From 378ac9074e393354253281fbe243860732d1f631 Mon Sep 17 00:00:00 2001
From: Alex Kocharin <alex@kocharin.ru>
Date: Fri, 25 Sep 2015 18:53:55 +0300
Subject: [PATCH] Change emphasis parsing algorithm

Tokenize first, and replace tokens with emphasis tags on
a second pass using an algorithm close to one used in CM.
---
 docs/4.0_migration.md                       |  10 +-
 docs/5.0_migration.md                       |  18 ++
 lib/index.js                                |   7 +
 lib/parser_inline.js                        |  35 ++-
 lib/presets/commonmark.js                   |   5 +
 lib/presets/zero.js                         |   4 +
 lib/rules_inline/balance_pairs.js           |  36 +++
 lib/rules_inline/emphasis.js                | 257 ++++++++------------
 lib/rules_inline/image.js                   |   6 +-
 lib/rules_inline/state_inline.js            |  72 +++++-
 lib/rules_inline/strikethrough.js           | 199 +++++++--------
 lib/rules_inline/text_collapse.js           |  33 +++
 test/fixtures/commonmark/bad.txt            |  14 --
 test/fixtures/commonmark/good.txt           |   9 +
 test/fixtures/markdown-it/strikethrough.txt |   9 +
 test/misc.js                                |   5 -
 16 files changed, 421 insertions(+), 298 deletions(-)
 create mode 100644 docs/5.0_migration.md
 create mode 100644 lib/rules_inline/balance_pairs.js
 create mode 100644 lib/rules_inline/text_collapse.js

diff --git a/docs/4.0_migration.md b/docs/4.0_migration.md
index 75a255d..f8abedf 100644
--- a/docs/4.0_migration.md
+++ b/docs/4.0_migration.md
@@ -6,14 +6,14 @@ authors should update their packages.
 
 ## For users
 
-External API did not changed.
+External API did not change.
 
-- If you used `markdown-it` with plugins - make shure to update those.
+- If you used `markdown-it` with plugins - make sure to update those.
 - If you modified renderer - see dev info below.
-- If you did not used plugins and renderer modification - no chanches needed.
+- If you did not use plugins and renderer modification - no changes needed.
 
 
-## For Developpers
+## For developers
 
 ### Tokens and renderer
 
@@ -25,7 +25,7 @@ External API did not changed.
     repos in this org.
 - [Renderer](https://github.com/markdown-it/markdown-it/blob/master/lib/renderer.js)
   methods were unified. Number of custom renderer rules were significantly reduced.
-  Custom renderer functions need update due tokens format change.
+  Custom renderer functions need update due to tokens format change.
 
 ### Other changes
 
diff --git a/docs/5.0_migration.md b/docs/5.0_migration.md
new file mode 100644
index 0000000..3346210
--- /dev/null
+++ b/docs/5.0_migration.md
@@ -0,0 +1,18 @@
+Migration to v5
+===============
+
+v5 has the same external API as v4, only internals were changed. Plugins that introduce block elements might need an update.
+
+## For users
+
+External API did not change.
+
+- If you use `markdown-it` with plugins, make sure to update them.
+
+
+## For plugin developers
+
+- added `stateBlock.sCount` to calculate indents instead of `stateBlock.tShift`, it only differs if tabs are present:
+  - `stateBlock.tShift` is used to calculate a number of *characters* (tab is 1 character)
+  - `stateBlock.sCount` is used to calculate the block *offset* (tab is 1-4 characters depending on position)
+
diff --git a/lib/index.js b/lib/index.js
index 6b301e8..d6ab79d 100644
--- a/lib/index.js
+++ b/lib/index.js
@@ -382,6 +382,9 @@ MarkdownIt.prototype.configure = function (presets) {
       if (presets.components[name].rules) {
         self[name].ruler.enableOnly(presets.components[name].rules);
       }
+      if (presets.components[name].rules2) {
+        self[name].ruler2.enableOnly(presets.components[name].rules2);
+      }
     });
   }
   return this;
@@ -414,6 +417,8 @@ MarkdownIt.prototype.enable = function (list, ignoreInvalid) {
     result = result.concat(this[chain].ruler.enable(list, true));
   }, this);
 
