[
https://issues.apache.org/jira/browse/SOLR-14185?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17265184#comment-17265184
]
David Smiley commented on SOLR-14185:
-------------------------------------
Now we have only two DocSet implementations (on master).
Proposed method on DocSet: {{DocIdSetIterator iterator(LeafReaderContext ctx)}}.
The implementation of that on BitDocSet is obvious; SortedIntDocSet is harder.
See SortedIntDocSet.getTopFilter.getDocIdSet which looks up the range of
indexes based on the ID range of the LeafReaderContext. We could basically
copy that logic to this new iterator method in order to get a DocIdSetIterator
of the corresponding slice. Straight-forward I think? But notice the
getTopFilter tracks "lastEndIdx" which we can't carry forward, as there is no
intermediary object to hold it. However, I imagine an optimization when the
SortedIntDocSet is built, in which the segment boundaries could be tracked _at
that time_ such that when iterator() is called, it merely looks up the
boundaries from an additional small array of leaf ordinal -> index into sorted
docIDs.
> add DocSet.getDocIdSetIterator(LeafReaderContext)
> -------------------------------------------------
>
> Key: SOLR-14185
> URL: https://issues.apache.org/jira/browse/SOLR-14185
> Project: Solr
> Issue Type: Sub-task
> Reporter: David Smiley
> Priority: Major
>
> Many callers of {{DocSet.getTopFilter()}} really just want to call
> {{getDocIdSet}} on the Filter and then call {{iterator()}}. The Bits 2nd arg
> is also always null or always live-docs so this arg doesn't matter since a
> Solr DocSet never contains deleted docs. The goal here is to reduce needless
> dependencies on the old Filter.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]