Merge pull request #1699 from PrincetonUniversity/devel

Devel

psyneulink/core/components/component.py

@@ -976,7 +976,7 @@ def _parse_variable(self, variable):
                 return variable
 
             try:
-                return np.asarray(variable)
+                return convert_to_np_array(variable)
             except ValueError:
                 return convert_all_elements_to_np_array(variable)
 
@@ -1507,7 +1507,8 @@ def checkAndCastInt(x):
                     variable = []
                     for s in size:
                         variable.append(np.zeros(s))
-                    variable = np.array(variable)
+                    variable = convert_to_np_array(variable)
+                # TODO: fix bare except
                 except:
                     raise ComponentError("variable (possibly default_variable) was not specified, but PsyNeuLink "
                                          "was unable to infer variable from the size argument, {}. size should be"

psyneulink/core/components/functions/combinationfunctions.py

@@ -43,7 +43,7 @@
     DEFAULT_VARIABLE, EXPONENTS, LINEAR_COMBINATION_FUNCTION, MULTIPLICATIVE_PARAM, OFFSET, OPERATION, \
     PREDICTION_ERROR_DELTA_FUNCTION, PRODUCT, REARRANGE_FUNCTION, REDUCE_FUNCTION, SCALE, SUM, WEIGHTS, \
     PREFERENCE_SET_NAME
-from psyneulink.core.globals.utilities import is_numeric, np_array_less_than_2d, parameter_spec
+from psyneulink.core.globals.utilities import convert_to_np_array, is_numeric, np_array_less_than_2d, parameter_spec
 from psyneulink.core.globals.context import Context, ContextFlags
 from psyneulink.core.globals.parameters import Parameter
 from psyneulink.core.globals.preferences.basepreferenceset import \
@@ -552,7 +552,7 @@ def _function(self,
                     for index in item:
                         stack.append(variable[index])
                     result.append(np.hstack(tuple(stack)))
-                result = np.array(result) * scale + offset
+                result = convert_to_np_array(result) * scale + offset
             except IndexError:
                 assert False, f"PROGRAM ERROR: Bad index specified in {repr(ARRANGEMENT)} arg -- " \
                     f"should have been caught in _validate_params or _instantiate_attributes_before_function"

psyneulink/core/components/functions/distributionfunctions.py

@@ -35,7 +35,7 @@
     EXPONENTIAL_DIST_FUNCTION, GAMMA_DIST_FUNCTION, HIGH, LOW, MULTIPLICATIVE_PARAM, NOISE, NORMAL_DIST_FUNCTION, \
     SCALE, STANDARD_DEVIATION, THRESHOLD, UNIFORM_DIST_FUNCTION, WALD_DIST_FUNCTION
 from psyneulink.core.globals.context import ContextFlags
-from psyneulink.core.globals.utilities import parameter_spec, get_global_seed
+from psyneulink.core.globals.utilities import convert_to_np_array, parameter_spec, get_global_seed
 from psyneulink.core.globals.preferences.basepreferenceset import is_pref_set
 
 from psyneulink.core.globals.parameters import Parameter
@@ -1265,9 +1265,15 @@ def _function(self,
                     exp_neg2_ztilde_atilde = np.exp(-2 * ztilde * atilde)
 
                     if self.shenhav_et_al_compat_mode:
-                        exp_neg2_x0tilde_atilde = np.nanmax([1e-12, exp_neg2_x0tilde_atilde])
-                        exp_2_ztilde_atilde = np.nanmin([1e12, exp_2_ztilde_atilde])
-                        exp_neg2_ztilde_atilde = np.nanmax([1e-12, exp_neg2_ztilde_atilde])
+                        exp_neg2_x0tilde_atilde = np.nanmax(
+                            convert_to_np_array([1e-12, exp_neg2_x0tilde_atilde])
+                        )
+                        exp_2_ztilde_atilde = np.nanmin(
+                            convert_to_np_array([1e12, exp_2_ztilde_atilde])
+                        )
+                        exp_neg2_ztilde_atilde = np.nanmax(
+                            convert_to_np_array([1e-12, exp_neg2_ztilde_atilde])
+                        )
 
                     rt = ztilde * np.tanh(ztilde * atilde) + \
                          ((2 * ztilde * (1 - exp_neg2_x0tilde_atilde)) / (

psyneulink/core/components/functions/function.py

@@ -161,7 +161,7 @@
 from psyneulink.core.globals.preferences.preferenceset import PreferenceEntry, PreferenceLevel
 from psyneulink.core.globals.registry import register_category
 from psyneulink.core.globals.utilities import (
-    get_global_seed, object_has_single_value, parameter_spec, safe_len
+    convert_to_np_array, get_global_seed, object_has_single_value, parameter_spec, safe_len
 )
 
