1 //===-- PathV2.cpp - Implement OS Path Concept ------------------*- C++ -*-===//
3 // The LLVM Compiler Infrastructure
5 // This file is distributed under the University of Illinois Open Source
6 // License. See LICENSE.TXT for details.
8 //===----------------------------------------------------------------------===//
10 // This file implements the operating system PathV2 API.
12 //===----------------------------------------------------------------------===//
14 #include "llvm/Support/PathV2.h"
15 #include "llvm/Support/FileSystem.h"
16 #include "llvm/Support/Endian.h"
17 #include "llvm/Support/ErrorHandling.h"
23 using llvm::StringRef;
24 using llvm::sys::path::is_separator;
27 const char *separators = "\\/";
28 const char prefered_separator = '\\';
30 const char separators = '/';
31 const char prefered_separator = '/';
34 StringRef find_first_component(StringRef path) {
35 // Look for this first component in the following order.
36 // * empty (in this case we return an empty string)
37 // * either C: or {//,\\}net.
40 // * {file,directory}name
47 if (path.size() >= 2 && std::isalpha(path[0]) && path[1] == ':')
48 return path.substr(0, 2);
52 if ((path.size() > 2) &&
53 is_separator(path[0]) &&
55 !is_separator(path[2])) {
56 // Find the next directory separator.
57 size_t end = path.find_first_of(separators, 2);
58 return path.substr(0, end);
62 if (is_separator(path[0]))
63 return path.substr(0, 1);
65 if (path.startswith(".."))
66 return path.substr(0, 2);
69 return path.substr(0, 1);
71 // * {file,directory}name
72 size_t end = path.find_first_of(separators, 2);
73 return path.substr(0, end);
76 size_t filename_pos(StringRef str) {
77 if (str.size() == 2 &&
78 is_separator(str[0]) &&
82 if (str.size() > 0 && is_separator(str[str.size() - 1]))
83 return str.size() - 1;
85 size_t pos = str.find_last_of(separators, str.size() - 1);
88 if (pos == StringRef::npos)
89 pos = str.find_last_of(':', str.size() - 2);
92 if (pos == StringRef::npos ||
93 (pos == 1 && is_separator(str[0])))
99 size_t root_dir_start(StringRef str) {
102 if (str.size() > 2 &&
104 is_separator(str[2]))
109 if (str.size() == 2 &&
110 is_separator(str[0]) &&
112 return StringRef::npos;
115 if (str.size() > 3 &&
116 is_separator(str[0]) &&
118 !is_separator(str[2])) {
119 return str.find_first_of(separators, 2);
123 if (str.size() > 0 && is_separator(str[0]))
126 return StringRef::npos;
129 size_t parent_path_end(StringRef path) {
130 size_t end_pos = filename_pos(path);
132 bool filename_was_sep = path.size() > 0 && is_separator(path[end_pos]);
134 // Skip separators except for root dir.
135 size_t root_dir_pos = root_dir_start(path.substr(0, end_pos));
138 (end_pos - 1) != root_dir_pos &&
139 is_separator(path[end_pos - 1]))
142 if (end_pos == 1 && root_dir_pos == 0 && filename_was_sep)
143 return StringRef::npos;
147 } // end unnamed namespace
153 const_iterator begin(StringRef path) {
156 i.Component = find_first_component(path);
161 const_iterator end(StringRef path) {
164 i.Position = path.size();
168 const_iterator &const_iterator::operator++() {
169 assert(Position < Path.size() && "Tried to increment past end!");
171 // Increment Position to past the current component
172 Position += Component.size();
175 if (Position == Path.size()) {
176 Component = StringRef();
180 // Both POSIX and Windows treat paths that begin with exactly two separators
182 bool was_net = Component.size() > 2 &&
183 is_separator(Component[0]) &&
184 Component[1] == Component[0] &&
185 !is_separator(Component[2]);
187 // Handle separators.
188 if (is_separator(Path[Position])) {
193 || Component.endswith(":")
196 Component = Path.substr(Position, 1);
200 // Skip extra separators.
201 while (Position != Path.size() &&
202 is_separator(Path[Position])) {
206 // Treat trailing '/' as a '.'.
207 if (Position == Path.size()) {
214 // Find next component.
215 size_t end_pos = Path.find_first_of(separators, Position);
216 Component = Path.slice(Position, end_pos);
221 const_iterator &const_iterator::operator--() {
222 // If we're at the end and the previous char was a '/', return '.'.