+  result = result.concat(this.inline.ruler2.enable(list, true));
+
   var missed = list.filter(function (name) { return result.indexOf(name) < 0; });
 
   if (missed.length && !ignoreInvalid) {
@@ -440,6 +445,8 @@ MarkdownIt.prototype.disable = function (list, ignoreInvalid) {
     result = result.concat(this[chain].ruler.disable(list, true));
   }, this);
 
+  result = result.concat(this.inline.ruler2.disable(list, true));
+
   var missed = list.filter(function (name) { return result.indexOf(name) < 0; });
 
   if (missed.length && !ignoreInvalid) {
diff --git a/lib/parser_inline.js b/lib/parser_inline.js
index a19a5de..e89a71b 100644
--- a/lib/parser_inline.js
+++ b/lib/parser_inline.js
@@ -17,8 +17,8 @@ var _rules = [
   [ 'newline',         require('./rules_inline/newline') ],
   [ 'escape',          require('./rules_inline/escape') ],
   [ 'backticks',       require('./rules_inline/backticks') ],
-  [ 'strikethrough',   require('./rules_inline/strikethrough') ],
-  [ 'emphasis',        require('./rules_inline/emphasis') ],
+  [ 'strikethrough',   require('./rules_inline/strikethrough').tokenize ],
+  [ 'emphasis',        require('./rules_inline/emphasis').tokenize ],
   [ 'link',            require('./rules_inline/link') ],
   [ 'image',           require('./rules_inline/image') ],
   [ 'autolink',        require('./rules_inline/autolink') ],
@@ -26,11 +26,20 @@ var _rules = [
   [ 'entity',          require('./rules_inline/entity') ]
 ];
 
+var _rules2 = [
+  [ 'balance_pairs',   require('./rules_inline/balance_pairs') ],
+  [ 'strikethrough',   require('./rules_inline/strikethrough').postProcess ],
+  [ 'emphasis',        require('./rules_inline/emphasis').postProcess ],
+  [ 'text_collapse',   require('./rules_inline/text_collapse') ]
+];
+
 
 /**
  * new ParserInline()
  **/
 function ParserInline() {
+  var i;
+
   /**
    * ParserInline#ruler -> Ruler
    *
@@ -38,9 +47,21 @@ function ParserInline() {
    **/
   this.ruler = new Ruler();
 
-  for (var i = 0; i < _rules.length; i++) {
+  for (i = 0; i < _rules.length; i++) {
     this.ruler.push(_rules[i][0], _rules[i][1]);
   }
+
+  /**
+   * ParserInline#ruler2 -> Ruler
+   *
+   * [[Ruler]] instance. Second ruler used for post-processing
+   * (e.g. in emphasis-like rules).
+   **/
+  this.ruler2 = new Ruler();
+
+  for (i = 0; i < _rules2.length; i++) {
+    this.ruler2.push(_rules2[i][0], _rules2[i][1]);
+  }
 }
 
 
@@ -119,9 +140,17 @@ ParserInline.prototype.tokenize = function (state) {
  * Process input string and push inline tokens into `outTokens`
  **/
 ParserInline.prototype.parse = function (str, md, env, outTokens) {
+  var i, rules, len;
   var state = new this.State(str, md, env, outTokens);
 
   this.tokenize(state);
+
+  rules = this.ruler2.getRules('');
+  len = rules.length;
+
+  for (i = 0; i < len; i++) {
+    rules[i](state);
+  }
 };
 
