Hi,
I have weird behavious since running 2.0.1 rc.
I have a parquet dataset containing files in multiple directories.
With acceleration enabled (in my case this is an aggregation), the query return completely absurd result when it uses acceleration
My query should enable to know the number of rows for the main directories.
SELECT dir0,count(*)
FROM “swn-sframe”.raw_events
group by dir0
With acceleration, I get
dir0 EXPR$1
d2466aae-877e-488b-b4bc-611bc2dc806e 127213696
But I have no such directory in my dataset, and correct result should be:
dir0 EXPR$1
year=2018 76827816
year=2017 50385880
It’s like the acceleration would have replaced my values, as the global sum is consistent.