From: Sean O'Dell <sean@...>
Date: 2003-11-22T09:25:04+09:00
Subject: Re: "stereotyping" (was: Re: Strong Typing (Re: Managing metadata about attribute types) )

On Friday 21 November 2003 03:53 pm, Yukihiro Matsumoto wrote:
> Hi,
>
> In message "Re: "stereotyping" (was: Re: Strong Typing (Re: Managing
> metadata about attribute types) )"
>
>     on 03/11/22, "Sean O'Dell" <sean@celsoft.com> writes:
> |Some of the things we gain were listed in the proposal.
> |
> |As to benefit: it depends on who you care about.  Programmers, and
> | especially library authors, gain a way to help people learn a little
> | about their code and to catch potential errors in pitfalls.  Project
> | developers gain pretty much the same things: the ability to share code
> | and help their co-workers understand a little more about their code and
> | to help prevent them from making some obvious mistakes.  We also gain
> | some speed in debugging: where interfaces are used by people unfamiliar
> | with the code they are calling into (through ignorance and lack of
> | documentation) it's faster responding to a nice error message about type
> | incompatibilities than it is to wonder about an error regarding method
> | calls.  Project managers are also far more likely to implement Ruby in
> | their projects if they feel it is as safe as what they need, and type
> | checking is one of the safety catches that a lot of people have grown
> | comfortable with.  It's also something Python doesn't have, and will be
> | one of the things you can flat-out say "we have it, they do not."
> |
> |Worth it?  In my opinion, absolutely.
>
> In my opinion, not much.  You've mentioned interface contract would:
>
>   (a) help learning code.
>   (b) gain speed in debugging by more check (and better message).
>   (c) make project manager feel better for more safety.
>   (d) allow us to say "we have something you don't have" to Python
>       people.
>
> Among these, only (b) appeal to me.  (a) can be done better by RDoc
> documents.  For (c) and (d), I don't care about selling Ruby to
> anyone.
>
> I prefer Jim Weirich's idea in [ruby-talk:86007], utilizing method
> combination.
>
> in the library code foo.rb:
>
>   require 'foo_if'
>   # which defines interface
>   # you can comment out from production code
>
>   class Foo
>     def greet(hi)
>       hi.display
>       print "\n"
>     end
>   end
>
> in the foo_if.rb:
>
>   class Foo
>     def greet:pre(hi)
>       assert hi.respond_to? :display
>     end
>   end
>
> with probably interface checking cosmetic.

Well, this is certainly much simpler to implement.

	Sean O'Dell