@peaceroad/markdown-it-strong-ja 0.4.5 → 0.5.0

package/index.js

@@ -1,848 +1,1371 @@
-const CHAR_ASTERISK = 0x2A    // *
-//const CHAR_UNDERSCORE = 0x5F  // _
-const CHAR_BACKSLASH = 0x5C   // \
-const CHAR_BACKTICK = 0x60    // `
-const CHAR_DOLLAR = 0x24      // $
-const CHAR_LT = 0x3C          // <
-const CHAR_GT = 0x3E          // >
-const CHAR_SLASH = 0x2F       // /
-const CHAR_SPACE = 0x20       // ' ' (space)
-
-const REG_ASTERISKS = /^\*+$/
-const REG_ATTRS = /{[^{}\n!@#%^&*()]+?}$/
-const REG_PUNCTUATION = /[!-/:-@[-`{-~ ]/
-const REG_JAPANESE = /\p{Script=Hiragana}|\p{Script=Katakana}|\p{Script=Han}|\p{General_Category=Punctuation}|\p{General_Category=Symbol}|\p{General_Category=Format}|\p{Emoji}/u // ひらがな|カタカナ|漢字|句読点|記号|フォーマット文字|絵文字
-
-const REG_MARKDOWN_HTML = /^\[[^\[\]]+\]\([^)]+\)$|^<([a-zA-Z][a-zA-Z0-9]*)[^>]*>([^<]+<\/\1>)$|^`[^`]+`$|^\$[^$]+\$$/ // for mixed-language context detection
-
-const hasBackslash = (state, start) => {
-  let slashNum = 0
-  let i = start - 1
-  const src = state.src
-  // Early exit if no backslash at all
-  if (i < 0 || src.charCodeAt(i) !== CHAR_BACKSLASH) {
-    return false
-  }
-  // Count consecutive backslashes efficiently
-  while (i >= 0 && src.charCodeAt(i) === CHAR_BACKSLASH) {
-    slashNum++
-    i--
+import Token from 'markdown-it/lib/token.mjs'
+
+const CHAR_ASTERISK = 0x2A    // *
+//const CHAR_UNDERSCORE = 0x5F  // _
+const CHAR_BACKSLASH = 0x5C   // \
+const CHAR_BACKTICK = 0x60    // `
+const CHAR_DOLLAR = 0x24      // $
+const CHAR_LT = 0x3C          // <
+const CHAR_GT = 0x3E          // >
+const CHAR_SLASH = 0x2F       // /
+const CHAR_SPACE = 0x20       // ' ' (space)
+const CHAR_OPEN_BRACKET = 0x5B // [
+const CHAR_CLOSE_BRACKET = 0x5D // ]
+
+const REG_ASTERISKS = /^\*+$/
+const REG_ATTRS = /{[^{}\n!@#%^&*()]+?}$/
+const REG_PUNCTUATION = /[!-/:-@[-`{-~ ]/
+const REG_JAPANESE = /\p{Script=Hiragana}|\p{Script=Katakana}|\p{Script=Han}|\p{General_Category=Punctuation}|\p{General_Category=Symbol}|\p{General_Category=Format}|\p{Emoji}/u // ひらがな|カタカナ|漢字|句読点|記号|フォーマット文字|絵文字
+
+const REG_MARKDOWN_HTML = /^\[[^\[\]]+\]\([^)]+\)$|^<([a-zA-Z][a-zA-Z0-9]*)[^>]*>([^<]+<\/\1>)$|^`[^`]+`$|^\$[^$]+\$$/ // for mixed-language context detection
+
+const hasBackslash = (state, start) => {
+  let slashNum = 0
+  let i = start - 1
+  const src = state.src
+  // Early exit if no backslash at all
+  if (i < 0 || src.charCodeAt(i) !== CHAR_BACKSLASH) {
+    return false
+  }
+  // Count consecutive backslashes efficiently
+  while (i >= 0 && src.charCodeAt(i) === CHAR_BACKSLASH) {
+    slashNum++
+    i--
+  }
+  return slashNum % 2 === 1
+}
+
+const findMatchingBracket = (state, start, max, openChar, closeChar) => {
+  let depth = 1
+  let pos = start + 1
+  const src = state.src
+  while (pos < max) {
+    const ch = src.charCodeAt(pos)
+    if (ch === openChar && !hasBackslash(state, pos)) {
+      depth++
+    } else if (ch === closeChar && !hasBackslash(state, pos)) {
+      depth--
+      if (depth === 0) return pos
+    }
+    pos++
+  }
+  return -1
+}
+
+const findRefRangeIndex = (pos, refRanges) => {
+  if (!refRanges || refRanges.length === 0) return -1
+  const cache = refRanges.__cache
+  if (cache && cache.has(pos)) return cache.get(pos)
+  let left = 0
+  let right = refRanges.length - 1
+  while (left <= right) {
+    const mid = left + Math.floor((right - left) / 2)
+    const range = refRanges[mid]
+    if (pos < range.start) {
+      right = mid - 1
+    } else if (pos > range.end) {
+      left = mid + 1
+    } else {
+      const result = range.hasReference ? mid : -1
+      if (cache) cache.set(pos, result)
+      return result
+    }
   }
-  return slashNum % 2 === 1
+  if (cache) cache.set(pos, -1)
+  return -1
 }
-
-const setToken = (state, inlines, opt) => {
+
+// Detect reference-link label ranges within the current inline slice
+const computeReferenceRanges = (state, start, max) => {
   const src = state.src
-  let i = 0
-  let attrsIsText = {
-    val: false,
-    tag: '',
-  }
-  while (i < inlines.length) {
-    let type = inlines[i].type
-    const tag = type.replace(/(?:_open|_close)$/, '')
-
-    if (/_open$/.test(type)) {
-      const startToken = state.push(type, tag, 1)
-      startToken.markup = tag === 'strong' ? '**' : '*'
-      attrsIsText = {
-        val: true,
-        tag: tag,
-      }
-    }
-
-    if (type === 'html_inline') {
-      type = 'text'
-    }
-    if (type === 'text') {
-      let content = src.slice(inlines[i].s, inlines[i].e + 1)
-      if (REG_ASTERISKS.test(content)) {
-        const asteriskToken = state.push(type, '', 0)
-        asteriskToken.content = content
-        i++
-        continue
-      }
-      if (opt.mditAttrs && attrsIsText.val && i + 1 < inlines.length) {
-        const hasImmediatelyAfterAsteriskClose = inlines[i+1].type === attrsIsText.tag + '_close'
-        if (hasImmediatelyAfterAsteriskClose && REG_ATTRS.test(content)) {
-          const attrsToken = state.push(type, '', 0)
-
-          const hasBackslashBeforeCurlyAttribute = content.match(/(\\+){/)
-          if (hasBackslashBeforeCurlyAttribute) {
-            if (hasBackslashBeforeCurlyAttribute[1].length === 1) {
-              attrsToken.content = content.replace(/\\{/, '{')
-            } else {
-              let backSlashNum = Math.floor(hasBackslashBeforeCurlyAttribute[1].length / 2)
-              let k = 0
-              let backSlash = ''
-              while (k < backSlashNum) {
-                backSlash +=  '\\'
-                k++
+  const references = state.env && state.env.references
+  const hasReferences = references && Object.keys(references).length > 0
+  const firstBracket = src.indexOf('[', start)
+  if (firstBracket === -1 || firstBracket >= max) return []
+  const ranges = []
+  let pos = start
+  while (pos < max) {
+    if (src.charCodeAt(pos) === CHAR_OPEN_BRACKET && !hasBackslash(state, pos)) {
+      const labelClose = findMatchingBracket(state, pos, max, CHAR_OPEN_BRACKET, CHAR_CLOSE_BRACKET)
+      if (labelClose !== -1) {
+        const nextPos = labelClose + 1
+        if (nextPos < max && src.charCodeAt(nextPos) === CHAR_OPEN_BRACKET && !hasBackslash(state, nextPos)) {
+          const refClose = findMatchingBracket(state, nextPos, max, CHAR_OPEN_BRACKET, CHAR_CLOSE_BRACKET)
+          if (refClose !== -1) {
+            let hasReference = false
+            if (hasReferences) {
+              if (refClose === nextPos + 1) {
+                const labelRaw = src.slice(pos + 1, labelClose)
+                const normalizedLabel = normalizeReferenceCandidate(state, labelRaw, { useClean: true })
+                hasReference = !!references[normalizedLabel]
+              } else {
+                const refRaw = src.slice(nextPos + 1, refClose)
+                const normalizedRef = normalizeReferenceCandidate(state, refRaw)
+                hasReference = !!references[normalizedRef]
               }
-              attrsToken.content = content.replace(/\\+{/, backSlash + '{')
             }
-          } else {
-            attrsToken.content = content
-          }
-          attrsIsText.val = false
-          i++
-          continue
-        }
-      }
-
-      const childTokens = state.md.parseInline(content, state.env)
-      if (childTokens[0] && childTokens[0].children) {
-        let j = 0
-        while (j < childTokens[0].children.length) {
-          const t = childTokens[0].children[j]
-          if (t.type === 'softbreak' && !opt.mdBreaks) {
-            t.type = 'text'
-            t.tag = ''
-            t.content = '\n'
-          }
-          if (!opt.mditAttrs && t.tag === 'br') {
-            t.tag = ''
-            t.content = '\n'
+            if (hasReference) {
+              ranges.push({ start: pos, end: labelClose, hasReference: true })
+              ranges.push({ start: nextPos, end: refClose, hasReference: true })
+            }
+            pos = refClose
+            continue
           }
-          const token = state.push(t.type, t.tag, t.nesting)
-          token.attrs = t.attrs
-          token.map = t.map
-          token.level = t.level
-          token.children = t.children
-          token.content = t.content
-          token.markup = t.markup
-          token.info = t.info
-          token.meta = t.meta
-          token.block = t.block
-          token.hidden = t.hidden
-          j++
         }
       }
     }
+    pos++
+  }
+  if (ranges.length) {
+    ranges.__cache = new Map()
+  }
+  return ranges
+}
+
+const copyInlineTokenFields = (dest, src) => {
+  if (src.attrs) dest.attrs = src.attrs
+  if (src.map) dest.map = src.map
+  dest.level = src.level
+  if (src.children) dest.children = src.children
+  dest.content = src.content
+  dest.markup = src.markup
+  if (src.info) dest.info = src.info
+  if (src.meta) dest.meta = src.meta
+  dest.block = src.block
+  dest.hidden = src.hidden
+}
 
-    if (/_close$/.test(type)) {
-      const closeToken = state.push(type, tag, -1)
-      closeToken.markup = tag === 'strong' ? '**' : '*'
-      attrsIsText = {
-        val: false,
-        tag: '',
-      }
-    }
-
-    i++
+const inlineHasCollapsedRef = (state) => {
+  if (state.__strongJaHasCollapsedRefs === undefined) {
+    state.__strongJaHasCollapsedRefs = /\[[^\]]*\]\s*\[[^\]]*\]/.test(state.src)
   }
+  return state.__strongJaHasCollapsedRefs
 }
 
-const pushInlines = (inlines, s, e, len, type, tag, tagType) => {
-  const inline = {
-    s: s,
-    sp: s,
-    e: e,
-    ep: e,
-    len: len,
-    type: type,
-    check: type === 'text',
+const registerCollapsedRefTarget = (state) => {
+  const env = state.env
+  if (!env.__strongJaCollapsedTargets) {
+    env.__strongJaCollapsedTargets = []
+    env.__strongJaCollapsedTargetSet = typeof WeakSet !== 'undefined' ? new WeakSet() : null
+  }
+  const targets = env.__strongJaCollapsedTargets
+  const targetSet = env.__strongJaCollapsedTargetSet
+  if (targetSet) {
+    if (targetSet.has(state.tokens)) return
+    targetSet.add(state.tokens)
+  } else if (targets.includes(state.tokens)) {
+    return
   }
-  if (tag) inline.tag = [tag, tagType]
-  inlines.push(inline)
+  targets.push(state.tokens)
 }
 
-const hasNextSymbol = (state, n, max, symbol, noMark) => {
-  let nextSymbolPos = -1
+const setToken = (state, inlines, opt) => {
+  const src = state.src
+  let i = 0
+  let attrsIsText = {
+    val: false,
+    tag: '',
+  }
+  while (i < inlines.length) {
+    let type = inlines[i].type
+    const tag = type.replace(/(?:_open|_close)$/, '')
+
+    if (/_open$/.test(type)) {
+      const startToken = state.push(type, tag, 1)
+      startToken.markup = tag === 'strong' ? '**' : '*'
+      attrsIsText = {
+        val: true,
+        tag: tag,
+      }
+    }
+
+    if (type === 'html_inline') {
+      type = 'text'
+    }
+    if (type === 'text') {
+      let content = src.slice(inlines[i].s, inlines[i].e + 1)
+      if (REG_ASTERISKS.test(content)) {
+        const asteriskToken = state.push(type, '', 0)
+        asteriskToken.content = content
+        i++
+        continue
+      }
+      if (opt.mditAttrs && attrsIsText.val && i + 1 < inlines.length) {
+        const hasImmediatelyAfterAsteriskClose = inlines[i+1].type === attrsIsText.tag + '_close'
+        if (hasImmediatelyAfterAsteriskClose && REG_ATTRS.test(content)) {
+          const attrsToken = state.push(type, '', 0)
+
+          const hasBackslashBeforeCurlyAttribute = content.match(/(\\+){/)
+          if (hasBackslashBeforeCurlyAttribute) {
+            if (hasBackslashBeforeCurlyAttribute[1].length === 1) {
+              attrsToken.content = content.replace(/\\{/, '{')
+            } else {
+              let backSlashNum = Math.floor(hasBackslashBeforeCurlyAttribute[1].length / 2)
+              let k = 0
+              let backSlash = ''
+              while (k < backSlashNum) {
+                backSlash +=  '\\'
+                k++
+              }
+              attrsToken.content = content.replace(/\\+{/, backSlash + '{')
+            }
+          } else {
+            attrsToken.content = content
+          }
+          attrsIsText.val = false
+          i++
+          continue
+        }
+      }
+
+      const childTokens = state.md.parseInline(content, state.env)
+      if (childTokens[0] && childTokens[0].children) {
+        let j = 0
+        while (j < childTokens[0].children.length) {
+          const t = childTokens[0].children[j]
+          if (t.type === 'softbreak' && !opt.mdBreaks) {
+            t.type = 'text'
+            t.tag = ''
+            t.content = '\n'
+          }
+          if (!opt.mditAttrs && t.tag === 'br') {
+            t.tag = ''
+            t.content = '\n'
+          }
+          const token = state.push(t.type, t.tag, t.nesting)
+          copyInlineTokenFields(token, t)
+          j++
+        }
+      }
+    }
+
+    if (/_close$/.test(type)) {
+      const closeToken = state.push(type, tag, -1)
+      closeToken.markup = tag === 'strong' ? '**' : '*'
+      attrsIsText = {
+        val: false,
+        tag: '',
+      }
+    }
+
+    i++
+  }
+}
+
+const pushInlines = (inlines, s, e, len, type, tag, tagType) => {
+  const inline = {
+    s: s,
+    sp: s,
+    e: e,
+    ep: e,
+    len: len,
+    type: type,
+    check: type === 'text',
+  }
+  if (tag) inline.tag = [tag, tagType]
+  inlines.push(inline)
+}
+
+const findNextSymbolPos = (state, n, max, symbol) => {
   const src = state.src
-  if (src.charCodeAt(n) === symbol && !hasBackslash(state, n)) {
-    for (let i = n + 1; i < max; i++) {
-      noMark += src[i]
-      if (src.charCodeAt(i) === symbol && !hasBackslash(state, i)) {
-        noMark += src.substring(n, i + 1)
-        nextSymbolPos = i
-        break
-      }
+  if (src.charCodeAt(n) !== symbol || hasBackslash(state, n)) return -1
+  for (let i = n + 1; i < max; i++) {
+    if (src.charCodeAt(i) === symbol && !hasBackslash(state, i)) {
+      return i
     }
   }
-  return [nextSymbolPos, noMark]
+  return -1
 }
 
 const processSymbolPair = (state, n, srcLen, symbol, noMark, textStart, pushInlines) => {
-  const [nextSymbolPos, newNoMark] = hasNextSymbol(state, n, srcLen, symbol, noMark)
-  if (nextSymbolPos !== -1) {
-    if (nextSymbolPos === srcLen - 1) {
-      pushInlines(textStart, nextSymbolPos, nextSymbolPos - textStart + 1, 'text')
-      return { shouldBreak: true, newN: nextSymbolPos + 1, newNoMark }
-    }
-    return { shouldBreak: false, shouldContinue: true, newN: nextSymbolPos + 1, newNoMark }
+  const nextSymbolPos = findNextSymbolPos(state, n, srcLen, symbol)
+  if (nextSymbolPos === -1) {
+    return { shouldBreak: false, shouldContinue: false, newN: n, newNoMark: noMark }
   }
-  return { shouldBreak: false, shouldContinue: false, newN: n, newNoMark }
-}
-
-const processTextSegment = (inlines, textStart, n, noMark) => {
-  if (n !== 0 && noMark.length !== 0) {
-    pushInlines(inlines, textStart, n - 1, n - textStart, 'text')
-    return ''
+  const src = state.src
+  const innerText = src.slice(n + 1, nextSymbolPos)
+  const markup = src.slice(n, nextSymbolPos + 1)
+  const newNoMark = noMark + innerText + markup
+  if (nextSymbolPos === srcLen - 1) {
+    pushInlines(textStart, nextSymbolPos, nextSymbolPos - textStart + 1, 'text')
+    return { shouldBreak: true, newN: nextSymbolPos + 1, newNoMark }
   }
-  return noMark
-}
-
-const createInlines = (state, start, max, opt) => {
+  return { shouldBreak: false, shouldContinue: true, newN: nextSymbolPos + 1, newNoMark }
+}
+
+const processTextSegment = (inlines, textStart, n, noMark) => {
+  if (n !== 0 && noMark.length !== 0) {
+    pushInlines(inlines, textStart, n - 1, n - textStart, 'text')
+    return ''
+  }
+  return noMark
+}
+
+const createInlines = (state, start, max, opt) => {
+  const src = state.src
+  const srcLen = max
+  const htmlEnabled = state.md.options.html
+  let n = start
+  let inlines = []
+  let noMark = ''
+  let textStart = n
+  
+  // Infinite loop prevention
+  const maxIterations = srcLen * 2 // Safe upper bound
+  let iterations = 0
+  
+  while (n < srcLen) {
+    // Prevent infinite loops
+    iterations++
+    if (iterations > maxIterations) {
+      // Add remaining text as-is and exit safely
+      if (textStart < srcLen) {
+        pushInlines(inlines, textStart, srcLen - 1, srcLen - textStart, 'text')
+      }
+      break
+    }
+    
+    const currentChar = src.charCodeAt(n)
+    
+    // Unified escape check
+    let isEscaped = false
+    if (currentChar === CHAR_ASTERISK || currentChar === CHAR_BACKTICK || 
+        (opt.dollarMath && currentChar === CHAR_DOLLAR) || 
+        (htmlEnabled && currentChar === CHAR_LT)) {
+      isEscaped = hasBackslash(state, n)
+    }
+
+    // Asterisk handling
+    if (currentChar === CHAR_ASTERISK) {
+      if (!isEscaped) {
+        noMark = processTextSegment(inlines, textStart, n, noMark)
+        if (n === srcLen - 1) {
+          pushInlines(inlines, n, n, 1, '')
+          break
+        }
+        let i = n + 1