223 if (Position == Path.size() &&
225 is_separator(Path[Position - 1])
227 && Path[Position - 2] != ':'
235 // Skip separators unless it's the root directory.
236 size_t root_dir_pos = root_dir_start(Path);
237 size_t end_pos = Position;
240 (end_pos - 1) != root_dir_pos &&
241 is_separator(Path[end_pos - 1]))
244 // Find next separator.
245 size_t start_pos = filename_pos(Path.substr(0, end_pos));
246 Component = Path.slice(start_pos, end_pos);
247 Position = start_pos;
251 bool const_iterator::operator==(const const_iterator &RHS) const {
252 return Path.begin() == RHS.Path.begin() &&
253 Position == RHS.Position;
256 bool const_iterator::operator!=(const const_iterator &RHS) const {
257 return !(*this == RHS);
260 ptrdiff_t const_iterator::operator-(const const_iterator &RHS) const {
261 return Position - RHS.Position;
264 const StringRef root_path(StringRef path) {
265 const_iterator b = begin(path),
269 bool has_net = b->size() > 2 && is_separator((*b)[0]) && (*b)[1] == (*b)[0];
277 if (has_net || has_drive) {
278 if ((++pos != e) && is_separator((*pos)[0])) {
279 // {C:/,//net/}, so get the first two components.
280 return path.substr(0, b->size() + pos->size());
282 // just {C:,//net}, return the first component.
287 // POSIX style root directory.
288 if (is_separator((*b)[0])) {
296 const StringRef root_name(StringRef path) {
297 const_iterator b = begin(path),
300 bool has_net = b->size() > 2 && is_separator((*b)[0]) && (*b)[1] == (*b)[0];
308 if (has_net || has_drive) {
309 // just {C:,//net}, return the first component.
314 // No path or no name.
318 const StringRef root_directory(StringRef path) {
319 const_iterator b = begin(path),
323 bool has_net = b->size() > 2 && is_separator((*b)[0]) && (*b)[1] == (*b)[0];
331 if ((has_net || has_drive) &&
332 // {C:,//net}, skip to the next component.
333 (++pos != e) && is_separator((*pos)[0])) {
337 // POSIX style root directory.
338 if (!has_net && is_separator((*b)[0])) {
343 // No path or no root.
347 const StringRef relative_path(StringRef path) {
348 StringRef root = root_path(path);
349 return path.substr(root.size());
352 void append(SmallVectorImpl<char> &path, const Twine &a,
356 SmallString<32> a_storage;
357 SmallString<32> b_storage;
358 SmallString<32> c_storage;
359 SmallString<32> d_storage;
361 SmallVector<StringRef, 4> components;
362 if (!a.isTriviallyEmpty()) components.push_back(a.toStringRef(a_storage));
363 if (!b.isTriviallyEmpty()) components.push_back(b.toStringRef(b_storage));
364 if (!c.isTriviallyEmpty()) components.push_back(c.toStringRef(c_storage));
365 if (!d.isTriviallyEmpty()) components.push_back(d.toStringRef(d_storage));
367 for (SmallVectorImpl<StringRef>::const_iterator i = components.begin(),
368 e = components.end();
370 bool path_has_sep = !path.empty() && is_separator(path[path.size() - 1]);
371 bool component_has_sep = !i->empty() && is_separator((*i)[0]);
372 bool is_root_name = has_root_name(*i);
375 // Strip separators from beginning of component.
376 size_t loc = i->find_first_not_of(separators);
377 StringRef c = i->substr(loc);
380 path.append(c.begin(), c.end());
384 if (!component_has_sep && !(path.empty() || is_root_name)) {
386 path.push_back(prefered_separator);
389 path.append(i->begin(), i->end());
393 void append(SmallVectorImpl<char> &path,
394 const_iterator begin, const_iterator end) {
395 for (; begin != end; ++begin)
396 path::append(path, *begin);
399 const StringRef parent_path(StringRef path) {
400 size_t end_pos = parent_path_end(path);
401 if (end_pos == StringRef::npos)
404 return path.substr(0, end_pos);
407 void remove_filename(SmallVectorImpl<char> &path) {
408 size_t end_pos = parent_path_end(StringRef(path.begin(), path.size()));
409 if (end_pos != StringRef::npos)
410 path.set_size(end_pos);
413 void replace_extension(SmallVectorImpl<char> &path, const Twine &extension) {
414 StringRef p(path.begin(), path.size());
415 SmallString<32> ext_storage;
416 StringRef ext = extension.toStringRef(ext_storage);
418 // Erase existing extension.
