url / gurl.h [blame]

// Copyright 2013 The Chromium Authors
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.

#ifndef URL_GURL_H_
#define URL_GURL_H_

#include <stddef.h>

#include <iosfwd>
#include <memory>
#include <string>
#include <string_view>

#include "base/component_export.h"
#include "base/debug/alias.h"
#include "base/debug/crash_logging.h"
#include "base/trace_event/base_tracing_forward.h"
#include "url/third_party/mozilla/url_parse.h"
#include "url/url_canon.h"
#include "url/url_canon_stdstring.h"
#include "url/url_constants.h"

// Represents a URL. GURL is Google's URL parsing library.
//
// A parsed canonicalized URL is guaranteed to be UTF-8. Any non-ASCII input
// characters are UTF-8 encoded and % escaped to ASCII.
//
// The string representation of a URL is called the spec(). Getting the
// spec will assert if the URL is invalid to help protect against malicious
// URLs. If you want the "best effort" canonicalization of an invalid URL, you
// can use possibly_invalid_spec(). Test validity with is_valid(). Data and
// javascript URLs use GetContent() to extract the data.
//
// This class has existence checkers and getters for the various components of
// a URL. Existence is different than being nonempty. "http://www.google.com/?"
// has a query that just happens to be empty, and has_query() will return true
// while the query getters will return the empty string.
//
// Prefer not to modify a URL using string operations (though sometimes this is
// unavoidable). Instead, use ReplaceComponents which can replace or delete
// multiple parts of a URL in one step, doesn't re-canonicalize unchanged
// sections, and avoids some screw-ups. An example is creating a URL with a
// path that contains a literal '#'. Using string concatenation will generate a
// URL with a truncated path and a reference fragment, while ReplaceComponents
// will know to escape this and produce the desired result.
//
// WARNING: While there is no length limit on GURLs, the Mojo serialization
// code will replace any very long URL with an invalid GURL.
// See url::mojom::kMaxURLChars for more details.
class COMPONENT_EXPORT(URL) GURL {
 public:
  using Replacements = url::StringViewReplacements<char>;
  using ReplacementsW = url::StringViewReplacements<char16_t>;

  // Creates an empty, invalid URL.
  GURL();

  // Copy construction is relatively inexpensive, with most of the time going
  // to reallocating the string. It does not re-parse.
  GURL(const GURL& other);
  GURL(GURL&& other) noexcept;

  // The strings to this constructor should be UTF-8 / UTF-16. They will be
  // parsed and canonicalized. For example, the host is lower cased, and
  // characters may be percent-encoded or percent-decoded to normalize the URL.
  explicit GURL(std::string_view url_string);
  explicit GURL(std::u16string_view url_string);

  // Constructor for URLs that have already been parsed and canonicalized. This
  // is used for conversions from KURL, for example. The caller must supply all
  // information associated with the URL, which must be correct and consistent.
  GURL(const char* canonical_spec,
       size_t canonical_spec_len,
       const url::Parsed& parsed,
       bool is_valid);
  // Notice that we take the canonical_spec by value so that we can convert
  // from WebURL without copying the string. When we call this constructor
  // we pass in a temporary std::string, which lets the compiler skip the
  // copy and just move the std::string into the function argument. In the
  // implementation, we use std::move to move the data into the GURL itself,
  // which means we end up with zero copies.
  GURL(std::string canonical_spec, const url::Parsed& parsed, bool is_valid);

  ~GURL();

  GURL& operator=(const GURL& other);
  GURL& operator=(GURL&& other) noexcept;

  // Returns true when this object represents a valid parsed URL. When not
  // valid, other functions will still succeed, but you will not get canonical
  // data out in the format you may be expecting. Instead, we keep something
  // "reasonable looking" so that the user can see how it's busted if
  // displayed to them.
  bool is_valid() const {
    return is_valid_;
  }

  // Returns true if the URL is zero-length. Note that empty URLs are also
  // invalid, and is_valid() will return false for them. This is provided
  // because some users may want to treat the empty case differently.
  bool is_empty() const {
    return spec_.empty();
  }

