Optimized scanner to avoid virtual calls for every character read.

test/cctest/test-parsing.cc

Index: test/cctest/test-parsing.cc
diff --git a/test/cctest/test-parsing.cc b/test/cctest/test-parsing.cc
index badbab5eeb1614bd487489a835ca92cebcc4d2e0..e642d1b622b2aa5a2ab4212c14c21d082834799a 100755
--- a/test/cctest/test-parsing.cc
+++ b/test/cctest/test-parsing.cc
@@ -260,10 +260,12 @@ TEST(StandAlonePreParser) {
   uintptr_t stack_limit = i::StackGuard::real_climit();
   for (int i = 0; programs[i]; i++) {
     const char* program = programs[i];
-    unibrow::Utf8InputBuffer<256> stream(program, strlen(program));
+    i::Utf8ToUC16CharacterStream stream(
+        reinterpret_cast<const i::byte*>(program),
+        static_cast<unsigned>(strlen(program)));
     i::CompleteParserRecorder log;
     i::V8JavaScriptScanner scanner;
-    scanner.Initialize(i::Handle<i::String>::null(), &stream);
+    scanner.Initialize(&stream);
 
     v8::preparser::PreParser::PreParseResult result =
         v8::preparser::PreParser::PreParseProgram(&scanner,
@@ -289,9 +291,10 @@ TEST(RegressChromium62639) {
   // and then used the invalid currently scanned literal. This always
   // failed in debug mode, and sometimes crashed in release mode.
 
-  unibrow::Utf8InputBuffer<256> stream(program, strlen(program));
+  i::Utf8ToUC16CharacterStream stream(reinterpret_cast<const i::byte*>(program),
+                                      static_cast<unsigned>(strlen(program)));
   i::ScriptDataImpl* data =
-      i::ParserApi::PreParse(i::Handle<i::String>::null(), &stream, NULL);
+      i::ParserApi::PreParse(&stream, NULL);
   CHECK(data->HasError());
   delete data;
 }
@@ -310,10 +313,10 @@ TEST(Regress928) {
       "try { } catch (e) { var foo = function () { /* first */ } }"
       "var bar = function () { /* second */ }";
 
-  unibrow::Utf8InputBuffer<256> stream(program, strlen(program));
+  i::Utf8ToUC16CharacterStream stream(reinterpret_cast<const i::byte*>(program),
+                                      static_cast<unsigned>(strlen(program)));
   i::ScriptDataImpl* data =
-      i::ParserApi::PartialPreParse(i::Handle<i::String>::null(),
-                                    &stream, NULL);
+      i::ParserApi::PartialPreParse(&stream, NULL);
   CHECK(!data->HasError());
 
   data->Initialize();
@@ -347,10 +350,12 @@ TEST(PreParseOverflow) {
 
   uintptr_t stack_limit = i::StackGuard::real_climit();
 
-  unibrow::Utf8InputBuffer<256> stream(*program, strlen(*program));
+  i::Utf8ToUC16CharacterStream stream(
+      reinterpret_cast<const i::byte*>(*program),
+      static_cast<unsigned>(kProgramSize));
   i::CompleteParserRecorder log;
   i::V8JavaScriptScanner scanner;
-  scanner.Initialize(i::Handle<i::String>::null(), &stream);
+  scanner.Initialize(&stream);
 
