Crates.io | tac-k-lib |
lib.rs | tac-k-lib |
version | 0.3.3 |
source | src |
created_at | 2024-05-30 12:04:03.404507 |
updated_at | 2024-05-31 20:52:38.890841 |
description | Print lines from file in reverse order |
homepage | https://github.com/my4ng/tack |
repository | https://github.com/my4ng/tack |
max_upload_size | |
id | 1256806 |
size | 22,838 |
Tack is a fork of the tac crate. It is available as a standalone binary, tac-k, and as a library, tac-k-lib.
Tack is a high-performance, simd-accelerated, cross-platform rewrite of the GNU tac
utility from Coreutils, released under MIT/Apache-2.0 licenses. tac
reads input from a file (or from stdin
, but see below) and then prints it line-by-line backwards.
This tac
implementation uses simd-acceleration for new line detection and utilizes memory-mapped files on all supported operating systems. It is additionally written in rust for maximum integrity and safety.
tac
anyway?Good question. Try grepping through a multi-gigabyte web access log file in reverse chronological order (tac --line-buffered access.log | grep foo
) and then get back to me.
Usage: tac [OPTIONS] [FILE]...
Arguments:
[FILE]... Files to be reversed.
Read from stdin if it is `-` or not specified.
Options:
-s, --separator <BYTE> Use BYTE as the separator instead of newline.
Only single-byte character is supported.
--line-buffered Always flush output after each line
-h, --help Print help
-V, --version Print version
Tack reads lines from any combination of stdin
and/or zero or more files and writes the lines to the output in reverse order.
$ echo -e "hello\nworld" | tac
world
hello
$ echo -e "hello\nworld" | tac --separator=o
rld
wohello%
The tack binary may be built installed via cargo, the rust package manager:
cargo install tac-k --locked
or installed with pre-built binaries via cargo-binstall
:
cargo binstall tac-k --locked
The tack library can be added to your project via:
cargo add tac-k-lib
This implementation of tac
uses SIMD instruction sets (AVX2, NEON) to accelerate the detection of new lines if available. The usage of memory-mapped files additionally boosts performance by avoiding slowdowns caused by context switches when reading from the input if speculative execution mitigations are enabled. It is significantly (2.55x if mitigations disabled, more otherwise) faster than the version of tac
that ships with GNU Coreutils, in addition to being more liberally licensed.
To obtain maximum performance:
tac
. e.g. instead of running cat /usr/share/dict/words | tac
, run tac /usr/share/dict/words
directly. Because tac
by definition must reach the end-of-file before it can emit its input with the lines reversed, if you use tac
's stdin
interface (e.g. cat foo | tac
), it must buffer all stdin
input before it can begin to process the results. tac
will try to buffer in memory, but once it exceeds a certain high-water mark (currently 4 MiB), it switches to disk-based buffering (because it can't know how large the input is or if it will end up exceeding the available free memory).tac
at the start of a pipeline where possible. Even if you can guarantee that the input to tac
will not exceed the in-memory buffering limit (see above), tac
is almost certainly faster than any other command in your pipeline, and if you are going to reverse the output, you will benefit most if you reverse it from the start, unless you are always going to run the command to completion. For example, instead of running grep foo /var/log/nginx/access.log | tac
, run tac /var/log/nginx/access.log | grep foo
. This will (significantly) reduce the amount of time/work before the first n matches are reported (because the file is first quickly reversed then searched in the desired order, vs slowly searched in its entirety and only then are the results reversed).tac --line-buffered
) if tac is piping into another command rather than writing to the tty directly. This gives you "live" streaming of results and lets you terminate much sooner if you're only looking for the first n matches. e.g. tac --line-buffered access.log | grep foo
will print its first match much, much sooner than tac access.log | grep foo
would.tac --line-buffered access.log | grep --line-buffered foo | head -n2
.Tack is licensed under either MIT and Apache-2.0 at your option.
Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in this crate by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.
As an open source project, Tack would not exist without the tireless efforts of its various contributors - see CONTRIBUTORS.md for full details.