[IMPALA-1458] group_concat does not obey ORDER BY

• Type: Bug
• Status: Resolved
• Priority: Minor
• Resolution: Not A Bug
• Affects Version/s: Impala 1.4.1
• Assignee: Unassigned
• Reporter: Aaron Harnly
• Created: 07/Nov/14 16:30
• Updated: 11/Dec/17 17:17
• Resolved: 07/Nov/14 19:10

Description

I have the impression that `group_concat` should follow the order of the rows it is aggregating. The documentation[1] does not specify that, but there is a test[2] which states "Use a subquery with an ORDER BY to ensure `group_concat` results are in a deterministic order".

[1] http://www.cloudera.com/content/cloudera/en/documentation/cloudera-impala/v1/latest/Installing-and-Using-Impala/ciiu_group_concat.html

[2] https://github.com/cloudera/Impala/blob/2dd31f698d4a18db9e99d4ce20da2bbc22ef748b/testdata/workloads/functional-query/queries/QueryTest/aggregation.test#L817

So, when I run the code below[3], I expect to get the result `a, b, c`.

Instead, I get `b, a, c`, which suggests the ORDER BY is having no effect.

Is this a problem with (1) my understanding of how to induce an order, (2) my and the test's expectations – `group_concat` was actually never intended to obey order, or (3) the behavior of `group_concat`?

[3]

select group_concat(name) 
from (
  select * from (
    select 'b' as name
    union all
    select 'a' as name
    union all
    select 'c' as name
  ) names
  order by name
) names