  // Returns the raw spec, i.e., the full text of the URL, in canonical UTF-8,
  // if the URL is valid. If the URL is not valid, this will assert and return
  // the empty string (for safety in release builds, to keep them from being
  // misused which might be a security problem).
  //
  // The URL will be ASCII (non-ASCII characters will be %-escaped UTF-8).
  //
  // The exception is for empty() URLs (which are !is_valid()) but this will
  // return the empty string without asserting.
  //
  // Use invalid_spec() below to get the unusable spec of an invalid URL. This
  // separation is designed to prevent errors that may cause security problems
  // that could result from the mistaken use of an invalid URL.
  const std::string& spec() const;

  // Returns the potentially invalid spec for a the URL. This spec MUST NOT be
  // modified or sent over the network. It is designed to be displayed in error
  // messages to the user, as the appearance of the spec may explain the error.
  // If the spec is valid, the valid spec will be returned.
  //
  // The returned string is guaranteed to be valid UTF-8.
  const std::string& possibly_invalid_spec() const {
    return spec_;
  }

  // Getter for the raw parsed structure. This allows callers to locate parts
  // of the URL within the spec themselves. Most callers should consider using
  // the individual component getters below.
  //
  // The returned parsed structure will reference into the raw spec, which may
  // or may not be valid. If you are using this to index into the spec, BE
  // SURE YOU ARE USING possibly_invalid_spec() to get the spec, and that you
  // don't do anything "important" with invalid specs.
  const url::Parsed& parsed_for_possibly_invalid_spec() const {
    return parsed_;
  }

  // Allows GURL to used as a key in STL (for example, a std::set or std::map).
  constexpr friend auto operator<=>(const GURL& lhs, const GURL& rhs) {
    return lhs.spec_ <=> rhs.spec_;
  }

  // Resolves a URL that's possibly relative to this object's URL, and returns
  // it. Absolute URLs are also handled according to the rules of URLs on web
  // pages.
  //
  // It may be impossible to resolve the URLs properly. If the input is not
  // "standard" (IsStandard() == false) and the input looks relative, we can't
  // resolve it. In these cases, the result will be an empty, invalid GURL.
  //
  // The result may also be a nonempty, invalid URL if the input has some kind
  // of encoding error. In these cases, we will try to construct a "good" URL
  // that may have meaning to the user, but it will be marked invalid.
  //
  // It is an error to resolve a URL relative to an invalid URL. The result
  // will be the empty URL.
  [[nodiscard]] GURL Resolve(std::string_view relative) const;
  [[nodiscard]] GURL Resolve(std::u16string_view relative) const;

  // Creates a new GURL by replacing the current URL's components with the
  // supplied versions. See the Replacements class in url_canon.h for more.
  //
  // These are not particularly quick, so avoid doing mutations when possible.
  // Prefer the 8-bit version when possible.
  //
  // It is an error to replace components of an invalid URL. The result will
  // be the empty URL.
  //
  // Note that this intentionally disallows direct use of url::Replacements,
  // which is harder to use correctly.
  [[nodiscard]] GURL ReplaceComponents(const Replacements& replacements) const;
  [[nodiscard]] GURL ReplaceComponents(const ReplacementsW& replacements) const;

  // A helper function that is equivalent to replacing the path with a slash
  // and clearing out everything after that. We sometimes need to know just the
  // scheme and the authority. If this URL is not a standard URL (it doesn't
  // have the regular authority and path sections), then the result will be
  // an empty, invalid GURL. Note that this *does* work for file: URLs, which
  // some callers may want to filter out before calling this.
  //
  // It is an error to get an empty path on an invalid URL. The result
  // will be the empty URL.
  [[nodiscard]] GURL GetWithEmptyPath() const;

  // A helper function to return a GURL without the filename, query values, and
  // fragment. For example,
  // GURL("https://www.foo.com/index.html?q=test").GetWithoutFilename().spec()
  // will return "https://www.foo.com/".
  // GURL("https://www.foo.com/bar/").GetWithoutFilename().spec()
  // will return "https://www.foo.com/bar/". If the GURL is invalid or missing a
  // scheme, authority or path, it will return an empty, invalid GURL.
  [[nodiscard]] GURL GetWithoutFilename() const;

  // A helper function to return a GURL without the Ref (also named Fragment
  // Identifier). For example,
  // GURL("https://www.foo.com/index.html#test").GetWithoutRef().spec()
  // will return "https://www.foo.com/index.html".
  // If the GURL is invalid or missing a
  // scheme, authority or path, it will return an empty, invalid GURL.
  [[nodiscard]] GURL GetWithoutRef() const;

