gnus-undo-mode and nnimap

gnus-undo-mode and nnimap

[Toby Speight]

Try this:

1. Enter a nnimap group with unread articles.

2. Read the articles and leave the group.  The *Group* buffer will
   show the group as having 0 unread articles.

3. M-x gnus-undo  (or C-_, by default).  The *Group* buffer will now
   show the group as having >0 unread articles.

4. Enter the group.  The articles you thought you had, er, unread (by
   undoing the read) are marked with the read mark.

I know this is the same effect as the generally useful feature of
accepting the server's marks over Gnus's, but WIBNI gnus-undo actually
told the backend to change its stored marks?  (Or if that's already
the case, WIBNI nnimap DTRT when it's told?)

Re: gnus-undo-mode and nnimap

[Simon Josefsson]

Toby Speight <streapadair@gmx.net> writes:

> I know this is the same effect as the generally useful feature of
> accepting the server's marks over Gnus's, but WIBNI gnus-undo actually
> told the backend to change its stored marks?

Yup.  See if it works and I could commit it.

2000-09-12  Simon Josefsson  <simon@josefsson.org>

	* gnus-sum.el (gnus-update-read-articles): Undo
	`gnus-request-set-mark' operation.

Index: gnus-sum.el
===================================================================
RCS file: /usr/local/cvsroot/gnus/lisp/gnus-sum.el,v
retrieving revision 5.131
diff -w -u -u -w -r5.131 gnus-sum.el
--- gnus-sum.el	2000/09/03 19:52:43	5.131
+++ gnus-sum.el	2000/09/12 13:56:38
@@ -9493,13 +9493,7 @@
       (if compute
 	  read
 	(save-excursion
-	  (set-buffer gnus-group-buffer)
-	  (gnus-undo-register
-	    `(progn
-	       (gnus-info-set-marks ',info ',(gnus-info-marks info) t)
-	       (gnus-info-set-read ',info ',(gnus-info-read info))
-	       (gnus-get-unread-articles-in-group ',info (gnus-active ,group))
-	       (gnus-group-update-group ,group t))))
+	  (let (setmarkundo)
 	;; Propagate the read marks to the backend.
 	(if (gnus-check-backend-function 'request-set-mark group)
 	    (let ((del (gnus-remove-from-range (gnus-info-read info) read))
@@ -9509,7 +9503,19 @@
 		  (error "Can't open server for %s" group))
 		(gnus-request-set-mark
 		 group (delq nil (list (if add (list add 'add '(read)))
-				       (if del (list del 'del '(read)))))))))
+					   (if del (list del 'del '(read))))))
+		    (setq setmarkundo
+			  `(gnus-request-set-mark
+			    ,group ',(delq nil (list (if del (list del 'add '(read)))
+						     (if add (list add 'del '(read))))))))))
+	    (set-buffer gnus-group-buffer)
+	    (gnus-undo-register
+	      `(progn
+		 (gnus-info-set-marks ',info ',(gnus-info-marks info) t)
+		 (gnus-info-set-read ',info ',(gnus-info-read info))
+		 (gnus-get-unread-articles-in-group ',info (gnus-active ,group))
+		 (gnus-group-update-group ,group t)
+		 ,setmarkundo))))
 	;; Enter this list into the group info.
 	(gnus-info-set-read info read)
 	;; Set the number of unread articles in gnus-newsrc-hashtb.

Re: gnus-undo-mode and nnimap

[Toby Speight]

0> Toby Speight <streapadair@gmx.net> writes:

>> I know this is the same effect as the generally useful feature of
>> accepting the server's marks over Gnus's, but WIBNI gnus-undo
>> actually told the backend to change its stored marks?


0> In article <iluem2paex5.fsf@barbar.josefsson.org>,
0> Simon Josefsson <URL:mailto:simon@josefsson.org> ("Simon") wrote:

Simon> Yup.  See if it works and I could commit it.

Thanks for the quick response, Simon!

