Read

Command Substitution: $(cmd) and backticks

Command substitution captures the stdout of a command and splices it into the current command line.

today=$(date +%Y-%m-%d)        # modern form
today=`date +%Y-%m-%d`         # legacy backtick form
files=$(ls *.py | head -5)     # full pipelines allowed
echo "$(uname) on $(hostname)"

$() vs backticks

POSIX shells, bash, zsh, and dash all support $(). There's no reason to write backticks in new code, but you'll read them in older scripts.

How the shell implements it

1. Parse phase: see $(, find the matching ) (handling nesting, quotes, and heredocs), capture the inner string as a "command".
2. Execution phase (before the surrounding command runs):
   • pipe(2) — create a pipe.
   • fork(2) — child runs the inner command with stdout dup2-ed to the pipe's write end.
   • Parent reads the pipe to EOF and waitpids for the child.
3. Word splitting & trimming:
   • All trailing newlines are stripped (but not leading whitespace).
   • The remaining bytes are subject to word splitting on $IFS if the substitution is unquoted.
   • Inside "...", the result is one literal string (no splitting).

$ touch "a b.txt"
$ for f in $(ls); do echo "[$f]"; done    # WRONG: splits "a b.txt"
[a]
[b.txt]
$ for f in "$(ls)"; do echo "[$f]"; done  # ALSO WRONG: one giant token
[a b.txt]
$ ls | while read -r f; do echo "[$f]"; done  # OK
[a b.txt]

Exit status

The exit status of $(cmd) is the exit status of the inner command, but in the surrounding context it's only the substituted bytes that flow out — the exit code is lost unless you check $? immediately or store the substitution result in a variable assignment.