enhancements for `<filesystem>` #3850

achabense · 2023-07-05T21:25:04Z

refactor path::lexically_normal (replacing list with vector)
drop <list> inclusion

Benchmark for lexically_normal

#include <filesystem>
#include <iostream>

#include "benchmark/benchmark.h"

using namespace std;
namespace fs = std::filesystem;

fs::path _lexically_normal_bef(const fs::path& path) {
    return path.lexically_normal();
}

fs::path _lexically_normal_now(const fs::path& path) {
    const auto& _Text = path.native();//path::_Text.
    constexpr auto preferred_separator = fs::path::preferred_separator;

    constexpr wstring_view _Dot = L"."sv;
    constexpr wstring_view _Dot_dot = L".."sv;

    // N4950 [fs.path.generic]/6:
    // "Normalization of a generic format pathname means:"

    // "1. If the path is empty, stop."
    if (path.empty()) {
        return {};
    }

    // "2. Replace each slash character in the root-name with a preferred-separator."
    const auto _First = _Text.data();
    const auto _Last = _First + _Text.size();
    const auto _Root_name_end = fs::_Find_root_name_end(_First, _Last);
    fs::path::string_type _Normalized(_First, _Root_name_end);
    _STD replace(_Normalized.begin(), _Normalized.end(), L'/', L'\\');

    // "3. Replace each directory-separator with a preferred-separator.
    // [ Note: The generic pathname grammar (29.11.7.1) defines directory-separator
    // as one or more slashes and preferred-separators. -end note ]"
    vector<wstring_view> _Vec; // Empty wstring_view means directory-separator
    // that will be normalized to a preferred-separator.
    // Non-empty wstring_view means filename.
    _Vec.reserve(13); // avoid frequent re-allocations
    bool _Has_root_directory = false; // true: there is a slash right after root name.
    auto _Pos = _Root_name_end;
    if (_Pos != _Last && _Is_slash(*_Pos)) {
        _Has_root_directory = true;
        _Normalized += preferred_separator;
        ++_Pos;
        while (_Pos != _Last && _Is_slash(*_Pos)) { ++_Pos; }
    }
    // _Vec will start with a filename (if not empty).
    while (_Pos != _Last) {
        if (_Is_slash(*_Pos)) {
            if (_Vec.empty() || !_Vec.back().empty()) {
                // collapse one or more slashes and preferred-separators to one empty wstring_view
                _Vec.emplace_back();
            }
            ++_Pos;
        }
        else {
            const auto _Filename_end = _STD find_if(_Pos + 1, _Last, _Is_slash);
            _Vec.emplace_back(_Pos, static_cast<size_t>(_Filename_end - _Pos));
            _Pos = _Filename_end;
        }
    }

    // "4. Remove each dot filename and any immediately following directory-separator."
    // "5. As long as any appear, remove a non-dot-dot filename immediately followed by a
    // directory-separator and a dot-dot filename, along with any immediately following directory-separator."
    // "6. If there is a root-directory, remove all dot-dot filenames
    // and any directory-separators immediately following them.
    // [ Note: These dot-dot filenames attempt to refer to nonexistent parent directories. -end note ]"
    auto _New_end = _Vec.begin();
    for (auto _Pos = _Vec.begin(); _Pos != _Vec.end();) {
        auto _Elem = *_Pos++; // _Pos points at end or a separator after ++.
        if (_Elem == _Dot) {
            // ignore dot (and following separator).
            if (_Pos == _Vec.end()) { break; }
        }
        else if (_Elem == _Dot_dot) {
            if (_New_end != _Vec.begin() && *prev(_New_end, 2) != _Dot_dot) {
                // note: _New_end == _Vec.begin() + 2n
                // remove preceding non-dot-dot filename and separator.
                _New_end -= 2;
                if (_Pos == _Vec.end()) { break; }
            }
            else if (!_Has_root_directory) {
                // due to 6, append dot-dot only when !_Has_root_directory.
                *_New_end++ = _Dot_dot;
                if (_Pos == _Vec.end()) { break; }
                *_New_end++ = {}; // as _Pos != _Vec.end(), it points at a separator; add it.
            }
            else {
                // ignore dot-dot (and following separator).
                if (_Pos == _Vec.end()) { break; }
            }
        }
        else {
            // append normal filename and separator.
            *_New_end++ = _Elem;
            if (_Pos == _Vec.end()) { break; }
            *_New_end++ = {}; // add separator.
        }
        ++_Pos; // _Pos points at a separator here in all cases; skip it.
    }
    _Vec.erase(_New_end, _Vec.end());

