[
https://issues.apache.org/jira/browse/LUCENE-7165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15222390#comment-15222390
]
Robert Muir commented on LUCENE-7165:
-------------------------------------
+1
I think the differences here are responsible for e.g. 4 documents different in
queries during benchmarking etc. It would be great to use 64-bits for geopoint
too. If we are going to change the encoding, maybe it is also worth looking at
using this full 64 bit encoding with a 1D KD tree? I would also consider not
using interleaved form in the docvalues storage: for the way in which docvalues
is used i think de-interleaving is just an unnecessary cost. LatLonPoint just
uses high/low bits and if GeoPointField is compatible, then it could use the
same (also re-use sort comparator!)
> Return GeoPointField Encoding to use full 64 bit space
> -------------------------------------------------------
>
> Key: LUCENE-7165
> URL: https://issues.apache.org/jira/browse/LUCENE-7165
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Nicholas Knize
>
> This was originally done in LUCENE-6170 but then reverted due to issues with
> GeoHash encoding (which has since been removed). This issue will revert the
> GeoPoint encoding back to full 64 bit space which will synchronize
> quantization TOLERANCE between {{LatLonPoint}} and {{GeoPointField}}.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