 __all__ = [
@@ -649,7 +649,7 @@ def convert_output_type(self, value, output_type=None):
             else:
                 output_type = self.output_type
 
-        value = np.asarray(value)
+        value = convert_to_np_array(value)
 
         # Type conversion (specified by output_type):
 
@@ -1006,7 +1006,7 @@ def get_matrix(specification, rows=1, cols=1, context=None):
 
     # Matrix provided (and validated in _validate_params); convert to array
     if isinstance(specification, (list, np.matrix)):
-        specification = np.array(specification)
+        return convert_to_np_array(specification)
 
     if isinstance(specification, np.ndarray):
         if specification.ndim == 2:

psyneulink/core/components/functions/interfacefunctions.py

@@ -27,6 +27,7 @@
 from psyneulink.core.globals.parameters import Parameter
 from psyneulink.core.globals.preferences.basepreferenceset import \
     PreferenceEntry, PreferenceLevel, is_pref_set, REPORT_OUTPUT_PREF
+from psyneulink.core.globals.utilities import convert_to_np_array
 
 
 __all__ = ['InterfaceFunction', 'InterfacePortMap']
@@ -173,7 +174,10 @@ def _function(
         if self.corresponding_input_port.owner.parameters.value._get(context) is not None:
 
             # If CIM's variable does not match its value, then a new pair of ports was added since the last execution
-            if not np.shape(self.corresponding_input_port.owner.get_input_values(context)) == np.shape(self.corresponding_input_port.owner.parameters.value._get(context)):
+            input_values = convert_to_np_array(
+                self.corresponding_input_port.owner.get_input_values(context)
+            )
+            if not np.shape(input_values) == np.shape(self.corresponding_input_port.owner.parameters.value._get(context)):
                 return self.corresponding_input_port.owner.defaults.variable[index]
 
             # If the variable is 1D (e.g. [0. , 0.], NOT [[0. , 0.]]), and the index is 0, then return whole variable

psyneulink/core/components/functions/learningfunctions.py

@@ -585,7 +585,7 @@ def _function(
 
 
         variable = self._check_args(
-            [np.atleast_2d(variable[0]), np.atleast_2d(variable[1])],
+            [convert_to_np_array(variable[0], dimension=2), convert_to_np_array(variable[1], dimension=2)],
             params,
             context,
         )

psyneulink/core/components/functions/optimizationfunctions.py

@@ -1639,7 +1639,7 @@ def _run_cuda_grid(self, ocm, variable, context):
         # Compiled evaluate expects the same variable as mech function
         new_variable = [np.asfarray(ip.parameters.value.get(context))
                         for ip in ocm.input_ports]
-        new_variable = np.atleast_2d(new_variable)
+        new_variable = np.array(new_variable, dtype=np.object)
         # Map allocations to values
         comp_exec = pnlvm.execution.CompExecution(ocm.agent_rep, [context.execution_id])
         ct_alloc, ct_values = comp_exec.cuda_evaluate(new_variable,

