60

When looping recursively through folders with files containing spaces the shell script I use is of this form, copied from the internet:

    while IFS= read -r -d $'\0' file; do
      dosomethingwith "$file"        # do something with each file
    done < <(find /bar -name *foo* -print0)

I think I understand the IFS bit, but I don't understand what the '< <(...)' characters mean. Obviously there's some sort of piping going on here.

It's very hard to Google "< <" or "<(", you see. I tried "angle bracket parenthesis" and "less-than parenthesis" but didn't find anything.

that other guy
  • 116,971
  • 11
  • 170
  • 194
stib
  • 3,346
  • 2
  • 30
  • 38

4 Answers4

72

<() is called process substitution in the manual, and is similar to a pipe but passes an argument of the form /dev/fd/63 instead of using stdin.

< reads the input from a file named on command line.

Together, these two operators function exactly like a pipe, so it could be rewritten as

find /bar -name *foo* -print0 | while read line; do
  ...
done
Josh Lee
  • 171,072
  • 38
  • 269
  • 275
  • 10
    it's not the same if you must not start a subshell – knittl Mar 14 '10 at 18:18
  • 7
    +1 because you have the correct name for the notation. As @knittl points out, it is not quite identical to your rewrite because the loop will run in a sub-shell, and any changes made in the variables by the loop only affect the sub-shell, not the main script. You can deal with that by redirecting the output of 'find' into '`{ while ...; do ...; done; ...rest of script...; }`', using the braces to group the whole of the rest of the script - loop and other material - into a single sub-shell. – Jonathan Leffler Mar 14 '10 at 18:39
  • 1
    thanks for that. I can feel my skull expanding over the part of the brain that does shell scripting. Having a precise term to google makes it much easier. Now I'm going to go find out what a subshell is. – stib Mar 15 '10 at 10:46
25

<( command ) is process substitution. Basically, it creates a special type of file called a "named pipe," then redirects the output of the command to be the named pipe. So for example, suppose you want to page through a list of files in an extra-big directory. You could do this:

ls /usr/bin | more

Or this:

more <( ls /usr/bin )

But NOT this:

more $( ls /usr/bin )

The reason for this becomes clear when you investigate further:

~$ echo $( ls /tmp )
gedit.maxtothemax.436748151 keyring-e0fuHW mintUpdate orbit-gdm orbit-maxtothemax plugtmp pulse-DE9F3Ei96ibD pulse-PKdhtXMmr18n ssh-wKHyBU1713 virtual-maxtothemax.yeF3Jo
~$ echo <( ls /tmp )
/dev/fd/63
~$ cat <( ls /tmp )
gedit.maxtothemax.436748151
keyring-e0fuHW
mintUpdate
orbit-gdm
orbit-maxtothemax
plugtmp
pulse-DE9F3Ei96ibD
pulse-PKdhtXMmr18n
ssh-wKHyBU1713
virtual-maxtothemax.yeF3Jo

/dev/fd/whatever acts like a text file with the output of the command between the parenthesis.

Max E.
  • 1,837
  • 13
  • 15
  • `more <( ls /usr/bin )` shows `/dev/fd/11 is not a regular file (use -f to see it)` on my zsh and shows `/dev/fd/63 is not a regular file (use -f to see it)` on my bash. – aafulei Aug 03 '22 at 08:18
  • 2
    The second shell should be `more < <(ls /usr/bin)`, since `<(ls /usr/bin)` converts stdout to file and return file path as args and ` < ` receives file path and convert it to stdin. – Yari Aug 09 '22 at 03:42
4

< redirects to stdin.

<() seems to be some sort of a reverse pipe, as mentioned on the page:

find /bar -name *foo* -print0 | \
while IFS= read -r -d $'\0' file; do
  dosomethingwith "$file"        # do something with each file
done

will not work, because the while loop will be executed in a subshell, and you'll lose changes made in the loop

knittl
  • 246,190
  • 53
  • 318
  • 364
  • This answer would benefit from an x-ref to 'process substitution' and a URL such as http://www.gnu.org/software/bash/manual/bashref.html#Process-Substitution that explains it. The key point is the sub-shell and changes made to variables in the sub-shell. The other way of dealing with it is: `find ... | { while ...; do ...; done; ...rest of script...; }`, using the braces to run all the rest of the script in a sub-shell instead of just the while loop. – Jonathan Leffler Mar 14 '10 at 18:35
0

You use the process substitution construct (as per https://www.gnu.org/software/bash/manual/html_node/Process-Substitution.html#Process-Substitution) mostly when you need to pass a file to a command but using the output of another command instead of a file, thus avoiding the need to create and populate that file in advance.

For ex. say that you want to compare the output of 2 commands, you can do like this:

diff <(command1) <(command2)

Luca De Rugeriis
  • 11
  • 3