That fixes the problem!

However, the reason I initially put the reset inside the `(unless
(buffer-live-p treesit--explorer-buffer) ...)`
in `treesit-explore-mode` was b/c it looked like there was an optimization
happening in
`treesit--explorer-refresh` where it does this check

           ;; If we didn't edit the buffer nor change the top-level
           ;; node, don't redraw the whole syntax tree.
           (highlight-only (treesit-node-eq
                            top-level treesit--explorer-last-node))

I don't know if that is something you'd want to keep, but just pointing it
out in case.  I think
the initial patch works as well, but still allows that check to work when
the explorer buffer hasn't
been killed.

Thankyou!

On Wed, Oct 18, 2023 at 9:36 PM Yuan Fu <casouri@gmail.com> wrote:

>
>
> > On Oct 14, 2023, at 9:20 PM, nvp <noah.v.peart@gmail.com> wrote:
> >
> > Hi,
> > The patch is supposed to reset `treesit--explorer-last-node` in the
> source buffer, just before the `(with-current-buffer
> treesit--explorer-buffer ...)`.
> > Upon trying to reproduce it now, I realized it's harder to reproduce
> than I had thought -- sorry about that.
> > I noticed the bug (if it is a bug) initially when I was adding a
> function to jump b/w source and explorer buffers, like the following:
> >
> > (defun my-treesit-explorer-jump ()
> >   "Pop b/w source and explorer buffers."
> >   (interactive)
> >   (let ((buf
> >          (cond
> >           ((eq major-mode 'treesit--explorer-tree-mode)
> >            (when (buffer-live-p treesit--explorer-source-buffer)
> >              treesit--explorer-source-buffer))
> >           (t
> >            (unless (and treesit-explore-mode
> >                         (buffer-live-p treesit--explorer-buffer))
> >              ;; *** Without the reset here, the explorer buffer doesn't
> >              ;; get redrawn the first time, when
> treesit--explorer-last-node
> >              ;; is non-nil in the source buffer ***
> >              ;; (setq-local treesit--explorer-last-node nil)
> >              (cl-letf (((symbol-function (function completing-read))
> >                         (lambda (&rest _) (symbol-name
> (treesit-language-at (point))))))
> >                (treesit-explore-mode 1)))
> >            treesit--explorer-buffer))))
> >     (pop-to-buffer buf)))
> >
> > Let me give a more precise recipe to reproduce:
> > 1. From a c++-ts-mode buffer, call `treesit-explorer-mode`, select
> `cpp`.  Now there should be an explorer buffer.
> > 2. Kill the associated explorer buffer.
> > 3. Now, back in the c++-ts-mode buffer, `treesit--explorer-last-node`
> should still have a value.
> > 4. From that c++-ts-mode buffer, call `my-treesit-explorer-jump`, and
> the explorer buffer should be empty, until
> > switching back to the source buffer.
> >
> > This seems to me to be caused by `treesit--explorer-post-command` not
> running until the source
> > buffer is active again.
>
> Thank you, I think I see the problem now. Could you try the below patch
> and see it fixes your problem? Also, sorry for the late reply, I meant to
> reply sooner but couldn’t find the time to figure out what exact was the
> cause :-) I was initially a bit confused since we already do set last-node
> to nil.
>
> Yuan
>
>
>
>