[PATCH] emacs-show: open excluded matches if no other matches

[Mark Walters]

On Fri, 13 Apr 2012, Jani Nikula <jani at nikula.org> wrote:
> On Thu, 12 Apr 2012 22:20:52 +0300, Tomi Ollila <tomi.ollila at iki.fi> wrote:
>> On Thu, Apr 12 2012, Jameson Graef Rollins wrote:
>> 
>> > On Thu, Mar 15 2012, Mark Walters <markwalters1009 at gmail.com> wrote:
>> >> Currently emacs show does not open matching but excluded
>> >> messages. This is normally the desired behaviour but is probably not
>> >> ideal if only excluded messages match. This patch opens all the
>> >> matching (necessarily excluded) messages in this case and goes to the
>> >> first one.
>> >
>> > Hi, Mark.  I have tested this patch and it LGTM.  I think this is the
>> > behavior we want, since it's no fun to open up a show buffer and have
>> > all messages in the thread closed.  That doesn't make much sense.
>> 
>> I disagree: If I open up a show buffer I'd like to have all messages
>> in the thread closed if search terms do not match. But that is my
>> opinion and inconsistent with current behaviour.
>> 
>> (Therefore I already +1:d this patch)
>> 
>> An example: I have thread of 100 messages and 1 of those is unread.
>> Search terms is tag:unread. I enter the thread: 99 messages closed
>> and this one open. Now unread is automatically removed. If I go back
>> to search view (by pressing q) and re-choose the same thread,
>> current behaviour is to open all 100 messages. I'd like to see all
>> 100 messages closed.
>
> IIUC this would still show all messages closed. Messages that don't
> match will be closed. The change is that if there are messages that
> match, but all of them are excluded, then open the messages anyway.

This patch does not affect the behaviour in this case, but the code
existing code does expand all messages in a thread if none match (the
use of basic-args and args in notmuch-show-build-buffer).

> In your example, only if you have *another* message that is both
> tag:unread and tag:deleted (or some other excluded tag) it would show up
> the second time you enter show view.
>
> I actually think I'd still want the above to show all messages closed,
> unless the excluded tag was explicitly specified in the search. The idea
> of exclusions is to reduce "noise" in the view, why open them? If you
> have those 100 messages in a thread, and you want to see the interesting
> (i.e. non-excluded) messages, I'd rather see 0 open if 0 are
> interesting, instead of 100 non-interesting if they are all excluded.

I think that is entirely plausible as a way to go (though I would also
be happy with the behaviour after this patch). But I think that should
be a subsequent patch because to be consistent it would have to change
the existing behaviour for no matches (and no excludes) to be all closed.

> Perhaps a function to toggle the open/close of excluded messages would
> be a better idea. 

That might be useful regardless. I could add a function to do this,
which could be called by the above patch if we want the behaviour it
introduces. Do you think that is worth doing?

> But then I don't use exclusions ATM, so maybe I just
> don't understand, and won't be affected either way. *shrug*.

That is one problem: there are lots of potential use cases and its not
clear which of them people will use and which we should encourage. I use
a `bizarre' one: I exclude all of my mailing list mail (including
notmuch) so that it doesn't show up in my personal mail searches. So I
guess I am using the excludes to return to the bad old days of
`folders'. (I am  not recommending this to anyone, I partly did it as a
good stress test for the exclude stuff, but it does seem to mostly
work.)


Best wishes

Mark