Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Date: Wed, 18 Jan 2012 14:48:23 -0000
To: Notmuch Mail <notmuch@notmuchmail.org>
Message-ID: <20120118144823.23822.46971@megatron>
From: Patrick Totzke <patricktotzke@googlemail.com>
Subject: [feature request] thread-based queries
Date: Wed, 18 Jan 2012 14:48:23 +0000
Precedence: list

Hi everyone,

A recent discussion on alot's issue tracker[0]  made it clear that it might=
 be useful to have a thread-based
matching for queries as an alternative to the current message based approac=
h:

Consider a thread with two messages, one has a single tag 'flagged', the ot=
her a single tag 'unread'.
Now the thread doesn't match "tag:flagged AND tag:unread" as it doesn't con=
tain any matching messages.
However, listing the tags of this thread accumulates to ['flagged', 'unread=
'].

As far as I know, there is no way one can tell notmuch to evaluate the quer=
y with threads as atomic entities
instead of messages so that the thread in the example above matches. I woul=
d like to see such a switch in =

libnotmuch and the python bindings, maybe as parameter to `notmuch_query_cr=
eate` or as something similar to =

`notmuch_query_set_sort`.

I am aware that I can simulate the intended behaviour by rewriting the quer=
y myself, but I think
such a feature makes sense not only for alot users and should be implemente=
d at a lower level.
Moreover, I wouldn't be surprised if doing this at a library level is less =
painful than higher up
as Xapian has to parse query formulae anyway and hence should offer some co=
nvenience methods for dealing with them.
Please correct me if I'm wrong here or have just overlooked previous discus=
sions/solutions for this issue.

Best,
/p

[o]: https://github.com/pazz/alot/issues/270