Design Decisions Regarding None-Types For Parameter Validation(s) Steps

[AFg6K7h4fhy2]

Goal

Determine design pattern for parameters of None type in the _assert_sample_and_rtype function (and possibly other functions)

Context

From TPM:

One uncertainty I had was the default behavior for the _assert_sample_and_rtype function generally. Is it a better pattern to not have a skip_if_none param and just instead have _assert_sample_and_rtype be able to handle these scenarios (latent_admissions v. observation_process, None or not None) internally?

From GVY:

That's a good question. I could be useful to post this as an issue/question. Right now, we have the NullRV class which somewhat deals with the None cases. Perhaps we could simply remove the option to skip checks as it should always be an instance of RandomVariable.

Code:

see here for example usage

_assert_sample_and_rtype(latent_infections, skip_if_none=False)
_assert_sample_and_rtype(observation_process, skip_if_none=False)

see here for the _assert_sample_and_rtype

Out of scope

• Complete redefinition of _assert_sample_and_rtype

[damonbayer]

Related: #314 & #350

[dylanhmorris]

Closing since the relevant validation patterns have now changed.