You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@hive.apache.org by "Brad Ruderman (JIRA)" <ji...@apache.org> on 2013/11/08 20:56:17 UTC

[jira] [Created] (HIVE-5785) Hive Metadata Thinks Table Partitions Are All Strings

Brad Ruderman created HIVE-5785:
-----------------------------------

             Summary: Hive Metadata Thinks Table Partitions Are All Strings
                 Key: HIVE-5785
                 URL: https://issues.apache.org/jira/browse/HIVE-5785
             Project: Hive
          Issue Type: Bug
          Components: Database/Schema
            Reporter: Brad Ruderman
            Priority: Minor


hive (bruderman)> CREATE TABLE test (a int, b int) partitioned by (dt int);
OK
Time taken: 0.101 seconds
hive (bruderman)> desc test;
OK
col_name	data_type	comment
a	int
b	int
dt	int
Time taken: 0.093 seconds
hive (bruderman)> CREATE VIEW v_test AS SELECT * FROM test;
OK
a	b	dt
Time taken: 0.042 seconds
hive (bruderman)> desc v_test;
OK
col_name	data_type	comment
a	int
b	int
dt	string
Time taken: 0.098 seconds
hive (bruderman)>
------------------------------------------------------------------------------

When I have a table which is partitioned by an int/bigint, and I go to import that table into Tableau, Tableau detects the partition column as being a string thus I cannot use it for incremental refreshes. I thought it was a tableau bug, however when creating a view: select * from table, then describing the view, I see that the partition column is a string, thus I think the issue is within hive.

Finally the issue extends when interfacing through hive server 1/hive server 2:

(hs2)➜  pyhs2 git:(master) python test.py
None
[{'comment': None, 'columnName': 'a', 'type': 'INT_TYPE'}, {'comment': None, 'columnName': 'b', 'type': 'INT_TYPE'}, {'comment': None, 'columnName': 'dt', 'type': 'STRING_TYPE'}]

Where the column is detected a string. 

The workaround is to create a view:

CREATE VIEW v_test AS 
SELECT t.*, CAST(t.dt as INT) dt_part
FROM test

And using that. This issue extended beyond Tableau and affects anything using the HiveServer1/2.

Thanks!



--
This message was sent by Atlassian JIRA
(v6.1#6144)