[
https://issues.apache.org/jira/browse/PHOENIX-3054?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15456078#comment-15456078
]
James Taylor commented on PHOENIX-3054:
---------------------------------------
I think that approach looks fine, [[email protected]], but you'll need to sort
through the failing tests:
- ProductMetricsIT.testDegenerateAggregation() I'm not sure about.
[~julianhyde] - what happens if there's a GROUP BY but there are no groups
(i.e. because the WHERE clause filters them all out)?
- ProductMetricsIT.testConstantFalseHaving() seems correct. With the HAVING
clause evaluating to FALSE, there should be no rows returned. We optimize out
the HAVING clause when possible, though, so this is a bit of an edge case.
- ViewIndexIT.testDeleteViewIndexSequences() doesn't look related to your
change, but conveniently fell right into your hands as it looks namespace
related. :-) Might be related to the recent change of running tests with
generated table names, but it'd be great if you could help shed some light on
this. FYI, [~samarthjain], [~prakul].
> Counting zero null rows returns an empty result set
> ---------------------------------------------------
>
> Key: PHOENIX-3054
> URL: https://issues.apache.org/jira/browse/PHOENIX-3054
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.7.0
> Reporter: Kevin Liew
> Assignee: Ankit Singhal
> Priority: Minor
> Labels: count, count(*), null, sql
> Fix For: 4.8.1
>
> Attachments: PHOENIX-3054.patch
>
>
> Execute the following query on a table with zero null cells.
> {code}
> select count(*) from table where column1 = null
> {code}
> The result set should consist of one cell with the value `0` but the result
> set is empty.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
