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

Basic implementation of recursive planning of non-colocated relation joins

recursively_plan_tables
Onder Kalaci 2018-11-23 13:41:12 +03:00
parent 6fd5c73444
commit 94ea93c0ae
5 changed files with 134 additions and 61 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
src/backend/distributed/planner/query_colocation_checker.c
View File

@ -31,10 +31,10 @@
#include "parser/parse_relation.h" #include "parser/parse_relation.h"
#include "optimizer/planner.h" #include "optimizer/planner.h"
#include "optimizer/prep.h" #include "optimizer/prep.h"
#include "utils/rel.h"
static RangeTblEntry * AnchorRte(Query *subquery); static RangeTblEntry * AnchorRte(Query *subquery);
static Query * WrapRteRelationIntoSubquery(RangeTblEntry *rteRelation);
static List * UnionRelationRestrictionLists(List *firstRelationList, static List * UnionRelationRestrictionLists(List *firstRelationList,
List *secondRelationList); List *secondRelationList);
@ -238,14 +238,16 @@ SubqueryColocated(Query *subquery, ColocatedJoinChecker *checker)
* projections. The returned query should be used cautiosly and it is mostly * projections. The returned query should be used cautiosly and it is mostly
* designed for generating a stub query. * designed for generating a stub query.
*/ */
static Query * Query *
WrapRteRelationIntoSubquery(RangeTblEntry *rteRelation) WrapRteRelationIntoSubquery(RangeTblEntry *rteRelation)
{ {
Query *subquery = makeNode(Query); Query *subquery = makeNode(Query);
RangeTblRef *newRangeTableRef = makeNode(RangeTblRef); RangeTblRef *newRangeTableRef = makeNode(RangeTblRef);
RangeTblEntry *newRangeTableEntry = NULL; RangeTblEntry *newRangeTableEntry = NULL;
Var *targetColumn = NULL;
TargetEntry *targetEntry = NULL; Relation relation = NULL;
int attributeNumber = 0;
int numberOfAttributes = 0;
subquery->commandType = CMD_SELECT; subquery->commandType = CMD_SELECT;
@ -258,14 +260,25 @@ WrapRteRelationIntoSubquery(RangeTblEntry *rteRelation)
newRangeTableRef->rtindex = 1; newRangeTableRef->rtindex = 1;
subquery->jointree = makeFromExpr(list_make1(newRangeTableRef), NULL); subquery->jointree = makeFromExpr(list_make1(newRangeTableRef), NULL);
/* Need the whole row as a junk var */ relation = relation_open(rteRelation->relid, AccessShareLock);
targetColumn = makeWholeRowVar(newRangeTableEntry, newRangeTableRef->rtindex, 0,
false);
/* create a dummy target entry */ numberOfAttributes = RelationGetNumberOfAttributes(relation);
targetEntry = makeTargetEntry((Expr *) targetColumn, 1, "wholerow", true);
for (attributeNumber = 1; attributeNumber <= numberOfAttributes; attributeNumber++)
{
Form_pg_attribute attributeTuple = TupleDescAttr(relation->rd_att,
attributeNumber - 1);
Var *targetColumn =
makeVar(1, attributeNumber, attributeTuple->atttypid,
attributeTuple->atttypmod, attributeTuple->attcollation, 0);
TargetEntry *targetEntry =
makeTargetEntry((Expr *) targetColumn, attributeNumber,
strdup(attributeTuple->attname.data), false);
subquery->targetList = lappend(subquery->targetList, targetEntry); subquery->targetList = lappend(subquery->targetList, targetEntry);
}
relation_close(relation, NoLock);
return subquery; return subquery;
} }

