Versions Compared

Old Version 2

changes.mady.by.user Julie Tibshirani

Saved on

compared with

New Version Current

changes.mady.by.user Julie Tibshirani

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

  • Merge on full flush is enabled now by default with a timeout of 500ms, giving the merge policy a chance to merge NRT segments together before publishing a new point-in-time view of the IndexReader. This should give queries a small performance boost in the near-realtime case, especially terms-dictionary-intensive queries like fuzzy queries.
  • Add getAllChildren functionality to facets.
  • Added facetsets module for high dimensional (hyper-rectangle) faceting. 

Optimizations

  • Implement Block-Max-Maxscore for 2 clauses disjunctionTop-level two-clause disjunctions sorted by score now use the block-max MAXSCORE algorithm, which introduced a 40%-75% speedup in our benchmarks.
  • BooleanQuery can return quick counts for simple boolean queries.

  • When running KnnVectorQuery with a filter, reuse the cached filter BitSet when possiblebit set.

Other

  • Update randomizedtesting to 2.8.0, hppc to 0.9.1, morfologik to 2.1.9.

...