You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@parquet.apache.org by "Tim Armstrong (JIRA)" <ji...@apache.org> on 2017/01/24 21:24:26 UTC
[jira] [Created] (PARQUET-840) Min-max for int96 is computed
incorrectly in parquet-mr
Tim Armstrong created PARQUET-840:
-------------------------------------
Summary: Min-max for int96 is computed incorrectly in parquet-mr
Key: PARQUET-840
URL: https://issues.apache.org/jira/browse/PARQUET-840
Project: Parquet
Issue Type: Bug
Components: parquet-mr
Affects Versions: 1.8.0
Reporter: Tim Armstrong
The min-max stats implementation computes the min and max values of an int96 column using the BinaryStatistics class, which uses the lexical ordering of signed bytes. Other integer types order based on the actual integer value.
The spec does not specify an ordering, but parquet-mr is internally inconsistent - there is no reason why int96 should be ordered in a completely different way from int64.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)