been merged into the IMP::em module. Thus, it is no longer necessary to obtain EMBED separately to build IMP::em. - All C++ methods which work with std::istream or std::ostream arguments should now be usable from Python: any Python file-like object can be used. - The SWIG interface now ties in to the reference counting mechanism used in the C++ code, for all classes derived from IMP::RefCounted. So, for %example, deleting a Python object should not free the underlying C++ object if another C++ object holds a reference to it, and unrefing a C++ object should not delete it if a Python object still holds a reference to it. There should be no further need to manually keep references to Python objects to prevent things from being prematurely cleaned up. This should be transparent to the user; please report any problems. - IMP::atom::AtomDecorator no longer inherits from IMP::core::XYZDecorator. ## Week of April 13 - The decorators IMP::atom::AtomDecorator, IMP::atom::ChainDecorator, IMP::atom::ResidueDecorator etc. now inherit from IMP::atom::MolecularHierarchyDecorator. This means they can set their IMP::atom::MolecularHierarchyDecorator::Type properly and saves a lot of typing in the common case. We can always make hierarchy decorators zero memory overhead if we find that the cost matters. - The base classes IMP::Comparable, IMP::NullDefault, IMP::ValidDefault and IMP::UninitializedDefault were added to document what the default constructors do for various objects in IMP. ## Week of April 6 - IMP::Object now defines pure virtual methods IMP::Object::get_version_info() and IMP::Object::show(). All non-abstract classes which inherit from IMP::Object must implement them. Doing this means they are more uniformly implemented and are documented in one place. - IMP::Object is now reference counted (the only objects this changes are IMP::Model and IMP::Optimizer). As a result optimizers and IMP::Model objects must be created on the heap. - IMP::RefCounted has had its name changed and no longer inherits from IMP::Object. This makes it easier to use independently. - the docs for reference counting (IMP::RefCounted) and and IMP::Decorator have been improved - running doxygen produces no warnings. Let's try to keep it that way. Run \command{scons doc > /dev/null} before you check in large blocks of new code to make sure that you didn't introduce warnings. - Helper classes have been added to make SingletonModifier and PairModifier objects look like functors for C++ and Python. This makes it so you can use \c map in Python and \c for_each in C++. See IMP::SingletonFunctor and IMP::PairFunctor for more information. - The recent changes page has moved to a doxygen page to make it easier to link changes back to the relevant source code. ## Week of March 20: - create_rigid_body(), create_rigid_bodies(), create_centroid() etc. now require you to add the score state to the model yourself. Whether the rigid body is enforced or not can be controlled by adding or removing the returned IMP::ScoreState. - All previously deprecated code has been removed. This includes indexes in particles, the no-argument constructor for particles and the functionality that was moved from core to atom. - The IMP::core::RigidBodyDecorator and IMP::core::RigidMemberDecorator have been simplified by removing the traits classes. There is now only one type of rigid body and it stores the list of member particles internally. - IMP::core::NameDecorator has been deprecated as IMP::Particles have built in names (use IMP::Particle::get_name(), IMP::Particle::set_name()). - IMP::atom::ChainDecorator has been added ## Week of March 13: - A new IMP::display::Writer, IMP::display::ChimeraWriter has been added which exports geometry to native Chimera constructs. It probably should be used instead of bild and cmm for most purposes as it combines the best of both types. - compose() for transformation is now a free function so compose(rotation, vector) will work via promotions. - the geometry classes have been cleaned up and are now more consistent with IMP standards (and more likely to do what their names/comments suggest). See IMP::algebra::Sphere3D, IMP::algebra::Cylinder3D, IMP::algebra::Segment3D, IMP::algebra::Plane3D etc. - ParticleRefiner has been renamed to IMP::Refiner since the Particle was long and unnecessary. eIF3 translation initiation complex](http://salilab.org/40S-eIF1-eIF3).
319 - IMP::domino::DiscreteSampler::set_restraints() now allows the set of
320  restraints used in DOMINO filtering to be controlled (previously, all
321  restraints in the IMP::Model were used).
322 - IMP::modeller::IMPRestraints now transfers an IMP::ScoringFunction into the
323  MODELLER score, not the implicit (and deprecated) score of the
324  entire IMP::Model.
325 - Bugfix #775: coordinates of voxels in EM maps were previously inconsistently
326  treated (they are supposed to be the centers of the voxels, but in some
327  cases they were the lower left edge). The fix may result in maps and their
328  bounding boxes appearing shifted by half the voxel size.
329 
330 # 2.4.0 - 2015-02-27 # {#changelog_2_4_0}
331 - Python 3 is now supported. The Windows package now works out of the box
332  with Python 3.3 or 3.4; a `-python3` subpackage is provided for Ubuntu
333  and Fedora. (Apple does not yet provide Python 3 so the Mac package is not
334  built with Python 3 support, but if you install Python 3 you can compile
335  IMP from source to work with it.)
336 - A new command line tool, `rrt_sample`, allows for running rapidly exploring
337  random tree (RRT) based sampling, e.g. to sample conformations of multiple
338  domains connected by flexible linkers.
339 - Command line tools are now placed in the `bin` directory of modules (the
340  old `bin` directory, used for utility programs that are not installed, is
341  now called `utility`). The `applications` directory is removed.
342 - Several Python modules that were only used by EMageFit have been moved from
343  IMP::em2d to IMP::EMageFit.
344 - A new tutorial that uses the IMP::pmi module is now available (modeling of
345  the RNA Polymerase II stalk).
346 
347 # 2.3.1 - 2014-12-01 # {#changelog_2_3_1}
348 - Minor bug fixes.
349 - Update to IMP::pmi to support the Nup84 and rnapolii biological systems.
350 
351 # 2.3.0 - 2014-10-24 # {#changelog_2_3_0}
352 - A new IMP::pmi module was added, containing high-level Python classes to simplify the representation, scoring, sampling and analyzing of models in IMP. This module is still under heavy development, and should be considered experimental.
353 - Binary installers for 64-bit Windows are now available.
354 - The `.py` extension has been removed from all Python applications, so that they are consistent with the C++ applications.
355 - IMP::atom::Selection objects can now be combined using basic set operations (union, intersection, difference, symmetric difference). In Python the equivalent operators (|, &, -, ^ respectively) can be used.
356 - The integrative docking (idock) application now employs SOAP to improve the scoring.
357 - IMP::atom::CHARMMParameters::create_topology() now works with a wider range of hierarchies (previously it would only accept Residues that were direct children of Chains). For example, it should do the right thing when given a Hierarchy that contains no Chain particles, or one where Residues are grouped under Fragments.
358 - Only Python 2.6 or later is now supported. We still provide IMP for RedHat Enterprise 5 (which ships with Python 2.4) but you will need to use the `python26` binary (part of EPEL) rather than the system default Python.
359 
360 # 2.2.1 - 2014-08-11 # {#changelog_2_2_1}
361 - Bugfix: PoseScore (see the ligand_score application) should now match that used by the Pose&Rank web server (http://salilab.org/poseandrank/)
362 
363 # 2.2.0 - 2014-04-09 # {#changelog_2_2_0}
364 - A module IMP.template was added to act as a template for creating new external modules. It includes setting up git hooks for format checks and things like that.
365 - The Python support for grids was cleaned up. Hopefully this makes it more robust. If you were using a grid in Python that no longer exists, ping us, we can add it back.
366 - The generic geometry functions were simplified slightly and renamed to IMP::algebra::get_vector_geometry().
367 - The IMP::base::Flag class was added to simplify adding flags in C++.
368 - The `--log_level` and `--check_level` command line argument now take strings (eg VERBOSE) to specify their values, rather than cryptic (and a bit unstable) numbers.
369 - Command line options are now divided into normal and advanced ones. You do `--help_advanced` to show the advanced ones.
370 - Support for computing Connolly surfaces was added with IMP::algebra::get_connolly_surface(). In addition, there is now IMP::algebra::get_uniform_surface_cover(const IMP::algebra::Sphere3Ds&,double) to sample the surface of a collection of balls.
371 - IMP::atom::reate_simplified_from_volume() was added to create a simplified representation of a hierarchy that preserves the surface are and occupied volume. It uses IMP::algebra::get_simplified_from_volume(). If you want to play with such representations, you can use the rmf_simply application.
372 - IMP::atom::get_rmsd(IMP::atom::Selection, IMP::atom::Selection, const IMP::algebra::Transformation3D&) was deprecated. Use IMP::atom::get_rmsd_transforming_first() instead.
373 - IMP::algebra::get_rmsd() and IMP::algebra::get_rmsd_transforming_first() were added.
374 - The family of macros IMP_BASE_[SMALL,LARGE]_[UNORDERED,ORDERED]_[MAP,SET] have been added (eg IMP_BASE_SMALL_ORDERED_SET ()). They provide platform and boost-version independent ways to take advantage of the best set/map for an application. In particular, they use boost::container::flat_set and kin when available.
375 - IMP::atom::State and IMP::atom::Representation decorators have been added to provide support for multiple states and multiple representation schemes. They are integrated with RMFs and IMP::atom::Selection. See atom/multiresolution.py and atom/multistate.py for examples.
376 - The container implementations has been cleaned up and simplified a bit. This should not change most usage.
377 - [RMF](http://salilab.github.io/rmf) has been updated. There are a few minor API changes, in addition to a new, faster format (`.rmf3`).
378 - [Eigen](#Eigen) is now included with IMP.
379 - The deprecated function IMP::kernel::Optimizer::set_restraints() was removed. Use IMP::kernel::Optimizer::set_scoring_function() instead.
380 - The deprecated constructor for IMP::kernel::ModelObject that does not take an IMP::kernel::Model was removed. This propagates to constructors IMP::kernel::Restraint, IMP::kernel::ScoreState and kin. Make sure you pass an IMP::kernel::Model as the first constructor argument. Related to this, deprecated functions like get_is_part_of_model(), set_model() were removed as they don't serve any purpose.
381 - The deprecated methods in IMP::kernel::Optimizer that pertain to attribute optimization were removed. Inherit from IMP::kernel::AttributeOptimizer instead if you want to use them in your optimizer.
382 - Some deprecated functionality was removed. If you encounter problems, replace IMP::base::OwnerPointer with IMP::base::PointerMember, find logging stuff in IMP/base/log.h, object in IMP/base/Object.h and random in IMP/base/random.h.
383 - IMP::atom::Chain now uses strings for the chain ids, to be consistent with changes in the PDB. PDB files are written using the first character of the chain string.
384 - Added orientation-dependent SOAP scoring (IMP::score_functor::OrientedSoap and IMP::atom::OrientedSoapPairScore). Orientation-dependent SOAP scores include those that score loops (SOAP-Loop), protein-peptide interfaces (SOAP-Peptide) and proteins (SOAP-Protein). The library files themselves are rather large (~1.5GB) and so are not included with IMP. They can be downloaded separately from http://salilab.org/SOAP/.
385 
386 # 2.1.1 - 2013-11-04 # {#changelog_2_1_1}
387 - Added support for Mac OS 10.9
388 
389 # 2.1.0 - 2013-10-24 # {#changelog_2_1_0}
390 - There is now some limited IMP-wide gathering of timing statistics. You can add `--statistics=filename` to the command line or use the IMP::base::show_timings() function to view a summary of where IMP has spent time.
