mirror of
https://github.com/apache/impala.git
synced 2026-02-01 21:00:29 -05:00
3f1d047224
IMPALA-5036 added an optimisation for count(star) in Parquet scans that avoids materialising dummy rows. This change provides similar optimization for Kudu tables. Instead of materializing empty rows when computing count star, we use the NumRows field from the Kudu API. The Kudu scanner tuple is modified to have one slot into which we will write the num rows statistic. The aggregate function is changed from count to a special sum function that gets initialized to 0. Tests: * Added end-to-end tests ̣* Added planner tests * Run performance tests on tpch.lineitem Kudu table with 25 set as scaling factor, on 1 node, with mt_dop set to 1, just to measure the speedup gained when scanning. Counting the rows before the optimization took around 400ms, and around 170ms after. Change-Id: Ic99e0f954d0ca65779bd531ca79ace1fcb066fb9 Reviewed-on: http://gerrit.cloudera.org:8080/14347 Reviewed-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com> Tested-by: Impala Public Jenkins <impala-public-jenkins@cloudera.com>
101 lines
1.8 KiB
Plaintext
101 lines
1.8 KiB
Plaintext
====
|
|
---- QUERY
|
|
# Tests the correctness of the Kudu count(*) optimization.
|
|
select count(1)
|
|
from functional_kudu.alltypes
|
|
---- RESULTS
|
|
7300
|
|
---- TYPES
|
|
bigint
|
|
=====
|
|
---- QUERY
|
|
# Kudu count(*) optimization with predicates on the partition columns.
|
|
select count(1)
|
|
from functional_kudu.alltypes where year < 2010 and month > 8
|
|
---- RESULTS
|
|
1220
|
|
---- TYPES
|
|
bigint
|
|
=====
|
|
---- QUERY
|
|
# Kudu count(*) optimization with group by partition columns.
|
|
select year, month, count(1)
|
|
from functional_kudu.alltypes group by year, month
|
|
---- RESULTS
|
|
2009,1,310
|
|
2009,2,280
|
|
2009,3,310
|
|
2009,4,300
|
|
2009,5,310
|
|
2009,6,300
|
|
2009,7,310
|
|
2009,8,310
|
|
2009,9,300
|
|
2009,10,310
|
|
2009,11,300
|
|
2009,12,310
|
|
2010,1,310
|
|
2010,2,280
|
|
2010,3,310
|
|
2010,4,300
|
|
2010,5,310
|
|
2010,6,300
|
|
2010,7,310
|
|
2010,8,310
|
|
2010,9,300
|
|
2010,10,310
|
|
2010,11,300
|
|
2010,12,310
|
|
---- TYPES
|
|
int, int, bigint
|
|
=====
|
|
---- QUERY
|
|
# Kudu count(*) optimization with both group by and predicates on partition columns.
|
|
select count(1)
|
|
from functional_kudu.alltypes where year < 2010 and month > 8
|
|
group by month
|
|
---- RESULTS
|
|
310
|
|
300
|
|
310
|
|
300
|
|
---- TYPES
|
|
bigint
|
|
=====
|
|
---- QUERY
|
|
# Kudu count(*) optimization with the result going into a join.
|
|
select x.bigint_col from functional.alltypes x
|
|
inner join (
|
|
select count(1) as a from functional_kudu.alltypes group by year
|
|
) t on x.id = t.a;
|
|
---- RESULTS
|
|
0
|
|
0
|
|
---- TYPES
|
|
bigint
|
|
=====
|
|
---- QUERY
|
|
# Kudu count(*) optimization with the agg function in the having clause.
|
|
select 1 from functional_kudu.alltypes having count(*) > 1
|
|
---- RESULTS
|
|
1
|
|
---- TYPES
|
|
tinyint
|
|
====
|
|
---- QUERY
|
|
# Verify that 0 is returned for count(*) on an empty table.
|
|
select count(1) from functional_kudu.emptytable
|
|
---- RESULTS
|
|
0
|
|
---- TYPES
|
|
bigint
|
|
=====
|
|
---- QUERY
|
|
# Verify that 0 is returned when all partitions are pruned.
|
|
select count(1) from functional_kudu.alltypes where year = -1
|
|
---- RESULTS
|
|
0
|
|
---- TYPES
|
|
bigint
|
|
=====
|