WIP: Restructuring to enable multiple rows, multiple access points, and unit tests

[zoghbi-a]

This is an attempt to go beyond the simple use cases and consider a realistic case of a user submitting a list of products as either an astropy.Table or pyvo.dal.DALResults.

Both a json column and datalinks are processed to create a list of AccessPoint objects for each row of the data product.

AccessPoint and its subclasses (for aws, gc etc) have a download method that downloads the data

[zoghbi-a]

I have an implementation that integrates this into pyvo. It is in my pyvo cloud-links branch.

[bsipocz]

@zoghbi-a - a lot of it is already massaged into pyvo in astropy/pyvo#369, I'll come back to that PR and chop it up to smaller pieces.

[zoghbi-a]

Thanks @bsipocz,
Yes astropy/pyvo#369 does the json part. The datalinks part that I added is clunky and I plugged it in as proof of concept only. This restructuring allows both (json and datalink methods; plus others if needed, e.g. a column with s3 uris) to be done cleanly both for a single row and for a set of rows. This latter point is important, because for datalinks, it is inefficient to make a datalink call for every row. It is more efficient to make a single call requesting all datalinks. The restructured code allows that.

One other difference is how to deal with services other than SIA, which we want to consider in the future. With what we have, it is not straight forward to expand it to SSA and other services. With the new code, it is a matter of adding a single inheritance to the relevant service or Query class.

[bsipocz]

Yes, I don't mean to have that as it's any more but am cutting it up to pieces.

[zoghbi-a]

This will be closed in favor of a other PRs (#21 and #22).