Add overlap option for cone search

[cdeil]

In Gammapy we currently call healpy.query_disc with both the inclusive=True and the inclusive=False option (see gammapy/gammapy#1167 (comment)).

As discussed in #41 , probably we don't want to spend the time to fully implement healpy.query_disc concerning the inclusive=False behaviour.

But maybe we could add an option to HEALPix.cone_search_lonlat that corresponds to the inclusive=True behaviour, i.e. restricts the pixels to the ones where the center is in the cone?

@astrofrog - OK to add this?

How about calling it overlap={'center', 'full'} which is maybe a bit more intuitive than inclusive = {True, False} and also more extensible should we want to add other options in the future?
Concerning implementation: should it just be done at the end of the function in an if overlap == 'center' block using a Numpy expression, subsetting with a mask that's true if the pixel center is in the cone? Or would you suggest this be added in the C or Cython code somehow?

[astrofrog]

I agree with this - I would maybe call the options overlap={'center', 'any'} because full to me sounds like the pixel has to overlap fully whereas really any overlap will do.

Implementation-wise, doing it in core.py in Python/Numpy is fine - if we want to avoid making temporary arrays then doing it in Cython would be better, but honestly I think just doing it in core.py would be fine for now and we can optimize later?

[tboch]

I agree with this - I would maybe call the options overlap={'center', 'any'} because full to me sounds like the pixel has to overlap fully whereas really any overlap will do.

+1 for using 'center' and 'any', sounds easier to understand to me

[cdeil]

Agreed that "full" was a complete misnomer and that overlap="any" is good.

Implementation-wise, doing it in core.py in Python/Numpy is fine - if we want to avoid making temporary arrays then doing it in Cython would be better, but honestly I think just doing it in core.py would be fine for now and we can optimize later?

Agreed. I'll try to find time to make a pull request later this week.

[taldcroft]

Discussed at astropy coordination meeting in Tucson. 👍 from @taldcroft and @adrn.