xargs in depth

Pipes feed standard input. That works for grep, sort or wc, which read their input. But rm, mkdir, cp or ls ignore standard input completely; they only look at their arguments. find . -name '*.tmp' | rm does nothing useful. xargs is the adapter: it reads words from its standard input and appends them as arguments to the command you give it. You met it briefly in lesson 37; this lesson covers the options that make it safe.

~% printf 'a\nb\nc\n' | xargs echo rm
rm a b c
~% echo "1 2 3 4 5" | xargs -n 2 echo
1 2
3 4
5
~% printf 'a.txt\nb.txt\n' | xargs -t -n 1 echo
echo a.txt
a.txt
echo b.txt
b.txt

By default xargs splits its input on any whitespace (spaces and newlines) and packs as many words as fit into one command line, which is fast: one rm for a thousand files instead of a thousand rm. -n N limits each run to N arguments. -t prints each command to standard error before running it, a great way to see what will happen. Putting echo in front of the real command is the classic dry run: read the output, then remove the echo.

Putting the argument in the middle: -I

Normally the words go at the end of the command. With -I{} each input line becomes one run, and every {} in the command is replaced with that line. This is how you copy each listed file into a directory, or build a new name from an old one. Because it works line by line, spaces inside a line are kept.

~% cat list.txt
/etc/hostname
/etc/fstab
~% xargs -I{} cp {} backup/ < list.txt
~% ls *.log | xargs -I{} mv {} {}.old

Spaces in file names: -print0 and -0

The default splitting is a trap with names like old file.tmp: xargs sees two words, old and file.tmp, and rm tries to delete two files that do not exist. The robust fix is to separate names with the NUL character, the one byte that can never appear in a file name. find -print0 writes names that way, and xargs -0 reads them that way. Make it a reflex whenever find feeds xargs.

~% find cache -name '*.tmp' | xargs rm
rm: can't remove 'cache/old': No such file or directory
rm: can't remove 'file.tmp': No such file or directory
~% find cache -name '*.tmp' -print0 | xargs -0 rm
~% find cache -name '*.tmp'
~%