+        while (i < srcLen && src.charCodeAt(i) === CHAR_ASTERISK) {
+          i++
+        }
+        if (i === srcLen) {
+          pushInlines(inlines, n, i - 1, i - n, '')
+        } else {
+          pushInlines(inlines, n, i - 1, i - n, '')
+          textStart = i
+        }
+        n = i
+        continue
+      }
+    }
+
+    // Inline code (backticks)
+    if (currentChar === CHAR_BACKTICK) {
+      if (!isEscaped) {
+        const result = processSymbolPair(state, n, srcLen, CHAR_BACKTICK, noMark, textStart, 
+          (start, end, len, type) => pushInlines(inlines, start, end, len, type))
+        if (result.shouldBreak) break
+        if (result.shouldContinue) {
+          n = result.newN
+          noMark = result.newNoMark
+          continue
+        }
+        noMark = result.newNoMark
+      }
+    }
+
+    // Inline math ($...$)
+    if (opt.dollarMath && currentChar === CHAR_DOLLAR) {
+      if (!isEscaped) {
+        const result = processSymbolPair(state, n, srcLen, CHAR_DOLLAR, noMark, textStart, 
+          (start, end, len, type) => pushInlines(inlines, start, end, len, type))
+        if (result.shouldBreak) break
+        if (result.shouldContinue) {
+          n = result.newN
+          noMark = result.newNoMark
+          continue
+        }
+        noMark = result.newNoMark
+      }
+    }
+
+    // HTML tags
+    if (htmlEnabled && currentChar === CHAR_LT) {
+      if (!isEscaped) {
+        let foundClosingTag = false
+        for (let i = n + 1; i < srcLen; i++) {
+          if (src.charCodeAt(i) === CHAR_GT && !hasBackslash(state, i)) {
+            noMark = processTextSegment(inlines, textStart, n, noMark)
+            let tag = src.slice(n + 1, i)
+            let tagType
+            if (tag.charCodeAt(0) === CHAR_SLASH) {
+              tag = tag.slice(1)
+              tagType = 'close'
+            } else {
+              tagType = 'open'
+            }
+            pushInlines(inlines, n, i, i - n + 1, 'html_inline', tag, tagType)
+            textStart = i + 1
+            n = i + 1
+            foundClosingTag = true
+            break
+          }
+        }
+        if (foundClosingTag) {
+          continue
+        }
+        // If no closing tag found, treat as regular character to prevent infinite loops
+      }
+    }
+
+    // Regular character
+    noMark += src[n]
+    if (n === srcLen - 1) {
+      pushInlines(inlines, textStart, n, n - textStart + 1, 'text')
+      break
+    }
+    n++
+  }
+  return inlines
+}
+
+const pushMark = (marks, opts) => {
+  // Maintain sorted order during insertion
+  const newMark = {
+    nest: opts.nest,
+    s: opts.s,
+    e: opts.e,
+    len: opts.len,
+    oLen: opts.oLen,
+    type: opts.type
+  }
+  // Binary search for insertion point to maintain sorted order
+  let left = 0
+  let right = marks.length
+  while (left < right) {
+    const mid = Math.floor((left + right) / 2)
+    if (marks[mid].s <= newMark.s) {
+      left = mid + 1
+    } else {
+      right = mid
+    }
+  }
+  
+  marks.splice(left, 0, newMark)
+}
+
+const setStrong = (state, inlines, marks, n, memo, opt, nestTracker, refRanges) => {
+  if (opt.disallowMixed === true) {
+    let i = n + 1
+    const inlinesLength = inlines.length
+    while (i < inlinesLength) {
+      if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
+      if (inlines[i].type !== '') { i++; continue }
+      
+      if (inlines[i].len > 1) {
+        const mixedCheck = checkMixedLanguagePattern(state, inlines, n, i, opt)
+        if (mixedCheck.shouldBlock) {
+          return [n, 0]
+        }
+        break
+      }
+      i++
+    }
+  }
+  
+  const strongOpenRange = findRefRangeIndex(inlines[n].s, refRanges)
+  let i = n + 1
+  let j = 0
+  let nest = 0
+  let insideTagsIsClose = 1
+  const inlinesLength = inlines.length
+  while (i < inlinesLength) {
+    if (inlines[i].type !== '') { i++; continue }
+    if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
+    if (inlines[i].type === 'html_inline') {
+      inlines[i].check = true
+      insideTagsIsClose = checkInsideTags(inlines, i, memo)
+      if (insideTagsIsClose === -1) return [n, nest]
+      if (insideTagsIsClose === 0) { i++; continue }
+    }
+
+    const closeRange = findRefRangeIndex(inlines[i].s, refRanges)
+    if (strongOpenRange !== closeRange) { i++; continue }
+
+    nest = checkNest(inlines, marks, n, i, nestTracker)
+    if (nest === -1) return [n, nest]
+
+    if (inlines[i].len === 1 && inlines[n].len > 2) {
+      pushMark(marks, {
+        nest: nest,
+        s: inlines[n].ep,
+        e: inlines[n].ep,
+        len: 1,
+        oLen: inlines[n].len - 1,
+        type: 'em_open'
+      })
+      pushMark(marks, {
+        nest: nest,
+        s: inlines[i].sp,
+        e: inlines[i].ep,
+        len: 1,
+        oLen: inlines[i].len - 1,
+        type: 'em_close'
+      })
+      inlines[n].len -= 1
+      inlines[n].ep -= 1
+      inlines[i].len -= 1
+      if (inlines[i].len > 0) inlines[i].sp += 1
+      if (insideTagsIsClose === 1) {
+        const [newN, newNest] = setEm(state, inlines, marks, n, memo, opt, null, nestTracker, refRanges)
+        n = newN
+        nest = newNest
+      }
+    }
+    let strongNum = Math.trunc(Math.min(inlines[n].len, inlines[i].len) / 2)
+
+    if (inlines[i].len > 1) {
+      if (hasPunctuationOrNonJapanese(state, inlines, n, i, opt, refRanges)) {
+        if (memo.inlineMarkEnd) {
+          marks.push(...createMarks(state, inlines, i, inlinesLength - 1, memo, opt, refRanges))
+          if (inlines[i].len === 0) { i++; continue }
+        } else {
+          return [n, nest]
+        }
+      }
+
+      j = 0
+      while (j < strongNum) {
+        pushMark(marks, {
+          nest: nest + strongNum - 1 - j,
+          s: inlines[n].ep - 1,
+          e: inlines[n].ep,
+          len: 2,
+          oLen: inlines[n].len - 2,
+          type: 'strong_open'
+        })
+        inlines[n].ep -= 2
+        inlines[n].len -= 2
+        pushMark(marks, {
+          nest: nest + strongNum - 1 - j,
+          s: inlines[i].sp,
+          e: inlines[i].sp + 1,
+          len: 2,
+          oLen: inlines[i].len - 2,
+          type: 'strong_close'
+        })
+        inlines[i].sp += 2
+        inlines[i].len -= 2
+        j++
+      }
+      if (inlines[n].len === 0) return [n, nest]
+    }
+
+    if (inlines[n].len === 1 && inlines[i].len > 0) {
+      nest++
+      const [newN, newNest] = setEm(state, inlines, marks, n, memo, opt, nest, nestTracker, refRanges)
+      n = newN
+      nest = newNest
+    }
+
+    i++
+  }
+
+  if (n == 0 && memo.inlineMarkEnd) {
+    marks.push(...createMarks(state, inlines, n + 1, inlinesLength - 1, memo, opt, refRanges))
+  }
+  return [n, nest]
+}
+
+const checkInsideTags = (inlines, i, memo) => {
+  if (inlines[i].tag === undefined) return 0
+  const tagName = inlines[i].tag[0].toLowerCase()
+  if (memo.htmlTags[tagName] === undefined) {
+    memo.htmlTags[tagName] = 0
+  }
+  if (inlines[i].tag[1] === 'open') {
+    memo.htmlTags[tagName] += 1
+  }
+  if (inlines[i].tag[1] === 'close') {
+    memo.htmlTags[tagName] -= 1
+  }
+  if (memo.htmlTags[tagName] < 0) {
+    return -1
+  }
+  
+  // Direct check instead of Object.values().every()
+  for (const count of Object.values(memo.htmlTags)) {
+    if (count !== 0) return 0
+  }
+  return 1
+}
+
+// Check if character is ASCII punctuation or space
+// Covers: !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~ and space
+const isPunctuation = (ch) => {
+  if (!ch) return false
+  const code = ch.charCodeAt(0)
+  // ASCII punctuation: !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~
+  return (code >= 33 && code <= 47) || (code >= 58 && code <= 64) || 
+         (code >= 91 && code <= 96) || (code >= 123 && code <= 126) || code === 32
+}
+
+// Check if character is Japanese (hiragana, katakana, kanji, punctuation, symbols, format chars, emoji)
+// Uses fast Unicode range checks for common cases, falls back to REG_JAPANESE for complex Unicode
+const isJapanese = (ch) => {
+  if (!ch) return false
+  const code = ch.charCodeAt(0)
+  // Fast ASCII check first
+  if (code < 128) return false
+  // Hiragana: U+3040-U+309F, Katakana: U+30A0-U+30FF, Kanji: U+4E00-U+9FAF
+  return (code >= 0x3040 && code <= 0x309F) || 
+         (code >= 0x30A0 && code <= 0x30FF) || 
+         (code >= 0x4E00 && code <= 0x9FAF) ||
+         // Fallback to regex for complex Unicode cases
+         REG_JAPANESE.test(ch)
+}