54
src/backend/distributed/planner/recursive_planning.c
View File

@ -78,6 +78,7 @@
#include "nodes/relation.h" #include "nodes/relation.h"
#include "utils/builtins.h" #include "utils/builtins.h"
#include "utils/guc.h" #include "utils/guc.h"
#include "utils/lsyscache.h"
#include "../../../include/distributed/query_pushdown_planning.h" #include "../../../include/distributed/query_pushdown_planning.h"
@ -154,6 +155,8 @@ static bool ShouldRecursivelyPlanSetOperation(Query *query,
static void RecursivelyPlanSetOperations(Query *query, Node *node, static void RecursivelyPlanSetOperations(Query *query, Node *node,
RecursivePlanningContext *context); RecursivePlanningContext *context);
static bool IsLocalTableRTE(Node *node); static bool IsLocalTableRTE(Node *node);
static void RecursivelyPlanRTERelation(RangeTblEntry *relationRte,
RecursivePlanningContext *planningContext);
static void RecursivelyPlanSubquery(Query *subquery, static void RecursivelyPlanSubquery(Query *subquery,
RecursivePlanningContext *planningContext); RecursivePlanningContext *planningContext);
static DistributedSubPlan * CreateDistributedSubPlan(uint32 subPlanId, static DistributedSubPlan * CreateDistributedSubPlan(uint32 subPlanId,
@ -474,12 +477,16 @@ RecursivelyPlanNonColocatedJoinWalker(Node *joinNode,
RangeTblEntry *rte = rt_fetch(rangeTableIndex, rangeTableList); RangeTblEntry *rte = rt_fetch(rangeTableIndex, rangeTableList);
Query *subquery = NULL; Query *subquery = NULL;
/* we're only interested in subqueries for now */ if (rte->rtekind == RTE_RELATION)
if (rte->rtekind != RTE_SUBQUERY)
{ {
return; subquery = WrapRteRelationIntoSubquery(rte);
if (!SubqueryColocated(subquery, colocatedJoinChecker))
{
RecursivelyPlanRTERelation(rte, recursivePlanningContext);
} }
}
else if (rte->rtekind == RTE_SUBQUERY)
{
/* /*
* If the subquery is not colocated with the anchor subquery, * If the subquery is not colocated with the anchor subquery,
* recursively plan it. * recursively plan it.
@ -490,6 +497,7 @@ RecursivelyPlanNonColocatedJoinWalker(Node *joinNode,
RecursivelyPlanSubquery(subquery, recursivePlanningContext); RecursivelyPlanSubquery(subquery, recursivePlanningContext);
} }
} }
}
else else
{ {
pg_unreachable(); pg_unreachable();
@ -1063,6 +1071,44 @@ IsLocalTableRTE(Node *node)
} }
/*
* RecursivelyPlanRTERelation wraps the relation with a subquery
* (e.g., users_table becomes (SELECT * FROM users_table) as users_table).
*
* Later, the subquery is recursively planned via RecursivelyPlanSubquery().
*/
static void
RecursivelyPlanRTERelation(RangeTblEntry *relationRte,
RecursivePlanningContext *planningContext)
{
Query *subquery = WrapRteRelationIntoSubquery(relationRte);
RangeTblEntry *wrappedSubqueryEntry = makeNode(RangeTblEntry);
ListCell *columnListCell = NULL;
List *colNames = NIL;
char *relationName = get_rel_name(relationRte->relid);
wrappedSubqueryEntry->rtekind = RTE_SUBQUERY;
wrappedSubqueryEntry->subquery = copyObject(subquery);
foreach(columnListCell, subquery->targetList)
{
TargetEntry *tle = (TargetEntry *) lfirst(columnListCell);
colNames =
lappend(colNames, makeString(pstrdup(tle->resname)));
}
wrappedSubqueryEntry->eref = makeAlias(pstrdup(relationName), copyObject(colNames));
wrappedSubqueryEntry->alias = makeAlias(pstrdup(relationName), copyObject(colNames));
wrappedSubqueryEntry->inFromCl = true;
memcpy(relationRte, wrappedSubqueryEntry, sizeof(RangeTblEntry));
RecursivelyPlanSubquery(wrappedSubqueryEntry->subquery, planningContext);
}
/* /*
* RecursivelyPlanQuery recursively plans a query, replaces it with a * RecursivelyPlanQuery recursively plans a query, replaces it with a
* result query and returns the subplan. * result query and returns the subplan.

