Source code for openmdao.components.explicit_func_comp

"""Define the FuncComponent class."""

import functools
import numpy as np
from numpy import asarray, isscalar
from itertools import chain
from openmdao.core.explicitcomponent import ExplicitComponent
from openmdao.core.constants import INT_DTYPE
import openmdao.func_api as omf
from openmdao.components.func_comp_common import _check_var_name, _copy_with_ignore, _add_options

[docs]class ExplicitFuncComp(ExplicitComponent): """ A component that wraps a python function. Parameters ---------- compute : function The function to be wrapped by this Component. compute_partials : function or None If not None, call this function when computing partials. **kwargs : named args Args passed down to ExplicitComponent. Attributes ---------- _compute : callable The function wrapper used by this component. _compute_partials : function or None If not None, call this function when computing partials. """
[docs] def __init__(self, compute, compute_partials=None, **kwargs): """ Initialize attributes. """ super().__init__(**kwargs) self._compute = omf.wrap(compute) # in case we're doing jit, force setup of wrapped func because we compute output shapes # during setup and that won't work on a jit compiled function if self._compute._call_setup: self._compute._setup() self._compute_partials = compute_partials
def _declare_options(self): """ Declare options before kwargs are processed in the init method. """ super()._declare_options() _add_options(self)
[docs] def setup(self): """ Define out inputs and outputs. """ optignore = {'is_option'} for name, meta in self._compute.get_input_meta(): _check_var_name(self, name) if 'is_option' in meta and meta['is_option']: kwargs = _copy_with_ignore(meta, omf._allowed_declare_options_args, ignore=optignore) self.options.declare(name, **kwargs) else: kwargs = omf._filter_dict(meta, omf._allowed_add_input_args) self.add_input(name, **kwargs) for i, (name, meta) in enumerate(self._compute.get_output_meta()): _check_var_name(self, name) kwargs = _copy_with_ignore(meta, omf._allowed_add_output_args, ignore=('resid',)) self.add_output(name, **kwargs)
[docs] def compute(self, inputs, outputs): """ Compute the result of calling our function with the given inputs. Parameters ---------- inputs : Vector Unscaled, dimensional input variables. outputs : Vector Unscaled, dimensional output variables. """ outputs.set_vals(self._compute(*self._func_values(inputs)))
[docs] def declare_partials(self, *args, **kwargs): """ Declare information about this component's subjacobians. Parameters ---------- *args : list Positional args to be passed to base class version of declare_partials. **kwargs : dict Keyword args to be passed to base class version of declare_partials. Returns ------- dict Metadata dict for the specified partial(s). """ return super().declare_partials(*args, **kwargs)
def _setup_partials(self): """ Check that all partials are declared. """ for kwargs in self._compute.get_declare_partials(): self.declare_partials(**kwargs) kwargs = self._compute.get_declare_coloring() if kwargs is not None: self.declare_coloring(**kwargs) super()._setup_partials()
[docs] def compute_partials(self, inputs, partials): """ Compute sub-jacobian parts. The model is assumed to be in an unscaled state. Parameters ---------- inputs : Vector Unscaled, dimensional input variables read via inputs[key]. partials : Jacobian Sub-jac components written to partials[output_name, input_name]. """ if self._compute_partials is None: return self._compute_partials(*self._func_values(inputs), partials)
def _func_values(self, inputs): """ Yield current function input args. Parameters ---------- inputs : Vector The input vector. Yields ------ object Value of current function input variable. """ inps = inputs.values() for name, meta in self._compute._inputs.items(): if 'is_option' in meta: yield self.options[name] else: yield next(inps)