 
diff --git a/lib/presets/commonmark.js b/lib/presets/commonmark.js
index 99e5b2b..b0aaabc 100644
--- a/lib/presets/commonmark.js
+++ b/lib/presets/commonmark.js
@@ -68,6 +68,11 @@ module.exports = {
         'link',
         'newline',
         'text'
+      ],
+      rules2: [
+        'balance_pairs',
+        'emphasis',
+        'text_collapse'
       ]
     }
   }
diff --git a/lib/presets/zero.js b/lib/presets/zero.js
index eda6dba..fd0376f 100644
--- a/lib/presets/zero.js
+++ b/lib/presets/zero.js
@@ -51,6 +51,10 @@ module.exports = {
     inline: {
       rules: [
         'text'
+      ],
+      rules2: [
+        'balance_pairs',
+        'text_collapse'
       ]
     }
   }
diff --git a/lib/rules_inline/balance_pairs.js b/lib/rules_inline/balance_pairs.js
new file mode 100644
index 0000000..4ec2845
--- /dev/null
+++ b/lib/rules_inline/balance_pairs.js
@@ -0,0 +1,36 @@
+// For each opening emphasis-like marker find a matching closing one
+//
+'use strict';
+
+
+module.exports = function link_pairs(state) {
+  var i, j, lastDelim, currDelim,
+      delimiters = state.delimiters,
+      max = state.delimiters.length;
+
+  for (i = 0; i < max; i++) {
+    lastDelim = delimiters[i];
+
+    if (!lastDelim.close) { continue; }
+
+    j = i - lastDelim.jump - 1;
+
+    while (j >= 0) {
+      currDelim = delimiters[j];
+
+      if (currDelim.open &&
+          currDelim.marker === lastDelim.marker &&
+          currDelim.end < 0 &&
+          currDelim.level === lastDelim.level) {
+
+        lastDelim.jump = i - j;
+        lastDelim.open = false;
+        currDelim.end  = i;
+        currDelim.jump = 0;
+        break;
+      }
+
+      j -= currDelim.jump + 1;
+    }
+  }
+};
diff --git a/lib/rules_inline/emphasis.js b/lib/rules_inline/emphasis.js
index 051d2d0..fa26e61 100644
--- a/lib/rules_inline/emphasis.js
+++ b/lib/rules_inline/emphasis.js
@@ -3,172 +3,121 @@
 'use strict';
 
 
-var isWhiteSpace   = require('../common/utils').isWhiteSpace;
-var isPunctChar    = require('../common/utils').isPunctChar;
-var isMdAsciiPunct = require('../common/utils').isMdAsciiPunct;
-
-
-// parse sequence of emphasis markers,
-// "start" should point at a valid marker
-function scanDelims(state, start) {
-  var pos = start, lastChar, nextChar, count, can_open, can_close,
-      isLastWhiteSpace, isLastPunctChar,
-      isNextWhiteSpace, isNextPunctChar,
-      left_flanking = true,
-      right_flanking = true,
-      max = state.posMax,
+// Insert each marker as a separate text token, and add it to delimiter list
+//
+module.exports.tokenize = function emphasis(state, silent) {
+  var i, scanned, token,
+      start = state.pos,
       marker = state.src.charCodeAt(start);
 
-  // treat beginning of the line as a whitespace
-  lastChar = start > 0 ? state.src.charCodeAt(start - 1) : 0x20;
-
-  while (pos < max && state.src.charCodeAt(pos) === marker) { pos++; }
-
-  count = pos - start;
-
-  // treat end of the line as a whitespace
-  nextChar = pos < max ? state.src.charCodeAt(pos) : 0x20;
-
-  isLastPunctChar = isMdAsciiPunct(lastChar) || isPunctChar(String.fromCharCode(lastChar));
-  isNextPunctChar = isMdAsciiPunct(nextChar) || isPunctChar(String.fromCharCode(nextChar));
-
-  isLastWhiteSpace = isWhiteSpace(lastChar);
-  isNextWhiteSpace = isWhiteSpace(nextChar);
-
-  if (isNextWhiteSpace) {
-    left_flanking = false;
-  } else if (isNextPunctChar) {
-    if (!(isLastWhiteSpace || isLastPunctChar)) {
-      left_flanking = false;
-    }
+  if (silent) { return false; }
+
+  if (marker !== 0x5F /* _ */ && marker !== 0x2A /* * */) { return false; }
+
+  scanned = state.scanDelims(state.pos, marker === 0x2A);
+
+  for (i = 0; i < scanned.length; i++) {
+    token         = state.push('text', '', 0);
+    token.content = String.fromCharCode(marker);
+
+    state.delimiters.push({
+      // Char code of the starting marker (number).
+      //
+      marker: marker,
+
+      // An amount of characters before this one that's equivalent to
+      // current one. In plain English: if this delimiter does not open
+      // an emphasis, neither do previous `jump` characters.
+      //
+      // Used to skip sequences like "*****" in one step, for 1st asterisk
+      // value will be 0, for 2nd it's 1 and so on.
+      //
+      jump:   i,
+
+      // A position of the token this delimiter corresponds to.
+      //
+      token:  state.tokens.length - 1,
+
+      // Token level.
+      //
+      level:  state.level,
+
+      // If this delimiter is matched as a valid opener, `end` will be
+      // equal to its position, otherwise it's `-1`.
+      //
+      end:    -1,
+
+      // Boolean flags that determine if this delimiter could open or close
+      // an emphasis.
+      //
+      open:   scanned.can_open,
+      close:  scanned.can_close
+    });
   }
 
-  if (isLastWhiteSpace) {
-    right_flanking = false;
-  } else if (isLastPunctChar) {
-    if (!(isNextWhiteSpace || isNextPunctChar)) {
-      right_flanking = false;
-    }
-  }
+  state.pos += scanned.length;
 
-  if (marker === 0x5F /* _ */) {
-    // "_" inside a word can neither open nor close an emphasis
-    can_open  = left_flanking  && (!right_flanking || isLastPunctChar);
-    can_close = right_flanking && (!left_flanking  || isNextPunctChar);
-  } else {
-    can_open  = left_flanking;
-    can_close = right_flanking;
-  }
+  return true;
+};
 
-  return {
-    can_open: can_open,
-    can_close: can_close,
-    delims: count
-  };
-}
-
-module.exports = function emphasis(state, silent) {
-  var startCount,
-      count,
-      found,
-      oldCount,
-      newCount,
-      stack,
-      res,
-      token,
-      max = state.posMax,
-      start = state.pos,
-      marker = state.src.charCodeAt(start);
 
-  if (marker !== 0x5F/* _ */ && marker !== 0x2A /* * */) { return false; }
-  if (silent) { return false; } // don't run any pairs in validation mode
+// Walk through delimiter list and replace text tokens with tags
+//
+module.exports.postProcess = function emphasis(state) {
+  var i,
+      startDelim,
+      endDelim,
+      token,
+      ch,
+      isStrong,
+      delimiters = state.delimiters,
+      max = state.delimiters.length;
 
-  res = scanDelims(state, start);
-  startCount = res.delims;
-  if (!res.can_open) {
-    state.pos += startCount;
-    // Earlier we checked !silent, but this implementation does not need it
-    state.pending += state.src.slice(start, state.pos);
-    return true;
-  }
+  for (i = 0; i < max; i++) {
+    startDelim = delimiters[i];
 
-  state.pos = start + startCount;
-  stack = [ startCount ];
-
-  while (state.pos < max) {
-    if (state.src.charCodeAt(state.pos) === marker) {
-      res = scanDelims(state, state.pos);
-      count = res.delims;
-      if (res.can_close) {
-        oldCount = stack.pop();
-        newCount = count;
-
-        while (oldCount !== newCount) {
-          if (newCount < oldCount) {
-            stack.push(oldCount - newCount);
-            break;
-          }
-
-          // assert(newCount > oldCount)
-          newCount -= oldCount;
-
-          if (stack.length === 0) { break; }
-          state.pos += oldCount;
-          oldCount = stack.pop();
-        }
-
-        if (stack.length === 0) {
-          startCount = oldCount;
-          found = true;
-          break;
-        }
-        state.pos += count;
-        continue;
-      }
-
-      if (res.can_open) { stack.push(count); }
-      state.pos += count;
+    if (startDelim.marker !== 0x5F/* _ */ && startDelim.marker !== 0x2A/* * */) {
       continue;
     }
 
-    state.md.inline.skipToken(state);
-  }
-
-  if (!found) {
-    // parser failed to find ending tag, so it's not valid emphasis
-    state.pos = start;
-    return false;
-  }
-
-  // found!
-  state.posMax = state.pos;
-  state.pos = start + startCount;
-
-  // Earlier we checked !silent, but this implementation does not need it
-
-  // we have `startCount` starting and ending markers,
-  // now trying to serialize them into tokens
-  for (count = startCount; count > 1; count -= 2) {
-    token        = state.push('strong_open', 'strong', 1);
-    token.markup = String.fromCharCode(marker) + String.fromCharCode(marker);
-  }
-  if (count % 2) {
-    token        = state.push('em_open', 'em', 1);
-    token.markup = String.fromCharCode(marker);
-  }
-
-  state.md.inline.tokenize(state);
+    // Process only opening markers
+    if (startDelim.end === -1) {
+      continue;
+    }
 
