Daniel Lemire

@richardstartin I am aware of why you wrote it... and to be clear, I think your implementation is quite fast in many cases... in fact, you have hard numbers... but...

@kishoreg Please see the following prototype: [HybridRoaringBitmap.java](https://github.com/RoaringBitmap/RoaringBitmap/blob/pinot/hybrid/roaringbitmap/src/main/java/org/roaringbitmap/buffer/HybridRoaringBitmap.java). It is on a separate branch.

@kishoreg The general idea is that we can already, easily, distinguish between the mapped and unmapped containers. Of course, mixing off-heap and heaped containers means that we need to check...

Do we have a benchmark stressing garbage collection? There are definitively reports that reducing GC pressure helps. I admit that I don’t know how we would make GC the bottleneck.

@blacelle Anything that might help motivate the flyweight pattern is welcome.

See the following commit for an application of this idea to the ``contains`` function: https://github.com/RoaringBitmap/RoaringBitmap/commit/a778e33adb9c651c8c39e60b1eba17a8edc4b0a0

@ppiotrow Yes, we expect to get better performance with ``IntIterator`` than ``Iterator``. It seems that we do too... cd RoaringBitmap cd jmh mvn -f ../pom.xml clean install -DskipTests -Dgpg.skip=true mvn...

@ppiotrow Please revisit https://github.com/RoaringBitmap/RoaringBitmap/blob/master/src/main/java/org/roaringbitmap/PeekableIntIterator.java I just entered more details in the comments as to what this is good for. The idea is to support more advanced operations efficiently.

@ppiotrow I have improved the iterator performance in C this week. In particular, the iteration over run containers wasn't great. It is now much better. I was able to double...

@ppiotrow I can replace failing operation with approach used currently in Java implementation I think that it would be fine. I don't think your goal should be to replace the...