What you wrote sends a list of newline separated file names (and paths) to rm
, but rm doesn't know what to do with that input. It's only expecting command line parameters.
xargs
takes input, usually separated by newlines, and places them on the command line, so adding xargs makes what you had work:
find . -name .svn | xargs rm -fr
xargs
is intelligent enough that it will only pass as many arguments to rm
as it can accept. Thus, if you had a million files, it might run rm
1,000,000/65,000 times (if your shell could accept 65,002 arguments on the command line {65k files + 1 for rm + 1 for -fr}).
As persons have adeptly pointed out, the following also work:
find . -name .svn -exec rm -rf {} \;
find . -depth -name .svn -exec rm -fr {} \;
find . -type d -name .svn -print0|xargs -0 rm -rf
The first two -exec
forms both call rm
for each folder being deleted, so if you had 1,000,000 folders, rm
would be invoked 1,000,000 times. This is certainly less than ideal. Newer implementations of rm
allow you to conclude the command with a +
indicating that rm
will accept as many arguments as possible:
find . -name .svn -exec rm -rf {} +
The last find/xargs version uses print0, which makes find generate output that uses \0
as a terminator rather than a newline. Since POSIX systems allow any character but \0
in the filename, this is truly the safest way to make sure that the arguments are correctly passed to rm
or the application being executed.
In addition, there's a -execdir
that will execute rm
from the directory in which the file was found, rather than at the base directory and a -depth
that will start depth first.