Date: Sat, 11 Jun 2016 16:32:14 +0000
From: Gaute Hope <eg@gaute.vetsj.com>
Subject: Re: [PATCH] WIP: regexp matching in 'subject' and 'from'
To: David Bremner <david@tethera.net>, Austin Clements
 <aclements@csail.mit.edu>
Cc: sfischme@uwaterloo.ca, notmuch <notmuch@notmuchmail.org>
References: <1465265149-7174-1-git-send-email-david@tethera.net>
 <1465525688-30913-1-git-send-email-david@tethera.net>
 <1465547660-astroid-0-nudmv20lbk-1296@strange>
 <87a8itxpu7.fsf@zancas.localnet>
In-Reply-To: <87a8itxpu7.fsf@zancas.localnet>
User-Agent: astroid/v0.5-221-g4c2c7173 (https://github.com/gauteh/astroid)
Message-Id: <1465662533-astroid-3-6vuqm3zu54-1296@strange>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: quoted-printable
Precedence: list

David Bremner writes on juni 10, 2016 13:09:
> Gaute Hope <eg@gaute.vetsj.com> writes:
>=20
>>
>> Cool!
>>
>> Would it break a lot of things if you just replace the original prefix?
>=20
> It would change the matching behaviour. I guess there are people that
> like the current "sloppy" matching of from: and subject:.  In my
> not-very-scientific tests, it is a factor of 5 to 10 times slower to do
> regexp search, which makes sense because it is effectively post
> processing the results from Xapian. At least on my system it seems fast
> enough to be usable interactively, but that is a pretty shocking
> performance regression. And I know there are people with more mail on
> slower systems.

Maybe we could check if the search string contains a regexp and decide
whether to pre-process it on the background of that? I think that would
make the interface more user-friendly. You'd just always use search
whether you decide that you need to put in some regexp or not.

>=20
>> Could it be made to work on the message body?
>=20
> See Austin's previous reply for the details, but basically no; these
> "values" index in terms of whole strings, while the body is indexed by
> terms (roughly, words). In principle we could add a value slot for the
> body, but I think that would at least double the size of the database
> (maybe more).
>=20

I would rather have double the db and be able wildcard beginning of
terms. If it is not too much maintaining overhead it might be made
optional?


Regards, Gaute

=