391 - Slack was being used inconsistently (and not always correctly) among the various containers and restraints. If you had heavily optimized the slack value for speed, you may want to revisit it.
392 - Class name OwnerPointer was switched to PointerMember, to denote more clearly a ref-counting pointer that is a class member.
393 - The dependency handling in IMP has been rewritten to simplify it and fix a bunch of bugs. There are now more checks of correctness, so behavior that might have silently worked (or not worked) before may now trigger errors. In particular, restraints/score states must have access to the IMP::kernel::Model to do much of anything, so either use constructors that take the model or call IMP::kernel::ModelObject::set_model(). Also, changing your set of inputs or outputs without calling IMP::kernel::ModelObject::set_has_dependencies(false) is now likely to trigger a check failure (it would just silently do the wrong thing before).
394 - Decorators have been cleaned up to make them behave consistently regarding IMP::kernel::Particle vs IMP::kernel::ParticleIndexes. This change shouldn't break existing code, but decorators should move to the new macros IMP_DECORATOR_METHODS(), IMP_DECORATOR_SETUP_0() etc.
395 - The IMP_GRAPH() and IMP_WEIGHTED_GRAPH() macros got an extra argument describing how to show the vertexes in the graph. They also now expose a method, eg IMP::kernel::show_as_graphviz() that writes the graph in graphviz format to an IMP::base::TextOutput.
396 - Support and rules for deprecating code in IMP have been updated. There are new macros and instructions on how to use them that should result in much better visibility about changes. See the C++ wiki for IMP for more information. And there is now an encoded policy of keeping deprecated things around for 1 release after deprecation.
397 - IMP is now much more selective about when it invalidates the dependency graph. See [Dependencies](http://github.com/salilab/imp/wiki/Dependencies) for more information. This change should make evaluation of isolated restraints much faster in certain situations where they were unexpectedly slow before.
398 - IMP::atom::Selection has been completely rewritten as it was very brittle. The set of particles returned may change in some cases, in some cases because it was buggy before, in some cases because of new bugs.
399 - Reminder: the IMP_PROTECTED macros didn't work out well, and are now deprecated. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading `_` on the method name, if you get `Swig director pure virtual method called IMP::kernel::PairScore::do_get_inputs` messages.
400 - The generated `cmake` files are no longer stored in the repository. Instead `tools/build/setup_cmake.py` is run after checkouts and other git events that change the repository tree (assuming you have run `setup_git.py`).
401 - Generation of documentation using `doxygen` has been refactored. It is now done on a per-module basis, via targets like `IMP.core-doc`. The main advantage is that changes to the doc for a module can be tested comparatively quickly, just by building that module's docs. On the down side, docs are no longer quite so heavily interconnected, so dependencies only work to things that the module actually depends on.
402 - The restraint and score state statistics functions were removed from IMP::Model. They had been broken for quite a while, with no one complaining. Statistics can be added back easily if requested.
403 - Added an IMP-specific fork of `git-flow`. It handles things like `README.md` files when you start and finish feature branches and provides a tool to nicely update IMP and display the change log. The main commands of use are `git imp feature start <feature_name>`, `git imp feature finish <feature_name>`, and `git imp update`. You may want to remove the `git-flow` lines from your `.git/config` file as you may accidentally type `git flow` instead of `git imp`. You need to run `setup_git.py` to set things up.
404 - [RMF](http://salilab.github.com/rmf) is now included as a git submodule rather than copied into the IMP repository. You should rerun `setup_git.py` to make sure submodule stuff is initialized.
405 
406 
407 # For IMP 2.0 and earlier # {#changelog_2_0}
408 ## Spring 2013
409 - Modules and applications now have a _Maintainer_ in their `README.md`, giving the github id of the person to whom issues should be assigned.
410 - `setup_git.py` supports the `--module` flag which will set up git hooks and things for a git repository that contains a module.
411 - To clean up for the release, the module dependencies are now described in a file called `dependencies.py` and the module docs in a `README.md`, both in the main module directory.
412 - IMP-wide support for progress bars was added with the IMP::base::set_progress_display() function and helpers. These are only displayed when the IMP::base::get_log_level() is IMP::base::PROGRESS.
413 
414 ## Winter 2012-2013
415 - Scons support was removed. Use cmake to build IMP, it is way better.
416 - Examples are run with a `--test` argument when IMP is built with internal checks. Expensive examples should check this flag and shorten what they do in order to try to test all their code without taking too long.
417 - The class IMP::core::NonRigidMember was added to allow particles that are not rigid but whose coordinates are specified via internal coordinates in some IMP::algebra::Reference3D that is part of an IMP::core::RigidBody.
418 - The IMP_BIG_ENDIAN and IMP_LITTLE_ENDIAN preprocessor symbols are deprecated in favor of the BOOST_BIG_ENDIAN and BOOST_LITTLE_ENDIAN symbols.
419 - Support for asymmetric moves was added to IMP::core::MonteCarlo. To do this, IMP::core::MonteCarloMover was added and IMP::core::Mover and IMP::core::MoverBase were deprecated and implemented in terms of the new class. This should require few or no changes in code that uses Monte Carlo support or implements movers (although movers should move to IMP::core::MonteCarlo). All IMP::core::MonteCarloMover objects keep track of statistics (number of moves and number of accepted moves), rather than that being done in the IMP::core::SerialMover. In addition, the poorly defined move probability was removed from IMP::core::MonteCarlo.
420 - IMP::base::WarningContext now outputs warnings immediately. This change was needed as the previous behavior resulted in occasional core dumps when the libraries for log support were unloaded before the libraries where static IMP::base::WarningContexts were declared. And there was no easy way to detect such problems.
421 - You can now (probably) build IMP with [CMake on IMP wiki](https://github.com/salilab/imp/wiki/Cmake).
422 - The maximum logging level and check level can now be controlled independently of the build mode. To, for example, enable progress level logging in a fast build, add maxlog="PROGRESS" to your config. The maximum check level for `release` has been reduced to IMP::USAGE as that speeds things up a bit.
423 - The way applications are structured and specified has been regularized and simplified. Python files are automatically copied/installed and binary executables are compiled from each found `.cpp` file linked with any `.cpp` files contained in a subdirectory called `lib`. Nothing needs to be put into the SConscript file.
424 - The Python constants like IMP.algebra.has_cgal went away as swig autogenerates ones like IMP.algebra.IMP_ALGEBRA_HAS_IMP_CGAL. While the latter are a bit uglier, the constants aren't used much and having them be auto generated is nice.
425 - There have been some major changes to the build system. The main one is that the SConscripts in the directories below the module directory aren't really used. They are being removed from the repository. Now everything really is off of file locations (eg all test_* in test is a test, each *.cpp in bin becomes a program etc). This greatly simplifies the build system, which has, historically been a disaster. You can now define dependencies by dropping files into a dependency subdirectory in your module, no need to mess with the SConscript.
426 - Building IMP now always requires specifying a `repository` argument. This was always required for out of source builds, which is what one should be doing, but is now required for in source builds to, so as to make accidentally polluting your source with an in-source build harder. Use `repository="."` in that case.
427 - The HDF5 functionality in RMF was reorganized. Specifically, the headers were moved into `RMF/HDF5`, the Python functionality into `RMF_HDF5` and the code into namespace `RMF::HDF5` and all mentions of `HDF5` or `hdf5` in the class and function names were removed.
428 - Support for IMP-wide command line flags have been added to IMP::base in the base/flags.h file. They provide a simple command line parsing interface that can be used from C++ and Python. In addition, various functionality in IMP is automatically exposed on the command line when those functions are used. For example, one can control logging, checks or profiling. See IMP/base/flags.h for (some) more info.
429 - OpenMP support in IMP is officially released. All tests pass. But there has been little specific testing so far.
430 - Access to attributes using IMP::ParticleIndex and the IMP::Model now works from Python. See IMP::Model::add_attribute() and kin.
431 - IMP::ParticleKey and IMP::ParticlesKey were renamed to IMP::ParticleIndexKey and IMP::ParticleIndexesKey respectively as that is what they store. Typedefs were provided to support old code.
432 - Indexes have been added to the docs connecting classes with all the methods that return them, examples that use them and methods that take them. And all methods that are used in examples with the examples they are used in. Please check them out and report issues and ways to improve things. You can find links on the main page.
433 - The docs for a modules/biological systems/applications should now be found in an `overview.dox` in the module `doc` folder. Modules in IMP svn have been updated. For external modules, you can just move a `overview.dox` from that was previously generated into your `doc` folder and (eventually) remove the lines from your `doc/SConscript`. This makes it easier to use the full range of doxygen markup, as well as removes the need to escape doxygen markup for Python.
434 - Examples have now been simplified into full blown doxygen examples. This involves merging the `.py` and `.readme` files (putting `\#\#` in front of each line of the `.readme`) and adding a `## \\example module/examplename.py` line at the beginning. You can use `./tools/updaters/update_examples.py modules/mymodule` to update a module that was not already updated.
435 - IMP::base::FailureHandler has been removed as it wasn't used.
436 - Handling of data and includes has been simplified. All appropriate files in the corresponding directories are linked into build/installed. This is only likely to be a problem if you had broken headers sitting around `include`. If so, just comment them out, or move them some place nicer.
437 - Modules now must be in a folder called `modules`, which must have a `description.py` file.
438 - The build system should now be a little better about cleaning up after itself.
439 - How headers/doc files/data/examples
440 - The IMP_PROTECTED macros didn't work out well, and are now deprecated. Sorry. If you had Python implementations on top of IMP base classes that implement protected virtual methods, you may need to remove a leading "_" from the method name.
441 - Nice version numbers will now be generated if you use git. You need to rerun
442 tools/git/developer_tools/init_git if you have an existing clone.
443 
444 ## Fall 2012
445 - 11/9: RMF is now an external library shipped with IMP. IMP will build it and install it automatically if you have cmake version 2.8 or higher installed.
446 - 11/9: Updated support macros for IMP::Restraint and IMP::ScoreState classes have been added (IMP_RESTRAINT_2() and IMP_SCORE_STATE_2()) that make it easier to return dependencies on things other than IMP::Particle and IMP::ScoreState objects. In addition, the dependency functions on scores and modifiers have been updated similarly and standardized. In general this should not cause backward compatibility problems unless you were overriding the IMP::ModelObject::do_get_inputs() in IMP::Restraint and using the IMP_RESTRAINT() macro. In that case update to the IMP_RESTRAINT_2() macro.
447 - 11/8: Support has been added for the C++11 keywords override (IMP_OVERRIDE) and final (IMP_FINAL). They allow you to tell the compiler to produce an error if the marked method does not override/implement a parent's method and if anything overrides a given method, respectively.
448 - 11/8: IMP::SingletonModifier::apply_indexes() and IMP::SingletonScore::evaluate_indexes() now require an upper and lower bound index (you can just pass 0 and the size of the array). We can make them option if there is a good reason.
