From: George Ogata <g_ogata@...>
Date: 2006-02-22T20:53:35+09:00
Subject: Re: dirty ranges

David Vallner <david@vallner.net> writes:

> Dňa Utorok 21 Február 2006 04:38 George Ogata napísal:
>> But I suggest not modifying range endpoints at all; Ranges themselves
>> are immutable, so changing their value indirectly like that is kinda
>> going against the grain.  And of course, if you ever modify your #end
>> while iterating over the range, it's nasal demons.
>>
>
> Pffft. Doesn't even flinch.
>
> ruby <<EOF
> rng1 = ("a".."g")
> rng1.each { |char|
>   if char == "d"
>     rng1.end[0] = "j"
>   end
>   puts char
> }
> rng2 = ("a".."j")
> rng2.each { |char|
>   if char == "g"
>     rng2.end[0] = "d"
>   end
>   puts char
> }
> END
>
> Outputs:
>
> a
> b
> c
> d
> e
> f
> g
> a
> b
> c
> d
> e
> f
> g
> h
> i
> j
>
>
> Of course, I have absolutely NO idea at all why, and don't particularly feel 
> like reading Ruby core source.
>
> David Vallner

You're right; modifying #end is okay.  Not documented, though.  Also
note that modifying #begin can break a loop.  Whether or not this is
counterintuitive depends on the person, I guess.

g@crash:~$ irb
irb(main):001:0> r = 'a'..'z'
=> "a".."z"
irb(main):002:0> r.each{|s| s << '!'; puts s}
a!
=> "a!".."z"

This is another face of Dirk/Tony's concern.  Perhaps the first
element should be cloned for symmetry and safety.  But how: #dup,
#clone, something else...?  You'd also be adding a new requirement
that non-immediate range elements must be copyable.  Hmmm...