+
+// Check if character is English (letters, numbers) or other non-Japanese characters
+// Uses REG_JAPANESE and REG_PUNCTUATION to exclude Japanese and punctuation characters
+const isEnglish = (ch) => {
+  if (!ch) return false
+  const code = ch.charCodeAt(0)
+  if ((code >= 65 && code <= 90) || (code >= 97 && code <= 122) || (code >= 48 && code <= 57)) {
+    return true
+  }
+  if (code < 128) {
+    return code === CHAR_SPACE || (code > 126)
+  }
+  return !REG_JAPANESE.test(ch) && !REG_PUNCTUATION.test(ch)
+}
+
+const checkMixedLanguagePattern = (state, inlines, n, i, opt) => {
+  const src = state.src
+  const openPrevChar = src[inlines[n].s - 1] || ''
+  const closeNextChar = src[inlines[i].e + 1] || ''
+  
+  const isEnglishPrefix = isEnglish(openPrevChar)
+  const isEnglishSuffix = isEnglish(closeNextChar)
+  if (!isEnglishPrefix && !isEnglishSuffix) {
+    return { hasEnglishContext: false, hasMarkdownOrHtml: false, shouldBlock: false }
+  }
+  
+  const contentBetween = src.slice(inlines[n].e + 1, inlines[i].s)
+  const hasMarkdownOrHtml = REG_MARKDOWN_HTML.test(contentBetween)
+  
+  return {
+    hasEnglishContext: true,
+    hasMarkdownOrHtml,
+    shouldBlock: hasMarkdownOrHtml
+  }
+}
+
+const hasPunctuationOrNonJapanese = (state, inlines, n, i, opt, refRanges) => {
   const src = state.src
-  const srcLen = max
-  const htmlEnabled = state.md.options.html
-  let n = start
-  let inlines = []
-  let noMark = ''
-  let textStart = n
-  
-  // Infinite loop prevention
-  const maxIterations = srcLen * 2 // Safe upper bound
-  let iterations = 0
-  
-  while (n < srcLen) {
-    // Prevent infinite loops
-    iterations++
-    if (iterations > maxIterations) {
-      // Add remaining text as-is and exit safely
-      if (textStart < srcLen) {
-        pushInlines(inlines, textStart, srcLen - 1, srcLen - textStart, 'text')
-      }
-      break
-    }
-    
-    const currentChar = src.charCodeAt(n)
-    
-    // Unified escape check
-    let isEscaped = false
-    if (currentChar === CHAR_ASTERISK || currentChar === CHAR_BACKTICK || 
-        (opt.dollarMath && currentChar === CHAR_DOLLAR) || 
-        (htmlEnabled && currentChar === CHAR_LT)) {
-      isEscaped = hasBackslash(state, n)
-    }
-
-    // Asterisk handling
-    if (currentChar === CHAR_ASTERISK) {
-      if (!isEscaped) {
-        noMark = processTextSegment(inlines, textStart, n, noMark)
-        if (n === srcLen - 1) {
-          pushInlines(inlines, n, n, 1, '')
-          break
-        }
-        let i = n + 1
-        while (i < srcLen && src.charCodeAt(i) === CHAR_ASTERISK) {
-          i++
-        }
-        if (i === srcLen) {
-          pushInlines(inlines, n, i - 1, i - n, '')
-        } else {
-          pushInlines(inlines, n, i - 1, i - n, '')
-          textStart = i
-        }
-        n = i
-        continue
-      }
-    }
-
-    // Inline code (backticks)
-    if (currentChar === CHAR_BACKTICK) {
-      if (!isEscaped) {
-        const result = processSymbolPair(state, n, srcLen, CHAR_BACKTICK, noMark, textStart, 
-          (start, end, len, type) => pushInlines(inlines, start, end, len, type))
-        if (result.shouldBreak) break
-        if (result.shouldContinue) {
-          n = result.newN
-          noMark = result.newNoMark
-          continue
-        }
-        noMark = result.newNoMark
-      }
-    }
-
-    // Inline math ($...$)
-    if (opt.dollarMath && currentChar === CHAR_DOLLAR) {
-      if (!isEscaped) {
-        const result = processSymbolPair(state, n, srcLen, CHAR_DOLLAR, noMark, textStart, 
-          (start, end, len, type) => pushInlines(inlines, start, end, len, type))
-        if (result.shouldBreak) break
-        if (result.shouldContinue) {
-          n = result.newN
-          noMark = result.newNoMark
-          continue
-        }
-        noMark = result.newNoMark
-      }
-    }
-
-    // HTML tags
-    if (htmlEnabled && currentChar === CHAR_LT) {
-      if (!isEscaped) {
-        let foundClosingTag = false
-        for (let i = n + 1; i < srcLen; i++) {
-          if (src.charCodeAt(i) === CHAR_GT && !hasBackslash(state, i)) {
-            noMark = processTextSegment(inlines, textStart, n, noMark)
-            let tag = src.slice(n + 1, i)
-            let tagType
-            if (tag.charCodeAt(0) === CHAR_SLASH) {
-              tag = tag.slice(1)
-              tagType = 'close'
-            } else {
-              tagType = 'open'
-            }
-            pushInlines(inlines, n, i, i - n + 1, 'html_inline', tag, tagType)
-            textStart = i + 1
-            n = i + 1
-            foundClosingTag = true
-            break
-          }
-        }
-        if (foundClosingTag) {
-          continue
-        }
-        // If no closing tag found, treat as regular character to prevent infinite loops
-      }
+  const openPrevChar = src[inlines[n].s - 1] || ''
+  const openNextChar = src[inlines[n].e + 1]  || ''
+  let checkOpenNextChar = isPunctuation(openNextChar)
+  if (checkOpenNextChar && (openNextChar === '[' || openNextChar === ']')) {
+    const openNextRange = findRefRangeIndex(inlines[n].e + 1, refRanges)
+    if (openNextRange !== -1) {
+      checkOpenNextChar = false
     }
-
-    // Regular character
-    noMark += src[n]
-    if (n === srcLen - 1) {
-      pushInlines(inlines, textStart, n, n - textStart + 1, 'text')
-      break
+  }
+  const closePrevChar = src[inlines[i].s - 1] || ''
+  let checkClosePrevChar = isPunctuation(closePrevChar)
+  if (checkClosePrevChar && (closePrevChar === '[' || closePrevChar === ']')) {
+    const closePrevRange = findRefRangeIndex(inlines[i].s - 1, refRanges)
+    if (closePrevRange !== -1) {
+      checkClosePrevChar = false
     }
-    n++
   }
-  return inlines
-}
+  const closeNextChar = src[inlines[i].e + 1] || ''
+  const checkCloseNextChar = (isPunctuation(closeNextChar) || i === inlines.length - 1)
+
+  if (opt.disallowMixed === false) {
+    if (isEnglish(openPrevChar) || isEnglish(closeNextChar)) {
+      const contentBetween = src.slice(inlines[n].e + 1, inlines[i].s)
+      if (REG_MARKDOWN_HTML.test(contentBetween)) {
+        return false
+      }
+    }
+  }
+
+  const result = (checkOpenNextChar || checkClosePrevChar) && !checkCloseNextChar && !(isJapanese(openPrevChar) || isJapanese(closeNextChar))
+  return result
+}
+
+const setEm = (state, inlines, marks, n, memo, opt, sNest, nestTracker, refRanges) => {
+  const emOpenRange = findRefRangeIndex(inlines[n].s, refRanges)
+  if (opt.disallowMixed === true && !sNest) {
+    let i = n + 1
+    const inlinesLength = inlines.length
+    while (i < inlinesLength) {
+      if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
+      if (inlines[i].type !== '') { i++; continue }
+      
+      if (inlines[i].len > 0) {
+        const mixedCheck = checkMixedLanguagePattern(state, inlines, n, i, opt)
+        if (mixedCheck.shouldBlock) {
+          return [n, 0]
+        }
+        break
+      }
+      i++
+    }
+  }
+  
+  let i = n + 1
+  let nest = 0
+  let strongPNum = 0
+  let insideTagsIsClose = 1
+  const inlinesLength = inlines.length
+  while (i < inlinesLength) {
+    if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
+    if (!sNest && inlines[i].type === 'html_inline') {
+      inlines.check = true
+      insideTagsIsClose = checkInsideTags(inlines, i, memo)
+      if (insideTagsIsClose === -1) return [n, nest]
+      if (insideTagsIsClose === 0) { i++; continue }
+    }
+    if (inlines[i].type !== '') { i++; continue }
+
+    const closeRange = findRefRangeIndex(inlines[i].s, refRanges)
+    if (emOpenRange !== closeRange) {
+      i++
+      continue
+    }
+
+    const emNum = Math.min(inlines[n].len, inlines[i].len)
+
+    if (!sNest && emNum !== 1) return [n, sNest, memo]
+
+    const hasMarkersAtStartAndEnd = (i) => {
+      let flag =  memo.inlineMarkStart
+      if (!flag) return false
+      inlinesLength - 1 === i ? flag = true : flag = false
+      if (!flag) return false
+      inlines[i].len > 1 ? flag = true : flag = false
+      return flag
+    }
+    if (!sNest && inlines[i].len === 2 && !hasMarkersAtStartAndEnd(i)) {
+      strongPNum++
+      i++
+      continue
+    }
+
+    if (sNest) {
+      nest  = sNest - 1
+    } else {
+      nest = checkNest(inlines, marks, n, i, nestTracker)
+    }
+    if (nest === -1) return [n, nest]
+
+    if (emNum === 1) {
+      if (hasPunctuationOrNonJapanese(state, inlines, n, i, opt, refRanges)) {
+        if (memo.inlineMarkEnd) {
