michael@0: // Copyright (c) 2008 The Chromium Authors. All rights reserved.
michael@0: // Use of this source code is governed by a BSD-style license that can be
michael@0: // found in the LICENSE file.
michael@0: 
michael@0: #include <fstream>
michael@0: 
michael@0: #include "base/file_path.h"
michael@0: #include "base/logging.h"
michael@0: 
michael@0: // These includes are just for the *Hack functions, and should be removed
michael@0: // when those functions are removed.
michael@0: #include "base/string_piece.h"
michael@0: #include "base/string_util.h"
michael@0: #include "base/sys_string_conversions.h"
michael@0: 
michael@0: #if defined(FILE_PATH_USES_WIN_SEPARATORS)
michael@0: const FilePath::CharType FilePath::kSeparators[] = FILE_PATH_LITERAL("\\/");
michael@0: #else  // FILE_PATH_USES_WIN_SEPARATORS
michael@0: const FilePath::CharType FilePath::kSeparators[] = FILE_PATH_LITERAL("/");
michael@0: #endif  // FILE_PATH_USES_WIN_SEPARATORS
michael@0: 
michael@0: const FilePath::CharType FilePath::kCurrentDirectory[] = FILE_PATH_LITERAL(".");
michael@0: const FilePath::CharType FilePath::kParentDirectory[] = FILE_PATH_LITERAL("..");
michael@0: 
michael@0: const FilePath::CharType FilePath::kExtensionSeparator = FILE_PATH_LITERAL('.');
michael@0: 
michael@0: 
michael@0: namespace {
michael@0: 
michael@0: // If this FilePath contains a drive letter specification, returns the
michael@0: // position of the last character of the drive letter specification,
michael@0: // otherwise returns npos.  This can only be true on Windows, when a pathname
michael@0: // begins with a letter followed by a colon.  On other platforms, this always
michael@0: // returns npos.
michael@0: FilePath::StringType::size_type FindDriveLetter(
michael@0:     const FilePath::StringType& path) {
michael@0: #if defined(FILE_PATH_USES_DRIVE_LETTERS)
michael@0:   // This is dependent on an ASCII-based character set, but that's a
michael@0:   // reasonable assumption.  iswalpha can be too inclusive here.
michael@0:   if (path.length() >= 2 && path[1] == L':' &&
michael@0:       ((path[0] >= L'A' && path[0] <= L'Z') ||
michael@0:        (path[0] >= L'a' && path[0] <= L'z'))) {
michael@0:     return 1;
michael@0:   }
michael@0: #endif  // FILE_PATH_USES_DRIVE_LETTERS
michael@0:   return FilePath::StringType::npos;
michael@0: }
michael@0: 
michael@0: bool IsPathAbsolute(const FilePath::StringType& path) {
michael@0: #if defined(FILE_PATH_USES_DRIVE_LETTERS)
michael@0:   FilePath::StringType::size_type letter = FindDriveLetter(path);
michael@0:   if (letter != FilePath::StringType::npos) {
michael@0:     // Look for a separator right after the drive specification.
michael@0:     return path.length() > letter + 1 &&
michael@0:         FilePath::IsSeparator(path[letter + 1]);
michael@0:   }
michael@0:   // Look for a pair of leading separators.
michael@0:   return path.length() > 1 &&
michael@0:       FilePath::IsSeparator(path[0]) && FilePath::IsSeparator(path[1]);
michael@0: #else  // FILE_PATH_USES_DRIVE_LETTERS
michael@0:   // Look for a separator in the first position.
michael@0:   return path.length() > 0 && FilePath::IsSeparator(path[0]);
michael@0: #endif  // FILE_PATH_USES_DRIVE_LETTERS
michael@0: }
michael@0: 
michael@0: }  // namespace
michael@0: 
michael@0: bool FilePath::IsSeparator(CharType character) {
michael@0:   for (size_t i = 0; i < arraysize(kSeparators) - 1; ++i) {
michael@0:     if (character == kSeparators[i]) {
michael@0:       return true;
michael@0:     }
michael@0:   }
michael@0: 
michael@0:   return false;
michael@0: }
michael@0: 
michael@0: // libgen's dirname and basename aren't guaranteed to be thread-safe and aren't
michael@0: // guaranteed to not modify their input strings, and in fact are implemented
michael@0: // differently in this regard on different platforms.  Don't use them, but
michael@0: // adhere to their behavior.
