bug#66988: 30.0.50; treesit-forward-sexp not working properly in js-ts-mode and tsx-ts-mode

[Loïc Lemaître]

Thanks for the patch ! It fixes the bug.
But unfortunatly, there is another similar bug in treesit-forward-sexp, that you can reproduce with that example:
({(<A></A>)});

(treesit-forward-sexp) does not work as expected for both the opening parentheses and the brace.

I have checked that it is not a regression due to the patch. That said,  the patch changes the results (that are not what we expect in any cases).

Loïc

Le 10/11/2023 à 02:41, Dmitry Gutov a écrit :

> Hi!
>       
>       
>       On 07/11/2023 16:56, Loïc Lemaître wrote:
>       
> > Hi Emacs team,
> >         
> >         
> >         Here the steps to demonstrate the bug :
> >         
> >         
> >         1. Compile Emacs from master branch with tree-sitter support
> >         
> >         2. Install _javascript_ and tsx languages
> >         
> >         3. Run Emacs
> >         
> >         4. Create a new buffer
> >         
> >         5. Turn major mode to either js-ts-mode or tsx-ts-mode
> >         
> >         6. Past the following content into the buffer:
> >         
> >         (
> >         
> >            <div>
> >         
> >            </div>
> >         
> >         );
> >         
> >         7. Place point before opening parenthese
> >         
> >         8. M-x forward-sexp (which will call treesit-forward-sexp)
> >         
> >         
> >         => New position is right after the semi-colon instead of
> >         being before the semi-colon.
> >         
> >         
> >         Note that the bug disappear if the buffer content is changed for
> >         :
> >         
> >         const component = (
> >         
> >            <div>
> >         
> >            </div>
> >         
> >         );
> >         
> >         
> >         But previous content, while not being very usefull, is valid
> >         JSX, as far as I know.
> >         
> >         I use this syntax for unit test purpose, since it is very short.
> >         
> >       
>       
>       Thanks for the report.
>       
>       
>       The patch below should fix it.
>       
>       
>       Yuan, what do you think? A similar change (bos and eos anchors)
>       might be useful for other things and other modes.
>       
>       
>       Alternatively, treesit-thing-settings could be interpreted to
>       imply full matches, then the code using it should not only match
>       against the regexps but also check that the entire string (type
>       name) is matched.
>       
>       
>       Also Cc'ing Theodor.
>       
>       
>       diff --git a/lisp/progmodes/js.el b/lisp/progmodes/js.el
>       
>       index 5a669fdbd42..d81fa9ed322 100644
>       
>       --- a/lisp/progmodes/js.el
>       
>       +++ b/lisp/progmodes/js.el
>       
>       @@ -3843,6 +3843,7 @@ js--treesit-sexp-nodes
>       
>            "undefined"
>       
>            "arguments"
>       
>            "pair"
>       
>       +    "parenthesized_expression"
>       
>            "jsx")
>       
>          "Nodes that designate sexps in _javascript_.
>       
>        See `treesit-thing-settings' for more information.")
>       
>       @@ -3886,7 +3887,7 @@ js-ts-mode
>       
>       
>            (setq-local treesit-thing-settings
>       
>                        `((_javascript_
>       
>       -                   (sexp ,(regexp-opt js--treesit-sexp-nodes))
>       
>       +                   (sexp ,(format "\\`%s\\'" (regexp-opt
>       js--treesit-sexp-nodes)))
>       
>                           (sentence ,(regexp-opt
>       js--treesit-sentence-nodes))
>       
>                           (text ,(regexp-opt '("comment"
>       
>                                                "template_string"))))))
>       
>       
>