Just keep track of the last three lines you read, and match against the one before the last you read.

  my $line1 = <>;
  chomp($line1);
  my $line2 = <>;
  chomp($line2);
  while (<>) {
    chomp($_);
    my $line3 = $_;
    if ($line2 =~/(SOMETHING)/) {
      my $something = $1;
      $line1 =~ /(DATE)/;
      my $date = $1;
      $line3 =~ /(SOMETHINGELSE)/;
      my $somethingElse = $1;
      print "$something,$date,$somethingElse\n";
    }
    $line1 = $line2;
    $line2 = $line3;
}
[download]

Note: this is untested code. It will fail on files shorter than three lines. And no, it's not elegant, but it's simplee.

Hope this helps, -gjb-

Update: This is a sliding buffer just as in BrowserUK's approach, but this should work for large files too.