+          marks.push(...createMarks(state, inlines, i, inlinesLength - 1, memo, opt, refRanges))
+
+          if (inlines[i].len === 0) { i++; continue }
+        } else {
+          return [n, nest]
+        }
+      }
+      if (inlines[i].len < 1) {
+        i++; continue;
+      }
+
+      pushMark(marks, {
+        nest: nest,
+        s: inlines[n].ep,
+        e: inlines[n].ep,
+        len: 1,
+        oLen: inlines[n].len - 1,
+        type: 'em_open'
+      })
+      inlines[n].ep -= 1
+      inlines[n].len -= 1
+
+      if (strongPNum % 2 === 0 || inlines[i].len < 2) {
+        pushMark(marks, {
+          nest: nest,
+          s: inlines[i].sp,
+          e: inlines[i].sp,
+          len: 1,
+          oLen: inlines[i].len - 1,
+          type: 'em_close'
+        })
+        inlines[i].sp += 1
+      } else {
+        pushMark(marks, {
+          nest: nest,
+          s: inlines[i].ep,
+          e: inlines[i].ep,
+          len: 1,
+          oLen: inlines[i].len - 1,
+          type: 'em_close'
+        })
+        inlines[i].sp = inlines[i].ep - 1
+        inlines[i].ep -= 1
+      }
+      inlines[i].len -= 1
+      if (inlines[n].len === 0) return [n, nest]
+    }
+
+    i++
+  }
+  return [n, nest]
+}
+
+const setText = (inlines, marks, n, nest) => {
+  pushMark(marks, {
+    nest: nest,
+    s: inlines[n].sp,
+    e: inlines[n].ep,
+    len: inlines[n].len,
+    oLen: -1,
+    type: 'text'
+  })
+  inlines[n].len = 0
+}
+
+// Nest state management
+const createNestTracker = () => {
+  return {
+    strongNest: 0,
+    emNest: 0,
+    markIndex: 0
+  }
+}
+
+const updateNestTracker = (tracker, marks, targetPos) => {
+  while (tracker.markIndex < marks.length && marks[tracker.markIndex].s <= targetPos) {
+    const mark = marks[tracker.markIndex]
+    if (mark.type === 'strong_open') tracker.strongNest++
+    else if (mark.type === 'strong_close') tracker.strongNest--
+    else if (mark.type === 'em_open') tracker.emNest++
+    else if (mark.type === 'em_close') tracker.emNest--
+    tracker.markIndex++
+  }
+}
+
+const checkNest = (inlines, marks, n, i, nestTracker) => {
+  if (marks.length === 0) return 1
+  // Update nest state up to current position
+  updateNestTracker(nestTracker, marks, inlines[n].s)
+
+  const parentNest = nestTracker.strongNest + nestTracker.emNest
+  // Check if there's a conflicting close mark before the end position
+  let parentCloseN = nestTracker.markIndex
+  while (parentCloseN < marks.length) {
+    if (marks[parentCloseN].nest === parentNest) break
+    parentCloseN++
+  }
+  if (parentCloseN < marks.length && marks[parentCloseN].s < inlines[i].s) {
+    return -1
+  }
+  return parentNest + 1
+}
+
+const createMarks = (state, inlines, start, end, memo, opt, refRanges) => {
+  let marks = []
+  let n = start
+  const nestTracker = createNestTracker()
+  
+  while (n < end) {
+    if (inlines[n].type !== '') { n++; continue }
+    let nest = 0
+    
+    if (inlines[n].len > 1) {
+      const [newN, newNest] = setStrong(state, inlines, marks, n, memo, opt, nestTracker, refRanges)
+      n = newN
+      nest = newNest
+    }
+    if (inlines[n].len !== 0) {
+      const [newN2, newNest2] = setEm(state, inlines, marks, n, memo, opt, null, nestTracker, refRanges)
+      n = newN2
+      nest = newNest2
+    }
+    if (inlines[n].len !== 0) {
+      setText(inlines, marks, n, nest)
+    }
+    n++
+  }
+  return marks
+}
+
+const mergeInlinesAndMarks = (inlines, marks) => {
+  // marks array is already sorted, skip sorting
+  const merged = []
+  let markIndex = 0
+  for (const token of inlines) {
+    if (token.type === '') {
+      while (markIndex < marks.length && marks[markIndex].s >= token.s && marks[markIndex].e <= token.e) {
+        merged.push(marks[markIndex])
+        markIndex++
+      }
+    } else {
+      merged.push(token)
+    }
+  }
+  while (markIndex < marks.length) {
+    merged.push(marks[markIndex++])
+  }
+  return merged
+}
+
+const isWhitespaceToken = (token) => token && token.type === 'text' && token.content.trim() === ''
+
+const strongJa = (state, silent, opt) => {
+  if (silent) return false
+  const start = state.pos
+  let max = state.posMax
+  const src = state.src
+  let attributesSrc
+  if (start > max) return false
+  if (src.charCodeAt(start) !== CHAR_ASTERISK) return false
+  if (hasBackslash(state, start)) return false
+
+  if (opt.mditAttrs) {
+    attributesSrc = src.match(/((\n)? *){([^{}\n!@#%^&*()]+?)} *$/)
+    if (attributesSrc && attributesSrc[3] !== '.') {
+      max = src.slice(0, attributesSrc.index).length
+      if (attributesSrc[2] === '\n') {
+        max = src.slice(0, attributesSrc.index - 1).length
+      }
+      if(hasBackslash(state, attributesSrc.index) && attributesSrc[2] === '' && attributesSrc[1].length === 0) {
+        max = state.posMax
+      }
+    } else {
+      let endCurlyKet = src.match(/(\n *){([^{}\n!@#%^&*()]*?)}.*(} *?)$/)
+      if (endCurlyKet) {
+        max -= endCurlyKet[3].length
+      }
+    }
+  }
+
+  const refRanges = computeReferenceRanges(state, start, max)
+  if (refRanges.length > 0) {
+    state.__strongJaHasCollapsedRefs = true
+  }
+  let inlines = createInlines(state, start, max, opt)
+
+  const memo = {
+    html: state.md.options.html,
+    htmlTags: {},
+    inlineMarkStart: src.charCodeAt(0) === CHAR_ASTERISK,
+    inlineMarkEnd: src.charCodeAt(max - 1) === CHAR_ASTERISK,
+  }
+
+  let marks = createMarks(state, inlines, 0, inlines.length, memo, opt, refRanges)
+
+  inlines = mergeInlinesAndMarks(inlines, marks)
+
+  setToken(state, inlines, opt)
 
-const pushMark = (marks, opts) => {
-  // Maintain sorted order during insertion
-  const newMark = {
-    nest: opts.nest,
-    s: opts.s,
-    e: opts.e,
-    len: opts.len,
-    oLen: opts.oLen,
-    type: opts.type
+  if (inlineHasCollapsedRef(state) && !state.__strongJaCollapsedRefRegistered) {
+    registerCollapsedRefTarget(state)
+    state.__strongJaCollapsedRefRegistered = true
   }
-  // Binary search for insertion point to maintain sorted order
-  let left = 0
-  let right = marks.length
-  while (left < right) {
-    const mid = Math.floor((left + right) / 2)
-    if (marks[mid].s <= newMark.s) {
-      left = mid + 1
-    } else {
-      right = mid
+
+  if (opt.mditAttrs && max !== state.posMax) {
+    if (!attributesSrc) {
+      state.pos = max
+      return true
     }
+    state.pos = attributesSrc[1].length > 1 ? max + attributesSrc[1].length : max
+    return true
   }
-  
-  marks.splice(left, 0, newMark)
+  state.pos = max
+  return true
+}
+
+// Collapsed reference helpers
+const buildReferenceLabel = (tokens) => {
+  let label = ''
+  for (const token of tokens) {
+    if (token.type === 'text' || token.type === 'code_inline') {
+      label += token.content
+    } else if (token.type === 'softbreak' || token.type === 'hardbreak') {
+      label += ' '
+    } else if (token.type && token.type.endsWith('_open') && token.markup) {
+      label += token.markup
+    } else if (token.type && token.type.endsWith('_close') && token.markup) {
+      label += token.markup
+    }
+  }
+  return label
+}
+
+const cleanLabelText = (label) => {
+  return label.replace(/^[*_]+/, '').replace(/[*_]+$/, '')
 }
 
-const setStrong = (state, inlines, marks, n, memo, opt, nestTracker) => {
-  if (opt.disallowMixed === true) {
-    let i = n + 1
-    const inlinesLength = inlines.length
-    while (i < inlinesLength) {
-      if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
-      if (inlines[i].type !== '') { i++; continue }
-      
-      if (inlines[i].len > 1) {
-        const mixedCheck = checkMixedLanguagePattern(state, inlines, n, i, opt)
-        if (mixedCheck.shouldBlock) {
-          return [n, 0]
-        }
-        break
-      }
-      i++
-    }
-  }
-  
-  let i = n + 1
-  let j = 0
-  let nest = 0
-  let insideTagsIsClose = 1
-  const inlinesLength = inlines.length
-  while (i < inlinesLength) {
-    if (inlines[i].type !== '') { i++; continue }
-    if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
-    if (inlines[i].type === 'html_inline') {
-      inlines[i].check = true
-      insideTagsIsClose = checkInsideTags(inlines, i, memo)
-      if (insideTagsIsClose === -1) return [n, nest]
-      if (insideTagsIsClose === 0) { i++; continue }
-    }
-
-    nest = checkNest(inlines, marks, n, i, nestTracker)
-    if (nest === -1) return [n, nest]
+const normalizeReferenceCandidate = (state, text, { useClean = false } = {}) => {
