[MERGE] faster extraction of plain texts from knits

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

James Westby wrote:
> On (09/08/07 09:48), Ian Clatworthy wrote:
>> John Arbash Meinel wrote:
>>
>>> Is this 0.19 worthy?
>> I haven't read the code so Aaron and Martin are better placed to
>> comment. My leaning would be to vote for inclusion, as long as James is
>> comfortable with that, given we still have enough(?) time for catching
>> weird side effects, if any, prior to 0.19 shipping.
>>
>> I wouldn't leave it any more than 24 hrs though. Low level changes late
>> in the cycle is inherently risky, even with our awesome test suite.
>>
> 
> This does seem pretty risky for a freeze exception. My inclination is to
> defer it until the next release, but I'm not going to stand in the way
> of it.
> 
> Therefore, if you think it should be in this release say so now.
> 
> Thanks,
> 
> James
> 

Well, I would like it to be, because it really does help, and it also
opens up the codebase for my pyrex optimization which helps even more.
(I believe it is 10% for this change, and another 20% for the pyrex, for
a total of 5.2s -> 3.5s for 'bzr checkout bzr.dev test').

However, I can be patient and wait for 0.91.

There will be a bit of conflicts with Robert's changes anyway, since he
was refactoring Knits at the same time I was. Not much of one, though,
and my actual code will actually benefit his current pack format. (It
optimizes how we get information out of a knit fulltext/delta, and his
current pack uses the *same* hunks on disk).

Anyway, I'll just wait until next week.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGu00xJdeBCYSNAAMRAqXZAJ9SiHHnTRgbQTNrDdpqsiXhE2Kr1wCgqoxk
N4ic0EYoOdqAJiU1WVQIQo0=
=50SY
-----END PGP SIGNATURE-----