1
src/include/distributed/query_colocation_checker.h
View File

@ -34,6 +34,7 @@ extern ColocatedJoinChecker CreateColocatedJoinChecker(Query *subquery,
PlannerRestrictionContext * PlannerRestrictionContext *
restrictionContext); restrictionContext);
extern bool SubqueryColocated(Query *subquery, ColocatedJoinChecker *context); extern bool SubqueryColocated(Query *subquery, ColocatedJoinChecker *context);
extern Query * WrapRteRelationIntoSubquery(RangeTblEntry *rteRelation);
#endif /* QUERY_COLOCATION_CHECKER_H */ #endif /* QUERY_COLOCATION_CHECKER_H */

10
src/test/regress/expected/multi_subquery_complex_reference_clause.out
View File

@ -1252,7 +1252,15 @@ WHERE
GROUP BY 1 GROUP BY 1
ORDER BY 2 DESC, 1 DESC ORDER BY 2 DESC, 1 DESC
LIMIT 5; LIMIT 5;
ERROR: complex joins are only supported when all distributed tables are joined on their distribution columns with equal operator user_id | count
---------+-------
5 | 936
4 | 644
6 | 390
3 | 289
2 | 144
(5 rows)
SELECT foo.user_id FROM SELECT foo.user_id FROM
( (
SELECT m.user_id, random() FROM users_table m JOIN events_reference_table r ON int4eq(m.user_id, r.user_id) SELECT m.user_id, random() FROM users_table m JOIN events_reference_table r ON int4eq(m.user_id, r.user_id)

