Add compression support for concat_on_disk about anndata HOT 2 OPEN

Thanks for the request! There are two things here:

h5ad files which are then gzip compressed

This is a weird file which we won't support. hdf5 supports internal compression with gzip, where each of in the internal chunks are compressed. This is what you get from AnnData.write_h5ad(..., compression="gzip"). We support this, and I concat_on_disk should already work with this type of file.

We, and h5py, don't support working with an entire hdf5 blob that has been compressed. E.g. the result of gzip file.h5ad.

• Writing a compressed file

While this is the default for zarr, this doesn't currently work with hdf5 and cannot be customized with the zarr backend. I think we could support this by passing though a dataset_kwargs argument as is done with write_elem.

Would you be up for a PR implementing this?

from anndata.

Thank you for the answer!

Hmm, interesting AnnData.write_h5ad(..., compression="gzip") is exactly what I did, but it caused an error. Maybe the problem was caused by me adding the ".gz" suffix to the files.

I can take a look at the possibility of opening PR next month.

from anndata.