Re: Recursive edits in `save-excursion'

[Stuart D. Herring]

> The explanation for this is that save-excursion saves the current
> buffer, but what you set when you switch buffers is the selected
> window's buffer.  save-excursion does not restore that.  The first
> save-excursion restores the original current buffer but doesn't
> switch to it.  The second recursive edit notices that the current
> buffer isn't the same one you've switched to, so it arranges to
> save and restore the current buffer.

So -- just to make sure that I understand correctly -- `recursive-edit'
acts like it was implemented as follows?

(defun recursive-edit ()
  "..."
  (let ((obuf (unless (eq (current-buffer) (window-buffer))
                (current-buffer)))
    (set-buffer (window-buffer))
    (unwind-protect
        (if (eq (catch 'exit (command-loop) nil) t) (signal 'quit nil))
      (if (obuf (set-buffer obuf))))))

In other words, it acts like it is a `save-excursion' embedded if and only
if the current buffer and current window's buffer differ?  So you could
defeat this by doing something like what follows?

(save-excursion (set-buffer (window-buffer)) (recursive-edit) ...)

My concern is not which buffer is current around this form, or in which
buffer the `recursive-edit' starts, but just that the `...' be evaluated
with the buffer current that was current when the recursive edit was
exited.

I also wonder if it would be profitable to use the original form, but
calling `window-buffer' instead of `current-buffer'?

(save-excursion (recursive-edit) (window-buffer))

A quick test with the 21.3 I have in front of me seems to support my
theory (both of the modified examples work as I would expect).  Assuming
I'm correct that they work reliably, is there a reason to prefer one or
the other?

Thanks,
Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or
too sparse, it is because mass-energy conversion has occurred during
shipping.