This sounds like a reasonable solution at the beginning, but once you examined what Tie::File does in this case, the solution is quickly invalidated.

Try this piece of testing code:

use Tie::File;
use Data::Dumper;

my $self = tie @array, "Tie::File", "a.pl";
print Dumper($self);
unshift @array, "abcd";
print Dumper($self);
[download]

If you look at the offset array in the dump, you will see that the entire file is loaded to memory (not at a given point, but across the time line)in order to archieve this, which is really a waste, so this task should be rather handed over to OS.

This is the test result against a copy of the demo program:

$VAR1 = bless( {
                 'autochomp' => 1,
                 'mode' => 258,
                 'deferred_s' => 0,
                 'autodefer_threshhold' => 3,
                 'sawlastrec' => undef,
                 'defer' => 0,
                 'dw_size' => 2097152,
                 'offsets' => [
                                0
                              ],
                 'deferred_max' => -1,
                 'autodeferring' => 0,
                 'recsep' => '

',
                 'rdonly' => '',
                 'memory' => 2097152,
                 'filename' => 'a.pl',
                 'fh' => \*Tie::File::FH,
                 'ad_history' => [],
                 'autodefer' => 1,
                 'deferred' => {},
                 'autodefer_filelen_threshhold' => 65536,
                 'recseplen' => 2,
                 'cache' => bless( [
                                     bless( [
                                              [
                                                0,
                                                $VAR1->{'cache'},
                                                0
                                              ]
                                            ], 'Tie::File::Heap' ),
                                     {},
                                     2097152,
                                     0
                                   ], 'Tie::File::Cache' )
               }, 'Tie::File' );
$VAR1 = bless( {
                 'autochomp' => 1,
                 'mode' => 258,
                 'deferred_s' => 0,
                 'autodefer_threshhold' => 3,
                 'sawlastrec' => undef,
                 'defer' => 0,
                 'dw_size' => 2097152,
                 'offsets' => [
                                0,
                                6,
                                22,
                                41,
                                43,
                                88,
                                110,
                                135,
                                157
                              ],
                 'deferred_max' => -1,
                 'autodeferring' => 0,
                 'recsep' => '

',
                 'rdonly' => '',
                 'memory' => 2097152,
                 'filename' => 'a.pl',
                 'eof' => 1,
                 'fh' => \*Tie::File::FH,
                 'ad_history' => [],
                 'autodefer' => 1,
                 'deferred' => {},
                 'autodefer_filelen_threshhold' => 65536,
                 'recseplen' => 2,
                 'cache' => bless( [
                                     bless( [
                                              [
                                                0,
                                                $VAR1->{'cache'},
                                                0
                                              ]
                                            ], 'Tie::File::Heap' ),
                                     {},
                                     2097152,
                                     0
                                   ], 'Tie::File::Cache' )
               }, 'Tie::File' );
[download]

So what?

No, seriously. *Something* has to copy the whole file at some point. I don't know of any mechanism that can avoid that without mucking about with inodes. I personally would only do that as a lark and I wouldn't recommend it as a serious solution to anyone asking this question.

Tie::File is well-maintained, has a nice interface, and is dead-simple to explain. If the original poster starts getting 86400 new messages per day (one hit per second), I'd start worrying about scalability.

I'm okay with "wasting" a few dozen milliseconds and kilobytes here and there if it makes my life easier. If that's unreasonable, so be it.