  // A helper function to return a GURL containing just the scheme, host,
  // and port from a URL. Equivalent to clearing any username and password,
  // replacing the path with a slash, and clearing everything after that. If
  // this URL is not a standard URL, then the result will be an empty,
  // invalid GURL. If the URL has neither username nor password, this
  // degenerates to GetWithEmptyPath().
  //
  // It is an error to get the origin of an invalid URL. The result
  // will be the empty URL.
  //
  // WARNING: Please avoid converting urls into origins if at all possible!
  // //docs/security/origin-vs-url.md is a list of gotchas that can result. Such
  // conversions will likely return a wrong result for about:blank and/or
  // in the presence of iframe.sandbox attribute. Prefer to get origins directly
  // from the source (e.g. RenderFrameHost::GetLastCommittedOrigin).
  [[nodiscard]] GURL DeprecatedGetOriginAsURL() const;

  // A helper function to return a GURL stripped from the elements that are not
  // supposed to be sent as HTTP referrer: username, password and ref fragment.
  // For invalid URLs or URLs that no valid referrers, an empty URL will be
  // returned.
  [[nodiscard]] GURL GetAsReferrer() const;

  // Returns true if the scheme for the current URL is a known "standard-format"
  // scheme. A standard-format scheme adheres to what RFC 3986 calls "generic
  // URI syntax" (https://tools.ietf.org/html/rfc3986#section-3). This includes
  // file: and filesystem:, which some callers may want to filter out explicitly
  // by calling SchemeIsFile[System].
  bool IsStandard() const;

  // Returns true when the url is of the form about:blank, about:blank?foo or
  // about:blank/#foo.
  bool IsAboutBlank() const;

  // Returns true when the url is of the form about:srcdoc, about:srcdoc?foo or
  // about:srcdoc/#foo.
  bool IsAboutSrcdoc() const;

  // Returns true if the given parameter (should be lower-case ASCII to match
  // the canonicalized scheme) is the scheme for this URL. Do not include a
  // colon.
  bool SchemeIs(std::string_view lower_ascii_scheme) const;

  // Returns true if the scheme is "http" or "https".
  bool SchemeIsHTTPOrHTTPS() const;

  // Returns true is the scheme is "ws" or "wss".
  bool SchemeIsWSOrWSS() const;

  // We often need to know if this is a file URL. File URLs are "standard", but
  // are often treated separately by some programs.
  bool SchemeIsFile() const {
    return SchemeIs(url::kFileScheme);
  }

  // FileSystem URLs need to be treated differently in some cases.
  bool SchemeIsFileSystem() const {
    return SchemeIs(url::kFileSystemScheme);
  }

  // Returns true if the scheme indicates a network connection that uses TLS or
  // some other cryptographic protocol (e.g. QUIC) for security.
  //
  // This function is a not a complete test of whether or not an origin's code
  // is minimally trustworthy. For that, see Chromium's |IsOriginSecure| for a
  // higher-level and more complete semantics. See that function's documentation
  // for more detail.
  bool SchemeIsCryptographic() const;

  // As above, but static. Parameter should be lower-case ASCII.
  static bool SchemeIsCryptographic(std::string_view lower_ascii_scheme);

  // Returns true if the scheme is "blob".
  bool SchemeIsBlob() const {
    return SchemeIs(url::kBlobScheme);
  }

  // Returns true if the scheme is a local scheme, as defined in Fetch:
  // https://fetch.spec.whatwg.org/#local-scheme
  bool SchemeIsLocal() const;

  // For most URLs, the "content" is everything after the scheme (skipping the
  // scheme delimiting colon) and before the fragment (skipping the fragment
  // delimiting octothorpe). For javascript URLs the "content" also includes the
  // fragment delimiter and fragment.
  //
  // It is an error to get the content of an invalid URL: the result will be an
  // empty string.
  //
  // Important note: The feature flag,
  // url::kStandardCompliantNonSpecialSchemeURLParsing, changes the behavior of
  // GetContent() and GetContentPiece() for some non-special URLs. See
  // GURLTest::ContentForNonStandardURLs for the differences.
  //
  // Until the flag becomes enabled by default, you'll need to manually check
  // the flag when using GetContent() and GetContentPiece() for non-special
  // URLs. See http://crbug.com/40063064 for more details.
  std::string GetContent() const;
  std::string_view GetContentPiece() const;

