[
https://issues.apache.org/jira/browse/LUCENE-7212?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15239494#comment-15239494
]
Karl Wright commented on LUCENE-7212:
-------------------------------------
bq. This is why i asked if we need two-phase. Maybe we need to extend the
benchmark first to understand if its really necessary?
+1 to that. I'm operating in the dark at the moment. I presume in order to do
that we have to have a naive implementation available?
bq. When i look at what geo3d computes per-doc for a distance query, it seems
to look much less expensive (e.g. just some multiplication) and I think this is
also why its faster at distances. For polygons we need to investigate too
(these get slower, linearly, as polygon complexity increases with 2D today).
If geo3d doesn't have these problems, and doesn't really need two-phase, then
we simplify the problem:
geo3d polygon performance will still degrade linearly with the number of sides,
but your assessment is accurate for the cheapness of evaluating any one of
those sides, yes, and of the cheapness of evaluating circle membership.
bq. and what goes in there need not be 100% consistent with the query (ok, it
could cause some confusion if its not, but it wont break things). This means
the encoding could be different, more lossy, etc.
Ok, you've lost me here, but that's my problem. :-)
> Add Geo3DPoint equivalents of LatLonPointDistanceComparator and
> LatLonPointSortField
> ------------------------------------------------------------------------------------
>
> Key: LUCENE-7212
> URL: https://issues.apache.org/jira/browse/LUCENE-7212
> Project: Lucene - Core
> Issue Type: Improvement
> Affects Versions: master
> Reporter: Karl Wright
> Assignee: Karl Wright
>
> Geo3D has a number of distance measurements and a generic way of computing
> interior distance. It would be great to take advantage of that for queries
> that return results ordered by interior distance.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