449 - 11/8: IMP::SingletonDerivativeModifier and kin were merged into IMP::SingletonModifier and kin as the only difference, the presence of the IMP::DerivativeAccumulator (whose weight was always 1) wasn't very interested and the split added lots of complexity. If this has an impact on your code, let us know.
450 - 11/8; Evaluation with containers was greatly simplified, moving many methods on containers. We can add backwards compatibility versions if that is interesting.
451 - 11/7: Predicates, scores and modifiers written in Python need to update their methods for returning inputs and outputs. They should now have a method \c _do_get_outputs(self, m, particle_indexes) which should probably just return \c [m.get_particle(i) for in in particle_indexes] to return all the passed particles.
452 - 11/7: It is now preferred that IMP::Restraint and IMP::ScoreState descendants report dependencies directly using IMP::ModelObject::do_get_inputs() and IMP::ModelObject::do_get_outputs(). To do this, you should use the IMP_RESTRAINT_2() and IMP_CONSTRAINT_2() macros instead. Methods like IMP::Restraint::get_input_particles() are now deprecated. This change was made as the old mechanism introduced a lot of extra complication and made it so that there was no way to directly depend on something that was not a Particle or a Container.
453 - 11/3: IMP::container::InContainerPairFilter and friends now treat pairs as unordered. A flag was added to control this.
454 - 9/5: IMP::base::Object classes now have a method IMP::base::Object::clear_caches() that clears all cache data in the object. This is not universally implemented, but now provides a correct path for doing that. The IMP_LIST() type macros call that method when their contents changes.
455 ## Summer 2012
456 - 7/3: IMP_PERIODIC_OPTIMIZER_STATE() was superceded by the class IMP::core::PeriodicOptimizerState as it is rather cleaner than the macro.
457 - 7/2: in order to make various base classes in IMP more Python friendly, macros IMP_PROTECTED_METHOD(), IMP_PROTECTED_CONSTRUCTOR() were added that properly expose such methods to Python. The standards checks now complain if \c protected is used an a class in the API.
458 - 6/24: Executables from \c module/bin are now put in \b build/module_bin/module. This removes a recurring source of errors in the build script due to the origin and destination directory have the same name in scons.
459 - 6/24: Benchmarks are now put int \c module/benchmark and build to \c build/benchmark/module to remove a recurring source of errors.
460 - 6/18: The IMP::benchmark module supports the GPerfTools profilers when available. Benchmark executables (things in \c bin whose name starts with \c benchmark_) that use the IMP_BENCHMARK() macro have command line flags to turn on and cpu and memory profiling.
461 - 6/12: The naming rules for .i files that are in module \c pyext directories changed. They should now be \c IMP_modulename.name.i instead of \c IMP_modulename_name.i. This is to better support underscores in names.
462 ## Winter 2011/2012
463 - 20/4: A method, IMP::atom::read_pdb(IMP::base::TextInput, int, IMP::atom::Hierarchy), was added to load coordinates from a pdb file into an existing IMP::atom::Hierarchy.
464 - 4/4: the filter types were superceded by the predicate types. You will only see a difference if you develop a filter class (eg IMP::PairFilter -derived class) or directly call methods on filters (specifically IMP::PairFilter::get_contains()).
465 - 29/3: IMP::Particle::get_has_model() has been changed to IMP::ModelObject::get_is_part_of_model() to be consistent with other classes.
466 - 28/3: IMP::core::Mover changed slightly, adding an extra method IMP::core::Mover::get_output_particles() and requiring the IMP::Model be passed to the constructor. All movers in IMP SVN were updated without any difficulty. This allows movers to use information contained in the dependency graph amongst particles so they can, for example, weight moves based on scores.
467 - 26/3: IMP::SingletonContainerInput etc were added. Methods that take these as their arguments can be passed either containers or lists of particles (or tuples), as convenient. All methods that take IMP::SingletonContainer objects should eventually move over to taking this (changing just requires changing the argument from a pointer to an Input).
468 - 22/3: Over the last week or so, which IMP headers include which other heads have been substantially simplified. This may result in compilation errors for code outside of \c svn that depended on IMP headers including other headers. In general, you want to explicitly include the header defining each thing you use.
469 - 15/3: The vector used throughout IMP is now exposed at IMP::base::Vector and what it does it documented.
470 - 10/3: IMP::domino::RestraintCache was introduced to provide a documented centralized place to handle restraint evaluation and caching in DOMINO. In the switch various bugs involving maximum scores and IMP::RestraintSet objects were fixed.
471 - 10/3: Scoring functions in IMP are now implemented in IMP::ScoringFunction objects. All restraint evaluation is handled by them and they can cache various useful information to aid in accelerating restraint evaluation. Backward compatibility interfaces have mostly been provided, please complain if some important interface went away. This has greatly simplified the code as well as fixed various bugs in various types of restraint evaluation (particularly in the handling of maximum scores on IMP::Restraint and IMP::RestraintSet objects.
472 - 29/2: The macros IMP_NAMED_TUPLE_1(),IMP_NAMED_TUPLE_2(),IMP_NAMED_TUPLE_3() were added for easy generation of simply struct for use in passing and returning values. The generated structs support comparison, output, hashing and have constructors. Use the macros instead of a boost::tuple or a a POD struct.
473 - 16/2: Restraints now are evaluated with the correct weights when evaluated by themselves as opposed to always returning their weight as seen by the model as a whole.
474 - 15/2: The IMP::algebra grid support was rearranged and simplified. Most of the changes should be transparent. However, the IMP::algebra::grids namespace got folded into IMP::algebra, which might break some code. And various (obscure) constructors of IMP::algebra::GridD were removed. All code in svn has been changed as needed.
475 - 3/2: The IMP::domino HDF5 container has been revamped. This involves splitting it into a write version (IMP::domino::WriteHDF5AssignmentContainer) and a read version (IMP::domino::ReadHDF5AssignmentContainer).
476 - 3/2: IMP::domino::WriteAssignmentContainer and IMP::domino::ReadAssignmentContainer were added to do direct binary file writing. This is faster than the HDF5 versions.
477 - 3/2: The installation instructions have been revamped and the developer instruction, in particular, have been greatly simplified.
478 - 24/1: \c tools/show-changes has been added that shows the changes in the history log since you last updated.
479 - 24/1: \c tools/make-source has been added that adds a source/header pair to a module
480 - 24/1: principal components analysis in IMP::algebra was generalized to work in any dimension. This required renaming the class to IMP::algebra::PrincipalComponentAnalysisD.
481 - 24/1: the output of the IMP::algebra::PrincipalComponentAnalysisD::show() was changed to reflect the standards (values produce compact, single line output on show). The old show has been moved to IMP::multifit
482 - 19/1: IMP::container::ConsecutivePairContainer had IMP::container::ExclusiveConsecutivePairContainer split off of it to make the two operating mode clearer and so that there could be a clear IMP::container::ExclusiveConsecutivePairFilter to use.
483 - IMP::algebra::Grid::apply was added to provide faster access to all (or, later, part of a grid). It can be an order of magnitude or more faster than using the iterators when doing simple operations.
484 - the scons config options \c pretty and \c color were added (defaulting to True). When they are true, build commands are mostly suppressed and a much briefer description of what is being done is printed out (in color). Note that \c color mode and compilation in emacs don't get along too well.
485 - IMP::algebra::ParabolicFit was renamed to IMP::algebra::ParabolicFit2D (a backwards compatibility typedef is currently provided)
486 - IMP::algebra::LinearFit was renamed to IMP::algebra::LinearFit2D. It now supports error bars on the y-values.
487 ## Fall 2011
488 - IMP::compatibility::checked_vector was renamed to IMP::compatibility::vector and made available to IMP as IMP::vector. Typedefs should use this type (IMP::vector).
489 - Various IMP::algebra plural typedefs which had been using std::vector (since they never were updated), have now been moved to IMP::vector.
490 - IMP::statistics::HistogramD now supports display with \c matplotlib.
491 - IMP::statistics::HistogramD supercedes IMP::statistics::Histogram
492 - predicate base classes have been added: IMP::SingletonPredicate, IMP::PairPredicate etc. These take particles and return an integer. Current functionality using them includes IMP::core::PredicateSingletonScore that uses a predicate to choose which score to apply.
493 - the core generic restraint and score state support was moved to the kernel to avoid breaking various scores outside of IMP svn.
494 - the IMP::display module was refactored and moved up to be a dependency of IMP::core and IMP::atom rather than vice versa. The main observable results are that you should now use IMP::atom::HierarchyGeometry, IMP::atom::SelectionGeometry, IMP::atom::HierarchiesGeometry, IMP::core::XYZRGeometry, IMP::atom::BondGeometry etc instead of finding them in IMP::display. This change means that, display can be used internally when testing and debugging functionality in IMP::core and IMP::atom and that tests that use IMP::display don't need to build all of IMP::atom.
495 - Decomposition of restraints now results in restraints that are faster to evaluate, using the generic restraint mechanism in IMP kernel.
496 - IMP::algebra::get_unit_bounding_box_d() was changed to return a bounding box with radius 1 to be consistent with IMP::algebra::get_unit_sphere_d()
497 - A general purpose scheme to associate data with tuples of hierarchy nodes was added to the RMF format. This deprecates the special purpose bond support.
498 - the way various types of geometry is stored in RMF files was changed, breaking backwards compatibility.
499 ## Summer 2011
500 - some functionality from the kernel has been spun off into a new module IMP.base. You probably won't notice anything. The only significant changes are that functions that accept or return IMP::base::TextInput and IMP::base::TextOutput must declare them with the \c base qualifier (eg base::TextInput) so that swig is happy.
501 - the source for the kernel has been moved from \c kernel to \c modules/kernel. This is unlikely to effect anyone.
502 - IMP::core::ExcludedVolumeRestraint now supports filters
503 - IMP::domino::HDF5AssignmentTable now caches internally, which makes it several times faster than before.
504 - you can now pick which Python to use with the \c python argument to scons.
505 - the units for BD diffusion coefficients changed to be consistent with other IMP units.
506 - IMP::domino::get_interaction_graph() now takes an IMP::RestraintsTemp, like other functions
507 - The assert failure handling function was documented and renamed to IMP::handle_error() so it is easier to find when you want to use gdb.
508 - to work around various issues with ref counting pointers and compiler eccentricities, we added emulation of the C++0x \c nullptr. You should prefer that to \c NULL in IMP code that you want to run cross platform.
509 - IMP::IntsList, IMP::FloatsList and IMP::StringsList have been added to replace the various scattered solutions for passing such types.
510 - IMP::Restraint::get_decomposition() and IMP::Restraint::get_current_decomposition() changed to return a single Restraint, instead of a list. This makes it easier to propagate the weight and maximum reliably.
511 - IMP now supports usage of restraints without adding them to the IMP::Model scoring function. To do this, do IMP::Restraint::set_model(), passing the model. The restraint can then be evaluated.
512 - IMP::domino::DominoSampler does not use the IMP::domino::OptimizeRestraints restraint optimization by default. Instead, if it is asked to create the IMP::domino::RestraintScoreSubsetFilterTable, it decomposes the restraints first (via IMP::create_decomposition()). This avoids several bugs and very fragile code.
513 - the IMP::core::DiameterRestraint was ignoring the radii of the particles involved and so was giving artificially low scores. It now is on the actual diameter of this molecule.
