16:52 <Anarchos> Can someone explain me this error : https://0x0.st/X6ry.txt

16:58 <dh`> PVar is the wrong type

16:59 <discocaml> <deepspacejohn> Reference PVar with the module namespace, F.PVar

16:59 <dh`> well yes, that much is obvious

16:59 <dh`> (what I said)

17:02 <discocaml> <deepspacejohn> because formula is an extensible variant, there could be multiple "PVar" constructors defined for it in different modules. That's why the compiler can't disambiguate it based on the type information like it could with a normal variant.

17:05 <Anarchos> <discocaml> <deepspacejohn> thanks for this clear explanation

17:19 <discocaml> <emiletrotignon> This one has an option, I am sure.

17:21 <discocaml> <emiletrotignon> I do not know if there is an option for that. This should probably be an option, for me parenthesis removal is one of the great features of ocamlformat because it remove the need to know about priority (it even teaches the priority of operations to some degree). Your need also seem legit.

20:14 <Anarchos> is it possible to have a.mly grammar whose token type is an extensible variant defined in a module A , to use the parser inside A ?

20:19 <discocaml> <contificate> You may be interested in the `--external-tokens` flag you can pass to menhir. That roughly lets you use a module that has a token type that you define. How flexible it is, I am unsure.

20:23 <Anarchos> <discocaml> <contificate>it is not menhir but it seems i can't use an extensible variant from within a module : https://0x0.st/X6zD.txt

20:24 <discocaml> <contificate> I don't think you can use polymorphic variants as constructors for menhir tokens anyway. I'm unfamiliar with dypgen but you can basically guarantee it'll be more inflexible than menhir.

20:25 <Anarchos> contificate, yes it seems i can't define the parser inside the module. I wanted it to have signature 'type t = .. ;; val of_string : string -> t val to_string : t-> string'

21:04 <Anarchos> Can A.ml{i} depends on B.mli and B.ml depends on module A ?

21:11 <discocaml> <deepspacejohn> No. You'd have to take the parts they both depend on and move it to a third module.

21:22 <discocaml> <getzapped.> are there any ppx derivers out there for binary serialisation and deserialisation for custom types

21:28 <discocaml> <getzapped.> or is the easiest option just compressed sexp strings

21:33 <discocaml> <yawaramin> there are a few, eg https://ocaml.org/p/ppx_protocol_conv_msgpack/latest

21:51 <Anarchos> <discocaml> <deepspacejohn> i find it weird cause A.ml{i} would only depends on B.cmi and B.ml just constains a renaming of a A.function

21:52 <discocaml> <gooby_clown> ocaml-protoc does binary as well iirc

21:52 <Anarchos> my usual trick is to use to circumvent recursive .ml files is to instantiate the concerned values to " ref None" that i instantiante later in the cycle

21:52 <discocaml> <gooby_clown> It's not a ppx but still

21:53 <companion_cube> Yes, protobuf does binary, thankfully

21:54 <discocaml> <gooby_clown> I think that was the whole point of the protocol

21:54 <discocaml> <gooby_clown> But I might be mistaken

21:57 <companion_cube> Absolutely, it has a json backend but more as a backup really

21:57 <companion_cube> Protobuf is the more efficient format

22:51 <discocaml> <darrenldl> @emiletrotignon not sure i follow why parenthesis removes the need to know about priority

23:37 <discocaml> <darrenldl> *parenthesis removal removes the need ...