Changelog

v1.6.5

(#241) Drop Python 3.9 support in favor of 3.12 and 3.13, bump required networkx to 3.4, fix issue with plot_topology on networkx 3.4+, and change default multiprocessing start method to safer forkserver method on Linux

v1.6.4

(#240) fix Pint Quantity import error by using more recent version.

v1.6.3

(#237) Add tests for MongoDB emitter and fix related bugs
(#238) apply global_time_precision to the emit_time

v1.6.2

(#236) Add data key to breakdown documents if missing

v1.6.1

(#234) Allow Numpy arrays to be deeply nested in Store schemas
(#233) Fix some bugs in emitter.py

v1.6.0

(#231) Update networkx error message for already deleted Step.

v1.5.9

(#230) add a test for '_move' updates. Extend '_move' updates to support targets going down a path rather than just a top-level port name

v1.5.8

(#227) Make Process.default_state parse ports schema recursively.
Add 'update' key to '_move' hierarchy update method.

v1.5.7

(#226) Drop Python 3.8 in favor of 3.11 and update dependencies.

v1.5.6

(#225) Update registry imports to store.py.

v1.5.5

(#224) Keep processes and steps in composite spec that is passed to Engine. Add test_hierarchy_update to keep this working.

v1.5.4

(#223) Update pint imports to support newest version
(#222) More breakdown fixes
(#220) Recalculate hashes for comparison of Units
(#219) Parallel queries. Ccreates helper methods to intelligently split large queries into smaller, approximately equal chunks that can be run in parallel on separate OS processes.
(#218) MongoClient optimizations
(#217) Create Mongo index for simulation time.
(#215) Serialization fixes for BSON.
(#214) Optimize dictionary key removal.
(#213) Clean up documentation for Serializer API.

v1.5.3

(#211) Fix a memory leak when removing parallel processes.
(#204) Implement a new serialization API based on PyMongo's TypeCodec interface.

v1.5.2

(#207) Fix a bug in Store.apply_update() that caused failures when self.value was a list and self.units was set.
(#205) Prevents formation of multi_updates when generating initial state.

v1.5.1

(#206) Support multiple emits for the same timestep and remove Engine.complete(). Instead, callers of Engine.run_for() must pass force_complete=True at the end of their caller-managed simulation loops.

v1.4.2

(#203) Make Store.sources more comprehensive, in particular by including dividers and flows.

v1.4.1

(#202) Add the filters argument to data_from_database() to allow further filtering of MongoDB query results.

v1.3.1

(#200) Inside Engine, store the Step execution layers as lists instead of sets to ensure deterministic execution order.
(#201) Restore ability to pass initial_state keys in settings dictionaries to composition.py functions.

v1.3.0

(#198) Introduce process commands to support more interactions with parallel processes. Now all Process methods of a parallelized process can be queried from the parent OS process. Users can also add support for custom methods of their processes.

This change also simplifies the way Engine handles parallel processes warns users when serializers are not being found efficiently.
(#192) Marks composition.py as deprecated and ensures that the rest of Vivarium Core doesn't depend on it.
(#199) Remove some Numpy dtypes that are not available on some platforms from serialize.py.

v1.2.8

(#186) Apply function to data from database emitter in get_history_data_db.

v1.2.7

(#190) makes global_time_precision an Engine parameter rather than Engine.run_for() keyword -- this changes the method introduced in 1.2.5. Technically a breaking API, but only from 2 versions ago.
(#191) don't use timestep for quiet paths. This allows Engine to skip over processes that don't meet their update condition.

v1.2.6

(#189) track composite updates in engine so that the composite topology can be plotted after being updated.
(#188) RAMEmitter returns query data only if it exists.

v1.2.5

(#183) floating-point timestep precision.
(#182) add pytest to setup.
(#180) Handle nan in units serializer.

v1.2.4

(#179) Register the injector process.
(#178) Remove compose_experiment(), which was replaced by Composite.merge() long ago.

v1.2.3

(#176) Fix UnitsSerializer to correctly handle lists and tuples.

v1.2.2

(#175) store_schema arg to Engine allows us to override schema directly.

v1.2.1

(#172) bug fix to keep assembly_ids when passing queries to the DB.

v1.2.0

(#167) Make serialization more structured and robust with serializer classes that know what kinds of data they can serialize and deserialize.

v1.1.1

(#169) get_history_data_db no longer requires a 'time' key. It asserts an 'assembly_id' key instead.

v1.1.0

(#164) Add a _no_original_parameters configuration to Process that, if True, disables the copying of parameters that Process.__init__() does by default. Disabling this copying decreases memory usage, but it puts the user in charge of ensuring that parameters are not mutated.

