[
https://issues.apache.org/jira/browse/HIVE-7177?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Tom Temple updated HIVE-7177:
-----------------------------
Environment: Redhat 5.10 running Cloudera 5.0.1 (was: Redhat 5.10 running
Cloudera 5.0.0)
> percentile_approx very inaccurate with high multiplicities in the data
> ----------------------------------------------------------------------
>
> Key: HIVE-7177
> URL: https://issues.apache.org/jira/browse/HIVE-7177
> Project: Hive
> Issue Type: Bug
> Components: UDF
> Affects Versions: 0.12.0
> Environment: Redhat 5.10 running Cloudera 5.0.1
> Reporter: Tom Temple
>
> To reproduce:
> 1) create a table with a single integer column
> 2) with values: 1 million, 2 million, 3 million, and 4 million each repeated
> a quarter million times.
> 3) percentile_approx(cast(col_0 as double), array(0.33,0.34),1000000)
> Expected results: [2000000.0,2000000.0]
> Actual results: [1280000.0,1320000.0] (I might be off by 40000 here)
--
This message was sent by Atlassian JIRA
(v6.2#6252)
