bug#62416: 30.0.50; Symbols skipped in the navigation in ruby-ts-mode

[Dmitry Gutov]

On 03/04/2023 01:34, Yuan Fu wrote:
> > On Mar 30, 2023, at 2:32 AM, Dmitry Gutov<dgutov@yandex.ru>  wrote:
> >
> > On 30/03/2023 10:47, Yuan Fu wrote:
> >
> > > > However, there are still a lot of more things that need fixing.
> > > > When point is on the left curly bracket in
> > > >
> > > >    b = %Q{This is a "string"}
> > > >
> > > > 'C-M-f' doesn't move to the right curly bracket.
> > > > Also double quotes inside the string are not matched by 'C-M-f'.
> > > >
> > > > In
> > > >
> > > >    d = %(hello (nested) world)
> > > >
> > > > 'C-M-f' doesn't move to the closing parens from opening parens.
> > > Have someone fixed these two cases? Because when I tried to invoke
> > > (treesit-forward-sexp), point moved to the closing bracket/paren.
> >  From which position? When point is right before '{', it doesn't move in my 
> > testing. It does move when it was before '%'.
> Ok, I see it. I’ll try to see what’s going on when I find some time.

There is nothing surprising in this behavior, given the current 
implementation: it doesn't examine the text in the buffer, just uses the 
parse tree,

And there is no trace of these parens/braces in the parse tree.

> > > Anyway, I just wonder if there’s any fundamental shortcoming with how
> > > treesit-beginning/end-of-thing works?
> > I don't know. Seems like this method is good for a lot of things, but some 
> > fiddly details are going to be different from the default forward-sexp.
> >
> > ruby-mode's sexp navigation is also not ideal in its own way, and it's been 
> > useful anyway.
> One thing I noticed is that treesit-forward-sexp uses 
> treesit-beginning/end-of-thing, which jumps out of the parent when there is no 
> more siblings to go to. The default forward-sexp obviously doesn’t do this. 
> Perhaps we should stay in the same level in tree-sitter-forward-sexp too.

That's a different aspect of its behavior. One that makes 
'backward-up-list' fail to work, IIUC.