fix doc comment interaction with ;;#2691
Conversation
|
There is a larger large diff in test branch, but thats because it used to crash and now it doesnt. |
* main: fix doc comment interaction with ;; (ocaml-ppx#2691) use ubuntu latest for ci (ocaml-ppx#2692) Simplify fmt_function (ocaml-ppx#2690)
| | [] -> [] | ||
| | item :: itms -> | ||
| let next_no_doc = | ||
| List.find itms ~f:(function |
There was a problem hiding this comment.
This doesn't seem very nice complexity wise. Can you explain what this code does in a comment ? I think it would be nice to have that as it makes this part of the code much more complicated.
There was a problem hiding this comment.
Yes its quadratic, but I think its find. There probably a way to make the algorithmic complexity better, but that would also be less readable.
A comment is of course a good idea
There was a problem hiding this comment.
To be clear the complexity is only quadratic when you have loads of floating doc comments which is pretty rare.
| | _ -> c | ||
| in | ||
| let fmt_item c ctx ~prev:_ ~next itm = | ||
| let rec get_semisemi itms = |
There was a problem hiding this comment.
There's a similar piece of code in fmt_structure (that take less cases into account), why are they different ?
There was a problem hiding this comment.
They are different because toplevel can have #require directive, so the types are different. Otherwise its the same logic. There might be a way to factor this.
There was a problem hiding this comment.
Factoring just a little part can be enough to make the code clearer and have a logical place to put an explanation comment. If the complexity needs to be improved later, it's also nice if the bad algorithm is at one place only.
* fix doc comment interaction with ;; * changelog * clearer changelog * fmt
* fix doc comment interaction with ;; * changelog * clearer changelog * fmt
….28.1) CHANGES: ### Highlight - \* Support for OCaml 5.4 (ocaml-ppx/ocamlformat#2717, ocaml-ppx/ocamlformat#2720, ocaml-ppx/ocamlformat#2732, ocaml-ppx/ocamlformat#2733, ocaml-ppx/ocamlformat#2735, @Julow, @Octachron, @cod1r, @EmileTrotignon) OCamlformat now supports OCaml 5.4 syntax. Module packing of the form `((module M) : (module S))` are no longer rewritten to `(module M : S)` because these are now two different syntaxes. - \* Reduce indentation after `|> map (fun` (ocaml-ppx/ocamlformat#2694, @EmileTrotignon) Notably, the indentation no longer depends on the length of the infix operator, for example: ```ocaml (* before *) v |>>>>>> map (fun x -> x ) (* after *) v |>>>>>> map (fun x -> x ) ``` `@@ match` can now also be on one line. ### Added - Added option `module-indent` option (ocaml-ppx/ocamlformat#2711, @HPRIOR) to control the indentation of items within modules. This affects modules and signatures. For example, module-indent=4: ```ocaml module type M = sig type t val f : (string * int) list -> int end ``` - `exp-grouping=preserve` is now the default in `default` and `ocamlformat` profiles. This means that its now possible to use `begin ... end` without tweaking ocamlformat. (ocaml-ppx/ocamlformat#2716, @EmileTrotignon) ### Deprecated - Starting in this release, ocamlformat can use cmdliner >= 2.0.0. When that is the case, the tool no longer accepts unambiguous option names prefixes. For example, `--max-iter` is not accepted anymore, you have to pass the full option `--max-iters`. This does not apply to the keys in the `.ocamlformat` configuration files, which have always required the full name. See dbuenzli/cmdliner#200. (ocaml-ppx/ocamlformat#2680, @emillon) ### Changed - \* The formatting of infix extensions is now consistent with regular formatting by construction. This reduces indentation in `f @@ match%e` expressions to the level of indentation in `f @@ match`. Other unknown inconsistencies might also be fixed. (ocaml-ppx/ocamlformat#2676, @EmileTrotignon) - \* The spacing of infix attributes is now consistent across keywords. Every keyword but `begin` `function`, and `fun` had attributes stuck to the keyword: `match[@A]`, but `fun [@A]`. Now its also `fun[@A]`. (ocaml-ppx/ocamlformat#2676, @EmileTrotignon) - \* The formatting of`let a = b in fun ...