bug#59662: 29.0.50; [PATCH] Add treesit--indent-defun

[Yuan Fu]

Theodor Thornhill <theo@thornhill.no> writes:

> Eli Zaretskii <eliz@gnu.org> writes:
>
>>> Cc: eliz@gnu.org, casouri@gmail.org
>>> From: Theodor Thornhill <theo@thornhill.no>
>>> Date: Mon, 28 Nov 2022 20:32:01 +0100
>>> 
>>> I've added a function to treesit.el, for use in
>>> treesit-major-mode-setup.  Because the treesit-defun-type-regexp gives
>>> us "defuns" for free, we can set fill-paragraph-function to this new
>>> function, thus enabling quick formatting, and some sensible default for
>>> fill-paragraph.  It aims to mirror c-indent-defun.
>>> 
>>> I think this is a nice way to get this functionality for free, but I'm
>>> not 100% whether this is considered ok or not.
>>
>> Sounds good, but why did you think it wouldn't be OK?  Anything here that
>> doesn't meet the eye?
>>
>
> Not really, but see below answer.
>
>>> An alternative could be to add a 'treesit-mode-map' where we can
>>> auto-enable such constructs.
>>
>> I think this is less desirable.
>>
>> Yuan, WDYT?
>>
>>> @@ -1698,7 +1713,10 @@ treesit-major-mode-setup
>>>    ;; Navigation.
>>>    (when treesit-defun-type-regexp
>>>      (setq-local beginning-of-defun-function #'treesit-beginning-of-defun)
>>> -    (setq-local end-of-defun-function #'treesit-end-of-defun)))
>>> +    (setq-local end-of-defun-function #'treesit-end-of-defun))
>>> +  ;; Filling
>>> +  (when (and treesit-defun-type-regexp treesit-simple-indent-rules)
>>> +    (setq-local fill-paragraph-function #'treesit--indent-defun)))
>>
>> I'm a bit confused: if the function's name is treesit--indent-defun, and it
>> uses treesit-indent-region to do its job, why do we assign it to
>> fill-paragraph-function, which is supposed to _fill_, not to _indent_?
>
> This is why I was thinking it would maybe be better to put it into a
> treesit-mode-map that major-modes can inherit from, thus binding it to
> things such as C-c C-q.  The reason I put it in filling was because that
> is a common key to press in everything _but_ prog-modes.  And prog-modes
> seem to mostly just turn it off if not inside of comments etc.  This
> would behave just like that, except we would reformat/reindent/refill
> code.
>
> In a way filling _is_ formatting/reindenting, at least that's how I look
> at it.

I see, so you want to implement C-c C-q in c-mode. But why don’t we
make treesit--indent-defun a command and bind it in C-c C-q in major modes? 
Filling is not indent IMO. If you fill a list in elisp code it wraps long lines 
instead of indent (IIRC).

Also, this is really not tree-sitter specific, it doesn’t require any
tree-sitter feature to accomplish: indent-defun only needs mark-defun
and indent-region, both are supported by practically any major mode.

Normally this kind of thing goes into lisp.el, alongside commands like
fill-paragraph, indent-region, beginning-of-defun, etc, and claim a
global keybinding. But maybe we only want it to live under C-c prefix,
in that case I guess we can bind it in prog-mode-map, under C-c C-q?

I’m thinking something like

(defun prog-mode-indent-defun ()
  (interactive)
  (mark-defun)
  (indent-region (region-beginning) (region-end)))

(defvar prog-mode-map
  (...
   (define-key map (kbd "C-c C-q") #'prog-mode-indent-defun)))

Yuan