Following up from Slack: Checking in the derived data isn't a best-practice... but sometimes changes in the derived data a surprise, so it also functions as a test. In this case, the real edits are in fields.tsv, and everything else is derived from that. One consequence is that in the portal we'll start getting errors about bad assay types because the API is giving us strings that don't match the enumeration in the spec.
Following up from Slack: Checking in the derived data isn't a best-practice... but sometimes changes in the derived data a surprise, so it also functions as a test. In this case, the real edits are in
fields.tsv, and everything else is derived from that. One consequence is that in the portal we'll start getting errors about bad assay types because the API is giving us strings that don't match the enumeration in the spec.