ScipyModel subclasses fail to restore constant values when using from_dict

[csala]

Environment Details

Please indicate the following details about the environment in which you found the bug:

• Copulas version: 0.4.0

Error Description

If ScipyModel subclasses are built using a dict specification that implies a constant value (such as scale=0), the constant value being passed is not properly restored, resulting on a model that samples None every time.

Steps to reproduce

To reproduce, a Univariate instance can be created passing a constant value specification, such as scale=0 on a GaussianUnivariate.

In [1]: from copulas.univariate import Univariate

In [2]: univariate = Univariate.from_dict({
   ...:     'type': 'copulas.univariate.gaussian.GaussianUnivariate',
   ...:     'loc': 10,
   ...:     'scale': 0
   ...: })

In [3]: univariate.sample(5)
Out[3]: array([None, None, None, None, None], dtype=object)

Fix

To fix this, a new method _extract_constant can be added to the ScipyModel subclasses which extracts the constant value encoded in the params and returns it, and then _set_constant_value must be used instead of _replace_constant_methods