support filtering event with url regex

chromium_src/extensions/common/url_pattern.cc

@@ -0,0 +1,619 @@
+// Copyright (c) 2012 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 "extensions/common/url_pattern.h"
+
+#include <ostream>
+
+#include "base/strings/pattern.h"
+#include "base/strings/string_number_conversions.h"
+#include "base/strings/string_piece.h"
+#include "base/strings/string_split.h"
+#include "base/strings/string_util.h"
+#include "base/strings/stringprintf.h"
+#include "content/public/common/url_constants.h"
+#include "net/base/registry_controlled_domains/registry_controlled_domain.h"
+#include "url/gurl.h"
+#include "url/url_util.h"
+
+const char extensions::URLPattern::kAllUrlsPattern[] = "<all_urls>";
+const char kExtensionScheme[] = "chrome-extension";
+
+namespace {
+
+// TODO(aa): What about more obscure schemes like data: and javascript: ?
+// Note: keep this array in sync with kValidSchemeMasks.
+const char* kValidSchemes[] = {
+    url::kHttpScheme,
+    url::kHttpsScheme,
+    url::kFileScheme,
+    url::kFtpScheme,
+    content::kChromeUIScheme,
+    kExtensionScheme,
+    url::kFileSystemScheme,
+};
+
+const int kValidSchemeMasks[] = {
+  extensions::URLPattern::SCHEME_HTTP,
+  extensions::URLPattern::SCHEME_HTTPS,
+  extensions::URLPattern::SCHEME_FILE,
+  extensions::URLPattern::SCHEME_FTP,
+  extensions::URLPattern::SCHEME_CHROMEUI,
+  extensions::URLPattern::SCHEME_EXTENSION,
+  extensions::URLPattern::SCHEME_FILESYSTEM,
+};
+
+static_assert(arraysize(kValidSchemes) == arraysize(kValidSchemeMasks),
+              "must keep these arrays in sync");
+
+const char kParseSuccess[] = "Success.";
+const char kParseErrorMissingSchemeSeparator[] = "Missing scheme separator.";
+const char kParseErrorInvalidScheme[] = "Invalid scheme.";
+const char kParseErrorWrongSchemeType[] = "Wrong scheme type.";
+const char kParseErrorEmptyHost[] = "Host can not be empty.";
+const char kParseErrorInvalidHostWildcard[] = "Invalid host wildcard.";
+const char kParseErrorEmptyPath[] = "Empty path.";
+const char kParseErrorInvalidPort[] = "Invalid port.";
+const char kParseErrorInvalidHost[] = "Invalid host.";
+
+// Message explaining each URLPattern::ParseResult.
+const char* const kParseResultMessages[] = {
+  kParseSuccess,
+  kParseErrorMissingSchemeSeparator,
+  kParseErrorInvalidScheme,
+  kParseErrorWrongSchemeType,
+  kParseErrorEmptyHost,
+  kParseErrorInvalidHostWildcard,
+  kParseErrorEmptyPath,
+  kParseErrorInvalidPort,
+  kParseErrorInvalidHost,
+};
+
+static_assert(extensions::URLPattern::NUM_PARSE_RESULTS == arraysize(kParseResultMessages),
+              "must add message for each parse result");
+
+const char kPathSeparator[] = "/";
+
+bool IsStandardScheme(const std::string& scheme) {
+  // "*" gets the same treatment as a standard scheme.
+  if (scheme == "*")
+    return true;
+
+  return url::IsStandard(scheme.c_str(),
+                         url::Component(0, static_cast<int>(scheme.length())));
+}
+
+bool IsValidPortForScheme(const std::string& scheme, const std::string& port) {
+  if (port == "*")
+    return true;
+
+  // Only accept non-wildcard ports if the scheme uses ports.
+  if (url::DefaultPortForScheme(scheme.c_str(), scheme.length()) ==
+      url::PORT_UNSPECIFIED) {
+    return false;
+  }
+
+  int parsed_port = url::PORT_UNSPECIFIED;
+  if (!base::StringToInt(port, &parsed_port))
+    return false;
+  return (parsed_port >= 0) && (parsed_port < 65536);
+}
+
+// Returns |path| with the trailing wildcard stripped if one existed.