    // "7. If the last filename is dot-dot, remove any trailing directory-separator."
    if (_Vec.size() >= 2 && _Vec.back().empty() && *(_STD prev(_Vec.end(), 2)) == _Dot_dot) {
        _Vec.pop_back();
    }

    // Build up _Normalized by flattening _Vec.
    for (const auto& _Elem : _Vec) {
        if (_Elem.empty()) {
            _Normalized += preferred_separator;
        }
        else {
            _Normalized += _Elem;
        }
    }

    // "8. If the path is empty, add a dot."
    if (_Normalized.empty()) {
        _Normalized = _Dot;
    }

    // "The result of normalization is a path in normal form, which is said to be normalized."
    return fs::path(_STD move(_Normalized));
}

using path = fs::path;
using _pfn = fs::path(*)(const fs::path&);

// modified from tests\P2018R1_filesystem\test.cpp
void _check(_pfn _lexically_normal, const path& path, wstring_view target) {
    fs::path result = _lexically_normal(path);
    if (result != target) [[unlikely]] {
        terminate();
        //wcerr << quoted(path.native()) << "->" << quoted(result.native()) << '\n';
    }
}

void test_1(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(cat/./dog/..)"sv), LR"(cat\)"sv);
    _check(_lexically_normal, path(LR"(cat/.///dog/../)"sv), LR"(cat\)"sv);
    _check(_lexically_normal, path(LR"(.)"sv), LR"(.)"sv);
    _check(_lexically_normal, path(LR"(.\)"sv), LR"(.)"sv);
    _check(_lexically_normal, path(LR"(.\.)"sv), LR"(.)"sv);
    _check(_lexically_normal, path(LR"(.\.\)"sv), LR"(.)"sv);
}

void test_2(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"()"sv), LR"()"sv);
    _check(_lexically_normal, path(LR"(X:)"sv), LR"(X:)"sv);
    _check(_lexically_normal, path(LR"(X:DriveRelative)"sv), LR"(X:DriveRelative)"sv);
    _check(_lexically_normal, path(LR"(X:\)"sv), LR"(X:\)"sv);
    _check(_lexically_normal, path(LR"(X:/)"sv), LR"(X:\)"sv);
    _check(_lexically_normal, path(LR"(X:\\\)"sv), LR"(X:\)"sv);
    _check(_lexically_normal, path(LR"(X:///)"sv), LR"(X:\)"sv);
}

void test_3(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(X:\DosAbsolute)"sv), LR"(X:\DosAbsolute)"sv);
    _check(_lexically_normal, path(LR"(X:/DosAbsolute)"sv), LR"(X:\DosAbsolute)"sv);
    _check(_lexically_normal, path(LR"(X:\\\DosAbsolute)"sv), LR"(X:\DosAbsolute)"sv);
    _check(_lexically_normal, path(LR"(X:///DosAbsolute)"sv), LR"(X:\DosAbsolute)"sv);
    _check(_lexically_normal, path(LR"(\RootRelative)"sv), LR"(\RootRelative)"sv);
    _check(_lexically_normal, path(LR"(/RootRelative)"sv), LR"(\RootRelative)"sv);
    _check(_lexically_normal, path(LR"(\\\RootRelative)"sv), LR"(\RootRelative)"sv);
    _check(_lexically_normal, path(LR"(///RootRelative)"sv), LR"(\RootRelative)"sv);
    _check(_lexically_normal, path(LR"(\\server\share)"sv), LR"(\\server\share)"sv);
    _check(_lexically_normal, path(LR"(//server/share)"sv), LR"(\\server\share)"sv);
    _check(_lexically_normal, path(LR"(\\server\\\share)"sv), LR"(\\server\share)"sv);
    _check(_lexically_normal, path(LR"(//server///share)"sv), LR"(\\server\share)"sv);
}