  // Returns true if the hostname is an IP address. Note: this function isn't
  // as cheap as a simple getter because it re-parses the hostname to verify.
  bool HostIsIPAddress() const;

  // Not including the colon. If you are comparing schemes, prefer SchemeIs.
  bool has_scheme() const { return parsed_.scheme.is_valid(); }
  std::string scheme() const {
    return ComponentString(parsed_.scheme);
  }
  std::string_view scheme_piece() const {
    return ComponentStringPiece(parsed_.scheme);
  }

  bool has_username() const { return parsed_.username.is_valid(); }
  std::string username() const {
    return ComponentString(parsed_.username);
  }
  std::string_view username_piece() const {
    return ComponentStringPiece(parsed_.username);
  }

  bool has_password() const { return parsed_.password.is_valid(); }
  std::string password() const {
    return ComponentString(parsed_.password);
  }
  std::string_view password_piece() const {
    return ComponentStringPiece(parsed_.password);
  }

  // The host may be a hostname, an IPv4 address, or an IPv6 literal surrounded
  // by square brackets, like "[2001:db8::1]". To exclude these brackets, use
  // HostNoBrackets() below.
  bool has_host() const {
    // Note that hosts are special, absence of host means length 0.
    return parsed_.host.is_nonempty();
  }
  std::string host() const {
    return ComponentString(parsed_.host);
  }
  std::string_view host_piece() const {
    return ComponentStringPiece(parsed_.host);
  }

  // The port if one is explicitly specified. Most callers will want IntPort()
  // or EffectiveIntPort() instead of these. The getters will not include the
  // ':'.
  bool has_port() const { return parsed_.port.is_valid(); }
  std::string port() const {
    return ComponentString(parsed_.port);
  }
  std::string_view port_piece() const {
    return ComponentStringPiece(parsed_.port);
  }

  // Including first slash following host, up to the query. The URL
  // "http://www.google.com/" has a path of "/".
  bool has_path() const { return parsed_.path.is_valid(); }
  std::string path() const {
    return ComponentString(parsed_.path);
  }
  std::string_view path_piece() const {
    return ComponentStringPiece(parsed_.path);
  }

  // Stuff following '?' up to the ref. The getters will not include the '?'.
  bool has_query() const { return parsed_.query.is_valid(); }
  std::string query() const {
    return ComponentString(parsed_.query);
  }
  std::string_view query_piece() const {
    return ComponentStringPiece(parsed_.query);
  }

  // Stuff following '#' to the end of the string. This will be %-escaped UTF-8.
  // The getters will not include the '#'.
  bool has_ref() const { return parsed_.ref.is_valid(); }
  std::string ref() const {
    return ComponentString(parsed_.ref);
  }
  std::string_view ref_piece() const {
    return ComponentStringPiece(parsed_.ref);
  }

  // Returns a parsed version of the port. Can also be any of the special
  // values defined in Parsed for ExtractPort.
  int IntPort() const;

  // Returns the port number of the URL, or the default port number.
  // If the scheme has no concept of port (or unknown default) returns
  // PORT_UNSPECIFIED.
  int EffectiveIntPort() const;

  // Extracts the filename portion of the path and returns it. The filename
  // is everything after the last slash in the path. This may be empty.
  std::string ExtractFileName() const;

  // Returns the path that should be sent to the server. This is the path,
  // parameter, and query portions of the URL. It is guaranteed to be ASCII.
  std::string PathForRequest() const;

  // Returns the same characters as PathForRequest(), avoiding a copy.
  std::string_view PathForRequestPiece() const;

  // Returns the host, excluding the square brackets surrounding IPv6 address
  // literals. This can be useful for passing to getaddrinfo().
  std::string HostNoBrackets() const;

  // Returns the same characters as HostNoBrackets(), avoiding a copy.
  std::string_view HostNoBracketsPiece() const;

  // Returns true if this URL's host matches or is in the same domain as
  // the given input string. For example, if the hostname of the URL is
  // "www.google.com", this will return true for "com", "google.com", and
  // "www.google.com".
  //
  // The input domain should match host canonicalization rules. i.e. the input
  // should be lowercase except for escape chars.
  //
  // This call is more efficient than getting the host and checking whether the
  // host has the specific domain or not because no copies or object
  // constructions are done.
  bool DomainIs(std::string_view canonical_domain) const;

  // Checks whether or not two URLs differ only in the ref (the part after
  // the # character).
  bool EqualsIgnoringRef(const GURL& other) const;

