paperlined.org
apps > bash

document updated 21 days ago, on Jan 31, 2025

Often I want to diff the output of two COMMANDS. That is, I want to run command A, then run command B, then diff the output of both of those.

For example:

sort file1 > file1.sorted
sort file2 > file2.sorted
diff file1.sorted file2.sorted

Can this be done without writing the output to a temporary file?

Answer: YES!

The trick is to use Bash's process substitution feature:

diff <(sort file1) <(sort file2)

Bash changes that into:

diff /dev/fd/63 /dev/fd/62

echo diff <(sort file1) <(sort file2)

It does that ^^^ after first running the two sort commands, and sending their output to filehandles 62 and 63.

You can write to subprocesses too:

# list all .so libraries that are loaded when a Perl script runs
strace -e trace=open,openat -o >( grep '\.so' | grep -v ENOENT ) ./script_under_test.pl

That way, strace's output gets grepped, while leaving ./script_under_test.pl's STDOUT alone.

show me more!

Want to do more magic with process substitution? See the following examples:

the x=> example here
the tee example here
a few more interesting examples
some important details about how process substitution works
you don't HAVE to use process substitution... you can manually shuffle around file descriptors too, that's equivalent

multitee is something else to look into

show me more!

There's lots of other I/O redirection tricks. [1] [2]