514 - the IMP::core::MonteCarlo optimizers now support incremental evaluation. This can be a lot faster when only a few particles are moved each step.
515 - all the plural types in IMP have moved to bounds checked vectors when using debug builds with gcc. If you get a non-imp error about an invalid access, it is probably some code that walks off the end of an array. Please report it.
516 - The IMP::core::MoverBase was changed to move it off of the IMP::SingletonContainer. This resulted in the IMP::core::BallMover and IMP::core::NormalMover having new constructors. Now create them from a list of particles, a list of float keys and a radius/standard deviation. This removes some complexity and inefficiencies as well is removes some scope for errors.
517 - Temp and non-temp decorator lists have been merged in C++. This was done to simplify code since no one was using the non-temp variants. Now all decorator lists are temp (non-reference counted) variants.
518 - restraints and restraint sets are now evaluated only using "canonical" weights. That is, each restraint has one weight for the model (the total over all the restraint sets it is contained in and will always be evaluated with that weight). This allows the evaluation of simple restraints to be faster.
519 - The IMP_LIST() macros have been changed slightly. If your usage needs to take action upon addition, removal or changes to the list, use the IMP_LIST_ACTION() variant in your header. The IMP_LIST_IMPL() macro no longer takes the actions, and the last three (generally empty arguments) will have to be removed from the call to get it to work. This is needed to move towards more flexible and efficient passing of lists of values.
520 - IMP::domino uses the new IMP::Model::evaluate_if_good() support to accelerate computations
521 - IMP::core::MonteCarlo can use the IMP::Model::evaluate_if_good() support to more quickly reject
522  very bad conformations. See IMP::core::MonteCarlo::set_maximum_difference().
523 - IMP::core::ExcludedVolumeRestraint has been rewritten to be faster. It is especially
524  faster when used with IMP::Model::evaluate_if_good()
525 - a new evaluation mode IMP::Model::evaluate_if_good() and IMP::Model::evaluate_if_below() has been added. It allows restraints to shortcut score computation when the score can be shown to be worse than a maximum value.
526 - as part of moving towards more efficient use of evaluate_if_good() various deprecated methods have been removed from IMP::core::ExcludedVolumeRestraint and IMP::core::RigidClosePairsFinder.
527 - incremental evaluation support was removed from IMP. The reason for doing this was that the code was complicated, created overhead even when incremental support was not used and was not commonly used. Further, decomposing the restraints and using the dependency graph yields similar results and is more transparent.
528 - the RMF helper binaries have been moved to their own application so that they get installed.
529 - IMP::display::Writer has been revised to better support writers that write multiple frames to a single file and ones that write binary data. The noticeable changes are that the existing IMP::display::Writer instances must be created with a file name. And the IMP::display::WriteOptimizerState just takes a writer rather than a write and a file name template. IMP::display::Writer::set_frame() has been added so that you can directly control the frame being written.
530 
531 ## Spring 2011
532 - An interactive mode has been added to IMP::domino::DominoSampler. The interactive mode is useful for understanding domino sampling runs and distributing domino computations across a cluster.
533 - Functions IMP::domino::set_assignments() and IMP::domino::get_assignments() were added which support
534  writing subset assignments to data fields in hdf5 files. These can be used to helping to distribute
535  domino on a cluster as well as debugging the sampling process.
536 - a Python function IMP.show_graphviz() has been added that attempts to render a graph with \c dot and then open the pdf.
537 - The Python functions IMP.show_altgraph() and IMP.get_networkx_graph() have been added to provide alternatives for quick viewing of IMP graphs. Unfortunately, they still pale compared to graphviz.
538 - IMP.atom.create_cover() was added to facilitate creating a summary particle for part of a hierarchy in order to accelerate restraints.
539 - IMP.hdf5 was renamed to IMP.rmf, the proposed name of the file format that it supports. That also better separates the hdf5 support from the file format support.
540 - The old I/O functions have been marked as deprecated in favor of IMP.hdf5. You may need to compile with deprecated="yes" if you use them.
541 - The Python graph interface now has a remove_vertex method.
542 - One can now access the model restraint evaluation statistics programmatically using IMP::Model::get_restaint_statistics(). Restraint statistics can be logged to a file during optimization using IMP::core::WriteRestraintScoreStatisticsOptimizerState.
543 - IMP::domino::DominoSampler now allows you to specify the tree on which merges occur directly. Also, you now need to specify the merge tree rather than the JunctionTree if you want to do non-loopy inference and specify something about how it is done. Sorry, it makes the code a lot easier.
544 - Preliminary symmetry support has been added via the IMP::core::TransformationSymmetry and IMP::core::TransformationAndReflectionSymmetry classes.
545 
546 ## Winter 2011
547 - There now is a common base class IMP::atom::Simulation for IMP::atom::BrownianDynamics and IMP::atom::MolecularDynamics. In merging them there were several minor changes made to both. The most noticeable may be that the function for setting the maximum time step length is now IMP::atom::Simulation::set_maximum_time_step(). The last time step length can be extracted by IMP::atom::Simulation::get_last_time_step()
548 - IMP::core::MonteCarlo::set_temperature() was renamed to IMP::core::MonteCarlo::set_kt() as that is what it is doing.
549 - A new module IMP.hdf5 has been added to provide more general support for saving and loading IMP::atom::Hierarchy structures to files. Multiple conformations of each hierarchy can be saved as well as information about geometry and restraints (although these can't be read at the moment, so they are not so useful).
550 - IMP::algebra::grids::GridD now supports dense grids in arbitrary dimension through use of IMP::algebra::grids::DenseGridStorageD.
551 - IMP::Optimizer now can use a list of restraint sets to compute the score rather than the complete Model score.
552 - each IMP::Restraint now has its own weight. This is lightly tested.
553 - A new log level IMP::PROGRESS has been added. The idea is that IMP::Sampler::get_sample() calls and other expensive loops can show a boost progress meter at this log level to give some idea when things are moving forward.
554 - IMP::domino::TrivialParticleStates renamed to IMP::domino::IndexParticleStates to be less judgmental
555 - IMP::domino::CompoundParticleStates and IMP::domino::RecursiveParticleStates added
556 - IMP::domino::DominoSampler now supports maximum scores on IMP::RestraintSets (untested).
557 - the name of the clustering methods have been changed to create_ from get_ since they create objects.
558 - the names of the recently added metric based clustering support has been changed to Metric rather than Distance.
559 - IMP now has the ability to have certain tests designated as expensive and to skip those tests when test-fast is run. The motivation for this is that the tests take a really long time making it impractical to run before every commit. And it is hard to tell which tests have not been run when running with the necessary multiple build processes. My thought is that tests that take about a second or less on a debug build should be left along (this is probably >95% of the tests), but the few that take longer should be marked as expensive. To do that, rename the test to expensive_text_XXX.py (from test_XXX.py) and add expensive_python_tests= env.IMPModuleGetExpensivePythonTests() to the IMPModuleTest call in the test SConscript.
560 - the fuction IMP::display::create_restraint_geometry() was added to do what its name says.
561 - runtime dimensional geometric objects have been added. These include IMP::algebra::VectorKD, IMP::algebra::BoundingBoxKD and IMP::algebra::SphereKD. The IMP::statistics::Embedding classes now use IMP::algebra::VectorKD instead of Floats. The sparse grids also support variable dimensions as to the nearest neighbor searches.
562 - IMP::atom::DopePairScore has been added to IMP.
563 - IMP::atom::CHARMMStereochemistryRestraint provides a high-level simplified
564 interface to the bonded part of the CHARMM forcefield.
565 - IMP modules are now built by first combining all the .cpp files into one, and then building that file. This greatly accelerates building of a module from scratch as build time is dominated by parsing headers. This does change the semantics of .cpp files as they see headers included by and functions declared by an arbitrary subset of other .cpp files. For now they are still required to be able to be compiled separately. To get the only behavior back either globally or on a per-module basis, see the percppcompilation parameter to scons.
566 - Instructions and scripts have been added for building modules and things externally to the IMP build. See the installation guide.
567 - IMP::atom::CHARMMParameters can now automatically map CHARMM-style atom and
568 residue naming to PDB-style, so CHARMM topology files can be used unmodified.
569 - IMP::atom::CHARMMTopology::add_sequence() provides a simple way to generate a topology from a string of amino acid one-letter codes.
570 - IMP::atom::CHARMMTopology::add_coordinates() will generate atomic Cartesian coordinates from CHARMM internal coordinate information.
571 - IMP::atom::CHARMMTopology::add_missing_atoms() and
572 IMP::atom::remove_charmm_untyped_atoms() can now be used to ensure that atoms
573 read from a PDB file match the CHARMM topology.
574 - building of deprecated functionality is now false by default
575 - the IMP_COMPARISONS(), IMP_COMPARISONS_1(), IMP_COMPARISONS_2() macros now take the name of the class as an additional (first) argument. This avoids the formerly needed, \c This typedef and makes them more consistent with the other macros in IMP.
576 - modules can now be built externally to the IMP svn in addition to applications and biological systems. To do so, create a directory with your module in a subdir containing links to the IMP SConscript file and scons_tools and a config.py that has information needed to run IMP (eg an appropriate \c pythonpath, \c ldpath).
577 - removed IMP::algebra::GridD::get_index(VectorD) as it wasn't very safe (in that vectors coordinates can get rounded between the IMP::algebra::GridD::get_has_index() and IMP::algebra::GridD::get_index() calls. Use IMP::algebra::GridD::get_nearest_index() or IMP::algebra::GridD::get_extended_index() or IMP::algebra::GridD::operator[]() instead.
578 - renamed IMP::domino::ParticleStatesTable::get_particles() to IMP::domino::ParticleStatesTable::get_subset() as it is more useful to be able to get the subset containing all the particles.
579 - rename IMP::restrainer::Restraint::print_all_restraints() to IMP::restrainer::Restraint::show_all_restraints() to be consistent with the rest of IMP.
580 - renamed IMP::atom::get_conect_record_string() to IMP::atom::get_pdb_conect_record_string() to make clearer what it is.
581 - class and method names are now spell checked. Add missing words to spelling_exceptions in the module test call.
582 - the plural names (eg IMP.Particles()) have been removed from the Python side as they are just Python lists.
583 - biological systems can now include internal modules. These provide a convenient place to put application specific C++ and Python code. See the example system for more info.
584 - IMP_VALUES() now includes IMP_OUTPUT_OPERATOR() so you no longer need two namespace macros for each value.
585 - the function IMP::core::Harmonic::k_from_standard_deviation was renamed to IMP::core::Harmonic::get_k_from_standard_deviation() to conform to the standards.
586 - the IMP::atom::PDBSelector and IMP::atom::Mol2Selector became IMP::Object classes. All this
587  means is that they need to be created with \c new in C++ code. It also means that one is
588  no longer prohibited from storing them (not that there is much use in doing that).
589 - the standards tests have been improved to test more useful aspects. Now class and function names are checked some as well as all modules have the checks for show and plural types. The error messages should explain how to fix things.
590 - IMP::core::RigidBody::lazy_set_reference_frame() was renamed to IMP::core::RigidBody::set_reference_frame_lazy() to fit with the standards better.
