[
https://issues.apache.org/jira/browse/LUCENE-7123?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Robert Muir updated LUCENE-7123:
--------------------------------
Attachment: LUCENE-7123.patch
Here is a patch:
* adds haversinMeters to SloppyMath and cuts everything over to it.
* adds a test that this is within 0.01MM of the actual haversin result (if you
were to use slower trig functions).
* moves out sin/tan stuff to the geoutils class that is using it. There is a
TODO to remove these further.
I did benchmarking and testing as well. I think its ready.
> deduplicate/cleanup spatial distance
> ------------------------------------
>
> Key: LUCENE-7123
> URL: https://issues.apache.org/jira/browse/LUCENE-7123
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Robert Muir
> Attachments: LUCENE-7123.patch
>
>
> Currently there is a bit of a mess here: SloppyMath.haversin is slightly
> different from GeoDistanceUtils.haversin, the latter is actually slightly
> faster and uses a simple fixed earth diameter (which makes calculations
> easier too).
> But one of these returns meters, the other kilometers. Furthermore
> lucene/spatial now uses some sin/tan functions that were added to sloppymath
> with some accuracy guarantees (which are untested, and not quite correct).
> Lucene/spatial queries also inconsistently mix the two different functions
> together for various purposes and this just causes headaches. Its tests did
> this recently too.
> We need to clean this up, otherwise users will be confused. E.G. they will
> see different results from expressions than from queries and not understand.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