michael@0: FilePath FilePath::DirName() const {
michael@0:   FilePath new_path(path_);
michael@0:   new_path.StripTrailingSeparatorsInternal();
michael@0: 
michael@0:   // The drive letter, if any, always needs to remain in the output.  If there
michael@0:   // is no drive letter, as will always be the case on platforms which do not
michael@0:   // support drive letters, letter will be npos, or -1, so the comparisons and
michael@0:   // resizes below using letter will still be valid.
michael@0:   StringType::size_type letter = FindDriveLetter(new_path.path_);
michael@0: 
michael@0:   StringType::size_type last_separator =
michael@0:       new_path.path_.find_last_of(kSeparators, StringType::npos,
michael@0:                                   arraysize(kSeparators) - 1);
michael@0:   if (last_separator == StringType::npos) {
michael@0:     // path_ is in the current directory.
michael@0:     new_path.path_.resize(letter + 1);
michael@0:   } else if (last_separator == letter + 1) {
michael@0:     // path_ is in the root directory.
michael@0:     new_path.path_.resize(letter + 2);
michael@0:   } else if (last_separator == letter + 2 &&
michael@0:              IsSeparator(new_path.path_[letter + 1])) {
michael@0:     // path_ is in "//" (possibly with a drive letter); leave the double
michael@0:     // separator intact indicating alternate root.
michael@0:     new_path.path_.resize(letter + 3);
michael@0:   } else if (last_separator != 0) {
michael@0:     // path_ is somewhere else, trim the basename.
michael@0:     new_path.path_.resize(last_separator);
michael@0:   }
michael@0: 
michael@0:   new_path.StripTrailingSeparatorsInternal();
michael@0:   if (!new_path.path_.length())
michael@0:     new_path.path_ = kCurrentDirectory;
michael@0: 
michael@0:   return new_path;
michael@0: }
michael@0: 
michael@0: FilePath FilePath::BaseName() const {
michael@0:   FilePath new_path(path_);
michael@0:   new_path.StripTrailingSeparatorsInternal();
michael@0: 
michael@0:   // The drive letter, if any, is always stripped.
michael@0:   StringType::size_type letter = FindDriveLetter(new_path.path_);
michael@0:   if (letter != StringType::npos) {
michael@0:     new_path.path_.erase(0, letter + 1);
michael@0:   }
michael@0: 
michael@0:   // Keep everything after the final separator, but if the pathname is only
michael@0:   // one character and it's a separator, leave it alone.
michael@0:   StringType::size_type last_separator =
michael@0:       new_path.path_.find_last_of(kSeparators, StringType::npos,
michael@0:                                   arraysize(kSeparators) - 1);
michael@0:   if (last_separator != StringType::npos &&
michael@0:       last_separator < new_path.path_.length() - 1) {
michael@0:     new_path.path_.erase(0, last_separator + 1);
michael@0:   }
michael@0: 
michael@0:   return new_path;
michael@0: }
michael@0: 
michael@0: FilePath::StringType FilePath::Extension() const {
michael@0:   // BaseName() calls StripTrailingSeparators, so cases like /foo.baz/// work.
michael@0:   StringType base = BaseName().value();
michael@0: 
michael@0:   // Special case "." and ".."
michael@0:   if (base == kCurrentDirectory || base == kParentDirectory)
michael@0:     return StringType();
michael@0: 
michael@0:   const StringType::size_type last_dot = base.rfind(kExtensionSeparator);
michael@0:   if (last_dot == StringType::npos)
michael@0:     return StringType();
michael@0:   return StringType(base, last_dot);
michael@0: }
michael@0: 
michael@0: FilePath FilePath::RemoveExtension() const {
michael@0:   StringType ext = Extension();
michael@0:   // It's important to check Extension() since that verifies that the
michael@0:   // kExtensionSeparator actually appeared in the last path component.
