You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Kurt Young (Jira)" <ji...@apache.org> on 2019/12/03 02:50:00 UTC
[jira] [Updated] (FLINK-14663) Distinguish unknown column stats and
zero
[ https://issues.apache.org/jira/browse/FLINK-14663?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Kurt Young updated FLINK-14663:
-------------------------------
Issue Type: Bug (was: Improvement)
> Distinguish unknown column stats and zero
> -----------------------------------------
>
> Key: FLINK-14663
> URL: https://issues.apache.org/jira/browse/FLINK-14663
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Hive, Table SQL / API
> Reporter: Kurt Young
> Priority: Major
>
> When converting from hive stats to flink's column stats, we didn't check whether some columns stats is really set or just an initial value. For example:
> {code:java}
> // code placeholder
> LongColumnStatsData longColStats = stats.getLongStats();
> return new CatalogColumnStatisticsDataLong(
> longColStats.getLowValue(),
> longColStats.getHighValue(),
> longColStats.getNumDVs(),
> longColStats.getNumNulls());
> {code}
> Hive `LongColumnStatsData` actually has information whether some stats is set through APIs like `isSetNumDVs()`. And the initial values are all 0, it will confuse us is it really 0 or just an initial value.
>
> We can use -1 to represent UNKNOWN value for column stats.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)