BETSE 0.8.0 (Kind Kaufmann) released.
Significant changes include:
* Official Anaconda release. This is the first major release of BETSE to
officially support our new conda-forge-hosted Anaconda package at:
https://anaconda.org/conda-forge/betse
* Installation instructions simplified. Installation instructions in our
front-facing "README.rst" documentation have been substantially
simplified to reflect our official Anaconda release.
* Fast simulation solver added. As an efficient alternative to the usual
full simulation solver, the newly developed fast solver implements a
less space- and time-intensive formalism for simulating a subset of
bioelectric phenomena. Specifically, the fast solver leverages the
well-known equivalent circuit model to simulate core bioelectricity
optionally coupled to a gene regulatory network (GRN). As a crude
heuristic, cursory inspection shows the fast solver to be
approximately 500—2000% faster than the full solver. Bioelectric
phenomena supported by the fast solver includes:
* Transmembrane potential (Vmem).
* Voltage-sensitive gap junctions.
* Intracellular currents and fields.
All other bioelectric phenomena should be assumed to be unsupported.
* Microtubule density maps and initial-condition alignment maps. These
maps are user-defined image mask files permitting any simulation to
preconfigure various properties pertaining to microtubules for that
simulation's cell cluster.
* Morris-Lecar ion channels. Ion channels based on the Morris-Lecar
formalism (including a so-called "slow" Morris-Lecar channel) have now
been added.
* Biomolecule degradation activation and inhibition. Degradation of
arbitrary biomolecules in gene regulatory networks (GRN) may now be
configured via customary activation and inhibition relations.
* Surgical interventions generalized. Cell clusters may now be
surgically cut at any arbitrary time step of a simulation. Previously,
surgical cuts were only supported at the first time step.
* "sim-grn" dynamic cutting support. The "betse sim-grn" subcommand
simulating a gene regulatory network (GRN) isolated from bioelectric
phenomena now supports all standard surgical interventions.
* Extracellular matrix (ECM) handling finalized. Main
electrophysiological properties of our standard planarian model have
now been successfully matched to experimental observation (e.g., as
compiled by several prominent planarian theses), validating our
handling of the extracellular matrix (ECM). These properties include:
* Transepithelial potential (TEP) level.
* Transepithelial potential (TEP) under an applied exogenous field.
* Current magnitude and direction.
* Lloyd's Voronoi mesh optimization. The pseudo-random Voronoi mesh
underlying each cell cluster has now been optimized using Lloyd's
algorithm, improving simulation accuracy and substantially reducing
the likelihood of mesh errors inducing computational instabilities.
* Realistic cell polarizability algorithm. An improved algorithm for
simulating the polarizability of cells has now been implemented.
* Microtubule Gaussian-noise distribution. Microtubules may now be
distributed according to a proper Gaussian-noise distribution.
* Numerous locomotion issues resolved, including:
* Mass conservation issues in motor protein transport.
* Gap junction-isolated profiles.
* Microtubule transport.
* Laplacians, including:
* Cell-grid Laplacians.
* Voronoi grid Laplacians.
* Comma-separated value (CSV) file exports pipelined. CSV files are now
exported via our standard pipeline infrastructure much like all other
exports (e.g., plots, animations), enabling the "plot init" and
"plot sim" subcommands to trivially support the fast solver.
* Seed exports generalized. All exports previously created via only the
"betse plot seed" subcommand are now general-purpose exports created
via all plotting subcommands (e.g., both the "betse plot init" and
"betse plot sim" subcommands). These exports includes:
* "cluster_mask" for the all-cells image mask defining the cell
cluster shape.
* "diffusion_extra" for final all-cells extracellular diffusion
weights (logarithm-scaled). Ignored if extracellular spaces are
disabled.
* "gj_connectivity" for the all-cells gap junction connectivity network.
* "gj_permeability" for final all-cells gap junction relative
permeabilities.
* "tissue_cuts" for all-cells tissue and cut profiles.
* Simulation configuration file documentation improved. To simplify
usage, the configuration file now documents all simulation features
required by each type of export -- which usually means the full solver.
* Simulation configuration creation generalized. Simulation
configurations created by the "betse config" subcommand now ignore all
existing files in target subdirectories (e.g., "geo/"), permitting
users to create multiple configurations in the same directory.
* Simulation runner callback API. The new
"betse.science.phase.phasecallbacks" class hierarchy provides a
general-purpose mechanism for interested callers (notably, the BETSEE
GUI) to register low-level callbacks with instances of the high-level
"SimRunner" class responsible for running simulation subcommands and
hence phases. Currently, these callbacks include methods to receive
periodic updates on simulation progress. The three principal
simulation subcommands (i.e., seed, initialization, and simulation)
now inform callers of incremental progress with respect to work
performed.
