Le 3/29/14 10:33 AM, Emmanuel Lécharny a écrit : > Hi guys, > > I just fixed a bug in the way the duplicate values were handled in the > persisted btrees : we weren't storing the new header into the value > after an insertion, so we wre not capable to get back any of the values. > > Having fixed this bug, all the tests are now passing green, but one (a > test where we construct a btree using a bulk load).
I was wrong, there was some bugs in the sub-btree handling. We weren't able to reaload them when the system was restarted. This is now fixed. > > Now, i'm not really pleased by the performance of the system when we > have sub-btrees. It works, but it's really slow. I have to investigate > this problem, and wil do that tomorrow in the plane. The performances are now way better, I removed the dump of free pages which were being done everytime we were adding some element in a btree... More to come soon. -- Regards, Cordialement, Emmanuel Lécharny www.iktek.com
