[
https://issues.apache.org/jira/browse/LUCENE-7871?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Mikhail Khludnev updated LUCENE-7871:
-------------------------------------
Attachment: LUCENE-7871.patch
bq. hasValue and seen seem t..
Ok. Thanks. I've collapsed them.
I did is non backward compatible due to child Query. But turning child Query to
DISI turned out soo hard. I had to reproduce
{{ValueSource.ValueSourceSortField}} trick with weight and context map. But now
{{ToParentBlockJoinSortField}} should be rewriten before searching. I find it
not really convenient, but looks like it's what ValueSourceSortField users live
with, see {{SolrIndexSearcher.weightSort(Sort)}} (I know), and
{{TestFunctionQuerySort}} as well. I wonder if we can do this simpler?
bq. Finally I'm still not a fan of the
Thankfully it doesn't sound like veto. Does it? I renamed it to the package
level {{ToParentDocValues}} and pack both twins (sorted and numerics) into it.
So, we can think that internally this code is duplicated.
I propose this OO-hairish stuff because the current duplicated code introduced
the bug, and I'm afraid it's caused exactly by this duplication.
> polish BlockJoinSelector
> -------------------------
>
> Key: LUCENE-7871
> URL: https://issues.apache.org/jira/browse/LUCENE-7871
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Mikhail Khludnev
> Attachments: LUCENE-7871.patch, LUCENE-7871.patch, LUCENE-7871.patch,
> LUCENE-7871.patch
>
>
> * fix false positive match for SortedSetDV
> * make {{children}} iterator instead of bitset.
> see [the
> comment|https://issues.apache.org/jira/browse/LUCENE-7407?focusedCommentId=16042640&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16042640]
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
