Adrien Grand created LUCENE-10089:
-------------------------------------
Summary: Add a way to disable the sort optimizations to leverage
points on numeric fields
Key: LUCENE-10089
URL: https://issues.apache.org/jira/browse/LUCENE-10089
Project: Lucene - Core
Issue Type: Improvement
Reporter: Adrien Grand
[~jimczi] [~romseygeek] [~mayya] and I were just looking at a case when it
would be useful to be able to disable the sort optimization.
Say you have a 8.x index that has a field that is indexed with IntPoint and
NumericDocValuesField. This field is used for index sorting, and the SortField
is created with SortField.Type.LONG.
This was accepted in 8.x, but this is something that Lucene 9 would complain
about: since the field is an integer, it should use SortField.Type.Int, not
SortField.Type.LONG. If the field was not used for index sorting, then you
could just switch to SortField.Type.INT and everything would work fine.
However since the field is used for index sorting, if you switch to
SortField.Type.INT, then index-time SortField objects are going to compare
differently from search-time SortField objects, which in-turn will disable
early termination of queries in TopFieldCollector.
To be able to migrate from indices that fall in this scenario, it would be
helpful to have an option that would disable the optimization to use points as
well as the associated validation logic, similarly to what we do in 8.x to
enable this optimization.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