+//
+// The functions that rely on this (OverlapsWith and Contains) are only
+// called for the patterns inside URLPatternSet. In those cases, we know that
+// the path will have only a single wildcard at the end. This makes figuring
+// out overlap much easier. It seems like there is probably a computer-sciency
+// way to solve the general case, but we don't need that yet.
+std::string StripTrailingWildcard(const std::string& path) {
+  size_t wildcard_index = path.find('*');
+  size_t path_last = path.size() - 1;
+  return wildcard_index == path_last ? path.substr(0, path_last) : path;
+}
+
+}  // namespace
+
+namespace extensions {
+// static
+bool URLPattern::IsValidSchemeForExtensions(const std::string& scheme) {
+  for (size_t i = 0; i < arraysize(kValidSchemes); ++i) {
+    if (scheme == kValidSchemes[i])
+      return true;
+  }
+  return false;
+}
+
+URLPattern::URLPattern()
+    : valid_schemes_(SCHEME_ALL),
+      match_all_urls_(false),
+      match_subdomains_(false),
+      port_("*") {}
+
+URLPattern::URLPattern(int valid_schemes)
+    : valid_schemes_(valid_schemes),
+      match_all_urls_(false),
+      match_subdomains_(false),
+      port_("*") {}
+
+URLPattern::URLPattern(int valid_schemes, const std::string& pattern)
+    // Strict error checking is used, because this constructor is only
+    // appropriate when we know |pattern| is valid.
+    : valid_schemes_(valid_schemes),
+      match_all_urls_(false),
+      match_subdomains_(false),
+      port_("*") {
+  ParseResult result = Parse(pattern);
+  if (PARSE_SUCCESS != result)
+    NOTREACHED() << "URLPattern invalid: " << pattern << " result " << result;
+}
+
+URLPattern::~URLPattern() {
+}
+
+bool URLPattern::operator<(const URLPattern& other) const {
+  return GetAsString() < other.GetAsString();
+}
+
+bool URLPattern::operator>(const URLPattern& other) const {
+  return GetAsString() > other.GetAsString();
+}
+
+bool URLPattern::operator==(const URLPattern& other) const {
+  return GetAsString() == other.GetAsString();
+}
+
+std::ostream& operator<<(std::ostream& out, const URLPattern& url_pattern) {
+  return out << '"' << url_pattern.GetAsString() << '"';
+}
+
+URLPattern::ParseResult URLPattern::Parse(const std::string& pattern) {
+  spec_.clear();
+  SetMatchAllURLs(false);
+  SetMatchSubdomains(false);
+  SetPort("*");
+
+  // Special case pattern to match every valid URL.
+  if (pattern == kAllUrlsPattern) {
+    SetMatchAllURLs(true);
+    return PARSE_SUCCESS;
+  }
+
+  // Parse out the scheme.
+  size_t scheme_end_pos = pattern.find(url::kStandardSchemeSeparator);
+  bool has_standard_scheme_separator = true;
+
+  // Some urls also use ':' alone as the scheme separator.
+  if (scheme_end_pos == std::string::npos) {
+    scheme_end_pos = pattern.find(':');
+    has_standard_scheme_separator = false;
+  }
+
+  if (scheme_end_pos == std::string::npos)
+    return PARSE_ERROR_MISSING_SCHEME_SEPARATOR;
+
+  if (!SetScheme(pattern.substr(0, scheme_end_pos)))
+    return PARSE_ERROR_INVALID_SCHEME;
+
+  bool standard_scheme = IsStandardScheme(scheme_);
+  if (standard_scheme != has_standard_scheme_separator)
+    return PARSE_ERROR_WRONG_SCHEME_SEPARATOR;
+
+  // Advance past the scheme separator.
+  scheme_end_pos +=
+      (standard_scheme ? strlen(url::kStandardSchemeSeparator) : 1);
+  if (scheme_end_pos >= pattern.size())
+    return PARSE_ERROR_EMPTY_HOST;
+
+  // Parse out the host and path.
+  size_t host_start_pos = scheme_end_pos;
+  size_t path_start_pos = 0;
+
+  if (!standard_scheme) {
+    path_start_pos = host_start_pos;
+  } else if (scheme_ == url::kFileScheme) {
+    size_t host_end_pos = pattern.find(kPathSeparator, host_start_pos);
+    if (host_end_pos == std::string::npos) {
+      // Allow hostname omission.
+      // e.g. file://* is interpreted as file:///*,
+      // file://foo* is interpreted as file:///foo*.
+      path_start_pos = host_start_pos - 1;
+    } else {
+      // Ignore hostname if scheme is file://.
+      // e.g. file://localhost/foo is equal to file:///foo.
+      path_start_pos = host_end_pos;
+    }
+  } else {
+    size_t host_end_pos = pattern.find(kPathSeparator, host_start_pos);
+
+    // Host is required.
+    if (host_start_pos == host_end_pos)
+      return PARSE_ERROR_EMPTY_HOST;
+
+    if (host_end_pos == std::string::npos)
+      return PARSE_ERROR_EMPTY_PATH;
+
+    host_ = pattern.substr(host_start_pos, host_end_pos - host_start_pos);
+
+    // The first component can optionally be '*' to match all subdomains.
+    std::vector<std::string> host_components = base::SplitString(
+        host_, ".", base::TRIM_WHITESPACE, base::SPLIT_WANT_ALL);
+
+    // Could be empty if the host only consists of whitespace characters.
+    if (host_components.empty() ||
+        (host_components.size() == 1 && host_components[0].empty()))
+      return PARSE_ERROR_EMPTY_HOST;
+
+    if (host_components[0] == "*") {
+      match_subdomains_ = true;
+      host_components.erase(host_components.begin(),
+                            host_components.begin() + 1);
+    }
+    host_ = JoinString(host_components, ".");
+
+    path_start_pos = host_end_pos;
+  }
+
+  SetPath(pattern.substr(path_start_pos));
+
+  size_t port_pos = host_.find(':');
+  if (port_pos != std::string::npos) {
+    if (!SetPort(host_.substr(port_pos + 1)))
+      return PARSE_ERROR_INVALID_PORT;
+    host_ = host_.substr(0, port_pos);
+  }
+
+  // No other '*' can occur in the host, though. This isn't necessary, but is
+  // done as a convenience to developers who might otherwise be confused and
+  // think '*' works as a glob in the host.
+  if (host_.find('*') != std::string::npos)
+    return PARSE_ERROR_INVALID_HOST_WILDCARD;
+
+  // Null characters are not allowed in hosts.
+  if (host_.find('\0') != std::string::npos)
+    return PARSE_ERROR_INVALID_HOST;
+
+  return PARSE_SUCCESS;
+}
+
+void URLPattern::SetValidSchemes(int valid_schemes) {
+  spec_.clear();
+  valid_schemes_ = valid_schemes;
+}
+
+void URLPattern::SetHost(const std::string& host) {
+  spec_.clear();
+  host_ = host;
+}
+
+void URLPattern::SetMatchAllURLs(bool val) {
+  spec_.clear();
+  match_all_urls_ = val;
+
+  if (val) {
+    match_subdomains_ = true;
+    scheme_ = "*";
+    host_.clear();
+    SetPath("/*");
+  }
+}
+
+void URLPattern::SetMatchSubdomains(bool val) {
+  spec_.clear();
+  match_subdomains_ = val;
+}
+
+bool URLPattern::SetScheme(const std::string& scheme) {
+  spec_.clear();
+  scheme_ = scheme;
+  if (scheme_ == "*") {
+    valid_schemes_ &= (SCHEME_HTTP | SCHEME_HTTPS);
+  } else if (!IsValidScheme(scheme_)) {
+    return false;
+  }
+  return true;
+}
+
+bool URLPattern::IsValidScheme(const std::string& scheme) const {
+  if (valid_schemes_ == SCHEME_ALL)
+    return true;
+
+  for (size_t i = 0; i < arraysize(kValidSchemes); ++i) {
+    if (scheme == kValidSchemes[i] && (valid_schemes_ & kValidSchemeMasks[i]))
+      return true;
+  }
+
+  return false;
+}
+
+void URLPattern::SetPath(const std::string& path) {
+  spec_.clear();
+  path_ = path;
+  path_escaped_ = path_;
+  base::ReplaceSubstringsAfterOffset(&path_escaped_, 0, "\\", "\\\\");
+  base::ReplaceSubstringsAfterOffset(&path_escaped_, 0, "?", "\\?");
+}
+
+bool URLPattern::SetPort(const std::string& port) {
+  spec_.clear();
+  if (IsValidPortForScheme(scheme_, port)) {
+    port_ = port;
+    return true;
+  }
+  return false;
+}
+
+bool URLPattern::MatchesURL(const GURL& test) const {
+  const GURL* test_url = &test;
+  bool has_inner_url = test.inner_url() != NULL;
+
+  if (has_inner_url) {
+    if (!test.SchemeIsFileSystem())
+      return false;  // The only nested URLs we handle are filesystem URLs.
