Well normally for something like this I would arrange to have an array of hash refs:

my @array = (
  {TYPE =>  "hole", SIZE => 126, STEP =>   "pcb", COMP =>  "none"},
  {TYPE =>  "hole", SIZE =>  93, STEP =>   "pcb", COMP =>  "none"},
  {TYPE => "chain", SIZE =>  96, STEP => "array", COMP => "right"},
  {TYPE =>  "slot", SIZE => 130, STEP => "panel", COMP =>  "left"},
);
[download]

and now sorting it is as easy as:

@sorted = sort {$a->{SIZE} <=> $b->{SIZE}} @array;
[download]

This kind of design is better for a number of reasons. But still in your case you have an unfriendly data structure for the operation. Well here is a way to do this:

my @indices = sort {$hash{SIZE}[$a] <=> $hash{SIZE}[$b]} 0..$#{$hash{S
+IZE}};
foreach my $key (keys %hash) {
  @{$hash{$key}} =  @{$hash{$key}}[@indices]; # Slice
}
[download]

But honest, reversing the structure of your data structure now is advised if at all possible.