[MERGE] Improve knit extraction time

[Robert Collins]

On Tue, 2007-01-30 at 18:50 -0600, John Arbash Meinel wrote:
> 
> By the way, under --lsprof the bulk of the remaining time is actually
> spent reading the knit .kndx file. (get_inventory_xml takes 23s, but
> of
> that time get_inventory_weave() takes 18.5s, extracting the text
> itself
> by applying 450 patches is only 5.5s). So I think spending some energy
> to improve our .kndx format could certainly be worthwhile. Admittedly
> most .kndx files don't have 215,660 lines. But trees with long
> histories
> will. (Real world testing shows
> branch.repository.get_inventory_weave()
> takes from 7-11s, which is rather long). 

So the kndx file is intended to be incrementally parsable - as long as
you read end-backwards.

The replacement of an entry by a new version complicates this a little,
because if you have two copies of an entry, you want to be sure you've
read the last one.

Rob

-- 
GPG key available at: <http://www.robertcollins.net/keys.txt>.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20070131/f164c56f/attachment.pgp