find: -exec vs xargs (aka Why does "find | xargs basename" break?)

Question

I was trying to find all files of a certain type spread out in subdirectories, and for my purposes I only needed the filename. I tried stripping out the path component via basename, but it did't work with xargs:

$ find . -name '*.deb' -print | xargs basename 
basename: extra operand `./pool/main/a/aalib/libaa1_1.4p5-37+b1_i386.deb'
Try `basename --help' for more information.

I get the same thing (exactly the same error) with either of these variations:

$ find . -name '*.deb' -print0 | xargs -0 basename 
$ find . -name '*.deb' -print | xargs basename {}

This, on the other hand, works as expected:

$ find . -name '*.deb' -exec basename {} \;
foo
bar
baz

This happens on up-to-date Cygwin and Debian 5.0.3. My diagnosis is that xargs is for some reason passing two input lines to basename, but why? What's going on here?

Solution 1

Because basename wants just one parameter... not LOTS of. And xargs creates a lot of parameters.

To solve your real problem (only list the filenames):

 find . -name '*.deb' -printf "%f\n"

Which prints just the 'basename' (man find):

 %f     File's name with any leading directories
        removed (only the last element).

Solution 2

Try this:

find . -name '*.deb' | xargs -n1 basename

Solution 3

xargs can be forced to just pass one argument as well...

find . -name '*.deb' -print | xargs -n1 basename

This works, however the accepted answer is using find in a more appropriate way. I found this question searching for xargs basename problems as I'm using another command to get a list of file locations. The -n1 flag for xargs was the ultimate answer for me.

Solution 4

basename only accepts a single argument. Using -exec works properly because each {} is replaced by the current filename being processed, and the command is run once per matched file, instead of trying to send all of the arguments to basename in one go.

View more solutions

Comments

[akira]

this is not the explanation, this is a workaround. and the workaround is as good as just calling 'basename' via -exec for any file found.

[quack quixote]

+1 ... while not an explanation, this would lead me to investigate the xargs switch you show, which would eventually lead me to the forehead-slapping motion i just used reading akira's and john t's answers...

[quack quixote]

oooh.... /slaps forehead again/ i think i need a "find for dummies" book...

[Ryan C. Thompson]

This is how I do it. I don't feel like learning all the ins and outs of the find command, so I only use it for finding and listing files, and I use xargs for everything else.

[WindowsMaker]

i thought the point of xargs is that it creates a list of arguments and feeds each to the command that comes after? otherwise what's the difference between that and find . -name '*.deb' | basename ?

[bishop]

GNU basename now has a -a option: "support multiple arguments and treat each as a name".

[8bittree]

@WindowsMaker xargs converts stdin to command arguments. In a way, it's the opposite of echo, which converts its arguments to stdout. The difference between find ... | xargs -n1 basename or find ... | xargs basename -a and find ... | basename is that the former two will work with implementations of basename that ignore stdin.

[soMuchToLearnAndShare]

@WindowsMaker, if my understanding is that you use -I {} xargs option, like this | xargs -I {} basename -s .json {} | , it will feed each argument separately