G'day soblanc,

The general solution for this is to do a primary sort and then, when primary elements are the same, do a secondary sort. The primary and secondary sorts are separated by a || operator. A typical application would be to sort on "lastname" then, for those with the same "lastname", sort on "firstname".

With the data you've presented, the aN values are already sorted within tN values; there's no way to use that to show how this works.

I ran three tests:

• using your OP data
• swapping t2-a1 with t2-a3
• further swapping t1-a1 with t1-a2

I manually changed @matrix for each run. Here's the final code.

#!/usr/bin/env perl

use strict;
use warnings;

my @matrix = (
    [qw{t1 t1 t2 t2 t1 t2}],
    [qw{a2 a1 a3 a2 a3 a1}],
    [qw{mis mis mis mis del del}],
);

print "Original\n";
print_matrix(\@matrix);

my @sorted_indices = sort {
    $matrix[0][$a] cmp $matrix[0][$b]
    ||
    $matrix[1][$a] cmp $matrix[1][$b]
} 0 .. $#{$matrix[0]};

print "Sorted indices\n";
print "@sorted_indices\n";

sub print_matrix {
    my ($matrix) = @_;

    for my $row (0 .. $#$matrix) {
        print join(' ', @{$matrix[$row]}), "\n";
    }
}
[download]

Here's the output for the three runs:

ken@titan ~/tmp
$ ./pm_11146491_sort_matrix.pl
Original
t1 t1 t2 t2 t1 t2
a1 a2 a1 a2 a3 a3
mis mis mis mis del del
Sorted indices
0 1 4 2 3 5

ken@titan ~/tmp
$ ./pm_11146491_sort_matrix.pl
Original
t1 t1 t2 t2 t1 t2
a1 a2 a3 a2 a3 a1
mis mis mis mis del del
Sorted indices
0 1 4 5 3 2

ken@titan ~/tmp
$ ./pm_11146491_sort_matrix.pl
Original
t1 t1 t2 t2 t1 t2
a2 a1 a3 a2 a3 a1
mis mis mis mis del del
Sorted indices
1 0 4 5 3 2
[download]

Armed with the sorted indices, I'll assume you can create sorted matrices. If you encounter problems with this, show us what you tried and where you encountered difficulties — we can provide further help when we know what problem you're having.