[
https://issues.apache.org/jira/browse/LUCENE-7872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16052115#comment-16052115
]
Hoss Man commented on LUCENE-7872:
----------------------------------
bq. Doesn't DocList only represent results from a single shard?
yes, but given how closely used it is with TopDocs, and since the end user APIs
are allready (supposed to be) "long" based, it seemed like a good idea to
follow it through -- and in doing so let to the problems/fixes in places like
SpellCheckCollation where "hits" was an int even though it was used to model
"totalHits" (for a spellcheck collation) across shards.
(which is the type of problem i was suspicious/worried about finding in usages
of DocList when i saw your first patch ... based on the changes you had to make
to TopGroupsResultTransformer)
ie: maybe it isn't strictly necessary to change DocList.matches() to be long,
but:
# doing so helped find problems with existing code making assumptions about
"total hits" (across shards) and should hopefully prevent similar problems in
future code as well
# now seems like as good a time as any to make this change given how closely
connected solr's usage of DocList/DocSlice is with TopDocs.
> TopDocs.totalHits should be a long
> ----------------------------------
>
> Key: LUCENE-7872
> URL: https://issues.apache.org/jira/browse/LUCENE-7872
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Adrien Grand
> Priority: Minor
> Attachments: LUCENE-7872.patch, LUCENE-7872.patch
>
>
> Even though a single index cannot have more than 2B documents, TopDocs.merge
> might merge TopDocs instances that have more than 2B documents in total.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
