Peter Bengtsson skrev:
From experience I find that BTrees are very fast to write to and pick out items from. Even in the millions. (Never gone into the tens of millions or further) Also, when it comes to browsing stuff I find SQL faster and easier to work with. An added advantage of a RDBMS is that you get the indexing seamlessly built in (no need to bridge zbrain.getObject()) and it makes it easier to optimize and figure out which indexes help and which indexes slow you down which is something that is far from obvious with a ZCatalog approach.
Right. But wouldn't profiling indexes in Zope be as easy as wrapping the index search method in a function that does time.time before and after the search? :-) -Morten -- Morten W. Petersen Manager Nidelven IT Ltd Phone: +47 45 44 00 69 Email: morten@nidelven-it.no