did2: daqmetadatareader migrator preserves tab_separated_file_parameter

[stevevanhooser]

Companion to waltham-data-science/did-schema#51 (which restores daqmetadatareader.tab_separated_file_parameter in V_delta as an optional pass-through field).

Summary

Update the daqmetadatareader migrator to preserve a populated v1 tab_separated_file_parameter value rather than silently dropping it. Real v1 corpora populate this "lazy hook" for TSV-per-epoch metadata sources; round-trip preservation lets them migrate cleanly without needing to synthesize a daqmetadatareader_tsv subclass per document.

What changed

• src/did/+did2/+convert/+migrators/daqmetadatareader.m — when the v1 block carries tab_separated_file_parameter, copy it into the V_delta newBlock under the same key. Idempotent if the v1 source lacks the field (the V_delta schema declares it mustBeNonEmpty: false, so absent is also valid). Docstring updated to reference the V_delta schema decision.

• tests/+did2/+unittest/testMigrators.m:

  • testDaqmetadatareaderRenamesReaderClass — updated to assert tab_separated_file_parameter is now preserved (previously asserted dropped). The reader_class rename and ndi_daqmetadatareader_class drop assertions are unchanged.
  • testDaqmetadatareaderTabSeparatedHookAbsentStaysAbsent — new: a v1 body without the optional hook does not gain an empty one in V_delta.
  • testDaqmetadatareaderTabSeparatedHookRoundtripsAlreadyVDelta — new: a body already in V_delta shape (with both reader_class and tab_separated_file_parameter) passes through unchanged.

CI dependency on did-schema#51

The validator at src/did/+did2/+schema/cache.m::validateDocument enforces a strict-fields check that raises did2:validation:undeclaredField for any property-block field not declared by the class schema. Until did-schema#51 merges to main (which this repo's CI workflow checks out at ref: main in .github/workflows/test-code.yml), the V_delta daqmetadatareader schema does not declare tab_separated_file_parameter, so any corpus test that runs a v1 doc with the TSV hook populated through did2.convert.v1_to_v2(..., Validate=true) will fail here.

Expected merge order: merge did-schema#51 first, then this PR will go green.

The migrator-level tests added here do not validate (they call did2.convert.migrators.daqmetadatareader and universalRenames directly), so they pass independently of the schema state. The risk is only in the larger end-to-end corpus tests (PRED, 20211116) if any of those v1 docs populate the TSV hook.

Test plan

CI will tell us. Local environment has no MATLAB per AGENTS.md.

---

Generated by Claude Code