db4o Developer Community

I have been working on extending the query API to enable reusing them  and cache results for a set of constraints. I already have about 10% improvement simply by reusing the queries. On any query that is ran multiple times the improvements can be massive however, these results won't be visible with the existing benchmarks. Could this still qualify for the contest?

An example of what I am trying to achieve is

String randVal = "child-" + randomPosition();
if (_randomValueQuery == null) {
	_randomValueQuery = newQuery().cached();
	_randomValueQuery.onCache().constrain(DBTree.class);
	_randomValueQuery.onCache().descend("_root").constrain(root1);
} else {
	_randomValueQuery.init();
}
_randomValueQuery.descend("_node").constrain(randVal);
Iterator it = _randomValueQuery.execute().iterator();

ErikNRC:

I have been working on extending the query API to enable reusing them  and cache results for a set of constraints. I already have about 10% improvement simply by reusing the queries. On any query that is ran multiple times the improvements can be massive however, these results won't be visible with the existing benchmarks. Could this still qualify for the contest?

 

Of course this would qualify!

Feel free to write a dedicated Poleposition circuit that measures rerunning the same query multiple times.

Carl Rosenberger:

Of course this would qualify!

Feel free to write a dedicated Poleposition circuit that measures rerunning the same query multiple times.

Ok will do! I finally got something working and with 10000 objects I have 45% improvement with the example above. Unfortunately, this won't be a pretty patch, I had to modify most of the Query classes and the cache itself is an object in the transaction; although I haven't changed the execution path without cache.

Erik.