On Oct 10, 2006, at 9:02 PM, Mike Orr wrote:

> What's surprising is that using
> Persistent objects in the BTree doubles the size of the database but
> does not affect performance.

Each pickle in a Durus database is, by default, compressed.
That's why Durus database is, in either case, much smaller
than the plain pickle.  If you compress the giant pickle file,
you can probably get it down to the 10MB range or lower.

I think you get better compression in the case of the BTree
of non-persisistent values because the pickles are bigger.
Compress has a better chance of removing more redundancy.

I think your BTree must not be loaded when you do
the searches for which you report times.  When it is fully loaded,
I'd expect the time to be much closer to the time for searching
the plain pickle version.