GNU bug report logs - #66989
30.0.50; tree-sitter: treesit-defun-at-point wrong behavior in python-ts-mode

Previous Next

Full log

Message #17 received at 66989 <at> debbugs.gnu.org (full text, mbox):

From: Yuan Fu <casouri <at> gmail.com>
To: Zubarev Dv <dvzubarev <at> yandex.ru>
Cc: 66989 <at> debbugs.gnu.org
Subject: Re: bug#66989: 30.0.50; tree-sitter: treesit-defun-at-point wrong
 behavior in python-ts-mode
Date: Thu, 9 Nov 2023 21:46:01 -0800

> On Nov 9, 2023, at 12:13 AM, Yuan Fu <casouri <at> gmail.com> wrote:
> 
> 
> 
>> On Nov 8, 2023, at 12:37 PM, Zubarev Dv <dvzubarev <at> yandex.ru> wrote:
>> 
>> Initially I thought that it is the second definition.  It seems to be more intuitive. But after reading code, I was sure its the first case.
>> 
>> But I actually don't sure that all three "things" are needed at once.
>> For example,  I think for `thing-at-point` only enclosing parent is needed.
>> If parent is nil, it may optionally fallback to the next sibling thing.
>> `treesit--navigate-thing` uses only `parent` and `next` or `prev`.
>> I was playing with creating evil text objects, based on your awesome
>> work with "things".
>> I ended up extracting the second part (;; 2. Find the parent) from `treesit--things-around`.
>> Since I only need to find enclosing parent and I don't want to pay for
>> what I don't use.
>> So it seems no function uses all results from `treesit--things-around`.
>> 
>> Maybe it makes sense to decompose `treesit--things-around` into two
>> functions:
>> 1. (treesit--enclosing-thing (pos thing)) - returns node or nil if no
>> enclosing parent
>> 2. (treesit--find-sibling (pos thing dir)) - return sibling node in
>> specified direction.
> 
> Makes sense. I think separating into two functions is ok. For treesit--enclosing-thing, we might as well call it treesit--thing-at-point ;-) Tentative plan: retire treesit—things-around, add treesit--enclosing-thing (as --thing-at-point), add treesit--find-sibling (as --sibling-thing), add treesit-node-enclose-p which tests whether a node encloses another node.
> 
>> 
>> BTW, I've found another unexpected behavior on the same code:
>> 
>> 
>> 1. python-ts-mode
>> 
>> def premain():
>>   pa|ss
>> def main():
>>   pass
>> 2. after
>> (goto-char (treesit--navigate-thing (point) 1 'end treesit-defun-type-regexp))
>> 
>> def premain():
>>   pass
>> def main():
>>   pass|
>> 
>> cursor in the end of the second function
>> 
>> 3. but if you move cursor to the beginning of pass
>> 
>> def premain():
>>   |pass
>> def main():
>>   pass
>> 
>> 4.  after
>> (goto-char (treesit--navigate-thing (point) 1 'end treesit-defun-type-regexp))
>> 
>> def premain():
>>   pass|
>> def main():
>>   pass
>> 
>> This behavior does not affected by the fix proposed in the first message.
> 
> In the first case, treesit--navigate-thing returns a non-nil NEXT node. And treesit—navigate will prioritize NEXT node over PARENT node (because it assumes PARENT node always encloses NEXT node). Fixing the original problem should fix this too.
> 
> Yuan

I’ve pushed to master four new functions: treesit--thing-prev, treesit--thing-next, treesit--thing-at, treesit-node-enclosed-p. And I updated other functions to use the new functions. Now your original example and the later ones should all work right.

Yuan

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.