From: Pieter Praet <pieter@praet.org>
To: Austin Clements <amdragon@MIT.EDU>
Subject: Re: [PATCH v2] emacs: bad regexp @ `notmuch-search-process-filter'
In-Reply-To: <20110713185721.GI25558@mit.edu>
References: <20110705214234.GA15360@mit.edu>
	<1310416993-31031-1-git-send-email-pieter@praet.org>
	<20110711210532.GC25558@mit.edu> <878vs28dvo.fsf@praet.org>
	<20110713185721.GI25558@mit.edu>
User-Agent: Notmuch/0.6-60-ga0910f1 (http://notmuchmail.org) Emacs/23.1.50.1
	(x86_64-pc-linux-gnu)
Date: Sat, 16 Jul 2011 17:07:12 +0200
Message-ID: <87oc0u6z8f.fsf@praet.org>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="=-=-="
Cc: Notmuch Mail <notmuch@notmuchmail.org>, David Edmondson <dme@dme.org>
Precedence: list

--=-=-=

On Wed, 13 Jul 2011 14:57:21 -0400, Austin Clements <amdragon@MIT.EDU> wrote:
> Quoth Pieter Praet on Jul 13 at  4:16 pm:
> > On Mon, 11 Jul 2011 17:05:32 -0400, Austin Clements <amdragon@MIT.EDU> wrote:
> > > Quoth Pieter Praet on Jul 11 at 10:43 pm:
> > > > TL;DR: I can haz regex pl0x?
> > > 
> > > Oof, what a pain.  I'm happy to change the output format of search; I
> > > hadn't realized how difficult it would be to parse.  In fact, I'm not
> > > sure it's even parsable by regexp, because the message ID's themselves
> > > could contain parens.
> > > 
> > > So what would be a good format?  One possibility would be to
> > > NULL-delimit the query part; as distasteful as I find that, this part
> > > of the search output isn't meant for user consumption.  Though I fear
> > > this is endemic to the dual role the search output currently plays as
> > > both user and computer readable.
> > > 
> > > I've also got the code to do everything using document ID's instead of
> > > message ID's.  As a side-effect, it makes the search output clean and
> > > readily parsable since document ID's are just numbers.  Hence, there
> > > are no quoting or escaping issues (plus the output is much more
> > > compact).  I haven't sent this to the list yet because I haven't had a
> > > chance to benchmark it and determine if the performance benefits make
> > > exposing document ID's worthwhile.
> > 
> > Jamie Zawinski once said/wrote [1]:
> >   'Some people, when confronted with a problem, think "I know,
> >   I'll use regular expressions." Now they have two problems.'
> > 
> > With this in mind, I set out to get rid of this whole regex mess altogether,
> > by populating the search buffer using Notmuch's JSON output instead of doing
> > brittle text matching tricks.
> > 
> > Looking for some documentation, I stumbled upon a long-forgotten gem [2].
> > 
> > David's already done pretty much all of the work for us!
> 
> Yes, similar thoughts were running through my head as I futzed with
> the formatting for this.  My concern with moving to JSON for search
> buffers is that parsing it is about *30 times slower* than the current
> regexp-based approach (0.6 seconds versus 0.02 seconds for a mere 1413
> result search buffer).  I think JSON makes a lot of sense for show
> buffers because there's generally less data and it has a lot of
> complicated structure.  Search results, on the other hand, have a very
> simple, regular, and constrained structure, so JSON doesn't buy us
> nearly as much.

That seems about right. Using the entire Notmuch mailing list archive,
processing JSON ends up taking 23x longer (see test in att).

> JSON is hard to parse because, like the text search output, it's
> designed for human consumption (of course, unlike the text search
> output, it's also designed for computer consumption).  There's
> something to be said for the debuggability and generality of this and
> JSON is very good for exchanging small objects, but it's a remarkably
> inefficient way to exchange large amounts of data between two
> programs.
> 
> I guess what I'm getting at, though it pains me to say it, is perhaps
> search needs a fast, computer-readable interchange format.  The
> structure of the data is so simple and constrained that this could be
> altogether trivial.

I guess that's our only option then. Could you implement it for me?
I'll make sure to rebase my patch series in an acceptable time frame.

An extra output format shouldn't be that much of a problem though, if we
further compartmentalize the code. What are your thoughts on (in the
long term) moving to a plugin-based architecture? Eg. enable something
like this:

  ./input/{Maildir, ...}
  ./output/{plain, JSON, ...}
  ./filters/{crypto, ...}
  ./backends/(Xapian, ...)
  ./uis/{Emacs, VIM, web, ...}

> Or maybe I need a faster computer.

That's what M$ Tech Support would want you to believe :)
What we need is slower computers, so devs are forced to count cycles again.
The rise of netbooks has thankfully done wonders in this respect.

> If anyone is curious, here's how I timed the parsing.
> 
> (defmacro time-it (code)
>   `(let ((start-time (get-internal-run-time)))
>      ,code
>      (float-time (time-subtract (get-internal-run-time) start-time))))
> 
> (with-current-buffer "json"
>   (goto-char (point-min))
>   (time-it (json-read)))
> 
> (with-current-buffer "text"
>   (goto-char (point-min))
>   (time-it
>    (while (re-search-forward "^\\(thread:[0-9A-Fa-f]*\\) \\([^][]*\\) \\(\\[[0-9/]*\\]\\) \\([^;]*\\); \\(.*\\) (\\([^()]*\\))$" nil t))))


Peace

-- 
Pieter

--=-=-=
Content-Type: application/octet-stream
Content-Disposition: attachment; filename=regexp-vs-json.org
Content-Transfer-Encoding: base64
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--=-=-=--