lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nicholas Knize (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (LUCENE-7179) GeoPoint and LatLonPoint test data should quantize once
Date Tue, 05 Apr 2016 18:35:25 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-7179?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15226868#comment-15226868
] 

Nicholas Knize commented on LUCENE-7179:
----------------------------------------

bq. that test is relatively worthless because it does not test, and will never test the values
that are buggy

I'm going to remove it if its worthless. LatLonPoint version should be the standard if the
TestEncodingUtils version omits important cases.

bq. This is unrelated to what I am talking about. I am only talking about the truncation of
the data provided by the user.

hmm. I miss the distinction then. The user provides lat lon values as 64 bit double precision.
For {{GeoPointField}} these floats are binned into unsigned 32 bit integer space so that all
values will be "sorted" (along the single dimension integer number line) correctly. Because
its a 32 bit space, though, data truncation is inevitable. "close" values will be binned together.
Similarly, {{LatLonPoint}} encodes these in signed 32 bit space and {{NumericUtils}} is left
to handle the signed bit for comparisons. But, again, because its 32 bit space truncation
is inevitable (e.g., 180.0 has to be stepped down). So in both cases there's data loss. We
just handle them differently. 

bq.  I think I cannot compromise on this!

The current patch isn't asking for a compromise. Its up for discussion and investigation for
everyone to have a look. There's likely a simple solution but I'm working a few things in
tandem. So if something comes to someone else's mind real quick I'm super happy to have the
collaboration. I'll get back to this one in a sec and check the LatLonPoint failing test case.

> GeoPoint and LatLonPoint test data should quantize once
> -------------------------------------------------------
>
>                 Key: LUCENE-7179
>                 URL: https://issues.apache.org/jira/browse/LUCENE-7179
>             Project: Lucene - Core
>          Issue Type: Bug
>            Reporter: Nicholas Knize
>         Attachments: LUCENE-7179.patch
>
>
> {{LatLonPoint}} and {{GeoPointField}} tests pre quantizes test data to ensure consistency
with indexed (encoded) data. The pre quantized data then becomes indexed, undergoing another
quantization. To guarantee numerical stability this should be changed such that the test data
is quantized after indexing.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

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


Mime
View raw message