419 size_t pos = p.find_last_of('.');
420 if (pos != StringRef::npos && pos >= filename_pos(p))
423 // Append '.' if needed.
424 if (ext.size() > 0 && ext[0] != '.')
428 path.append(ext.begin(), ext.end());
431 void native(const Twine &path, SmallVectorImpl<char> &result) {
435 SmallString<128> path_storage;
436 StringRef p = path.toStringRef(path_storage);
437 result.reserve(p.size());
438 for (StringRef::const_iterator i = p.begin(),
443 result.push_back('\\');
445 result.push_back(*i);
448 path.toVector(result);
452 const StringRef filename(StringRef path) {
453 return *(--end(path));
456 const StringRef stem(StringRef path) {
457 StringRef fname = filename(path);
458 size_t pos = fname.find_last_of('.');
459 if (pos == StringRef::npos)
462 if ((fname.size() == 1 && fname == ".") ||
463 (fname.size() == 2 && fname == ".."))
466 return fname.substr(0, pos);
469 const StringRef extension(StringRef path) {
470 StringRef fname = filename(path);
471 size_t pos = fname.find_last_of('.');
472 if (pos == StringRef::npos)
475 if ((fname.size() == 1 && fname == ".") ||
476 (fname.size() == 2 && fname == ".."))
479 return fname.substr(pos);
482 bool is_separator(char value) {
485 case '\\': // fall through
487 case '/': return true;
488 default: return false;
492 void system_temp_directory(bool erasedOnReboot, SmallVectorImpl<char> &result) {
495 // Check whether the temporary directory is specified by an environment
497 const char *EnvironmentVariable;
499 EnvironmentVariable = "TEMP";
501 EnvironmentVariable = "TMPDIR";
503 if (char *RequestedDir = getenv(EnvironmentVariable)) {
504 result.append(RequestedDir, RequestedDir + strlen(RequestedDir));
508 // Fall back to a system default.
509 const char *DefaultResult;
511 (void)erasedOnReboot;
512 DefaultResult = "C:\\TEMP";
515 DefaultResult = "/tmp";
517 DefaultResult = "/var/tmp";
519 result.append(DefaultResult, DefaultResult + strlen(DefaultResult));
522 bool has_root_name(const Twine &path) {
523 SmallString<128> path_storage;
524 StringRef p = path.toStringRef(path_storage);
526 return !root_name(p).empty();
529 bool has_root_directory(const Twine &path) {
530 SmallString<128> path_storage;
531 StringRef p = path.toStringRef(path_storage);
533 return !root_directory(p).empty();
536 bool has_root_path(const Twine &path) {
537 SmallString<128> path_storage;
538 StringRef p = path.toStringRef(path_storage);
540 return !root_path(p).empty();
543 bool has_relative_path(const Twine &path) {
544 SmallString<128> path_storage;
545 StringRef p = path.toStringRef(path_storage);
547 return !relative_path(p).empty();
550 bool has_filename(const Twine &path) {
551 SmallString<128> path_storage;
552 StringRef p = path.toStringRef(path_storage);
554 return !filename(p).empty();
557 bool has_parent_path(const Twine &path) {
558 SmallString<128> path_storage;
559 StringRef p = path.toStringRef(path_storage);
561 return !parent_path(p).empty();
564 bool has_stem(const Twine &path) {
565 SmallString<128> path_storage;
566 StringRef p = path.toStringRef(path_storage);
568 return !stem(p).empty();
571 bool has_extension(const Twine &path) {
572 SmallString<128> path_storage;
573 StringRef p = path.toStringRef(path_storage);
575 return !extension(p).empty();
578 bool is_absolute(const Twine &path) {
579 SmallString<128> path_storage;
580 StringRef p = path.toStringRef(path_storage);
582 bool rootDir = has_root_directory(p),
584 rootName = has_root_name(p);
589 return rootDir && rootName;
592 bool is_relative(const Twine &path) {
593 return !is_absolute(path);
596 } // end namespace path
600 error_code make_absolute(SmallVectorImpl<char> &path) {
601 StringRef p(path.data(), path.size());
603 bool rootDirectory = path::has_root_directory(p),
605 rootName = path::has_root_name(p);
611 if (rootName && rootDirectory)
612 return error_code::success();
614 // All of the following conditions will need the current directory.
615 SmallString<128> current_dir;
616 if (error_code ec = current_path(current_dir)) return ec;
618 // Relative path. Prepend the current directory.
