int32 columns in obs break writing of adata to disk

[le-ander]

if I convert a singlecellexperiment object with anndata2ri, obs columns which should be int32 have a weird dtype in pandas. this breaks writing the adata to disk with the fillowing h5py error:

TypeError: Object dtype dtype('O') has no native HDF5 equivalent
Above error raised while writing key 'nFeature_RNA' of <class 'h5py._hl.group.Group'> from /.
Above error raised while writing key 'nFeature_RNA' of <class 'h5py._hl.group.Group'> from /.
Above error raised while writing key 'obs' of <class 'h5py._hl.files.File'> from /.

this fixes the error:

>>> adata.obs["nFeature_RNA"].dtype
Int32Dtype()
>>> adata.obs["nFeature_RNA"] = adata.obs["nFeature_RNA"].astype(np.int32)
>>> adata.obs["nFeature_RNA"].dtype
dtype('int32')

no more error when writing the adata to disk after this

[ivirshup]

I think the solution here would be for this library to not allow the nullable pandas dtypes. However, these do work in anndata 0.8.0, which you can try the release candidate of now.

The issue is that Int32Dtype() can have null values while dtype('int32') can't.