Review logic & use cases for reconciling CAS and Anndata files.

[dosumis]

TDT Workflows to support.

1. Starting point = spreadsheet + Anndata file + config. Workflow steps:
   • Ingest spreadsheet to generate CAS w/o cell_ids
   • run checks to test before populating cell_ids - check labels and values of all labelsets in config are in CAS and AnnData.
     • If cluster (rank 0) not in Anndata - fail.
     • If other labelsets in config are present in Anndata - checks keys and values match & warn if not. If they are not present. Pass.
   • populate cell_ids
   • If labelsets other than cluster are present - check that the hierarchy matches (compare cell sets for all). If they don't match. warn.
   • If everything passes to this point
     • record validation status in schema?
     • generate checksum for whole AnnData file and store.
   • Editing --> save changes to AnnData - no further checks are needed. BUT Save changes to AnnData should flush and replace labelset (Flush = remove column in dataframe). Warning should make this clear.
   • Generate checksum for whole AnnData file for later use in validation.
2. Starting point: existing TDT repo with config, CAS file + cell_ids and linked AnnData file.
   • Run checks in 1 on startup ().
3. Starting point: CAS files with cell_ids shared by PURL.
   • Download AnnData (over-ridden if in local)

Review CAS tools functionality in light of these workflows.