Cleanup latin-1 conversion check in regexp engine

R=yangguo@chromium.org
BUG=

Review URL: https://chromiumcodereview.appspot.com/11880045
Patch from Dan Carney <dcarney@google.com>.

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@13400 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

diff --git a/src/jsregexp.cc b/src/jsregexp.cc
index a3efb85..f6e2e7f 100644 (file)
--- a/src/jsregexp.cc
+++ b/src/jsregexp.cc
@@ -2875,23 +2875,9 @@ RegExpNode* TextNode::FilterASCII(int depth, bool ignore_case) {
         if (!ignore_case) return set_replacement(NULL);
         // Here, we need to check for characters whose upper and lower cases
         // are outside the Latin-1 range.
-        // TODO(dcarney): Replace this code with a simple
-        // table lookup in unibrow::Latin-1.
-        // TODO(dcarney): Test cases!.
-        unibrow::uchar result;
-        int chars;
-        chars = unibrow::ToLowercase::Convert(quarks[j], 0, &result, NULL);
-        if (chars > 1 ||
-            (chars == 1 && result <= String::kMaxOneByteCharCodeU)) {
-          continue;
-        }
-        chars = unibrow::ToUppercase::Convert(quarks[j], 0, &result, NULL);
-        if (chars > 1 ||
-            (chars == 1 && result <= String::kMaxOneByteCharCodeU)) {
-          continue;
+        if (!unibrow::Latin1::NonLatin1CanBeConvertedToLatin1(quarks[j])) {
+          return set_replacement(NULL);
         }
-        // This character is definitely not in the Latin-1 range.
-        return set_replacement(NULL);
 #endif
       }
     } else {

diff --git a/src/unicode-inl.h b/src/unicode-inl.h
index c3a00ed..b4e2cb5 100644 (file)
--- a/src/unicode-inl.h
+++ b/src/unicode-inl.h
@@ -79,6 +79,36 @@ template <class T, int s> int Mapping<T, s>::CalculateValue(uchar c, uchar n,
 }
 
 
+bool Latin1::NonLatin1CanBeConvertedToLatin1(uint16_t c) {
+  ASSERT(c > Latin1::kMaxChar);
+  switch (c) {
+    case 0x130:
+    case 0x131:
+    case 0x149:
+    case 0x178:
+    case 0x17f:
+    case 0x1f0:
+    case 0x1e96:
+    case 0x1e97:
+    case 0x1e98:
+    case 0x1e99:
+    case 0x1e9a:
+    case 0x1e9e:
+    case 0x212a:
+    case 0x212b:
+    case 0xfb00:
+    case 0xfb01:
+    case 0xfb02:
+    case 0xfb03:
+    case 0xfb04:
+    case 0xfb05:
+    case 0xfb06:
+      return true;
+  }
+  return false;
+}
+
+
 unsigned Utf8::Encode(char* str, uchar c, int previous) {
   static const int kMask = ~(1 << 6);
   if (c <= kMaxOneByteChar) {

diff --git a/src/unicode.h b/src/unicode.h
index 61dbab4..0278576 100644 (file)
--- a/src/unicode.h
+++ b/src/unicode.h
@@ -140,6 +140,7 @@ class Latin1 {
 #else
   static const unsigned kMaxChar = 0xff;
 #endif
+  static inline bool NonLatin1CanBeConvertedToLatin1(uint16_t);
 };
 
 class Utf8 {

diff --git a/test/cctest/test-strings.cc b/test/cctest/test-strings.cc
index 0d2971b..3684b87 100644 (file)
--- a/test/cctest/test-strings.cc
+++ b/test/cctest/test-strings.cc
@@ -1275,3 +1275,40 @@ TEST(IsAscii) {
   CHECK(String::IsAscii(static_cast<char*>(NULL), 0));
   CHECK(String::IsOneByte(static_cast<uc16*>(NULL), 0));
 }
+
+
+static bool CanBeConvertedToLatin1(uint16_t c) {
+  CHECK(c > unibrow::Latin1::kMaxChar);
+  uint32_t result[4];
+  int chars;
+  chars = unibrow::ToLowercase::Convert(c, 0, result, NULL);
+  if (chars > 0) {
+    CHECK_LE(chars, static_cast<int>(sizeof(result)));
+    for (int i = 0; i < chars; i++) {
+      if (result[i] <= unibrow::Latin1::kMaxChar) {
+        return true;
+      }
+    }
+  }
+  chars = unibrow::ToUppercase::Convert(c, 0, result, NULL);
+  if (chars > 0) {
+    CHECK_LE(chars, static_cast<int>(sizeof(result)));
+    for (int i = 0; i < chars; i++) {
+      if (result[i] <= unibrow::Latin1::kMaxChar) {
+        return true;
+      }
+    }
+  }
+  return false;
+}
+
+
+TEST(Latin1) {
+#ifndef ENABLE_LATIN_1
+    if (true) return;
+#endif
+  for (uint16_t c = unibrow::Latin1::kMaxChar + 1; c != 0; c++) {
+    CHECK_EQ(CanBeConvertedToLatin1(c),
+             unibrow::Latin1::NonLatin1CanBeConvertedToLatin1(c));
+  }
+}