fix: keep CJK opening brackets with annotations

chenglou · mayrang · chenglou · commit d9f2dffeeef8 · 2026-04-18T14:16:28.000+08:00
Skip backward punctuation merges when the previous segment contains CJK so opening brackets can flow through to the forward-sticky pass and attach to the following annotation text. Closes #145 Refs #148. Co-authored-by: mayrang <pkss0626@naver.com>
diff --git a/src/analysis.ts b/src/analysis.ts
@@ -1008,6 +1008,7 @@ function buildMergedSegmentation(
         !piece.isWordLike &&
         mergedLen > 0 &&
         mergedKinds[prevIndex] === 'text' &&
+        !mergedContainsCJK[prevIndex] &&
         (
           isLeftStickyPunctuationSegment(piece.text) ||
           (piece.text === '-' && mergedWordLike[prevIndex]!)
@@ -1056,7 +1057,8 @@ function buildMergedSegmentation(
       mergedKinds[i] === 'text' &&
       !mergedWordLike[i]! &&
       isEscapedQuoteClusterSegment(mergedTexts[i]!) &&
-      mergedKinds[i - 1] === 'text'
+      mergedKinds[i - 1] === 'text' &&
+      !mergedContainsCJK[i - 1]
     ) {
       mergedTexts[i - 1] += mergedTexts[i]!
       mergedWordLike[i - 1] = mergedWordLike[i - 1]! || mergedWordLike[i]!
diff --git a/src/layout.test.ts b/src/layout.test.ts
@@ -622,6 +622,14 @@ describe('prepare invariants', () => {
     expect(isCJK('hello')).toBe(false)
   })
 
+  test('keeps opening brackets after CJK attached to following annotation text', () => {
+    expect(prepareWithSegments('서울(Seoul)과', FONT).segments).toEqual(['서', '울', '(Seoul)', '과'])
+    expect(prepareWithSegments('東京(Tokyo)と', FONT).segments).toEqual(['東', '京', '(Tokyo)', 'と'])
+    expect(prepareWithSegments('北京(Beijing)和', FONT).segments).toEqual(['北', '京', '(Beijing)', '和'])
+    expect(prepareWithSegments('참조[1]와', FONT).segments).toEqual(['참', '조', '[1]', '와'])
+    expect(prepareWithSegments('AB(CD)', FONT).segments).toEqual(['AB(', 'CD)'])
+  })
+
   test('prepare and prepareWithSegments agree on layout behavior', () => {
     const plain = prepare('Alpha beta gamma', FONT)
     const rich = prepareWithSegments('Alpha beta gamma', FONT)
