rust-parallel 0.4.1

Fast command line app in rust/tokio to execute commands in parallel. Similar interface to GNU parallel or xargs.
rust-parallel-0.4.1 is not a library.

rust-parallel

Command-line utility to execute commands in parallel and aggregate their output.

Similar interface to GNU Parallel or xargs but implemented in rust and tokio.

Always executes 1 process for each input line, similar to xargs -n1 or -L1 options.

Being written in asynchronous rust it is quite fast - see benchmarks.

Crates.io CI workflow

Contents:

Usage:

$ rust-parallel --help
Execute commands in parallel

By Aaron Riekenberg <aaron.riekenberg@gmail.com>

https://github.com/aaronriekenberg/rust-parallel
https://crates.io/crates/rust-parallel

Usage: rust-parallel [OPTIONS] [COMMAND_AND_INITIAL_ARGUMENTS]...

Arguments:
  [COMMAND_AND_INITIAL_ARGUMENTS]...
          Optional command and initial arguments to run for each input line

Options:
  -i, --input <INPUT>
          Input file or - for stdin.  Defaults to stdin if no inputs are specified

  -j, --jobs <JOBS>
          Maximum number of commands to run in parallel, defauts to num cpus

          [default: 12]

  -0, --null-separator
          Use null separator for reading input instead of newline

  -s, --shell
          Use shell for running commands.

          If $SHELL environment variable is set use it else use /bin/bash.

          Each input line is passed to $SHELL -c <line> as a single argument.

  -h, --help
          Print help (see a summary with '-h')

  -V, --version
          Print version

Installation:

Recommended:

  1. Download a pre-built release from Github Releases.
  2. Extract the executable and put somewhere in your $PATH.

For manual installation/update:

  1. Install Rust
  2. Install the latest version of this app from crates.io:
$ cargo install rust-parallel
  1. The same cargo install rust-parallel command will also update to the latest version after initial installation.

Demos:

  1. Small demo of 5 echo commands
  2. Debug Logging
  3. Specifying command and intial arguments on command line
  4. Using awk to form complete commands
  5. Using as part of a shell pipeline
  6. Working on a set of files from find command
  7. Reading multiple inputs
  8. Calling a bash function

Small demo of 5 echo commands.

With -j5 all 5 commands are run in parallel. With -j1 commands are run sequentially:

$ cat >./test <<EOL
echo hi
echo there
echo how
echo are
echo you
EOL

$ cat test | rust-parallel -j5
are
hi
there
how
you

$ cat test | rust-parallel -j1
hi
there
how
are
you

Debug logging.

Set environment variable RUST_LOG=debug to see debug output.

This will log information about command line arguments and commands being run.

Recommend enabling debug logging for all demos to understand what is happening in more detail.

$ cat test | RUST_LOG=debug rust-parallel

Specifying command and intial arguments on command line:

Here md5 -s will be prepended to each input line to form a command like md5 -s aal

$ head -100 /usr/share/dict/words | rust-parallel md5 -s
MD5 ("aal") = ff45e881572ca2c987460932660d320c
MD5 ("A") = 7fc56270e7a70fa81a5935b72eacbe29
MD5 ("aardvark") = 88571e5d5e13a4a60f82cea7802f6255
MD5 ("aalii") = 0a1ea2a8d75d02ae052f8222e36927a5
MD5 ("aam") = 35c2d90f7c06b623fe763d0a4e5b7ed9
MD5 ("aa") = 4124bc0a9335c27f086f24ba207a4912
MD5 ("a") = 0cc175b9c0f1b6a831c399e269772661
MD5 ("Aani") = e9b22dd6213c3d29648e8ad7a8642f2f
MD5 ("Aaron") = 1c0a11cc4ddc0dbd3fa4d77232a4e22e
MD5 ("aardwolf") = 66a4a1a2b442e8d218e8e99100069877

Using awk to form complete commands:

$ head -100 /usr/share/dict/words | awk '{printf "md5 -s %s\n", $1}' | rust-parallel
MD5 ("Abba") = 5fa1e1f6e07a6fea3f2bb098e90a8de2
MD5 ("abaxial") = ac3a53971d52d9ce3277eadf03f13a5e
MD5 ("abaze") = 0b08c52aa63d947b6a5601ee975bc3a4
MD5 ("abaxile") = 21f5fc27d7d34117596e41d8c001087e
MD5 ("abbacomes") = 76640eb0c929bc97d016731bfbe9a4f8
MD5 ("abbacy") = 08aeac72800adc98d2aba540b6195921
MD5 ("Abbadide") = 7add1d6f008790fa6783bc8798d8c803
MD5 ("abb") = ea01e5fd8e4d8832825acdd20eac5104

Using as part of a shell pipeline.

stdout and stderr from each command run are copied to stdout/stderr of the rust-parallel process.

$ head -100 /usr/share/dict/words | rust-parallel md5 -s | grep -i abba
MD5 ("Abba") = 5fa1e1f6e07a6fea3f2bb098e90a8de2
MD5 ("abbacomes") = 76640eb0c929bc97d016731bfbe9a4f8
MD5 ("abbacy") = 08aeac72800adc98d2aba540b6195921
MD5 ("Abbadide") = 7add1d6f008790fa6783bc8798d8c803

Working on a set of files from find command.

The -0 option works nicely with find -print0 to handle filenames with newline or whitespace characters:

$ mkdir testdir

$ touch 'testdir/a b' 'testdir/b c' 'testdir/c d'

$ find testdir -type f -print0 | rust-parallel -0 gzip -f -k

$ ls testdir
'a b'  'a b.gz'  'b c'  'b c.gz'  'c d'  'c d.gz'

Reading multiple inputs.

By default rust-parallel reads input from stdin only. The -i option can be used 1 or more times to override this behavior. -i - means read from stdin, -i ./test means read from the file ./test:

$ cat >./test <<EOL
foo
bar
baz
EOL

$ head -5 /usr/share/dict/words | rust-parallel -i - -i ./test echo
A
aalii
aa
a
aal
bar
foo
baz

Calling a bash function.

Use -s shell mode so that each input line is passed to /bin/bash -c as a single argument:

$ doit() {
  echo Doing it for $1
  sleep 2
  echo Done with $1
}

$ export -f doit

$ cat >./test <<EOL
doit 1
doit 2
doit 3
EOL

$ cat test | rust-parallel -s
Doing it for 1
Done with 1
Doing it for 3
Done with 3
Doing it for 2
Done with 2

Benchmarks:

See the wiki page for benchmarks.

Features:

  • Use only safe rust.
  • Use only asynchronous operations supported by tokio, do not use any blocking operations. This includes writing to stdout and stderr.
  • Support arbitrarily large number of input lines, avoid O(number of input lines) memory usage. In support of this:
    • tokio::sync::Semaphore is used carefully to limit the number of commands that run concurrently. Do not spawn tasks for all input lines immediately to limit memory usage.
  • Support running commands on local machine only, not on remote machines.

Tech Stack: