From mboxrd@z Thu Jan  1 00:00:00 1970
From: rlane@club.cc.cmu.edu (Rich Lane)
Date: Thu, 03 Sep 2009 14:44:06 -0400
Subject: [sup-talk] On making inbox-mode and search-results-mode more
	similar
In-Reply-To: <1252002783-sup-3659@masanjin.net>
References: <1251323747-sup-1595@yoom.home.cworth.org>
	<1251341286-sup-9400@zyrg.net> <1252002783-sup-3659@masanjin.net>
Message-ID: <1252003316-sup-2379@zyrg.net>

Excerpts from William Morgan's message of Thu Sep 03 14:36:42 -0400 2009:
> Reformatted excerpts from Rich Lane's message of 2009-08-26:
> > We had a thread a while back about applying label changes immediately
> > and I think the consensus was that that's a good idea. Doing this
> > could simplify a great deal of thread-index-mode code because a fast
> > is_relevant? could be implemented using existing index operations
> > without the special cases for archiving/etc and without creating an
> > inmemory database.
> 
> But it's not just a label issue. Determining whether a given message
> belongs in a general search buffer requires the full engine. And even if
> the query is just on labels, it can contain arbitrary boolean
> expressions, and I don't want to be the one having to parse that, and
> parse it in such a way that it matches what the search engine is doing.

You're right that it require the full engine. If we're immediately
saving the label changes to the (on-disk) index, we can simply query it
and get the correct answer.