[
https://issues.apache.org/jira/browse/IMPALA-9080?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Csaba Ringhofer updated IMPALA-9080:
------------------------------------
Description:
Most tests ignore the order of the expected / returned results, but in some
cases we expect them to be in the same order (decided by
https://github.com/apache/impala/blob/e05a5323785ecb09e45bdb5dfc96533e68256175/tests/common/test_result_verifier.py#L463
)
The logic above simply checks for the presence of ORDER BY in the query, while
the results should be in a deterministic order only if there is an ORDER BY in
the outermost query.
The following query caused a test failure while the rows returned were correct,
just in different order:
{code}
select leftSide.id, rnk
from complextypestbl leftSide
left outer join (
select t.id, rank() over (order by t.id) as rnk
from complextypestbl t left outer join t.int_array
) rightSide
on leftSide.id = rightSide.id;
{code}
The ORDER BY is in an analytical function, which sorts the results, but those
results are joined with a table, and the order of rows returned by JOIN can be
non-deterministic.
was:
Most tests ignore the order of the expected / returned results, but in some
cases we expect them to be in the same order (decided by
https://github.com/apache/impala/blob/e05a5323785ecb09e45bdb5dfc96533e68256175/tests/common/test_result_verifier.py#L463
)
The logic above simply checks for the presence of ORDER BY in the query, while
the results should be in a deterministic order only if there is an ORDER BY in
the outermost query.
The following query caused test a failure while the rows returned were correct,
just in different order:
{code}
select leftSide.id, rnk
from complextypestbl leftSide
left outer join (
select t.id, rank() over (order by t.id) as rnk
from complextypestbl t left outer join t.int_array
) rightSide
on leftSide.id = rightSide.id;
{code}
The ORDER BY is in an analytical function, which sorts the results, but those
results are joined with a table, and the order of rows returned by JOIN can be
non-deterministic.
> RESULTS section in .test files can unnecessarily check order
> ------------------------------------------------------------
>
> Key: IMPALA-9080
> URL: https://issues.apache.org/jira/browse/IMPALA-9080
> Project: IMPALA
> Issue Type: Bug
> Components: Infrastructure
> Reporter: Csaba Ringhofer
> Priority: Major
> Labels: flaky-test
>
> Most tests ignore the order of the expected / returned results, but in some
> cases we expect them to be in the same order (decided by
> https://github.com/apache/impala/blob/e05a5323785ecb09e45bdb5dfc96533e68256175/tests/common/test_result_verifier.py#L463
> )
> The logic above simply checks for the presence of ORDER BY in the query,
> while the results should be in a deterministic order only if there is an
> ORDER BY in the outermost query.
> The following query caused a test failure while the rows returned were
> correct, just in different order:
> {code}
> select leftSide.id, rnk
> from complextypestbl leftSide
> left outer join (
> select t.id, rank() over (order by t.id) as rnk
> from complextypestbl t left outer join t.int_array
> ) rightSide
> on leftSide.id = rightSide.id;
> {code}
> The ORDER BY is in an analytical function, which sorts the results, but those
> results are joined with a table, and the order of rows returned by JOIN can
> be non-deterministic.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