 
   v8::preparser::PreParser::PreParseResult result =
@@ -360,3 +365,283 @@ TEST(PreParseOverflow) {
                                                 stack_limit);
   CHECK_EQ(v8::preparser::PreParser::kPreParseStackOverflow, result);
 }
+
+
+class TestExternalResource: public v8::String::ExternalStringResource {
+ public:
+  explicit TestExternalResource(uint16_t* data, int length)
+      : data_(data), length_(static_cast<size_t>(length)) { }
+
+  ~TestExternalResource() { }
+
+  const uint16_t* data() const {
+    return data_;
+  }
+
+  size_t length() const {
+    return length_;
+  }
+ private:
+  uint16_t* data_;
+  size_t length_;
+};
+
+
+#define CHECK_EQU(v1, v2) CHECK_EQ(static_cast<int>(v1), static_cast<int>(v2))
+
+void TestCharacterStream(const char* ascii_source,
+                         unsigned length,
+                         unsigned start = 0,
+                         unsigned end = 0) {
+  if (end == 0) end = length;
+  unsigned sub_length = end - start;
+  i::HandleScope test_scope;
+  i::SmartPointer<i::uc16> uc16_buffer(new i::uc16[length]);
+  for (unsigned i = 0; i < length; i++) {
+    uc16_buffer[i] = static_cast<i::uc16>(ascii_source[i]);
+  }
+  i::Vector<const char> ascii_vector(ascii_source, static_cast<int>(length));
+  i::Handle<i::String> ascii_string(
+      i::Factory::NewStringFromAscii(ascii_vector));
+  TestExternalResource resource(*uc16_buffer, length);
+  i::Handle<i::String> uc16_string(
+      i::Factory::NewExternalStringFromTwoByte(&resource));
+
+  i::ExternalTwoByteStringUC16CharacterStream uc16_stream(
+      i::Handle<i::ExternalTwoByteString>::cast(uc16_string), start, end);
+  i::GenericStringUC16CharacterStream string_stream(ascii_string, start, end);
+  i::Utf8ToUC16CharacterStream utf8_stream(
+      reinterpret_cast<const i::byte*>(ascii_source), end);
+  utf8_stream.SeekForward(start);
+
+  unsigned i = start;
+  while (i < end) {
+    // Read streams one char at a time
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+    int32_t c0 = ascii_source[i];
+    int32_t c1 = uc16_stream.Advance();
+    int32_t c2 = string_stream.Advance();
+    int32_t c3 = utf8_stream.Advance();
+    i++;
+    CHECK_EQ(c0, c1);
+    CHECK_EQ(c0, c2);
+    CHECK_EQ(c0, c3);
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+  }
+  while (i > start + sub_length / 4) {
+    // Pushback, re-read, pushback again.
+    int32_t c0 = ascii_source[i - 1];
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+    uc16_stream.PushBack(c0);
+    string_stream.PushBack(c0);
+    utf8_stream.PushBack(c0);
+    i--;
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+    int32_t c1 = uc16_stream.Advance();
+    int32_t c2 = string_stream.Advance();
+    int32_t c3 = utf8_stream.Advance();
+    i++;
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+    CHECK_EQ(c0, c1);
+    CHECK_EQ(c0, c2);
+    CHECK_EQ(c0, c3);
+    uc16_stream.PushBack(c0);
+    string_stream.PushBack(c0);
+    utf8_stream.PushBack(c0);
+    i--;
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+  }
+  unsigned halfway = start + sub_length / 2;
+  uc16_stream.SeekForward(halfway - i);
+  string_stream.SeekForward(halfway - i);
+  utf8_stream.SeekForward(halfway - i);
+  i = halfway;
+  CHECK_EQU(i, uc16_stream.pos());
+  CHECK_EQU(i, string_stream.pos());
+  CHECK_EQU(i, utf8_stream.pos());
+
+  while (i < end) {
+    // Read streams one char at a time
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+    int32_t c0 = ascii_source[i];
+    int32_t c1 = uc16_stream.Advance();
+    int32_t c2 = string_stream.Advance();
+    int32_t c3 = utf8_stream.Advance();
+    i++;
+    CHECK_EQ(c0, c1);
+    CHECK_EQ(c0, c2);
+    CHECK_EQ(c0, c3);
+    CHECK_EQU(i, uc16_stream.pos());
+    CHECK_EQU(i, string_stream.pos());
+    CHECK_EQU(i, utf8_stream.pos());
+  }
+
+  int32_t c1 = uc16_stream.Advance();
+  int32_t c2 = string_stream.Advance();
+  int32_t c3 = utf8_stream.Advance();
+  CHECK_LT(c1, 0);
+  CHECK_LT(c2, 0);
+  CHECK_LT(c3, 0);
+}
+
+
+TEST(CharacterStreams) {
+  v8::HandleScope handles;
+  v8::Persistent<v8::Context> context = v8::Context::New();
+  v8::Context::Scope context_scope(context);
+
+  TestCharacterStream("abc\0\n\r\x7f", 7);
+  static const unsigned kBigStringSize = 4096;
+  char buffer[kBigStringSize + 1];
+  for (unsigned i = 0; i < kBigStringSize; i++) {
