Adrien Grand created LUCENE-10123:
-------------------------------------
Summary: Avoid double wrapping of doc values at merge time
Key: LUCENE-10123
URL: https://issues.apache.org/jira/browse/LUCENE-10123
Project: Lucene - Core
Issue Type: Improvement
Reporter: Adrien Grand
If you have a SORTED_SET/SORTED_NUMERIC doc values that are effectively
single-valued, the Lucene90 doc-values format internally stores them the same
way that it stores SORTED/NUMERIC doc values, and they get wrapped with a
singleton wrapper ({{DocValues::singleton}}) at read time.
However {{DocValuesConsumer}} doesn't optimize for singletons, which means that
the merging process will create a merged view of the singleton wrappers, and
that Lucene90DocValuesConsumer will then wrap this back to a single-valued view
using {{SortedSetSelector}}.
I think it would be simpler if {{DocValuesConsumer}} detected singleton
wrappers, unwrapped them, created a SORTED/NUMERIC merged view and only then
wrapped it with a singleton wrapper so that {{SortedSetSelector}} would just
unwrap it?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
