How about using negative lookahead? This way, you check to see whether what follows 'request(' actually contains the BADSTRING, and fails if that is the case.

use strict;

$/ = "";           # Set up for slurp mode
my $data = <DATA>; # Grab the entire file to a string.
my %matches = $data =~
         /^request\(      # start of record
            (?!BADSTRING) # Fails if BADSTRING is present
            ([^\n]+)      # Captures $blah1
            \n            # Match the newline
            [^,]+,        # Match anything up to the comma
            ([^\n]+)      # Capture $blah2. Everything from
         /mgx;            # the comma to the next \n

foreach ( keys %matches ) {
    print "$_ : $matches{$_}\n";
}

__DATA__
request(goodstring
START_TIME,some goodness
request(BADSTRING
START_TIME,some bad stuff
request(randomstring
START_TIME,more goodness
request(whatever you are
[download]

This gives:

goodstring : some goodness
randomstring : more goodness
[download]

Thanks to pzbagel for his excellent sample data :)

This uses several neat tricks. /g repeats the match until the string is empty. /m allows embedded newlines in the string to match. Calling the regex in list context and assigning it to a hash gives us a nice little summary to dump afterwards.

Disadvantages to this approach is of course the amount of memory used by $data and %matches. Malformed data may break /g, and of course the relative illegibility of the regex may be a problem. If your datasamples aren't too big, it might work, though.

Hope this helps, and I welcome (and appreciate) any criticism on my regex-programming style.

pernod
--
Mischief. Mayhem. Soap.