void test_4(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(\\?\device)"sv), LR"(\\?\device)"sv);
    _check(_lexically_normal, path(LR"(//?/device)"sv), LR"(\\?\device)"sv);
    _check(_lexically_normal, path(LR"(\??\device)"sv), LR"(\??\device)"sv);
    _check(_lexically_normal, path(LR"(/??/device)"sv), LR"(\??\device)"sv);
    _check(_lexically_normal, path(LR"(\\.\device)"sv), LR"(\\.\device)"sv);
    _check(_lexically_normal, path(LR"(//./device)"sv), LR"(\\.\device)"sv);
}

void test_5(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(\\?\UNC\server\share)"sv), LR"(\\?\UNC\server\share)"sv);
    _check(_lexically_normal, path(LR"(//?/UNC/server/share)"sv), LR"(\\?\UNC\server\share)"sv);
    _check(_lexically_normal, path(LR"(C:\a/b\\c\/d/\e//f)"sv), LR"(C:\a\b\c\d\e\f)"sv);
}

void test_6(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(C:\meow\)"sv), LR"(C:\meow\)"sv);
    _check(_lexically_normal, path(LR"(C:\meow/)"sv), LR"(C:\meow\)"sv);
    _check(_lexically_normal, path(LR"(C:\meow\\)"sv), LR"(C:\meow\)"sv);
    _check(_lexically_normal, path(LR"(C:\meow\/)"sv), LR"(C:\meow\)"sv);
    _check(_lexically_normal, path(LR"(C:\meow/\)"sv), LR"(C:\meow\)"sv);
    _check(_lexically_normal, path(LR"(C:\meow//)"sv), LR"(C:\meow\)"sv);
}

void test_7(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(C:\a\.\b\.\.\c\.\.\.)"sv), LR"(C:\a\b\c\)"sv);
    _check(_lexically_normal, path(LR"(C:\a\.\b\.\.\c\.\.\.\)"sv), LR"(C:\a\b\c\)"sv);
    _check(_lexically_normal, path(LR"(C:\a\b\c\d\e\..\f\..\..\..\g\h)"sv), LR"(C:\a\b\g\h)"sv);
    _check(_lexically_normal, path(LR"(C:\a\b\c\d\e\..\f\..\..\..\g\h\..)"sv), LR"(C:\a\b\g\)"sv);
    _check(_lexically_normal, path(LR"(C:\a\b\c\d\e\..\f\..\..\..\g\h\..\)"sv), LR"(C:\a\b\g\)"sv);
}

void test_8(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(..\..\..)"sv), LR"(..\..\..)"sv);
    _check(_lexically_normal, path(LR"(..\..\..\)"sv), LR"(..\..\..)"sv);
    _check(_lexically_normal, path(LR"(..\..\..\a\b\c)"sv), LR"(..\..\..\a\b\c)"sv);
    _check(_lexically_normal, path(LR"(\..\..\..)"sv), LR"(\)"sv);
    _check(_lexically_normal, path(LR"(\..\..\..\)"sv), LR"(\)"sv);
    _check(_lexically_normal, path(LR"(\..\..\..\a\b\c)"sv), LR"(\a\b\c)"sv);
    _check(_lexically_normal, path(LR"(a\..)"sv), LR"(.)"sv);
    _check(_lexically_normal, path(LR"(a\..\)"sv), LR"(.)"sv);
}

void test_9(_pfn _lexically_normal) {
    _check(_lexically_normal, path(LR"(/\server/\share/\a/\b/\c/\./\./\d/\../\../\../\../\../\../\../\other/x/y/z/.././..\meow.txt)"sv)
        , LR"(\\server\other\x\meow.txt)"sv);
}

template<auto _test>
void BM_bef(benchmark::State& state) {
    for (auto _ : state) {
        _test(_lexically_normal_bef);
    }
}

template<auto _test>
void BM_now(benchmark::State& state) {
    for (auto _ : state) {
        _test(_lexically_normal_now);
    }
}

void BM_bef_Tot(benchmark::State& state) {
    for (auto _ : state) {
        test_1(_lexically_normal_bef);
        test_2(_lexically_normal_bef);
        test_3(_lexically_normal_bef);
        test_4(_lexically_normal_bef);
        test_5(_lexically_normal_bef);
        test_6(_lexically_normal_bef);
        test_7(_lexically_normal_bef);
        test_8(_lexically_normal_bef);
        test_9(_lexically_normal_bef);
    }
}

void BM_now_Tot(benchmark::State& state) {
    for (auto _ : state) {
        test_1(_lexically_normal_now);
        test_2(_lexically_normal_now);
        test_3(_lexically_normal_now);
        test_4(_lexically_normal_now);
        test_5(_lexically_normal_now);
        test_6(_lexically_normal_now);
        test_7(_lexically_normal_now);
        test_8(_lexically_normal_now);
        test_9(_lexically_normal_now);
    }
}