v1.0.2

(#163) Fix two bugs:
- In divide_condition.py, use a divider that sets the division variable to False instead of 0.
- In engine.py, fix the function that checks that every dependency in the flow is also in the dictionary of steps. This function did not correctly handle cases where steps were nested in sub-dictionaries.

v1.0.1

(#151) Make Engine._run_steps() public.
(#155) Add error messages for flows that include steps not found in Engine.steps and for dividers specified as strings not found in the divider registry.
(#157) Add flows and steps to the value returned by Composite.generate_store().

v1.0.0

(#131) Do not assume that a process's initial_state should be used when using the Store API.
(#132) Improve error message when no path exists between nodes in the hierarchy.
(#133) Remove Process.local_timestep(), which is no longer being used.
(#134) Support using ** in a ports schema to connect to an entire subtree of the hierarchy.
(#143) Make more methods private.
(#147) Fix bug in process serialization to support changing the parameter argument name in the constructor.

v0.4.20

deep_merge_combine_lists recursive call.

v0.4.19

(#130) Raise an exception when a user specifies two different dividers for the same variable, and when serializing a process, use its original parameters without any changes that may have occurred since the process was initialized.

v0.4.18

(#127) build a store argument's topology views in Engine constructor to support the store API.

v0.4.17

(#126) A new method, Engine.run_for, can be called iteratively without completing processes on the front. Engine.update keeps the same behavior as before. Engine.complete forces all processes to complete at the current global time.

v0.4.16

(#125) Use the null divider by default for processes so that upon division, processes are generated by the composer, not a divider.

v0.4.15

(#123) Extend Composite.merge to use steps and flows.
(#124) Use set divider by default.

v0.4.14

(#121) Fix a bug in Store.divide() where daughter cell states were by default the mother cell state. Instead, add support for an initial_state key in the _divide dictionary.
(#122) Add an _output flag option to Process.ports_schema, which marks ports as output-only.

v0.4.13

Improve topology_view caching mechanism to speed up simulations that have a lot of division.

v0.4.12

Allow schema overrides to override step schemas.

v0.4.11

Assert that when Engine treats a process like a step, that process is actually a step.
Use Process.is_step() to check whether a process is a step instead of using isinstance.

v0.4.10

Fix Engine to support flows that are nested dictionaries.

v0.4.9

Add profile argument to the Engine constructor. When this argument is set to True, the simulation, including any parallel processes, will be profiled by cProfile. Once Engine.end() is called, the profile stats will be saved to Engine.stats for analysis.

v0.4.8

Remove self._run_steps() from Engine.apply_update if view_expire.

v0.4.7

view_expire flag for _add and _delete updates as well.

v0.4.6

deepcopy mother state upon division so daughter states are ensured to not reference the same objects.
improve _get_composite_state to avoid recursive deepcopy.
add view_expire flag in Store.apply_update, and have Engine use this to trigger self.state.build_topology_views.

v0.4.5

Fixed a major bug introduced by topology_view in v0.4.3, which had glob (*) schema unable to view sub-stores.
dict_value updater added to updater_registry.

v0.4.4

Add queries to Emitter, and to get_history_data_db. This allows you to selectively retrieve from paths in the data, and can save a lot of time when retrieving data from large experiments.

v0.4.3

Bug fix in the new topology_view, return Store for '*' schema.

v0.4.2

Add topology_view caching in Store to improve performance.

v0.4.1

Fix a bug in Store.generate() that caused conflicts between a user-provided initial state and a schema to raise an error instead of the initial state taking priority.

v0.4.0

Replaces Derivers with Steps. While derivers were executed sequentially, steps are executed in topological generations according to a dependency graph. This lets some derivers run in parallel. This change mostly preserves backwards-compatibility since Deriver is now an alias for Step, and we still support legacy derivers that are specified without dependencies. These legacy derivers are executed sequentially before any steps. However, the minor version is incremented because the following public interfaces have changed (though we don't expect this to break dependent code):
- Composites now have 2 more keys: steps and flow.
- Engine.run_derivers() has been replaced by Engine._run_steps().
- Engine.deriver_paths has been replaced by Engine._step_paths.
- New lists of step and flow updates have been added to the tuple returned by Store.apply_update().
- Store.EMPTY_UPDATES has two more None values and is now private (Store._EMPTY_UPDATES).
- Store.get_processes() no longer returns steps (formerly "derivers"). Instead, these are returned by get_steps().
- Makes Store.generate_paths() private (now Store._generate_paths()).
- Adds required step and flow arguments to Store.generate().
- Adds metadata argument to Engine.
Fixes a bug where parallel derivers were re-instantiated every timestep.
Marks the Store API as experimental, including the public use of Store.move(), Store.insert(), Store.divide(), and Store.delete().

