Even if you can't load the entire log into memory, loading it in chunks should speed things up.

1. You can take the hash with the entries and compact a few thousand of them (or however much memory you want to use). OR you can grab a chunk of the log data from a file with newlines, and read to the next line <IN>.

2. Run search/replace for each regex over the entire buffer, by evaluating the s/foo/bar/g in list context. It can still be found how many replacements were done for each regex.

Code:

Get a chunk of data:

my $bufflen = 4 * 1024; #or w/e

do {
    $result = read ( IN, $buffer, 
            $bufflen-length($buffer), length($buffer) );
} while ( $result && ( length($buffer) < $bufflen ) );
[download]

If the chunk ends in the middle of a line, strip off the remainder and save it for the next chunk: (not needed if each entry is separated beforehand)

my $newline = "\n"; #or some other unique record separator

## if we're not at eof
if ($result > 0) {
    my $last_newline = rindex $buffer, $newline;
    my $remainderlen = length($buffer)-$last_newline-length($newline);
    if ($remainderlen <= 0) {
        $remainder = ''; 
    } else { 
        $remainder = substr($buffer, $last_newline+length($newline), 
            $remainderlen, ''); 
    }
}

## this is important: prefix the remainder before next chunk
$buffer = $remainder;
[download]

__EDIT__: Or instead of the above you could just do readline like BrowserUk did. D'oh!

$buffer .= <IN>;
[download]

Then apply your regexes: (and count how many replacements were done)

foreach my $regex (@conversions) {
    my @results = ( $buffer =~ s/$regex->{from}/$regex->{to}/g );
    my $reps_done = 0;
    grep { $reps_done += $_ } @results;
    $regex->{count} += $reps_done;    
}
## and do whatever with the result
print OUT $buffer;
[download]

The above would be inside a block which loops over each chunk until the end of the log file is reached