[ https://issues.apache.org/jira/browse/IGNITE-7482?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Zbyszek B updated IGNITE-7482: ------------------------------ Description: In org.apache.ignite.internal.processors.query.h2.opt.GridLuceneIndex#query in line 285 one executes query: docs = searcher.search(query, Integer.MAX_VALUE) what causes all data to be fetched and deserialised regardless the size of the cursor. TopDocs searchAfter(final ScoreDoc after, Query query, int numHits) is not used to implement cursor advancing. This causes the issues with 10 mln. entities when (depending on term) execution takes way above 7 sec. to retrieve first row, hence making this impossible to use TextQuery to implement auto-complete. See [http://apache-ignite-users.70518.x6.nabble.com/Cursor-in-TextQuery-first-hasNex-is-slow-td19265.html] for more details. was: In org.apache.ignite.internal.processors.query.h2.opt.GridLuceneIndex#query in line 285 one executes query docs = searcher.search(query, Integer.MAX_VALUE) what causes all data to be fetched and deserialised regardless the size of the cursor. TopDocs searchAfter(final ScoreDoc after, Query query, int numHits) is not used to implement cursor advancing. This causes the issues with 10 mln. entities when (depending on term) execution takes way above 7 sec. to retrieve first row, hence making this impossible to use TextQuery to implement auto-complete. See [http://apache-ignite-users.70518.x6.nabble.com/Cursor-in-TextQuery-first-hasNex-is-slow-td19265.html] for more details. > Cursor in TextQuery fetches all data in first call to next() or hasNext() > ------------------------------------------------------------------------- > > Key: IGNITE-7482 > URL: https://issues.apache.org/jira/browse/IGNITE-7482 > Project: Ignite > Issue Type: Improvement > Components: general > Affects Versions: 2.3 > Reporter: Zbyszek B > Priority: Major > > In org.apache.ignite.internal.processors.query.h2.opt.GridLuceneIndex#query > in line 285 one executes query: > docs = searcher.search(query, Integer.MAX_VALUE) what causes all data to be > fetched and deserialised regardless the size of the cursor. > TopDocs searchAfter(final ScoreDoc after, Query query, int numHits) is not > used to implement cursor advancing. This causes the issues with 10 mln. > entities when (depending > on term) execution takes way above 7 sec. to retrieve first row, hence > making this impossible to use TextQuery to implement auto-complete. > See > [http://apache-ignite-users.70518.x6.nabble.com/Cursor-in-TextQuery-first-hasNex-is-slow-td19265.html] > for more details. -- This message was sent by Atlassian JIRA (v7.6.3#76005)