Krzysztof Findeisen, mind reviewing this set of small changes to butler tests and test utility code, and some downstream tests that use the latter?

The bottom line is that DM-21840 is soon going to add some NOT NULL constraints for implied dimensions, and that breaks test code that has thus far gotten away with e.g. declaring visit data IDs without any filters. I think you'll agree that having the test code this dependent on the dimensions schema is not ideal, at least for testing code that doesn't depend on the data IDs it's given (of course, some might), but I also don't see any real alternatives (short of making the test utility code much more reliant on mocks), and at least the fixes are small.

Changes are spread across multiple packages. I identified these packages as needing fixes by running Jenkins on DM-21840 and seeing what broke; I then just rebased the test fixes to this ticket.

• daf_butler
• pipe_base
• obs_base
• verify
• pipe_tasks

Krzysztof Findeisen, I think this is ready for another look. I've addressed the big problem with the original change: the test utility code now invents data ID values for dependencies when they're needed. The changes to daf_butler were more substantial than our discussion on the PR predicted, but the PRs for most other packages have been closed and their branches deleted. The pipe_tasks one is still there because it didn't use the test utility code and hence still needs more manual dimension record insertions to be compatible with the upcoming NOT NULL constraint.

Thank you for your patience, I'm fine with these changes now.