Re: Recent changes in Range#step behavior

Dave,

Your figures surprised me and they make it sound as if 1.9 is still 
really in flux, so I took a closer look...

Dave Thomas wrote:
> 
> You're right about array: it's only added 5 instance methods since 
> 12/25. 

But these are all just specialized versions of Enumerable methods that 
array already supported. So this is mostly an implementation change and 
not much of a specification change.

The builtins overall have added about 160 methods, though.
> Numeric has added 14, 

These Numeric methods there because Complex and Rational were merged 
into core, so they're not really that new.  Does your 160 figure include 
the 63 methods of Complex and Rational?

String 5, and so on.(*)

These are encoding-related things, and that is one area that was still 
in flux when 1.9.0 was released.

So in reality, we're still
> in the rock-and-roll phase. 

Since you're in the position of writing an API reference, these are 
details that matter to you. But I argue that most of them are 
implementation-level details, not specification-level details, and that 
1.9 is more stable than your figures make it sound.

Seems like a good time to fix stuff, while
> the glue is still workable...

Matz has said, on more than one occasion, (paraphrasing from memory 
here) that the 1.9.x series is intended to be stable, not a series of 
development releases.  He's said that 1.9.0 won't be as stable as he had 
hoped, and that it won't be ready for production use.  But that API 
changes after the 1.9.0 release will happen only if he feels that 
serious mistakes have been made.

Given that we've been living with the current behavior of Range.member? 
for at least 4 years, I just don't think that this change rises to that 
level.  And again, unless you believe that the current behavior is a bug 
introduced in 1.8.x, then you've got to go back and figure out why the 
change was made and explain your argument for changing it back in that 
context.  I still think there are probably very good reasons for making 
numeric ranges uniformly continous.

Furthermore, I think I've started repeating myself on this thread, so 
I'll shut up now. :-)

	David
> 
> Dave
> 
> 
> 
> (*) rough count by grepping for rb_define_method(rb_cXxxx)
> 
>