|Perl: the Markov chain saw|
I disagree slightly with tye that Moose forces [public accessors] on you
I find it entertaining that I must parse that as "chromatic believes that Moose forces public accessors upon you", since I don't believe that Moose does that and chromatic claims to disagree with me on that point. :)
As I previously noted:
Now, I'd be more convinced by the standard "don't blame the tool" arguments if I had argued that one shouldn't use Moose because it fails to prevent you from doing stupid things.
Moose doesn't force you to do stupid things. I haven't even come close enough to such a claim as to have claimed that Moose should be faulted for failing to prevent you from doing stupid things.
It is easy to see that you are quite free to do nothing at all, despite choosing to use Moose:
(Untested, however, as I don't plan on waiting for hour(s) while Moose and all of its dependencies try to install.)
But much more to the point, it is quite possible to fight against Moose and avoid violating encapsulation in your class design by not making your implementation details about what attributes to store in each object manifest in the public interface to your class.
But doing that requires eschewing so much of the functionality of Moose that it starkly raises the question asked in the prior code comment.
stvn himself says that the only benefit of using Moose that way is:
What this buys you over doing it by hand is [...] proper constructor initialization for any subclasses (via BUILD).
But inheritance (in Perl) is just anti-modular so claiming some benefit based on supporting inheritance doesn't actually demonstrate a benefit when trying to do good OO design.
So that leaves us with using Moose to automatically generate "private accessors". Since private accessors have underscores at the front of their names, they don't actually leak implementation details into the public interface. Well, I half buy that argument.
But even generated private accessors encourage poor class design. When you've got a generated accessor, you end up wanting to add a small bit of functionality to one of the accessors. So you end up defining a "data type" for that attribute and/or you declare a 'before' or 'after' wrapper around the method.
And this leads to: "Your logic for a single class will be split apart into tons of tiny pieces where the order and interactions will become almost impossible to see, predict, adjust, and debug. But your code will look pretty. Just don't try to understand it on any sort of deep level (such as when you need to fix something)."
To which stvn noted: "Nothing about Moose forces you into doing something as idiotic as you describe."
So, yes, you don't have to use generated private accessors with Moose since the only way to adjust/extend them is "idiotic" design.
So, the bottom line is that using Moose but not in any of the ways that lead to non-modular or idiotic design also means that Moose offers no benefits. So there is no point using it that way.
And having repeatedly had to deal with the pain of the non-modular design that takes great discipline to avoid slipping into once you establish a practice of using a tool that makes generating accessors easy or encourages the use of inheritance to partially re-use code (despite working with a group of very smart and capable Perl developers), I very much don't want to revisit that pain.
And the pain of fragmented code from "idiotic" design from when you've established a practice of using a tool that encourages generated accessors that you customize with different types of wrappers ("datatype", "before", "after", etc.) was so acute that having experienced when that becomes painful only a few times was still enough to make me quite eager to avoid it.
So, yes, I find using Moose is very likely to lead to painful situations caused by poor class design in the long run. But only if you actually use any of the features that Moose provides. So, indeed, "Nothing about Moose forces you into doing" this. (: