[RFC] Faster load of full inventory for development6-rich-root?

Ian Clatworthy ian.clatworthy at internode.on.net
Fri Jun 5 06:25:55 BST 2009

Aaron Bentley wrote:

>> If the latter is sufficient, I can skip the path
>> calculations, hence the children lookups, by using iter_just_entries()
>> and sorting by file-id?
> You have to iterate over the whole inventory anyhow.  It shouldn't be a
> win to skip generating paths.


So send is 3 times quicker in the latest bzr.dev than it was 2 weeks
ago. It now takes 63 seconds instead of 3+ minutes on OOo. Hooray to
whoever fixed that!

Path calculation is *still* a dominant factor though. As an experiment,
the following patch drops the time down to 32 seconds, simply by
avoiding path calculation while generating a testament sha.

I don't know much about testaments and perhaps this sort of change is
semantically invalid, e.g. it will give different testament sha's for
different inventory formats. So, stepping up a layer, a gather the
purpose of the testament.sha is an integrity check that the receiving
repository think revision X really is the same as what the source
repository thinks it is? Is that right? If so, are there other
alternatives to generating a magic number?

Ian C.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: faster-testament-sha.diff
Type: text/x-patch
Size: 1415 bytes
Desc: not available
Url : https://lists.ubuntu.com/archives/bazaar/attachments/20090605/c5e60273/attachment.bin 

More information about the bazaar mailing list