[IMPALA-3084] Incorrect assignment of NULL checking predicate through an outer join of a nested collection.

• Type: Bug
• Status: Resolved
• Priority: Critical
• Resolution: Fixed
• Affects Version/s: Impala 2.5.0
• Fix Version/s: Impala 2.7.0
• Component/s: Frontend
• Labels: correctness downgraded query_generator
• Target Version: Impala 2.7.0
• Assignee: Alexander Behm
• Reporter: Taras Bobrovytsky
• Created: 25/Feb/16 21:30
• Updated: 22/Mar/18 23:00
• Resolved: 23/Jul/16 04:02

Description

The number of NULLS in this query is 57:

SELECT
a2.pos, count(*)
FROM table_1.field_19 a1
LEFT JOIN a1.value.item.field_22.item.field_36.field_38 a2
group by a2.pos
order by 1, 2

However, when we add a WHERE predicate, this causes the number of NULL to increase to 94:

SELECT
a2.pos, count(*)
FROM table_1.field_19 a1
LEFT JOIN a1.value.item.field_22.item.field_36.field_38 a2
WHERE
(a2.pos) <=> (a2.item)
group by a2.pos
order by 1, 2

To reproduce:

ssh [e-mail] -p 22222 (pw: cloudera)
Impala db name: random_nested_db_0

Issue Links

• is duplicated by: IMPALA-4385 Left join doesn't work properly for array