open(my $INFILE, ...) or die "unable to open file: $!\n";
chomp(my @lines = <$INFILE>);
close($INFILE);
[download]

print "Number of lines read: ", 1+$#lines, "\n";
for my $i (0..$#lines) {
  print "line $i is: $lines[$i]\n";
}
[download]

Update: Fixed bad chomp usage.

• Work in minutes, not HH:MM
• read arrival times into an array

which leads to something like...

use constant START_MINS => 180;
use constant PERIOD     => 20;

my @arrivals = qw/03:05 03:17 03:23/;

@arrivals = map { as_minutes( $_ ) } @arrivals;

for my $time ( START_MINS+1 .. START_MINS+8 ) {
    my @arrivals_this_period = grep
      { $_ > $time && $_ <= $time + PERIOD } @arrivals;
    print as_hhmm( $time ) , ": ", scalar @arrivals_this_period, "\n";
}

sub as_minutes {
    my ( $hh, $mm ) = split /:/, shift;
    return $hh*60 + $mm;
}

sub as_hhmm {
    my $minutes = shift;
    my $hh = int $minutes / 60;
    my $mm = $minutes - $hh * 60;
    return sprintf "%02d:%02d", $hh, $mm;
}
[download]

Output:

03:01: 2
03:02: 2
03:03: 3
03:04: 3
03:05: 2
03:06: 2
03:07: 2
03:08: 2
[download]

There's still the problem of your 20 minute period crossing midnight though

update: Fixed the for my $time ... line to match the output I gave

For your solution it would probably be better to work in 'epoch minutes' (i.e., minutes some some epoch...i.e., reference...date) rather than 'daily minutes' (i.e., minutes since midnight).

There are time-handling modules in CPAN that can handle it and there are plenty of formulas out on the internet that are pretty simple to compute Julian Dates (i.e., time since some 'standard' epoch, where Julian Date is usually expressed in elapsed decimal days since a 'standard' date...e.g., days since noon on 1 Jan 2000). This avoids the ...crossing of midnight... issue that you noted can be tricky.

ack Albuquerque, NM

Another possibility would be to maintain a hash with time as the key and number of arrivals in that minute as the value. Grabbing any arbitrary set of minutes and adding up the total number of arrivals for those times would then be trivial. It would also provide much easier ways to deal with rescheduled arrivals and out-of-order input data.

How much data are you likely to deal with? Is this a static file or is it updated in real-time?

If it's a static file, and not too large, just read the entire file into memory. If it's being updated dynamically (say it's a pipe), use a data structure (a hash perhaps?) and keep as much data as required in memory. Depending on the number of flights you've got to process, you could probably keep the past twenty-four hours of data (iirc, the busiest airport in the US handles no more than a couple of thousand flights per day).

• File is static and too large to fit in memory
• Records are constant width (as example given in OP implies)

Then...

File can be read in block mode (assign  $/ a reference to block size; see perlvar) and "going back" is just a matter of doing a  seek() by the appropriate negative offset (number of blocks * block size).