From: Run Paint Run Run <runrun@...>
Date: 2009-08-02T14:45:51+09:00
Subject: [ruby-core:24684] Re: [Bug #1852] Enumerable's #hash Raises 	ArgumentError When Recursive Values are Present

> * several real applications are found

This change broke a lot of specs which was how I noticed it. For
example, Array#&, Array#-, Array#|, Array#sort, Array.flatten,
Array#uniq, and Array#uniq! all now raise ArgumentErrors when
previously they didn't. And that's just the Array methods for which
this scenario was being verified. That's quite wide-scale breakage,
and I still don't understand the reason. Why is it beneficial for
these methods to raise exceptions when previously they dealt with
recursive structures?

From what you're saying the plan is to allow recursive values but
leave it to the individual methods to decide whether they want to
behave or not. This will lead to more exception handling code
scattered around because previously compliant functions are now
temperamental. Or something equally ugly like "ary.uniq! unless
ary.include?(ary)".

> * O(n) behavior by constant hash value is not a problem for the applications
> * someone write a patch

This would be a reasonable requirement if it were a feature being
suggested, but that's not the case here. A multitude of methods have
suddenly started raising exceptions because of what appears to be a
new feature, and there has been no explanation why. That confuses me.

-- 
Run Paint Run Run