I want to find the largest 10 folders in a current directory with human-readable folder size? The examples I've seen, all are listing files along with the folders. I just want to see which folders are using most spaces, because each folder is being used by different users.
Following is the code I tried
du -hs * | sort -rh | head -10
du -hs * in your code contains bare * globbing pattern. Your shell expands it to filenames of almost all files in the current directory, so in general du gets multiple arguments: directories and non-directories.
"Almost", because * doesn't match names starting with a dot (.). These are hidden. In Bash you can change this behavior by invoking shopt -s dotglob.
To match directories only, append a trailing slash to the pattern:
du -hs */ | sort -rh | head -10
This simple solution will fail if there are too many directories. The error will be argument list too long.
findA generic way to do something to files matching some criteria is find. A find command solving your case is more complicated than the above. I'm posting it here mainly as a starting point for users who want to add more criteria (especially criteria that cannot be easily included in the solution with */):
find . -type d ! -name . -prune -exec du -hs {} + | sort -rh | head -10
It means: starting from . find all files of the type directory (-type d) that are not . (! -name .). Do not descend into them (-prune). Execute du -hs zero or more times, passing possibly multiple pathnames of found files to a single invocation of du (-exec du … {} +).
This solution won't trigger argument list too long because find is smart enough to run more than one du, if needed, so each du gets a list that is not too long. If you need a robust solution (e.g. in a script that should work without supervision) then prefer find over globbing.
ncduAn interactive approach is ncdu. It's true it shows non-directories along with directories, but you can easily tell them apart (observe leading slashes) or toggle "dirs before [other] files when sorting" (t).
