You are using a Schwartzian Transform which is good. In this example my expensive function splits the name into three parts (word chars, digits or ., and then everything else). split is a good way to do it as 1) if it fails it will return the whole string (so we can sort on that). If it succeeds then we need to remember that it will split at all the number . sequences so we put the string back together after the first split. If we were to use a straight regex here and it failed to match we would need to have all sorts of fallback cases split() handles it neatly.

The sort then proceeds on first bit, numbers, last bit. If you just wanted to sort foo-bar-0.01.tar.gz then you only need to return @bits[0,1] from func() for it to work correctly.

sub sort_n {
  return
  map { $_->[0] } 
  sort { $a->[1] cmp $b->[1] || $a->[2] <=> $b->[2] || $a->[3] cmp $b-
+>[3] } 
  map { [$_, func($_)] } @_;
  sub func { my @bits = split /([\d\.]+)/, $_[0]; return @bits[0,1], j
+oin '', @bits[2..$#bits] }
}

my @ary = qw( z1z z2a z2z a1a a2b a1.1a a2.1a a0.1a a0.01a b bbb b2b2b
+2b 2b2c a2bc2 );

print "This is what func() does for us\n";
$" = ', ';
for (@ary) { my @a = func($_); print "@a\n" }

print "\nHere is the sorted list\n";
print "$_\n" for sort_n(@ary);

__DATA__
This is what func() does for
z, 1, z
z, 2, a
z, 2, z
a, 1, a
a, 2, b
a, 1.1, a
a, 2.1, a
a, 0.1, a
a, 0.01, a
b, ,
bbb, ,
b, 2, b2b2b
, 2, b2c
a, 2, bc2

Here is the sorted list
2b2c
a0.01a
a0.1a
a1a
a1.1a
a2b
a2bc2
a2.1a
b
b2b2b2b
bbb
z1z
z2a
z2z
[download]

cheers

tachyon

s&&rsenoyhcatreve&&&s&n.+t&"$'$`$\"$\&"&ee&&y&srve&&d&&print

jdporter
The 6th Rule of Perl Club is -- There is no Rule #6.

Apparently it just wouldn't be a day on PerlMonks if no one asked this question. (*sigh*)

I know, I know... But I really did have to ask, and now I'm smarter for it :-)

—John

sub sort_n {
   return
   map  {
           $_->[0]
        }
   sort {
           my $l = @{$a->[1]} < @{$b->[1]} ? @{$a->[1]} : @{$b->[1]};
           for (my $n=0; $n<$l; $n++) {
              my $A = $a->[1][$n];
              my $B = $b->[1][$n];
              my $tmp;
              if ($A =~ /\d/) {
                 if ($B =~ /\d/) { $tmp = $A <=> $B; }
                            else { $tmp = -1;        }
              } else {
                 if ($B =~ /\d/) { $tmp = 1;         }
                            else { $tmp = $A cmp $B; }
              }
              return $tmp if $tmp;
           }
           return @{$a->[1]} <=> @{$b->[1]};
        }
   map  {
           [ $_ , [ split( /(\d+)/, $_ ) ] ]
        }
   @_;
}
[download]