| commit | ef650465a87f452b01546ff3423062c51ed936e9 | [log] [tgz] |
|---|---|---|
| author | Kaushik Lingarkar <kaushik.lingarkar@linaro.org> | Wed Aug 17 13:57:27 2022 -0700 |
| committer | Kaushik Lingarkar <kaushik.lingarkar@linaro.org> | Tue Sep 13 16:18:13 2022 -0700 |
| tree | aeb40d9b537286382ee6395ef41e7a579db3a512 | |
| parent | 19f7e67c86a8eca1dcca72c76ab7076451dd649c [diff] |
Introduce a SEARCH_AFTER index pagination type Current Paginated interface only allows to restart a query from a given offset. Update it to also allow restarting queries using a searchAfter key. searchAfter can help with performance of queries such as [1] where the visibility is being queried for an account that cannot see most/all changes. [1] finishes in ~45 mins with OFFSET and ~10 mins with SEARCH_AFTER when the site has around 1M abandoned changes. Index config has been updated to add a paginationType entry with OFFSET and SEARCH_AFTER as supported values. The default is set to OFFSET to keep the current pagination type unchanged. Gerrit's Lucene index implementation has been updated to support this new pagination type with search-after[2]. Elasticsearch index implementation has also been updated to paginate using search-after[3] and can be further improved to use PIT[4]. Lucene and Elasticsearch tests have been updated to run with both pagination types. Also, Elasticsearch tests now use the official docker image as it can handle running tests with multiple index config suites. Note that, searchAfter will not impact using offsets in Gerrit query APIs. [1] gerrit query 'status:abandoned visibleto:guest' [2] https://lucene.apache.org/core/6_6_5/core/org/apache/lucene/search/IndexSearcher.html#search-org.apache.lucene.search.Query-int-org.apache.lucene.search.Sort-boolean-boolean- [3] https://www.elastic.co/guide/en/elasticsearch/reference/current/paginate-search-results.html#search-after [4] https://www.elastic.co/guide/en/elasticsearch/reference/current/point-in-time-api.html Release-Notes: Index searches now support search-after pagination Change-Id: If3f8d914d5fd3f350c026cf099f08cf9a13f2195