external
/
citus
mirror of https://github.com/citusdata/citus.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Push down LIMIT and HAVING when grouped by partition key. (#1641) 

We can do this because all rows belonging to a group are in the same shard when grouping by distribution column on a range/hash distributed table.
pull/1682/head
Hadi Moshayedi 2017-10-02 20:17:51 -04:00 committed by GitHub
parent 03bddcbfab
commit 11adb9b034
7 changed files with 479 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

114
src/backend/distributed/planner/multi_logical_optimizer.c
View File

@ -109,12 +109,14 @@ static void RemoveUnaryNode(MultiUnaryNode *unaryNode);
static void PullUpUnaryNode(MultiUnaryNode *unaryNode);
static void ParentSetNewChild(MultiNode *parentNode, MultiNode *oldChildNode,
MultiNode *newChildNode);
static bool GroupedByDisjointPartitionColumn(List *tableNodeList,
MultiExtendedOp *opNode);
/* Local functions forward declarations for aggregate expressions */
static void ApplyExtendedOpNodes(MultiExtendedOp *originalNode,
MultiExtendedOp *masterNode,
MultiExtendedOp *workerNode);
static void TransformSubqueryNode(MultiTable *subqueryNode);
static void TransformSubqueryNode(MultiTable *subqueryNode, List *tableNodeList);
static MultiExtendedOp * MasterExtendedOpNode(MultiExtendedOp *originalOpNode);
static Node * MasterAggregateMutator(Node *originalNode,
MasterAggregateWalkerContext *walkerContext);
@ -123,7 +125,8 @@ static Expr * MasterAggregateExpression(Aggref *originalAggregate,
static Expr * MasterAverageExpression(Oid sumAggregateType, Oid countAggregateType,
AttrNumber *columnId);
static Expr * AddTypeConversion(Node *originalAggregate, Node *newExpression);
static MultiExtendedOp * WorkerExtendedOpNode(MultiExtendedOp *originalOpNode);
static MultiExtendedOp * WorkerExtendedOpNode(MultiExtendedOp *originalOpNode,
bool groupedByDisjointPartitionColumn);
static bool WorkerAggregateWalker(Node *node,
WorkerAggregateWalkerContext *walkerContext);
static List * WorkerAggregateExpressionList(Aggref *originalAggregate,
@ -151,8 +154,10 @@ static bool TablePartitioningSupportsDistinct(List *tableNodeList,
static bool GroupedByColumn(List *groupClauseList, List *targetList, Var *column);
/* Local functions forward declarations for limit clauses */
static Node * WorkerLimitCount(MultiExtendedOp *originalOpNode);
static List * WorkerSortClauseList(MultiExtendedOp *originalOpNode);
static Node * WorkerLimitCount(MultiExtendedOp *originalOpNode,
bool groupedByDisjointPartitionColumn);
static List * WorkerSortClauseList(MultiExtendedOp *originalOpNode,
bool groupedByDisjointPartitionColumn);
static bool CanPushDownLimitApproximate(List *sortClauseList, List *targetList);
static bool HasOrderByAggregate(List *sortClauseList, List *targetList);
static bool HasOrderByAverage(List *sortClauseList, List *targetList);
@ -177,6 +182,7 @@ void
MultiLogicalPlanOptimize(MultiTreeRoot *multiLogicalPlan)
{
bool hasOrderByHllType = false;
bool groupedByDisjointPartitionColumn = false;
List *selectNodeList = NIL;
List *projectNodeList = NIL;
List *collectNodeList = NIL;
@ -251,19 +257,23 @@ MultiLogicalPlanOptimize(MultiTreeRoot *multiLogicalPlan)
