find and md5sum not yielding any output (find -o limit?)

Question

I ran into a situation that I don't really understand. I have a bunch of backup files in a recursive structure of which I want to calculate the md5. When I add some additional file extensions the process exits (exit code 0) without yielding any output.

find . -type f -iname "*.3gp" -o -iname "*.avi" -o -iname "*.mov" -o -iname "*.mp4" -print0 | xargs -0 md5sum
find . -type f -iname "*.3gp" -o -iname "*.avi" -o -iname "*.mov" -o -iname "*.mp4" -o -iname "*.mpg" -print0 | xargs -0 md5sum

The first one works fine, the second one doesn't yield any output. I've even tried it in a directory where there are no mpg files, same behavior.

Is there a limit on the number of arguments to find? I am running OSX and I installed md5sum from Macports.

Extra information

There seems to be something odd with the pipe and I'm inclined to blame the filenames. Further investigation in another folder shows me that the find command seems to work and there are 129 video files, 1 of which is .mpg. When I try the find+md5sum it returns after only 1 file. I ran a similar command in other folder that only contains pictures and it worked fine (found 80k files, yield 80k hashes).

Pictures@2006$ find . -type f -iname "*.3gp" -o -iname "*.avi" -o -iname "*.mov" -o -iname "*.mp4" | wc -l
 128
Pictures@2006$ find . -type f -iname "*.3gp" -o -iname "*.avi" -o -iname "*.mov" -o -iname "*.mp4" -o -iname "*.mpg" | wc -l
 129
Pictures@2006$ find . -type f -iname "*.3gp" -o -iname "*.avi" -o -iname "*.mov" -o -iname "*.mp4" -o -iname "*.mpg" -print0 | xargs -0 md5sum
 c21a78f2b2d5ca773b47647315ad91f8  ./pending photos/Video [%]/P007.MPG
Pictures@2006$

I also noticed that the second filename to process contained punctuation, a plus sign and non-ascii characters. Is it possible that the error may be due to file naming? Is there any workaround?

/Esplai/+Nou/20060604 Dinar d'últim dia d'esplai[Barbacoa al torrent de l'Escaiola]/MVI_7702.AVI

Answer 1

If the operator expressions in a find command are not separated by -o (meaning or) or -a (and), there's an implicit -a between them. And -a binds more tightly than -o, so

find . -type f -iname "*.3gp" -o -iname "*.avi" -o -iname "*.mov" \
     -o -iname "*.mp4" -o -iname "*.mpg" -print0 

is going to be parsed as

find . '(' -type f -a -iname "*.3gp" ')' -o -iname "*.avi" -o -iname "*.mov" \
     -o -iname "*.mp4" -o '(' -iname "*.mpg" -a -print0 ')'

So the -print0 will only print files (and directories!) matching *.mpg. That's why, with this command, you're only processing one file rather than the 129 files you expect.

You can use parentheses (surrounded by quotes because they're also special to the shell) to change the grouping:

find . -type f '(' -iname "*.3gp" -o -iname "*.avi" -o -iname "*.mov" \
     -o -iname "*.mp4" -o -iname "*.mpg" ')' -print0 | xargs -0 md5sum

Answer 2

Let me first mention that -print0 is non standard and not the best solution. Better is to use "execplus", e.g.

find dir -type f -exec cmd {} +

Your main problem however is that the operators have precedence and your -print is "anded" with the last -name primary only.

So the right method is to put the -o red primaries in parenthesis:

find dir ( -name '*.x1' -o -name '*.x2' ) -exec cmd {} +

You may of course add more -o -type operators if you need.

Answer 3

I think that you have a file in the current directory that matches the "*.mpg" wildcard, and expands to something that confuses find, e.g. ... -o iname A Movie about Birds.mpg. If you use single quotes, e.g. '*.mpg' rather than double quotes "*.mpg", wildcard expansion will not be done, and find will see *.mpg.

To see what is happening, either put an echo at the beginning (echo find ...), or pipe the result to | tr "\000" "\n"