*** 143,148 **** --- 143,150 ---- This is only of moderate value, because we'll only save a maximum of a byte per term. But it's a free byte, and not too complicated at all to implement. Additionally, awhile back I tested delta-encoding for interior nodes, and for most nodes it *increased* the space needed, because adjacent terms tended to have no shared prefix. It did save about 2% for the 64kdoc segment's height-1 interior nodes. This encoding might counter that. + *UPDATE:* A reasonable encoding is nSuffix*iScale+nPrefix when the result is <128,and nSuffix>0, nSuffix*128+nPrefix when nPrefix<128, and 0, nPrefix, nSuffix for other cases. Unfortunately, measurements show the results to be mixed. Very slight reduction to index size (.4% to 1%). More interesting is that a certain number of the segment trees end up one level shorter, which I'm not sure how to quantify. Having a one-level-shorter btree 1% of the time is probably much more interesting than a 1% gain in storage size. + ---- *store interior nodes inline*
Small. Fast. Reliable.
Choose any three.
Choose any three.
|
|