+  const source = useClean ? cleanLabelText(text) : text.replace(/\s+/g, ' ').trim()
+  return normalizeRefKey(state, source)
+}
 
-    if (inlines[i].len === 1 && inlines[n].len > 2) {
-      pushMark(marks, {
-        nest: nest,
-        s: inlines[n].ep,
-        e: inlines[n].ep,
-        len: 1,
-        oLen: inlines[n].len - 1,
-        type: 'em_open'
-      })
-      pushMark(marks, {
-        nest: nest,
-        s: inlines[i].sp,
-        e: inlines[i].ep,
-        len: 1,
-        oLen: inlines[i].len - 1,
-        type: 'em_close'
-      })
-      inlines[n].len -= 1
-      inlines[n].ep -= 1
-      inlines[i].len -= 1
-      if (inlines[i].len > 0) inlines[i].sp += 1
-      if (insideTagsIsClose === 1) {
-        const [newN, newNest] = setEm(state, inlines, marks, n, memo, opt, null, nestTracker)
-        n = newN
-        nest = newNest
+const normalizeRefKey = (state, label) => {
+  const normalize = state.md && state.md.utils && state.md.utils.normalizeReference
+    ? state.md.utils.normalizeReference
+    : (str) => str.trim().replace(/\s+/g, ' ').toUpperCase()
+  return normalize(label)
+}
+
+const adjustTokenLevels = (tokens, startIdx, endIdx, delta) => {
+  for (let i = startIdx; i < endIdx; i++) {
+    if (tokens[i]) tokens[i].level += delta
+  }
+}
+
+const cloneTextToken = (source, content) => {
+  const newToken = new Token('text', '', 0)
+  newToken.content = content
+  newToken.level = source.level
+  newToken.markup = source.markup
+  newToken.info = source.info
+  newToken.meta = source.meta ? {...source.meta} : null
+  newToken.block = source.block
+  newToken.hidden = source.hidden
+  return newToken
+}
+
+// Split only text tokens that actually contain bracket characters
+const splitBracketToken = (tokens, index) => {
+  const token = tokens[index]
+  if (!token || token.type !== 'text') return false
+  const content = token.content
+  if (!content || (content.indexOf('[') === -1 && content.indexOf(']') === -1)) {
+    return false
+  }
+  const segments = []
+  let buffer = ''
+  let pos = 0
+  while (pos < content.length) {
+    if (content.startsWith('[]', pos)) {
+      if (buffer) {
+        segments.push(buffer)
+        buffer = ''
       }
+      segments.push('[]')
+      pos += 2
+      continue
     }
-    let strongNum = Math.trunc(Math.min(inlines[n].len, inlines[i].len) / 2)
-
-    if (inlines[i].len > 1) {
-      if (hasPunctuationOrNonJapanese(state, inlines, n, i, opt)) {
-        if (memo.inlineMarkEnd) {
-          marks.push(...createMarks(state, inlines, i, inlinesLength - 1, memo, opt))
-          if (inlines[i].len === 0) { i++; continue }
-        } else {
-          return [n, nest]
-        }
+    const ch = content[pos]
+    if (ch === '[' || ch === ']') {
+      if (buffer) {
+        segments.push(buffer)
+        buffer = ''
       }
-
-      j = 0
-      while (j < strongNum) {
-        pushMark(marks, {
-          nest: nest + strongNum - 1 - j,
-          s: inlines[n].ep - 1,
-          e: inlines[n].ep,
-          len: 2,
-          oLen: inlines[n].len - 2,
-          type: 'strong_open'
-        })
-        inlines[n].ep -= 2
-        inlines[n].len -= 2
-        pushMark(marks, {
-          nest: nest + strongNum - 1 - j,
-          s: inlines[i].sp,
-          e: inlines[i].sp + 1,
-          len: 2,
-          oLen: inlines[i].len - 2,
-          type: 'strong_close'
-        })
-        inlines[i].sp += 2
-        inlines[i].len -= 2
-        j++
-      }
-      if (inlines[n].len === 0) return [n, nest]
-    }
-
-    if (inlines[n].len === 1 && inlines[i].len > 0) {
-      nest++
-      const [newN, newNest] = setEm(state, inlines, marks, n, memo, opt, nest, nestTracker)
-      n = newN
-      nest = newNest
+      segments.push(ch)
+      pos++
+      continue
     }
-
-    i++
-  }
-
-  if (n == 0 && memo.inlineMarkEnd) {
-    marks.push(...createMarks(state, inlines, n + 1, inlinesLength - 1, memo, opt))
-  }
-  return [n, nest]
-}
-
-const checkInsideTags = (inlines, i, memo) => {
-  if (inlines[i].tag === undefined) return 0
-  const tagName = inlines[i].tag[0].toLowerCase()
-  if (memo.htmlTags[tagName] === undefined) {
-    memo.htmlTags[tagName] = 0
+    buffer += ch
+    pos++
   }
-  if (inlines[i].tag[1] === 'open') {
-    memo.htmlTags[tagName] += 1
+  if (buffer) segments.push(buffer)
+  if (segments.length <= 1) return false
+  token.content = segments[0]
+  let insertIdx = index + 1
+  for (let s = 1; s < segments.length; s++) {
+    const newToken = cloneTextToken(token, segments[s])
+    tokens.splice(insertIdx, 0, newToken)
+    insertIdx++
   }
-  if (inlines[i].tag[1] === 'close') {
-    memo.htmlTags[tagName] -= 1
-  }
-  if (memo.htmlTags[tagName] < 0) {
-    return -1
-  }
-  
-  // Direct check instead of Object.values().every()
-  for (const count of Object.values(memo.htmlTags)) {
-    if (count !== 0) return 0
-  }
-  return 1
-}
-
-// Check if character is ASCII punctuation or space
-// Covers: !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~ and space
-const isPunctuation = (ch) => {
-  if (!ch) return false
-  const code = ch.charCodeAt(0)
-  // ASCII punctuation: !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~
-  return (code >= 33 && code <= 47) || (code >= 58 && code <= 64) || 
-         (code >= 91 && code <= 96) || (code >= 123 && code <= 126) || code === 32
+  return true
 }
-
-// Check if character is Japanese (hiragana, katakana, kanji, punctuation, symbols, format chars, emoji)
-// Uses fast Unicode range checks for common cases, falls back to REG_JAPANESE for complex Unicode
-const isJapanese = (ch) => {
-  if (!ch) return false
-  const code = ch.charCodeAt(0)
-  // Fast ASCII check first
-  if (code < 128) return false
-  // Hiragana: U+3040-U+309F, Katakana: U+30A0-U+30FF, Kanji: U+4E00-U+9FAF
-  return (code >= 0x3040 && code <= 0x309F) || 
-         (code >= 0x30A0 && code <= 0x30FF) || 
-         (code >= 0x4E00 && code <= 0x9FAF) ||
-         // Fallback to regex for complex Unicode cases
-         REG_JAPANESE.test(ch)
+
+const isBracketToken = (token, bracket) => {
+  return token && token.type === 'text' && token.content === bracket
 }
 
-// Check if character is English (letters, numbers) or other non-Japanese characters
-// Uses REG_JAPANESE and REG_PUNCTUATION to exclude Japanese and punctuation characters
-const isEnglish = (ch) => {
-  if (!ch) return false
-  const code = ch.charCodeAt(0)
-  if ((code >= 65 && code <= 90) || (code >= 97 && code <= 122) || (code >= 48 && code <= 57)) {
-    return true
-  }
-  if (code < 128) {
-    return code === CHAR_SPACE || (code > 126)
-  }
-  return !REG_JAPANESE.test(ch) && !REG_PUNCTUATION.test(ch)
-}
-
-const checkMixedLanguagePattern = (state, inlines, n, i, opt) => {
-  const src = state.src
-  const openPrevChar = src[inlines[n].s - 1] || ''
-  const closeNextChar = src[inlines[i].e + 1] || ''
-  
-  const isEnglishPrefix = isEnglish(openPrevChar)
-  const isEnglishSuffix = isEnglish(closeNextChar)
-  if (!isEnglishPrefix && !isEnglishSuffix) {
-    return { hasEnglishContext: false, hasMarkdownOrHtml: false, shouldBlock: false }
-  }
-  
-  const contentBetween = src.slice(inlines[n].e + 1, inlines[i].s)
-  const hasMarkdownOrHtml = REG_MARKDOWN_HTML.test(contentBetween)
-  
-  return {
-    hasEnglishContext: true,
-    hasMarkdownOrHtml,
-    shouldBlock: hasMarkdownOrHtml
+const findLinkCloseIndex = (tokens, startIdx) => {
+  let depth = 0
+  for (let idx = startIdx; idx < tokens.length; idx++) {
+    const token = tokens[idx]
+    if (token.type === 'link_open') depth++
+    if (token.type === 'link_close') {
+      depth--
+      if (depth === 0) return idx
+    }
   }
+  return -1
 }
 
-const hasPunctuationOrNonJapanese = (state, inlines, n, i, opt) => {
-  const src = state.src
-  const openPrevChar = src[inlines[n].s - 1] || ''
-  const openNextChar = src[inlines[n].e + 1]  || ''
-  const checkOpenNextChar = isPunctuation(openNextChar)
-  const closePrevChar = src[inlines[i].s - 1] || ''
-  const checkClosePrevChar = isPunctuation(closePrevChar)
-  const closeNextChar = src[inlines[i].e + 1] || ''
-  const checkCloseNextChar = (isPunctuation(closeNextChar) || i === inlines.length - 1)
+const convertCollapsedReferenceLinks = (tokens, state) => {
+  const references = state.env && state.env.references
+  if (!references || Object.keys(references).length === 0) return
 
-  if (opt.disallowMixed === false) {
-    if (isEnglish(openPrevChar) || isEnglish(closeNextChar)) {
