Approaching Join Index for Lucene #bbuzz
summary
Compares three concrete Lucene approaches and shows a working prototype: a join index prototype built with updatable doc values that stores Lucene docIDs to avoid expensive external-id lookups. Benchmarks add the built-in block-join (fastest, often by orders of magnitude) and a global ordinal join (query-time but compact and with nearly zero update cost). Result: join index speeds lookups vs plain query-time joins and enables both join directions, but current prototype forces full-column rewrites so incremental updates do not yet work.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 2:36Advantage Section
- 3:13Plan
- 4:52Explanation
- 8:57Summary
- 10:38Query Time Join
- 12:33Block Join
- 14:06Query Time
- 14:40leapfrog
- 16:50index
- 20:19finding the algorithm
- 22:09Joint operations
- 23:54Global original query time join
- 26:17Benchmarks
related talks
43mTo find a needle in a haystack
50mID, EGO and Super EGO of your entity
29mThe maintainability of open source code 20 years from today
01h 16mHow do to do research with Systems Researchers
01h 09mThere's Treasure Everywhere
24mExtending OpenSearch: A Practical Guide To Building Custom Plugins
56mKeynote: Benchee: 9 Years of Benchmarking on the BEAM
58mMultidimensional Index Sets for Data Locality in HPC Applications