Process Substitution

Bash · Reference cheat sheet

📋 Overview

Process substitution (<(…) and >(…)) exposes a command’s output or input as a temporary filename (usually /dev/fd/N). It lets commands that expect files work with streams—and avoids pipeline subshell issues when reading into the current shell.

🔧 Core concepts

Requires bash/zsh/ksh support and /dev/fd (or named temps). Not POSIX sh.

💡 Examples

Compare outputs:

diff -u <(sort a.txt) <(sort b.txt)
comm <(ls dir1) <(ls dir2)

Read in current shell:

while IFS= read -r line; do
  echo "$line"
done < <(curl -fsS "$url")

Multiple inputs:

paste <(cut -f1 data.tsv) <(cut -f2 data.tsv)

Tee-like with >():

echo hello > >(tee saved.txt)
# or
cmd | tee >(sha256sum >&2) > out.bin

mapfile:

mapfile -t lines < <(grep -v '^#' config)

⚠️ Pitfalls

• Not available in POSIX sh or some restricted environments.
• The path is ephemeral; don’t store it for later use after the command ends.
• <(cmd) runs asynchronously; ensure consumers read fully.
• Combining with sudo may break /dev/fd visibility—run the whole pipeline elevated carefully.
• Prefer process substitution over cmd >tmp && use tmp for cleanliness, but temp files still win for huge reusable data.
• Quoting: "<(cmd)" becomes a literal—don’t quote the substitution form.

🔗 Related

• pipes_redirection.md
• subshells.md
• loops.md
• debugging.md
• scripts_best_practices.md