[jira] [Created] (HIVE-19484) 'IN' & '=' do not behave the same way for Date/Timestamp comparison.

["Venu Yanamandra (JIRA)" <ji...@apache.org>]

Venu Yanamandra created HIVE-19484:
--------------------------------------

             Summary: 'IN' & '=' do not behave the same way for Date/Timestamp comparison.
                 Key: HIVE-19484
                 URL: https://issues.apache.org/jira/browse/HIVE-19484
             Project: Hive
          Issue Type: Bug
            Reporter: Venu Yanamandra


We find that there is a difference in the way '=' operator and 'IN' behave when operating on timestamps.

The issue could be demonstrated using below -
   i) create table test_table (test_date timestamp);
  ii) insert into test_table values('2018-01-01');
 iii) select * from test_table where test_date='2018-01-01'; -- Works
 iv) select * from test_table where test_date in ('2018-01-01'); -- Fails with error [1]
  v) However, casting works - 
       select * from test_table where test_date in (cast ('2018-01-01' as timestamp));

As per url [2], we find no references to limitations when '=' or 'IN' are used.

As per the url [3], we find that there are implicit type conversions defined. However, '=' operates in a different way than the 'IN' operator.

We would like to see if 'IN' could be made to behave the same way as '='.


[1]:
 Caused by: org.apache.hadoop.hive.ql.exec.UDFArgumentException: The arguments for IN should be the same type! Types are: {timestamp IN (string)}

[2]:
 https://cwiki.apache.org/confluence/display/Hive/LanguageManual+UDF#LanguageManualUDF-LogicalOperators

[3]:
 https://cwiki.apache.org/confluence/display/Hive/LanguageManual+Types#LanguageManualTypes-AllowedImplicitConversions




--
This message was sent by Atlassian JIRA
(v7.6.3#76005)