It works a little better now - read articles re-appear with the "M"
mark (duplicate suppressed), which I guess is a bug with duplicate
suppression, not nnimap...

But if I tick an article and then undo it, it still appears as
ticked.  Though I haven't tried, I expect the same is true of other
changes (ticked->dormant, perhaps?).

Re: gnus-undo-mode and nnimap

[Simon Josefsson]

Toby Speight <streapadair@gmx.net> writes:

> It works a little better now - read articles re-appear with the "M"
> mark (duplicate suppressed), which I guess is a bug with duplicate
> suppression, not nnimap...

Yes, probably.

> But if I tick an article and then undo it, it still appears as
> ticked.  Though I haven't tried, I expect the same is true of other
> changes (ticked->dormant, perhaps?).

No, only read-marks are undoable, other marks have never been undoable
even in other backends.  Perhaps we should change this?

It has some complications though.  Should it be possible to undo a
"Answered" flag?  What would that mean?  And if you undo a "dormant"
flag, should you remove the article from the cache?  What about
"Saved" flags?  "Downloadable" flags?  "Bookmark" flags?  "Expire"
flags when using total/auto-expire?

Perhaps we should only allow undo of tick flags.

Re: gnus-undo-mode and nnimap

[Toby Speight]

0> In article <iluhf7kqxut.fsf@barbar.josefsson.org>,
0> Simon Josefsson <URL:mailto:simon@josefsson.org> ("Simon") wrote:

Simon> Toby Speight <streapadair@gmx.net> writes:

>> ... if I tick an article and then undo it, it still appears as
>> ticked.  Though I haven't tried, I expect the same is true of other
>> changes (ticked->dormant, perhaps?).

Simon> No, only read-marks are undoable, other marks have never been
Simon> undoable even in other backends.

Ah.  I didn't check this.


Simon> Perhaps we should change this?
Simon>
Simon> It has some complications though.  Should it be possible to
Simon> undo a "Answered" flag?  What would that mean?  And if you undo
Simon> a "dormant" flag, should you remove the article from the cache?
Simon> What about "Saved" flags?  "Downloadable" flags?  "Bookmark"
Simon> flags?  "Expire" flags when using total/auto-expire?
Simon>
Simon> Perhaps we should only allow undo of tick flags.

I think it would be nice (but not essential) to support as much as is
meaningful ;-)

The question of undoing cache entry for ticked/dormant (or, I
guess, manually caching) is particularly tricky, given that it may
be impossible to re-do (if the article has been expired in NNTP,
or deleted from another client in IMAP).

Undoing answers doesn't seem reasonable, given that there's no way to
cancel mail messages - IMO the flag should reflect the state of the
Real World.

It does look like the undoability of flags and marks needs to be
considered on a case-by-case basis :-(.  As a start, I'd suggest all
the marks in "Read Articles" and "Unread Articles" in the manual, and
probably none of the ones in "Other Marks".

Having nnimap consistent with the other backends is a Good Thing, of
course, so I'm pleased with your patch :-).

Re: gnus-undo-mode and nnimap

[Kai Großjohann]

On 13 Sep 2000, Simon Josefsson wrote:

> It has some complications though.  Should it be possible to undo a
> "Answered" flag?  What would that mean?  And if you undo a "dormant"
> flag, should you remove the article from the cache?  What about
> "Saved" flags?  "Downloadable" flags?  "Bookmark" flags?  "Expire"
> flags when using total/auto-expire?
> 
> Perhaps we should only allow undo of tick flags.

Note that the ticked and dormant marks have the same issue w.r.t. the
cache.  Hm.  When gnus-use-cache is t and a message was entered into
the cache because it was ticked or marked as dormant, then marking it
as read does not immediately remove it from the cache.  I'm not sure
when it is removed from the cache, though.  Maybe when the group is
exited?

I think as a first approach, the unread, read, and expirable flags
should be taken into consideration for undoing.  It does not make
sense to undo an answered flag, and undoing a saved flag would mean
deleting the file in question.  Not what we want, I think.

kai
-- 
I like BOTH kinds of music.