add the dataclasses for object metadata and conftest

[alorenzo175]

with examples of what metadata, observations, and forecasts will look like

closes #12

[alorenzo175]

@lboeman @wholmgren what do you think about dataclasses? They are nice because they can be "frozen" and be accessed easily. They will require python 3.7 though.

[wholmgren]

Based only on an hour of research into dataclasses, I'd say that they seem useful for defining the containers in the datamodel. Alternatives are standard classes, dicts, namedtuples. Aside from version compatibility, I don't see an advantage to a standard class or namedtuple for this application. dicts are nice because they are so simple, but perhaps too flexible.

requiring 3.7 is pretty aggressive, but I'm not opposed.

[alorenzo175]

@cwhanse @dplarson @awig Do any of you have reactions to using dataclasses for the site, forecast, and observation metadata? They provide a nice (almost) immutable container but Python 3.7 is required. I want to make sure 3.7 doesn't hinder development (although a conda create -n sfacore python=3.7 should get an environment going easily).

[cwhanse]

We have had issues testing pvlib in py3.7, although off the top of my head, I can't recall what they are. Something related to numpy functions.

[wholmgren]

@cwhanse I think pvlib/pvlib-python#634 solved those issues. I've been using 3.7 in some conda environments for the last few months without issues.

[cwhanse]

That was my reaction. For me objections is None from a brief scan of the py 3.7.2 release note. I'm not currently using py 3.7.

[alorenzo175]

@wholmgren Seems like 3.7 is ok. Availability and curtailment were added. anything else?

[wholmgren]

merge/rebase on master so your test runs