Prevent Emacs from asking "kill anyway" when only text properties have changed

Question

I have a read-only file that I open in Emacs. I have created a keyboard shortcut to hide some regions of the file. For example:

(defun hide-beginning ()
  (interactive)
  (setq buffer-read-only nil)
  (put-text-property 1 10 'invisible t)
  (setq buffer-read-only t)
  )

However, when I exit the buffer with kill-buffer or rather ido-kill-buffer, I get the question

Buffer myfile.txt modified; kill anyway? (yes or no)

How can this question be avoided?

If you only care about changes that occur in this function, you can probably use buffer-modified-p and set-buffer-modified-p to store the modified state and restore it after changing the properties. — ChrisR, Dec 02 '14 at 19:59
Thanks it works perfectly! I just put (set-buffer-modified-p nil) after I changed the property.. — Håkon Hægland, Dec 02 '14 at 20:02
@ChrisR: I didn't see your comment before posting. Could you post your comment as an answer so that it could be accepted? — Dan, Dec 02 '14 at 20:05
@HåkonHægland You might want to save value of calling buffer-modified-p at the beginning of the function. That way, if something else modifies the buffer, you will still be prompted about it. — zck, Dec 02 '14 at 21:36
If you're setting text properties which aren't meant to be saved in a file or copied and yanked, you should probably use overlays instead of text properties. See also What are overlays for, and how do they differ from text properties? — Gilles 'SO- stop being evil', Dec 04 '14 at 02:20

score 5 · Accepted Answer · answered Dec 02 '14 at 20:07

5

If you only care about changes that occur in this function, you can use buffer-modified-p and set-buffer-modified-p to store the modified state and restore it after changing the properties.

answered Dec 02 '14 at 20:07

ChrisR

404
2
7

score 2 · Answer 2 · edited Dec 04 '14 at 00:24

2

The manual notes that you can use set-buffer-modified-p with a nil argument to flag the buffer as unmodified. Hence, if you want your function not to modify the buffer, you could do the following:

(defun hide-beginning ()
  (interactive)
  (let ((buffer-read-only nil)
        (buffer-modified (buffer-modified-p)))
    (put-text-property 1 10 'invisible t)
    (set-buffer-modified-p buffer-modified)))

edited Dec 04 '14 at 00:24

Gilles 'SO- stop being evil'

22,318
4
56
92

answered Dec 02 '14 at 20:03

Dan

32,980
7
102
169

Thanks! Don't you need to restore buffer-read-only at the end of the function? – Håkon Hægland Dec 02 '14 at 20:09
1

@HåkonHægland: when you let-bind buffer-read-only, you are binding it dynamically for the lifespan of the function -- basically, you're temporarily overriding its value. – Dan Dec 02 '14 at 20:12
2

@HåkonHægland the best approach is to letbind inhibit-read-only to t. That will cover more cases than just setting buffer-read-only to nil. – Malabarba Dec 02 '14 at 20:48
2

Unconditionally marking the buffer as unmodified is very dangerous. You should do that only if the buffer was unmodified in the first place, otherwise this has a high risk of resulting in unsaved changes sooner or later. – Gilles 'SO- stop being evil' Dec 04 '14 at 00:23

score 2 · Answer 3 · answered Jan 06 '15 at 19:34

The macro with-silent-modifications (in subr.el) seems to exist for exactly this use case:

Execute BODY, pretending it does not modify the buffer.
...
Typically used around modifications of text-properties which do
not really affect the buffer's content.

Prevent Emacs from asking "kill anyway" when only text properties have changed

3 Answers3