619 if (!rootName && !rootDirectory) {
620 // Append path to the current directory.
621 path::append(current_dir, p);
622 // Set path to the result.
623 path.swap(current_dir);
624 return error_code::success();
627 if (!rootName && rootDirectory) {
628 StringRef cdrn = path::root_name(current_dir);
629 SmallString<128> curDirRootName(cdrn.begin(), cdrn.end());
630 path::append(curDirRootName, p);
631 // Set path to the result.
632 path.swap(curDirRootName);
633 return error_code::success();
636 if (rootName && !rootDirectory) {
637 StringRef pRootName = path::root_name(p);
638 StringRef bRootDirectory = path::root_directory(current_dir);
639 StringRef bRelativePath = path::relative_path(current_dir);
640 StringRef pRelativePath = path::relative_path(p);
642 SmallString<128> res;
643 path::append(res, pRootName, bRootDirectory, bRelativePath, pRelativePath);
645 return error_code::success();
648 llvm_unreachable("All rootName and rootDirectory combinations should have "
652 error_code create_directories(const Twine &path, bool &existed) {
653 SmallString<128> path_storage;
654 StringRef p = path.toStringRef(path_storage);
656 StringRef parent = path::parent_path(p);
657 if (!parent.empty()) {
659 if (error_code ec = fs::exists(parent, parent_exists)) return ec;
662 if (error_code ec = create_directories(parent, existed)) return ec;
665 return create_directory(p, existed);
668 bool exists(file_status status) {
669 return status_known(status) && status.type() != file_type::file_not_found;
672 bool status_known(file_status s) {
673 return s.type() != file_type::status_error;
676 bool is_directory(file_status status) {
677 return status.type() == file_type::directory_file;
680 error_code is_directory(const Twine &path, bool &result) {
682 if (error_code ec = status(path, st))
684 result = is_directory(st);
685 return error_code::success();
688 bool is_regular_file(file_status status) {
689 return status.type() == file_type::regular_file;
692 error_code is_regular_file(const Twine &path, bool &result) {
694 if (error_code ec = status(path, st))
696 result = is_regular_file(st);
697 return error_code::success();
700 bool is_symlink(file_status status) {
701 return status.type() == file_type::symlink_file;
704 error_code is_symlink(const Twine &path, bool &result) {
706 if (error_code ec = status(path, st))
708 result = is_symlink(st);
709 return error_code::success();
712 bool is_other(file_status status) {
713 return exists(status) &&
714 !is_regular_file(status) &&
715 !is_directory(status) &&
719 void directory_entry::replace_filename(const Twine &filename, file_status st) {
720 SmallString<128> path(Path.begin(), Path.end());
721 path::remove_filename(path);
722 path::append(path, filename);
727 error_code has_magic(const Twine &path, const Twine &magic, bool &result) {
728 SmallString<32> MagicStorage;
729 StringRef Magic = magic.toStringRef(MagicStorage);
730 SmallString<32> Buffer;
732 if (error_code ec = get_magic(path, Magic.size(), Buffer)) {
733 if (ec == errc::value_too_large) {
734 // Magic.size() > file_size(Path).
736 return error_code::success();
741 result = Magic == Buffer;
742 return error_code::success();
745 /// @brief Identify the magic in magic.
746 file_magic identify_magic(StringRef magic) {
747 switch ((unsigned char)magic[0]) {
748 case 0xDE: // 0x0B17C0DE = BC wraper
749 if (magic[1] == (char)0xC0 && magic[2] == (char)0x17 &&
750 magic[3] == (char)0x0B)
751 return file_magic::bitcode;
754 if (magic[1] == 'C' && magic[2] == (char)0xC0 && magic[3] == (char)0xDE)
755 return file_magic::bitcode;
758 if (magic.size() >= 8)
759 if (memcmp(magic.data(),"!<arch>\n",8) == 0)
760 return file_magic::archive;
764 if (magic[1] == 'E' && magic[2] == 'L' && magic[3] == 'F') {
765 if (magic.size() >= 18 && magic[17] == 0)
768 case 1: return file_magic::elf_relocatable;
769 case 2: return file_magic::elf_executable;
770 case 3: return file_magic::elf_shared_object;
771 case 4: return file_magic::elf_core;
777 if (magic[1] == char(0xFE) && magic[2] == char(0xBA) &&
778 magic[3] == char(0xBE)) {
779 // This is complicated by an overlap with Java class files.
