Yi Zhang created HIVE-28642:
-------------------------------
Summary: PARTITION_PARAMS value too long due to column stats state
unnecessarily created when hive.stats.column.autogather=false
Key: HIVE-28642
URL: https://issues.apache.org/jira/browse/HIVE-28642
Project: Hive
Issue Type: Bug
Security Level: Public (Viewable by anyone)
Components: Hive
Affects Versions: 3.1.3
Reporter: Yi Zhang
when loadPartition, if it is a new partiton
```
if (this.getConf().getBoolVar(HiveConf.ConfVars.HIVE_STATS_AUTOGATHER)) {
StatsSetupConst.setStatsStateForCreateTable(newTPart.getParameters(),
MetaStoreUtils.getColumnNames(tbl.getCols()), StatsSetupConst.TRUE);
}
```
this creates a query string to metastore
INSERT INTO `PARTITION_PARAMS` (`PARAM_VALUE`,`PART_ID`,`PARAM_KEY`) VALUES
(<'\{"BASIC_STATS":"true","COLUMN_STATS":{"col1":"true","col2":"true",
....}}'>,<593>,<'COLUMN_STATS_ACCURATE'>)
if there are many columns, it can run into PARAM_VALUE too long error.
though in hive-4, HIVE-20221 increased PARAM_VALUE width to blob, for hive-3,
the width is too small.
besides, this seems unnecessary to set the column stats state when
hive.stats.column.autogather=false
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
