Source code for brian2.core.names

import re
import uuid

from brian2.core.tracking import Trackable
from brian2.utils.logger import get_logger

__all__ = ["Nameable"]

logger = get_logger(__name__)

[docs] def find_name(name, names=None): """ Determine a unique name. If the desired ``name`` is already taken, will try to use a derived ``name_1``, ``name_2``, etc. Parameters ---------- name : str The desired name. names : Iterable, optional A set of names that are already taken. If not provided, will use the names of all Brian objects as stored in `Nameable`. Returns ------- unique_name : str A name based on ``name`` or ``name`` itself, unique with respect to the names in ``names``. """ if not name.endswith("*"): # explicitly given names are used as given. Network.before_run (and # the device in case of standalone) will check for name clashes later return name name = name[:-1] if names is None: instances = set(Nameable.__instances__()) allnames = {obj().name for obj in instances if hasattr(obj(), "name")} else: allnames = names # Try the name without any additions first: if name not in allnames: return name # Name is already taken, try _1, _2, etc. i = 1 while f"{name}_{str(i)}" in allnames: i += 1 return f"{name}_{str(i)}"
[docs] class Nameable(Trackable): """ Base class to find a unique name for an object If you specify a name explicitly, and it has already been taken, a `ValueError` is raised. You can also specify a name with a wildcard asterisk in the end, i.e. in the form ``'name*'``. It will then try ``name`` first but if this is already specified, it will try ``name_1``, `name__2``, etc. This is the default mechanism used by most core objects in Brian, e.g. `NeuronGroup` uses a default name of ``'neurongroup*'``. Parameters ---------- name : str An name for the object, possibly ending in ``*`` to specify that variants of this name should be tried if the name (without the asterisk) is already taken. If (and only if) the name for this object has already been set, it is also possible to call the initialiser with ``None`` for the `name` argument. This situation can arise when a class derives from multiple classes that derive themselves from `Nameable` (e.g. `Group` and `CodeRunner`) and their initialisers are called explicitely. Raises ------ ValueError If the name is already taken. """ def __init__(self, name): if getattr(self, "_name", None) is not None and name is None: # name has already been specified previously return self.assign_id() if not isinstance(name, str): raise TypeError( "'name' argument has to be a string, is type " f"{repr(type(name))} instead" ) if not re.match(r"[_A-Za-z][_a-zA-Z0-9]*\*?$", name): raise ValueError(f"Name {name} not valid variable name") self._name = find_name(name) logger.diagnostic( f"Created object of class {self.__class__.__name__} with name {self._name}" )
[docs] def assign_id(self): """ Assign a new id to this object. Under most circumstances, this method should only be called once at the creation of the object to generate a unique id. In the case of the `MagicNetwork`, however, the id should change when a new, independent set of objects is simulated. """ self._id = uuid.uuid4()
name = property( fget=lambda self: self._name, doc=""" The unique name for this object. Used when generating code. Should be an acceptable variable name, i.e. starting with a letter character and followed by alphanumeric characters and ``_``. """, ) id = property( fget=lambda self: self._id, doc=""" A unique id for this object. In contrast to names, which may be reused, the id stays unique. This is used in the dependency checking to not have to deal with the chore of comparing weak references, weak proxies and strong references. """, )