[
https://issues.apache.org/jira/browse/LUCENE-6256?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ryan Ernst updated LUCENE-6256:
-------------------------------
Attachment: LUCENE-6256.patch
I was reminded that {{nextPosition()}} should be called {{freq()}} times only.
Having a {{NO_MORE_POSITIONS}} constant doesn't make sense. The case that
should return -1 is only for when positions do not exist.
This patch removes asserts in the places that should just return -1, and
updates the javadocs for {{nextPositions()}}. All tests pass.
> PostingsEnum impls should respect documented behavior when no positions exist
> -----------------------------------------------------------------------------
>
> Key: LUCENE-6256
> URL: https://issues.apache.org/jira/browse/LUCENE-6256
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Ryan Ernst
> Attachments: LUCENE-6256.patch
>
>
> {{PostingsEnum.nextPositions}} says that if no positions exist,
> NO_MORE_POSITIONS will be returned on the first call (actually it refers to
> DocsEnum.NO_MORE_DOCS, which should be changed since DocsEnum doesn't exist
> on trunk). At least one impl (BlockDocsEnum) does {{assert false}} inside
> {{nextPosition()}}. This means if you have assertions turned on (e.g. in a
> test) you get an assertion here, when the behavior should return
> NO_MORE_POSITIONS. I'm still going through all the implementations, but I
> also see {{MultiTermHighlighting}} which returns {{Integer.MAX_VALUE}}. I
> think we should clean up all these implementations which have no positions
> (including maybe the fake scorers that are copied around in a lot of places?)
> to return NO_MORE_POSITIONS.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
