Add get_observer_state

[moeyensj]

Two main additions:

• Added get_observer state with unit tests (we are able to get observer states to within ~10 meters or so, which is pretty good)
• Added to_dataframe, from_dataframe definitions for Times that stores the time scale as part of the column name (this is important for round-tripping from a dataframe back to an instance).

Note that the line change counts are dominated by the addition of test files that each contain a few hundred data points.

The next PR will focus on making an actual Observers quivr table that will store these states.

[moeyensj]

I've got another commit in the works that will optimize the linear algebra in get_observer_state. Not that is super slow right now but it's kind of fun to make it faster.

[moeyensj]

Benchmarks at commit 8e2da8c:

Benchmarks at commit c2718ae:

[spenczar]

Intermediate comments, thanks for thoroughly commenting get_observer_state

[spenczar]

It's useful to know about typing.Self for cases like this.

Suppose I subclass Times, making my own class TDBTimes(Times) or something. What is TDBTimes.from_dataframe's return type?

As you have written here, it's a Times - the superclass. Which is not quite right, but mypy and other type checkers will just trust you on that, which could lead to spurious type checker errors.

The way out of this is to say that from_dataframe returns type Self. That way, you're telling mypy that the return type is "whatever cls is". So it would all work.

typing.Self was added only in Python 3.11, but it's backported to be available via typing_extensions as well.

Now - I don't think it matters practically for this case, nobody is really going to subclass Times, but I notice whenever a constructor method like this doesn't return Self.

[spenczar]

Could you add a little note on what these physically represent? Don't go crazy writing a lot, you can just link to a wikipedia article or whatever if you want.

[spenczar]

As of quivr v0.4.2, you don't need to use .values on pandas.Series objects anymore, you can pass them in directly to from_kwargs.

[spenczar]

Is there anything that you're aware of that you don't take into account? I don't have anything in mind, but it would be useful to spell out if so.

[moeyensj]

Great question, I got this from the SPICE documentation so I'm not sure what is missing. It should be a more robust frame than the standard IAU_EARTH that comes defined in DE440.

[spenczar]

Instead of the really big else block, return early here, which makes it clear that you really will do no more work.

[spenczar]

Is this slow? This looks slow.

I don't think it ought to be necessary to call spicepy for this. I think it should be possible to do this conversion yourself, since it's really just a change of epoch from JD epoch up to J2000. It might even be a constant shift for all values?

This is only worth investigating if it is indeed slow.

[moeyensj]

Yes, it's definitely slow. The conversion to ephemeris time in spiceypy requires the leapseconds kernel so I've just assumed it's doing something a bit more complex than what I can do outside of spiceypy, but TDB to ET should be nearly the same... Let me change it and see if the unit test tolerance changes at all.

[akoumjian]

Haven't had a check to review tests yet. Partial feedback.

[akoumjian]

Any reason to validate scale against an enum / set of some kind? Might catch problems earlier.