Source code for brian2.stateupdaters.base

"""
This module defines the `StateUpdateMethod` class that acts as a base class for
all stateupdaters and allows to register stateupdaters so that it is able to
return a suitable stateupdater object for a given set of equations. This is used
for example in `NeuronGroup` when no state updater is given explicitly.
"""

import time
from abc import ABCMeta, abstractmethod
from collections.abc import Iterable

from brian2.utils.caching import cached
from brian2.utils.logger import get_logger

__all__ = ["StateUpdateMethod"]

logger = get_logger(__name__)


[docs] class UnsupportedEquationsException(Exception): pass
[docs] def extract_method_options(method_options, default_options): """ Helper function to check ``method_options`` against options understood by this state updater, and setting default values for all unspecified options. Parameters ---------- method_options : dict or None The options that the user specified for the state update. default_options : dict The default option values for this state updater (each admissible option needs to be present in this dictionary). To specify that a state updater does not take any options, provide an empty dictionary as the argument. Returns ------- options : dict The final dictionary with all the options either at their default or at the user-specified value. Raises ------ KeyError If the user specifies an option that is not understood by this state updater. Examples -------- >>> options = extract_method_options({'a': True}, default_options={'b': False, 'c': False}) # doctest: +IGNORE_EXCEPTION_DETAIL Traceback (most recent call last): ... KeyError: 'method_options specifies "a", but this is not an option for this state updater. Avalaible options are: "b", "c".' >>> options = extract_method_options({'a': True}, default_options={}) # doctest: +IGNORE_EXCEPTION_DETAIL Traceback (most recent call last): ... KeyError: 'method_options specifies "a", but this is not an option for this state updater. This state updater does not accept any options.' >>> options = extract_method_options({'a': True}, default_options={'a': False, 'b': False}) >>> sorted(options.items()) [('a', True), ('b', False)] """ if method_options is None: method_options = {} for key in method_options: if key not in default_options: if len(default_options): keys = sorted(default_options.keys()) options = ( "Available options are: " + ", ".join(f"'{key}'" for key in keys) + "." ) else: options = "This state updater does not accept any options." raise KeyError( f"method_options specifies '{key}', but this " "is not an option for this state updater. " f"{options}" ) filled_options = dict(default_options) filled_options.update(method_options) return filled_options
[docs] class StateUpdateMethod(metaclass=ABCMeta): stateupdaters = dict()
[docs] @abstractmethod def __call__(self, equations, variables=None, method_options=None): """ Generate abstract code from equations. The method also gets the the variables because some state updaters have to check whether variable names reflect other state variables (which can change from timestep to timestep) or are external values (which stay constant during a run) For convenience, this arguments are optional -- this allows to directly see what code a state updater generates for a set of equations by simply writing ``euler(eqs)``, for example. Parameters ---------- equations : `Equations` The model equations. variables : dict, optional The `Variable` objects for the model variables. method_options : dict, optional Additional options specific to the state updater. Returns ------- code : str The abstract code performing a state update step. """ pass
[docs] @staticmethod def register(name, stateupdater): """ Register a state updater. Registered state updaters can be referred to via their name. Parameters ---------- name : str A short name for the state updater (e.g. `'euler'`) stateupdater : `StateUpdaterMethod` The state updater object, e.g. an `ExplicitStateUpdater`. """ # only deal with lower case names -- we don't want to have 'Euler' and # 'euler', for example name = name.lower() if name in StateUpdateMethod.stateupdaters: raise ValueError( f"A stateupdater with the name '{name}' has already been registered" ) if not isinstance(stateupdater, StateUpdateMethod): raise ValueError( f"Given stateupdater of type {type(stateupdater)} does " "not seem to be a valid stateupdater." ) StateUpdateMethod.stateupdaters[name] = stateupdater
[docs] @staticmethod @cached def apply_stateupdater( equations, variables, method, method_options=None, group_name=None ): """ apply_stateupdater(equations, variables, method, method_options=None, group_name=None) Applies a given state updater to equations. If a `method` is given, the state updater with the given name is used or if is a callable, then it is used directly. If a `method` is a list of names, all the methods will be tried until one that doesn't raise an `UnsupportedEquationsException` is found. Parameters ---------- equations : `Equations` The model equations. variables : `dict` The dictionary of `Variable` objects, describing the internal model variables. method : {callable, str, list of str} A callable usable as a state updater, the name of a registered state updater or a list of names of state updaters. Returns ------- abstract_code : str The code integrating the given equations. """ if isinstance(method, Iterable) and not isinstance(method, str): the_method = None start_time = time.time() for one_method in method: try: one_method_start_time = time.time() code = StateUpdateMethod.apply_stateupdater( equations, variables, one_method, group_name=group_name ) the_method = one_method one_method_time = time.time() - one_method_start_time break except UnsupportedEquationsException: pass except TypeError: raise TypeError( "Each element in the list of methods has " "to be a string or a callable, got " f"{type(one_method)}." ) total_time = time.time() - start_time if the_method is None: raise ValueError( "No stateupdater that is suitable for the " "given equations has been found." ) # If only one method was tried if method[0] == the_method: timing = f"took {one_method_time:.2f}s" else: timing = ( f"took {one_method_time:.2f}s, trying other methods took " f"{total_time - one_method_time:.2f}s" ) if group_name is not None: msg_text = ( "No numerical integration method specified for group " f"'{group_name}', using method '{the_method}' ({timing})." ) else: msg_text = ( "No numerical integration method specified, " f"using method '{the_method}' ({timing})." ) logger.info(msg_text, "method_choice") else: if callable(method): # if this is a standard state updater, i.e. if it has a # can_integrate method, check this method and raise a warning if it # claims not to be applicable. stateupdater = method method = getattr( stateupdater, "__name__", repr(stateupdater) ) # For logging, get a nicer name elif isinstance(method, str): method = method.lower() # normalize name to lower case stateupdater = StateUpdateMethod.stateupdaters.get(method, None) if stateupdater is None: raise ValueError( "No state updater with the name '{method}' is known." ) else: raise TypeError( "method argument has to be a string, a " "callable, or an iterable of such objects. " f"Got {type(method)}" ) start_time = time.time() code = stateupdater(equations, variables, method_options) method_time = time.time() - start_time timing = "took %.2fs" % method_time if group_name is not None: logger.debug( f"Group {group_name}: using numerical integration " f"method {method} ({timing})", "method_choice", ) else: logger.debug( f"Using numerical integration method: {method} f({{timing}})", "method_choice", ) return code