Bridge bug: ICML filter extracts <Note> content with extractNotes=false (default)

[asgeirf]

Surfaced by the spec contract audit (PR for spec/icml).

Symptom

Per the bridge composite schema (dist/plugin/formats/okf_icml/schema.json) and the upstream Parameters.reset() in net.sf.okapi.filters.icml.Parameters, the default value of extractNotes is false. With this default, <Note> element content (editorial annotations authored in InCopy / InDesign) should be excluded from the translatable part stream.

The Okapi ICMLFilter instead emits <Note> content as a translatable text unit whenever a Note appears inside a CharacterStyleRange — and, while doing so, splits the surrounding visible text at the Note boundary so the trailing Content run after the Note is dropped from the part stream entirely.

Reproduction

core/formats/icml/spec.yaml feature note_extraction_default::notes_skipped over testdata/notes.icml:

<Story Self=\"story1\" ...>
  <ParagraphStyleRange AppliedParagraphStyle=\"ParagraphStyle/Body\">
    <CharacterStyleRange AppliedCharacterStyle=\"CharacterStyle/$ID/[No character style]\">
      <Content>Visible text</Content>
      <Note>
        <ParagraphStyleRange ...><CharacterStyleRange ...>
          <Content>Note content here</Content>
        </CharacterStyleRange></ParagraphStyleRange>
      </Note>
      <Content> continues</Content>
    </CharacterStyleRange>
  </ParagraphStyleRange>
</Story>

expected (extractNotes=false): 1 block, "Visible text continues"     (native, correct per schema default)
bridge:                        2 blocks: "Visible text" + "Note content here"   (note extracted; trailing \" continues\" dropped)

Spec tag

The example is tagged expected_fail: in the spec.yaml so the parity gate downgrades it to a warning until the bridge is fixed. Once the bridge fix lands, drop the expected_fail: block.

Repro test

KAPI_PARITY_SANDBOX=$PWD/.parity \
  go test -tags parity -run TestParityIcmlSpec/note_extraction_default -count=1 -v ./parity/...

(from cli/, on branch spec/icml)

Affected upstream code

okf_icml (okapi-filter-icml) — net.sf.okapi.filters.icml.ICMLFilter's Note-handling path, all currently-shipped Okapi versions affected.

[asgeirf]

Moved from neokapi/neokapi#506

Transferred during tracker housekeeping. The defect stays open here rather than being closed: okapi-bridge continues as a standalone product — the reference implementation of a third-party kapi plugin in a non-Go language (neokapi/neokapi#1073) — so a divergence in the okf_icml filter remains actionable independently of what neokapi ships natively.

What changed on the neokapi side

neokapi/neokapi#1079 removed the native icml format package. The reproduction path recorded in the body above therefore no longer resolves:

• core/formats/icml/spec.yaml (and its expected_fail: tag) was deleted with the package
• TestParityIcmlSpec and the spec/icml branch no longer exist

The report itself is unaffected. Its reference point is this repository's own composite schema for okf_icml, which records extractNotes defaulting to false (matching Parameters.reset() upstream), so the expectation can be checked here without a neokapi counterpart.

Reproducing here

Exercise net.sf.okapi.filters.icml.ICMLFilter directly from this repository's test suite with the notes.icml story fragment recorded above and default parameters. Two assertions capture the report: that no part carries the <Note> content, and that the visible <Content> runs flanking the Note arrive as a single unbroken text unit (the trailing run is currently dropped).