Re: RCR allow indexing last n items

On 01/01/2008, Jonas Pfenniger <[e-mail]> wrote:

> And there is yet another case : "abc"[0,0] #=> ""
>
> if "abc"[0,0] would return nil, then you could coerce the range so
> that [-10, -5] would give [0,0], which would return nil.
>

Yes, I do not understand the reasoning for returning "" in some cases
and nil in other cases.

Returning nil when the range is invalid, and false when there is no
intersection (or something like that) would be more consistent.
However, it would make using of the ranges harder.

I guess it is reasonable that you can use the same range to dissect
strings of different length:
"abcde"[2..-3] => "e" #(2..2)
"abc"[2..-3] => "" #(2..0)

Thanks

Michal