-  if (count % 2) {
-    token        = state.push('em_close', 'em', -1);
-    token.markup = String.fromCharCode(marker);
-  }
-  for (count = startCount; count > 1; count -= 2) {
-    token        = state.push('strong_close', 'strong', -1);
-    token.markup = String.fromCharCode(marker) + String.fromCharCode(marker);
+    endDelim = delimiters[startDelim.end];
+
+    // If the next delimiter has the same marker and is adjacent to this one,
+    // merge those into one strong delimiter.
+    //
+    // `<em><em>whatever</em></em>` -> `<strong>whatever</strong>`
+    //
+    isStrong = i + 1 < max &&
+               delimiters[i + 1].end === startDelim.end - 1 &&
+               delimiters[i + 1].token === startDelim.token + 1 &&
+               delimiters[startDelim.end - 1].token === endDelim.token - 1 &&
+               delimiters[i + 1].marker === startDelim.marker;
+
+    ch = String.fromCharCode(startDelim.marker);
+
+    token         = state.tokens[startDelim.token];
+    token.type    = isStrong ? 'strong_open' : 'em_open';
+    token.tag     = isStrong ? 'strong' : 'em';
+    token.nesting = 1;
+    token.markup  = isStrong ? ch + ch : ch;
+    token.content = '';
+
+    token         = state.tokens[endDelim.token];
+    token.type    = isStrong ? 'strong_close' : 'em_close';
+    token.tag     = isStrong ? 'strong' : 'em';
+    token.nesting = -1;
+    token.markup  = isStrong ? ch + ch : ch;
+    token.content = '';
+
+    if (isStrong) {
+      state.tokens[delimiters[i + 1].token].content = '';
+      state.tokens[delimiters[startDelim.end - 1].token].content = '';
+      i++;
+    }
   }
-
-  state.pos = state.posMax + startCount;
-  state.posMax = max;
-  return true;
 };
diff --git a/lib/rules_inline/image.js b/lib/rules_inline/image.js
index 32355e0..167b861 100644
--- a/lib/rules_inline/image.js
+++ b/lib/rules_inline/image.js
@@ -136,16 +136,12 @@ module.exports = function image(state, silent) {
   // so all that's left to do is to call tokenizer.
   //
   if (!silent) {
-    state.pos = labelStart;
-    state.posMax = labelEnd;
-
-    var newState = new state.md.inline.State(
+    state.md.inline.parse(
       state.src.slice(labelStart, labelEnd),
       state.md,
       state.env,
       tokens = []
     );
-    newState.md.inline.tokenize(newState);
 
     token          = state.push('image', 'img', 0);
     token.attrs    = attrs = [ [ 'src', href ], [ 'alt', '' ] ];
diff --git a/lib/rules_inline/state_inline.js b/lib/rules_inline/state_inline.js
index 8b3a5fc..2847496 100644
--- a/lib/rules_inline/state_inline.js
+++ b/lib/rules_inline/state_inline.js
@@ -3,7 +3,11 @@
 'use strict';
 
 
-var Token = require('../token');
+var Token          = require('../token');
+var isWhiteSpace   = require('../common/utils').isWhiteSpace;
+var isPunctChar    = require('../common/utils').isPunctChar;
+var isMdAsciiPunct = require('../common/utils').isMdAsciiPunct;
+
 
 function StateInline(src, md, env, outTokens) {
   this.src = src;
@@ -19,6 +23,8 @@ function StateInline(src, md, env, outTokens) {
 
   this.cache = {};        // Stores { start: end } pairs. Useful for backtrack
                           // optimization of pairs parse (emphasis, strikes).
+
+  this.delimiters = [];   // Emphasis-like delimiters
 }
 
 
@@ -53,6 +59,70 @@ StateInline.prototype.push = function (type, tag, nesting) {
   return token;
 };
 
