open(FILE, "< $filename"); # use < for taint reasons
$count += tr/\n/\n/ while sysread(FILE, $_, 2 ** 16);
[download]

I'm not sure what your performance requirements are, but the following will scan a single directory of 100+ .pl files (for example) on my system, count & print the number of lines in each in the format file:nn in less 1 second. including printing. With the output re-directed to the null device it does several thousand files in less than 5 seconds.

Update:Scrunched to prevent wrapping

perl -e"@_=map{glob}@ARGV;print$_,':'and$_=do{local(@ARGV,$/)=$_;<>},p
+rint$£=s!$/!!g,$/for@_" c:\*\*.pl
[download]

It is set up for use on a Win32 system so has extra stuff for getting around CMD's lack of globbing.

I did some very unscientific tests on zip files (using WINZIP 7.0) and discovered that for simple text files there appears to be a relationship between the number of NL sequences in the zip and the number of lines in the file ,

ie. text-file-lines +3 = zip-file-NL's.

Maybe you could find a similar relationship using your favorite zip utility on your system.

Update: It seems that at least one person felt that the above was too obfuscated, so here's an expanded version as a normal program with a few comments.

#! perl -sw
use strict;

die "Usage: $0 filespec|glob [[filespec|glob] ... ]\n" unless @ARGV;

# On Win32 CMD doesn't do globing so we need to expand @ARGV ourselves
+.
@_ = map{glob}@ARGV;

# print the name of the file and a colon

    print $_, ':'

# Slurp the contents of the file into $_

    and $_ = do{ local( @ARGV, $/ ) = $_; <>; },

# Use s///g to look for $/ (Newline for your system) in the scalar rea
+d above
# by assigning s///g to a scalar it returns the number of matches
# So we print that and a newline

    print $£ = s!$/!!g, $/

# For all the files we found by globing the command line arguments.

    for @_;
__END__
[download]

thor