feat: update validation for obs['in_tissue'] to include descendants of Visiium

[ejmolinelli]

Reason for Change

• cellxgene-schema CLI must update validation for obs['in_tissue'] #1104

Changes

• update _validate_spatial_tissue_position to support descendants of Visium
• refactor _is_visium_including_descendants to work around non-idempotent dependence on ONTOLOGY_PARSER COG#246

Testing

• test_schema_compliance.test_column_presence_in_tissue tests descendants and non descendants

pip install git+https://github.com/chanzuckerberg/single-cell-curation/@main#subdirectory=cellxgene_schema_cli

Notes for Reviewer

• the implementation for checking descendants relies on a new utility method. This was just a way to have a single method implementation that can be used as an apply() function and be cached to avoid unnecessary graph traversals in ontology

[joyceyan]

i would prefer passing in the ONTOLOGY_PARSER that we create an instance of in validate.py into is_ontological_descendant_of, that way we only have one place to update when we bump the schema version

[ejmolinelli]

yes, I see your point. I definitely agree we should minimize (to 1) the number of places a bump needs to be reflected. I'll do that.

One question I have is, why is the version in the code at all and not as a configuration option?

[joyceyan]

this was previously set to something like get_current_schema_version(). however, the current schema version is 5.2.0, because we bump the version number as part of the automated release process. and in the past, we would usually bump the ontology versions in COG completing all the feature changes to the validator. there's some more context in this slack thread but basically i decided hard-coding this was easier than needing to change some of our automated release process to make sure we're not double-bumping the version number before creating the new 5.3.0 release.

[ejmolinelli]

I've followed your suggestion. Good looking out @joyceyan