[
https://issues.apache.org/jira/browse/LUCENE-6699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14699543#comment-14699543
]
Michael McCandless commented on LUCENE-6699:
--------------------------------------------
Thanks [~daddywri], I'll commit shortly.
I ran the same "225 bboxes around London on the 61M OpenStreetMap
lat/lon points" perf tests with the current branch, plus the 3 other
existing options:
* Geo3D + BKD (PointInGeo3DShapeQuery, this issue/branch): index is
936 MB, took 423 + 96 (close, waiting for merges) sec to build.
7.01 sec to run 225 queries, total hits 221,118,860.
* 2D BKD (BKDPointInBBoxQuery): index is 704 MB, took 163 + 97 sec
to build. 2.32 sec to run 225 queries, total hits 221,118,844.
It's OK (I think) that number of hits is a wee bit smaller than
geo3d because the geo3d bbox when projected to 2D space is a bit
bowed out (I think?) and can therefore include a few more
points.
* Postings + DV (GeoPointInBBoxQuery): index is 3.4 GB, took 349 +
11 sec to build. 3.60 sec to run 225 queries, total 221,120,393
hits (hmm: I'm not sure why this is not identical to 2D BKD?)
* PackedQuadPrefixTree (RecursivePrefixTreeStrategy), with maxLevels=25:
990 + 11 seconds to build, 7.7 GB, 3.82 seconds to run 225 queries, total
221,123,206 hits.
All these benchmarks are in luceneutil.
> Integrate lat/lon BKD and spatial3d
> -----------------------------------
>
> Key: LUCENE-6699
> URL: https://issues.apache.org/jira/browse/LUCENE-6699
> Project: Lucene - Core
> Issue Type: New Feature
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Attachments: Geo3DPacking.java, LUCENE-6699.patch, LUCENE-6699.patch,
> LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch,
> LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch, LUCENE-6699.patch
>
>
> I'm opening this for discussion, because I'm not yet sure how to do
> this integration, because of my ignorance about spatial in general and
> spatial3d in particular :)
> Our BKD tree impl is very fast at doing lat/lon shape intersection
> (bbox, polygon, soon distance: LUCENE-6698) against previously indexed
> points.
> I think to integrate with spatial3d, we would first need to record
> lat/lon/z into doc values. Somewhere I saw discussion about how we
> could stuff all 3 into a single long value with acceptable precision
> loss? Or, we could use BinaryDocValues? We need all 3 dims available
> to do the fast per-hit query time filtering.
> But, second: what do we index into the BKD tree? Can we "just" index
> earth surface lat/lon, and then at query time is spatial3d able to
> give me an enclosing "surface lat/lon" bbox for a 3d shape? Or
> ... must we index all 3 dimensions into the BKD tree (seems like this
> could be somewhat wasteful)?
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