` is now consistent with other contexts like `a ; fun ...`. A check for the syntax `let a = fun ... in ...` was made more precise. (ocaml-ppx/ocamlformat#2705, @EmileTrotignon) - \* `|> begin`, `~arg:begin`, `begin if`, `lazy begin`, `begin match`, `begin fun` and `map li begin fun` can now be printed on the same line, with one less indentation level for the body of the inner expression. (ocaml-ppx/ocamlformat#2664, ocaml-ppx/ocamlformat#2666, ocaml-ppx/ocamlformat#2671, ocaml-ppx/ocamlformat#2672, ocaml-ppx/ocamlformat#2681, ocaml-ppx/ocamlformat#2685, ocaml-ppx/ocamlformat#2693, @EmileTrotignon) For example : ```ocaml (* before *) begin fun x -> some code end (* after *) begin fun x -> some code end ``` - \* `break-struct=natural` now also applies to `sig ... end`. (ocaml-ppx/ocamlformat#2682, @EmileTrotignon) ### Fixed - Fixed `wrap-comments=true` not working with the janestreet profile (ocaml-ppx/ocamlformat#2645, @Julow) Asterisk-prefixed comments are also now formatted the same way as with the default profile. - Fixed `nested-match=align` not working with `match%ext` (ocaml-ppx/ocamlformat#2648, @EmileTrotignon) - Fixed the AST generated for bindings of the form `let pattern : type = function ...` (ocaml-ppx/ocamlformat#2651, @v-gb) - Print valid syntax for the corner case (1).a (ocaml-ppx/ocamlformat#2653, @v-gb) - `Ast_mapper.default_mapper` now iterates on the location of `in` in `let+ .. in ..` (ocaml-ppx/ocamlformat#2658, @v-gb) - Fix missing parentheses in `let+ (Cstr _) : _ = _` (ocaml-ppx/ocamlformat#2661, @Julow) This caused a crash as the generated code wasn't valid syntax. - Fix bad indentation of `let%ext { ...` (ocaml-ppx/ocamlformat#2663, @EmileTrotignon) with `dock-collection-brackets` enabled. - ocamlformat is now more robust when used as a library to print modified ASTs (ocaml-ppx/ocamlformat#2659, @v-gb) - Fix crash due to edge case with asterisk-prefixed comments (ocaml-ppx/ocamlformat#2674, @Julow) - Fix crash when formatting `mld` files that cannot be lexed as ocaml (e.g. containing LaTeX or C code) (ocaml-ppx/ocamlformat#2684, @emillon) - \* Fix double parens around module constraint in functor application : `module M = F ((A : T))` becomes `module M = F (A : T)`. (ocaml-ppx/ocamlformat#2678, @EmileTrotignon) - Fix misplaced `;;` due to interaction with floating doc comments. (ocaml-ppx/ocamlformat#2691, @EmileTrotignon) - The formatting of attributes of expression is now aware of the attributes infix or postix positions: `((fun [@A] x -> y) [@b])` is formatted without moving attributes. (ocaml-ppx/ocamlformat#2676, @EmileTrotignon) - `begin%e ... end` and `begin [@A] ... end` nodes are always preserved. (ocaml-ppx/ocamlformat#2676, @EmileTrotignon) - `begin end` syntax for `()` is now preserved. (ocaml-ppx/ocamlformat#2676, @EmileTrotignon) - Fix a crash on `type 'a t = A : 'a. {a: 'a} -> 'a t`. (ocaml-ppx/ocamlformat#2710, @EmileTrotignon) - Fix a crash where `type%e nonrec t = t` was formatted as `type nonrec%e t = t`, which is invalid syntax. (ocaml-ppx/ocamlformat#2712, @EmileTrotignon) - Fix commandline parsing being quadratic in the number of arguments (ocaml-ppx/ocamlformat#2724, @let-def) - \* Fix `;;` being added after a documentation comment (ocaml-ppx/ocamlformat#2683, @EmileTrotignon) This results in more `;;` being inserted, for example: ```ocaml (* before *) print_endline "foo" let a = 3 (* after *) print_endline "foo" ;; let a = 3 ``` - Fix dropped comment in `if then (* comment *) begin .. end` (ocaml-ppx/ocamlformat#2734, @Julow)
the issue is that floating doc comments are a structure item in the ast, but not in the parser
Therefore the printer sometimes printed
;;after floating doc comment, but did not do so after the item that was before the doc comment.This solves this with a explicit loop over the structure item that is allowed to look at more than one item.
I tried to cst-ize
;;, but thats very annoying. It can probably be done, but is of very limited utility in my opinion.The added test used to cause a crash.