michael@0:   if (ext.empty())
michael@0:     return FilePath(path_);
michael@0:   // Since Extension() verified that the extension is in fact in the last path
michael@0:   // component, this substr will effectively strip trailing separators.
michael@0:   const StringType::size_type last_dot = path_.rfind(kExtensionSeparator);
michael@0:   return FilePath(path_.substr(0, last_dot));
michael@0: }
michael@0: 
michael@0: FilePath FilePath::InsertBeforeExtension(const StringType& suffix) const {
michael@0:   if (suffix.empty())
michael@0:     return FilePath(path_);
michael@0: 
michael@0:   if (path_.empty())
michael@0:     return FilePath();
michael@0: 
michael@0:   StringType base = BaseName().value();
michael@0:   if (base.empty())
michael@0:     return FilePath();
michael@0:   if (*(base.end() - 1) == kExtensionSeparator) {
michael@0:     // Special case "." and ".."
michael@0:     if (base == kCurrentDirectory || base == kParentDirectory) {
michael@0:       return FilePath();
michael@0:     }
michael@0:   }
michael@0: 
michael@0:   StringType ext = Extension();
michael@0:   StringType ret = RemoveExtension().value();
michael@0:   ret.append(suffix);
michael@0:   ret.append(ext);
michael@0:   return FilePath(ret);
michael@0: }
michael@0: 
michael@0: FilePath FilePath::ReplaceExtension(const StringType& extension) const {
michael@0:   if (path_.empty())
michael@0:     return FilePath();
michael@0: 
michael@0:   StringType base = BaseName().value();
michael@0:   if (base.empty())
michael@0:     return FilePath();
michael@0:   if (*(base.end() - 1) == kExtensionSeparator) {
michael@0:     // Special case "." and ".."
michael@0:     if (base == kCurrentDirectory || base == kParentDirectory) {
michael@0:       return FilePath();
michael@0:     }
michael@0:   }
michael@0: 
michael@0:   FilePath no_ext = RemoveExtension();
michael@0:   // If the new extension is "" or ".", then just remove the current extension.
michael@0:   if (extension.empty() || extension == StringType(1, kExtensionSeparator))
michael@0:     return no_ext;
michael@0: 
michael@0:   StringType str = no_ext.value();
michael@0:   if (extension[0] != kExtensionSeparator)
michael@0:     str.append(1, kExtensionSeparator);
michael@0:   str.append(extension);
michael@0:   return FilePath(str);
michael@0: }
michael@0: 
michael@0: FilePath FilePath::Append(const StringType& component) const {
michael@0:   DCHECK(!IsPathAbsolute(component));
michael@0:   if (path_.compare(kCurrentDirectory) == 0) {
michael@0:     // Append normally doesn't do any normalization, but as a special case,
michael@0:     // when appending to kCurrentDirectory, just return a new path for the
michael@0:     // component argument.  Appending component to kCurrentDirectory would
michael@0:     // serve no purpose other than needlessly lengthening the path, and
michael@0:     // it's likely in practice to wind up with FilePath objects containing
michael@0:     // only kCurrentDirectory when calling DirName on a single relative path
michael@0:     // component.
michael@0:     return FilePath(component);
michael@0:   }
michael@0: 
michael@0:   FilePath new_path(path_);
michael@0:   new_path.StripTrailingSeparatorsInternal();
michael@0: 
michael@0:   // Don't append a separator if the path is empty (indicating the current
michael@0:   // directory) or if the path component is empty (indicating nothing to
michael@0:   // append).
michael@0:   if (component.length() > 0 && new_path.path_.length() > 0) {
michael@0: 
michael@0:     // Don't append a separator if the path still ends with a trailing
michael@0:     // separator after stripping (indicating the root directory).
