[devin-ai-integration]

What and why

Naming a member by its naming feature (#1304) raised the api-json differential's matched elements on every corpus root but docs/manual/examples, which fell 6834 → 6663. This PR adds the referee diagnostic that names the moved matches, uses it to account for every one of them, and records the finding.

No derivation of ours disagrees with the spec. The 171 is the net of 357 matches lost and 186 gained:

movement	elements	shape	verdict
lost	357 = 87 members + the 270 elements they own	84 AttributeUsage + 3 ReferenceUsage, each attribute redefines caption|title|text|term|description|kind|direction|style|groupBy = … or ref redefines source|target = … of a feature of DocumentQueries (internal/workspace/libs/stdlib/OpenSysML Libraries/), with their Redefinition, FeatureValue and literal / FeatureReferenceExpression value	referee limitation. Ours keys them …::coverage::title per KerML 8.3.3.3.4 Feature::effectiveName (namingFeature().effectiveName() when nothing is declared) with the default namingFeature() = ownedRedefinition->at(1).redefinedFeature. The pilot's export shows name/qualifiedName null and the Redefinition's redefinedFeature pointing at an id outside its key table — the bridge loads only the pinned sysml.library, so it cannot resolve this repository's own libraries and keys the member structurally (…::coverage/AttributeUsage[0]). The earlier match was that structural key lining up with our then-positional one. All 87 verified in the pilot's JSON.
gained	186 = 46 members + 138 owned + 2	46 attribute :>> x = … redefining features the model declares (with Redefinition, FeatureValue, 27 LiteralString / 19 LiteralRational), plus render asInterconnectionDiagram; (SysML 8.3.26.6 RenderingUsage::namingFeature) and its ReferenceSubsetting	the intended effect of #1304; both sides now derive the same name

Specification basis

Clauses read for the adjudication: KerML 8.3.2.1.2 Element (effectiveName, name = effectiveName()), 8.3.3.3.4 Feature (effectiveName, namingFeature); SysML 8.3.6.4 Usage (variant), 8.3.6.3 ReferenceUsage (transition payload), 8.3.17.14 PerformActionUsage, 8.3.20.4 ConstraintUsage, 8.3.26.6 RenderingUsage. Nothing was made stricter than the spec; no derivation code changed.

The same shape on the other seven roots

Before → after #1304, fresh cache per tree, every root (the -lost-against listing of the after run against the before run's inventory):

root	matched before → after	lost	gained	lost members and their shape
training	6097 → 7287	0	1190	—
pilot-examples	19299 → 25778	0	6479	—
pilot-validation	6997 → 9246	0	2249	—
kerml-examples	5399 → 6387	0	988	—
spec-examples	6654 → 6768	57	171	13 members whose naming feature the pinned excerpt leaves undeclared (perform VehicleActions::move :> … 7.17.6/2, exhibit VehicleStates::monitoring 7.18.4/2, attribute :>> massActual under an undefined MaximumMass 7.21.2/3, render asMyTreeDiagram 7.26.2/2–3, frame 'modularity' 7.26.3/2, individual … redefines aircraft|preflight|… 7.9.4/2, step redefines focus KerML 7.4.7.3/3) + what they own; the pilot resolves nothing and names nothing, we name by the reference as written (the same reference we export as @ref)
testdata	1124 → 1165	0	41	—
examples	21488 → 23379	914	2805	205 DocumentQueries / AnalysisRecords redefinitions (OpenSysMLDocument, Records, Reporting, RoverChecks, RelationshipMatrixReport) + 2 render asMixedDiagram|asCaseDiagram of OpenSysMLRenderings (LanderViews), + what they own
manual-examples	6834 → 6663	357	186	the 87 above

Every lost member on every root is the one shape — a naming feature the pilot does not resolve (an OpenSysML-only library it does not load, or a feature the excerpt never declares) — verified element by element in the pilot's JSON (name/qualifiedName null, redefinedFeature/referencedFeature id outside its key table). No lost member has a naming feature both sides resolve.