591 - the transform based methods on reference frame which have been deprecated for a while were removed
592 - the hierarchy visiting and searching functions have been renamed to better fit the IMP naming scheme as we don't expect they were used outside of the IMP source.
593 - IMP::atom::NonhydrogenMol2Selector was renamed IMP::atom::NonHydrogenMol2Selector to be consistent with the PDB version
594 
595 ## Fall 2010
596 - IMP::display::PlaneGeometry was added to display planes
597 - the optimize methods on optimizers was moved to a "non-virtual interface". Usage has not changed, but optimizers should implement IMP::Optimizer::do_optimize() instead of IMP::Optimizer::optimize() to
598 benefit from better checking.
599 - the two-vector constructor for IMP::algebra::Plane3D now expects the second vector to be a unit vector as otherwise one could switch the order around without noticing.
600 - the io functions in the kernel have been cleaned up a bit (the old ones are there for backwards compatibility). See IMP::write_particles(), IMP::write_particles_binary(), IMP::write_particles_to_buffer() and IMP::read_particles(), IMP::read_particles_binary(), IMP::read_particles_from_buffer()
601 - IMP::core::WriteModelOptimizerState is going away as it duplicated IMP::WriteParticlesOptimizerState.
602 - the IMP::algebra::Grid3D has been changed to make it support any (fixed) dimension when used with sparse storage. The template arguments have also been changed to remove a redundant one. See IMP::algebra::GridD.
603 - the biological systems specification has been improved a little adding ability to run some rudimentary tests to make sure the scripts are ok and to not require that everything be run from the svn directory.
604 - A new method IMP::atom::create_compatible_rigid_body() which creates a rigid body from a hierarchy which has the same internal coordinates as another rigid body. A way to make the rigid body construction more canonical would be nice (so we don't need this function), but I'm not sure how to do that.
605 - Logging has been improved to better report the name of methods producing the log messages
606 - a macro IMP_FUNCTION_LOG has been added which should be added to the start of non-member functions which produce significant logging output.
607 - IMP::algebra::get_distance(const Rotation3D&,const Rotation3D&) was changed to use the Euclidean distance in R4 as that seems better behaved than the prior metric.
608 - IMP.domino2 was renamed to IMP.domino and the old IMP.domino was moved
609 - the documentation for classes and some methods now includes a list of all examples which use the class. Please report any incorrect links or phantom classes appearing in the docs. Unfortunately, functions have to be added individually. Please request any functions of particular interest.
610 - applications and biological systems now have their own documentation pages
611 - IMP::atom::Hierarchy requires radii on its leaves
612 - IMP::atom::read_pdb add radii to the atoms
613 - IMP::helper was merged into IMP::restrainer as all the functionality there was primarily to provide support for restrainer
614 - IMP::domino2::SubsetEvaluatorTable was removed. Pass the model and IMP::domino2::ParticleStatesTable to the IMP::domino2::RestraintScoreSubsetFilterTable instead.
615 - IMP::atom::Selection was added to allow specification of parts of IMP::atom::Hierarchy using biological sorts of names. A few functions to create restraints from them were added too.
616 - a new example was added in the kernel to show IMP used on a biological system
617 - IMP::display::PymolWriter now supports adding labels to things
618 - IMP::display::PairRestraintGeometry and IMP::display::ConnectivityRestraintGeometry were added to aid in visualizing those restraints.
619 - IMP::bullet was added with an optimizer to use hard body physics to resolve local collisions.
620 - IMP::atom::setup_as_rigid_body() has been deprecated and replaced by IMP::atom::create_rigid_body(). The latter creates the rigid body as a separate particle so that the molecular hierarchy doesn't have to worry about having its bounding sphere large enough to enclose all the members.
621 - IMP::core::KClosePairsPairScore was added to have count based cutoffs for close pair scoring.
622 - IMP::algebra::get_surface_area_and_volume() was added to compute the two numbers for a set of balls
623 
624 ## Summer 2010
625 - The IMP::core::RigidDistancePairScore and IMP::core::ClosePairsPairScore have been revised to remove certain inconsistencies and ambiguities and to make them faster. Existing code using them will need to be slightly tweaked to work.
626 - Javi added his 2D EM scoring functionality as IMP::em2D
627 - some new ways to speed up scons have been added. See IMP::core::HarmonicDistancePairScore, IMP::core::SoftSpherePairScore and, for C++ users, the generic restraint support IMP::container::create_restraint() and IMP::core::create_restraint(). These each offer between 2x and 1.2x speedup when using simple scores (like harmonics on distances). In addition, a function IMP::container::get_slack_estimate() has been added which will estimate the correct slack to use with a IMP::container::ClosePairContainer. This will be extended to support IMP::core::ExcludedVolumeRestraint eventually.
628 - support for dense and sparse grids have been added to IMP::algebra. See the IMP::algebra module page (I'm still working on how to make the docs clearer).
629 - all write functions now have associated IMP::OptimizerState and IMP::FailureHandler classes. For examples, see IMP::WriteOptimizerState, IMP::WriteBinaryOptimizerState, IMP::WriteFailureHandler, IMP::atom::WritePDBOptimizerState etc.
630 - functions IMP::write_binary_model(), IMP::read_binary_model() and IMP::read_configurations() have been added to support space and filesystem efficient storage of model configurations. Multiple configuration can be written to a single file. This functionality requires NetCDF be installed (NetCDF is a standard library for reading and writing binary data).
631 - domino2 has been added. It provides a simpler interface on Keren's domino optimize as well as a number of new optimization and hooks for customizing the optimization process
632 - a better explanation of how to use reference counted pointers has been added to IMP::Pointer.
633 - all simple, planar polygons are now supported for display when CGAL is present
634 
635 ## Week of April 1, 2010
636 
637 - All plural objects in Python are now simply Python lists rather than C++ recreations of Python lists. This makes compilation of the wrappers much, much, much faster. The visible changes are
638  - types are not checked until the lists are passed to C++
639  - passing std::vector types by pointer and storing passed std::vector types by pointer will not work (but should never have been done anyway from code exposed to Python)
640  - swig has issues with some overloads (not always for obvious reasons). These are being sorted out. None are commonly used.
641  - get_particles() on a plural object doesn't work in Python (and is not needed in C++). Implicit conversions will always work.
642 
643 - All containers need to know the model at time of construction. If, after updating, you get an error on a container constructor, just add the model as the first argument. This also means that empty lists of particles in the constructors cause check failures. We can figure out a work around if this is too annoying.
644 
645 ## Week of March 13, 2010
646 - function get_rotation_in_radians_about_axis() was renamed to get_rotation_about_axis() as it didn't parse and we don't have the "in radians" elsewhere.
647 
648 - Added IMP::algebra::ReferenceFrame3D. As part of the cleanup, IMP::core::RigidBody::set_reference_frame() replaces set_transformation(). And the reference-frame related transformation construction functions were replaced. IMP::algebra::get_rotation_from_x_y_axes() was added to aid in the construction of reference frames (and replaces the algebra function taking three vectors).
649 
650 ## Week of March 6, 2010
651 - Publications now have a description field to give browsers an indication of what is in the publication to aid selection of what to read/cite.
652 
653 
654 ## Week of Feb 27, 2010
655 - IMP::algebra::decompose_rotation_into_axis_angle() was renamed to IMP::algebra::get_angle_and_axis().
656 
657 - renamed IMP::atom::bond() and IMP::atom::unbond() to IMP::atom::create_bond() and IMP::atom::destroy_bond() to be more consistent with things
658 
659 - renamed IMP::core::root() to IMP::core::get_root()
660 
661 
662 ## Week of Feb 21, 2010
663 
664 - IMP::Object::set_was_owned() was changed to IMP::Object::set_was_used() as that better reflects what the flag means.
665 - various more interpolation and endian methods had their names fixed (by adding \c get_)
666 - the function get_mean() on vectors was removed as it duplicated IMP::algebra::get_centroid()
667 - build_Transformation3D_from_transformation3D was replaced by IMP.algebra.get_transformation_3d()
668 - IMP::algebra::almost_equal() is now IMP::algebra::get_are_almost_equal() to fit the naming conventions.
669 - the random vector generators have been cleaned up to remove redundant ones and make the names more consistent. They now have the form of either
670  - IMP::algebra::get_random_vector_in() or IMP::algebra::get_random_vector_on() where there is one argument describing the volume or shape (such as an IMP::algebra::SphereD<4>).
671  - or IMP::algebra::get_uniform_surface_cover() where the first argument is the primitive whose surface we are covering and the second is the number of points.
672  Some good sed strings are
673  - "s\#random_vector_in_unit_sphere()\#get_random_vector_in(IMP.algebra.get_unit_sphere_3d())\#g"
674  - "s\#random_vector_in_unit_box()\#get_random_vector_in(IMP.algebra.get_unit_bounding_box_3d())\#g"
675  - "s\#random_vector_on_unit_sphere()\#get_random_vector_on(IMP.algebra.get_unit_sphere_3d())\#g"
676  - "s\#random_vector_on_unit_box()\#get_random_vector_on(IMP.algebra.get_unit_bounding_box_3d())\#g"
677  - "s\#random_vector_in_box(\#get_random_vector_in(\#g"
678  - "s\#random_vector_on_box(\#get_random_vector_on(\#g"
679  - "s\#random_vector_in_sphere(\#get_random_vector_in(\#g"
680  - "s\#random_vector_on_sphere(\#get_random_vector_on(\#g"
681  - "s\#grid_cover\#get_grid_surface_cover\#g"
682  - "s\#uniform_coverr\#get_uniform_surface_cover\#g"
683  .
684 - Spheres in IMP::algebra are now supported in general dimensions. See IMP::algebra::SphereD.
685 - Many methods in IMP::algebra were renamed to be more consistent with other methods or general conventions. Most of these simply involved adding \c get_ as a prefix and/or add a \c _3d suffix when they did not take any arguments.
686  - build_Rotation2D_from_Vector2D became IMP::algebra::get_rotation_to_x_axis()
687  - most of the Rotation3D and Transformation3D functions (in addition, IMP::algebra::get_identity_rotation_3d() and IMP::algebra::get_identity_transformation_3d() added the \c _3d so we can add 2D versions at some point)
688  - IMP::algebra::Transformation3D::get_transformed() and IMP::algebra::Rotation3D::get_rotated(). Replacing \c ".transform(" worked in the IMP code base.
689  - IMP::algebra::get_distance(), this also carries over to the IMP::core::get_distance()
690  The following global replacements are probably safe (in sed expressions)
691  - "s\#core.distance(\#core.get_distance\#g"
692  - "s\#algebra.distance(\#algebra.get_distance\#g"
693  - "s\#\.transform(\#.get_transformed(\#g" Only in C++
694  - "s\#\.rotate(\#.get_rotated(\#g" Only in C++
695  - "s\#rotation_from\#get_rotation_from\#g"
696  - "s\#rotation_in\#get_rotation_in\#g"
697  - "s\#identity_rotation(\#get_identity_rotation_3d("
698  - "s\#::distance(\#::get_distance(\#g"
699  .
