10:38 <d_bot> <Cyclomatic Complexity> Hi

10:39 <d_bot> <Cyclomatic Complexity> Is there a ppx that lets you declare which module declarations are private/public, and generate its interface from it?

10:40 <d_bot> <Cyclomatic Complexity> Let's say that there is no such ppx, and I wanted to make one, what would be the best way to do so? I already made some ppx deriving in the past if that helps.

10:58 <d_bot> <octachron> If you are generating another file, this is not a ppx. It can be a compilerlibs based executable.

10:59 <d_bot> <octachron> And yes that is probably doable, and probably an anti-feature.

12:51 <d_bot> <Cyclomatic Complexity> @octachron Not an other file, just a module signature

12:51 <d_bot> <Cyclomatic Complexity> Wondering why this would be an anti-feature. Worst case, if one is attached to the .mli because of legacy/aesthetics, one could just generate it. Right now, I have to annotate my functions both in the .ml / struct, to get good error messages, _and_ in the signature / interface

12:53 <d_bot> <octachron> Writing a mli is a minor percentage of the work required to write a barebone documentation.

12:53 <d_bot> <octachron> It is trading a minor local annoyance for far less visible interfaces and documentation.

12:54 <d_bot> <octachron> You are writing the interface for the ml file, that requires generating the mli file.

12:58 <d_bot> <roddy> "I have to annotate my functions both in the .ml / struct, to get good error messages" as in, to get good type errors? Or something else? Personally I don't find annotations in structs to be necessary very often

12:59 <d_bot> <Cyclomatic Complexity> A lot of the time, you are writing code that is not meant for other people, and you can use whatever guidelines you want when in teams

13:00 <d_bot> <Cyclomatic Complexity> Yes. If you violate a type annotation in an .mli / module sig, you get an usually much more bloated error message (in the case of merlin, the whole module + sig become red) than if you violate a type annotation on a function

15:23 <User4321> Hi everyone. I have a problem with types. I have a function that returns a pair of elements and I want these elements to be passed as parameters to another function which takes 2 separate arguments...

15:23 <User4321> Could someone have a look at my code? Thanks :)

15:25 <d_bot> <undu> ```ocaml

15:25 <d_bot> <undu> let arg_1, arg_2 = function_that_returns_a_pair () in

15:25 <d_bot> <undu> another_function arg_1 arg_2

15:25 <d_bot> <undu> ```

17:24 <d_bot> <EduardoRFS> it's not being attached to the mli, mli provides docs and also performance improvements, if you don't have an mli at least you need to wait the `.ml` to type to start building another file, in practice you will be waiting for the ml to fully build then start building the next one, having the mli in opaque mode allows everything to build in parallel

17:25 <d_bot> <EduardoRFS> but yeah it's possible to generate this, you will need to run the typechecker so a staged_pps, which is also slower, because it then runs the typer 2 times, but it is feasible

17:51 <flux> so, I haven't done OCaml for a while, but now I have a job for it. how come `opam switch --full` doesn't list 4.12?

17:51 <flux> also my ocamlrun binary is failing due to glibc version.. :)

17:54 <octachron> "opam switch --full" lists the installed switches.

17:56 <flux> ah, it was list-available :)