Implement WebSocketExtensionParser

net/websockets/websocket_extension_parser.cc

Index: net/websockets/websocket_extension_parser.cc
diff --git a/net/websockets/websocket_extension_parser.cc b/net/websockets/websocket_extension_parser.cc
new file mode 100644
index 0000000000000000000000000000000000000000..ea9987b077e1055668bb4998488a167ebbcee433
--- /dev/null
+++ b/net/websockets/websocket_extension_parser.cc
@@ -0,0 +1,200 @@
+// Copyright 2013 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "net/websockets/websocket_extension_parser.h"
+
+#include "base/strings/string_util.h"
+
+namespace net {
+
+// Backup the head pointer if the parser has error.

[Adam Rice, 2013/09/17 04:30:33]

I think this grammar can be parsed with single-character lookahead. Backtracking
should not be necessary.

[yhirano, 2013/09/17 06:03:31]

I deleted some |has_error_ = false| statements.
The rule "extension-list = 1#extension" needs potential backtracking, though
nobody uses it. At any rate, it doesn't hurt performance.

+class WebSocketExtensionParser::CurrentPointerBackup {
+ public:
+  explicit CurrentPointerBackup(WebSocketExtensionParser* parser);
+  ~CurrentPointerBackup();
+
+ private:
+  WebSocketExtensionParser* parser_;
+  const char* original_head_;
+};
+
+WebSocketExtensionParser::CurrentPointerBackup::CurrentPointerBackup(
+    WebSocketExtensionParser* parser)
+    : parser_(parser), original_head_(parser->current_) {}
+
+WebSocketExtensionParser::CurrentPointerBackup::~CurrentPointerBackup() {
+  if (parser_->has_error_)
+     parser_->current_ = original_head_;
+}
+
+WebSocketExtensionParser::WebSocketExtensionParser() {}
+
+WebSocketExtensionParser::~WebSocketExtensionParser() {}
+
+void WebSocketExtensionParser::Parse(const char* data, size_t size) {
+  current_ = data;
+  end_ = &data[size];
+  extensions_.clear();
+  has_error_ = false;
+  CurrentPointerBackup backup(this);
+
+  ConsumeExtensionList();
+  has_error_ = has_error_ || (current_ != end_);
+  if (has_error_)
+    extensions_.clear();
+}
+
+void WebSocketExtensionParser::Consume(const char* data, size_t size) {
+  DCHECK(!has_error_);
+  CurrentPointerBackup backup(this);
+  ConsumeLWS();
+  DCHECK(!has_error_);
+  if (UnconsumedBytes() < size) {
+    has_error_ = true;
+    return;
+  }
+  if (memcmp(data, current_, size)) {
+    has_error_ = true;
+    return;
+  }
+  current_ += size;
+}
+
+void WebSocketExtensionParser::ConsumeExtensionList() {
+  DCHECK(!has_error_);
+  WebSocketExtension extension("");
+  ConsumeExtension(&extension);
+  if (has_error_) return;
+  do {
+    extensions_.push_back(extension);
+    ConsumeExtension(&extension);
+  } while (!has_error_);
+  has_error_ = false;
+}
+
+void WebSocketExtensionParser::ConsumeExtension(WebSocketExtension* extension) {
+  DCHECK(!has_error_);
+  CurrentPointerBackup backup(this);
+  base::StringPiece name;
+  ConsumeToken(&name);
+  if (has_error_) return;
+  *extension = WebSocketExtension(name.as_string());
+
+
+  while (true) {
+    Consume(";", 1);
+    if (has_error_) {
+      has_error_ = false;
+      return;
+    }
+    WebSocketExtension::Parameter parameter("");
+    ConsumeExtensionParameter(&parameter);
+    if (has_error_) return;
+    extension->Add(parameter);
+  }
+}
+
+void WebSocketExtensionParser::ConsumeExtensionParameter(
+    WebSocketExtension::Parameter* parameter) {
+  DCHECK(!has_error_);
+  CurrentPointerBackup backup(this);
+  base::StringPiece name, value;
+  std::string value_string;
+
+  ConsumeToken(&name);
+  if (has_error_) return;
+  Consume("=", 1);
+  if (has_error_) {
+    has_error_ = false;
+    *parameter = WebSocketExtension::Parameter(name.as_string());
+    return;
+  }
+
+  ConsumeToken(&value);
+  if (has_error_) {
+    has_error_ = false;
+    ConsumeQuotedToken(&value_string);
+    if (has_error_) return;
+  } else {
+    value_string = value.as_string();
+  }
+  *parameter = WebSocketExtension::Parameter(name.as_string(), value_string);
+}
+
+void WebSocketExtensionParser::ConsumeToken(base::StringPiece* token) {
+  DCHECK(!has_error_);
+  CurrentPointerBackup backup(this);
+  ConsumeLWS();
+  DCHECK(!has_error_);
+  const char* head = current_;
+  while (current_ < end_ &&
+         !IsControl(current_[0]) && !IsSeparator(current_[0]))
+    ++current_;
+  if (current_ == head) {
+    has_error_ = true;
+    return;
+  }
+  *token = base::StringPiece(head, current_ - head);
+}
+
+void WebSocketExtensionParser::ConsumeQuotedToken(std::string* token) {
+  DCHECK(!has_error_);
+  CurrentPointerBackup backup(this);
+  Consume("\"", 1);
+  if (has_error_) return;
+  *token = "";
+  while (current_ < end_ && !IsControl(current_[0])) {
+    if (UnconsumedBytes() >= 2 && current_[0] == '\\') {
+      char next = current_[1];
+      if (IsControl(next) || IsSeparator(next)) break;
+      *token += next;
+      current_ += 2;
+    } else if (IsSeparator(current_[0])) {
+      break;
+    } else {
+      *token += current_[0];
+      ++current_;
+    }
+  }
+  Consume("\"", 1);
+  has_error_ = has_error_ || token->empty();
+}
+
+// Unlike the "implied *LWS" specification in RFC2616, this parser
+// consumes *LWS, because there are no adjacent tokens in WebSocket
+// extension grammer.
+void WebSocketExtensionParser::ConsumeLWS() {
+  DCHECK(!has_error_);
+  size_t counter = 0;
+  while (current_ < end_) {
+    if (current_[0] == ' ' || current_[0] == '\t') {
+      while (current_ < end_ && (current_[0] == ' ' || current_[0] == '\t'))
+        ++current_;
+    } else if (UnconsumedBytes() >= 3 &&
+               current_[0] == '\r' && current_[1] == '\n' &&
+               (current_[2] == ' ' || current_[2] == '\t')) {
+      current_ += 2;
+      while (current_ < end_ && (current_[0] == ' ' || current_[0] == '\t'))
+        ++current_;
+      break;
+    } else {
+      break;
+    }
+    ++counter;
+  }
+  return;
+}
+
+// static
+bool WebSocketExtensionParser::IsControl(char c) {
+  return (0 <= c && c <= 31) || c == 127;
+}
+
+// static
+bool WebSocketExtensionParser::IsSeparator(char c) {
+  const char separators[] = "()<>@,;:\\\"/[]?={} \t";
+  return strchr(separators, c);
+}
+
+}  // namespace net