  // Swaps the contents of this GURL object with |other|, without doing
  // any memory allocations.
  void Swap(GURL* other);

  // Returns a reference to a singleton empty GURL. This object is for callers
  // who return references but don't have anything to return in some cases.
  // If you just want an empty URL for normal use, prefer GURL(). This function
  // may be called from any thread.
  static const GURL& EmptyGURL();

  // Returns the inner URL of a nested URL (currently only non-null for
  // filesystem URLs).
  //
  // TODO(mmenke): inner_url().spec() currently returns the same value as
  // caling spec() on the GURL itself. This should be fixed.
  // See https://crbug.com/619596
  const GURL* inner_url() const {
    return inner_url_.get();
  }

  // Estimates dynamic memory usage.
  // See base/trace_event/memory_usage_estimator.h for more info.
  size_t EstimateMemoryUsage() const;

  // Helper used by GURL::IsAboutUrl and KURL::IsAboutURL.
  static bool IsAboutPath(std::string_view actual_path,
                          std::string_view allowed_path);

  void WriteIntoTrace(perfetto::TracedValue context) const;

 private:
  // Variant of the string parsing constructor that allows the caller to elect
  // retain trailing whitespace, if any, on the passed URL spec, but only if
  // the scheme is one that allows trailing whitespace. The primary use-case is
  // for data: URLs. In most cases, you want to use the single parameter
  // constructor above.
  enum RetainWhiteSpaceSelector { RETAIN_TRAILING_PATH_WHITEPACE };
  GURL(const std::string& url_string, RetainWhiteSpaceSelector);

  template <typename T, typename CharT = typename T::value_type>
  void InitCanonical(T input_spec, bool trim_path_end);

  void InitializeFromCanonicalSpec();

  // Helper used by IsAboutBlank and IsAboutSrcdoc.
  bool IsAboutUrl(std::string_view allowed_path) const;

  // Returns the substring of the input identified by the given component.
  std::string ComponentString(const url::Component& comp) const {
    return std::string(ComponentStringPiece(comp));
  }
  std::string_view ComponentStringPiece(const url::Component& comp) const {
    if (comp.is_empty())
      return std::string_view();
    return std::string_view(spec_).substr(static_cast<size_t>(comp.begin),
                                          static_cast<size_t>(comp.len));
  }

  void ProcessFileSystemURLAfterReplaceComponents();

  // The actual text of the URL, in canonical ASCII form.
  std::string spec_;

  // Set when the given URL is valid. Otherwise, we may still have a spec and
  // components, but they may not identify valid resources (for example, an
  // invalid port number, invalid characters in the scheme, etc.).
  bool is_valid_;

  // Identified components of the canonical spec.
  url::Parsed parsed_;

  // Used for nested schemes [currently only filesystem:].
  std::unique_ptr<GURL> inner_url_;
};

// Stream operator so GURL can be used in assertion statements.
COMPONENT_EXPORT(URL)
std::ostream& operator<<(std::ostream& out, const GURL& url);

COMPONENT_EXPORT(URL) bool operator==(const GURL& x, const GURL& y);

// Equality operator for comparing raw spec_. This should be used in place of
// url == GURL(spec) where |spec| is known (i.e. constants). This is to prevent
// needlessly re-parsing |spec| into a temporary GURL.
COMPONENT_EXPORT(URL)
bool operator==(const GURL& x, std::string_view spec);

// DEBUG_ALIAS_FOR_GURL(var_name, url) copies |url| into a new stack-allocated
// variable named |<var_name>|.  This helps ensure that the value of |url| gets
// preserved in crash dumps.
#define DEBUG_ALIAS_FOR_GURL(var_name, url) \
  DEBUG_ALIAS_FOR_CSTR(var_name, (url).possibly_invalid_spec().c_str(), 128)

namespace url::debug {

class COMPONENT_EXPORT(URL) ScopedUrlCrashKey {
 public:
  ScopedUrlCrashKey(base::debug::CrashKeyString* crash_key, const GURL& value);
  ~ScopedUrlCrashKey();

  ScopedUrlCrashKey(const ScopedUrlCrashKey&) = delete;
  ScopedUrlCrashKey& operator=(const ScopedUrlCrashKey&) = delete;

 private:
  base::debug::ScopedCrashKeyString scoped_string_value_;
};

}  // namespace url::debug

#endif  // URL_GURL_H_