+    test_url = test.inner_url();
+  }
+
+  if (!MatchesScheme(test_url->scheme()))
+    return false;
+
+  if (match_all_urls_)
+    return true;
+
+  std::string path_for_request = test.PathForRequest();
+  if (has_inner_url)
+    path_for_request = test_url->path() + path_for_request;
+
+  return MatchesSecurityOriginHelper(*test_url) &&
+         MatchesPath(path_for_request);
+}
+
+bool URLPattern::MatchesSecurityOrigin(const GURL& test) const {
+  const GURL* test_url = &test;
+  bool has_inner_url = test.inner_url() != NULL;
+
+  if (has_inner_url) {
+    if (!test.SchemeIsFileSystem())
+      return false;  // The only nested URLs we handle are filesystem URLs.
+    test_url = test.inner_url();
+  }
+
+  if (!MatchesScheme(test_url->scheme()))
+    return false;
+
+  if (match_all_urls_)
+    return true;
+
+  return MatchesSecurityOriginHelper(*test_url);
+}
+
+bool URLPattern::MatchesScheme(const std::string& test) const {
+  if (!IsValidScheme(test))
+    return false;
+
+  return scheme_ == "*" || test == scheme_;
+}
+
+bool URLPattern::MatchesHost(const std::string& host) const {
+  std::string test(url::kHttpScheme);
+  test += url::kStandardSchemeSeparator;
+  test += host;
+  test += "/";
+  return MatchesHost(GURL(test));
+}
+
+bool URLPattern::MatchesHost(const GURL& test) const {
+  // If the hosts are exactly equal, we have a match.
+  if (test.host() == host_)
+    return true;
+
+  // If we're matching subdomains, and we have no host in the match pattern,
+  // that means that we're matching all hosts, which means we have a match no
+  // matter what the test host is.
+  if (match_subdomains_ && host_.empty())
+    return true;
+
+  // Otherwise, we can only match if our match pattern matches subdomains.
+  if (!match_subdomains_)
+    return false;
+
+  // We don't do subdomain matching against IP addresses, so we can give up now
+  // if the test host is an IP address.
+  if (test.HostIsIPAddress())
+    return false;
+
+  // Check if the test host is a subdomain of our host.
+  if (test.host().length() <= (host_.length() + 1))
+    return false;
+
+  if (test.host().compare(test.host().length() - host_.length(),
+                          host_.length(), host_) != 0)
+    return false;
+
+  return test.host()[test.host().length() - host_.length() - 1] == '.';
+}
+
+bool URLPattern::ImpliesAllHosts() const {
+  // Check if it matches all urls or is a pattern like http://*/*.
+  if (match_all_urls_ ||
+      (match_subdomains_ && host_.empty() && port_ == "*" && path_ == "/*")) {
+    return true;
+  }
+
+  // If this doesn't even match subdomains, it can't possibly imply all hosts.
+  if (!match_subdomains_)
+    return false;
+
+  // If |host_| is a recognized TLD, this will be 0. We don't include private
+  // TLDs, so that, e.g., *.appspot.com does not imply all hosts.
+  size_t registry_length = net::registry_controlled_domains::GetRegistryLength(
+      host_,
+      net::registry_controlled_domains::EXCLUDE_UNKNOWN_REGISTRIES,
+      net::registry_controlled_domains::EXCLUDE_PRIVATE_REGISTRIES);
+  // If there was more than just a TLD in the host (e.g., *.foobar.com), it
+  // doesn't imply all hosts.
