Changelog

Adapted from keepachangelog.com for org-mode.

See the git commit log history for full details.

1.1.0 - 2021-03-22

  • Fixes uninitialized discontinuities when not using boundary conditions in a ContigTree. (Fixes: 10.ghissue)

  • Fixes mistake in the ContigTree.lineages function. (Fixes: 10.ghissue)

  • Adds an option to account for discontinuities in ContigTree.lineages.

  • Fixes the WepyHDF5.observable_field_names function.

  • Fixes the ContigTreeProfiler functions for partitioning FE profiles into tranches: _tranche_cumulative_partitions

  • Fixes the setting of continuations in ContigTree initialization.

  • Fixes computation of ContigTreeProfiler.fe_profile_all.

  • Fix the arguments to the mapped function in MacroStateNetwork.node_fields_map

  • Removed all implicit file opening in MacroStateNetwork except for in the BaseMacroStateNetwork constructor. This allows for parallel processing.

  • Improves calculations of transition counts when constructing BaseMacroStateNetwork objects. Now instead of only calculating the “weighted counts” for edges the “unweighted counts” are also calculated and the full set of transition sample weights are also saved as edge attributes. This change in interface is the reason for the breakages and regressions, see below.

  • Addition of the edge_attribute_to_matrix to the BaseMacroStateNetwork (and MacroStateNetwork) classes. This generalizes and supercedes the use of the precomputed countsmat and probmat attributes. And makes the source-to-sink orientation of the matrix explicit and user controlled.

  • Addition of edge attribute getting methods similar to the node ones: get_edge_attribute, get_edge_attributes, get_edges_attributes.

  • Fixes the setting of the platform in the OpenMMRunner when the pre_cycle() is called without a platform.

  • Moves the set_loglevel function into wepy.util.util thus making it a general purpose function rather than a helper to the orchestration CLI (see breakages, but we don’t expect anyone to have used this and is minor).

Breakages

  • The two attributes countsmat and probmat for (Base)MacroStateNetwork classes have been removed, and superceded by the edge_attribute_to_matrix method.

  • The values of the columns of the nodes and edges dataframe have changed. They no longer have the ‘Weight’ column, which was for support of gephi tables, and was too tight a coupling. The edges dataframe will no longer have the ‘transition probability’ or ‘type’ columns. The latter for similar reasons and the former because this value is no longer calculated.

  • The set_loglevel function was removed from wepy.orchestration.cli to wepy.util.util, risk of breaking other peoples code is low (we can put a stub in for backwards compatibility if needed).

Regression

  • As stated in the Growth items, a potentially large amount of transition sample weights are stored in the BaseMacroStateNetwork object now.

1.0.7 - 2020-11-12

  • Adds a check to make sure the platform for the runners.openmm.OpenMMRunner is a string and not an actual Platform object.

  • Fixes the tranchecumulativepartition function in the analysis.profiles module.

  • Pins version of h5py to less than version 3, which introduced breaking API changes.

1.0.6 - 2020-09-28

  • Fixes the WepyHDF5.get_trace_field function.

    Previously I had implemented a chunking strategy to improve performance for situations with very long traces. This has a bug somewhere in it. For the time being we have just reverted this optimization to the naive (but correct) approach. If an optimized version is introduced it will be made explicit (either in the get_trace_field call, or as a wrapper or separate method).

1.0.5 - 2020-09-14

  • Work around fix for final snapshot in orchestrator being deleted if it is the same hash as the last checkpoint snapshot.

1.0.4 - 2020-08-27

  • Fixes the broken CLI interface. Moves it into the standard __main__.py module.

1.0.3 - 2020-08-19

  • This fixes an issue with the MacroStateNetwork that allows for

different shapes of state assignment arrays. This was previously an issue but did not completely solve the issue.

1.0.2 - 2020-08-13

  • Fixes too-large chunk size when copying observable run data to other files in WepyHDF5.copy_run_slice

  • Improves docstrings for the MacroStateNetwork methods

1.0.1 - 2020-08-07

This fixes an issue with the MacroStateNetwork that allows for different shapes of state assignment arrays.

1.0.0 - 2020-08-04

Changes in rc6:

  • adds feature to make weighted transition counts from simulation data. This was also made the way that the MacroStateNetwork calculates the counts and transition probabilities matrices.

  • Fixed initialization errors in the TaskMapper which would cause a zombie job.

  • Fixed implementation of the CPU Task process for OpenMM.

  • Fixed sim maker details in wepytools

Changes in rc5:

  • adds simulation monitoring via prometheus (or whatever) as a feature

Changes in rc4:

  • fixes bugs introduced to work mappers from rc2

Changes in rc3:

  • removed integration with openmmtools.testsystems in favor of openmm_systems fork

Changes in rc2:

  • adds support for setting the OpenMM Platform at run_segment or in pre_cycle

  • add support for parametrizing apparatus components (runner, resampler, or BC) in the orchestration Configuration object.

  • Fixes wepy_tools.sim_makers for no longer default arguments for resamplers.

  • minor bugfixes found during testing.

Changes in rc1:

  • fixes bug in WorkerMapper

1.0.0rc1 - 2020-04-09

Changes in rc2:

  • adds support for setting the OpenMM Platform at run_segment or in pre_cycle

  • add support for parametrizing apparatus components (runner, resampler, or BC) in the orchestration Configuration object.

  • Fixes wepy_tools.sim_makers for no longer default arguments for resamplers.

  • minor bugfixes found during testing.

Changes in rc1:

  • fixes bug in WorkerMapper

1.0.0rc0

Prehistory