[
https://issues.apache.org/jira/browse/LUCENE-5200?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13760294#comment-13760294
]
ASF subversion and git services commented on LUCENE-5200:
---------------------------------------------------------
Commit 1520615 from [~rcmuir] in branch 'dev/trunk'
[ https://svn.apache.org/r1520615 ]
LUCENE-5200: HighFreqTerms has confusing behavior with -t option
> HighFreqTerms has confusing behavior with -t option
> ---------------------------------------------------
>
> Key: LUCENE-5200
> URL: https://issues.apache.org/jira/browse/LUCENE-5200
> Project: Lucene - Core
> Issue Type: Bug
> Components: modules/other
> Reporter: Robert Muir
> Attachments: LUCENE-5200.patch
>
>
> {code}
> * <code>HighFreqTerms</code> class extracts the top n most frequent terms
> * (by document frequency) from an existing Lucene index and reports their
> * document frequency.
> * <p>
> * If the -t flag is given, both document frequency and total tf (total
> * number of occurrences) are reported, ordered by descending total tf.
> {code}
> Problem #1:
> Its tricky what happens with -t: if you ask for the top-100 terms, it
> requests the top-100 terms (by docFreq), then resorts the top-N by
> totalTermFreq.
> So its not really the top 100 most frequently occurring terms.
> Problem #2:
> Using the -t option can be confusing and slow: the reported docFreq includes
> deletions, but totalTermFreq does not (it actually walks postings lists if
> there is even one deletion).
> I think this is a relic from 3.x days when lucene did not support this
> statistic. I think we should just always output both TermsEnum.docFreq() and
> TermsEnum.totalTermFreq(), and -t just determines the comparator of the PQ.
--
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]