Re: Problems with sorting

What you want here is called a topological sort. First create a directed dependency graph of tasks and then apply a topological sort to that graph to produce an ordered list of tasks. Happily, there is a module Sort::Topological that accomplishes this task:

my %children = (
                  'a' => [ 'b', 'c' ],
                  'c' => [ 'x' ],
                  'b' => [ 'x' ],
                  'x' => [ 'y' ],
                  'y' => [ 'z' ],
                  'z' => [ ],
                  );
sub children { @{$children{$_[0]} || []}; } 
my @unsorted = ( 'z', 'a', 'x', 'c', 'b', 'y' );
my @sorted = toposort(\&children, \@unsorted);
[download]

%children gives the dependency graph, and the sub children takes care of the null case.

-Mark

Comment on Re: Problems with sorting Select or Download Code

Replies are listed 'Best First'.

Re^2: Problems with sorting
by Tanktalus (Canon) on Feb 22, 2005 at 00:45 UTC

Thank you - that looks extremely promising. I really, really like that it uses a call-back to get the dependancies rather than asking for a complete list of dependancies up front. Since S::T has to loop through everything anyway, why loop through it in the calling code, too, when a callback will implicitly perform the loop?

It does seem to perform the sort in the reverse order of what I wanted, but thankfully perl makes that pretty darned trivial.

For posterity, my example has changed to:

use Sort::Topological;

use strict;
use Data::Dumper;

my %before =
    (
     a => _s(qw(c)),
     b => _s(qw(d e)),
     c => _s(qw(l)),
     d => _s(qw(e a)),
     e => _s(qw(c)),
     f => _s(qw(g d)),
     g => _s(qw(c)),
     h => _s(qw(g i)),
     i => _s(qw()),
     j => _s(qw(c)),
     k => _s(qw()),
     l => _s(qw()),
     n => _s(qw(c)),
     o => _s(qw()),
    );

# print Dumper(\%before);

my @order = reverse
    Sort::Topological::toposort(
                                sub {
                                    @{$before{$_[0]}};
                                },
                                [ keys %before ],
                               );

print "@order\n";

sub _s
{
    [ @_ ]; #my %h = map { $_ => 1 } @_;
    #\%h
}
[download]

l c a e i g d o f b h k j n
[download]

The only issue is that the module drags in too much - too bad it wasn't separated out into its own distribution. Oh well, can't win 'em all!

[reply]
[d/l]
[select]


Problems? Is your data what you think it is?
	PerlMonks