cell_census.get_anndata should return obs columns and var columns used for filtering

[pablo-gar]

Issue

SOMADataFrame.read_as_pandas_all() retrieves all columns used for filtering for example:

in this case feature_id was included in the final object as it was used for filtering

However for the cell_census.get_anndata() method, this is not the case:

Solution

In cell_census.get_anndata() all columns used for filtering in obs_value_filter and var_value_filter must be included in the corresponding anndata.AnnData.obs and anndata.Anndata.var respectively

[atolopko-czi]

There is some effort involved with implementing this, as it requires parsing the declarative query string. We do have the BNF grammar for this, but is it worth doing this?

Also, I think there will be cases where a user might not want all of the query columns in the result.

[bkmartinjr]

I think this is a SOMA bug/inconsistency - shouldn't be hard to fix. I will get to it soon.

[bkmartinjr]

Root cause: single-cell-data/TileDB-SOMA#755

[bkmartinjr]

this is fixed in tiledbsoma, and will be available in the next release.