naiveproxy/url/url_canon_fileurl.cc

191 lines
7.2 KiB
C++
Raw Normal View History

2018-12-10 05:59:24 +03:00
// 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.
// Functions for canonicalizing "file:" URLs.
#include "base/strings/string_util.h"
#include "url/url_canon.h"
#include "url/url_canon_internal.h"
#include "url/url_file.h"
#include "url/url_parse_internal.h"
namespace url {
namespace {
#ifdef WIN32
// Given a pointer into the spec, this copies and canonicalizes the drive
// letter and colon to the output, if one is found. If there is not a drive
// spec, it won't do anything. The index of the next character in the input
// spec is returned (after the colon when a drive spec is found, the begin
// offset if one is not).
template<typename CHAR>
int FileDoDriveSpec(const CHAR* spec, int begin, int end,
CanonOutput* output) {
// The path could be one of several things: /foo/bar, c:/foo/bar, /c:/foo,
// (with backslashes instead of slashes as well).
int num_slashes = CountConsecutiveSlashes(spec, begin, end);
int after_slashes = begin + num_slashes;
if (!DoesBeginWindowsDriveSpec(spec, after_slashes, end))
return begin; // Haven't consumed any characters
// A drive spec is the start of a path, so we need to add a slash for the
// authority terminator (typically the third slash).
output->push_back('/');
// DoesBeginWindowsDriveSpec will ensure that the drive letter is valid
// and that it is followed by a colon/pipe.
// Normalize Windows drive letters to uppercase
if (base::IsAsciiLower(spec[after_slashes]))
output->push_back(static_cast<char>(spec[after_slashes] - 'a' + 'A'));
else
output->push_back(static_cast<char>(spec[after_slashes]));
// Normalize the character following it to a colon rather than pipe.
output->push_back(':');
return after_slashes + 2;
}
#endif // WIN32
template<typename CHAR, typename UCHAR>
bool DoFileCanonicalizePath(const CHAR* spec,
const Component& path,
CanonOutput* output,
Component* out_path) {
// Copies and normalizes the "c:" at the beginning, if present.
out_path->begin = output->length();
int after_drive;
#ifdef WIN32
after_drive = FileDoDriveSpec(spec, path.begin, path.end(), output);
#else
after_drive = path.begin;
#endif
// Copies the rest of the path, starting from the slash following the
// drive colon (if any, Windows only), or the first slash of the path.
bool success = true;
if (after_drive < path.end()) {
// Use the regular path canonicalizer to canonicalize the rest of the
// path. Give it a fake output component to write into. DoCanonicalizeFile
// will compute the full path component.
Component sub_path = MakeRange(after_drive, path.end());
Component fake_output_path;
success = CanonicalizePath(spec, sub_path, output, &fake_output_path);
} else {
// No input path, canonicalize to a slash.
output->push_back('/');
}
out_path->len = output->length() - out_path->begin;
return success;
}
template<typename CHAR, typename UCHAR>
bool DoCanonicalizeFileURL(const URLComponentSource<CHAR>& source,
const Parsed& parsed,
CharsetConverter* query_converter,
CanonOutput* output,
Parsed* new_parsed) {
// Things we don't set in file: URLs.
new_parsed->username = Component();
new_parsed->password = Component();
new_parsed->port = Component();
// Scheme (known, so we don't bother running it through the more
// complicated scheme canonicalizer).
new_parsed->scheme.begin = output->length();
output->Append("file://", 7);
new_parsed->scheme.len = 4;
// Append the host. For many file URLs, this will be empty. For UNC, this
// will be present.
// TODO(brettw) This doesn't do any checking for host name validity. We
// should probably handle validity checking of UNC hosts differently than
// for regular IP hosts.
bool success = CanonicalizeHost(source.host, parsed.host,
output, &new_parsed->host);
success &= DoFileCanonicalizePath<CHAR, UCHAR>(source.path, parsed.path,
output, &new_parsed->path);
CanonicalizeQuery(source.query, parsed.query, query_converter,
output, &new_parsed->query);
// Ignore failure for refs since the URL can probably still be loaded.
CanonicalizeRef(source.ref, parsed.ref, output, &new_parsed->ref);
return success;
}
} // namespace
bool CanonicalizeFileURL(const char* spec,
int spec_len,
const Parsed& parsed,
CharsetConverter* query_converter,
CanonOutput* output,
Parsed* new_parsed) {
return DoCanonicalizeFileURL<char, unsigned char>(
URLComponentSource<char>(spec), parsed, query_converter,
output, new_parsed);
}
bool CanonicalizeFileURL(const base::char16* spec,
int spec_len,
const Parsed& parsed,
CharsetConverter* query_converter,
CanonOutput* output,
Parsed* new_parsed) {
return DoCanonicalizeFileURL<base::char16, base::char16>(
URLComponentSource<base::char16>(spec), parsed, query_converter,
output, new_parsed);
}
bool FileCanonicalizePath(const char* spec,
const Component& path,
CanonOutput* output,
Component* out_path) {
return DoFileCanonicalizePath<char, unsigned char>(spec, path,
output, out_path);
}
bool FileCanonicalizePath(const base::char16* spec,
const Component& path,
CanonOutput* output,
Component* out_path) {
return DoFileCanonicalizePath<base::char16, base::char16>(spec, path,
output, out_path);
}
bool ReplaceFileURL(const char* base,
const Parsed& base_parsed,
const Replacements<char>& replacements,
CharsetConverter* query_converter,
CanonOutput* output,
Parsed* new_parsed) {
URLComponentSource<char> source(base);
Parsed parsed(base_parsed);
SetupOverrideComponents(base, replacements, &source, &parsed);
return DoCanonicalizeFileURL<char, unsigned char>(
source, parsed, query_converter, output, new_parsed);
}
bool ReplaceFileURL(const char* base,
const Parsed& base_parsed,
const Replacements<base::char16>& replacements,
CharsetConverter* query_converter,
CanonOutput* output,
Parsed* new_parsed) {
RawCanonOutput<1024> utf8;
URLComponentSource<char> source(base);
Parsed parsed(base_parsed);
SetupUTF16OverrideComponents(base, replacements, &utf8, &source, &parsed);
return DoCanonicalizeFileURL<char, unsigned char>(
source, parsed, query_converter, output, new_parsed);
}
} // namespace url