Loading Combobulate into non-tree-sitter-enabled major modes

[tmcgilchrist]

The Readme.md mentions:

Combobulate works in non-tree-sitter-enabled major modes as well as the new Emacs 29 tree-sitter modes.

I would like to understand how I can make this work and what is expected to work. I'm working on Combobulate support for a few of my favourite languages (OCaml and Haskell) and neither of their main modes support tree-sitter just yet.

I have a basic skeleton like so, when I activate Combobulate and try using it I get an error saying "No parsers available"

(eval-and-compile
  (defconst combobulate-ocaml-definitions
    ;; ... DEFINITIONS ...

    )))

 (define-combobulate-language
  :name ocaml
  :language ocaml
  :major-modes (tuareg-mode)
  :custom combobulate-ocaml-definitions
  :extra-defcustoms combobulate-ocaml-definitions
  :setup-fn combobulate-ocaml-setup)

 (defun combobulate-ocaml-setup (_))

(provide 'combobulate-ocaml)

That makes sense, I probably need to load the tree-sitter grammar for OCaml. The main example I could find was rust-mode which provides a defcustom for rust-mode-treesitter-derive which loads a full rust-ts-mode. Is there an option to just load the parser using something like treesit-parser-create while leaving the non-tree-sitter-enabled major mode in charge?

WIP code tmcgilchrist#1

[mickeynp]

You need a valid combobulate configuration for it to work in the non-ts modes.

[tmcgilchrist]

Apologies for the noise I was overthinking the solution. Adding the right mode into :major-modes (tuareg-mode) works if there is one grammar to load. I have a partially working implementation in that linked code.

There is a mismatch between the treesitter grammars I'm using for OCaml, there are two grammars I need loaded depending on what type of file is open. Regular source files are in .ml files and interface files are in .mli. I have two defconst for each type of source file and a setup of:

(define-combobulate-language
 :name ocaml
 :language ocaml
 :major-modes (ocaml-ts-mode neocaml-mode tuareg-mode)
 :custom combobulate-ocaml-definitions
 :setup-fn combobulate-ocaml-setup)

(define-combobulate-language
 :name ocaml-interface
 :language ocaml-interface
 :major-modes (ocamli-ts-mode neocamli-mode tuareg-mode)
 :custom combobulate-ocaml-interface-definitions
 :setup-fn combobulate-ocaml-setup)

Perhaps I need to combine the two sets of rules into one and combobulate will know which grammar is loaded?

[mickeynp]

I don't know. I do not know this language nor why you would want two different grammars loaded for the same file.

[jgarvin]

@mickeynp I believe @TMGilchrist is saying the language has 2 file types that each have different grammars, not that they want 2 grammars loaded for the same file. Ocaml has a distinction between the file containing the implementations of functions and the files that contain the declarations/interface, similar to the difference between source files and header files in C, except that unlike C the syntax is also slightly different between the two file types. Emacs ocaml extensions typically handle both types with a single major mode by just changing some internal flags rather than having two separate modes. So the problem is 2 grammars for 1 mode, not 2 grammars for 1 file.

[mickeynp]

Aha. Well then you'll need two major modes -- one for interface and one for implementation -- and two combobulate define language I think?

Unless the TS grammars are somehow merged in which case you presumably can get away with one.

[tmcgilchrist]

This is what I hacked in this PR tmcgilchrist#7 along with some changes to register grammars with underscores in their names e.g. ocaml_interface. It's not pretty code right now, but it seems to work. It'll get tidied up before I attempt to upstream that particular change.