[
https://issues.apache.org/jira/browse/LUCENE-5028?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13671942#comment-13671942
]
Han Jiang commented on LUCENE-5028:
-----------------------------------
hmm, maybe not? I suppose, for TwoLong case, when the first long is 'moved'
towards root and minus to zero, the second one might still be greater than
zero? So, when output is shared, we might create an extra output.
> doShare is pointless in PositiveIntOutputs
> ------------------------------------------
>
> Key: LUCENE-5028
> URL: https://issues.apache.org/jira/browse/LUCENE-5028
> Project: Lucene - Core
> Issue Type: Improvement
> Components: core/FSTs
> Reporter: Han Jiang
> Assignee: Han Jiang
> Priority: Minor
> Fix For: 5.0
>
> Attachments: LUCENE-5028.patch
>
>
> We recently use this in oal.core.fst.PositiveIntOutputs to indicate whether
> to share outputs. The comment mentioned 'with doShare=false, in some case
> this may result in a smaller FST'. However, this is not intuitive, as for
> long type, we always have the smallest output reduced to NO_OUTPUT, thus the
> smallest one is 'moved' towards root, and no extra output is created.
> However, if there are many many small outputs around root arcs, when we share
> outputs, a large output might be pushed into the root arcs. When root arcs
> are packed as fixed-array, yes the size of FST is increased. But, I suppose
> this should invoke other intuitive heuristics, instead of the confusing
> 'doShare'?
> Besides, this only exist in PositiveIntOutputs.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
