Use pandas or astropy.tables in Lightcurve?

[dhuppenkothen]

There was a suggestion today to use either pandas or astropy.tables within the Lightcurve class. pandas has the advantage that it has lots of functionality built-in.

astropy.tables has the advantage that it connects to other useful astropy concepts such as units and time concepts (which might allow us to provide time conversions more easily).

Not something that needs to be fixed instantaneously, more a call for discussion.

[nithinsingh61]

1. Pandas as in such doesn’t support fits and VOTable format(which are very relevant to us ),
2. Column cant be multidimensional (if we wish wish to store data(say bispectrum viz (nfft,nfft) ) of a fixed time range or energy range Additions to Lightcurve class #90 ) we would fail
3. Storing metadata (Additions to Lightcurve class #90) is not that flexible (Panda issue)

But pandas does have some good functionalities (groupby etc ) , May be we should use both pandas and astropy.tables?

[fustiman]

Tables fit better in Lightcurve and they also support grouping. I don't think necessary nor desirable including also pandas.

[abigailStev]

Keep an eye on this astropy feature request for a time series module based on astropy tables: astropy/astropy-APEs#12

[eteq]

@dhuppenkothen @nithinsingh61 - Not sure if you've actually done something with this or not, but something that might be relevant: Astropy tables support conversion to/from pandas.

So my thinking is that you might be better off having the "real" version be Astropy tables, because they support more complex data than pandas (e.g. times, coordinates), free metadata, and so on, but when you want pandas-compatible analysis tools/features, you can just do table.to_pandas() and use it as needed.

[evandromr]

Hi,

I've been playing around with Astropy tables and it really works well, including the group_by method which makes binning really easy.

Do we still want to use it for our Class objects (Lightcurve, Crosspectrum, Powerspectrum) ? If so I can give this issue a shot.

ping @dhuppenkothen @matteobachetti

[matteobachetti]

@dhuppenkothen I think we should pin this and think about the use of Astropy's timeseries objects