xargs

# 1) The basic idea — pipe stdin → command arguments
echo 'a b c' | xargs touch                         # touch a b c
ls *.tmp | xargs rm                                  # delete .tmp files (DANGEROUS — see -0)
find . -name '*.log' | xargs cat                     # cat all log files (NEWLINE bug; see -0)

# By default xargs splits on whitespace AND newlines — filenames with spaces break it.

# 2) The safe pattern — null-delimited
find . -name '*.log' -print0 | xargs -0 cat
find . -name '*.log' -print0 | xargs -0 rm

# -print0 separates with NUL bytes; xargs -0 reads them. Handles ANY filename safely.

# 3) Limit arguments per command
echo 'a b c d e f' | xargs -n 2 echo
# echo a b
# echo c d
# echo e f

# Useful for batch ops:
find . -name '*.jpg' -print0 | xargs -0 -n 1 -I {} convert {} {}.webp

# 4) Replace token with -I
ls | xargs -I {} mv {} /backup/{}.bak
find . -name '*.md' -print0 | xargs -0 -I FILE wc -l FILE

# 5) Parallel execution
find . -name '*.jpg' -print0 | xargs -0 -n 1 -P 4 -I {} convert {} {}.webp
# -P 4 = 4 jobs at once; -n 1 = one filename per job.
# Great for CPU-bound batch work; check available cores with 'nproc'.

# 6) Print the commands without running
echo 'a b' | xargs -t echo
# echo a b
# a b   (the actual output)

find . -name '*.tmp' -print0 | xargs -0 -t rm
# Lets you preview what xargs WOULD run.

# 7) Confirm before each run (-p)
ls *.tmp | xargs -p rm
# rm a.tmp b.tmp c.tmp ?...    (type 'y' to proceed)

# 8) Empty input handling — -r / --no-run-if-empty
find . -name '*.no-match' | xargs rm                  # would run 'rm' with no args (sometimes errors)
find . -name '*.no-match' | xargs -r rm                # skips the command entirely

# GNU xargs has -r; BSD xargs (macOS) doesn't but treats empty input safely. Install gxargs via brew.

# 9) Real-world recipes

# Find + ESLint changed files
git diff --name-only origin/main | xargs -r npx eslint

# Find + remove old log files (older than 30 days)
find /var/log -type f -mtime +30 -print0 | xargs -0 rm

# Search files for a pattern (skip binaries)
find . -type f -print0 | xargs -0 grep -lI 'TODO'

# Compress large files
find . -type f -size +10M -print0 | xargs -0 -n 1 -P 8 gzip

# Convert .jpg to .webp in parallel
find . -type f -name '*.jpg' -print0 | xargs -0 -n 1 -P 4 -I {} \\
    sh -c 'convert "$1" "${1%.jpg}.webp"' _ {}

# Remove untracked files (verify with -t first)
git ls-files --others --exclude-standard -z | xargs -0 -t rm

# 10) Read from a file
xargs -a files.txt -n 1 -I {} npm publish {}

# 11) Limit max-args for command-line length
ls /usr/bin | xargs -n 100 echo 'batch:'

# 12) Substitute multiple times in one command
docker images -q | xargs -n 1 -I {} sh -c 'docker inspect --format="{}{{.Created}}" {}'

# 13) parallel as an alternative
# GNU parallel has richer parallelism + auto-quoting
ls *.jpg | parallel convert {} {.}.webp
# parallel --citation     once to acknowledge
# parallel is amazing but heavier than xargs; both have their place.

# 14) Common bugs
# • Filenames with spaces blow up — ALWAYS use -print0 / -0
# • Empty input running command with no args — use -r (GNU)
# • Quoting nightmares with -I — wrap shell commands in 'sh -c ...' explicitly
# • Race conditions on parallel writes — make sure each job writes to a UNIQUE target
# • Long arg list exceeds ARG_MAX → xargs splits automatically; -n controls batch size
# • macOS xargs differs from GNU xargs — install GNU via 'brew install findutils' as gxargs
# • Forgetting that xargs INHERITS errors silently — use 'set -e' in scripts; check exit codes
# • Piping ls output is fragile — use 'find ... -print0' for safety
# • -P without -n 1 → xargs batches files per process; subtle for some commands

find . -name '*.tmp' | xargs rm
ls *.txt | xargs -I{} cp {} backup/