False positive warning for find

[arth1]

This warning is invalid:

find trunk -type f -name '*.gcda' -print0 |\
^-- Don't use find | xargs cmd. find -exec cmd {} + handles whitespace.

find ... -print0 | xargs -0 ... handles whitespace.
And piping to xargs can be orders of magnitude(!) faster than using -exec {}, which is single-threaded, blocking, and not file system cache friendly:

$ rm -f /tmp/foo; time find /usr/lib64 -type f 2>/dev/null -print0 | xargs -0 md5sum >>/tmp/foo 2>/dev/null

real 0m3.304s
user 0m2.971s
sys 0m0.373s
$ rm -f /tmp/foo; time find /usr/lib64 -type f 2>/dev/null -exec md5sum >>/tmp/foo {} ;

real 0m14.842s
user 0m4.928s
sys 0m5.174s

If anything, the warning should be the other way - never use exec {} if there's an opportunity to use -print0 | xargs -0

[koalaman]

You're quite right: it should not be complaining when using -print0. I'll fix this shortly.

You did ignore its advice and used "-exec {} ;" instead of "-exec {} +" though, which is why your alternative is slower :P

[arth1]

Thanks for fixing!

Actually, even with exec {} +, find is notably slower than xargs. The main reason is that it blocks the find process while doing a block of exec.

$ time find . -type f -exec ls -d >/dev/null {} +

real 0m3.790s
user 0m2.368s
sys 0m1.346s
$ time find . -type f -exec ls -d >/dev/null {} +

real 0m3.779s
user 0m2.404s
sys 0m1.295s
$ time find . -type f -exec ls -d >/dev/null {} +

real 0m3.806s
user 0m2.412s
sys 0m1.315s
$ time find . -type f -print0 | xargs -0 ls -d >/dev/null

real 0m3.122s
user 0m2.644s
sys 0m1.372s
$ time find . -type f -print0 | xargs -0 ls -d >/dev/null

real 0m3.157s
user 0m2.628s
sys 0m1.343s
$ time find . -type f -print0 | xargs -0 ls -d >/dev/null

real 0m3.185s
user 0m2.633s
sys 0m1.387s
$ find . -type f | wc -l
175730

... and never mind that you can also pass -P NNN to xargs :P

$ time find . -type f -print0 | xargs -0 -P 8 ls -d >/dev/null

real 0m0.810s
user 0m3.503s
sys 0m1.585s

[koalaman]

Shellcheck no longer warns about piping find to xargs -0