v0.3.14

De/serializer for np.bool_

v0.3.13

Fix topology plot to avoid using plt.figure so that plots work correctly when other plotting functions are used in the same Python session (e.g. Jupyter notebooks).

v0.3.12

add composite specification

v0.3.11

fix bug in serializing list

v0.3.10

add emitter_registry, so that emitters can be added from different repositories by importing the registry and registering their emitter.
add _condition key to process parameters, that lets them by conditionally triggered by connecting to a boolean states in the hierarchy.

v0.3.9

reduce mongo document limit for the DatabaseEmitter

v0.3.8

revert rename of divider argument to "state"
do not allow _add updates with repeat keys

v0.3.7

fix custom dividers to use topology argument

v0.3.5

remove unneeded deepcopy from update step, leading to runtime boost.
runtime_profile experiment added.

v0.3.4

Engine uses keyword arguments rather than a config dict. To fix this do Engine(**config_dict) instead of Engine(config_dict). Additional emit_config options are added to selectively emit metadata.

v0.3.3

Engine has emit_config option. Is set to False, this keeps the serialized processes from being emitted and can save a lot of compute time.

v0.3.2

DatabaseEmitter supports arbitrary emit sizes by breaking up large emits into chunks, and reassembles them seamlessly upon retrieval.
libary.wrappers make_logging_process adds a logging port to any process
Improve the Store API by adding methods for creating new stores and connecting processes through their ports.

v0.3.1

Fix plot_topology to use Process.generate() now that Process no longer inherits from Composer

v0.3.0

Store API supports simplified scripting of a bigraph
Breaking import changes: Composite and Composer are now under vivarium.core.composer, and Experiment has been renamed Engine under vivarium.core.engine.

v0.2.16

add more flexible kwarg passing in Control

v0.2.15

make Experiment's invoke_process method public again

v0.2.14

fix bug to clear Composite of processes and topology instance variables
more general divider declaration
add time_display arg to agents_multigen
update docs

v0.2.13

handle large emit_configuration to database emitter by checking the size of the packet, and removing process parameters if necessary.

v0.2.12

plotting tweaks.
provide information about update failure.
improve data_from_database to return experiment config.

v0.2.11

process.schema = None in base class constructor

v0.2.10

string-based path declaration for topologies.
custom node border widths in plot_topology.

v0.2.9

composite.merge() uses the provided topology to override merged composite topology

v0.2.8

improved plot_topology

v0.2.7

plot_topology includes new the graph_options: horizontal and hierarchy, and new process_colors and store_colors for coloring individual nodes.

v0.2.6

new Composite and AggregateComposer classes in core/process.py

v0.2.5

remove future annotations to continue support of python 3.6

v0.2.3

bring back compatibility for python 3.6

v0.2.2

generalize mass_adaptors and molarity_derivers

v0.2.1

add mass_adaptor

v0.2.0

breaking API changes:
- Composer class instead of Composite.
- several composition.py functions have been renamed to improve self-description.

v0.1.12

bug fix when handling complex_topology. store.apply_update() can now go up paths with '..'.
calculate_timestep() for Process supports adaptive timesteps.

v0.1.11

clock process initial state

v0.1.10

add plot_variables plotting function to simulation_output

v0.1.9

clock process for keeping track of global time

v0.1.7

growth_rate process for generic exponential growth of variables

v0.1.6

topology grapher option to leave edge colors black

v0.1.5

topology grapher uses flowchart symbols

v0.1.4

add options to agents_multigen plot, for improved look on Jupyter notebooks

v0.1.3

fix Composite generate() to get correct path

v0.1.2

rework the class structure of Processes and Composites. Replace Generator with Factory.

v0.1.1

get_networkx_graph() and graph_figure() in vivarium/plot/topology.py

v0.0.39

improved units handling, with a units serializer for emit_data, and convert data to timeseries stripping the units.

v0.0.36

multi_update for experiment allos multiple updates to a variable from the same process
fix some inspection errors & warnings
merge for generator class

v0.0.34

Control object in vivarium/cor/control for streamlining experiments
progress bar and display in experiment
disintegrate process
divide_condition process
swap_compartment process

v0.0.29

timeline process configured with paths for ports

v0.0.28

vivarium/plots directory
register in module init
get_initialize_state for Generator
generalize timeline process

v0.0.20

Update documentation with:
- A template repository for getting started.
- Changes to reflect recent updates to the project.
- Tests for the documentation.
Add injector process.
Add demo process, compartment, and experiment modeling glucose phosphorylation.