Re: Tree-sitter integration on feature/tree-sitter

[Yuan Fu]

> On Jun 20, 2022, at 4:44 AM, Eli Zaretskii <eliz@gnu.org> wrote:
> 
>> From: Yuan Fu <casouri@gmail.com>
>> Date: Sun, 19 Jun 2022 20:00:49 -0700
>> Cc: emacs-devel@gnu.org
>> 
>> 
>> I added navigation functions like treesit-beginning/end-of-defun, and added 
>> search functions like treesit-search-beginning/end. Now I wonder where 
>> should I put them in the manual, do I put them under the treesit section 
>> (Parsing Program Source), or under the relevant existing sections in the 
>> manual?
> 
> The latter, please.
> 
> But why do we need a separate description for the tree-sitter
> variants?  Shouldn't that be automatically supported by
> beginning/end-of-defun, once some switch is thrown to enable
> tree-sitter?
> 
> And if beginning/end-of-defun is for some reason too low-level/basic
> for this role (but if you think so, please explain why), then I think
> we need higher-level functions that by default are just thin wrappers
> around beginning/end-of-defun, and will call tree-sitter versions when
> Emacs is configured to do so.
> 
> I mean, it would be very cumbersome to request that each and every
> major mode which wants to use tree-sitter will have to explicitly call
> treesit-SOMETHING everywhere.

Major mode should set beginning-of-defun-function to 
treesit-beginning-of-defun, not unlike what they already do with major 
mode-specific beginning-of-defun functions. This way major mode has the freedom 
to decide which treesit features it wants to leverage.

> 
>> Treesit-beginning/end-of-defun jumps to the beginning/end of the current 
>> defun form, treesit-search-beginning searches for a query and stops at the 
>> beginning/end of the node that matched the query.
> 
> So you are saying treesit-beginning/end-of-defun don't actually look
> for beginning and end of a function, but for beginning and end of a
> more abstract entity?  Then perhaps it would be wrong to have "defun"
> in their names?  And in that case, maybe a separate section (under
> "Motion") is better after all, since this is no longer "List Motion",
> strictly speaking.

The second sentence describes treesit-search-beginning, not 
treesit-beginning-of-defun, I think you confused the two?

Thanks,
Yuan