extendedOpNodeList = FindNodesOfType(logicalPlanNode, T_MultiExtendedOp);
extendedOpNode = (MultiExtendedOp *) linitial(extendedOpNodeList);
tableNodeList = FindNodesOfType(logicalPlanNode, T_MultiTable);
groupedByDisjointPartitionColumn = GroupedByDisjointPartitionColumn(tableNodeList,
extendedOpNode);
masterExtendedOpNode = MasterExtendedOpNode(extendedOpNode);
workerExtendedOpNode = WorkerExtendedOpNode(extendedOpNode);
workerExtendedOpNode = WorkerExtendedOpNode(extendedOpNode,
groupedByDisjointPartitionColumn);
ApplyExtendedOpNodes(extendedOpNode, masterExtendedOpNode, workerExtendedOpNode);
tableNodeList = FindNodesOfType(logicalPlanNode, T_MultiTable);
foreach(tableNodeCell, tableNodeList)
{
MultiTable *tableNode = (MultiTable *) lfirst(tableNodeCell);
if (tableNode->relationId == SUBQUERY_RELATION_ID)
{
ErrorIfContainsUnsupportedAggregate((MultiNode *) tableNode);
TransformSubqueryNode(tableNode);
TransformSubqueryNode(tableNode, tableNodeList);
}
}
@ -1150,14 +1160,17 @@ ApplyExtendedOpNodes(MultiExtendedOp *originalNode, MultiExtendedOp *masterNode,
* operator node.
*/
static void
TransformSubqueryNode(MultiTable *subqueryNode)
TransformSubqueryNode(MultiTable *subqueryNode, List *tableNodeList)
{
MultiExtendedOp *extendedOpNode =
(MultiExtendedOp *) ChildNode((MultiUnaryNode *) subqueryNode);
MultiNode *collectNode = ChildNode((MultiUnaryNode *) extendedOpNode);
MultiNode *collectChildNode = ChildNode((MultiUnaryNode *) collectNode);
bool groupedByDisjointPartitionColumn =
GroupedByDisjointPartitionColumn(tableNodeList, extendedOpNode);
MultiExtendedOp *masterExtendedOpNode = MasterExtendedOpNode(extendedOpNode);
MultiExtendedOp *workerExtendedOpNode = WorkerExtendedOpNode(extendedOpNode);
MultiExtendedOp *workerExtendedOpNode =
WorkerExtendedOpNode(extendedOpNode, groupedByDisjointPartitionColumn);
MultiPartition *partitionNode = CitusMakeNode(MultiPartition);
List *groupClauseList = extendedOpNode->groupClauseList;
List *targetEntryList = extendedOpNode->targetList;
@ -1759,7 +1772,8 @@ AddTypeConversion(Node *originalAggregate, Node *newExpression)
* list of worker extended operator.
*/
static MultiExtendedOp *
WorkerExtendedOpNode(MultiExtendedOp *originalOpNode)
WorkerExtendedOpNode(MultiExtendedOp *originalOpNode,
bool groupedByDisjointPartitionColumn)
{
MultiExtendedOp *workerExtendedOpNode = NULL;
MultiNode *parentNode = ParentNode((MultiNode *) originalOpNode);
@ -1913,8 +1927,19 @@ WorkerExtendedOpNode(MultiExtendedOp *originalOpNode)
workerExtendedOpNode->groupClauseList = groupClauseList;
/* if we can push down the limit, also set related fields */
workerExtendedOpNode->limitCount = WorkerLimitCount(originalOpNode);
workerExtendedOpNode->sortClauseList = WorkerSortClauseList(originalOpNode);
workerExtendedOpNode->limitCount = WorkerLimitCount(originalOpNode,
groupedByDisjointPartitionColumn);
workerExtendedOpNode->sortClauseList =
WorkerSortClauseList(originalOpNode, groupedByDisjointPartitionColumn);
/*
* If grouped by a partition column whose values are shards have disjoint sets
* of partition values, we can push down the having qualifier.
*/
if (havingQual != NULL && groupedByDisjointPartitionColumn)
{
workerExtendedOpNode->havingQual = originalOpNode->havingQual;
}
return workerExtendedOpNode;
