Rewrite internals

[purcell]

• Allows rectangle-mark-mode to work as expected (fixes #9)
• Uses yank-handler, so no need to replace the yank command
• Improved preservation of point, generally to improve correctness. However, after yanking full lines, the point is now preserved, not placed at the line beginning, so feedback would be appreciated regarding how this feels.
• Default override for comment-dwim is changed, and now uses prefix to denote the number of lines to comment/uncomment, rather than comment level
• Removed the obsolete whole-line-or-region-mode alias: use either the -local-mode or the -global-mode now.
• Removed the old-fashioned "turn-on" function
• Greatly improved tests

Plan would be to release this as v2.0

[purcell]

save column position, move to the beginning of the line, insert the text, restore column position

Yes, that's reasonable, and I can get that working sanely here, but the issue of mark position after C-w C-w C-y is still there. Can you share the code for your yank-handler?

[gusbrs]

Sure:

(defun gb/kill-dwim-whole-line-yank-handler (str)
  "Yank-handler for `gb/kill-dwim-base-call'."
  (let ((saved-column (current-column))
        (whole-line))
    ;; A whole-line killed from end of file may not have a trailing newline,
    ;; add one if missing.
    (unless (string-match "\n\\'" str)
      (setq str (concat str "\n")))
    ;; Check for active region and delsel before moving to BOL.
    ;; But we cannot check if the region is empty here, because
    ;; 'delete-selection-pre-hook' has already deleted the selection at this
    ;; stage, so that '(region-beginning)' and '(region-end)' will always be
    ;; equal.
    (unless (and (region-active-p) delete-selection-mode)
      (setq whole-line t)
      (beginning-of-line))
    (let ((beg-pos (point))
          (end-pos))
      ;; So we can honor the behavior of "C-u" to 'yank'.
      (push-mark (point) t)
      (insert str)
      (setq end-pos (point))
      ;; Remove excluded properties from yanked text.  It is necessary to use
      ;; the 'NOEXCLUDE' argument of 'yank-handler' and do this here, because
      ;; the handler moves point after inserting the text (see
      ;; 'image-file-yank-handler' for an example, and the docstring of
      ;; 'yank').
      (remove-yank-excluded-properties beg-pos end-pos)
      ;; Set yank-undo-function in order to support yank-pop properly (from
      ;; 'whole-line-or-region.el')
      (setq yank-undo-function
            (lambda (_beg _end) (delete-region beg-pos end-pos))))
    (when whole-line
      (move-to-column saved-column))))

[purcell]

I pushed some changes to use that approach, and now I feel like my yank handler is functionally identical to yours. But still, if I add the final assertion (should (equal (point-min) (mark))) to the wlr-consecutive-kill-region-combines-them test, the mark is not left in the desired place. So I'm not really sure what the issue is there.

[purcell]

Also I realised yesterday I'm not doing anything to test my handling of delete-selection-mode, which is annoying to use in tests because it's a global mode.

[gusbrs]

Try:

(defmacro whole-line-or-region-filter-with-yank-handler (&rest body)
  "Execute BODY with `filter-buffer-substring-function' bound.
The binding ensure killed strings have a yank handler attached."
  (let ((orig (cl-gensym)))
    `(let* ((,orig filter-buffer-substring-function)
            (filter-buffer-substring-function
             (lambda (&rest args)
               (let ((s (apply ,orig args)))
                 (put-text-property 0 (length s) 'yank-handler
                                    '(whole-line-or-region-yank-handler nil t)
                                    s)
                 s))))
       ,@body)))

But don't ask me why.... :shrug:

[purcell]

O M G

[gusbrs]

O M G

:smile: That's called "coding by luck".

I essentially tried to see what was different here. And that was it.

Again, I have no idea why propertize didn't work...

[purcell]

It looks like propertize adds the property to a copy of the original string, but it's indeed hard to see why that should make a difference.

Anyhow, I'm going to merge this now and tag it as 2.0. Thanks so much!

[gusbrs]

Anyhow, I'm going to merge this now and tag it as 2.0. Thanks so much!

Great! Thank you!

[andersjohansson]

This happened quickly! I offered to help testing changes in #9, but this seemed to involve many subtleties beyond my experience of Emacs’ killing and yanking functionality. And all this sorted out while I hadn’t even got to trying out the new code :slightly_smiling_face:. I’ll gladly try out 2.0 now and report any issues. Thank you!

[purcell]

And all this sorted out while I hadn’t even got to trying out the new code 🙂. I’ll gladly try out 2.0 now and report any issues. Thank you!

I cheated by doing most of the work months ago and switching to it locally over the last month before quickly releasing it. I'm a compulsive tidier.