Merge pull request #1528 from SpiNNakerManchester/AbstractProvidesDef…

…aults

Abstract provides defaults

spynnaker/pyNN/external_devices_models/munich_spinnaker_link_motor_device.py

@@ -21,7 +21,7 @@
     AbstractOneAppOneMachineVertex)
 from spinn_front_end_common.abstract_models import (
     AbstractVertexWithEdgeToDependentVertices)
-from spynnaker.pyNN.models.defaults import defaults
+from spynnaker.pyNN.models.defaults import AbstractProvidesDefaults
 from spynnaker.pyNN.models.common import PopulationApplicationVertex
 from .machine_munich_motor_device import MachineMunichMotorDevice
 
@@ -35,11 +35,11 @@ def __init__(self, spinnaker_link_id, board_address=None):
             label="External Munich Motor", board_address=board_address)
 
 
-@defaults
 class MunichMotorDevice(
         AbstractOneAppOneMachineVertex,
         AbstractVertexWithEdgeToDependentVertices,
-        PopulationApplicationVertex):
+        PopulationApplicationVertex,
+        AbstractProvidesDefaults):
     """
     An Omnibot motor control device. This has a real vertex and an
     external device vertex.

spynnaker/pyNN/models/abstract_pynn_model.py

@@ -15,21 +15,20 @@
 from collections import defaultdict
 import sys
 from typing import (
-    Any, Callable, cast, Dict, FrozenSet, Optional, Mapping, Sequence, Tuple,
-    TYPE_CHECKING)
+    Any, Dict, Optional, Sequence, Tuple, TYPE_CHECKING)
 import numpy
 from pyNN import descriptions
 from spinn_utilities.classproperty import classproperty
 from spinn_utilities.abstract_base import (
     AbstractBase, abstractmethod)
-from spynnaker.pyNN.models.defaults import get_map_from_init
+from spynnaker.pyNN.models.defaults import AbstractProvidesDefaults
 from spynnaker.pyNN.exceptions import SpynnakerException
 if TYPE_CHECKING:
     from spynnaker.pyNN.models.common.population_application_vertex import (
         PopulationApplicationVertex)
 
 
-class AbstractPyNNModel(object, metaclass=AbstractBase):
+class AbstractPyNNModel(AbstractProvidesDefaults, metaclass=AbstractBase):
     """
     A Model that can be passed in to a Population object in PyNN.
     """
@@ -89,44 +88,6 @@ def absolute_max_atoms_per_core(  # pylint: disable=no-self-argument
         """
         return sys.maxsize
 
-    @staticmethod
-    def __get_init_params_and_svars(the_cls: type) -> Tuple[
-            Callable, Optional[FrozenSet[str]], Optional[FrozenSet[str]]]:
-        init = getattr(the_cls, "__init__")
-        while hasattr(init, "_method"):
-            init = getattr(init, "_method")
-        params = None
-        if hasattr(init, "_parameters"):
-            params = getattr(init, "_parameters")
-        svars = None
-        if hasattr(init, "_state_variables"):
-            svars = getattr(init, "_state_variables")
-        return init, params, svars
-
-    @classproperty
-    def default_parameters(  # pylint: disable=no-self-argument
-            cls) -> Mapping[str, Any]:
-        """
-        Get the default values for the parameters of the model.
-
-        :rtype: dict(str, Any)
-        """
-        init, params, svars = cls.__get_init_params_and_svars(cast(type, cls))
-        return get_map_from_init(init, skip=svars, include=params)
-
-    @classproperty
-    def default_initial_values(  # pylint: disable=no-self-argument
-            cls) -> Mapping[str, Any]:
-        """
-        Get the default initial values for the state variables of the model.
-
-        :rtype: dict(str, Any)
-        """
-        init, params, svars = cls.__get_init_params_and_svars(cast(type, cls))
-        if params is None and svars is None:
-            return {}
-        return get_map_from_init(init, skip=params, include=svars)
-
     @classmethod
     def get_parameter_names(cls) -> Sequence[str]:
         """