}
@ -2277,6 +2302,46 @@ TypeOid(Oid schemaId, const char *typeName)
}
/*
* GroupedByDisjointPartitionColumn returns true if the query is grouped by the
* partition column of a table whose shards have disjoint sets of partition values.
*/
static bool
GroupedByDisjointPartitionColumn(List *tableNodeList, MultiExtendedOp *opNode)
{
bool result = false;
ListCell *tableNodeCell = NULL;
foreach(tableNodeCell, tableNodeList)
{
MultiTable *tableNode = (MultiTable *) lfirst(tableNodeCell);
Oid relationId = tableNode->relationId;
char partitionMethod = 0;
if (relationId == SUBQUERY_RELATION_ID || !IsDistributedTable(relationId))
{
continue;
}
partitionMethod = PartitionMethod(relationId);
if (partitionMethod != DISTRIBUTE_BY_RANGE &&
partitionMethod != DISTRIBUTE_BY_HASH)
{
continue;
}
if (GroupedByColumn(opNode->groupClauseList, opNode->targetList,
tableNode->partitionColumn))
{
result = true;
break;
}
}
return result;
}
/*
* CountDistinctHashFunctionName resolves the hll_hash function name to use for
* the given input type, and returns this function name.
@ -3178,9 +3243,11 @@ ReplaceColumnsInOpExpressionList(List *opExpressionList, Var *newColumn)
* If they can, the function returns the limit count.
*
* The limit push-down decision tree is as follows:
* group by?
* 1/ \0
* order by? (exact pd)
* group by?
* 1/ \0
* group by partition column? (exact pd)
* 0/ \1
* order by? (exact pd)
* 1/ \0
* has order by agg? (no pd)
* 1/ \0
@ -3195,7 +3262,8 @@ ReplaceColumnsInOpExpressionList(List *opExpressionList, Var *newColumn)
* returns null.
*/
static Node *
WorkerLimitCount(MultiExtendedOp *originalOpNode)
WorkerLimitCount(MultiExtendedOp *originalOpNode,
bool groupedByDisjointPartitionColumn)
{
Node *workerLimitNode = NULL;
List *groupClauseList = originalOpNode->groupClauseList;
@ -3221,11 +3289,12 @@ WorkerLimitCount(MultiExtendedOp *originalOpNode)
IsA(originalOpNode->limitOffset, Const));
/*
* If we don't have group by clauses, or if we have order by clauses without
* aggregates, we can push down the original limit. Else if we have order by
* clauses with commutative aggregates, we can push down approximate limits.
* If we don't have group by clauses, or we have group by partition column,
* or if we have order by clauses without aggregates, we can push down the
* original limit. Else if we have order by clauses with commutative aggregates,
* we can push down approximate limits.
*/
if (groupClauseList == NIL)
if (groupClauseList == NIL || groupedByDisjointPartitionColumn)
{
canPushDownLimit = true;
}
@ -3293,7 +3362,8 @@ WorkerLimitCount(MultiExtendedOp *originalOpNode)
* the function returns null.
*/
static List *
WorkerSortClauseList(MultiExtendedOp *originalOpNode)
WorkerSortClauseList(MultiExtendedOp *originalOpNode,
bool groupedByDisjointPartitionColumn)
{
List *workerSortClauseList = NIL;
List *groupClauseList = originalOpNode->groupClauseList;
@ -3314,7 +3384,7 @@ WorkerSortClauseList(MultiExtendedOp *originalOpNode)
* in different task results. By ordering on the group by clause, we ensure
* that query results are consistent.
*/
if (groupClauseList == NIL)
if (groupClauseList == NIL || groupedByDisjointPartitionColumn)
{
workerSortClauseList = originalOpNode->sortClauseList;
}

