11.7 Command Substitution

The Bourne shell can redirect a command’s standard output back to the shell’s own command line. That is, you can use a command’s output as an argument to another command, or you can store the command output in a shell variable by enclosing a command in $().

This example stores a command’s output inside the FLAGS variable. The bold code in the second line shows the command substitution.

#!/bin/sh
FLAGS=$(grep ^flags /proc/cpuinfo | sed 's/.*://' | head -1)
echo Your processor supports:
for f in $FLAGS; do
    case $f in
        fpu)    MSG="floating point unit"
                ;;
        3dnow)  MSG="3DNOW graphics extensions"
                ;;
        mtrr)   MSG="memory type range register"
                ;;
        *)      MSG="unknown"
                ;;
    esac
    echo $f: $MSG
done

This example is somewhat complicated because it demonstrates that you can use both single quotes and pipelines inside the command substitution. The result of the grep command is sent to the sed command (more about sed in Section 11.10.3), which removes anything matching the expression .*:, and the result of sed is passed to head.

It’s easy to go overboard with command substitution. For example, don’t use $(ls) in a script, because using the shell to expand * is faster. Also, if you want to invoke a command on several filenames that you get as a result of a find command, consider using a pipeline to xargs rather than command substitution, or use the -exec option (both are discussed in Section 11.10.4).

note

The traditional syntax for command substitution is to enclose the command in backticks (``), and you’ll see this in many shell scripts. The $() syntax is a newer form, but it is a POSIX standard and is generally easier (for humans) to read and write.