+
+// Scan a sequence of emphasis-like markers, and determine whether
+// it can start an emphasis sequence or end an emphasis sequence.
+//
+//  - start - position to scan from (it should point at a valid marker);
+//  - canSplitWord - determine if these markers can be found inside a word
+//
+StateInline.prototype.scanDelims = function (start, canSplitWord) {
+  var pos = start, lastChar, nextChar, count, can_open, can_close,
+      isLastWhiteSpace, isLastPunctChar,
+      isNextWhiteSpace, isNextPunctChar,
+      left_flanking = true,
+      right_flanking = true,
+      max = this.posMax,
+      marker = this.src.charCodeAt(start);
+
+  // treat beginning of the line as a whitespace
+  lastChar = start > 0 ? this.src.charCodeAt(start - 1) : 0x20;
+
+  while (pos < max && this.src.charCodeAt(pos) === marker) { pos++; }
+
+  count = pos - start;
+
+  // treat end of the line as a whitespace
+  nextChar = pos < max ? this.src.charCodeAt(pos) : 0x20;
+
+  isLastPunctChar = isMdAsciiPunct(lastChar) || isPunctChar(String.fromCharCode(lastChar));
+  isNextPunctChar = isMdAsciiPunct(nextChar) || isPunctChar(String.fromCharCode(nextChar));
+
+  isLastWhiteSpace = isWhiteSpace(lastChar);
+  isNextWhiteSpace = isWhiteSpace(nextChar);
+
+  if (isNextWhiteSpace) {
+    left_flanking = false;
+  } else if (isNextPunctChar) {
+    if (!(isLastWhiteSpace || isLastPunctChar)) {
+      left_flanking = false;
+    }
+  }
+
+  if (isLastWhiteSpace) {
+    right_flanking = false;
+  } else if (isLastPunctChar) {
+    if (!(isNextWhiteSpace || isNextPunctChar)) {
+      right_flanking = false;
+    }
+  }
+
+  if (!canSplitWord) {
+    can_open  = left_flanking  && (!right_flanking || isLastPunctChar);
+    can_close = right_flanking && (!left_flanking  || isNextPunctChar);
+  } else {
+    can_open  = left_flanking;
+    can_close = right_flanking;
+  }
+
+  return {
+    can_open:  can_open,
+    can_close: can_close,
+    length:    count
+  };
+};
+
+
 // re-export Token class to use in block rules
 StateInline.prototype.Token = Token;
 
diff --git a/lib/rules_inline/strikethrough.js b/lib/rules_inline/strikethrough.js
index 5787687..0347241 100644
--- a/lib/rules_inline/strikethrough.js
+++ b/lib/rules_inline/strikethrough.js
@@ -3,138 +3,115 @@
 'use strict';
 
 
-var isWhiteSpace   = require('../common/utils').isWhiteSpace;
-var isPunctChar    = require('../common/utils').isPunctChar;
-var isMdAsciiPunct = require('../common/utils').isMdAsciiPunct;
-
-
-// parse sequence of markers,
-// "start" should point at a valid marker
-function scanDelims(state, start) {
-  var pos = start, lastChar, nextChar, count,
-      isLastWhiteSpace, isLastPunctChar,
-      isNextWhiteSpace, isNextPunctChar,
-      can_open = true,
-      can_close = true,
-      max = state.posMax,
+// Insert each marker as a separate text token, and add it to delimiter list
+//
+module.exports.tokenize = function strikethrough(state, silent) {
+  var i, scanned, token, len, ch,
+      start = state.pos,
       marker = state.src.charCodeAt(start);
 
-  // treat beginning of the line as a whitespace
-  lastChar = start > 0 ? state.src.charCodeAt(start - 1) : 0x20;
+  if (silent) { return false; }
 
-  while (pos < max && state.src.charCodeAt(pos) === marker) { pos++; }
+  if (marker !== 0x7E/* ~ */) { return false; }
 
-  if (pos >= max) {
-    can_open = false;
-  }
+  scanned = state.scanDelims(state.pos, true);
+  len = scanned.length;
+  ch = String.fromCharCode(marker);
 
-  count = pos - start;
+  if (len < 2) { return false; }
 
-  // treat end of the line as a whitespace
-  nextChar = pos < max ? state.src.charCodeAt(pos) : 0x20;
+  if (len % 2) {
+    token         = state.push('text', '', 0);
+    token.content = ch;
+    len--;
+  }
 
-  isLastPunctChar = isMdAsciiPunct(lastChar) || isPunctChar(String.fromCharCode(lastChar));
-  isNextPunctChar = isMdAsciiPunct(nextChar) || isPunctChar(String.fromCharCode(nextChar));
+  for (i = 0; i < len; i += 2) {
+    token         = state.push('text', '', 0);
+    token.content = ch + ch;
+
+    state.delimiters.push({
+      marker: marker,
+      jump:   i,
+      token:  state.tokens.length - 1,
+      level:  state.level,
+      end:    -1,
+      open:   scanned.can_open,
+      close:  scanned.can_close
+    });
+  }
 