* Simulation runner API generalized. The SimRunner.__init__() method now
accepts a "Parameters" object rather than a simulation configuration
filename, enabling callers (notably, the BETSEE GUI) to reconfigure
simulation parameters in-memory without writing changes back to disk.
* Simulation phase API sanitized. The "Simulator.dyna" and
"SimPhase.dyna" tissue handlers are now synonyms rather than
duplicates of one another, improving sanity throughout the codebase.
Although the former has been temporarily preserved, all code requiring
a tissue handler should now use the canonical "phase.dyna" object.
* Script API removed. The "betse.script" subpackage has been removed.
This API had begun to unsafely bit-rot, had remained unused by end
users for several years, and could no longer be reasonably maintained.
It is dead.
* REPL API disabled. The "betse.cli.repl" API has been moved to
"betse.util.cli.repl" and temporarily disabled until further notice.
This API had also begun to unsafely bit-rot, but could yield tangible
benefit at some future date. This API has been preserved in cold
storage until it can be resuscitated by a future development effort.
It isn't dead; it's merely sleeping for eternity.
* Simulation phase requirements API generalized. The existing
"betse.science.phase.require" subpackage has been refactored to
leverage immutable sets, permitting two or more lower-level simulation
phase requirements to be readily composed into a high-level simulation
phase requirement.
* KDTree -> cKDTree optimization. For efficiency, all prior usage of the
pure-Python "KDTree" class has been refactored to leverage the C-based
"cKDTree" class instead. The latter is a drop-in replacement for the
former commonly providing speedups in kd-tree queries of up to 4000%.
Doing so also permits all kd-tree instances to be trivially pickled.
* Numpy array validation API. The new "betse.lib.numpy.nptest" submodule
now provides optimized routines for testing and validating Numpy
arrays -- notably, whether a given array contains one or more
Not-a-Number (NaN) values or not.
* Numpy-based image loading API improved. The existing
"betse.lib.pil.pilnumpy" submodule deserializing arbitrary image files
into Numpy arrays has now resolved a critical defect pertaining to
signed versus unsigned data types, reducing the likelihood of
computational errors on the resulting Numpy arrays.
* Simulation phase enumerations isolated. For maintainability, the central
"betse.science.phase.phasecls.SimPhaseKind" enumeration has been shifted
into the newly created "betse.science.phase.phaseenum" submodule.
* @log_time_seconds decorator. All simulation subcommand methods (e.g.,
betse.science.simrunner.SimRunner.seed()) are now decorated by the new
@log_time_seconds decorator defined by the new
"betse.util.type.decorator.decprof" submodule, logging wall clock time
in fractional seconds consumed by each such subcommand.
* @abstractproperty decorator de-deprecated. Python >= 3.3 deprecated
the @abc.abstractproperty decorator, for unwise and frankly
incomprehensible reasons. This release locally reverts this
deprecation by restoring the @abstractproperty decorator in perpetuity
under the rubric of the "betse.util.type.decorators.deccls" submodule.
* @type_check decorator generalized. The @type_check decorator widely used
throughout the codebase to perform type checking of callable parameters
now supports tuple annotations containing both class objects and
fully-qualified classnames, finally finalizing this decorator.
* Matplotlib >= 2.2.0 compatibility. Matplotlib 2.2.0, the most recently
released version of matplotlib as of this writing, accidentally broke
backward compatibility in a horrifying manner. The long-standing
public "matplotlib.verbose" API was removed in favour of the standard
"logging" API -- a dramatic upheaval so terrifying it even broke
PyCharm. This release permanently resolves the breakage on our end by
dynamically detecting the current version of matplotlib at runtime
such that:
* For matplotlib >= 2.2.0, matplotlib logging is configured via the
standard "logging" API.
* For matplotlib < 2.2.0, matplotlib logging is configured via the
non-standard "matplotlib.verbose" API. Since this API is now deprecated
and expected to be properly removed in six months, this API is only
safely accessible under older versions of matplotlib.
* Test suite expanded. Continuous integration (CI) now exercises most
simulation features with comprehensive tests, including:
* Fast simulation solver.
* The "betse sim-grn" subcommand simulating a gene regulatory network
(GRN) isolated from bioelectric phenomena. Exercised features include:
* Piggybacking onto an uninitialized, unsimulated cell cluster.
* Piggybacking onto an initialized, unsimulated cell cluster.
* Piggybacking onto an initialized, simulated cell cluster.
* Restarting from the last time step of a prior GRN run.
* Project logo adopted. The official logo for this release of BETSE is
an aesthetic flat icon bundled with Maxim Kulikov's Noun
Project-hosted "Cows" collection, kindly released under the permissive
BSD-compatible CC BY 3.0 license.