lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "David Smiley (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (LUCENE-4916) Spatial WithinPrefixTreeFilter can too eagerly find false-positive
Date Mon, 08 Apr 2013 15:15:16 GMT

     [ https://issues.apache.org/jira/browse/LUCENE-4916?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

David Smiley updated LUCENE-4916:
---------------------------------

    Attachment: LUCENE-4916__Spatial_Within_RPT_and_shape_simplification.patch

The attached patch fixes this issue, and includes a testcase.  The approach has room for further
optimization.

Although unrelated, I found it possible for AbstractVisitingPrefixTreeFilter to have a minimum
configurable scan level of 0 (formerly 1).  I made this change too; it enabled the aforementioned
test to be a little more thorough. 
                
> Spatial WithinPrefixTreeFilter can too eagerly find false-positive 
> -------------------------------------------------------------------
>
>                 Key: LUCENE-4916
>                 URL: https://issues.apache.org/jira/browse/LUCENE-4916
>             Project: Lucene - Core
>          Issue Type: Bug
>          Components: modules/spatial
>            Reporter: David Smiley
>            Assignee: David Smiley
>            Priority: Minor
>         Attachments: LUCENE-4916__Spatial_Within_RPT_and_shape_simplification.patch
>
>
> The recently committed WithinPrefixTreeFilter can sometimes match a false-positive due
to the grid approximation technique.  Some of this is normal.  But if an indexed shape has
any of its leaf detail levels "simplified" (removed) up to a higher level, then a false-positive
match can occur that is a little too eager when one of those cells that were simplified-away
don't intersect the query shape, and thus shouldn't be considered a match for this predicate.

> Marking as "Minor" because it should be quite rare, and when it does occur, the answer
(a match) will usually appear close-enough due to the grid approximation that it's not a totally
inaccurate answer.  I found this problem via extensive randomized testing.  I want to fix
it because it should make my new tests pass :-)

--
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: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message