Re: vector-last / vector-ref with negative indices?

[lloda <lloda@sarc.name>]

Besides the standard which applies to vector- operations, negative indices wouldn't work for Guile arrays to count from the end since negative indices, like in say Fortran, can be valid [e.g. (make-array 0 '(-1 4))]. In older versions of Guile you could use vector-ref on such arrays.

I don't think allowing negative indices is worthwhile for all the complications it introduces. But I don't like negative indices to count from the end either. Most of the time vectors aren't meant to have that wraparound property so it can turn a obvious error into an obscure one.

The solution I like best is to have a keyword that means 'the end', like in Octave, that can be used generically. Barring that you can easily define vector-last or even something like vector-ref-from-end I guess. I think it's weird that srfi-43 doesn't have vector-last, when srfi-1 has last. Maybe Guile could add that as an extension, although I don't like srfi-43 in general (I think the map functions are wrongly designed).

regards

	Daniel


> On 20 Dec 2020, at 11:05, Christopher Lam <christopher.lck@gmail.com> wrote:
> 
> Easy - vector-ref with a negative index is not defined in r[5|7]rs.
> 
> On Sun, 20 Dec 2020 at 06:59, Aleix Conchillo Flaqué <aconchillo@gmail.com>
> wrote:
> 
>> Hi,
>> 
>> This month I'm trying to go through Advent Of Code one more year
>> (previous years I didn't get too far) and I've been finding myself
>> writing the same patterns multiple times that could be avoided by just
>> having a helper function.
>> 
>> One of them is getting the last element of a vector. It is a quite
>> common operation when solving these types of problems. For example
>> Python as you might know uses negative indices.
>> 
>> I have looked around and haven't seen it (unless I completely missed
>> it which would be shameful) but wouldn't it be good to have this
>> built-in by default?
>> 
>> Instead of having to write:
>> 
>> (vector-ref v (- (vector-length v) 1))
>> 
>> you would write:
>> 
>> (vector-last v) or even better (vector-ref v -1).
>> 
>> Interestingly Racket doesn't offer those functions either as far as I can
>> tell.
>> 
>> Basic use case? Get the maximum element of a sorted vector which would
>> be done in constant time.
>> 
>> I'm trying to solve the problems idiomatically, to the best of my
>> limited knowledge, with Scheme. It is possible that the times I'm
>> using vectors I should think about it differently and not use them,
>> but I'm not 100% sure.
>> 
>> What am I missing?
>> 
>> Thank you in advance,
>> 
>> Aleix
>> 
>>