BENCHMARK(BM_bef<test_1>);
BENCHMARK(BM_now<test_1>);
BENCHMARK(BM_bef<test_2>);
BENCHMARK(BM_now<test_2>);
BENCHMARK(BM_bef<test_3>);
BENCHMARK(BM_now<test_3>);
BENCHMARK(BM_bef<test_4>);
BENCHMARK(BM_now<test_4>);
BENCHMARK(BM_bef<test_5>);
BENCHMARK(BM_now<test_5>);
BENCHMARK(BM_bef<test_6>);
BENCHMARK(BM_now<test_6>);
BENCHMARK(BM_bef<test_7>);
BENCHMARK(BM_now<test_7>);
BENCHMARK(BM_bef<test_8>);
BENCHMARK(BM_now<test_8>);
BENCHMARK(BM_bef<test_9>);
BENCHMARK(BM_now<test_9>);
BENCHMARK(BM_bef_Tot);
BENCHMARK(BM_now_Tot);

// typical path
void test_10(_pfn _lexically_normal) {
    const path p(LR"(C:\Program Files\Azure Data Studio\resources\app\extensions\bat\snippets\batchfile.code-snippets)");
    benchmark::DoNotOptimize(_lexically_normal(p));
}

void test_11(_pfn _lexically_normal) {
    const path p(LR"(..\snippets\batchfile.code-snippets)");
    benchmark::DoNotOptimize(_lexically_normal(p));
}

BENCHMARK(BM_bef<test_10>);
BENCHMARK(BM_now<test_10>);
BENCHMARK(BM_bef<test_11>);
BENCHMARK(BM_now<test_11>);

BENCHMARK_MAIN();

Result

The main bottleneck of the original implementation was memory allocation. As the benchmark shows, the new approach works much better for "long" paths(like cat/./dog/.. or longer). For very-short paths(like X:///) the efficiency gain will not be significant.

…to _For_filesystem

stl/inc/filesystem

StephanTLavavej · 2023-07-05T21:39:33Z

As lexically_normal is pure computation, it would seem reasonable to add your benchmark to this PR. (It only needs to measure the improved implementation of lexically_normal, not provide the old implementation for comparison.)

Additionally, { break; } will fail clang-format, which adds our conventional newlines here. As a reminder, we recommend configuring your editor to clang-format on save, to avoid having such formatting issues detected by the CI system. Opening the repo in VSCode with a clang-format extension will automatically do this:

STL/.vscode/settings.json

Lines 5 to 8 in 2261f7e

    
           "editor.formatOnSave": false, 
        
           "[cpp]": { 
        
               "editor.formatOnSave": true 
        
           },

achabense · 2023-07-05T21:59:25Z

I will update these benchmarks once I have time.

It only needs to measure the improved implementation of lexically_normal, not provide the old implementation for comparison.

Does this mean that I only need to provide test cases in the benchmark, and call library functions directly, roughly like this?

Example

#include <filesystem>

#include "benchmark/benchmark.h"

using namespace std;
namespace fs = std::filesystem;

void BM_lexically_normal(benchmark::State& state) {
  const fs::path& p(
      LR"(C:\Program Files\Azure Data Studio\resources\app\extensions\bat\snippets\batchfile.code-snippets)");
  for (auto _ : state) {
    benchmark::DoNotOptimize(p.lexically_normal());
  }
}

BENCHMARK(BM_lexically_normal);
// more tests...

BENCHMARK_MAIN();

StephanTLavavej · 2023-07-05T22:17:08Z

Thanks!

Does this mean that I only need to provide test cases in the benchmark, and call library functions directly, roughly like this?

Yep, then we can compare the results before and after proposed changes.

frederick-vs-ja

We should make calls to std::prev ADL-proof. As you've changed to use vector now, I believe we can use _New_end[-2]/_Vec.end()[-2].

stl/inc/filesystem

Co-authored-by: A. Jiang <[email protected]>

…to _For_filesystem

strega-nil-ms

That is the only thing I see; this is fantastic otherwise. Thanks so much!

stl/inc/filesystem

…iew..."; try to improve comment

benchmarks/src/path_lexically_normal.cpp

stl/inc/filesystem

benchmarks/src/path_lexically_normal.cpp

StephanTLavavej · 2023-07-07T22:16:09Z

Thanks, looks great! 😻 🗃️ I pushed minor changes, the most significant being an update to benchmarks/CMakeLists.txt so the new benchmark is automatically built by the suite. FYI @strega-nil-ms as you had already approved.

achabense · 2023-07-08T05:59:50Z

benchmarks/src/path_lexically_normal.cpp

+            LR"(X:DriveRelative)"sv,
+            LR"(\\server\\\share)"sv,
+            LR"(STL/.github/workflows/../..)"sv,
+            LR"(C:\Program Files\Azure Data Studio\resources\app\extensions\bat\snippets\batchfile.code-snippets)"sv,


Is it better to replace these two test cases with more generic ones?

Seems fine to me since it's a realistic Microsoft path. If it were from some other program then I'd request a generic meow-style example.

yes, it is a real path in my computer.

for C:\Program Files\Azure Data Studio... I find a better one:
C:\Program Files\Microsoft Visual Studio\2022\Community\Common7\IDE\VC\Snippets (it has the same number of segments, and looks shorter than the last test case(/\server/\share/\a..., so that it's more obvious that the test cases are increasing in segment size).

(I'm not sure whether I can make pushes to ready-to-merge prs)

If you want to add more tests, it's probably better to do it in a new PR to not reset the review process.

Generally, when a PR has been moved to Ready To Merge, changes should only be pushed when they're big enough to be worth it (fixing bugs or major omissions), or they're small enough to be unquestionably safe like comment typo fixes. You should ping the maintainers who approved, on GitHub and/or Discord, so they notice and can take another look.

Basically, we maintainers love doing work, but we want to focus it on necessary work that keeps PRs flowing and the codebase moving forward. We try to minimize avoidable work, so we prefer to avoid scenarios like:

When a PR that was ready to go is changed in a way that requires further rework.

If the change was an attempt to fix a real bug, then this is not annoying - we'd rather pull a PR at the last minute than merge a bug. However, if the change wasn't really necessary, then iterating on it again is extra work.

When changes are pushed after Ready To Merge without maintainers being notified.

This is a potential process loophole so we don't like it - we require all codebase modifications to be reviewed by another maintainer so last-minute changes that try to sneak in are not cool. (This is applied uniformly; if maintainer X creates a PR, and maintainer Y approves, then X doesn't get to push further changes without Y taking a look. The only exceptions are fixing build breaks and test failures during mirroring, in which case we notify but don't block on getting re-approval.)

When changes are pushed after a maintainer has self-assigned the PR and commented "I'm mirroring this to the MSVC-internal codebase, please notify me of any changes", and they aren't for a really good reason.

Resetting this mirroring process is extra extra work and adds a lot of delay.

hm, thanks for clarification 👀

StephanTLavavej · 2023-07-13T02:43:39Z

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

StephanTLavavej · 2023-07-14T02:06:07Z

Thanks for improving the performance of my old implementation here! 😻 🚀 🐇

achabense added 4 commits July 6, 2023 05:01

optimize lexically_normal and drop <list> inclusion

799099e

Merge branch 'microsoft:main' into _For_filesystem

13bdc66

nit

eaa143f

Merge branch '_For_filesystem' of https://github.com/achabense/STL in…

0a1a255

…to _For_filesystem

achabense requested a review from a team as a code owner July 5, 2023 21:25

StephanTLavavej added performance Must go faster filesystem C++17 filesystem labels Jul 5, 2023

achabense commented Jul 5, 2023

View reviewed changes

stl/inc/filesystem Show resolved Hide resolved

StephanTLavavej assigned strega-nil-ms Jul 5, 2023

formatting

0693958

achabense added 3 commits July 6, 2023 06:46

renaming outerscope '_Pos'

