I think it makes sense to refine the contract to guarantee a fixnum result for `vector-length`.
This fact is currently documented in `unsafe-vetcor-length`, because that's the layer where it has seemed sensible to talk about fixnums in the past, but that's not where anyone would think to look. Meanwhile, the contract for unsafe operation doesn't specify a nonnegative fixnum as it should. At Wed, 4 Dec 2019 14:07:47 -0500, George Neuner wrote: > > On 12/4/2019 11:03 AM, Dominik Pantůček wrote > > looking at vector-length[1] documentation, it returns > > (exact-nonnegative-integer?). However, as far as I can tell, it returns > > (fixnum?). Also for subsequent contracts in the vector's documentation > > all indices are assumed to be (exact-nonnegative-integer?) but usually > > it is impossible on given platform to use anything larger than (fixnum?) > > in reality > > You are correct that (exact-nonnegative-integer?) doesn't limit values > to fixnum range, but the point of using it is more to exclude negative > values than to allow super large positive ones. > > This question - or something substantially similar - has come up > before. Your question is about contracts - which can be used from plain > Racket - but it is noteworthy that typed/Racket provides types for > positive, negative, nonpositive, and nonnegative fixnums - but there are > no built-in tests for any of these, so a suitable contract or type > declaration has to be synthesized using multiple tests. > > > > For example on 64bit platforms the (fixnum?) could store numbers from (- > > (expt 2 62)) to (sub1 (expt 2 62)). Vector slots cannot be smaller than > > 64bits (machine-dependent pointers) which means 8 bytes = 3 bits of > > address. Given 64bit (VERY theoretical) size of the address space, this > > leaves the possibility of storing a single vector with (expt 2 61) > > elements into the memory. > > > > If I did not overlook anything, the contracts could be safely changed to > > (fixnum?). > > It would be more correct to use (and/c fixnum? (or/c zero? positive?)) > to explicitly limit the value. > > > > And yes, I found this issue while cleaning up my futures-sort[2] package > > discussed a few months ago here. If I assume > > (exact-nonnegative-integer?), I need to manually check for (fixnum?). > > Even though it - given the information above - does not really make much > > sense. > > > > > > Should I open a PR? > > I don't remember how the previous discussions went. I'm not sure > anything ever really was resolved. > > > > Cheers, > > Dominik > > > > > > [1] > > > https://docs.racket-lang.org/reference/vectors.html#(def._((quote._~23~25kernel > )._vector-length)) > > [2] https://docs.racket-lang.org/futures-sort/index.html > > YMMV, > George > > -- > You received this message because you are subscribed to the Google Groups > "Racket Users" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to racket-users+unsubscr...@googlegroups.com. > To view this discussion on the web visit > https://groups.google.com/d/msgid/racket-users/295427b5-9390-43b7-b7d5-cf25cce7 > fd4b%40comcast.net. -- You received this message because you are subscribed to the Google Groups "Racket Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to racket-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/racket-users/20191204192133.5463E65018B%40mail-svr1.cs.utah.edu.
