The cache does not remember match length but next position after the match.

This leads to simpler and faster code because it avoids one addition and
subtraction.

lib/compiler.js

@@ -267,7 +267,7 @@ PEG.Grammar.Rule.prototype.compile = function() {
     "  this._cache[${name|string}] = this._cache[${name|string}] || [];",
     "  var cachedResult = this._cache[${name|string}][this._pos];",
     "  if (cachedResult !== undefined) {",
-    "    this._pos += cachedResult.length;",
+    "    this._pos = cachedResult.nextPos;",
     "    return cachedResult.result;",
     "  }",
     "  ",
@@ -279,8 +279,8 @@ PEG.Grammar.Rule.prototype.compile = function() {
     "  ${reportMatchFailureCode}",
     "  ",
     "  this._cache[${name|string}][pos] = {",
-    "    length: this._pos - pos,",
-    "    result: ${resultVar}",
+    "    nextPos: this._pos,",
+    "    result:  ${resultVar}",
     "  };",
     "  return ${resultVar};",
     "};",