-      const contentBetween = src.slice(inlines[n].e + 1, inlines[i].s)
-      if (REG_MARKDOWN_HTML.test(contentBetween)) {
-        return false
-      }
+  let i = 0
+  while (i < tokens.length) {
+    if (splitBracketToken(tokens, i)) {
+      continue
     }
-  }
-
-  const result = (checkOpenNextChar || checkClosePrevChar) && !checkCloseNextChar && !(isJapanese(openPrevChar) || isJapanese(closeNextChar))
-  return result
-}
-
-const setEm = (state, inlines, marks, n, memo, opt, sNest, nestTracker) => {
-  if (opt.disallowMixed === true && !sNest) {
-    let i = n + 1
-    const inlinesLength = inlines.length
-    while (i < inlinesLength) {
-      if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
-      if (inlines[i].type !== '') { i++; continue }
-      
-      if (inlines[i].len > 0) {
-        const mixedCheck = checkMixedLanguagePattern(state, inlines, n, i, opt)
-        if (mixedCheck.shouldBlock) {
-          return [n, 0]
-        }
+    if (!isBracketToken(tokens[i], '[')) {
+      i++
+      continue
+    }
+    let closeIdx = i + 1
+    while (closeIdx < tokens.length && !isBracketToken(tokens[closeIdx], ']')) {
+      if (splitBracketToken(tokens, closeIdx)) {
+        continue
+      }
+      if (tokens[closeIdx].type === 'link_open') {
+        closeIdx = -1
         break
       }
+      closeIdx++
+    }
+    if (closeIdx === -1 || closeIdx >= tokens.length) {
+      i++
+      continue
+    }
+
+    if (closeIdx === i + 1) {
       i++
+      continue
     }
-  }
-  
-  let i = n + 1
-  let nest = 0
-  let strongPNum = 0
-  let insideTagsIsClose = 1
-  const inlinesLength = inlines.length
-  while (i < inlinesLength) {
-    if (inlines[i].len === 0 || inlines[i].check) { i++; continue }
-    if (!sNest && inlines[i].type === 'html_inline') {
-      inlines.check = true
-      insideTagsIsClose = checkInsideTags(inlines, i, memo)
-      if (insideTagsIsClose === -1) return [n, nest]
-      if (insideTagsIsClose === 0) { i++; continue }
-    }
-    if (inlines[i].type !== '') { i++; continue }
-
-    const emNum = Math.min(inlines[n].len, inlines[i].len)
 
-    if (!sNest && emNum !== 1) return [n, sNest, memo]
-
-    const hasMarkersAtStartAndEnd = (i) => {
-      let flag =  memo.inlineMarkStart
-      if (!flag) return false
-      inlinesLength - 1 === i ? flag = true : flag = false
-      if (!flag) return false
-      inlines[i].len > 1 ? flag = true : flag = false
-      return flag
+    const labelTokens = tokens.slice(i + 1, closeIdx)
+    const labelText = buildReferenceLabel(labelTokens)
+    const cleanedLabel = cleanLabelText(labelText)
+    const whitespaceStart = closeIdx + 1
+    let refRemoveStart = whitespaceStart
+    while (refRemoveStart < tokens.length && isWhitespaceToken(tokens[refRemoveStart])) {
+      refRemoveStart++
     }
-    if (!sNest && inlines[i].len === 2 && !hasMarkersAtStartAndEnd(i)) {
-      strongPNum++
-      i++
+    if (splitBracketToken(tokens, refRemoveStart)) {
       continue
     }
-
-    if (sNest) {
-      nest  = sNest - 1
-    } else {
-      nest = checkNest(inlines, marks, n, i, nestTracker)
+    const whitespaceCount = refRemoveStart - whitespaceStart
+    let refKey = null
+    let refRemoveCount = 0
+    let existingLinkOpen = null
+    let existingLinkClose = null
+    const nextToken = tokens[refRemoveStart]
+    if (process.env.DEBUG_COLLAPSED === 'wide') {
+      const debugSlice = tokens.slice(i, Math.min(tokens.length, refRemoveStart + 3)).map((t) => `${t.type}:${t.content || ''}`)
+      console.log('debug collapsed ctx:', debugSlice)
+      console.log('next token info:', nextToken && nextToken.type, nextToken && JSON.stringify(nextToken.content))
     }
-    if (nest === -1) return [n, nest]
-
-    if (emNum === 1) {
-      if (hasPunctuationOrNonJapanese(state, inlines, n, i, opt)) {
-        if (memo.inlineMarkEnd) {
-          marks.push(...createMarks(state, inlines, i, inlinesLength - 1, memo, opt))
 
-          if (inlines[i].len === 0) { i++; continue }
-        } else {
-          return [n, nest]
-        }
+    if (isBracketToken(nextToken, '[]')) {
+      refKey = normalizeReferenceCandidate(state, cleanedLabel)
+      refRemoveCount = 1
+    } else if (isBracketToken(nextToken, '[')) {
+      let refCloseIdx = refRemoveStart + 1
+      while (refCloseIdx < tokens.length && !isBracketToken(tokens[refCloseIdx], ']')) {
+        refCloseIdx++
       }
-      if (inlines[i].len < 1) {
-        i++; continue;
+      if (refCloseIdx >= tokens.length) {
+        i++
+        continue
       }
-
-      pushMark(marks, {
-        nest: nest,
-        s: inlines[n].ep,
-        e: inlines[n].ep,
-        len: 1,
-        oLen: inlines[n].len - 1,
-        type: 'em_open'
-      })
-      inlines[n].ep -= 1
-      inlines[n].len -= 1
-
-      if (strongPNum % 2 === 0 || inlines[i].len < 2) {
-        pushMark(marks, {
-          nest: nest,
-          s: inlines[i].sp,
-          e: inlines[i].sp,
-          len: 1,
-          oLen: inlines[i].len - 1,
-          type: 'em_close'
-        })
-        inlines[i].sp += 1
+      const refTokens = tokens.slice(refRemoveStart + 1, refCloseIdx)
+      if (refTokens.length === 0) {
+        refKey = normalizeReferenceCandidate(state, cleanedLabel)
       } else {
-        pushMark(marks, {
-          nest: nest,
-          s: inlines[i].ep,
-          e: inlines[i].ep,
-          len: 1,
-          oLen: inlines[i].len - 1,
-          type: 'em_close'
-        })
-        inlines[i].sp = inlines[i].ep - 1
-        inlines[i].ep -= 1
+        const refLabelText = buildReferenceLabel(refTokens)
+        refKey = normalizeReferenceCandidate(state, refLabelText)
       }
-      inlines[i].len -= 1
-      if (inlines[n].len === 0) return [n, nest]
-    }
-
-    i++
-  }
-  return [n, nest]
-}
-
-const setText = (inlines, marks, n, nest) => {
-  pushMark(marks, {
-    nest: nest,
-    s: inlines[n].sp,
-    e: inlines[n].ep,
-    len: inlines[n].len,
-    oLen: -1,
-    type: 'text'
-  })
-  inlines[n].len = 0
-}
-
-// Nest state management
-const createNestTracker = () => {
-  return {
-    strongNest: 0,
-    emNest: 0,
-    markIndex: 0
-  }
-}
-
-const updateNestTracker = (tracker, marks, targetPos) => {
-  while (tracker.markIndex < marks.length && marks[tracker.markIndex].s <= targetPos) {
-    const mark = marks[tracker.markIndex]
-    if (mark.type === 'strong_open') tracker.strongNest++
-    else if (mark.type === 'strong_close') tracker.strongNest--
-    else if (mark.type === 'em_open') tracker.emNest++
-    else if (mark.type === 'em_close') tracker.emNest--
-    tracker.markIndex++
-  }
-}
-
-const checkNest = (inlines, marks, n, i, nestTracker) => {
-  if (marks.length === 0) return 1
-  // Update nest state up to current position
-  updateNestTracker(nestTracker, marks, inlines[n].s)
-
-  const parentNest = nestTracker.strongNest + nestTracker.emNest
-  // Check if there's a conflicting close mark before the end position
-  let parentCloseN = nestTracker.markIndex
-  while (parentCloseN < marks.length) {
-    if (marks[parentCloseN].nest === parentNest) break
-    parentCloseN++
-  }
-  if (parentCloseN < marks.length && marks[parentCloseN].s < inlines[i].s) {
-    return -1
-  }
-  return parentNest + 1
-}
-
-const createMarks = (state, inlines, start, end, memo, opt) => {
-  let marks = []
-  let n = start
-  const nestTracker = createNestTracker()
-  
-  while (n < end) {
-    if (inlines[n].type !== '') { n++; continue }
-    let nest = 0
-    
-    if (inlines[n].len > 1) {
-      const [newN, newNest] = setStrong(state, inlines, marks, n, memo, opt, nestTracker)
-      n = newN
-      nest = newNest
-    }
-    if (inlines[n].len !== 0) {
-      const [newN2, newNest2] = setEm(state, inlines, marks, n, memo, opt, null, nestTracker)
-      n = newN2
-      nest = newNest2
-    }
-    if (inlines[n].len !== 0) {
-      setText(inlines, marks, n, nest)
-    }
-    n++
-  }
-  return marks
-}
-
-const mergeInlinesAndMarks = (inlines, marks) => {
-  // marks array is already sorted, skip sorting
-  const merged = []
-  let markIndex = 0
-  for (const token of inlines) {
-    if (token.type === '') {
-      while (markIndex < marks.length && marks[markIndex].s >= token.s && marks[markIndex].e <= token.e) {
-        merged.push(marks[markIndex])
-        markIndex++
+      refRemoveCount = refCloseIdx - refRemoveStart + 1
+    } else if (nextToken && nextToken.type === 'link_open') {