700 - IMP::algebra::Vector3D and IMP::algebra::Vector3Ds were replaced by the expanded (non-typedef) version in functions parsed by SWIG as the current situation was extremely brittle. For the moment, you should not use the typedefs in definitions that swig sees. Sorry.
701 - IMP::algebra::Sphere3DPatch was renamed to IMP::algebra::SpherePatch3D to be consistent with the other geometry and allow us to support higher dimensions in the future.
702 - rigid_align_first_to_second was renamed to IMP::algebra::get_transformation_aligning_first_to_second() and build_Transformation2D_from_point_sets was renamed to IMP::algebra::get_transformation_aligning_pair()
703 
704 ## Week of Feb 14, 2010
705 
706 - The IMP::atom::Bond types were rearranged to reconcile the protein-centric classification with ligand support.
707 
708 - IMP::atom::Atom no longer has charge since we have IMP::atom::Charged.
709 
710 - IMP::atom::read_mol2() and IMP::atom::write_mol2() were added
711 
712 - IMP::atom::Selector and descendants renamed to IMP::atom::PDBSelector since we will soon have Mol2Selectors. In addition, the IMP::atom::IgnoreAlteriativesSelector is now IMP::atom::NonAlternativeSelector to keep the name positive, like the others. In the IMP body, all instances of the string Selector were for PDBSelectors and so sed was safe to use.
713 
714 - IMP::atom::read_pdb() lost the "ignore_alternatives" argument. Instead, all of the provided selectors, other than IMP::atom::AllSelector ignore alternatives internally. Code that uses the IMP::atom::AllSelector should probably change to the IMP::atom::NonAlternativeSelector.
715 
716 - IMP::example renamed (from examples)
717 
718 - The IMP_USAGE_CHECK() macro lost the (unused) third argument.
719 
720 ## Week of February 3, 2010
721 
722 - The model io code as moved into the kernel so it is now IMP::write_model() and IMP::read_model()
723 
724 - Remove the version_info argument from the implementation macros. Instead, just pick up the module version info automatically.
725 
726 - Add a function Object::get_type_name() and Object::get_module() to get the respective bits of information about each object.
727 
728 - Change Object::show() to be a non-virtual wrapper which prints the object name and other info before passing control off to Object::do_show().
729 
730 - The containers were moved to IMP.container. See that module page for a full list of what moved. In the process, the backwards compatibility functions were removed from IMP::container::PairsRestraint and kin.
731 
732 - IMP.atom.create_protein() now returns the created protein
733 
734 - the simplify functions have been moved from IMP.helper. They are now IMP.atom.create_simplified_along_backbone().
735 
736 - IMP.helper.clone() has been renamed to IMP.atom.create_clone()
737 
738 - IMP.misc.ConnectingPairContainer has been moved to IMP.core.ConnectingPairContainer
739 
740 - The locations for pixels in IMP::em::DensityMap have been moved to the center of the corresponding voxel from the lower left corner.
741 
742 - IMP::em::get_transformed(), IMP::em::get_resampled() and IMP::em::get_density() have been added.
743 
744 - The documentation has been majorly cleaned up and simplified. Please look around and see if you like it. There are now two main pages, the introduction and the developers guide, rather than the previous plethora of pages.
745 
746 - IMP::algebra::Cone3D::get_bounding_sphere() has been removed since it was not of general use, not returning the bounding sphere you would expect and was trivial to implement.
747 
748 - IMP::algebra::projection() was removed since it didn't actually project onto the segment and the name didn't follow the normal conventions. And it wasn't used.
749 
750 - Most geometry now has namespace methods IMP::algebra::get_bounding_box(), IMP::algebra::get_surface_area(), IMP::algebra::get_volume().
751 
752 ## Week of January 22, 2010
753 
754 - A generic geometry layer has been added (see \ref genericgeometry). Using it, IMP::atom::rmsd(),
755  IMP::algebra::alignment_taking_first_to_second() and the new IMP::atom::native_overlap() all
756  can work on either collections of either IMP::core::XYZ particles or IMP::algebra::Vector3Ds.
757 
758 - Deprecated functionality has been removed from IMP.core. For those still using it, IMP::core::ClosePairsScoreState was replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairsScoreState was replaced by IMP::core::CloseBipartitePairContainer.
759 
760 - The macros for Scores changed slightly. Scores not in IMP svn which use the IMP_*_SCORE macros now have to implement the get_is_changed() function outside of the class declaration and scores which use the IMP_SIMPLE_*_SCORE now don't have to implement the functions to get interactions and used particles.
761 
762 - Rename IMP::core::MinimumPairScoreRestraint (and kin, Maximum, Singleton, Triplet and Quad) to IMP::core::MinimumPairRestraint.
763 
764 - Add IMP::core::MinumumPairScore (and kin).
765 
766 ## Week of January 1, 2010
767 
768 - IMP::em is now using Particles to resample density maps and for fitting restraints. The previous ParticleAccessPoint mechanism is removed.
769 
770 - IMP::UsageException and IMP::InternalException are now reserved for their respective
771  checks (and cannot be thrown by IMP_THROW()). This makes the distinction between API
772  exceptions (which are thrown in any build mode and should be documented and tested)
773  and helper exceptions which are only thrown when checks are turned on (and hence
774  should not be documented) clearer.
775 
776 - IMP::core::RestraintSet moved to IMP::RestraintSet. The restraint sets now simply define a tree
777  with weights which the model traverses as opposed to being an opaque IMP::Restraint themselves.
778  This means that incremental evaluation of restraints in restraint sets now works and that
779  restraints in restraint sets can be individually evaluated. Existing working usage should not
780  have changed (so \c sed should manage updates just fine), but the internal
781  implementation is completely new, so pay attention.
782 
783 - IMP now has a text i/o layer consisting of IMP::TextInput and IMP::TextOutput. These classes are
784  convertible to C++ streams and implicitly constructable from both streams and file paths. As a result,
785  functions that take them will work when called with a string for a file path, a Python file or a
786  C++ stream.
787 
788 - The IMP.examples module now documents how to use the needed aspects of scons and SWIG.
789 
790 - The IMP.examples module is used as the basis for new modules created by the make-module script, so those include more documentation and are tested.
791 
792 - Module overview pages are now generated from the information passed in the SConscript file, rather than from the (now removed) modulename.dox file. See the examples file for an example of the things you can do there.
793 
794 - boost program options is now a checked for dependency.
795 
796 - The IMP::Model now uses dependency information from IMP::ScoreStates to automatically reorder the IMP::ScoreStates so that they are evaluated in the right order. Also, IMP::Model::evaluate() on a subset of the restraints now only updates the needed score states, not all of them.
797 
798 - Functions on rigid bodies no longer provide the IMP::core::RigidMemberRefiner as a default as it is, pretty much always, the wrong thing to use. That class may go away without notice in the future and is no longer documented.
799 
800 - the IMP::helper::create_simple_connectivity_on_rigid_bodies() and IMP::helper::create_simple_excluded_volume_on_rigid_bodies() now both take an IMP::Refiner as an (optional) final argument. This makes it so they can be used with the rigid bodies created from molecular hierarchies. This code still needs to be updated and the docs don't reflect what the code was or is doing, so use with caution (see notes in the header file).
801 
802 - The IMP::helper create* functions now all take non-Object inputs as const& to make clear what is input and that there are no memory management gotchas.
803 
804 - a new base class IMP::Constraint has been added for constraints. Various
805 other classes have been renamed to be consistent.
806 
807 ## Week of December 11
808 
809 - The interface of IMP::PairScore changed slightly to make the function signatures more uniform across various methods. Now all methods that take a pair of IMP::Particle objects take a const IMP::ParticlePair&. Previously some took that, some took an IMP::ParticlePair and some took two IMP::Particle*s. If you have an implementation of an IMP::PairScore outside of svn, you will have to update your function signatures (the compiler will tell you where). Backward compatible IMP::PairScore::evaluate() methods have been provided, so simple users of pair scores shouldn't see a difference.
810 
811 - IMP::core::PairRestraint takes a IMP::ParticlePair in the constructor instead of two IMP::Particle*s. A backwards compatible constructor can be added if needed.
812 
813 - IMP::ParticlePair::first and second went away since no one claimed to be using it and it complicated the implementation.
814 
815 - The generic restraints/score states/scores infrastructure has been extended to triplets and quads (for angles and torsion angles).
816 
817 ## Week of December 4
818 
819 - IMP now makes use of Boost.Filesystem to manipulate paths. Things may not work well on systems where boost libs are not available which does not support posix paths. But libless boost installs generally don't exist outside of our test systems.
820 
821 - new functions like IMP::core::get_example_path(filename) and IMP::core::get_data_path(filename) allow one to easily find a data file named 'filename' for the examples and library internal data.
822 
823 - The IMP::display module has been revamped in order to support non-linear geometry. VRML support has been dropped. CMM support will be dropped unless someone requests it (as IMP::display::ChimeraWriter does all it does and more).
824 
825 ## Week of November 19
826 
827 - imp.xcodeproj was moved into tools and updated accordingly
828 
829 - the bin directory finally went away
830 
831 - IMP::display::ChimeraWriter now supports the latest version of Chimera
832 
833 - IMP::display::CGOWriter was renamed to IMP::display::PymolWriter as it can do more than just CGO.
834 
835 - IMP::VectorOfRefCounted::operator[] can now be used for assignments just like a normal vector. No more need to use the IMP::VectorOfRefCounted::set() method.
836 
837 - the IMP::atom::Hierarchy constructor from an IMP::Particle* is now explicit (as were the decorator constructors)
838 
839 
840 ## Week of November 12
841 
842 - All exception types are now mapped to equivalent classes in Python. The
843 class hierarchy is similar (e.g. IMP::ModelException derives from IMP::Exception
844 in both C++ and Python) so all IMP exceptions can be caught in Python using
845 "except IMP.Exception". For convenience, the IMP::IOException,
846 IMP::ValueException and IMP::IndexException classes in Python also derive from
847 the similarly-named standard Python classes (IOError, ValueError and IndexError
848 respectively).
849 
850 ## week of October 29 Update
851 
852 - You can now add members to rigid bodies via IMP::core::RigidBody::add_member(). These members can be other rigid bodies (although this functionality is only lightly tested).
853 
854 - The code to get bounding volumes for IMP::atom::Hierarchy has been moved to the atom module, as as the clone and destroy code. They can be found at IMP::atom::clone(), IMP::atom::destroy(), IMP::atom::get_bounding_box().
855 
856 - Added IMP::core::VolumeRestraint.
857 
858 - Several functions have been added to helper in creating restraints. They are create_simple_connectivity_on_rigid_bodies(), create_simple_connectivity_on_molecules(), create_simple_distance(), create_simple_diameter(), create_simple_em_fit(), and create_simple_excluded_volume_on_rigid_bodies(). Functions load_em_density_map() and set_rigid_bodies() are added as well. See IMP::helper for more information.
859 
860 
861 - IMP::ScoreStates are further hidden. Namely, IMP::core::RigidBody constructs the needed score states internally, doing away with the need for the create functions (which have been done away with). In addition there are new decorators IMP::core::Cover and IMP::core::Centroid which manage their respective score states, further doing away with more create methods. The hierarchy-based create method for rigid bodies has been moved to the atom module and is called IMP::atom::rigid_body_setup_hierarchy().
