[MERGE] optimize annotate

[Wouter van Heyst]

On Tue, Nov 28, 2006 at 06:05:01PM -0500, Aaron Bentley wrote:
> Wouter van Heyst wrote:
> > Even though the sizable amount of annotations is listified first.
> > I have trouble seeing when that would become a problem, as so often with
> > these kind of tradeoffs.
> 
> Knits are delta compressed, so each text 'version' must be built from
> older revisions, up to the last fulltext.  If any of the intermediate
> versions would be useful in generating another needed version, it's more
> efficient to generate both versions at once.  So retrieving multiple
> versions of the revision knit at the same time is generally faster than
> retrieving the same revisions in multiple steps.

So what would this do to a 5MB file with 200 versions?

> There's also disk-caching advantages.
> 
> Retrieving multiple revisions at once is the same strategy I used to
> speed up 'log'.  It applies to revision_trees also.
> 
> Plus, my code avoids retrieving the same revision twice, which the old
> code did not do.

Aye, those help to get it faster, it's memory consumption (and cpu-cache
though not all that relevant atm) I'm after.

Wouter van Heyst
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 307 bytes
Desc: Digital signature
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20061129/7abff302/attachment.pgp