02:34 <discocaml> <Kali> there are precious few resources about ocaml online

04:38 <discocaml> <sim642> It's too small to make any difference at this point in one day

07:12 <discocaml> <RegularSpatula> Anyone seen this message when running inline_tests under javascript before? "Warning: overriding the purity of the primitive Base_am_testing: pure -> mutator"

09:54 <discocaml> <kakadu18> Folks, do we have PPX-based preprocessor that allows to define something from command line? ppx_optcomp doesn't...

12:51 <discocaml> <xavierm02> Hi. I'm still on my idea of having two versions of an inductive: the normal one and another with some constructors disabled. My latest attempt defines the inductive in a functor and wraps arguments of inductive type in polymorphic types given as arguments to the functor. The full inductive is recovered by passing a module with all types equal to type 'a t = 'a, and some constructors can be disabled by defining the corresponding polymorphic t

12:51 <discocaml> <xavierm02> The only drawback I can see is that the restricted inductive is not a subtype of the full inductive, but this is only a problem when considering functions between the two types (because everything else can be placed in a functor to define it on both at once), and for my use case where the inductive represents an AST, the only such function simplifies the AST from the "full syntax" to the "core syntax", and would make no use of a subtyping r

12:51 <discocaml> <xavierm02> Did I miss an obvious drawback ?

13:40 <discocaml> <xavierm02> Nevermind, this does not work, not all functions can be written in similarly parameterized functors.

13:41 <discocaml> <octachron> @xavierm02 , you are aware that your GADT version works, if you use correctly GADTs?

13:53 <discocaml> <octachron> For the functor version you also probably want to lift the type definition outside of the functor to keep type equalities: https://gist.github.com/Octachron/dea0ed81b2884546da239aa00bf159cb .

13:55 <discocaml> <xavierm02> @octachron In the last version with GADT ( https://pastebin.com/38ZSnUxy ), I could not get a polymorphic-enough type for replace_A_by_C so I kind of gave up

13:57 <discocaml> <octachron> You are not using GADTs.

13:58 <companion_cube> my_type seems ilike it does?

13:58 <companion_cube> -i

13:58 <discocaml> <octachron> GADTs can only refine abstract types, and you are using unification variables rather than can never be refined (because they are global variables)

13:58 <discocaml> <octachron> GADTs can only refine abstract types, and you are using unification variables that cannot be refined (because they are global variables)

13:59 <companion_cube> but the recursive function most like needs a type annotation like `let rec replace_A_by_C : type a b c f. …`

14:00 <discocaml> <octachron> Indeed, and without those type annotations, GADTs are equivalent to normal variants + type constraints.

14:00 <discocaml> <octachron> See https://v2.ocaml.org/manual/gadts-tutorial.html .

14:05 <discocaml> <xavierm02> It works! replace_A_by_C has type "type a b c f. < a: a; b: b; c: c; f: f > my_type -> < a: no; b: b; c: yes; f: f > my_type". Thanks!

14:07 <discocaml> <xavierm02> (The c:yes would ideally be "a or c", but that's most likely impossible. And for the real use case, I won't record whether the "core" constructors are there or not, so it's not a problem)

14:09 <discocaml> <octachron> If you really *really* want to suffer rather than a "yes" tag, the field could carry a proof term for `a or b`.

14:09 <discocaml> <xavierm02> I had tried "let rec replace_A_by_C (type a b c f) (x : < a: a; b: b; c: c; f: f > my_type) : < a: no; b: b; c: yes; f: f > my_type = match x with", and had gotten an error that "b escapes its scope". I was so close >_<

14:11 <discocaml> <octachron> Ah yes, you often need recursive polymorphism with recursive functions + GADTS. Overall, my recommendation is to always start with `let f : type a b c . .... ` when using GADTs. This tends to avoid a lot of troubles.

18:37 <discocaml> <psychokitty> Is it possible to `./configure` the OCaml compiler in such a way that it uses a different `asmcomp/{arch}` folder than the host? Even if it's to the extent of the `ocamlopt` invocations being to actually assemble and link anything (like, say I don't have a cross compiler and merely want to see `-dsel` etc.).

18:37 <discocaml> <psychokitty> Is it possible to `./configure` the OCaml compiler in such a way that it uses a different `asmcomp/{arch}` folder than the host? Even if it's _not_ to the extent of the `ocamlopt` invocations being to actually assemble and link anything (like, say I don't have a cross compiler and merely want to see `-dsel` etc.).

18:51 <hexology> is there a library to iterate over unicode codepoints and/or unicode extended grapheme clusters in text? assuming it's encoded in either utf-8 or utf-32

18:52 <hexology> it looks like ocaml strings are "strings of bytes". maybe i'm looking for a general "unicode strings" library?

18:53 <companion_cube> uutf can iterate on codepoints

20:24 <octachron> there is `uuseg` for general segmentation

23:17 <discocaml> <Et7f3 (@me on reply)> Rename your default compiler and it should work

23:17 <discocaml> <Et7f3 (@me on reply)> You will not produce a cross platform compiler but at least it should run

23:29 <discocaml> <psychokitty> can you elaborate on what you mean