New index format

[John Arbash Meinel]

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

Aaron Bentley wrote:
> James Westby wrote:
>> Hi,
> 
>> I might be wrong here, but it seems like the commands that the
>> Emacs thread is discussing all have a bottleneck in the index
>> format.
> 
> The problem people are discussing is due to the in-memory bisect code,
> not the on-disk format.  It's quite likely that the on-disk format could
> be improved, but it's not the bottleneck at the moment.
> 
> Aaron

Well... some of it is just because of problems with using whole-history
operations.

For this specific case, the "buffer_all" patch I proposed a while ago
would help a lot. But it is just because they have 1 huge project with a
completely linear history.

For 'log' we could just punt and introduce a "flatlog" that wouldn't do
revision numbers and would do simple topo-sorted output. Which you can
get by just walking the graph (no sorting or post processing, etc.)

I *would* like to see a quick fix for fast-to-get lookups for "-1" etc,
since we already have the mechanism in place. I'll actually try putting
something together for it.

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

iD8DBQFH2p8YJdeBCYSNAAMRAsWeAKCrIwJ89AQhOg+IIOJPrpUHRN4yjACgmCx+
gABgIE8FNahCPf08Si62YS8=
=Re02
-----END PGP SIGNATURE-----