Version 8.0.0

Release Notes

1. Improvements of bit-vector compression algorithms (still based on binary interpolative coding). Current generation of compression shows compression ration of 3.64 bits per int on Gov2 sorted benchmark set (previos cersion was 3.66 bits per int). New improvement gives significant reduction in sizes of inverted index storages and improved compression of succinct vectors produced by bit-slicing. There is a minor reduction in speed of compression due to more complicated algorithms, serialization APIs improved to better tune compression settings for either speed or compression.

   New serialization format is backward compatible and based on "protocol evolution" principles. New code will be able to read and deserialize BLOBs from older versions. New serializations will produce new compression format. Old code will not be able to read new BLOBs.

2. Serialization API examples for bit-vectors and sparse vectors improved to illustrate new performance tuning settings.

GitHub