Model info as function attribute

[stestoll]

Currently, model info is returned by calling the model function without arguments, and an internal conditional statement distinguishes this call from the actual model call.

It would be cleaner (and more Pythonic) to use a function attribute for storing and providing these model metadata. See "User-defined functions" in the official Python doc.

Two possibilities:

• Store the info dict as a function attribute, e.g. bg_hom3d.info = dict(...).
• Store the key-value pairs from the info dict separately as function attributes, e.g. bg_hom3d.Parameters = ..., bg_hom3d.Units = ... etc.

The second possibility seems a lot cleaner. There is no need to introduce an additional dict layer as done in the first possibility.

In either case, we need to watch whether this structural change has any impact on performance when evaluating the model.

[mtessmer]

Using a dictionary should allow for individual access through the info attribute so you should be able to call bg_hom3d.info.Parameters and have the best of both worlds.

[stestoll]

The function attributes are stored as a dict, so introducing another layer of dict via info might not be necessary. If one wants to the whole dict, one can do info = bg_hom3d.__dict__.

[mtessmer]

That works as well; but, .__dict__ may also return any other (user-defined) attributes. An unlikely event that shouldn't disrupt any DeerLab internals if implemented in a safe way, but still one that can occur. There is still the issue of mutability too. You could make info a named tuple, but then we're back to the extra layer.