185
src/test/regress/expected/multi_having_pushdown.out Normal file
View File

@ -0,0 +1,185 @@
--
-- MULTI_HAVING_PUSHDOWN
--
ALTER SEQUENCE pg_catalog.pg_dist_shardid_seq RESTART 590000;
CREATE TABLE lineitem_hash (LIKE lineitem);
SELECT create_distributed_table('lineitem_hash', 'l_orderkey', 'hash');
create_distributed_table
--------------------------
(1 row)
CREATE TABLE orders_hash (LIKE orders);
SELECT create_distributed_table('orders_hash', 'o_orderkey', 'hash');
create_distributed_table
--------------------------
(1 row)
-- push down when table is distributed by hash and grouped by partition column
EXPLAIN (COSTS FALSE)
SELECT l_orderkey, sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash
GROUP BY l_orderkey HAVING sum(l_quantity) > 24
ORDER BY 2 DESC, 1 ASC LIMIT 3;
QUERY PLAN
--------------------------------------------------------------------------------------------------------
Limit
-> Sort
Sort Key: sum((sum(remote_scan.revenue))) DESC, remote_scan.l_orderkey
-> HashAggregate
Group Key: remote_scan.l_orderkey
Filter: (sum(remote_scan.worker_column_3) > '24'::numeric)
-> Custom Scan (Citus Real-Time)
Task Count: 32
Tasks Shown: One of 32
-> Task
Node: host=localhost port=57637 dbname=regression
-> Limit
-> Sort
Sort Key: (sum((l_extendedprice * l_discount))) DESC, l_orderkey
-> HashAggregate
Group Key: l_orderkey
Filter: (sum(l_quantity) > '24'::numeric)
-> Seq Scan on lineitem_hash_590000 lineitem_hash
(18 rows)
-- but don't push down when table is distributed by append
EXPLAIN (COSTS FALSE)
SELECT l_orderkey, sum(l_extendedprice * l_discount) as revenue
FROM lineitem
GROUP BY l_orderkey HAVING sum(l_quantity) > 24
ORDER BY 2 DESC, 1 ASC LIMIT 3;
QUERY PLAN
--------------------------------------------------------------------------------
Limit
-> Sort
Sort Key: sum((sum(remote_scan.revenue))) DESC, remote_scan.l_orderkey
-> HashAggregate
Group Key: remote_scan.l_orderkey
Filter: (sum(remote_scan.worker_column_3) > '24'::numeric)
-> Custom Scan (Citus Real-Time)
Task Count: 8
Tasks Shown: One of 8
-> Task
Node: host=localhost port=57637 dbname=regression
-> HashAggregate
Group Key: l_orderkey
-> Seq Scan on lineitem_290001 lineitem
(14 rows)
-- and don't push down when not grouped by partition column
EXPLAIN (COSTS FALSE)
SELECT l_shipmode, sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash
GROUP BY l_shipmode HAVING sum(l_quantity) > 24
ORDER BY 2 DESC, 1 ASC LIMIT 3;
QUERY PLAN
------------------------------------------------------------------------------------
Limit
-> Sort
Sort Key: sum((sum(remote_scan.revenue))) DESC, remote_scan.l_shipmode
-> HashAggregate
Group Key: remote_scan.l_shipmode
Filter: (sum(remote_scan.worker_column_3) > '24'::numeric)
-> Custom Scan (Citus Real-Time)
Task Count: 32
Tasks Shown: One of 32
-> Task
Node: host=localhost port=57637 dbname=regression
-> HashAggregate
Group Key: l_shipmode
-> Seq Scan on lineitem_hash_590000 lineitem_hash
(14 rows)
-- push down if grouped by multiple rows one of which is partition column
EXPLAIN (COSTS FALSE)