spynnaker/pyNN/models/defaults.py

@@ -19,7 +19,9 @@
 import inspect
 import logging
 from types import MappingProxyType
-from typing import Any, Callable, FrozenSet, Iterable, List, Mapping, Optional
+from typing import (
+    Any, Callable, FrozenSet, Iterable, List, Mapping, Optional)
+from spinn_utilities.classproperty import classproperty
 from spinn_utilities.log import FormatAdapter
 
 logger = FormatAdapter(logging.getLogger(__name__))
@@ -156,6 +158,8 @@ def defaults(cls: type) -> type:
     If neither are specified, it is assumed that all default arguments are
     parameters.
     """
+    logger.warning("@defaults is deprecated! "
+                   "Extend AbstractProvidesDefaults instead")
     if not inspect.isclass(cls):
         raise TypeError(f"{cls} is not a class")
     if not hasattr(cls, "__init__"):
@@ -182,3 +186,103 @@ def defaults(cls: type) -> type:
         cls.default_parameters = get_map_from_init(init, include=params)
         cls.default_initial_values = get_map_from_init(init, include=svars)
     return cls
+
+
+class AbstractProvidesDefaults(object):
+    """
+    Provides the default_parameters and default_initial_values properties
+
+    These will be filled in based on the @default_parameters and
+    @default_initial_values decorators with values read from the init.
+    """
+
+    @classmethod
+    def __fill_in_defaults(cls):
+        """
+        Fills in default_parameters and default_initial_values attributes
+        """
+        # get the init method
+        init = getattr(cls, "__init__")
+        # Find the real method as there may be decorators
+        while hasattr(init, "_method"):
+            init = getattr(init, "_method")
+
+        # read the values from the init method
+        init_args = inspect.getfullargspec(init)
+        n_defaults = 0 if init_args.defaults is None else len(
+            init_args.defaults)
+        n_args = 0 if init_args.args is None else len(init_args.args)
+        default_args = ([] if init_args.args is None else
+                        init_args.args[n_args - n_defaults:])
+        if init_args.defaults is None:
+            default_values = []
+        else:
+            default_values = init_args.defaults
+
+        # get the keys based on the decorators
+        if hasattr(init, "_parameters"):
+            params = getattr(init, "_parameters")
+            if hasattr(init, "_state_variables"):
+                svars = getattr(init, "_state_variables")
+                assert len(params.intersection(svars)) == 0
+            else:
+                svars = frozenset(svar for svar in default_args
+                                  if svar not in params)
+        else:
+            if hasattr(init, "_state_variables"):
+                svars = getattr(init, "_state_variables")
+                params = frozenset(param for param in default_args
+                                   if param not in svars)
+            else:
+                svars = frozenset()
+                params = set(default_args)
+
+        # Check all decorator values used
+        _check_args(params.union(svars), default_args, init)
+
+        # fill in the defaults so this method is only called once
+        cls.default_parameters: Mapping[str, Any] = {}
+        cls.default_initial_values: Mapping[str, Any] = {}
+        for arg, value in zip(default_args, default_values):
+            if arg in params:
+                cls.default_parameters[arg] = value
+            elif arg in svars:
+                cls.default_initial_values[arg] = value
+        cls.default_parameters = MappingProxyType(cls.default_parameters)
+        cls.default_initial_values = (
+            MappingProxyType(cls.default_initial_values))
+
+    @classproperty
+    def default_parameters(  # pylint: disable=no-self-argument
+            cls) -> Mapping[str, Any]:
+        """
+        Get the default values for the parameters of the model.
+
+        If a @default_parameters decorator is used
+        this will be the init default values for those keys
+
+        If no @default_parameters decorator is used
+        this will be all the init parameters with a default value
+        less any defined in @default_initial_values
+        """
+        cls.__fill_in_defaults()
+        return cls.default_parameters
+
+    @classproperty
+    def default_initial_values(  # pylint: disable=no-self-argument
+            cls) -> Mapping[str, Any]:
+        """
+        Get the default initial values for the state variables of the model.
+
+        If @default_initial_values decorator is used
+        this will be the init default values for those keys
+
+        If no @default_initial_values decorator is used
+        but a @default_parameters decorator was used
+        this will be all the init parameters with a default value
+        less any defined in @default_parameters
+
+        If neither decorator is used this will be an empty Mapping
+        """
+        cls.__fill_in_defaults()
+        return cls.default_initial_values