psyneulink/core/components/functions/statefulfunctions/integratorfunctions.py

@@ -44,7 +44,7 @@
     INCREMENT, INITIALIZER, INPUT_PORTS, INTEGRATOR_FUNCTION, INTEGRATOR_FUNCTION_TYPE, \
     INTERACTIVE_ACTIVATION_INTEGRATOR_FUNCTION, LEAKY_COMPETING_INTEGRATOR_FUNCTION, \
     MULTIPLICATIVE_PARAM, NOISE, OFFSET, OPERATION, ORNSTEIN_UHLENBECK_INTEGRATOR_FUNCTION, OUTPUT_PORTS, PRODUCT, \
-    RATE, REST, SIMPLE_INTEGRATOR_FUNCTION, SUM, TIME_STEP_SIZE, THRESHOLD
+    RATE, REST, SIMPLE_INTEGRATOR_FUNCTION, SUM, TIME_STEP_SIZE, THRESHOLD, VARIABLE
 from psyneulink.core.globals.parameters import Parameter
 from psyneulink.core.globals.utilities import parameter_spec, all_within_range, iscompatible, get_global_seed
 from psyneulink.core.globals.context import Context, ContextFlags, handle_external_context
@@ -1203,7 +1203,12 @@ def _function(self,
         previous_value = self.get_previous_value(context)
         # MODIFIED 6/14/19 END
 
-        value = self._EWMA_filter(previous_value, rate, variable) + noise
+        try:
+            value = self._EWMA_filter(previous_value, rate, variable) + noise
+        except TypeError:
+            # TODO: this should be standardized along with the other instances
+            # of this error
+            raise FunctionError("Unrecognized type for {} of {} ({})".format(VARIABLE, self.name, variable))
 
         adjusted_value = value + offset
 

psyneulink/core/components/functions/statefulfunctions/memoryfunctions.py

@@ -38,7 +38,7 @@
 from psyneulink.core.globals.keywords import \
     ADDITIVE_PARAM, BUFFER_FUNCTION, MEMORY_FUNCTION, COSINE, ContentAddressableMemory_FUNCTION, \
     MIN_INDICATOR, MULTIPLICATIVE_PARAM, NEWEST, NOISE, OLDEST, OVERWRITE, RATE, RANDOM
-from psyneulink.core.globals.utilities import all_within_range, parameter_spec, get_global_seed
+from psyneulink.core.globals.utilities import all_within_range, convert_to_np_array, parameter_spec, get_global_seed
 from psyneulink.core.globals.context import Context, ContextFlags, handle_external_context
 from psyneulink.core.globals.parameters import Parameter
 from psyneulink.core.globals.preferences.basepreferenceset import is_pref_set
@@ -330,7 +330,7 @@ def _function(self,
 
         # Apply rate and/or noise, if they are specified, to all stored items
         if len(previous_value):
-            previous_value = previous_value * rate + noise
+            previous_value = convert_to_np_array(previous_value) * rate + noise
 
         previous_value = deque(previous_value, maxlen=self.parameters.history._get(context))
 
@@ -995,7 +995,7 @@ def _initialize_previous_value(self, initializer, context=None):
                     warnings.warn(f"Attempt to initialize memory of {self.__class__.__name__} with an entry ({entry}) "
                                   f"that has the same key as a previous one, while 'duplicate_keys'==False; "
                                   f"that entry has been skipped")
-            return np.asarray(self._memory)
+            return convert_to_np_array(self._memory)
 
     def _instantiate_attributes_before_function(self, function=None, context=None):
         self.parameters.previous_value._set(
@@ -1125,7 +1125,7 @@ def _function(self,
         # Return 3d array with keys and vals as lists
         # IMPLEMENTATION NOTE:  if try to create np.ndarray directly, and keys and vals have same length
         #                       end up with array of arrays, rather than array of lists
-        ret_val = np.array([list(memory[0]),[]])
+        ret_val = convert_to_np_array([list(memory[0]),[]])
         ret_val[1] = list(memory[1])
         return ret_val
 
@@ -1326,7 +1326,7 @@ def delete_from_memory(self, memories:tc.any(list, np.ndarray), key_only:bool= T
 
     def _parse_memories(self, memories, method, context=None):
         """Parse passing of single vs. multiple memories, validate memories, and return ndarray"""
-        memories = np.array(memories)
+        memories = convert_to_np_array(memories)
         if not 1 <= memories.ndim <= 3:
             raise FunctionError(f"'memories' arg for {method} method of {self.__class__.__name__} "
                                 f"must be a 2-item list or 2d array, or a list or 3d array containing those")

psyneulink/core/components/mechanisms/mechanism.py

@@ -2397,7 +2397,7 @@ def execute(self,
 
                         return return_value
                 else:
-                    converted_to_2d = np.atleast_2d(return_value)
+                    converted_to_2d = convert_to_np_array(return_value, dimension=2)
                 # If return_value is a list of heterogenous elements, return as is
                 #     (satisfies requirement that return_value be an array of possibly multidimensional values)
                 if converted_to_2d.dtype == object:
@@ -2411,7 +2411,7 @@ def execute(self,
                 return_value = super()._execute(variable=self.defaults.variable,
                                                 context=context,
                                                 runtime_params=runtime_params)
-                return np.atleast_2d(return_value)
+                return convert_to_np_array(return_value, dimension=2)
 
