I actually code the other way around. I do for example something like this:

sub foo
  {
  #takes 2 arguments, a scalar(number between x and y) and a hashref

  #make the scalar jump through a hoop

  #make a copy of the hash referenced by the hashref
  
  #set the hashref on fire

  #multiply each value in the newly created hash by the scalar 

  #return a reference to said newly created hash

  }
[download]

And the next step is to actually turn said train of thought into code, which would yield something like this:

sub foo
  {
  #takes 2 arguments, a scalar(number between x and y) and a hashref
  my($scalar, $hashref) = @_;
  unless(($scalar =~ /^\d+$/) && ($scalar >= 7) && ($scalar <= 25) && 
+ref($hashref) eq 'HASH'))
    { print "you're an idiot!\n"; return; }
    #gee, maybe i could be a little more informative here?

  #make the scalar jump through a hoop
  make_jump_through_hoop($scalar);

  #make a copy of the hash referenced by the hashref
  my %newhash = %{$hashref};
  
  #set the hashref on fire
  set_on_fire($hashref);

  #multiply each value in the newly created hash by the 
  foreach my $key (keys(%newhash))
    { $newhash{$key} = $newhash{$key} * $scalar; }

  #return a reference to said newly created hash
  return \%newhash; 
  }
[download]

So in this case(usually when I'm doing something really complicated(for me anyway)) code follows comment, not the other way around.