I'm on version 0.8.1+b1110, the AWS AMI you so kindly supplied.
Given this query:
select a.db, a.tab, b.tab, a.partition_date, b.partition_date
from (select * from hive.lance.counts_pd where tab not like 'tmp_%') a
left join
(select * from hive.lance.counts_pd where tab like 'tmp_%') b
on a.db = b.db and concat('tmp_', a.tab) = b.tab and a.partition_date = b.partition_date
where a.partition_date is not null and b.partition_date is null
The point is to find rows where there is a 'a.tab=x' field but no 'b.tab=tmp_x' field.
This works in Presto.
db | tab | tab | partition_date | partition_date
-----------------------+--------------------------------------+------+----------------+----------------
district_analytics_v2 | agg_registered_group | NULL | 2016-01-14 | NULL
When I run it in re:dash, the results have the correct contents in the a.db' field, but the other fields are all null.
db | tab | tab | partition_date | partition_date
-----------------------+--------------------------------------+------+----------------+----------------
district_analytics_v2 | null | null | null | null
If I give an alias for "b.tab", the a.tab column has the correct values.
If I give an alias for "b.partition_date", the a.partition_date column has the correct values.
So, if two successive columns have the same name, the first will show the value from the second?