Hive：当插入分区表时，在大多数行中，hive double url-encode分区键列

Question

I created a partitioned table:

create table  t1 ( amount double) partitioned by ( events_partition_key string) stored as paquet;

added some data to tmp_table, where column 'events_partition_key' contains timestamp (string type) in the following format: "2018-02-25 00:00:00"

Then I insert some data into the partitioned table.

insert into table t1 partition (events_partition_key)
select amount, events_partition_key
from tmp_table

when selecting from the new partitioned table t1, in some cases events_partition_key column is presented regularly the same it appeared in tmp_table ie "2018-02-25 00:00:00", but in most cases, it appears URL-encoded, ie "2018-02-25 00%3A00%3A00"

In any case, the original column partition_key has no difference in cases where it was URL-encoded or not,

when showing partitions for the new table:

show partitions t1;

I got everything URL-encoded twice (ie "2018-02-25 00%253A00%253A00"), but some cases it was only once (ie "2018-02-25 00%3A00%3A00")

there is nothing wrong in the original values, only after it has become a partition key.

Answer 1

This Jira HIVE-3679 adds hive.decode.partition.name property to the HiveConf.java in release 0.10.0

Try to set it true.

set hive.decode.partition.name=true;