+    buffer[i] = static_cast<char>(i & 0x7f);
+  }
+  TestCharacterStream(buffer, kBigStringSize);
+
+  TestCharacterStream(buffer, kBigStringSize, 576, 3298);
+
+  TestCharacterStream("\0", 1);
+  TestCharacterStream("", 0);
+}
+
+
+TEST(Utf8CharacterStream) {
+  static const unsigned kMaxUC16CharU = unibrow::Utf8::kMaxThreeByteChar;
+  static const int kMaxUC16Char = static_cast<int>(kMaxUC16CharU);
+
+  static const int kAllUtf8CharsSize =
+      (unibrow::Utf8::kMaxOneByteChar + 1) +
+      (unibrow::Utf8::kMaxTwoByteChar - unibrow::Utf8::kMaxOneByteChar) * 2 +
+      (unibrow::Utf8::kMaxThreeByteChar - unibrow::Utf8::kMaxTwoByteChar) * 3;
+  static const unsigned kAllUtf8CharsSizeU =
+      static_cast<unsigned>(kAllUtf8CharsSize);
+
+  char buffer[kAllUtf8CharsSizeU];
+  unsigned cursor = 0;
+  for (int i = 0; i <= kMaxUC16Char; i++) {
+    cursor += unibrow::Utf8::Encode(buffer + cursor, i);
+  }
+  ASSERT(cursor == kAllUtf8CharsSizeU);
+
+  i::Utf8ToUC16CharacterStream stream(reinterpret_cast<const i::byte*>(buffer),
+                                      kAllUtf8CharsSizeU);
+  for (int i = 0; i <= kMaxUC16Char; i++) {
+    CHECK_EQU(i, stream.pos());
+    int32_t c = stream.Advance();
+    CHECK_EQ(i, c);
+    CHECK_EQU(i + 1, stream.pos());
+  }
+  for (int i = kMaxUC16Char; i >= 0; i--) {
+    CHECK_EQU(i + 1, stream.pos());
+    stream.PushBack(i);
+    CHECK_EQU(i, stream.pos());
+  }
+  int i = 0;
+  while (stream.pos() < kMaxUC16CharU) {
+    CHECK_EQU(i, stream.pos());
+    unsigned progress = stream.SeekForward(12);
+    i += progress;
+    int32_t c = stream.Advance();
+    if (i <= kMaxUC16Char) {
+      CHECK_EQ(i, c);
+    } else {
+      CHECK_EQ(-1, c);
+    }
+    i += 1;
+    CHECK_EQU(i, stream.pos());
+  }
+}
+
+#undef CHECK_EQU
+
+void TestStreamScanner(i::UC16CharacterStream* stream,
+                       i::Token::Value* expected_tokens,
+                       int skip_pos = 0,  // Zero means not skipping.
+                       int skip_to = 0) {
+  i::V8JavaScriptScanner scanner;
+  scanner.Initialize(stream, i::JavaScriptScanner::kAllLiterals);
+
+  int i = 0;
+  do {
+    i::Token::Value expected = expected_tokens[i];
+    i::Token::Value actual = scanner.Next();
+    CHECK_EQ(i::Token::String(expected), i::Token::String(actual));
+    if (scanner.location().end_pos == skip_pos) {
+      scanner.SeekForward(skip_to);
+    }
+    i++;
+  } while (expected_tokens[i] != i::Token::ILLEGAL);
+}
+
+TEST(StreamScanner) {
+  const char* str1 = "{ foo get for : */ <- \n\n /*foo*/ bib";
+  i::Utf8ToUC16CharacterStream stream1(reinterpret_cast<const i::byte*>(str1),
+                                       static_cast<unsigned>(strlen(str1)));
+  i::Token::Value expectations1[] = {
+      i::Token::LBRACE,
+      i::Token::IDENTIFIER,
+      i::Token::IDENTIFIER,
+      i::Token::FOR,
+      i::Token::COLON,
+      i::Token::MUL,
+      i::Token::DIV,
+      i::Token::LT,
+      i::Token::SUB,
+      i::Token::IDENTIFIER,
+      i::Token::EOS,
+      i::Token::ILLEGAL
+  };
+  TestStreamScanner(&stream1, expectations1, 0, 0);
+
+  const char* str2 = "case default const {THIS\nPART\nSKIPPED} do";
+  i::Utf8ToUC16CharacterStream stream2(reinterpret_cast<const i::byte*>(str2),
+                                       static_cast<unsigned>(strlen(str2)));
+  i::Token::Value expectations2[] = {
+      i::Token::CASE,
+      i::Token::DEFAULT,
+      i::Token::CONST,
+      i::Token::LBRACE,
+      // Skipped part here
+      i::Token::RBRACE,
+      i::Token::DO,
+      i::Token::EOS,
+      i::Token::ILLEGAL
+  };
+  ASSERT_EQ('{', str2[19]);
+  ASSERT_EQ('}', str2[37]);
+  TestStreamScanner(&stream2, expectations2, 20, 37);
+
+  const char* str3 = "{}}}}";
+  i::Token::Value expectations3[] = {
+      i::Token::LBRACE,
+      i::Token::RBRACE,
+      i::Token::RBRACE,
+      i::Token::RBRACE,
+      i::Token::RBRACE,
+      i::Token::EOS,
+      i::Token::ILLEGAL
+  };
+  // Skip zero-four RBRACEs.
+  for (int i = 0; i <= 4; i++) {
+     expectations3[6 - i] = i::Token::ILLEGAL;
+     expectations3[5 - i] = i::Token::EOS;
+     i::Utf8ToUC16CharacterStream stream3(
+         reinterpret_cast<const i::byte*>(str3),
+         static_cast<unsigned>(strlen(str3)));
+     TestStreamScanner(&stream3, expectations3, 1, 1 + i);
+  }
+}