2.7.1. sdepy.kfunc

sdepy.kfunc(f=None, *, nvar=None)

Decorator to wrap classes or functions as objects with managed keyword arguments.

This decorator, intended as an aid to interactive and notebook sessions, wraps a callable, class or function, as a “kfunc” object that handles separately its parameters (keyword-only), whose values are stored in the object, and its variables (positional or keyword), always provided upon evaluation.

Syntax:

@kfunc
class my_class:
    def __init___(self, **kwparams):
        ...
    def __call__(self, *var, **kwvar):
        ...

@kfunc(nvar=k)
def my_function(*var, **kwargs):
    ...

After decoration, my_class is a kfunc with kwparams as parameters, and with var and kwvar as variables, and my_function is a kfunc with the first k of var, kwargs as variables, and the remaining kwargs as parameters. For usage, see examples below.

Examples

Wrap wiener_source into a kfunc, named dw:

>>> from sdepy import wiener_source, kfunc
>>> dw = kfunc(wiener_source)

Instantiate dw and evaluate it (this is business as usual):

>>> my_instance = dw(paths=100, dtype=np.float32)
>>> x = my_instance(t=0, dt=1)
>>> x.shape, x.dtype
((100,), dtype('float32'))

Inspect kfunc parameters stored in my_instance:

>>> my_instance.params  # doctest: +SKIP
{'paths': 100, 'vshape': (), 'dtype': <class 'numpy.float32'>,         'corr': None, 'rho': None}

Evaluate my_instance changing some parameters (call the instance with one or more):

>>> x = my_instance(t=0, dt=1, paths=999)
>>> x.shape, x.dtype
((999,), dtype('float32'))

Parameters stored in my_instance are not affected:

>>> my_instance.paths == my_instance.params['paths'] == 100
True

Create a new instance, changing some parameters and keeping those already set (call the instance without passing any variables):

>>> new_instance = my_instance(vshape=2, rho=0.5)
>>> new_instance.params  # doctest: +SKIP
{'paths': 100, 'vshape': 2, 'dtype': <class 'numpy.float32'>,         'corr': None, 'rho': 0.5}

Instantiate and evaluate at once (pass one or more variables to the class constructor):

>>> x = dw(0, 1, paths=100, dtype=np.float32)
>>> x.shape, x.dtype
((100,), dtype('float32'))

As long as variables are passed by name, order doesn’t matter (omitted variables take default values, if any):

>>> x = dw(paths=100, dtype=np.float32, dt=1, t=0)
>>> x.shape, x.dtype
((100,), dtype('float32'))

Attributes:

params : dictionary

Parameter values stored in the instance (read-only). For wrapped SDE subclasses, also includes default values of all SDE-specific parameters, as stored in the args attribute.