SELECT l_shipmode, l_orderkey, sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash
GROUP BY l_shipmode, l_orderkey HAVING sum(l_quantity) > 24
ORDER BY 3 DESC, 1, 2 LIMIT 3;
QUERY PLAN
--------------------------------------------------------------------------------------------------------------------
Limit
-> Sort
Sort Key: sum((sum(remote_scan.revenue))) DESC, remote_scan.l_shipmode, remote_scan.l_orderkey
-> HashAggregate
Group Key: remote_scan.l_shipmode, remote_scan.l_orderkey
Filter: (sum(remote_scan.worker_column_4) > '24'::numeric)
-> Custom Scan (Citus Real-Time)
Task Count: 32
Tasks Shown: One of 32
-> Task
Node: host=localhost port=57637 dbname=regression
-> Limit
-> Sort
Sort Key: (sum((l_extendedprice * l_discount))) DESC, l_shipmode, l_orderkey
-> HashAggregate
Group Key: l_shipmode, l_orderkey
Filter: (sum(l_quantity) > '24'::numeric)
-> Seq Scan on lineitem_hash_590000 lineitem_hash
(18 rows)
-- couple more checks with joins
EXPLAIN (COSTS FALSE)
SELECT sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash, orders_hash
WHERE o_orderkey = l_orderkey
GROUP BY l_orderkey, o_orderkey, l_shipmode HAVING sum(l_quantity) > 24
ORDER BY 1 DESC LIMIT 3;
QUERY PLAN
-----------------------------------------------------------------------------------------------------------------------------------
Limit
-> Sort
Sort Key: sum((sum(remote_scan.revenue))) DESC
-> HashAggregate
Group Key: remote_scan.worker_column_2, remote_scan.worker_column_3, remote_scan.worker_column_4
Filter: (sum(remote_scan.worker_column_5) > '24'::numeric)
-> Custom Scan (Citus Real-Time)
Task Count: 32
Tasks Shown: One of 32
-> Task
Node: host=localhost port=57637 dbname=regression
-> Limit
-> Sort
Sort Key: (sum((lineitem_hash.l_extendedprice * lineitem_hash.l_discount))) DESC
-> HashAggregate
Group Key: lineitem_hash.l_orderkey, orders_hash.o_orderkey, lineitem_hash.l_shipmode
Filter: (sum(lineitem_hash.l_quantity) > '24'::numeric)
-> Hash Join
Hash Cond: (orders_hash.o_orderkey = lineitem_hash.l_orderkey)
-> Seq Scan on orders_hash_590032 orders_hash
-> Hash
-> Seq Scan on lineitem_hash_590000 lineitem_hash
(22 rows)
EXPLAIN (COSTS FALSE)
SELECT sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash, orders_hash
WHERE o_orderkey = l_orderkey
GROUP BY l_shipmode, o_clerk HAVING sum(l_quantity) > 24
ORDER BY 1 DESC LIMIT 3;
QUERY PLAN
------------------------------------------------------------------------------------------------------
Limit
-> Sort
Sort Key: sum((sum(remote_scan.revenue))) DESC
-> HashAggregate
Group Key: remote_scan.worker_column_2, remote_scan.worker_column_3
Filter: (sum(remote_scan.worker_column_4) > '24'::numeric)
-> Custom Scan (Citus Real-Time)
Task Count: 32
Tasks Shown: One of 32
-> Task
Node: host=localhost port=57637 dbname=regression
-> HashAggregate
Group Key: lineitem_hash.l_shipmode, orders_hash.o_clerk
-> Hash Join
Hash Cond: (orders_hash.o_orderkey = lineitem_hash.l_orderkey)
-> Seq Scan on orders_hash_590032 orders_hash
-> Hash
-> Seq Scan on lineitem_hash_590000 lineitem_hash
(18 rows)
DROP TABLE lineitem_hash;
DROP TABLE orders_hash;

