Fix typos and ambiguities

[minnerbe]

Hi Kevin, thanks for putting together this addition to the OME-NGFF spec!

I've been reading the table spec very carefully lately and came across some minor typos and (perceived) ambiguities. Therefore, I suggest a few changes that hopefully contribute to the spec being more precise and complete:

• Nullable integers seemed to be at the wrong hierarchy level within the var group, so I put them one level higher.
• The wording of the chunking recommendations for columns in the obs and var dataframes seemed ambiguous to me. I tried to make it as clear as possible and consistent with obsm and varm.
• I added some recommendations for chunking of arrays within non-basic dataframe columns (nullable integers and categoricals) as well as for sparse arrays. The rationale behind my changes is that arrays that are likely to be read together should ideally share chunk sizes (which was your original intention, I believe). In particular:
  • the codes, mask and values array should be chunked the same way as the corresponding dimension of X;
  • the data and indices arrays of sparse matrices should be chunked in the same way.

Furthermore, I am confused by the absence of nullable integers and categoricals in the obs group. In my understanding, obs and var should be qualitatively the same, but associated with different dimensions of X, is this correct?
If this absence is unintentional, I suggest to add nullable integers and categoricals also to obs and I'd be happy to do so.

[github-actions]

Automated Review URLs

• render latest/index.bs
• diff latest modified