db751a5

add benchmark

6264009

update benchmark

9b90aec

frederick-vs-ja reviewed Jul 6, 2023

View reviewed changes

stl/inc/filesystem Outdated Show resolved Hide resolved

stl/inc/filesystem Outdated Show resolved Hide resolved

achabense and others added 5 commits July 6, 2023 15:53

refactor benchmark

37a0393

apply []

c93c0c5

Co-authored-by: A. Jiang <[email protected]>

apply []

c5ed88c

Co-authored-by: A. Jiang <[email protected]>

Merge branch '_For_filesystem' of https://github.com/achabense/STL in…

04ee85d

…to _For_filesystem

nits

f948375

strega-nil-ms suggested changes Jul 6, 2023

View reviewed changes

stl/inc/filesystem Outdated Show resolved Hide resolved

StephanTLavavej unassigned strega-nil-ms Jul 7, 2023

review feedback

a81598e

StephanTLavavej assigned StephanTLavavej and strega-nil-ms Jul 7, 2023

restore indentation of "// that will be..." & "// Non-empty wstring_v…

74f2aa8

…iew..."; try to improve comment

strega-nil-ms approved these changes Jul 7, 2023

View reviewed changes

StephanTLavavej unassigned strega-nil-ms Jul 7, 2023

StephanTLavavej added 7 commits July 7, 2023 14:17

Add new benchmark to benchmarks/CMakeLists.txt.

9e58b1d

Include <string_view> instead of <xstring>.

9f89fe9

Style: Add a trailing comma to get one path per line.

fef6883

Update quoted Standardese.

df24c8d

Comment nitpicks.

dac591c

_Elem can be const.

753f9b6

Use DenseRange to avoid verbosity.

426b64e

StephanTLavavej reviewed Jul 7, 2023

View reviewed changes

StephanTLavavej approved these changes Jul 7, 2023

View reviewed changes

StephanTLavavej removed their assignment Jul 7, 2023

achabense commented Jul 8, 2023

View reviewed changes

frederick-vs-ja mentioned this pull request Jul 11, 2023

Optimize stacktrace by avoid large allocation #3859

Open

StephanTLavavej self-assigned this Jul 13, 2023

StephanTLavavej merged commit f6401b1 into microsoft:main Jul 14, 2023

achabense deleted the _For_filesystem branch July 14, 2023 13:02

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

enhancements for `<filesystem>` #3850

enhancements for `<filesystem>` #3850

achabense commented Jul 5, 2023

StephanTLavavej commented Jul 5, 2023

achabense commented Jul 5, 2023

StephanTLavavej commented Jul 5, 2023

frederick-vs-ja left a comment

strega-nil-ms left a comment

StephanTLavavej commented Jul 7, 2023

achabense Jul 8, 2023 •

edited

Loading

StephanTLavavej Jul 8, 2023

achabense Jul 8, 2023 •

edited

Loading

sam20908 Jul 8, 2023

StephanTLavavej Jul 8, 2023

achabense Jul 9, 2023

StephanTLavavej commented Jul 13, 2023

StephanTLavavej commented Jul 14, 2023

enhancements for <filesystem> #3850

enhancements for <filesystem> #3850

Conversation

achabense commented Jul 5, 2023

StephanTLavavej commented Jul 5, 2023

achabense commented Jul 5, 2023

StephanTLavavej commented Jul 5, 2023

frederick-vs-ja left a comment

Choose a reason for hiding this comment

strega-nil-ms left a comment

Choose a reason for hiding this comment

StephanTLavavej commented Jul 7, 2023

achabense Jul 8, 2023 • edited Loading

Choose a reason for hiding this comment

StephanTLavavej Jul 8, 2023

Choose a reason for hiding this comment

achabense Jul 8, 2023 • edited Loading

Choose a reason for hiding this comment

sam20908 Jul 8, 2023

Choose a reason for hiding this comment

StephanTLavavej Jul 8, 2023

Choose a reason for hiding this comment

achabense Jul 9, 2023

Choose a reason for hiding this comment

StephanTLavavej commented Jul 13, 2023

StephanTLavavej commented Jul 14, 2023

enhancements for `<filesystem>` #3850

enhancements for `<filesystem>` #3850

achabense Jul 8, 2023 •

edited

Loading

achabense Jul 8, 2023 •

edited

Loading