Linux Command Reference ยท Bash Syntax
Bash process substitution
Reference for Bash process substitution: use the output of a command as a file with <(cmd), feed a command with >(cmd), and avoid temporary files and pipe subshells.
Process substitution lets a command's output look like a file. <(command) runs the command and gives you a file name (usually /dev/fd/63) from which you can read its output. >(command) does the opposite: it gives a file name that, when written to, feeds the command's input.
It is useful when a program insists on file arguments, such as diff, comm or paste, and you have commands instead of files. It also keeps a loop in the current shell, so variables survive.
diff <(sort a.txt) <(sort b.txt) while read -r line; do ...; done < <(some_command) tar cf >(gzip > out.tgz) dir
Try it
Press Run (or Ctrl + Enter) to see the result here.
Program input ยท what you'd type while the program runs
Type the answers before you press Run, one per line. Each time the program reads a line (with read or cat), it gets the next line from this box.
Press Run (or Ctrl + Enter) to see the result here.
Program input ยท what you'd type while the program runs
Type the answers before you press Run, one per line. Each time the program reads a line (with read or cat), it gets the next line from this box.
Syntax
<(list) # read the output of list through a file name >(list) # write to the input of list through a file name
| Form | Description |
|---|---|
<(command) |
Replaced by the name of a file or named pipe that provides the command's standard output. Use it wherever a file to read is expected. |
>(command) |
Replaced by a file name; whatever is written there becomes the command's standard input. |
< <(command) |
Redirect standard input from the substituted file; note the space between the two < characters. |
Returns: a path such as /dev/fd/63 (or a path in /tmp on systems without /dev/fd). The commands run asynchronously, in parallel with the main command.
Errors / edge cases: the feature exists in Bash, Zsh and Ksh but not in POSIX sh or dash. It requires /dev/fd or named pipes. The substituted path is a pipe, so it cannot be seeked or read twice.
Usage notes
- Single use. The data is a stream: a program that reads the file twice or seeks backwards (some editors,
unzip,file) will fail. Use a temporary file for those. - Space in
< <(.<<(cmd)is a syntax error because<<is the here-document operator. Write< <(cmd). - Better than a pipe into
while. Because the loop is not part of a pipeline, assignments inside it stay visible afterwards. See Pipelines. >(...)timing. The shell does not wait for>(command)processes in older Bash versions. Output files may be incomplete immediately afterwards; Bash 4.4 and later let youwaitfor the last one, and a shortsleepis only a demo shortcut.- Exit status is lost. The status of the substituted command is not available in
$?. If it matters, write its output to a file or use a coprocess. - Not in POSIX scripts. For
#!/bin/shscripts, use named pipes (mkfifo) or temporary files. - Several inputs:
diff <(cmd1) <(cmd2)works because each substitution has its own file name; the same pattern fitscomm,joinandpaste. - Quote the whole expression when you hand it to
eval,sshor another shell: the substitution is expanded by the shell that parses it.
Related
- Command substitution, Redirection, Pipelines,
tee,diff,comm - Lessons: Pipes, Subshells, Advanced patterns