fd

Bugfixes

• Fixed performance regression due to --ignore-contain; see #1913 and #1914

This is just a re-release of 10.4.0 due to an issue with the 10.4.0 release.

Features

• Add --ignore-contain option to ignore directories containing a named entry (e.g. to ignore CACHEDIR.TAG); see #1727 (@fischman).

Bugfixes

• Fix Windows hyperlink generation for paths with spaces. (#1872)

• --print0 combined with --exec will now print a \0 between the output of each entry. Note that if there are multiple instances of --exec, the \0 will be between each set of commands, not between each individual command run. Fixes #1797.

• Several bugs were fixed by an update to the ignore library used for handling ignore rules

  • #1506
  • #1667
  • #1813

Changes

• Minimum required rust version has been increased to 1.90.0. Notably, this means dropping fully support for intel Mac and Windows 7.
• Statically link the CRT for MSVC builds via Cargo config to avoid runtime DLL dependencies, see #1874 (@FidelSch)

Features

• Add a hidden --mindepth alias for --min-depth. (#1617)

Bugfixes

Changes

• Replace humantime crate and chrono crate with jiff crate, see #1690 (@sorairolake). This has some small changes to the way dates given to options such --changed-within and --changed-before including:
  • 'M' no longer means "month", as that could be confusing with minutes. Use "mo", "mos", "month" or "months" instead.
  • month and year now account for variability in the calander rather than being a hard-coded number of seconds. That is probably what you would expect, but it is a slight change in behavior.
• aarch64 Windows was added to CI and release artifacts
• Many dependencies were updated
• Better support building on Illumos (there is no automated testing, but some known issues were fixed)

Other

This will be the last release that has been tested on x86_64 Mac OS, since GitHub is dropping support for runners with that hardware.

It may also be the last release to use a version of Rust with tier-1 support for x86_64/intel Macs and Windows 7.

Features

• Add --hyperlink option to add OSC 8 hyperlinks to output

Changes

• Build windows releases with rust 1.77 so windows 7 is still supported
• Deb packages now include symlink for fdfind to be more consistent with official packages

New Contributors

• @bryanhonof made their first contribution in https://github.com/sharkdp/fd/pull/1574
• @JCHacking made their first contribution in https://github.com/sharkdp/fd/pull/1602
• @RossSmyth made their first contribution in https://github.com/sharkdp/fd/pull/1596

Full Changelog: https://github.com/sharkdp/fd/compare/v10.1.0...v10.2.0

Features

• Allow passing an optional argument to --strip-cwd-prefix of "always", "never", or "auto". to force whether the cwd prefix is stripped or not.
• Add a --format option which allows using a format template for direct ouput similar to the template used for --exec. (#1043)

Bugfixes

• Fix aarch64 page size again. This time it should actually work. (#1085, #1549) (@tavianator)

Other

• aarch64-apple-darwin target added to builds on the release page. Note that this is a tier 2 rust target.

Features

• Add dir as an alias to directory when using -t \ --type, see #1460 and #1464 (@Ato2207).
• Add support for @%s date format in time filters similar to GNU date (seconds since Unix epoch for --older/--newer), see #1493 (@nabellows)
• Breaking: No longer automatically ignore .git when using --hidden with vcs ignore enabled. This reverts the change in v9.0.0. While this feature was often useful, it also broke some existing workflows, and there wasn't a good way to opt out of it. And there isn't really a good way for us to add a way to opt out of it. And you can easily get similar behavior by adding .git/ to your global fdignore file. See #1457.

Bugfixes

• Respect NO_COLOR environment variable with --list-details option. (#1455)
• Fix bug that would cause hidden files to be included despite gitignore rules if search path is "." (#1461, BurntSushi/ripgrep#2711).
• aarch64 builds now use 64k page sizes with jemalloc. This fixes issues on some systems, such as ARM Macs that have a larger system page size than the system that the binary was built on. (#1547)
• Address CVE-2024-24576, by increasing minimum rust version.

Changes

• Minimum supported rust version is now 1.77.2

Performance

• This release is all about performance, which has been significantly improved, both due to optimizations in the underlying ignore crate (#1429 by @tavianator), and in fd itself (#1422, #1408, #1362 by @tavianator). Benchmark results show gains of 6-8x for full traversals of smaller directories (100k files) and up to 13x for larger directories (1M files).

• The default number of threads is now constrained to be at most 64. This should improve startup time on systems with many CPU cores. (#1203, #1410, #1412, #1431) - @tmccombs and @tavianator

• New flushing behavior when writing output to stdout, providing better performance for TTY and non-TTY use cases, see #1452 and #1313 (@tavianator).

Features

• Support character and block device file types, see #1213 and #1336 (@cgzones)
• Breaking: .git/ is now ignored by default when using --hidden / -H, use --no-ignore / -I or --no-ignore-vcs to override, see #1387 and #1396 (@skoriop)

Bugfixes

• Fix NO_COLOR support, see #1421 (@acuteenvy)

Other

• Fixed documentation typos, see #1409 (@marcospb19)

Thanks

Special thanks to @tavianator for his incredible work on performance in the ignore crate and fd itself.

Bugfixes

• -1 properly conflicts with the exec family of options.
• --max-results overrides -1
• --quiet properly conflicts with the exec family of options. This used to be the case, but broke during the switch to clap-derive
• --changed-within now accepts a space as well as a "T" as the separator between date and time (due to update of chrono dependency)

Other

• Many dependencies were updated
• Some documentation was updated and fixed

Features

• Add flag --no-require-git to always respect gitignore files, see #1216 (@vegerot)

Bugfixes

• Fix logic for when to use global ignore file. There was a bug where the only case where the global ignore file wasn't processed was if --no-ignore was passed, but neither --unrestricted nor --no-global-ignore-file is passed. See #1209