780 // See the Mach-O section in /usr/share/file/magic for details.
781 if (magic.size() >= 8 && magic[7] < 43)
782 // FIXME: Universal Binary of any type.
783 return file_magic::macho_dynamically_linked_shared_lib;
787 // The two magic numbers for mach-o are:
788 // 0xfeedface - 32-bit mach-o
789 // 0xfeedfacf - 64-bit mach-o
794 if (magic[0] == char(0xFE) && magic[1] == char(0xED) &&
795 magic[2] == char(0xFA) &&
796 (magic[3] == char(0xCE) || magic[3] == char(0xCF))) {
798 if (magic.size() >= 16) type = magic[14] << 8 | magic[15];
799 } else if ((magic[0] == char(0xCE) || magic[0] == char(0xCF)) &&
800 magic[1] == char(0xFA) && magic[2] == char(0xED) &&
801 magic[3] == char(0xFE)) {
803 if (magic.size() >= 14) type = magic[13] << 8 | magic[12];
807 case 1: return file_magic::macho_object;
808 case 2: return file_magic::macho_executable;
809 case 3: return file_magic::macho_fixed_virtual_memory_shared_lib;
810 case 4: return file_magic::macho_core;
811 case 5: return file_magic::macho_preload_executabl;
812 case 6: return file_magic::macho_dynamically_linked_shared_lib;
813 case 7: return file_magic::macho_dynamic_linker;
814 case 8: return file_magic::macho_bundle;
815 case 9: return file_magic::macho_dynamic_linker;
816 case 10: return file_magic::macho_dsym_companion;
820 case 0xF0: // PowerPC Windows
821 case 0x83: // Alpha 32-bit
822 case 0x84: // Alpha 64-bit
823 case 0x66: // MPS R4000 Windows
825 case 0x4c: // 80386 Windows
826 if (magic[1] == 0x01)
827 return file_magic::coff_object;
829 case 0x90: // PA-RISC Windows
830 case 0x68: // mc68K Windows
831 if (magic[1] == 0x02)
832 return file_magic::coff_object;
835 case 0x4d: // Possible MS-DOS stub on Windows PE file
836 if (magic[1] == 0x5a) {
838 *reinterpret_cast<const support::ulittle32_t*>(magic.data() + 0x3c);
839 // PE/COFF file, either EXE or DLL.
840 if (off < magic.size() && memcmp(magic.data() + off, "PE\0\0",4) == 0)
841 return file_magic::pecoff_executable;
845 case 0x64: // x86-64 Windows.
846 if (magic[1] == char(0x86))
847 return file_magic::coff_object;
853 return file_magic::unknown;
856 error_code identify_magic(const Twine &path, file_magic &result) {
857 SmallString<32> Magic;
858 error_code ec = get_magic(path, Magic.capacity(), Magic);
859 if (ec && ec != errc::value_too_large)
862 result = identify_magic(Magic);
863 return error_code::success();
867 error_code remove_all_r(StringRef path, file_type ft, uint32_t &count) {
868 if (ft == file_type::directory_file) {
869 // This code would be a lot better with exceptions ;/.
871 directory_iterator i(path, ec);
873 for (directory_iterator e; i != e; i.increment(ec)) {
876 if (error_code ec = i->status(st)) return ec;
877 if (error_code ec = remove_all_r(i->path(), st.type(), count)) return ec;
879 bool obviously_this_exists;
880 if (error_code ec = remove(path, obviously_this_exists)) return ec;
881 assert(obviously_this_exists);
882 ++count; // Include the directory itself in the items removed.
884 bool obviously_this_exists;
885 if (error_code ec = remove(path, obviously_this_exists)) return ec;
886 assert(obviously_this_exists);
890 return error_code::success();
892 } // end unnamed namespace
894 error_code remove_all(const Twine &path, uint32_t &num_removed) {
895 SmallString<128> path_storage;
896 StringRef p = path.toStringRef(path_storage);
899 if (error_code ec = status(path, fs))
902 return remove_all_r(p, fs.type(), num_removed);
905 error_code directory_entry::status(file_status &result) const {
906 return fs::status(Path, result);
909 } // end namespace fs
910 } // end namespace sys
911 } // end namespace llvm
913 // Include the truly platform-specific parts.
914 #if defined(LLVM_ON_UNIX)
915 #include "Unix/PathV2.inc"
917 #if defined(LLVM_ON_WIN32)
918 #include "Windows/PathV2.inc"