From: Yukihiro Matsumoto <matz@...>
Date: 2009-08-17T07:20:08+09:00
Subject: Re: Bizarre Range behavior

Hi,

In message "Re: Bizarre Range behavior"
    on Thu, 6 Aug 2009 01:44:56 +0900, James Coglan <jcoglan@googlemail.com> writes:

|Quite. I think consistency is important, and this is currently broken in
|1.9:
|
|$ irb1.8
|>> ('!'..']').to_a
|=> ["!", "\"", "#", "$", "%", "&", "'", "(", ")", "*", "+", ",", "-", ".",
|"/", "0", "1", "2", "3", "4", "5", "6", "7", "8", "9"]
|>> '9'.succ
|=> "10"
|
|$ irb1.9
|>> ('!'..']').to_a
|=> ["!", "\"", "#", "$", "%", "&", "'", "(", ")", "*", "+", ",", "-", ".",
|"/", "0", "1", "2", "3", "4", "5", "6", "7", "8", "9", ":", ";", "<", "=",
|">", "?", "@", "A", "B", "C", "D", "E", "F", "G", "H", "I", "J", "K", "L",
|"M", "N", "O", "P", "Q", "R", "S", "T", "U", "V", "W", "X", "Y", "Z", "[",
|"\\", "]"]
|>> '9'.succ
|=> "10"

|There is obviously room for debate over the above examples: the 1.9 one
|could be viewed as more complete even though it doesn't apply #succ
|consistently. A lot of these odd examples are edge cases that are unlikely
|to come up in production code, but personally I find the rules used by 1.8
|easier to understand. If I define my own class, Range will use its #succ and
|#<=> methods to carry out its logic and I expect the same for built-in
|classes.

I admit 1.8 behavior in above example is simpler to implement, but 1.9
behavior follows our (or more precisely "my") unwritten expectation of
character iteration.  Simple #upto (a la 1.8) can act as weirdly,
since #succ don't know about the edge values.  I want to make #upto
work smart.  I know it's magical, but #upto (and Range#each) already
contains magic.  So I consider the weird (even unexpected for OP)
behavior is a result from less magic.  I am not sure the new behavior
should be back-ported to 1.8 or not.

							matz.