Request to make Dyn.float more precise

[mbarbin]

Expected Behavior

When serializing floats you have to decide which precision to use. The sexp (sexplib0) and json (yojson) serializers for example will use a varying precision to guarantee that the value could in theory be parsed back to the same value.

Granted Dyn is not meant to be "parsed back", but I've taken a liking to this module as of late and would like to use more generally as a printer to debug values, in expect tests, etc. I'd like to request to import a similar logic from the sexp or yojson libs to dyn and print floats with the precision they require.

Actual Behavior

Currently Dyn.float is printed with a hard coded "12g" (via Stdlib.string_of_float) which loses information. It will print as equal floats that are not, causing some confusion.

Reproduction

This should be visible with a value that doesn't round-trip with 12g.

Specifications

• Version of dune (output of dune --version): latest
• Version of ocaml (output of ocamlc --version): 5.3
• Operating system (distribution and version): ubuntu-something

[mbarbin]

Thanks @Alizter for fixing this ! Another thing I noticed using Dyn in expect tests is that Dyn.char are printed as just the char without the surrounding quotes, thus they're not valid OCaml literals. Making Dyn print literals makes sense to me, but maybe this is not the consensus on what this module is for. I thought I'd ask you the question. Thanks!

[Alizter]

@mbarbin We don't use this constructor internally, at least not for printing. I'm happy to make such a change. Could you create a new issue about this?

[mbarbin]

@Alizter thank you! I created the issue linked above.