fix: align line walkers and Hangul jamo breaks

chenglou · mayrang · voidborne-d · chenglou · commit c28eecb0e362 · 2026-04-18T13:26:59.000+08:00
Normalize chunked batch line starts through the same segment-kind policy used by streaming so layoutWithLines(), walkLineRanges(), layoutNextLine(), and layout() stay aligned after zero-width break opportunities and collapsible spaces. Classify Hangul Compatibility Jamo (U+3130..U+318F) as CJK so common Korean compatibility jamo runs break like browser text. Refs #129, #135, #141. Closes #121 Closes #142 Co-authored-by: mayrang <pkss0626@naver.com> Co-authored-by: voidborne-d <voidborne-d@users.noreply.github.com> Co-authored-by: lttlin <lttlin@gmail.com>
diff --git a/src/analysis.ts b/src/analysis.ts
@@ -121,6 +121,7 @@ function isCJKCodePoint(codePoint: number): boolean {
     (codePoint >= 0x3000 && codePoint <= 0x303F) ||
     (codePoint >= 0x3040 && codePoint <= 0x309F) ||
     (codePoint >= 0x30A0 && codePoint <= 0x30FF) ||
+    (codePoint >= 0x3130 && codePoint <= 0x318F) ||
     (codePoint >= 0xAC00 && codePoint <= 0xD7AF) ||
     (codePoint >= 0xFF00 && codePoint <= 0xFFEF)
   )
diff --git a/src/layout.test.ts b/src/layout.test.ts
@@ -79,6 +79,7 @@ function isWideCharacter(ch: string): boolean {
     (code >= 0x3000 && code <= 0x303F) ||
     (code >= 0x3040 && code <= 0x309F) ||
     (code >= 0x30A0 && code <= 0x30FF) ||
+    (code >= 0x3130 && code <= 0x318F) ||
     (code >= 0xAC00 && code <= 0xD7AF) ||
     (code >= 0xFF00 && code <= 0xFFEF)
   )
@@ -552,6 +553,19 @@ describe('prepare invariants', () => {
     expect(prepareWithSegments('&#53580;&#49828;&#53944;&#51077;&#45768;&#45796;.', FONT).segments.at(-1)).toBe('&#45796;.')
   })
 
+  test('treats Hangul compatibility jamo as CJK break units', () => {
+    const prepared = prepareWithSegments('&#12619;&#12619;&#12619; &#51652;&#51676;', FONT)
+    expect(prepared.segments).toEqual(['&#12619;', '&#12619;', '&#12619;', ' ', '&#51652;', '&#51676;'])
+
+    const width = measureWidth('&#12619;&#12619;', FONT) + 0.1
+    const lines = layoutWithLines(prepared, width, LINE_HEIGHT)
+    expect(lines.lines.map(line => line.text)).toEqual(['&#12619;&#12619;', '&#12619; ', '&#51652;&#51676;'])
+    expect(layout(prepared, width, LINE_HEIGHT)).toEqual({
+      lineCount: 3,
+      height: LINE_HEIGHT * 3,
+    })
+  })
+
   test('keeps non-CJK glue-connected runs intact before CJK text', () => {
     const prepared = prepareWithSegments('foo\u00A0&#19990;&#30028;', FONT)
     expect(prepared.segments).toEqual(['foo\u00A0', '&#19990;', '&#30028;'])
@@ -600,7 +614,8 @@ describe('prepare invariants', () => {
     }
   })
 
-  test('isCJK covers the newer CJK extension blocks', () => {
+  test('isCJK covers Hangul compatibility jamo and the newer CJK extension blocks', () => {
+    expect(isCJK('&#12619;')).toBe(true)
     expect(isCJK('\u{2EBF0}')).toBe(true)
     expect(isCJK('\u{31350}')).toBe(true)
     expect(isCJK('\u{323B0}')).toBe(true)
@@ -840,6 +855,16 @@ describe('layout invariants', () => {
     expect(layoutWithLines(prepared, width, LINE_HEIGHT).lines).toEqual(collectStreamedLines(prepared, width))
   })
 
+  test('chunked batch line walking normalizes spaces after zero-width breaks like streaming', () => {
+    const prepared = prepareWithSegments('x\u00AD A\u200B B', FONT)
+    const width = measureWidth('x A', FONT) + 0.1
+    const batched = layoutWithLines(prepared, width, LINE_HEIGHT)
+
+    expect(batched.lines.map(line => line.text)).toEqual(['x A\u200B', 'B'])
+    expect(collectStreamedLines(prepared, width)).toEqual(batched.lines)
+    expect(layout(prepared, width, LINE_HEIGHT).lineCount).toBe(batched.lineCount)
+  })
+
   test('layoutNextLine can resume from any fixed-width line start without hidden state', () => {
     const prepared = prepareWithSegments('foo trans\u00ADatlantic said "hello" to &#19990;&#30028; and waved. alpha\u200Bbeta &#128640;', FONT)
     const width = 90
diff --git a/src/line-break.ts b/src/line-break.ts
@@ -38,13 +38,28 @@ type InternalLineVisitor = (
   endGraphemeIndex: number,
 ) => void
 