+      const linkCloseIdx = findLinkCloseIndex(tokens, refRemoveStart)
+      if (linkCloseIdx === -1) {
+        i++
+        continue
       }
+      existingLinkOpen = tokens[refRemoveStart]
+      existingLinkClose = tokens[linkCloseIdx]
+      refRemoveCount = linkCloseIdx - refRemoveStart + 1
     } else {
-      merged.push(token)
+      i++
+      continue
     }
-  }
-  while (markIndex < marks.length) {
-    merged.push(marks[markIndex++])
-  }
-  return merged
-}
-
-const strongJa = (state, silent, opt) => {
-  if (silent) return false
-  const start = state.pos
-  let max = state.posMax
-  const src = state.src
-  let attributesSrc
-  if (start > max) return false
-  if (src.charCodeAt(start) !== CHAR_ASTERISK) return false
-  if (hasBackslash(state, start)) return false
-
-  if (opt.mditAttrs) {
-    attributesSrc = src.match(/((\n)? *){([^{}\n!@#%^&*()]+?)} *$/)
-    if (attributesSrc && attributesSrc[3] !== '.') {
-      max = src.slice(0, attributesSrc.index).length
-      if (attributesSrc[2] === '\n') {
-        max = src.slice(0, attributesSrc.index - 1).length
+    if (process.env.DEBUG_COLLAPSED === '1') {
+      const context = tokens.slice(Math.max(0, i - 2), Math.min(tokens.length, closeIdx + 3))
+      console.log('[collapsed-ref] context:',
+        context.map((t) => t.type + ':' + (t.content || '')))
+    }
+    let linkOpenToken = null
+    let linkCloseToken = null
+    if (existingLinkOpen && existingLinkClose) {
+      if (whitespaceCount > 0) {
+        tokens.splice(whitespaceStart, whitespaceCount)
+        refRemoveStart -= whitespaceCount
       }
-      if(hasBackslash(state, attributesSrc.index) && attributesSrc[2] === '' && attributesSrc[1].length === 0) {
-        max = state.posMax
+      if (refRemoveCount > 0) {
+        tokens.splice(refRemoveStart, refRemoveCount)
       }
+      linkOpenToken = existingLinkOpen
+      linkCloseToken = existingLinkClose
     } else {
-      let endCurlyKet = src.match(/(\n *){([^{}\n!@#%^&*()]*?)}.*(} *?)$/)
-      if (endCurlyKet) {
-        max -= endCurlyKet[3].length
+      if (!refKey) {
+        i++
+        continue
       }
+      const ref = references[refKey]
+      if (!ref) {
+        i++
+        continue
+      }
+      if (whitespaceCount > 0) {
+        tokens.splice(whitespaceStart, whitespaceCount)
+        refRemoveStart -= whitespaceCount
+      }
+      if (refRemoveCount > 0) {
+        tokens.splice(refRemoveStart, refRemoveCount)
+      }
+      linkOpenToken = new Token('link_open', 'a', 1)
+      linkOpenToken.attrs = [['href', ref.href]]
+      if (ref.title) linkOpenToken.attrPush(['title', ref.title])
+      linkOpenToken.markup = '[]'
+      linkOpenToken.info = 'auto'
+      linkCloseToken = new Token('link_close', 'a', -1)
+      linkCloseToken.markup = '[]'
+      linkCloseToken.info = 'auto'
     }
-  }
-
-  let inlines = createInlines(state, start, max, opt)
-
-  const memo = {
-    html: state.md.options.html,
-    htmlTags: {},
-    inlineMarkStart: src.charCodeAt(0) === CHAR_ASTERISK,
-    inlineMarkEnd: src.charCodeAt(max - 1) === CHAR_ASTERISK,
-  }
-
-  let marks = createMarks(state, inlines, 0, inlines.length, memo, opt)
-
-  inlines = mergeInlinesAndMarks(inlines, marks)
-
-  setToken(state, inlines, opt)
-
-  if (opt.mditAttrs && max !== state.posMax) {
-    if (!attributesSrc) {
-      state.pos = max
-      return true
-    }
-    if (attributesSrc[1].length > 1) {
-      state.pos = max + attributesSrc[1].length
-    } else {
-      state.pos = max
+    tokens.splice(closeIdx, 1)
+    tokens.splice(i, 1)
+
+    let labelStartIdx = i
+    let labelEndIdx = i + labelTokens.length - 1
+    if (labelStartIdx > labelEndIdx) {
+      i++
+      continue
+    }
+
+    const wrapperPairs = []
+    while (labelStartIdx > 0) {
+      const prevToken = tokens[labelStartIdx - 1]
+      const nextToken = tokens[labelEndIdx + 1]
+      if (!prevToken || !nextToken) break
+      if (!/_close$/.test(prevToken.type)) break
+      const expectedOpen = prevToken.type.replace('_close', '_open')
+      if (nextToken.type !== expectedOpen) break
+      if (process.env.DEBUG_COLLAPSED === '1') {
+        console.log('[collapsed-ref] wrapper pair:', prevToken.type, nextToken.type)
+      }
+      wrapperPairs.push({
+        base: prevToken.type.replace('_close', ''),
+        tag: prevToken.tag,
+        markup: prevToken.markup
+      })
+      tokens.splice(labelEndIdx + 1, 1)
+      tokens.splice(labelStartIdx - 1, 1)
+      labelStartIdx -= 1
+      labelEndIdx -= 1
+    }
+
+    if (labelStartIdx > labelEndIdx) {
+      i++
+      continue
+    }
+
+    let labelLength = labelEndIdx - labelStartIdx + 1
+    const firstLabelToken = tokens[labelStartIdx]
+    const linkLevel = firstLabelToken ? Math.max(firstLabelToken.level - 1, 0) : 0
+    linkOpenToken.level = linkLevel
+    linkCloseToken.level = linkLevel
+    tokens.splice(labelStartIdx, 0, linkOpenToken)
+    tokens.splice(labelStartIdx + labelLength + 1, 0, linkCloseToken)
+
+    adjustTokenLevels(tokens, labelStartIdx + 1, labelStartIdx + labelLength + 1, 1)
+
+    if (wrapperPairs.length > 0) {
+      let insertIdx = labelStartIdx + 1
+      for (let wp = 0; wp < wrapperPairs.length; wp++) {
+        const pair = wrapperPairs[wp]
+        const innerOpen = new Token(pair.base + '_open', pair.tag, 1)
+        innerOpen.markup = pair.markup
+        innerOpen.level = linkLevel + 1 + wp
+        tokens.splice(insertIdx, 0, innerOpen)
+        insertIdx++
+        labelLength++
+      }
+      let linkClosePos = labelStartIdx + labelLength + 1
+      for (let wp = wrapperPairs.length - 1; wp >= 0; wp--) {
+        const pair = wrapperPairs[wp]
+        const innerClose = new Token(pair.base + '_close', pair.tag, -1)
+        innerClose.markup = pair.markup
+        innerClose.level = linkLevel + 1 + wp
+        tokens.splice(linkClosePos, 0, innerClose)
+        labelLength++
+      }
+    }
+
+    i = labelStartIdx + labelLength + 2
+  }
+}
+
+// Link cleanup helpers
+const mergeBrokenMarksAroundLinks = (tokens) => {
+  let i = 0
+  while (i < tokens.length) {
+    const closeToken = tokens[i]
+    if (!closeToken || !/_close$/.test(closeToken.type)) {
+      i++
+      continue
+    }
+    const openType = closeToken.type.replace('_close', '_open')
+    let j = i + 1
+    while (j < tokens.length && isWhitespaceToken(tokens[j])) j++
+    if (j >= tokens.length || tokens[j].type !== 'link_open') {
+      i++
+      continue
+    }
+    let linkDepth = 1
+    j++
+    while (j < tokens.length && linkDepth > 0) {
+      if (tokens[j].type === 'link_open') linkDepth++
+      if (tokens[j].type === 'link_close') linkDepth--
+      j++
+    }
+    if (linkDepth !== 0) {
+      i++
+      continue
+    }
+    while (j < tokens.length && isWhitespaceToken(tokens[j])) j++
+    if (j >= tokens.length) {
+      i++
+      continue
+    }
+    const reopenToken = tokens[j]
+    if (reopenToken.type !== openType || reopenToken.level !== closeToken.level) {
+      i++
+      continue
+    }
+    tokens.splice(j, 1)
+    tokens.splice(i, 1)
+  }
+}
+
+
+const mditStrongJa = (md, option) => {
+  const opt = {
+    dollarMath: true, //inline math $...$
+    mditAttrs: true, //markdown-it-attrs
+    mdBreaks: md.options.breaks,
+    disallowMixed: false, //Non-Japanese text handling
+  }
+  if (option) Object.assign(opt, option)
+
+  md.inline.ruler.before('emphasis', 'strong_ja', (state, silent) => {
+    return strongJa(state, silent, opt)
+  })
+
+  md.core.ruler.after('inline', 'strong_ja_collapsed_refs', (state) => {
+    const targets = state.env.__strongJaCollapsedTargets
+    if (!targets || targets.length === 0) return
+    for (const tokens of targets) {
+      if (!tokens || !tokens.length) continue
+      convertCollapsedReferenceLinks(tokens, state)
+      mergeBrokenMarksAroundLinks(tokens)
     }
-  } else {
-    state.pos = max
-  }
-  return true
-}
-
-const mditStrongJa = (md, option) => {
-  const opt = {
-    dollarMath: true, //inline math $...$
-    mditAttrs: true, //markdown-it-attrs
-    mdBreaks: md.options.breaks,
-    disallowMixed: false, //Non-Japanese text handling
-  }
-  if (option) Object.assign(opt, option)
-
-  md.inline.ruler.before('emphasis', 'strong_ja', (state, silent) => {
-    return strongJa(state, silent, opt)
+    delete state.env.__strongJaCollapsedTargets
+    delete state.env.__strongJaCollapsedTargetSet
   })
-}
-
-export default mditStrongJa
+}
+
+export default mditStrongJa