-  isLastWhiteSpace = isWhiteSpace(lastChar);
-  isNextWhiteSpace = isWhiteSpace(nextChar);
+  state.pos += scanned.length;
 
-  if (isNextWhiteSpace) {
-    can_open = false;
-  } else if (isNextPunctChar) {
-    if (!(isLastWhiteSpace || isLastPunctChar)) {
-      can_open = false;
-    }
-  }
+  return true;
+};
 
-  if (isLastWhiteSpace) {
-    can_close = false;
-  } else if (isLastPunctChar) {
-    if (!(isNextWhiteSpace || isNextPunctChar)) {
-      can_close = false;
-    }
-  }
 
-  return {
-    can_open: can_open,
-    can_close: can_close,
-    delims: count
-  };
-}
-
-
-module.exports = function strikethrough(state, silent) {
-  var startCount,
-      count,
-      tagCount,
-      found,
-      stack,
-      res,
+// Walk through delimiter list and replace text tokens with tags
+//
+module.exports.postProcess = function strikethrough(state) {
+  var i, j,
+      startDelim,
+      endDelim,
       token,
-      max = state.posMax,
-      start = state.pos,
-      marker = state.src.charCodeAt(start);
+      loneMarkers = [],
+      delimiters = state.delimiters,
+      max = state.delimiters.length;
 
-  if (marker !== 0x7E/* ~ */) { return false; }
-  if (silent) { return false; } // don't run any pairs in validation mode
-
-  res = scanDelims(state, start);
-  startCount = res.delims;
-  if (!res.can_open) {
-    state.pos += startCount;
-    // Earlier we checked !silent, but this implementation does not need it
-    state.pending += state.src.slice(start, state.pos);
-    return true;
-  }
+  for (i = 0; i < max; i++) {
+    startDelim = delimiters[i];
 
-  stack = Math.floor(startCount / 2);
-  if (stack <= 0) { return false; }
-  state.pos = start + startCount;
-
-  while (state.pos < max) {
-    if (state.src.charCodeAt(state.pos) === marker) {
-      res = scanDelims(state, state.pos);
-      count = res.delims;
-      tagCount = Math.floor(count / 2);
-      if (res.can_close) {
-        if (tagCount >= stack) {
-          state.pos += count - 2;
-          found = true;
-          break;
-        }
-        stack -= tagCount;
-        state.pos += count;
-        continue;
-      }
-
-      if (res.can_open) { stack += tagCount; }
-      state.pos += count;
+    if (startDelim.marker !== 0x7E/* ~ */) {
       continue;
     }
 
-    state.md.inline.skipToken(state);
-  }
+    if (startDelim.end === -1) {
+      continue;
+    }
 
-  if (!found) {
-    // parser failed to find ending tag, so it's not valid emphasis
-    state.pos = start;
-    return false;
-  }
+    endDelim = delimiters[startDelim.end];
+
+    token         = state.tokens[startDelim.token];
+    token.type    = 's_open';
+    token.tag     = 's';
+    token.nesting = 1;
+    token.markup  = '~~';
+    token.content = '';
 
-  // found!
-  state.posMax = state.pos;
-  state.pos = start + 2;
+    token         = state.tokens[endDelim.token];
+    token.type    = 's_close';
+    token.tag     = 's';
+    token.nesting = -1;
+    token.markup  = '~~';
+    token.content = '';
 
-  // Earlier we checked !silent, but this implementation does not need it
-  token        = state.push('s_open', 's', 1);
-  token.markup = '~~';
+    if (state.tokens[endDelim.token - 1].type === 'text' &&
+        state.tokens[endDelim.token - 1].content === '~') {
 
-  state.md.inline.tokenize(state);
+      loneMarkers.push(endDelim.token - 1);
+    }
+  }
 
-  token        = state.push('s_close', 's', -1);
-  token.markup = '~~';
+  // If a marker sequence has an odd number of characters, it's splitted
+  // like this: `~~~~~` -> `~` + `~~` + `~~`, leaving one marker at the
+  // start of the sequence.
+  //
+  // So, we have to move all those markers after subsequent s_close tags.
+  //
+  while (loneMarkers.length) {
+    i = loneMarkers.pop();
+    j = i + 1;
+
+    while (j < state.tokens.length && state.tokens[j].type === 's_close') {
+      j++;
+    }
 