         # SET UP RUNTIME PARAMS if any
 
@@ -2491,7 +2491,7 @@ def execute(self,
                                 for item in value))):
                         pass
                 else:
-                    converted_to_2d = np.atleast_2d(value)
+                    converted_to_2d = convert_to_np_array(value, dimension=2)
                     # If return_value is a list of heterogenous elements, return as is
                     #     (satisfies requirement that return_value be an array of possibly multidimensional values)
                     if converted_to_2d.dtype == object:
@@ -2537,7 +2537,7 @@ def execute(self,
         return value
 
     def _get_variable_from_input(self, input, context=None):
-        input = np.atleast_2d(input)
+        input = convert_to_np_array(input, dimension=2)
         num_inputs = np.size(input, 0)
         num_input_ports = len(self.input_ports)
         if num_inputs != num_input_ports:
@@ -2559,7 +2559,7 @@ def _get_variable_from_input(self, input, context=None):
                                      f"required length ({len(input_port.defaults.variable)}) for input "
                                      f"to {InputPort.__name__} {repr(input_port.name)} of {self.name}.")
 
-        return np.array(self.get_input_values(context))
+        return convert_to_np_array(self.get_input_values(context))
 
     def _update_input_ports(self, runtime_input_port_params=None, context=None):
         """Update value for each InputPort in self.input_ports:
@@ -2574,7 +2574,7 @@ def _update_input_ports(self, runtime_input_port_params=None, context=None):
             port= self.input_ports[i]
             port._update(params=runtime_input_port_params,
                          context=context)
-        return np.array(self.get_input_values(context))
+        return convert_to_np_array(self.get_input_values(context))
 
     def _update_parameter_ports(self, runtime_parameter_port_params=None, context=None):
 
@@ -3569,7 +3569,7 @@ def add_ports(self, ports, context=None):
                 else:
                     old_variable = self.defaults.variable
                 old_variable.extend(added_variable)
-                self.defaults.variable = np.array(old_variable)
+                self.defaults.variable = convert_to_np_array(old_variable)
             instantiated_input_ports = _instantiate_input_ports(self,
                                                                   input_ports,
                                                                   added_variable,
@@ -3993,7 +3993,7 @@ def _is_mechanism_spec(spec):
 from collections import UserList
 class MechanismList(UserList):
     """Provides access to Mechanisms and their attributes in a list Mechanisms of an owner.
-    
+
     Properties return dicts with item : attribute pairs.
     Recursively process any item that itself is a MechanismList (e.g., a `Nested Composition <Composition_Nested>`.
 

psyneulink/core/components/mechanisms/modulatory/control/controlmechanism.py

@@ -587,7 +587,7 @@
 from psyneulink.core.globals.parameters import Parameter
 from psyneulink.core.globals.preferences.basepreferenceset import is_pref_set
 from psyneulink.core.globals.preferences.preferenceset import PreferenceLevel
-from psyneulink.core.globals.utilities import ContentAddressableList, convert_to_list, copy_iterable_with_shared, is_iterable
+from psyneulink.core.globals.utilities import ContentAddressableList, convert_to_list, convert_to_np_array, copy_iterable_with_shared, is_iterable
 
