This is a fine thing. It would be nice in theory if the texinfo stuff could be fixed, but I'll admit that even I use it so rarely that I probably wouldn't notice.

Thanks! Do you have an HTML rendering of the new version?

Thanks! Do you have an HTML rendering of the new version?

Built it, yes. I should mention I noticed a couple of oddities (the hyperlinks for (&&) for example come out a different color than the others, and I have no idea why.) Also, in the table itself, you should say "A few other syntactic constructions" rather than "Few other syntactic constructions". Mostly it seems good. I'm a bit distracted right now so I can't do a tight proofread though.

@gasche , here is the new html rendering .

@pmetzger , typo fixed, thanks! Does all (&&)s render in a different color?

It looks weird to have the list of submodules without transition, as if they were part of the appendix. I suppose it was already weird before the appendix was added. Can it be placed before, with a title/section name? Or should there at least be a section for it after the appendix?

Good point. I moved the appendix to the end of the documentation and added a section for the list of modules of the standard library.

I'm not completely sure that we really need to recall the levels in the text itself (we could just say (see the {!reference} for associativity and precedence))

Except for associativity which concerns the operator itself and not its relationship to others, I would tend to agree with @gasche.

I think precedence relationships among operators is better understood by the table than with the text and the idea of classifying by domain doesn't work that well since for example @@ and |> concern all domains.

Also repeatedly seeing the whole table might help people (at least visual ones) to eventually get it in their brain (personally I'm a lost cause).

I agree that the textual description of precedence is not really useful for the html version, I am less sure for the manpage version where it is not possible to follow the link to the table.

I am less sure for the manpage version where it is not possible to follow the link to the table.

One option is to have a man page specifically for precedence. Some man page readers (including the one I use in emacs) let you click from one man page to another when there are cross references.

Indeed, it seems that lifting the table to its own page might be the more portable choice.

I'm not sure what are valid manpage names, but one that comes to mind is ocaml-operator-grammar.

Most Unix systems have an operator(7) man page with C operators and their precedence. An ocaml-operator page might be fine.

I have split the operator table to its own documentation-only module ocaml_operators.mld inside the stdlib directory and replaced the above ... and below ... indication by a simpler reference

see {!Ocaml_operators} for more information

I don't remember everything; is the PR mergeable in the current state (modulo minor fix discussed above)?

Some check-typo fiddling was missing, but this should be fixed in the last commit.
Otherwise, if the location of the mld file and the reference message are fine, the PR should be mergeable.

Well then I would suggest that you merge when you are satisfied with the results / typo checks / CI results :-) (I already approved.)

I will merge tomorrow to let some time for potential reaction.

I have fixed some missing reference in the pdf manual, and some missing cleanup; I will merge once CIs pass.