Quickly checking if a directory is empty

[Anonymous Monk]

I need to check and see if certain directories are empty (without deleting them) as part of a larger program and I use the code below. Some of the directories have a lot of files so I thought I would optimize it by including the last if line to keep it from counting all files but it makes no difference in the time this sample code takes to run? Note I have about 10,000 files in the directory I am testing this on and it takes about 4 seconds for the code to complete with or without the last if line.

$start = time();
chdir "c:/temp";
while (<*>) {
  $num++;
  last if ($num==1); # makes no difference in speed
}
$stop = time();
$elapsed=$stop-$start;
print "elapsed = $elapsed seconds\n";

What am I missing and/or is there a better way to do this? Again, all I need to know is if the directory is empty or not without deleting anything.

[Fletch]

Rather than using glob which probably is going ahead and reading the entire directory under the covers, you might try just calling opendir and then check if readdir returns more than 2 items (last'ing out on the 3rd; presuming of course a POSIX-y at least "." and ".." even in an empty dir). Alternately stat the directory itself and see if the nlink field is greater than 2.

And as a somewhat related aside: don't put thousands of files in a single directory if you can avoid it as for most filesystems it'll cause performance problems (as you've found). Break up files into groups (e.g. based on part of the filename, or artificially generate a key (Digest::MD5 is useful) and break that key up (directories 00, 01, ..., fe, ff, files with md5s starting with ab go in the ab directory, yadda yadda yadda)).

[Your Mother]

Path::Class offers a way. The module is portable across different file systems though the example is *nix. The main advantage to using this approach is it grows nicely and lets you pass things around easily.

moo@cow~>mkdir temp
moo@cow~>perl -MPath::Class -le 'print Path::Class::Dir->new("./temp")
+->children == 0 ? "none" : "some"'
none
moo@cow~>touch temp/fish
moo@cow~>perl -MPath::Class -le 'print Path::Class::Dir->new("./temp")
+->children == 0 ? "none" : "some"'
some

[leocharre]

Might be of use.. File::Find::Rule::DirectoryEmpty

What you need to do is use readdir() but make sure you read one at at time, that is, not in array context

[alexm]

Using File::Slurp::read_dir() as in:

use File::Slurp;

if ( read_dir('/path/to/dir') == 0 ) {
   print "is empty!\n";
}

Update: using File::Slurp or Path::Class will have bad performance on a directory with a large number of entries. However, File::Find::Rule::DirectoryEmpty is addressing the problem the right way, as mentioned by Fletch, by looking for any other entry aside from "." and ".." (which means 3 readdir calls at most whatever the number of entries).