From: Franz Fellner <alpine.art.de@gmail.com>
To: <notmuch@notmuchmail.org>
Subject: Re: Performance-issues with large threads
Date: Wed, 24 Jul 2013 07:39:48 +0200
User-Agent: trojita/v0.3.93-100-gbbf4022-dirty; Qt/4.8.4; X11; Linux; 
MIME-Version: 1.0
Message-ID: <418182bd-9a04-4e1d-8392-ce16d64681c0@gmail.com>
In-Reply-To: <87wqoh77zh.fsf@qmul.ac.uk>
References: <027633b9-200d-4174-a8b5-72daff923b93@gmail.com>
	<87wqoh77zh.fsf@qmul.ac.uk>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: quoted-printable
Precedence: list

Hej,=0A=0A> How long does =0A>=0A> notmuch show --format=3Djson thread:=
thread-id  > /dev/null=0A>=0A> take? Is it much faster after the first =
run? What about the same with=0A> the added cli option --body=3Dfalse?=0A=
=0A[7:23] $ time notmuch show --format=3Djson thread:00000000000002d1 >=
|/dev/null=0Anotmuch show --format=3Djson thread:00000000000002d1 >| /d=
ev/null  0,33s user 0,09s system 8% cpu 4,571 total=0A[7:25] $ time not=
much show --format=3Djson thread:00000000000002d1 >|/dev/null=0Anotmuch=
 show --format=3Djson thread:00000000000002d1 >| /dev/null  0,12s user =
0,02s system 99% cpu 0,139 total=0A[7:25] $ time notmuch show --format=3D=
json --body=3Dfalse thread:00000000000002d1 >|/dev/null=0Anotmuch show =
--format=3Djson --body=3Dfalse thread:00000000000002d1 >| /dev/null  0,=
30s user 0,08s system 8% cpu 4,446 total=0A[7:25] $ time notmuch show -=
-format=3Djson --body=3Dfalse thread:00000000000002d1 >|/dev/null=0Anot=
much show --format=3Djson --body=3Dfalse thread:00000000000002d1 >| /de=
v/null  0,10s user 0,01s system 99% cpu 0,119 total=0A[7:26] $=0A=0ASo =
fetching without cache takes 4.5 secs, with only 0.1 -> The issue could=
 be in parsing the result inside notmuch-vim/alot/...=0A=0A> Do the mes=
sages have html parts?=0AYes, most of them do (unless you reply by mail=
 and only use plaintext).=0A=0A> Also what disk are you using (and does=
 it have any other significant=0A> activity happening)? And what versio=
n of notmuch are you using?=0AI use a HDD with 5400 RPM. The most inter=
esting activity is ~80-100% CPU usage during loading, all the 13 secs -=
 quite a battery killer ;)=0AI regularly build from latest git master (=
due to the vim-client).=0A=0A>=0A> Best wishes=0A>=0A> Mark=0A>=0A>=0A>=
=0A>=0A>=0A>=0A> Franz Fellner <alpine.art.de@gmail.com> writes:=0A>=0A=
>> Hi,=0A>> =0A>> "Large" in the context of this issue is >20 replies, =
I have to =0A>> wait ~1 sec until the thread shows up. With larger =0A>=
> conversations it get's worse (a bug report on bugs.kde.org with =0A>>=
 ~400 replies takes ~13 secs. to load).=0A>> This is due to the fact, t=
hat notmuch-{vim,emacs} load the =0A>> complete mail including the body=
.=0A>> ner does a great job here, it loads all headers but only the =0A=
>> body of the first unread mail, resulting in instant loading =0A>> ev=
en of gigantic conversations. It would be terrific to see =0A>> somethi=
ng like that, at least in my favored vim-client ;)=0A>> =0A>> Thx ...=0A=
>=0A>=0A