Diagnostic

Every run now writes pilot-apijson-diff-matches.json (per root: every key under matched / ours-only / pilot-only with its metaclass). -lost-against <earlier run>/pilot-apijson-diff-matches.json writes pilot-apijson-diff-movements.txt|json: the matches lost and gained since that run, grouped by root and metaclass, each with the key's fate in the run that does not match it. Tested without Java (TestMatchMovementsNameEveryMovedKey, TestCompareRootRecordsTheInventory) and end to end: a run of this branch against the pre-#1304 inventory lists exactly the 357 / 186 above.

Record

docs/project/api-json-differential.md gains "Where the pilot cannot name a member" (the limitation, the clauses, the counts) and documents the new files and flag. The baseline is unchanged (already 6663 on the base branch). Changelog fragment changes/unreleased/api-json-match-movements.added.md.

Behaviour changes

None in the writer or the semantics; the referee's report and baseline shape are unchanged (the inventory is an extra file).

Overlapping work checked

• Open PRs touching internal/translate/export / internal/semantic/semantics naming or the baseline: feat(export): owned absence and evaluator-derived properties in API JSON #1232 (feature/api-json-derived-properties-layered, api-json writer — nothing here depends on it, nothing duplicated), feat(corpus): gate the specification's own textual-notation examples #1302 (base of this PR), fix(tools): re-record the API JSON differential baseline after the errata overlay change #1321 / docs(project): re-record the API JSON differential baseline over current develop #1322 (docs-only baseline re-records), feat(semantics): usages subset the base library usage of their kind #1319 / fix(runtime): nested action and state usages are performances their owner holds #1310 / fix(check): report redefinition-not-end only inside associations and connectors #1305 / feat(parser): parse the five productions the specification examples use #1315 / fix(libs): record a document whose metadata names undeclared types or elements #1314 (semantics, no naming-feature overlap).
• Remote branches without a PR: none touch export or semantics naming.
• One concurrent referee change records the exporter bridge's source digest in the provenance (tools/referee/apijson/provenance.go); this PR does not touch that file.

Follow-up (not in this PR)

Loading OpenSysML Libraries/ into the bridge alongside the pinned sysml.library would let the pilot resolve these features and turn the 87 into real comparisons; it changes the bridge source (and so its recorded digest) and the baseline of two roots, so it is a change of its own.

How it was verified

• go build ./..., go vet ./..., gofmt -l . (empty), go test ./..., go test -C tools ./..., make lint, make docs-counts, python3 scripts/changelog.py check — all pass.
• go run -C tools ./cmd/pilot-apijson-diff -check reproduces the committed baseline on every root.
• OPENSYSML_REQUIRE_TRAINING_CORPUS=1 OPENSYSML_REQUIRE_PILOT_CORPORA=1 go test -count=1 ./tests/corpus -run TestCorpusRoundTrip passes (no export change, nothing moved).
• The inventory and -lost-against are pinned without Java by TestMatchMovementsNameEveryMovedKey and TestCompareRootRecordsTheInventory, and exercised end to end as described above (a run of this branch over docs/manual/examples with -lost-against the pre-fix(export): name a view's render tree; and other naming-feature members by their effective name #1304 inventory lists exactly 357 lost / 186 gained, every lost member re-read in the pilot's own JSON).

Checklist

• make test and make lint pass locally
• Tests added or updated for the change
• Documentation extended where it already covers the surface (see CONTRIBUTING.md)
• Changelog entry added as changes/unreleased/<slug>.<section>.md, not as an edit to CHANGELOG.md
• baselines regenerated and make docs-counts run if a gate count moved (no count moved; the baseline is unchanged)
• No internal work-item labels (waves, slices, F4, K5) in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/556a00af681842bf9d52a532275fc761
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/556a00af681842bf9d52a532275fc761?variant=devin
Requested by: @HuiJun

[devin-ai-integration]

I'll fix CI failures and address comments from users with write access. I'll skip comments containing "(aside)".

• Disable automatic comment, CI, and merge conflict monitoring