95
src/test/regress/expected/multi_limit_clause.out
View File

@ -1,6 +1,14 @@
--
-- MULTI_LIMIT_CLAUSE
--
CREATE TABLE lineitem_hash (LIKE lineitem);
SELECT create_distributed_table('lineitem_hash', 'l_orderkey', 'hash');
create_distributed_table
--------------------------
(1 row)
INSERT INTO lineitem_hash SELECT * FROM lineitem;
-- Display debug messages on limit clause push down.
SET client_min_messages TO DEBUG1;
-- Check that we can correctly handle the Limit clause in distributed queries.
@ -203,4 +211,91 @@ DEBUG: push down of limit count: 1
1.00 | 0.00 | 99167.304347826087
(1 row)
-- We can push down LIMIT clause when we group by partition column of a hash
-- partitioned table.
SELECT l_orderkey, count(DISTINCT l_partkey)
FROM lineitem_hash
GROUP BY l_orderkey
ORDER BY 2 DESC, 1 DESC LIMIT 5;
DEBUG: push down of limit count: 5
l_orderkey | count
------------+-------
14885 | 7
14884 | 7
14821 | 7
14790 | 7
14785 | 7
(5 rows)
SELECT l_orderkey
FROM lineitem_hash
GROUP BY l_orderkey
ORDER BY l_orderkey LIMIT 5;
DEBUG: push down of limit count: 5
l_orderkey
------------
1
2
3
4
5
(5 rows)
-- Don't push down if not grouped by partition column.
SELECT max(l_orderkey)
FROM lineitem_hash
GROUP BY l_linestatus
ORDER BY 1 DESC LIMIT 2;
max
-------
14947
14916
(2 rows)
-- Don't push down if table is distributed by append
SELECT l_orderkey, max(l_shipdate)
FROM lineitem
GROUP BY l_orderkey
ORDER BY 2 DESC, 1 LIMIT 5;
l_orderkey | max
------------+------------
4678 | 11-27-1998
12384 | 11-26-1998
1124 | 11-25-1998
11523 | 11-22-1998
14694 | 11-21-1998
(5 rows)
-- Push down if grouped by multiple columns one of which is partition column.
SELECT
l_linestatus, l_orderkey, max(l_shipdate)
FROM lineitem_hash
GROUP BY l_linestatus, l_orderkey
ORDER BY 3 DESC, 1, 2 LIMIT 5;
DEBUG: push down of limit count: 5
l_linestatus | l_orderkey | max
--------------+------------+------------
O | 4678 | 11-27-1998
O | 12384 | 11-26-1998
O | 1124 | 11-25-1998
O | 11523 | 11-22-1998
O | 14694 | 11-21-1998
(5 rows)
-- Don't push down if grouped by multiple columns none of which is partition column.
SELECT
l_linestatus, l_shipmode, max(l_shipdate)
FROM lineitem_hash
GROUP BY l_linestatus, l_shipmode
ORDER BY 3 DESC, 1, 2 LIMIT 5;
l_linestatus | l_shipmode | max
--------------+------------+------------
O | AIR | 11-27-1998
O | RAIL | 11-26-1998
O | SHIP | 11-21-1998
O | REG AIR | 11-19-1998
O | TRUCK | 11-17-1998
(5 rows)
SET client_min_messages TO NOTICE;
DROP TABLE lineitem_hash;

6
src/test/regress/multi_schedule
View File

@ -57,9 +57,9 @@ test: multi_subquery_union multi_subquery_in_where_clause multi_subquery_misc
test: multi_reference_table
test: multi_outer_join_reference
test: multi_single_relation_subquery
test: multi_agg_distinct multi_agg_approximate_distinct multi_limit_clause multi_limit_clause_approximate
test: multi_average_expression multi_working_columns
test: multi_array_agg
test: multi_agg_distinct multi_agg_approximate_distinct multi_limit_clause_approximate
test: multi_average_expression multi_working_columns multi_having_pushdown
test: multi_array_agg multi_limit_clause
test: multi_agg_type_conversion multi_count_type_conversion
test: multi_partition_pruning
test: multi_join_pruning multi_hash_pruning

4
src/test/regress/multi_task_tracker_extra_schedule
View File

@ -30,9 +30,9 @@ test: multi_load_data
# Miscellaneous tests to check our query planning behavior
# ----------
test: multi_basic_queries multi_complex_expressions
test: multi_agg_distinct multi_limit_clause multi_limit_clause_approximate
test: multi_agg_distinct multi_limit_clause_approximate
test: multi_average_expression multi_working_columns
test: multi_array_agg
test: multi_array_agg multi_limit_clause
test: multi_agg_type_conversion multi_count_type_conversion
test: multi_hash_pruning
test: multi_query_directory_cleanup