83
src/test/regress/expected/non_colocated_subquery_joins.out
View File

@ -638,9 +638,14 @@ SELECT true AS valid FROM explain_json_2($$
FROM FROM
(users_table u1 JOIN users_table u2 using(value_1)) a JOIN (SELECT value_1, random() FROM users_table) as u3 USING (value_1); (users_table u1 JOIN users_table u2 using(value_1)) a JOIN (SELECT value_1, random() FROM users_table) as u3 USING (value_1);
$$); $$);
DEBUG: generating subplan 66_1 for subquery SELECT value_1, random() AS random FROM public.users_table DEBUG: generating subplan 66_1 for subquery SELECT user_id, "time", value_1, value_2, value_3, value_4 FROM public.users_table u2
DEBUG: Plan 66 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((public.users_table u1 JOIN public.users_table u2 USING (value_1)) a(value_1, user_id, "time", value_2, value_3, value_4, user_id_1, time_1, value_2_1, value_3_1, value_4_1) JOIN (SELECT intermediate_result.value_1, intermediate_result.random FROM read_intermediate_result('66_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer, random double precision)) u3 USING (value_1)) DEBUG: generating subplan 66_2 for subquery SELECT value_1, random() AS random FROM public.users_table
ERROR: complex joins are only supported when all distributed tables are joined on their distribution columns with equal operator DEBUG: Plan 66 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((public.users_table u1 JOIN (SELECT intermediate_result.user_id, intermediate_result."time", intermediate_result.value_1, intermediate_result.value_2, intermediate_result.value_3, intermediate_result.value_4 FROM read_intermediate_result('66_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, "time" timestamp without time zone, value_1 integer, value_2 integer, value_3 double precision, value_4 bigint)) users_table(user_id, "time", value_1, value_2, value_3, value_4) USING (value_1)) a(value_1, user_id, "time", value_2, value_3, value_4, user_id_1, time_1, value_2_1, value_3_1, value_4_1) JOIN (SELECT intermediate_result.value_1, intermediate_result.random FROM read_intermediate_result('66_2'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer, random double precision)) u3 USING (value_1))
valid
-------
t
(1 row)
-- a very similar query to the above -- a very similar query to the above
-- however, this time we users a subquery instead of join alias, and it works -- however, this time we users a subquery instead of join alias, and it works
SELECT true AS valid FROM explain_json_2($$ SELECT true AS valid FROM explain_json_2($$
@ -651,8 +656,8 @@ SELECT true AS valid FROM explain_json_2($$
(SELECT * FROM users_table u1 JOIN users_table u2 using(value_1)) a JOIN (SELECT value_1, random() FROM users_table) as u3 USING (value_1); (SELECT * FROM users_table u1 JOIN users_table u2 using(value_1)) a JOIN (SELECT value_1, random() FROM users_table) as u3 USING (value_1);
$$); $$);
DEBUG: cannot use real time executor with repartition jobs DEBUG: cannot use real time executor with repartition jobs
DEBUG: generating subplan 68_1 for subquery SELECT u1.value_1, u1.user_id, u1."time", u1.value_2, u1.value_3, u1.value_4, u2.user_id, u2."time", u2.value_2, u2.value_3, u2.value_4 FROM (public.users_table u1 JOIN public.users_table u2 USING (value_1)) DEBUG: generating subplan 69_1 for subquery SELECT u1.value_1, u1.user_id, u1."time", u1.value_2, u1.value_3, u1.value_4, u2.user_id, u2."time", u2.value_2, u2.value_3, u2.value_4 FROM (public.users_table u1 JOIN public.users_table u2 USING (value_1))
DEBUG: Plan 68 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT intermediate_result.value_1, intermediate_result.user_id, intermediate_result."time", intermediate_result.value_2, intermediate_result.value_3, intermediate_result.value_4, intermediate_result.user_id_1 AS user_id, intermediate_result.time_1 AS "time", intermediate_result.value_2_1 AS value_2, intermediate_result.value_3_1 AS value_3, intermediate_result.value_4_1 AS value_4 FROM read_intermediate_result('68_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer, user_id integer, "time" timestamp without time zone, value_2 integer, value_3 double precision, value_4 bigint, user_id_1 integer, time_1 timestamp without time zone, value_2_1 integer, value_3_1 double precision, value_4_1 bigint)) a(value_1, user_id, "time", value_2, value_3, value_4, user_id_1, time_1, value_2_1, value_3_1, value_4_1) JOIN (SELECT users_table.value_1, random() AS random FROM public.users_table) u3 USING (value_1)) DEBUG: Plan 69 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT intermediate_result.value_1, intermediate_result.user_id, intermediate_result."time", intermediate_result.value_2, intermediate_result.value_3, intermediate_result.value_4, intermediate_result.user_id_1 AS user_id, intermediate_result.time_1 AS "time", intermediate_result.value_2_1 AS value_2, intermediate_result.value_3_1 AS value_3, intermediate_result.value_4_1 AS value_4 FROM read_intermediate_result('69_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer, user_id integer, "time" timestamp without time zone, value_2 integer, value_3 double precision, value_4 bigint, user_id_1 integer, time_1 timestamp without time zone, value_2_1 integer, value_3_1 double precision, value_4_1 bigint)) a(value_1, user_id, "time", value_2, value_3, value_4, user_id_1, time_1, value_2_1, value_3_1, value_4_1) JOIN (SELECT users_table.value_1, random() AS random FROM public.users_table) u3 USING (value_1))
