Qifan Chen has uploaded a new patch set (#28). (
http://gerrit.cloudera.org:8080/16720 )
Change subject: [WIP] IMPALA-10325: Parquet scan should use min/max statistics
to skip pages based on equi-join predicate
......................................................................
[WIP] IMPALA-10325: Parquet scan should use min/max statistics to skip pages
based on equi-join predicate
This patch adds a new class of predicates called overlap predicates
to aid in the determination of whether a Parquet row group or a page
overlap with a range computed from an equi hash join. If not, then
the entire Parquet row group or the page are skipped. The new class
of overlap predicates exist as in the form of min/max filters.
For the following query, the min and max in the min/max filter are
computed after the hash table is popuated with data from table 'b'.
These two values are compared against the min/max of each row group
or page at the scan node for 'a'.
select straight_join count(*)
from lineitem_sorted_l_shipdate a join [SHUFFLE]
lineitem_sorted_l_shipdate b
where a.l_shipdate = b.l_receiptdate
and b.l_commitdate = "1992-01-31";
An overlap predicate associated with a join column of type J and a scan
column type of S will be formed provided the following is true:
Both J and S are Booleans
Both J and S are Integers (tinyint, smallint, int, or bigint)
Both J and S are approximate numeric (float or double)
Both J and S are Decimals with the same precision and scale
Both J and S are strings (STRING, CHAR or VARCHAR)
Both J and S are date
Both J and S are timestamp
Like any existing min/max filters, MAX_NUM_RUNTIME_FILTERS query option
does not apply to the min/max filters created for overlap predicates.
The overlap predicates will always be evaluated, after the min/max
conjuncts (if any).
Two new run-time profile counters are added to report the number of row
groups or pages filtered out via the overlap predicates respectively:
1. NumMinMaxFilteredRowGroups
2. NumMinMaxFilteredPages
Testing:
1. Added data type specific overlap method tests in
min-max-filter-test.cc (boolean, int, string, date, timestamp and
decimal);
2. Unit tested on various column types (int, bigint, string
and decimal) with TPCH and TPCDS tables. Benefits were significant
when the join column on the outer table is sorted, or when the
min/max boundary values of the pages or row groups are monotonic;
3. Added new tests in min_max_filters.test to demonstrate the number of
filtered out pages and row groups.
4. Core testing.
TBD:
1. Performance measurement;
Change-Id: I379405ee75b14929df7d6b5d20dabc6f51375691
---
M be/src/exec/exec-node.h
M be/src/exec/hdfs-scan-node-base.cc
M be/src/exec/hdfs-scan-node-base.h
M be/src/exec/parquet/hdfs-parquet-scanner.cc
M be/src/exec/parquet/hdfs-parquet-scanner.h
M be/src/exec/parquet/parquet-column-stats.cc
M be/src/exec/parquet/parquet-column-stats.h
M be/src/exec/partitioned-hash-join-builder.cc
M be/src/exec/scan-node.cc
M be/src/runtime/coordinator.cc
M be/src/runtime/date-value.cc
M be/src/runtime/date-value.h
M be/src/runtime/decimal-value.h
M be/src/runtime/timestamp-value.cc
M be/src/runtime/timestamp-value.h
M be/src/util/min-max-filter-test.cc
M be/src/util/min-max-filter.cc
M be/src/util/min-max-filter.h
M common/thrift/PlanNodes.thrift
M fe/src/main/java/org/apache/impala/analysis/TupleDescriptor.java
M fe/src/main/java/org/apache/impala/planner/HdfsScanNode.java
M fe/src/main/java/org/apache/impala/planner/RuntimeFilterGenerator.java
M testdata/workloads/functional-query/queries/QueryTest/min_max_filters.test
23 files changed, 1,090 insertions(+), 153 deletions(-)
git pull ssh://gerrit.cloudera.org:29418/Impala-ASF refs/changes/20/16720/28
--
To view, visit http://gerrit.cloudera.org:8080/16720
To unsubscribe, visit http://gerrit.cloudera.org:8080/settings
Gerrit-Project: Impala-ASF
Gerrit-Branch: master
Gerrit-MessageType: newpatchset
Gerrit-Change-Id: I379405ee75b14929df7d6b5d20dabc6f51375691
Gerrit-Change-Number: 16720
Gerrit-PatchSet: 28
Gerrit-Owner: Qifan Chen <[email protected]>
Gerrit-Reviewer: Csaba Ringhofer <[email protected]>
Gerrit-Reviewer: Impala Public Jenkins <[email protected]>
Gerrit-Reviewer: Qifan Chen <[email protected]>
Gerrit-Reviewer: Tim Armstrong <[email protected]>
Gerrit-Reviewer: Zoltan Borok-Nagy <[email protected]>
