[IMPALA-2966] Incorrect results - Analytic functions

• Type: Bug
• Status: Resolved
• Priority: Blocker
• Resolution: Not A Bug
• Affects Version/s: Impala 2.5.0
• Component/s: Frontend
• Labels: correctness query_generator
• Environment: DB: functional
• Assignee: Alexander Behm
• Reporter: Taras Bobrovytsky
• Created: 08/Feb/16 22:59
• Updated: 09/Feb/16 01:10
• Resolved: 09/Feb/16 01:10

Description

The following query returns 1000 rows with no NULLS.

Query:

SELECT
  COALESCE(
    LEAD('251', 70) OVER (ORDER BY a3.smallint_col DESC, a3.int_col DESC), 
    LEAD('168', 64) OVER (ORDER BY a3.smallint_col DESC, a4.double_col ASC)
  ) AS char_col
FROM 
  alltypes a3
  INNER JOIN alltypesagg a4 ON ((a4.date_string_col) = (a3.date_string_col)) AND ((a4.day) = (a3.bigint_col))

However the following 2 queries, which are subsets return NULLS.

SELECT
  LEAD('251', 70) OVER (ORDER BY a3.smallint_col DESC, a3.int_col DESC)
FROM 
  alltypes a3
  INNER JOIN alltypesagg a4 ON ((a4.date_string_col) = (a3.date_string_col)) AND ((a4.day) = (a3.bigint_col))

SELECT
  LEAD('168', 64) OVER (ORDER BY a3.smallint_col DESC, a4.double_col ASC)
FROM 
  alltypes a3
  INNER JOIN alltypesagg a4 ON ((a4.date_string_col) = (a3.date_string_col)) AND ((a4.day) = (a3.bigint_col))

Also when the original query is executed in Postgres, some NULLS are returned.