hyoungjunkim created TAJO-702:
---------------------------------
Summary: ORDER BY DESC query sends all datas to single task.
Key: TAJO-702
URL: https://issues.apache.org/jira/browse/TAJO-702
Project: Tajo
Issue Type: Improvement
Reporter: hyoungjunkim
Priority: Minor
I ran TPC-H Q10 with scale 100. Last execution block for ORDER BY had 10 tasks.
But single task received all datas.
{code}
select c_custkey, c_name, sum(l_extendedprice * (1 - l_discount)) as revenue,
c_acctbal, n_name, c_address, c_phone, c_comment from customer as c join nation
as n on c.c_nationkey = n.n_nationkey join orders as o on c.c_custkey =
o.o_custkey and o.o_orderdate >= '1993-10-01' and o.o_orderdate < '1994-01-01'
join lineitem as l on l.l_orderkey = o.o_orderkey and l.l_returnflag = 'R'
group by c_custkey, c_name, c_acctbal, c_phone, n_name, c_address, c_comment
order by revenue desc
{code}
I found some bug in UniformRangePartition class. UniformRangePartition's
increment() method always add value to inc regardless of the sort order. So,
Range values are generated as follows.
{noformat}
(0=>936768.8469000002) ~ (0=>1030364.8469000002)
(0=>1030364.8469000002) ~ (0=>1123960.8469000002)
(0=>1123960.8469000002) ~ (0=>1217556.8469000002)
(0=>1217556.8469000002) ~ (0=>1311152.8469000002)
(0=>1311152.8469000002) ~ (0=>1404748.8469000002)
(0=>1404748.8469000002) ~ (0=>1498344.8469000002)
(0=>1498344.8469000002) ~ (0=>1591940.8469000002)
(0=>1591940.8469000002) ~ (0=>1685536.8469000002)
(0=>1685536.8469000002) ~ (0=>1779132.8469000002)
(0=>1779132.8469000002) ~ (0=>817.902)
{noformat}
--
This message was sent by Atlassian JIRA
(v6.2#6252)