-  state.pos = state.posMax + 2;
-  state.posMax = max;
-  return true;
+    j--;
+
+    if (i !== j) {
+      token = state.tokens[j];
+      state.tokens[j] = state.tokens[i];
+      state.tokens[i] = token;
+    }
+  }
 };
diff --git a/lib/rules_inline/text_collapse.js b/lib/rules_inline/text_collapse.js
new file mode 100644
index 0000000..3104c0c
--- /dev/null
+++ b/lib/rules_inline/text_collapse.js
@@ -0,0 +1,33 @@
+// Merge adjacent text nodes into one, and re-calculate all token levels
+//
+'use strict';
+
+
+module.exports = function text_collapse(state) {
+  var curr, last,
+      level = 0,
+      tokens = state.tokens,
+      max = state.tokens.length;
+
+  for (curr = last = 0; curr < max; curr++) {
+    // re-calculate levels
+    level += tokens[curr].nesting;
+    tokens[curr].level = level;
+
+    if (tokens[curr].type === 'text' &&
+        curr + 1 < max &&
+        tokens[curr + 1].type === 'text') {
+
+      // collapse two adjacent text nodes
+      tokens[curr + 1].content = tokens[curr].content + tokens[curr + 1].content;
+    } else {
+      if (curr !== last) { tokens[last] = tokens[curr]; }
+
+      last++;
+    }
+  }
+
+  if (curr !== last) {
+    tokens.length = last;
+  }
+};
diff --git a/test/fixtures/commonmark/bad.txt b/test/fixtures/commonmark/bad.txt
index 9a799f9..e69de29 100644
--- a/test/fixtures/commonmark/bad.txt
+++ b/test/fixtures/commonmark/bad.txt
@@ -1,14 +0,0 @@
-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-src line: 6241
-
-.
-*foo __bar *baz bim__ bam*
-.
-<p><em>foo <strong>bar *baz bim</strong> bam</em></p>
-.
-
-error:
-
-<p>*foo <strong>bar *baz bim</strong> bam*</p>
-
-
diff --git a/test/fixtures/commonmark/good.txt b/test/fixtures/commonmark/good.txt
index bdb86d3..5df0d9b 100644
--- a/test/fixtures/commonmark/good.txt
+++ b/test/fixtures/commonmark/good.txt
@@ -5444,6 +5444,15 @@ src line: 6235
 <p><em><em>foo</em>bar</em>*</p>
 .
 
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+src line: 6241
+
+.
+*foo __bar *baz bim__ bam*
+.
+<p><em>foo <strong>bar *baz bim</strong> bam</em></p>
+.
+
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
 src line: 6249
 
diff --git a/test/fixtures/markdown-it/strikethrough.txt b/test/fixtures/markdown-it/strikethrough.txt
index 299dc3a..fb1463f 100644
--- a/test/fixtures/markdown-it/strikethrough.txt
+++ b/test/fixtures/markdown-it/strikethrough.txt
@@ -22,6 +22,15 @@ x ~~~~foo~~~~
 <p>x <s><s>foo</s></s></p>
 .
 
+.
+x ~~a ~~foo~~~~~~~~~~~bar~~ b~~
+
+x ~~a ~~foo~~~~~~~~~~~~bar~~ b~~
+.
+<p>x <s>a <s>foo</s></s>~~~<s><s>bar</s> b</s></p>
+<p>x <s>a <s>foo</s></s>~~~~<s><s>bar</s> b</s></p>
+.
+
 
 Strikeouts have the same priority as emphases:
 .
diff --git a/test/misc.js b/test/misc.js
index 82e2381..e35b3de 100644
--- a/test/misc.js
+++ b/test/misc.js
@@ -261,11 +261,6 @@ describe('Links validation', function () {
 
 describe('maxNesting', function () {
 
-  it('Inline parser should not nest above limit', function () {
-    var md = markdownit({ maxNesting: 2 });
-    assert.strictEqual(md.render('*foo *bar *baz* bar* foo*'), '<p><em>foo <em>bar *baz* bar</em> foo</em></p>\n');
-  });
-
   it('Block parser should not nest above limit', function () {
     var md = markdownit({ maxNesting: 2 });
     assert.strictEqual(