DM-40381: Deduplicate when merging DatastoreRecordData and document preconditions.

[TallJimbo]

Checklist

• ran Jenkins
• added a release note for user-visible changes to doc/changes

[codecov]

Codecov Report

Patch coverage: 50.00% and project coverage change: -0.01% ⚠️

Comparison is base (41038f7) 87.71% compared to head (363042d) 87.70%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #879      +/-   ##
==========================================
- Coverage   87.71%   87.70%   -0.01%     
==========================================
  Files         274      274              
  Lines       36107    36108       +1     
  Branches     7552     7553       +1     
==========================================
  Hits        31670    31670              
  Misses       3264     3264              
- Partials     1173     1174       +1     

Files Changed	Coverage Δ
python/lsst/daf/butler/core/datastore.py	94.80% <ø> (ø)
python/lsst/daf/butler/core/datastoreRecordData.py	85.88% <50.00%> (-1.03%)	⬇️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[timj]

Looks okay once I realized that the idea here is that there are fully populated subsets of DatastoreRecordData floating around and when you call update you are noting that, assuming these are all from the same datastore, that if one record for a dataset_id is already present there is no need to copy over the records from the incoming object. If there is a mismatch in definitions for that record we have bigger problems.

Ideally DatastoreRecordData would have some idea about the primary keys that StoredDatastoreItemsInfo could report to it.

[TallJimbo]

Looks okay once I realized that the idea here is that there are fully populated subsets of DatastoreRecordData floating around and when you call update you are noting that, assuming these are all from the same datastore, that if one record for a dataset_id is already present there is no need to copy over the records from the incoming object. If there is a mismatch in definitions for that record we have bigger problems.

Yes, that's exactly it.

Ideally DatastoreRecordData would have some idea about the primary keys that StoredDatastoreItemsInfo could report to it.

For our current limited use cases for this export API (i.e. just QBB), I like keeping DatastoreRecordData being unaware of the real primary key, and just declaring that only full datasets (not components) can be exported this way. It's a lot simpler than any alternative I can think of, and I can't think of a scenario where it'd be a serious limitation.