57
src/test/regress/sql/multi_having_pushdown.sql Normal file
View File

@ -0,0 +1,57 @@
--
-- MULTI_HAVING_PUSHDOWN
--
ALTER SEQUENCE pg_catalog.pg_dist_shardid_seq RESTART 590000;
CREATE TABLE lineitem_hash (LIKE lineitem);
SELECT create_distributed_table('lineitem_hash', 'l_orderkey', 'hash');
CREATE TABLE orders_hash (LIKE orders);
SELECT create_distributed_table('orders_hash', 'o_orderkey', 'hash');
-- push down when table is distributed by hash and grouped by partition column
EXPLAIN (COSTS FALSE)
SELECT l_orderkey, sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash
GROUP BY l_orderkey HAVING sum(l_quantity) > 24
ORDER BY 2 DESC, 1 ASC LIMIT 3;
-- but don't push down when table is distributed by append
EXPLAIN (COSTS FALSE)
SELECT l_orderkey, sum(l_extendedprice * l_discount) as revenue
FROM lineitem
GROUP BY l_orderkey HAVING sum(l_quantity) > 24
ORDER BY 2 DESC, 1 ASC LIMIT 3;
-- and don't push down when not grouped by partition column
EXPLAIN (COSTS FALSE)
SELECT l_shipmode, sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash
GROUP BY l_shipmode HAVING sum(l_quantity) > 24
ORDER BY 2 DESC, 1 ASC LIMIT 3;
-- push down if grouped by multiple rows one of which is partition column
EXPLAIN (COSTS FALSE)
SELECT l_shipmode, l_orderkey, sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash
GROUP BY l_shipmode, l_orderkey HAVING sum(l_quantity) > 24
ORDER BY 3 DESC, 1, 2 LIMIT 3;
-- couple more checks with joins
EXPLAIN (COSTS FALSE)
SELECT sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash, orders_hash
WHERE o_orderkey = l_orderkey
GROUP BY l_orderkey, o_orderkey, l_shipmode HAVING sum(l_quantity) > 24
ORDER BY 1 DESC LIMIT 3;
EXPLAIN (COSTS FALSE)
SELECT sum(l_extendedprice * l_discount) as revenue
FROM lineitem_hash, orders_hash
WHERE o_orderkey = l_orderkey
GROUP BY l_shipmode, o_clerk HAVING sum(l_quantity) > 24
ORDER BY 1 DESC LIMIT 3;
DROP TABLE lineitem_hash;
DROP TABLE orders_hash;

45
src/test/regress/sql/multi_limit_clause.sql
View File

@ -3,6 +3,10 @@
--
CREATE TABLE lineitem_hash (LIKE lineitem);
SELECT create_distributed_table('lineitem_hash', 'l_orderkey', 'hash');
INSERT INTO lineitem_hash SELECT * FROM lineitem;
-- Display debug messages on limit clause push down.
SET client_min_messages TO DEBUG1;
@ -59,4 +63,45 @@ SELECT l_quantity, l_discount, avg(l_partkey) FROM lineitem
GROUP BY l_quantity, l_discount
ORDER BY l_quantity, l_discount LIMIT 1;
-- We can push down LIMIT clause when we group by partition column of a hash
-- partitioned table.
SELECT l_orderkey, count(DISTINCT l_partkey)
FROM lineitem_hash
GROUP BY l_orderkey
ORDER BY 2 DESC, 1 DESC LIMIT 5;
SELECT l_orderkey
FROM lineitem_hash
GROUP BY l_orderkey
ORDER BY l_orderkey LIMIT 5;
-- Don't push down if not grouped by partition column.
SELECT max(l_orderkey)
FROM lineitem_hash
GROUP BY l_linestatus
ORDER BY 1 DESC LIMIT 2;
-- Don't push down if table is distributed by append
SELECT l_orderkey, max(l_shipdate)
FROM lineitem
GROUP BY l_orderkey
ORDER BY 2 DESC, 1 LIMIT 5;
-- Push down if grouped by multiple columns one of which is partition column.
SELECT
l_linestatus, l_orderkey, max(l_shipdate)
FROM lineitem_hash
GROUP BY l_linestatus, l_orderkey
ORDER BY 3 DESC, 1, 2 LIMIT 5;
-- Don't push down if grouped by multiple columns none of which is partition column.
SELECT
l_linestatus, l_shipmode, max(l_shipdate)
FROM lineitem_hash
GROUP BY l_linestatus, l_shipmode
ORDER BY 3 DESC, 1, 2 LIMIT 5;
SET client_min_messages TO NOTICE;
DROP TABLE lineitem_hash;