Putting aside the mail reading bits, the regex for that would be: /START(.{140)/)/ with modifiers to fit your data format (which could have used a tighter description here). If you have several keywords, you might be better served by Parse::RecDescent.

Update: Having seen your data now, you might be better off splitting on whitespace. You may also be able to exploit the structure of five-character groups. The data you show doesn't seem to fit the description in the root node, am I missing something?

After Compline,
Zaxo

Thanks. I'll have a play with those. The data format is a group of numbers/letters/punctuation. It looks something like this, but may be set out out on any number of lines. The '/' is sometimes a letter in some circumstances.

UFOFH 10504 91001 /0600
6036/ 7049/ 8055/ 9065/ 0068/
1073/ 2075/ 3076/ 4076/ 5073/
6079/ 7073/ 8068/ 9064/ 0060/
1046/ 2042/ 3040/ 0037/ 1035/
2035/ 3035/ 4031/ 5030/
[download]

use strict;

my $keyword = "START";
my $length  = 20;

my $file = <<EOF;
something START text1
 text2 ........................ text 3
text 4 START more text
still more text
EOF

$file =~ s/(?:\s+|\n+)//gc;

my @hits = $file =~ /$keyword(.{$length})/g;

for (@hits) {
  print $_."\n";
}
[download]

Just a little note on efficiency...

use Benchmark;

sub a {
    $_ = "this is a\ntest\n";
    s/(?:\s+|\n+)//gc;
}
sub b {
    $_ = "this is a\ntest\n";
    s/\s+//gc;
}
sub c {
    $_ = "this is a\ntest\n";
    tr/\n\r\t //d;
}

timethese(250000,{ a => \&a, b => \&b, c => \&c });


Benchmark: timing 250000 iterations of a, b, c...
         a:  5 wallclock secs ( 4.18 usr +  0.01 sys =  4.19 CPU) @ 59
+665.87/s (n=250000)
         b:  1 wallclock secs ( 1.61 usr +  0.03 sys =  1.64 CPU) @ 15
+2439.02/s (n=250000)
         c:  1 wallclock secs ( 0.66 usr +  0.02 sys =  0.68 CPU) @ 36
+7647.06/s (n=250000)
[download]

All of these have the same effect on the string...

for one... \s includes \n, so the right part of that doesn't ever actually do anything useful, but I believe it still gets checked each time to make sure it doesn't make a better match... I believe in most cases a character class would be better for that, but that is really irrelevant, since stripping single characters is much faster with tr///d

Just thought I would point it out...

                - Ant
                - Some of my best work - (1 2 3)

I would like to extract, the next 140 characters (excluding whitespace and newlines) after a keyword eg. "START".

my ($foo) = $mailbox =~ /START((?:\s*\S){140})/;
[download]

- Yes, I reinvent wheels.
- Spam: Visit eurotraQ.

Why a regex? For extracting fixed strings, substr() is the handiest way. To find the index from where to extract, we'd use index().

So, combining them gives us:

    my $data   = "... whatever ...";
    my $key    = "START";
    my $length =  140;

    my $text   =  substr $data, index ($data, $key) + length ($key), $
+length;
[download]