+  if (registry_length > 0)
+    return false;
+
+  // At this point the host could either be just a TLD ("com") or some unknown
+  // TLD-like string ("notatld"). To disambiguate between them construct a
+  // fake URL, and check the registry. This returns 0 if the TLD is
+  // unrecognized, or the length of the recognized TLD.
+  registry_length = net::registry_controlled_domains::GetRegistryLength(
+      base::StringPrintf("foo.%s", host_.c_str()),
+      net::registry_controlled_domains::EXCLUDE_UNKNOWN_REGISTRIES,
+      net::registry_controlled_domains::EXCLUDE_PRIVATE_REGISTRIES);
+  // If we recognized this TLD, then this is a pattern like *.com, and it
+  // should imply all hosts. Otherwise, this doesn't imply all hosts.
+  return registry_length > 0;
+}
+
+bool URLPattern::MatchesSingleOrigin() const {
+  // Strictly speaking, the port is part of the origin, but in URLPattern it
+  // defaults to *. It's not very interesting anyway, so leave it out.
+  return !ImpliesAllHosts() && scheme_ != "*" && !match_subdomains_;
+}
+
+bool URLPattern::MatchesPath(const std::string& test) const {
+  // Make the behaviour of OverlapsWith consistent with MatchesURL, which is
+  // need to match hosted apps on e.g. 'google.com' also run on 'google.com/'.
+  if (test + "/*" == path_escaped_)
+    return true;
+
+  return base::MatchPattern(test, path_escaped_);
+}
+
+const std::string& URLPattern::GetAsString() const {
+  if (!spec_.empty())
+    return spec_;
+
+  if (match_all_urls_) {
+    spec_ = kAllUrlsPattern;
+    return spec_;
+  }
+
+  bool standard_scheme = IsStandardScheme(scheme_);
+
+  std::string spec = scheme_ +
+      (standard_scheme ? url::kStandardSchemeSeparator : ":");
+
+  if (scheme_ != url::kFileScheme && standard_scheme) {
+    if (match_subdomains_) {
+      spec += "*";
+      if (!host_.empty())
+        spec += ".";
+    }
+
+    if (!host_.empty())
+      spec += host_;
+
+    if (port_ != "*") {
+      spec += ":";
+      spec += port_;
+    }
+  }
+
+  if (!path_.empty())
+    spec += path_;
+
+  spec_ = spec;
+  return spec_;
+}
+
+bool URLPattern::OverlapsWith(const URLPattern& other) const {
+  if (match_all_urls() || other.match_all_urls())
+    return true;
+  return (MatchesAnyScheme(other.GetExplicitSchemes()) ||
+          other.MatchesAnyScheme(GetExplicitSchemes()))
+      && (MatchesHost(other.host()) || other.MatchesHost(host()))
+      && (MatchesPortPattern(other.port()) || other.MatchesPortPattern(port()))
+      && (MatchesPath(StripTrailingWildcard(other.path())) ||
+          other.MatchesPath(StripTrailingWildcard(path())));
+}
+
+bool URLPattern::Contains(const URLPattern& other) const {
+  if (match_all_urls())
+    return true;
+  return MatchesAllSchemes(other.GetExplicitSchemes()) &&
+         MatchesHost(other.host()) &&
+         (!other.match_subdomains_ || match_subdomains_) &&
+         MatchesPortPattern(other.port()) &&
+         MatchesPath(StripTrailingWildcard(other.path()));
+}
+
+bool URLPattern::MatchesAnyScheme(
+    const std::vector<std::string>& schemes) const {
+  for (std::vector<std::string>::const_iterator i = schemes.begin();
+       i != schemes.end(); ++i) {
+    if (MatchesScheme(*i))
+      return true;
+  }
+
+  return false;
+}
+
+bool URLPattern::MatchesAllSchemes(
+    const std::vector<std::string>& schemes) const {
+  for (std::vector<std::string>::const_iterator i = schemes.begin();
+       i != schemes.end(); ++i) {
+    if (!MatchesScheme(*i))
+      return false;
+  }
+
+  return true;
+}
+
+bool URLPattern::MatchesSecurityOriginHelper(const GURL& test) const {
+  // Ignore hostname if scheme is file://.