spynnaker/pyNN/models/neuron/builds/eif_cond_alpha_isfa_ista.py

@@ -13,12 +13,12 @@
 # limitations under the License.
 
 from spynnaker.pyNN.exceptions import SpynnakerException
-from spynnaker.pyNN.models.defaults import defaults, default_initial_values
+from spynnaker.pyNN.models.defaults import (
+    AbstractProvidesDefaults, default_initial_values)
 
 
 # pylint: disable=wrong-spelling-in-docstring
-@defaults
-class EIFConductanceAlphaPopulation(object):
+class EIFConductanceAlphaPopulation(AbstractProvidesDefaults):
     """
     Exponential integrate and fire neuron with spike triggered and
     sub-threshold adaptation currents (isfa, ista reps.)

spynnaker/pyNN/models/neuron/builds/hh_cond_exp.py

@@ -13,11 +13,11 @@
 # limitations under the License.
 
 from spynnaker.pyNN.exceptions import SpynnakerException
-from spynnaker.pyNN.models.defaults import defaults, default_initial_values
+from spynnaker.pyNN.models.defaults import (
+    AbstractProvidesDefaults, default_initial_values)
 
 
-@defaults
-class HHCondExp(object):
+class HHCondExp(AbstractProvidesDefaults):
     """
     Single-compartment Hodgkin-Huxley model with exponentially decaying
     current input.

spynnaker/pyNN/models/neuron/builds/if_cond_alpha.py

@@ -13,11 +13,11 @@
 # limitations under the License.
 
 from spynnaker.pyNN.exceptions import SpynnakerException
-from spynnaker.pyNN.models.defaults import defaults, default_initial_values
+from spynnaker.pyNN.models.defaults import (
+    AbstractProvidesDefaults, default_initial_values)
 
 
-@defaults
-class IFCondAlpha(object):
+class IFCondAlpha(AbstractProvidesDefaults):
     """
     Leaky integrate and fire neuron with an alpha-shaped current input.
 

spynnaker/pyNN/models/neuron/builds/if_facets_hardware1.py

@@ -13,11 +13,11 @@
 # limitations under the License.
 
 from spynnaker.pyNN.exceptions import SpynnakerException
-from spynnaker.pyNN.models.defaults import defaults, default_initial_values
+from spynnaker.pyNN.models.defaults import (
+    AbstractProvidesDefaults, default_initial_values)
 
 
-@defaults
-class IFFacetsConductancePopulation(object):
+class IFFacetsConductancePopulation(AbstractProvidesDefaults):
     """
     Leaky integrate and fire neuron with conductance-based synapses and
     fixed threshold as it is resembled by the FACETS Hardware Stage 1.

unittests/model_tests/neuron/test_abstract_neuron_impl.py

@@ -19,11 +19,10 @@
 from spynnaker.pyNN.models.abstract_pynn_model import AbstractPyNNModel
 from spynnaker.pyNN.models.neuron.abstract_pynn_neuron_model import (
     AbstractPyNNNeuronModel)
-from spynnaker.pyNN.models.defaults import default_initial_values, defaults
+from spynnaker.pyNN.models.defaults import default_initial_values
 from spynnaker.pyNN.exceptions import SpynnakerException
 
 
-@defaults
 class _MyPyNNModelImpl(AbstractPyNNModel):
 
     default_population_parameters = {}