 __all__ = [
     'CONTROL_ALLOCATION', 'GATING_ALLOCATION', 'ControlMechanism', 'ControlMechanismError', 'ControlMechanismRegistry',
@@ -674,9 +674,13 @@ def validate_monitored_port_spec(owner, spec_list):
 def _control_mechanism_costs_getter(owning_component=None, context=None):
     # NOTE: In cases where there is a reconfiguration_cost, that cost is not returned by this method
     try:
-        costs = [c.compute_costs(c.parameters.value._get(context), context=context)
-                 for c in owning_component.control_signals
-                 if hasattr(c, 'compute_costs')] # GatingSignals don't have cost fcts
+        costs = [
+            convert_to_np_array(
+                c.compute_costs(c.parameters.value._get(context), context=context)
+            )
+            for c in owning_component.control_signals
+            if hasattr(c, 'compute_costs')
+        ] # GatingSignals don't have cost fcts
         return costs
 
     except TypeError:
@@ -692,8 +696,10 @@ def _net_outcome_getter(owning_component=None, context=None):
     # NOTE: In cases where there is a reconfiguration_cost, that cost is not included in the net_outcome
     try:
         c = owning_component
-        return c.compute_net_outcome(c.parameters.outcome._get(context),
-                                     c.combine_costs(c.parameters.costs._get(context)))
+        return c.compute_net_outcome(
+            c.parameters.outcome._get(context),
+            c.combine_costs()
+        )
     except TypeError:
         return [0]
 

psyneulink/core/components/mechanisms/modulatory/control/optimizationcontrolmechanism.py

@@ -413,6 +413,7 @@
     DEFAULT_VARIABLE, EID_FROZEN, FUNCTION, INTERNAL_ONLY, NAME, \
     OPTIMIZATION_CONTROL_MECHANISM, OBJECTIVE_MECHANISM, OUTCOME, PRODUCT, PARAMS, \
     CONTROL, AUTO_ASSIGN_MATRIX
+from psyneulink.core.globals.utilities import convert_to_np_array
 from psyneulink.core.globals.parameters import Parameter, ParameterAlias
 from psyneulink.core.globals.preferences.preferenceset import PreferenceLevel
 from psyneulink.core.globals.context import handle_external_context
@@ -873,7 +874,7 @@ def _update_input_ports(self, runtime_params=None, context=None):
             port = self.input_ports[i]
             port._update(params=runtime_params, context=context)
             port_values.append(port.parameters.value._get(context))
-        return np.array(port_values)
+        return convert_to_np_array(port_values)
         # # MODIFIED 5/8/20 NEW:
         # input_port_values = super()._update_input_ports(runtime_params, context)
         # port_values.append(input_port_values)

psyneulink/core/components/mechanisms/modulatory/learning/learningmechanism.py

@@ -550,7 +550,7 @@
 from psyneulink.core.globals.parameters import Parameter
 from psyneulink.core.globals.preferences.basepreferenceset import is_pref_set
 from psyneulink.core.globals.preferences.preferenceset import PreferenceLevel
-from psyneulink.core.globals.utilities import ContentAddressableList, is_numeric, parameter_spec, convert_to_list
+from psyneulink.core.globals.utilities import ContentAddressableList, convert_to_np_array, is_numeric, parameter_spec, convert_to_list
 
 __all__ = [
     'ACTIVATION_INPUT', 'ACTIVATION_INPUT_INDEX', 'ACTIVATION_OUTPUT', 'ACTIVATION_OUTPUT_INDEX',
@@ -1352,9 +1352,13 @@ def _execute(
 
         # Compute learning_signal for each error_signal (and corresponding error-Matrix):
         for error_signal_input, error_matrix in zip(error_signal_inputs, error_matrices):
-            function_variable = np.array([variable[ACTIVATION_INPUT_INDEX],
-                                          variable[ACTIVATION_OUTPUT_INDEX],
-                                          error_signal_input])
+            function_variable = convert_to_np_array(
+                [
+                    variable[ACTIVATION_INPUT_INDEX],
+                    variable[ACTIVATION_OUTPUT_INDEX],
+                    error_signal_input
+                ]
+            )
             learning_signal, error_signal = super()._execute(variable=function_variable,
             # MODIFIED CROSS_PATHWAYS 7/22/19 END
                                                              context=context,

psyneulink/core/components/mechanisms/processing/transfermechanism.py

@@ -1080,6 +1080,10 @@ class Parameters(ProcessingMechanism_Base.Parameters):
             structural=True,
         )
 
+        def _validate_variable(self, variable):
+            if 'U' in str(variable.dtype):
+                return 'may not contain non-numeric entries'
+
         def _validate_integrator_mode(self, integrator_mode):
             if not isinstance(integrator_mode, bool):
                 return 'may only be True or False.'