Make RegExp character class match JSC.

See http://trac.webkit.org/changeset/73594

Review URL: http://codereview.chromium.org/5723002

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5974 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

diff --git a/src/parser.cc b/src/parser.cc
index fd93aae4241d7494dc26522ca4904081e746c4eb..94ad57c9c28ba02a40b111562c56aa1f243129b1 100644 (file)
--- a/src/parser.cc
+++ b/src/parser.cc
@@ -4449,6 +4449,22 @@ CharacterRange RegExpParser::ParseClassAtom(uc16* char_class) {
 }
 
 
+static const uc16 kNoCharClass = 0;
+
+// Adds range or pre-defined character class to character ranges.
+// If char_class is not kInvalidClass, it's interpreted as a class
+// escape (i.e., 's' means whitespace, from '\s').
+static inline void AddRangeOrEscape(ZoneList<CharacterRange>* ranges,
+                                    uc16 char_class,
+                                    CharacterRange range) {
+  if (char_class != kNoCharClass) {
+    CharacterRange::AddClassEscape(char_class, ranges);
+  } else {
+    ranges->Add(range);
+  }
+}
+
+
 RegExpTree* RegExpParser::ParseCharacterClass() {
   static const char* kUnterminated = "Unterminated character class";
   static const char* kRangeOutOfOrder = "Range out of order in character class";
@@ -4462,12 +4478,8 @@ RegExpTree* RegExpParser::ParseCharacterClass() {
   }
   ZoneList<CharacterRange>* ranges = new ZoneList<CharacterRange>(2);
   while (has_more() && current() != ']') {
-    uc16 char_class = 0;
+    uc16 char_class = kNoCharClass;
     CharacterRange first = ParseClassAtom(&char_class CHECK_FAILED);
-    if (char_class) {
-      CharacterRange::AddClassEscape(char_class, ranges);
-      continue;
-    }
     if (current() == '-') {
       Advance();
       if (current() == kEndMarker) {
@@ -4475,15 +4487,17 @@ RegExpTree* RegExpParser::ParseCharacterClass() {
         // following code report an error.
         break;
       } else if (current() == ']') {
-        ranges->Add(first);
+        AddRangeOrEscape(ranges, char_class, first);
         ranges->Add(CharacterRange::Singleton('-'));
         break;
       }
-      CharacterRange next = ParseClassAtom(&char_class CHECK_FAILED);
-      if (char_class) {
-        ranges->Add(first);
+      uc16 char_class_2 = kNoCharClass;
+      CharacterRange next = ParseClassAtom(&char_class_2 CHECK_FAILED);
+      if (char_class != kNoCharClass || char_class_2 != kNoCharClass) {
+        // Either end is an escaped character class. Treat the '-' verbatim.
+        AddRangeOrEscape(ranges, char_class, first);
         ranges->Add(CharacterRange::Singleton('-'));
-        CharacterRange::AddClassEscape(char_class, ranges);
+        AddRangeOrEscape(ranges, char_class_2, next);
         continue;
       }
       if (first.from() > next.to()) {
@@ -4491,7 +4505,7 @@ RegExpTree* RegExpParser::ParseCharacterClass() {
       }
       ranges->Add(CharacterRange::Range(first.from(), next.to()));
     } else {
-      ranges->Add(first);
+      AddRangeOrEscape(ranges, char_class, first);
     }
   }
   if (!has_more()) {

diff --git a/test/mjsunit/regexp.js b/test/mjsunit/regexp.js
index b57b86d2d88bbf2a882e0116e3c8be38b13f3cd7..4c1d2e315f856ed9e241d52eb2cbcf98e987e2c4 100644 (file)
--- a/test/mjsunit/regexp.js
+++ b/test/mjsunit/regexp.js
@@ -202,6 +202,17 @@ assertFalse(re.test('\n'));
 assertFalse(re.test('a'));
 assertFalse(re.test('Z'));
 
+// First - is treated as range operator, second as literal minus.
+// This follows the specification in parsing, but doesn't throw on
+// the \s at the beginning of the range.
+re = /[\s-0-9]/;
+assertTrue(re.test(' '));
+assertTrue(re.test('\xA0'));
+assertTrue(re.test('-'));
+assertTrue(re.test('0'));
+assertTrue(re.test('9'));
+assertFalse(re.test('1'));
+
 // Test beginning and end of line assertions with or without the
 // multiline flag.
 re = /^\d+/;
@@ -647,3 +658,4 @@ assertEquals(4, re.exec("zimzamzumba").index);
 assertEquals(["bc"], re.exec("zimzomzumbc"));
 assertFalse(re.test("c"));
 assertFalse(re.test(""));
+