862 
863 - IMP::em::read_map() will figure out the file type to use automatically from the extension. Other such methods should follow suite.
864 
865 - A class IMP::core::StructureSet has been added to provide a convenient way to manage solutions of optimization runs.
866 
867 - IMP::core::RigidBodyMover now takes its angle argument in radians and is documented as such.
868 
869 - IMP::core::ClosePairsScoreState and IMP::core::CloseBipartitePairsScoreState have been replaced by IMP::core::ClosePairContainer and IMP::core::CloseBipartitePairContainer. The score states have been marked as deprecated and are mostly still there, but are more limited than they were.
870 
871 - IMP::core::AllPairsPairContainer has been renamed to IMP::core::AllPairContainer and a bipartite version has been added.
872 
873 - All containers now keep track of what they looked like on the previous time step.
874 
875 - The IMP::core::ListPairContainer and IMP::core::ListSingletonContainer have had their interface reduced a bit in order to make implementing change tracking easier.
876 
877 - The set of exceptions has been simplified to go along with simplification of the checks. We now have IMP::InternalException and IMP::UsageException which are thrown by the respective checks (and the former is thrown by IMP_FAILURE()). When you want to to pass an exception back to Python use one of IMP::ValueException or IMP::IndexException or IMP::IOException. There is also a new macro, IMP_THROW() to be used when you want to throw an exception with a nice error message. Please look at code you have worked on and make sure that the exceptions make sense given the new clarifications.
878 
879 - An IMP::ModelException is now thrown when IMP::Particle attributes attain invalid values. This exception can be caught to recover from failures of IMP::core::ConjugateGradients or difficult restraints.
880 
881 ## Week of October 15
882 
883 - The set of functions needed to get meta information about IMP::Restraint and IMP::ScoreState objects has hopefully stabilized. Sorry about the thrashing. If you have restraints or score states not in SVN, see the corresponding IMP_RESTRAINT type macros, the IMP::Interaction class and base classes for what exactly you need to implement.
884 
885 ## Week of October 8
886 
887 - IMP::Interaction was modified to split IMP::interaction::get_used_particles() into IMP::interaction::get_read_particles() and IMP::interaction::get_write_particles(). This makes it possible to check for IMP::ScoreState objects being in an inconsistent order (so that the input to one is updated after it is read). Later, we will be able to automatically order the score states ensure that there are no problems.
888 
889 - IMP::atom::Hierarchy has been modifier to get rid of the IMP::atom::Hierarchy::Type. Instead, all types are contained in the decorators. IMP::atom::get_by_type() was also updated to reflect this.
890 
891 - IMP_check() and IMP_assert() have been
892 replaced with IMP_USAGE_CHECK() and IMP_INTERNAL_CHECK()
893 respectively. Hopefully this makes it clearer when to use
894 which. IMP::set_check_level() has been revised accordingly (to use
895 IMP::USAGE and IMP::USAGE_AND_INTERNAL). Also, IMP_failure() and
896 IMP_not_implemented() have been renamed to IMP_FAILURE() and
897 IMP_NOT_IMPLEMENTED() to make all macros be all caps. To update
898 external code,
899  - replace IMP.EXPENSIVE with IMP.USAGE_AND_INTERNAL
900  - replace IMP.CHEAP with IMP.USAGE
901  - replace IMP_assert and IMP_check as above.
902 
903 ## Week of October 1
904 - IMP::ScoreState and IMP::Restraint now better track the set of particles that they use and the dependencies between the particles. To this end, a new base class IMP::Interaction has been added with methods to get the inter-particle interactions and the set of particles used.
905 
906 - .cpp, .h, .py and SConscript files which contain the string "DO NOT COMMIT" will be rejected on checkin. Use this to make experimental or debugging changes you are making so you don't accidentally commit them.
907 
908 - The search module is no longer built by default and is not longer supported. It will get moved out of svn shortly.
909 
910 - IMP::ScoreState objects now also have IMP::Interacting::get_interacting_particles() methods. Soon IMP::PairScore and IMP::SingletonScore objects will too.
911 
912 ## Week of September 27
913 
914 - Some support for gather statistics has been added to IMP::Model (see IMP::Model::set_gather_statistics()). If requested, an API can be provided so that the individual data points can be returned to programs.
915 
916 - A class IMP::core::RigidBodyDistancePairScore has been added which provides accelerated computations for the distance between two rigid bodies.
917 
918 - The SWIG input files were renamed from swig.i to swig.i-in to aid in dependency tracking.
919 
920 - rigid bodies now have two sets of particles associated with them in the various places they are used. One is the IMP::core::RigidBodyMember particles, as was previously the case. The other is defined through a refiner (and could be the RigidBodyMember particles if a IMP::core::RigidMembersRefiner is used). The idea is that you sometimes want some particles which are transformed along with the rigid body, but which are not really part of the representation of the rigid body. Examples would include marker points or multiresolution representations of proteins.
921 
922 ## Week of September 20
923 
924 - all \c .i files have been renamed to start with \c IMP_ to help make dependency tracking easier.
925 
926 - static builds (builds with \c static=True) build both the static and dynamic libraries and just static executables
927 
928 - the \c profile build target has been removed as \c static=True does the important bit of what is needed. Just do \c linkflags=['-pg'] to add the needed link flag for \c gprof.
929 
930 - Restraints should now implement IMP::Restraint::unprotected_evaluate() instead of IMP::Restraint::evaluate(). The latter can now be called outside of IMP::Model::evaluate() and will return the right answer always.
931 
932 - \c bin/imppy.sh no longer works, as was previously warned. Use \c tools/imppy.sh instead. The bin directory will go away eventually, for now the script just tells you to use the tools version.
933 
934 ## Week of August 30
935 
936 - A new score IMP::core::BoundingBox3DSingletonScore has been added to keep particles
937  in a box.
938 
939 - str should now work better in Python when applied to IMP::Object objects and a few others. Objects
940  should use the new IMP_SHOWABLE and IMP_SHOWABLE_INLINE macros instead of declaring show
941  functions directory to make it so that they work in Python. It is now a bug to \c print or \c str
942  to not return something useful in Python for any IMP object.
943 
944 ## Week of August 23
945 
946 - each module now has a version, eg IMP::get_module_version_info() which is also returned by the various restraints of the module. The Python code checks that loaded libraries have the expected version numbers to try to provide better messages when runtime linking problems occur.
947 
948 - the various bounding_box methods have been renamed to ::get_bounding_box() to be more consistent with the other names.
949 
950 - Python documentation is now parsed by doxygen
951 
952 ## Week of August 15
953 
954 - The Object::get_version_info() result now includes the SVN revision info when building from SVN. This means you can use it to more easily tell exactly what a built version of IMP was built from.
955 
956 ## Week of August 8
957 
958 - Treatment of hetatoms in hierarchies (and hence pdbs) is now documented. See IMP::atom::Hierarchy. This involved broadening IMP::atom::Residue to handle ligands and having IMP::atom::Hierarchy::LIGAND, IMP::atom::Hierarchy::AMINOACID and IMP::atom::Hierarchy::NUCLEICACID be handled by IMP::atom::Residue.
959 
960 - IMP::atom::read_pdb() currently drops HOH ATOM records on the floor. We should add a IMP::atom::Water decorator if anyone wants to handle them (since they aren't atoms and don't have elements).
961 
962 - IMP::atom::Mass() has been added for decorating particles with mass. IMP::atom::Atom mass has been moved to that.
963 
964 - The scheme for generating atom types is now fully specified. See the IMP::atom::AtomType documentation.
965 
966 - IMP::atom::Atom and IMP::atom::Residue have been somewhat restructured. Some of the helper functions have been renamed to be more descriptive or removed (if they duplicated other easily accessible functionality). get_residue_index() has been removed (since you can just call get_index() on the return result of the IMP::atom::get_residue()). IMP::atom::get_chain() now returns an IMP::atom::Chain. root() was renamed to IMP::atom::get_root() for consistency.
967 
968 - SWIG files are now generated from a file called swig.i in the pyext directory. This file only has to contain module-specific commands such as lines to wrap headers and instantiate module templates. No more boilerplate.
969 
970 - The contents of bin are moving to tools since the two directories both contained assorted development tools
971  (not all of which were executable). For now, bin/imppy.sh will still work, but that will go away soon.
972 
973 - IMP::Decorator::show() now just takes the stream (it used to take a prefix also, which was often ignored).
974  Anyone who implemented a decorator will have to delete the second argument of their show function.
975 
976 ## Week of August 1
977 
978 - Containers now have methods IMP::SingletonContainer::evaluate(), IMP::PairContainer::evaluate(),
979  IMP::SingletonContainer::apply() and IMP::PairContainer::apply(). Using these can be significantly
980  faster than looping over the particles externally. Similarly, there is IMP::SingletonScore::evaluate(),
981  IMP::PairScore::evaluate(), IMP::SingletonModifier::apply(), IMP::PairModifier::apply().
982 
983 - The decorator functions have been renamed to IMP::Decorator::setup_particle(), IMP::Decorator::decorate_particle()
984  and IMP::Decorator::particle_is_instance(). A shell script \c tools/update-decorators is provided which
985  updates the names in all code in subdirectories of the current directory. Please backup things first.
986 
987 - a function, IMP::Restraint::safe_evaluate() has been added to be used
988  when requesting the value of restraints outside of the IMP::Model::evaluate().
989 
990 ## Week of July 18
991 
992 - The atom::BondPairContainer has been renamed to atom::BondPairFilter
993  since it wasn't a real container. A real atom::BondPairContainer has
994  been added.
995 
996 - Out of source builds mostly works. To do an out of source build, make a link
997  to \c scons_tools and \c SConstruct from the source directory in the build
998  directory and then add the line \command{repository="../path_to_src_dir"} in
999  the \c config.py in the build directory. Then running \c scons in the build
1000  directory will put generated files there. This allows you to make both a
1001  \c release and \c fast build off of the same set of sources. Proper docs
1002  will be added soon.
1003 
1004 - The SConscript setup for a module has been greatly simplified so it should
1005  be easy enough for users to maintain their own module and updates to the
1006  build system shouldn't involve changes to user SConscripts any more. This does
1007  mean that any non-svn modules need to be updated. Look at IMP.core or another
1008  module and copy what is there. Again, proper docs will be added soon.
1009 
1010 
1011 ## Week of July 11
1012 
1013 - Per Javi's suggestion, the conventions have been split into
1014  "Coding conventions" and "API conventions" and linked from the main page.
1015 
1016 - benchmarks got moved to their own module, IMP::benchmark and will
1017  now be tracked over time
1018 
1019 ## Week of July 4
1020 
1021 - FilteredListSingletonContainer and FilteredPairSingletonContainer
1022  went away as there was too much hard to factor code shared with
1023  IMP::core::ListSingletonContainer and IMP::core::PairSingletonContainer. The
1024  IMP::core::ClosePairsScoreState is now handled by a list of filters
1025  on the class itself. That is, do
1026  IMP::core::ClosePairsScoreState::add_close_pair_filter() to add a
1027  filter.
