B-Tree performance analysis

[Robert Collins]

iter_entries_by_prefix can be made lazy for sure - its not used in
general operation anyhow, now.

I agree aiming for 1-pass is important.

Yes, GraphIndex does its own dictionary compression which is why the B
+Tree design is larger if you don't entropy compress the nodes.

The add_node[s] calls have not been tuned, reusing the base class one
was a bootstrap process optimisation, not a performance one - I'm in
complete agreement about doing less work there; I'd be inclined to
implement the method specific to the index class, or move some stuff out
into per-type methods, or something like that.

-Rob
-------------- 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/20080824/df8bf523/attachment.pgp