[devin-ai-integration]

What and why

Follow-up to #1807 (merged); originally stacked on it, now merged with develop and targeting develop.

#1807 gives a port definition's implied ConjugatedPortDefinition a native element. For a port definition with a short name, the conjugate was named after the short name in three places that agreed with each other but not with the specification: native Name/ShortName ~s/~q, RDF declaredShortName "~s"/"~q" and an invented address segment P::@0::'~s', and a reverse converter that required them. (The ~q short name for port def <q> Q; is also present on develop, and develop could not round-trip port def <s>; through RDF.)

Now the conjugate is named ~ plus the original's name when the original has a name and is otherwise anonymous. It never has a short name:

• native (symbols/implied.go buildConjugatedPortDefinition): the conjugate's identification has Name ~Name only for a named original and never has a ShortName. The conjugate is still built for a short-name-only port. Its ids, OwningMembership and PortConjugation from fix(semantics): a port definition owns the ConjugatedPortDefinition it implies — its OwningMembership and PortConjugation read natively as the conversion writes them #1807 are unchanged. port x : ~s; still resolves, because the ~ typing resolves the original s and then conjugates it, so it never looks up the conjugate's own name. A test covers this case.
• positional identities (export/positional.go): an implied symbol's declaration span is its original's span, so an anonymous conjugate would have claimed its original's positional name (P::@0). positionalSymbol now skips implied symbols, which get their ids from ElementIDs.conjugateOf.
• RDF writer (export/rdf_normative.go conjugatedPortDefinition): writes no declaredShortName. It writes declaredName "~P" and sysml:qualifiedName <owner>::'~P' only for a named original. An anonymous conjugate gets no declared name and the address sysx:positionalName <owner>::@conjugated. That address follows the minted-segment form fix(export): an unnamed element's address is sysx:positionalName; sysml:qualifiedName states only the qualified name the metamodel derives #1805 documents in address.go for implied elements without a name (@trigger for a transition's trigger action). The unnamed-member form @N does not fit, because P::@0::@0 already names the first unnamed member of <s>'s own body.
• reverse converter (export/rdf_in_normative.go verifyConjugated): requires declaredName "~"+original name when the original is named, through its declaredName or a named address segment. It requires no declaredName when the original is unnamed. It accepts an absent declaredShortName, or exactly ~<original short> (the legacy value earlier writers emitted). It rejects anything else.

TestConversionNamesAShortNamedPortsConjugateByItsShortName asserted declaredShortName "~s" and an address ending ::'~s', which is the non-conformant behavior. It is replaced by TestConversionNamesConjugatesFromOriginalDeclaredNames, which asserts the spec behavior for all three originals. The ~q/~s expectations in TestPortDefinitionOwnsItsConjugate change for the same reason.

Specification basis

• SysML v2.1 Beta 2 §8.3.12.2 ConjugatedPortDefinition::effectiveName(): let originalName : String = originalPortDefinition.name in if originalName = null then null else '~' + originalName endif.
• KerML Element::shortName is derived by effectiveShortName(), whose default body is declaredShortName. ConjugatedPortDefinition does not redefine it, and the implied conjugate declares nothing, so its shortName is null in every case.
• KerML Element::qualifiedName is null if there is not a complete ownership chain of named Namespaces. The original <s> is unnamed.

The pinned pilot agrees. For package P { port def <s>; port def <q> Q; } its api-json exporter gives: the conjugate of <q> Q has name ~Q, shortName null, qualifiedName P::Q::'~Q' and declaredName/declaredShortName null; the conjugate of <s> has name, shortName and qualifiedName all null.

No docs/project/spec-compliance.md row moves.

How it was verified

• TestConjugatedPortDefinitionNamesFollowTheOriginalName (semantics) runs over <s>, <q> Q and R and a part typing ~s/~Q/~R. It checks the reflective name/declaredName/shortName/declaredShortName/qualifiedName of each conjugate and that each port's ConjugatedPortTyping reaches its original with no resolver diagnostics. A metamodel-evaluator test checks the derived Element::qualifiedName against the pilot's P::Q::'~Q'.
• TestConversionNamesConjugatesFromOriginalDeclaredNames checks the RDF properties and address for each original. TestElementIDReaderSeparatesShortPortAndAnonymousConjugate and TestPositionalIdentityIgnoresImpliedConjugate check that <s> and its conjugate read distinct graph-written ids.
• TestShortNamedPortDefinitionRoundTripsThroughRDF runs over port def <s>;, port def <q> Q;, port def Q; and the ~s/~Q/~R model. Each does ToRDF → ToSysML with sysx:sourceText stripped, then re-converts, and checks that the graph is identical. Reverse-converter tests cover the legacy declaredShortName "~q" (accepted), a wrong declaredName/declaredShortName (rejected) and a declaredName on the anonymous conjugate (rejected).
• Generated artifacts compared with the base: pilot-apijson-diff -update moved only the writer sourceDigest in docs/project/api-json-differential-baseline.json, and the scorecard did not move. make stdlib-snapshot-check is clean, so stdlib.snapshot did not move. Grep shows no port definition with a short name (port def <…>, short-only or name plus short) in examples/, tests/ or the downloaded training, pilot and spec-example corpora. That is why the corpus round-trip and scorecard do not move. TestCorpusRoundTrip passes with OPENSYSML_REQUIRE_TRAINING_CORPUS=1 OPENSYSML_REQUIRE_PILOT_CORPORA=1.
• go build ./..., go vet ./..., gofmt -l . (empty), go test -count=1 ./internal/semantic/... ./internal/translate/... ./tests/export/... ./tests/corpus/... ./internal/check/... ./internal/doc/queryexec/... ./internal/frontend/repl/... ./internal/workspace/libs/..., go -C tools test -count=1 ./referee/apijson/... ./oracle/baseline/....

Known difference not changed here: the semantics reflective qualifiedName feature joins raw names for every element (P::Q::~Q, as it would P::a b), while the metamodel evaluator escapes segments as the pilot does.

Checklist

• make test and make lint pass locally (the targeted gate above)
• Tests added or updated for the change
• Documentation extended where it already covers the surface (docs/reference/rdf-mapping.md)
• Changelog entry added as changes/unreleased/conjugated-port-definition-names.fixed.md
• baselines regenerated (pilot-apijson-diff -update: writer digest only)
• No internal work-item labels in the body, docs, or changelog

Link to Devin session: https://nasa-jpl-demo.devinenterprise.com/sessions/0e3626a6a66d49e49fc83db32d479123
Open in Devin Desktop: https://nasa-jpl-demo.devinenterprise.com/desktop/session/0e3626a6a66d49e49fc83db32d479123?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

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no bugs or issues to report.