[jesshaley]

Builds three decisions from did-schema V_eta_go_forward_class_audit.md (2026-10-29). Nothing here has been run in MATLAB; CI is the first run.

[image label goes to the observation's notes] (supersedes [image label goes to notes])

• +migrators_j/image_stack.m: when the source is a v1 image, the statement notes of the intensity observation is now the image block's label. It is written only when the label is non-empty, exactly as v1 wrote it. An imageStack still uses its own label. The data body gets no new field.
• +migrators_j/image.m: removes the carry-051 carry of a labelled image, and the sourceId helper that only that carry used.
• Tests (testMigratorsJ.m): testDidV1ImageWithALabelIsCarriedAsV1 is inverted to testDidV1ImageLabelIsTheObservationsNotes. New test testDidV1ImageWithNoLabelHasEmptyNotes.
• Round trip not updated: there is no did2.convert.roundtrip.image. A v1 image is not read as raw v1 (rawV1Classes.m says why), so there is no inverse to change. Whether to build one is an open question.

[ngrid migrates to sampled bodies] + [ngrid with no statement carried]

• +migrators/ngrid.m: a standalone ngrid is now carried whole as v1, needs_mapping. Before, the block was reshaped and data_size and coordinates were deleted (did-schema sqlitedb: make each write operation one transaction, recovered by replay #227). Under every V2 target, a document with an ngrid superclass gets +migrators_j/+super/ngrid.m, so this base migrator only runs on a document whose class is ngrid. That makes sqlitedb: make each write operation one transaction, recovered by replay #227's loss the same code, and it is fixed here.
• +migrators_j/ontology_image.m: a vintage-B ontologyImage whose only non-empty edge is ontologyTableRow_id is now carried, needs_mapping, with a carried_as_v1 note. Before, it passed through with no note. The emitted body is the same as before.
• The fold into sampled_body was already built wherever a statement exists (jNgridBody: hartley_calc, and the ontology_image raster arm). It is unchanged.
• Tests:
  • testMigrators.m: the 3 ngrid tests that checked the reshape are inverted. Added a dispatcher test showing a standalone ngrid keeps all four v1 fields.
  • testNgridCoordinates.m: testBaseNgridMigratorStillReshapesAndDrops is inverted to testBaseNgridMigratorCarriesTheBlockUnchanged.
  • testRawV1RoundTripOntologyImage.m: testTheWriterDocumentRoundTripsAsAPassthrough is inverted to testTheWriterDocumentIsCarriedAndRoundTrips. New test: a writer document with an empty table row still passes through.
  • testNgridSampledBodyFold.m and testMigratorsJ.m: added assertions that the carry is noted.
• Not covered by the decision, so left unchanged and asked: a vintage-B ontologyImage with no non-empty edge (imageDocMaker writes one when it is given no row), or with a subject edge next to the row.

[binaryseries parameters carried] (with [binaryseries parameters fold as decided])

• +migrators_j/binaryseries_parameters.m: every v1 document is now carried whole as v1, needs_mapping. The invented-field guard and the raw read (carry-049 / carry-100) still run first. The placeholder normalising and the char-integer refusal are removed.
• +roundtrip/binaryseries_parameters.m: header updated.
• Tests:
  • testMiscSingletons.m: testBinaryseriesTemplatePlaceholdersAreDropped and testBinaryseriesRejectsNonEmptyCharInAnIntegerField are inverted. The populated-body test now checks the carry.
  • testRawV1RoundTripDemo.m: the binaryseries fixture's not-carried round trip is inverted to testBinaryseriesParametersIsCarriedAndRoundTrips.

Carry register

• 3 new carry sites have no register row yet. Each marker says none, and check_carry_citations.py lists them:
  • +migrators/ngrid.m: standalone ngrid
  • +migrators_j/ontology_image.m: ontologyImage whose only edge is the table row
  • +migrators_j/binaryseries_parameters.m
• carry-051 (image.label) is no longer cited by any carry site.

Python checks

• check_carry_citations.py: OK. It reports 74 sites and 7 with no row; 3 of those 7 are new.
• test_check_carry_citations.py: 19 tests, 0 failures.
• check_migrator_roster.py: OK, after running --write.
• The other tools/test_*.py files also pass.

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 0.00%. Comparing base (e6bb66e) to head (163f602).
⚠️ Report is 7 commits behind head on V2.

Additional details and impacted files

@@    Coverage Diff    @@
##   V2   #288   +/-   ##
=========================
=========================

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.