-function normalizeSimpleLineStartSegmentIndex(
+function consumesAtLineStart(kind: SegmentBreakKind): boolean {
+  return kind === 'space' || kind === 'zero-width-break' || kind === 'soft-hyphen'
+}
+
+function breaksAfter(kind: SegmentBreakKind): boolean {
+  return (
+    kind === 'space' ||
+    kind === 'preserved-space' ||
+    kind === 'tab' ||
+    kind === 'zero-width-break' ||
+    kind === 'soft-hyphen'
+  )
+}
+
+function normalizeLineStartSegmentIndex(
   prepared: PreparedLineBreakData,
   segmentIndex: number,
+  endSegmentIndex = prepared.widths.length,
 ): number {
-  while (segmentIndex < prepared.widths.length) {
+  while (segmentIndex < endSegmentIndex) {
     const kind = prepared.kinds[segmentIndex]!
-    if (kind !== 'space' && kind !== 'zero-width-break' && kind !== 'soft-hyphen') break
+    if (!consumesAtLineStart(kind)) break
     segmentIndex++
   }
   return segmentIndex
@@ -113,14 +128,11 @@ function normalizeLineStartInChunk(
   }
 
   if (segmentIndex < chunk.startSegmentIndex) segmentIndex = chunk.startSegmentIndex
-  while (segmentIndex < chunk.endSegmentIndex) {
-    const kind = prepared.kinds[segmentIndex]!
-    if (kind !== 'space' && kind !== 'zero-width-break' && kind !== 'soft-hyphen') {
-      cursor.segmentIndex = segmentIndex
-      cursor.graphemeIndex = 0
-      return chunkIndex
-    }
-    segmentIndex++
+  segmentIndex = normalizeLineStartSegmentIndex(prepared, segmentIndex, chunk.endSegmentIndex)
+  if (segmentIndex < chunk.endSegmentIndex) {
+    cursor.segmentIndex = segmentIndex
+    cursor.graphemeIndex = 0
+    return chunkIndex
   }
 
   if (chunk.consumedEndSegmentIndex >= prepared.widths.length) return -1
@@ -274,13 +286,13 @@ function walkPreparedLinesSimple(
   let i = 0
   while (i < widths.length) {
     if (!hasContent) {
-      i = normalizeSimpleLineStartSegmentIndex(prepared, i)
+      i = normalizeLineStartSegmentIndex(prepared, i)
       if (i >= widths.length) break
     }
 
     const w = widths[i]!
     const kind = kinds[i]!
-    const breakAfter = kind === 'space' || kind === 'preserved-space' || kind === 'tab' || kind === 'zero-width-break' || kind === 'soft-hyphen'
+    const breakAfter = breaksAfter(kind)
 
     if (!hasContent) {
       if (w > maxWidth && breakableFitAdvances[i] !== null) {
@@ -525,8 +537,13 @@ export function walkPreparedLinesRaw(
 
     let i = chunk.startSegmentIndex
     while (i < chunk.endSegmentIndex) {
+      if (!hasContent) {
+        i = normalizeLineStartSegmentIndex(prepared, i, chunk.endSegmentIndex)
+        if (i >= chunk.endSegmentIndex) break
+      }
+
       const kind = kinds[i]!
-      const breakAfter = kind === 'space' || kind === 'preserved-space' || kind === 'tab' || kind === 'zero-width-break' || kind === 'soft-hyphen'
+      const breakAfter = breaksAfter(kind)
       const w = kind === 'tab' ? getTabAdvance(lineW, tabStopAdvance) : widths[i]!
 
       if (kind === 'soft-hyphen') {
@@ -800,7 +817,7 @@ function stepPreparedChunkLineGeometry(
 
   for (let i = cursor.segmentIndex; i < chunk.endSegmentIndex; i++) {
     const kind = kinds[i]!
-    const breakAfter = kind === 'space' || kind === 'preserved-space' || kind === 'tab' || kind === 'zero-width-break' || kind === 'soft-hyphen'
+    const breakAfter = breaksAfter(kind)
     const startGraphemeIndex = i === cursor.segmentIndex ? cursor.graphemeIndex : 0
     const w = kind === 'tab' ? getTabAdvance(lineW, tabStopAdvance) : widths[i]!
 
@@ -902,7 +919,7 @@ function stepPreparedSimpleLineGeometry(
   for (let i = cursor.segmentIndex; i < widths.length; i++) {
     const w = widths[i]!
     const kind = kinds[i]!
-    const breakAfter = kind === 'space' || kind === 'preserved-space' || kind === 'tab' || kind === 'zero-width-break' || kind === 'soft-hyphen'
+    const breakAfter = breaksAfter(kind)
     const startGraphemeIndex = i === cursor.segmentIndex ? cursor.graphemeIndex : 0
     const breakableFitAdvance = breakableFitAdvances[i]
 

Original file line number	Diff line number	Diff line change
`@@ -121,6 +121,7 @@ function isCJKCodePoint(codePoint: number): boolean {`
`121`	`121`	`(codePoint >= 0x3000 && codePoint <= 0x303F) \|\|`
`122`	`122`	`(codePoint >= 0x3040 && codePoint <= 0x309F) \|\|`
`123`	`123`	`(codePoint >= 0x30A0 && codePoint <= 0x30FF) \|\|`
	`124`	`+ (codePoint >= 0x3130 && codePoint <= 0x318F) \|\|`
`124`	`125`	`(codePoint >= 0xAC00 && codePoint <= 0xD7AF) \|\|`
`125`	`126`	`(codePoint >= 0xFF00 && codePoint <= 0xFFEF)`
`126`	`127`	`)`