From: David Flanagan <david@...>
Date: 2008-03-18T13:14:02+09:00
Subject: Re: Range#member? semantics seem wrong

Dave Thomas wrote:
> Range#member? has been changed so that it the start and end of the range 
> are scalars, then
> 
>    member?(val)  is the same as min <= val <= max
> 
> This means that (1..5).member?(3.5) returns true. In 1.8, member meant 
> that the value was one of the possible elements in the range (so was 
> effectively the same as range.to.include?(val).
> 
> It seem to me that the old way was more intuitive. We now have three 
> methods on range (===, include?, and member?) that all do the same 
> thing, and none allow you to test for membership.
> 
> 
> 
> Dave
> 

Actually your sample code also returns true in both 1.8 (at least 1.8.6) 
and 1.9.0.

And 1.9.0 has also introduced a new method named cover?.

My understanding (and it takes me all of page 70 to explain this in my 
book) is that in 1.9 include? and member? test what I call "discrete 
membership" (and what you call membership) using succ when the endpoints 
of the range are not numbers.  For numbers they test continuous 
membership using <=>.  And the new method cover? always uses the 
continuous membership test using <=>.

	David