+  if (scheme_ != url::kFileScheme && !MatchesHost(test))
+    return false;
+
+  if (!MatchesPortPattern(base::IntToString(test.EffectiveIntPort())))
+    return false;
+
+  return true;
+}
+
+bool URLPattern::MatchesPortPattern(const std::string& port) const {
+  return port_ == "*" || port_ == port;
+}
+
+std::vector<std::string> URLPattern::GetExplicitSchemes() const {
+  std::vector<std::string> result;
+
+  if (scheme_ != "*" && !match_all_urls_ && IsValidScheme(scheme_)) {
+    result.push_back(scheme_);
+    return result;
+  }
+
+  for (size_t i = 0; i < arraysize(kValidSchemes); ++i) {
+    if (MatchesScheme(kValidSchemes[i])) {
+      result.push_back(kValidSchemes[i]);
+    }
+  }
+
+  return result;
+}
+
+std::vector<URLPattern> URLPattern::ConvertToExplicitSchemes() const {
+  std::vector<std::string> explicit_schemes = GetExplicitSchemes();
+  std::vector<URLPattern> result;
+
+  for (std::vector<std::string>::const_iterator i = explicit_schemes.begin();
+       i != explicit_schemes.end(); ++i) {
+    URLPattern temp = *this;
+    temp.SetScheme(*i);
+    temp.SetMatchAllURLs(false);
+    result.push_back(temp);
+  }
+
+  return result;
+}
+
+// static
+const char* URLPattern::GetParseResultString(
+    URLPattern::ParseResult parse_result) {
+  return kParseResultMessages[parse_result];
+}
+
+}  // namespace extensions

chromium_src/extensions/common/url_pattern.h

@@ -0,0 +1,264 @@
+// Copyright (c) 2012 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.
+#ifndef EXTENSIONS_COMMON_URL_PATTERN_H_
+#define EXTENSIONS_COMMON_URL_PATTERN_H_
+
+#include <functional>
+#include <iosfwd>
+#include <string>
+#include <vector>
+
+class GURL;
+
+namespace extensions {
+// A pattern that can be used to match URLs. A URLPattern is a very restricted
+// subset of URL syntax:
+//
+// <url-pattern> := <scheme>://<host><port><path> | '<all_urls>'
+// <scheme> := '*' | 'http' | 'https' | 'file' | 'ftp' | 'chrome' |
+//             'chrome-extension' | 'filesystem'
+// <host> := '*' | '*.' <anychar except '/' and '*'>+
+// <port> := [':' ('*' | <port number between 0 and 65535>)]
+// <path> := '/' <any chars>
+//
+// * Host is not used when the scheme is 'file'.
+// * The path can have embedded '*' characters which act as glob wildcards.
+// * '<all_urls>' is a special pattern that matches any URL that contains a
+//   valid scheme (as specified by valid_schemes_).
+// * The '*' scheme pattern excludes file URLs.
+//
+// Examples of valid patterns:
+// - http://*/*
+// - http://*/foo*
+// - https://*.google.com/foo*bar
+// - file://monkey*
+// - http://127.0.0.1/*
+//
+// Examples of invalid patterns:
+// - http://* -- path not specified
+// - http://*foo/bar -- * not allowed as substring of host component
+// - http://foo.*.bar/baz -- * must be first component
+// - http:/bar -- scheme separator not found
+// - foo://* -- invalid scheme
+// - chrome:// -- we don't support chrome internal URLs
+class URLPattern {
+ public:
+  // A collection of scheme bitmasks for use with valid_schemes.
+  enum SchemeMasks {
+    SCHEME_NONE       = 0,
+    SCHEME_HTTP       = 1 << 0,
+    SCHEME_HTTPS      = 1 << 1,
+    SCHEME_FILE       = 1 << 2,
+    SCHEME_FTP        = 1 << 3,
+    SCHEME_CHROMEUI   = 1 << 4,
+    SCHEME_EXTENSION  = 1 << 5,
+    SCHEME_FILESYSTEM = 1 << 6,
+
+    // IMPORTANT!
+    // SCHEME_ALL will match every scheme, including chrome://, chrome-
+    // extension://, about:, etc. Because this has lots of security
+    // implications, third-party extensions should usually not be able to get
+    // access to URL patterns initialized this way. If there is a reason
+    // for violating this general rule, document why this it safe.
+    SCHEME_ALL      = -1,
+  };
+
+  // Error codes returned from Parse().
+  enum ParseResult {
+    PARSE_SUCCESS = 0,
+    PARSE_ERROR_MISSING_SCHEME_SEPARATOR,
+    PARSE_ERROR_INVALID_SCHEME,
+    PARSE_ERROR_WRONG_SCHEME_SEPARATOR,
+    PARSE_ERROR_EMPTY_HOST,
+    PARSE_ERROR_INVALID_HOST_WILDCARD,
+    PARSE_ERROR_EMPTY_PATH,
+    PARSE_ERROR_INVALID_PORT,
+    PARSE_ERROR_INVALID_HOST,
+    NUM_PARSE_RESULTS
+  };
+
+  // The <all_urls> string pattern.
+  static const char kAllUrlsPattern[];
+
+  // Returns true if the given |scheme| is considered valid for extensions.
+  static bool IsValidSchemeForExtensions(const std::string& scheme);
+
+  explicit URLPattern(int valid_schemes);
+
+  // Convenience to construct a URLPattern from a string. If the string is not
+  // known ahead of time, use Parse() instead, which returns success or failure.
+  URLPattern(int valid_schemes, const std::string& pattern);
+
+  URLPattern();
+  ~URLPattern();
+
+  bool operator<(const URLPattern& other) const;
+  bool operator>(const URLPattern& other) const;
+  bool operator==(const URLPattern& other) const;
+
+  // Initializes this instance by parsing the provided string. Returns
+  // URLPattern::PARSE_SUCCESS on success, or an error code otherwise. On
+  // failure, this instance will have some intermediate values and is in an
+  // invalid state.
+  ParseResult Parse(const std::string& pattern_str);
+
+  // Gets the bitmask of valid schemes.
+  int valid_schemes() const { return valid_schemes_; }
+  void SetValidSchemes(int valid_schemes);
+
+  // Gets the host the pattern matches. This can be an empty string if the
+  // pattern matches all hosts (the input was <scheme>://*/<whatever>).
+  const std::string& host() const { return host_; }
+  void SetHost(const std::string& host);
+
+  // Gets whether to match subdomains of host().
+  bool match_subdomains() const { return match_subdomains_; }
+  void SetMatchSubdomains(bool val);
+
+  // Gets the path the pattern matches with the leading slash. This can have
+  // embedded asterisks which are interpreted using glob rules.
+  const std::string& path() const { return path_; }
+  void SetPath(const std::string& path);
+
+  // Returns true if this pattern matches all urls.
+  bool match_all_urls() const { return match_all_urls_; }
+  void SetMatchAllURLs(bool val);
+
+  // Sets the scheme for pattern matches. This can be a single '*' if the
+  // pattern matches all valid schemes (as defined by the valid_schemes_
+  // property). Returns false on failure (if the scheme is not valid).
+  bool SetScheme(const std::string& scheme);
+  // Note: You should use MatchesScheme() instead of this getter unless you
+  // absolutely need the exact scheme. This is exposed for testing.
+  const std::string& scheme() const { return scheme_; }
+
+  // Returns true if the specified scheme can be used in this URL pattern, and
+  // false otherwise. Uses valid_schemes_ to determine validity.
+  bool IsValidScheme(const std::string& scheme) const;
+
+  // Returns true if this instance matches the specified URL.
+  bool MatchesURL(const GURL& test) const;
+
+  // Returns true if this instance matches the specified security origin.
+  bool MatchesSecurityOrigin(const GURL& test) const;
+
+  // Returns true if |test| matches our scheme.
+  // Note that if test is "filesystem", this may fail whereas MatchesURL
+  // may succeed.  MatchesURL is smart enough to look at the inner_url instead
+  // of the outer "filesystem:" part.
+  bool MatchesScheme(const std::string& test) const;
+
+  // Returns true if |test| matches our host.