michael@0:     if (!IsSeparator(new_path.path_[new_path.path_.length() - 1])) {
michael@0: 
michael@0:       // Don't append a separator if the path is just a drive letter.
michael@0:       if (FindDriveLetter(new_path.path_) + 1 != new_path.path_.length()) {
michael@0:         new_path.path_.append(1, kSeparators[0]);
michael@0:       }
michael@0:     }
michael@0:   }
michael@0: 
michael@0:   new_path.path_.append(component);
michael@0:   return new_path;
michael@0: }
michael@0: 
michael@0: FilePath FilePath::Append(const FilePath& component) const {
michael@0:   return Append(component.value());
michael@0: }
michael@0: 
michael@0: FilePath FilePath::AppendASCII(const std::string& component) const {
michael@0:   DCHECK(IsStringASCII(component));
michael@0: #if defined(OS_WIN)
michael@0:   return Append(ASCIIToWide(component));
michael@0: #elif defined(OS_POSIX)
michael@0:   return Append(component);
michael@0: #endif
michael@0: }
michael@0: 
michael@0: bool FilePath::IsAbsolute() const {
michael@0:   return IsPathAbsolute(path_);
michael@0: }
michael@0: 
michael@0: #if defined(OS_POSIX)
michael@0: // See file_path.h for a discussion of the encoding of paths on POSIX
michael@0: // platforms.  These *Hack() functions are not quite correct, but they're
michael@0: // only temporary while we fix the remainder of the code.
michael@0: // Remember to remove the #includes at the top when you remove these.
michael@0: 
michael@0: // static
michael@0: FilePath FilePath::FromWStringHack(const std::wstring& wstring) {
michael@0:   return FilePath(base::SysWideToNativeMB(wstring));
michael@0: }
michael@0: std::wstring FilePath::ToWStringHack() const {
michael@0:   return base::SysNativeMBToWide(path_);
michael@0: }
michael@0: #elif defined(OS_WIN)
michael@0: // static
michael@0: FilePath FilePath::FromWStringHack(const std::wstring& wstring) {
michael@0:   return FilePath(wstring);
michael@0: }
michael@0: std::wstring FilePath::ToWStringHack() const {
michael@0:   return path_;
michael@0: }
michael@0: #endif
michael@0: 
michael@0: void FilePath::OpenInputStream(std::ifstream& stream) const {
michael@0:   stream.open(
michael@0: #ifndef __MINGW32__
michael@0:               path_.c_str(),
michael@0: #else
michael@0:               base::SysWideToNativeMB(path_).c_str(),
michael@0: #endif
michael@0:               std::ios::in | std::ios::binary);
michael@0: }
michael@0: 
michael@0: FilePath FilePath::StripTrailingSeparators() const {
michael@0:   FilePath new_path(path_);
michael@0:   new_path.StripTrailingSeparatorsInternal();
michael@0: 
michael@0:   return new_path;
michael@0: }
michael@0: 
michael@0: void FilePath::StripTrailingSeparatorsInternal() {
michael@0:   // If there is no drive letter, start will be 1, which will prevent stripping
michael@0:   // the leading separator if there is only one separator.  If there is a drive
michael@0:   // letter, start will be set appropriately to prevent stripping the first
michael@0:   // separator following the drive letter, if a separator immediately follows
michael@0:   // the drive letter.
michael@0:   StringType::size_type start = FindDriveLetter(path_) + 2;
michael@0: 
michael@0:   StringType::size_type last_stripped = StringType::npos;
michael@0:   for (StringType::size_type pos = path_.length();
michael@0:        pos > start && IsSeparator(path_[pos - 1]);
michael@0:        --pos) {
michael@0:     // If the string only has two separators and they're at the beginning,
michael@0:     // don't strip them, unless the string began with more than two separators.
michael@0:     if (pos != start + 1 || last_stripped == start + 2 ||
michael@0:         !IsSeparator(path_[start - 1])) {
michael@0:       path_.resize(pos - 1);
michael@0:       last_stripped = pos;
michael@0:     }
michael@0:   }
michael@0: }