my %skip = (
   'gif'  => 1,
   'jpg'  => 1,
   'jpeg' => 1,
   'png'  => 1
);
[download]

my %skip_for;
@skip_for{qw( gif jpg jpeg png )} = ();
[download]

and later test using exists $skip_for{$ext}

Next note: you can just use $File::Find::name rather than "$File::Find::dir/$_"

Then we have a case of redundant syntax: in \&{ sub { ... } } the sub{ ... } already gives you a reference. Then your &{} goes and dereferences it, only to feed it back to the \ which makes a reference from the result again. You can drop the surrounding \&{} and simply write sub { ... } here.

I am a bit puzzled by this:

If you throw away the first capture, why capture at all?

which is better written as

(In words: I want as many non-dot characters as there are in front of the end of the string, update: but only if there's a dot in the filename.)

The $ext = '' unless defined $ext; can be avoided if you put the $skip{$ext} inside an if(/match here/)

Lastly, since you're not interested in the individual lines of your input, but separating the input costs effort, it would be better to unconditionally slurp large chunks of X bytes instead.

The next point is a maneuvre critique. Why would one first fetch a list of directories and then go and read each directory manually, when the same first search already gives you all the file names on a silver plate? (And why are counting something, when you never use that count? :-))

And lastly, rather than hardcode the directory in the script, it's preferrable to take them as parameters from the commandline.

So here's an updated version:

#!/usr/bin/perl -w
use strict;
use Fcntl;
use File::Find;

my %skip_for;
@skip_for{qw( gif jpg jpeg png )} = ();

find(
    sub {
        next if -d or /^[.]/;
        next if /[.]([^.]+)$/ and exists $skip_for{$1};

        my $content = "";

        # gobble and mangle 64k chunks at a time
        sysopen FH, $_, O_RDWR;
        s/\r//g, $content .= $_ while sysread FH, $_, 65536;

        # go back to top of file
        sysseek FH, 0, 0;
        syswrite FH, $content, length $content;

        # the file still has its original length,
        # because we didn't clobber it with an open FH, ">file"
        # so we need to fix that
        truncate FH, tell FH;

        close FH;
    },
    (@ARGV) || "." # NB: parens required
);
[download]

Further improvement might be to use some Getopt:: module to allow the user to change the $skip_for rules.

Update: I must have been asleep as well. Kudos to Zaxo for pointing out my regex would return the whole filename for extensionless files. Also, I need to go flaggelate myself for a while:

sysopen FH, $_, O_RDWR
    or (warn "Couldn't open $File::Find::name: $!\n", return);
s/\r//g, $content .= $_ while (
    defined (sysread FH, $_, 65536)
    or (warn "Couldn't open $File::Find::name: $!\n", return)
);
[download]

return if -d or /^[.]/;
return if /[.]([^.]+)$/ and exists $skip_for{$1};
[download]

Makeshifts last the longest.

Heheh, apparently I was either sleeping, on some kind of drugs or in just some kind of hurry, throwing that script together with anything that did the job :)

Thanks for the improvement (okay fine, improvementS) that you made to it. Must look at perldoc -f sysopen and related docs. Once again, thanks! :)

*goes off, pretending that his code is perfect and cannot be improved upon*

$ perl -MFile::Find -e'find(sub{next if -d or /^\.\.?/ or /\.(jpe?g|gi
+f|png)$/;print "$File::Find::name "},".")' | xargs perl -pi.bak -e 's
+/\r//g'
[download]

find . -type f -print | xargs perl -pi -e 's/\r//g'
[download]

..slightly different:

perl -pi.orig -e 's/\cM//g' `find . -type f`
[download]