+  bool MatchesHost(const std::string& test) const;
+  bool MatchesHost(const GURL& test) const;
+
+  // Returns true if |test| matches our path.
+  bool MatchesPath(const std::string& test) const;
+
+  // Returns true if the pattern is vague enough that it implies all hosts,
+  // such as *://*/*.
+  // This is an expensive method, and should be used sparingly!
+  // You should probably use URLPatternSet::ShouldWarnAllHosts(), which is
+  // cached.
+  bool ImpliesAllHosts() const;
+
+  // Returns true if the pattern only matches a single origin. The pattern may
+  // include a path.
+  bool MatchesSingleOrigin() const;
+
+  // Sets the port. Returns false if the port is invalid.
+  bool SetPort(const std::string& port);
+  const std::string& port() const { return port_; }
+
+  // Returns a string representing this instance.
+  const std::string& GetAsString() const;
+
+  // Determines whether there is a URL that would match this instance and
+  // another instance. This method is symmetrical: Calling
+  // other.OverlapsWith(this) would result in the same answer.
+  bool OverlapsWith(const URLPattern& other) const;
+
+  // Returns true if this pattern matches all possible URLs that |other| can
+  // match. For example, http://*.google.com encompasses http://www.google.com.
+  bool Contains(const URLPattern& other) const;
+
+  // Converts this URLPattern into an equivalent set of URLPatterns that don't
+  // use a wildcard in the scheme component. If this URLPattern doesn't use a
+  // wildcard scheme, then the returned set will contain one element that is
+  // equivalent to this instance.
+  std::vector<URLPattern> ConvertToExplicitSchemes() const;
+
+  static bool EffectiveHostCompare(const URLPattern& a, const URLPattern& b) {
+    if (a.match_all_urls_ && b.match_all_urls_)
+      return false;
+    return a.host_.compare(b.host_) < 0;
+  }
+
+  // Used for origin comparisons in a std::set.
+  class EffectiveHostCompareFunctor {
+   public:
+    bool operator()(const URLPattern& a, const URLPattern& b) const {
+      return EffectiveHostCompare(a, b);
+    }
+  };
+
+  // Get an error string for a ParseResult.
+  static const char* GetParseResultString(URLPattern::ParseResult parse_result);
+
+ private:
+  // Returns true if any of the |schemes| items matches our scheme.
+  bool MatchesAnyScheme(const std::vector<std::string>& schemes) const;
+
+  // Returns true if all of the |schemes| items matches our scheme.
+  bool MatchesAllSchemes(const std::vector<std::string>& schemes) const;
+
+  bool MatchesSecurityOriginHelper(const GURL& test) const;
+
+  // Returns true if our port matches the |port| pattern (it may be "*").
+  bool MatchesPortPattern(const std::string& port) const;
+
+  // If the URLPattern contains a wildcard scheme, returns a list of
+  // equivalent literal schemes, otherwise returns the current scheme.
+  std::vector<std::string> GetExplicitSchemes() const;
+
+  // A bitmask containing the schemes which are considered valid for this
+  // pattern. Parse() uses this to decide whether a pattern contains a valid
+  // scheme.
+  int valid_schemes_;
+
+  // True if this is a special-case "<all_urls>" pattern.
+  bool match_all_urls_;
+
+  // The scheme for the pattern.
+  std::string scheme_;
+
+  // The host without any leading "*" components.
+  std::string host_;
+
+  // Whether we should match subdomains of the host. This is true if the first
+  // component of the pattern's host was "*".
+  bool match_subdomains_;
+
+  // The port.
+  std::string port_;
+
+  // The path to match. This is everything after the host of the URL, or
+  // everything after the scheme in the case of file:// URLs.
+  std::string path_;
+
+  // The path with "?" and "\" characters escaped for use with the
+  // MatchPattern() function.
+  std::string path_escaped_;
+
+  // A string representing this URLPattern.
+  mutable std::string spec_;
+};
+
+std::ostream& operator<<(std::ostream& out, const URLPattern& url_pattern);
+
+typedef std::vector<URLPattern> URLPatternList;
+
+}  // namespace extensions
+
+#endif  // EXTENSIONS_COMMON_URL_PATTERN_H_