http://qs1969.pair.com?node_id=418063


in reply to Re: Mutator chaining considered harmful
in thread Mutator chaining considered harmful

If you don't return $self, you're taking the decision for the user. Which isn't very friendly.

That's actually a better argument than it might seem like, but, as I wrote in my root node update, if you don't write separate mutators, but rather a single central property setter method, you completely obviate the need for chaining to begin with:

$window->set( title => 'foo', border => 20 );

And I was almost going to point out that you were picking up the wrong widget in your rewrite of the chaining because the en passant assignment in

my $button = $window->child->border(20);

misled me.

Really, don't mix these things. It seriously makes reading the code an excercise in frustration.

Makeshifts last the longest.

Replies are listed 'Best First'.
Re^3: Mutator chaining NOT considered harmful
by Ovid (Cardinal) on Dec 30, 2004 at 03:36 UTC

    Let's compare:

    $window->title('foo')->border(20); # versus $window->set(title => 'foo', border => 20);

    Clearly, I like method chaining. I use it frequently and my primary argument to people is also "don't use it if you don't want it." However I would never encourage method chaining if the mutator didn't return an object of the same class as the object that received the message (or maybe with an isa relationship and with constructors being a special case). That's just begging Demeter to beat you to death. However, when I look at your code, I see one great big "catch-all" mutator where now I no longer can merely validate the values, but I also have to validate the names of the attributes. It would be easy to type this:

    $window->set(title => 'foo', border => 20);

    Depending on how someone codes that, it could easily silently fail when a direct method call would not.

    I suppose we could argue all day long about which technique is better, but both of our preferred styles have strengths and weaknesses. For now I'll continue to return $self and you can continue to ignore it :)

    Cheers,
    Ovid

    New address of my CGI Course.

      Assuming you aren't writing all of your separate setter/getters manually, which means you're doing some of the typical method generation or AUTOLOAD monkeying, then the separate setter/getters don't buy you a whole lot. I concede that they can make mistakes apparent a little sooner. It's not at all hard to write a unified setter in such a fashion that it blows up just as quickly, though — actually it's trivial enough that I'll bet money on getting it right the first time. Mostly because it's not a point I needed to be made aware of either; I already do that all the time.

      If you want me to ignore your chainable mutators, allow me to have a unified setter and I gleefully shall. :-)

      If you're writing code that uses (rather than provides) mutator chaining though, and I'm going to be maintaining it later, then I shall keep arguing. My experience so far has been frustrating enough, I'd really rather avoid more of that. :-(

      Makeshifts last the longest.