1028 
1029 - Hooks have been added to allow actions to be taken on an assertion
1030  or check failure. See IMP::HandleFailure and its descendants for
1031  more details.
1032 
1033 - IMP::display::CGOWriter has been much improved and now allows
1034  provides lots of display-time control and easy building of
1035  animations.
1036 
1037 - Generic optimizer states have been added
1038  (IMP::core::SingletonsOptimizerState and
1039  IMP::core::PairsOptimizerState).
1040 
1041 ## Week of June 15
1042 
1043 - box distributions (IMP::algebra::random_vector_on_box(),
1044  IMP::algebra::random_vector_in_box()) take bounding boxes
1045 
1046 - IMP::em::DensityMap are now IMP::Object objects. Some current
1047  allocations on the stack may need to be changed (since ref counted
1048  objects can't be stack allocated).
1049 
1050 - IMP::em::DensityMap::Read() is now deprecated (as is
1051  IMP::em::DensityMap::Write) as they violated a slew of
1052  conventions. Use IMP::em::read_map() and IMP::em::write_map()
1053  instead.
1054 
1055 
1056 ## Week of June 8
1057 
1058 - new methods IMP::atom::add_radii() and IMP::atom::add_bonds();
1059 
1060 - Hierarchys to IMP::atom::Hierarchies
1061 
1062 - IMP::Particles has been changed to a new type. It no longer plays
1063  quite so well with Python lists (you can't create one from an
1064  arbitrary Python list).
1065 
1066 - The containers for multiple decorators have been revised to make them
1067  more useful. See IMP::Decorators. You can now implicitly convert,
1068  for %example, a IMP::core::XYZs to an IMP::Particles (as well as use
1069  IMP::Decorators::get_particles()). All methods that take a
1070  collection of particles with a certain type now instead take a
1071  IMP::Decorators.
1072 
1073 - A method that has expectations about the particles it takes should
1074  take decorators rather than particles. That is, anything which
1075  expects IMP::core::XYZ particles, should take an IMP::core::XYZs
1076  instead of IMP::Particles.
1077 
1078 - IMP_LIST() now takes an extra argument (the name of the plural
1079  container)
1080 
1081 - In C++ IMP::Decorator objects are implicitly convertible to
1082  IMP::Particle* objects. Unfortunately, making this work in Python is
1083  problematic.
1084 
1085 - More methods take and return decorators rather than
1086  IMP::Particles. This means it is clearer what their expectations are
1087  and what they produce.
1088 
1089 ## Week of May 15
1090 
1091 - The IMP::helper module was added and various functionality moved
1092  there. See its documentation page for a list of what is there.
1093 
1094 - IMP::algebra::SphericalCoords was renamed to
1095  IMP::algebra::SphericalVector3D to be consistent with other algebra
1096  classes.
1097 
1098 - IMP::algebra::Rotation2D is now IMP::InvalidDefault to be consistent
1099  with the other similar classes.
1100 
1101 - We are trying to be a bit more specific about what is expected from
1102  a IMP::atom::Hierarchy. Towards that end, there are two methods for
1103  extracting a representation from a hierarchy,
1104  IMP::atom::get_detailed_representation(), and
1105  IMP::atom::get_simplified_representation()
1106 
1107 - A method has been added to produce a simplified copy of an
1108  IMP::atom::Hierarchy. The method is still a bit crude and could
1109  benefit from other people inspecting the output and saying how it
1110  should be improved.
1111 
1112 - There is now an IMP::algebra::BoundingBoxD (and
1113  IMP::algebra::BoundingBox3D) to track bounding boxes. There are
1114  methods like IMP::algebra::bound_box(Sphere3D) which gets the
1115  bounding box of a sphere.
1116 
1117 - Each IMP::Object has its own log level. If
1118  IMP::Object::set_log_level() is called with a value other than
1119  IMP::DEFAULT, the value is used for code within and called by the
1120  object. To support this in your objects you need to add
1121  IMP_OBJECT_LOG to the start of each major method. It has been added
1122  to some of the core objects.
1123 
1124 ## Week of May 8
1125 
1126 - IMP::atom::MolecularHierarchy has been renamed to
1127  IMP::atom::Hierarchy. Use \command{sed -i.old
1128  "s\#MolecularHierarchy\#Hierarchy\#g" my_files}
1129 
1130 - the long-deprecated Name(Decorator) was removed
1131 
1132 - The decorator suffix has been removed from decorator names.
1133 
1134 - Unit conventions have been established and documented for distance,
1135  energy, force and angle.
1136 
1137 - there is a function IMP::atom::create_protein() to create a coarse
1138  grained protein. The function is crude so far, but the idea is that
1139  we should have standard ways of generating such things so that
1140  everyone can benefit from best practices.
1141 
1142 - a class IMP::atom::BondGraph has been added so Boost.Graph
1143  algorithms can be used with bond graphs
1144 
1145 - the IMP-specific doxygen commands have been documented.
1146 
1147 - the doxygen commands unstable{name} and untested{name} have been
1148  added to doxygen to be used for marking methods and functions as
1149  likely to change or not well tested. Pass the name of the class or
1150  method as name.
1151 
1152 - bonds can now be added to molecular hierarchies automatically see
1153  IMP::atom::add_bonds()
1154 
1155 ## Week of May 1
1156 
1157 - There is a IMP::Rotation2D
1158 
1159 - You can now extract fixed Euler ZYZ angles from a
1160  IMP::Rotation3D. This code can serve as a template for extracting
1161  other Euler angle conventions if they are needed.
1162 
1163 - IMP::algebra::rotation_between_two_vectors() changed to
1164  IMP::algebra::rotation_taking_first_to_second() to make it clear
1165  which direction the rotation goes.
1166 
1167 ## Week of April 27
1168 
1169 - IMP::Decorator derived objects are now convertible to bool. So code
1170 like \code if (d) { blah } \endcode works.
1171 
1172 - IMP::Decorator::cast now returns a null decorator (\c
1173  DecoratorName()) rather than throwing an exception. This might make
1174  it actually useful.
1175 
1176 - New IMP::statistics module for statistical methods
1177  (e.g. clustering).
1178 
1179 ## Week of April 20
1180 
1181 - Support for taking advantage of rigid bodies in collision detection
1182  has been added via the IMP::core::RigidClosePairsFinder.
1183 
1184 - All IMP::core::ClosePairsScoreState-related functionality has been
1185  optimized and should be noticeably faster.
1186 
1187 - The EMBED project has been merged into the IMP::em module. Thus, it
1188  is no longer necessary to obtain EMBED separately to build IMP::em.
1189 
1190 - All C++ methods which work with std::istream or std::ostream
1191  arguments should now be usable from Python: any Python file-like
1192  object can be used.
1193 
1194 - The SWIG interface now ties in to the reference counting mechanism
1195  used in the C++ code, for all classes derived from
1196  IMP::RefCounted. So, for %example, deleting a Python object should
1197  not free the underlying C++ object if another C++ object holds a
1198  reference to it, and unrefing a C++ object should not delete it if a
1199  Python object still holds a reference to it. There should be no
1200  further need to manually keep references to Python objects to
1201  prevent things from being prematurely cleaned up. This should be
1202  transparent to the user; please report any problems.
1203 
1204 - IMP::atom::AtomDecorator no longer inherits from
1205  IMP::core::XYZDecorator.
1206 
1207 ## Week of April 13
1208 
1209 - The decorators IMP::atom::AtomDecorator, IMP::atom::ChainDecorator,
1210  IMP::atom::ResidueDecorator etc. now inherit from
1211  IMP::atom::MolecularHierarchyDecorator. This means they can set
1212  their IMP::atom::MolecularHierarchyDecorator::Type properly and
1213  saves a lot of typing in the common case. We can always make
1214  hierarchy decorators zero memory overhead if we find that the cost
1215  matters.
1216 
1217 - The base classes IMP::Comparable, IMP::NullDefault,
1218  IMP::ValidDefault and IMP::UninitializedDefault were added to
1219  document what the default constructors do for various objects in
1220  IMP.
1221 
1222 ## Week of April 6
1223 
1224 - IMP::Object now defines pure virtual methods
1225  IMP::Object::get_version_info() and IMP::Object::show(). All
1226  non-abstract classes which inherit from IMP::Object must implement
1227  them. Doing this means they are more uniformly implemented and are
1228  documented in one place.
1229 
1230 - IMP::Object is now reference counted (the only objects this changes
1231  are IMP::Model and IMP::Optimizer). As a result optimizers and
1232  IMP::Model objects must be created on the heap.
1233 
1234 - IMP::RefCounted has had its name changed and no longer inherits from
1235  IMP::Object. This makes it easier to use independently.
1236 
1237 - the docs for reference counting (IMP::RefCounted) and and IMP::Decorator
1238  have been improved
1239 
1240 - running doxygen produces no warnings. Let's try to keep it that
1241  way. Run \command{scons doc > /dev/null} before you check in large
1242  blocks of new code to make sure that you didn't introduce warnings.
1243 
1244 - Helper classes have been added to make SingletonModifier and
1245  PairModifier objects look like functors for C++ and Python. This
1246  makes it so you can use \c map in Python and \c for_each in C++. See
1247  IMP::SingletonFunctor and IMP::PairFunctor for more information.
1248 
1249 - The recent changes page has moved to a doxygen page to make it
1250  easier to link changes back to the relevant source code.
1251 
1252 ## Week of March 20:
1253 
1254 - create_rigid_body(), create_rigid_bodies(), create_centroid()
1255  etc. now require you to add the score state to the model
1256  yourself. Whether the rigid body is enforced or not can be
1257  controlled by adding or removing the returned IMP::ScoreState.
1258 
1259 - All previously deprecated code has been removed. This includes
1260  indexes in particles, the no-argument constructor for particles and
1261  the functionality that was moved from core to atom.
1262 
1263 - The IMP::core::RigidBodyDecorator and
1264  IMP::core::RigidMemberDecorator have been simplified by removing the
1265  traits classes. There is now only one type of rigid body and it
1266  stores the list of member particles internally.
1267 
1268 - IMP::core::NameDecorator has been deprecated as IMP::Particles have
1269  built in names (use IMP::Particle::get_name(),
1270  IMP::Particle::set_name()).
1271 
1272 - IMP::atom::ChainDecorator has been added
1273 
1274 ## Week of March 13:
1275 
1276 - A new IMP::display::Writer, IMP::display::ChimeraWriter has been
1277  added which exports geometry to native Chimera constructs. It
1278  probably should be used instead of bild and cmm for most purposes as
1279  it combines the best of both types.
1280 
1281 - compose() for transformation is now a free function so
1282  compose(rotation, vector) will work via promotions.
1283 
1284  - the geometry classes have been cleaned up and are now more
1285  consistent with IMP standards (and more likely to do what their
1286  names/comments suggest). See IMP::algebra::Sphere3D,
1287  IMP::algebra::Cylinder3D, IMP::algebra::Segment3D,
1288  IMP::algebra::Plane3D etc.
1289 
1290 - ParticleRefiner has been renamed to IMP::Refiner since the Particle
1291  was long and unnecessary.
def temporary_working_directory
def temporary_directory