Larry Wall wrote:
> David Green wrote:
> : I actually found things I liked in pretty much all the suggested
> : alternatives, but none of them reached out and grabbed me by the
> : throat the way "nth" did. It just seems more Perlish.
>
> Yow. Presumably "nth" without an argument would mean the last.
If it means the last, why not just use C<last>? It's only one character
longer, and its meaning can't be mistaken or confused. So
@ints[1st..last]
means
@ints[*]
There is the question about what C<last+1> would mean; intuitively, moving
forward from the end takes you into limbo; but I can see an advantage in
wrapping things around such that last+1 == 1st, and 1st-1 == last. You
could even go so far as to allow something like postfix:'th for whole
number scalars, so that $n'th == last + $n. This would imply that 0th ==
last, that -1st == last - 1 (or "next to last"), and that -3rd == last - 3
(or "third from last"). While the idea that 0th == last is
counterintuitive to me, the idea that negative ordinals correspond to
offsets from the last element seems rather nice. All in all, this
salvages the perl5 $ints[-1] notion, albeit requiring a +1 adjustment - or
this could be translated as @ints[1st-1] (a bit bulkier, but more
legible).
Would we be able to extend this notion? That is, would the following be
viable:
@ints.shape[0].1st
# returns the index that corresponds to 1st for dimension zero
Regardless, the following _would_ be viable:
@ints.shape[1st]
# returns information that corresponds to the first dimension
And that's worth a lot.
--
Where else could ordinals come in handy? They're already being used in
REs, but I'd recommend expanding the usage there to include C<last> and
possibly C<$n'th> as well as C<1st>, C<2nd>, C<3rd>, etc. Where else?
Would there be any use for C[<STDIN>.3rd], frex?
=====
Jonathan "Dataweaver" Lang
_______________________________
Do you Yahoo!?
Win 1 of 4,000 free domain names from Yahoo! Enter now.
http://promotions.yahoo.com/goldrush