valid valid
------- -------
t t
@ -670,8 +675,8 @@ SELECT true AS valid FROM explain_json_2($$
events_table events_table
using (value_2); using (value_2);
$$); $$);
DEBUG: generating subplan 70_1 for subquery SELECT value_2, random() AS random FROM public.users_table DEBUG: generating subplan 71_1 for subquery SELECT value_2, random() AS random FROM public.users_table
DEBUG: Plan 70 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT intermediate_result.value_2, intermediate_result.random FROM read_intermediate_result('70_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer, random double precision)) u1 JOIN public.events_table USING (value_2)) DEBUG: Plan 71 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT intermediate_result.value_2, intermediate_result.random FROM read_intermediate_result('71_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer, random double precision)) u1 JOIN public.events_table USING (value_2))
valid valid
------- -------
t t
@ -688,8 +693,8 @@ SELECT true AS valid FROM explain_json_2($$
(SELECT value_2, random() FROM users_table) as u2 (SELECT value_2, random() FROM users_table) as u2
USING(value_2); USING(value_2);
$$); $$);
DEBUG: generating subplan 72_1 for subquery SELECT value_2, random() AS random FROM public.users_table DEBUG: generating subplan 73_1 for subquery SELECT value_2, random() AS random FROM public.users_table
DEBUG: Plan 72 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT users_table.value_2, random() AS random FROM public.users_table) u1 LEFT JOIN (SELECT intermediate_result.value_2, intermediate_result.random FROM read_intermediate_result('72_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer, random double precision)) u2 USING (value_2)) DEBUG: Plan 73 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT users_table.value_2, random() AS random FROM public.users_table) u1 LEFT JOIN (SELECT intermediate_result.value_2, intermediate_result.random FROM read_intermediate_result('73_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer, random double precision)) u2 USING (value_2))
valid valid
------- -------
t t
@ -708,8 +713,8 @@ SELECT true AS valid FROM explain_json_2($$
(SELECT value_2, random() FROM users_table) as u2 (SELECT value_2, random() FROM users_table) as u2
USING(value_2); USING(value_2);
$$); $$);
DEBUG: generating subplan 74_1 for subquery SELECT value_2, random() AS random FROM public.users_table DEBUG: generating subplan 75_1 for subquery SELECT value_2, random() AS random FROM public.users_table
DEBUG: Plan 74 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT users_table.value_2, random() AS random FROM public.users_table) u1 RIGHT JOIN (SELECT intermediate_result.value_2, intermediate_result.random FROM read_intermediate_result('74_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer, random double precision)) u2 USING (value_2)) DEBUG: Plan 75 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT users_table.value_2, random() AS random FROM public.users_table) u1 RIGHT JOIN (SELECT intermediate_result.value_2, intermediate_result.random FROM read_intermediate_result('75_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer, random double precision)) u2 USING (value_2))
ERROR: cannot pushdown the subquery ERROR: cannot pushdown the subquery
-- set operations may produce not very efficient plans -- set operations may produce not very efficient plans
-- although we could have picked a as our anchor subquery, -- although we could have picked a as our anchor subquery,
@ -727,11 +732,11 @@ SELECT true AS valid FROM explain_json_2($$
(SELECT value_1 FROM users_table) as foo ON (a.user_id = foo.value_1) (SELECT value_1 FROM users_table) as foo ON (a.user_id = foo.value_1)
); );
$$); $$);
DEBUG: generating subplan 77_1 for subquery SELECT user_id FROM public.users_table DEBUG: generating subplan 78_1 for subquery SELECT user_id FROM public.users_table
DEBUG: generating subplan 77_2 for subquery SELECT user_id FROM public.users_table DEBUG: generating subplan 78_2 for subquery SELECT user_id FROM public.users_table
DEBUG: Plan 77 query after replacing subqueries and CTEs: SELECT intermediate_result.user_id FROM read_intermediate_result('77_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer) UNION SELECT intermediate_result.user_id FROM read_intermediate_result('77_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer) DEBUG: Plan 78 query after replacing subqueries and CTEs: SELECT intermediate_result.user_id FROM read_intermediate_result('78_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer) UNION SELECT intermediate_result.user_id FROM read_intermediate_result('78_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)
DEBUG: generating subplan 76_1 for subquery SELECT users_table.user_id FROM public.users_table UNION SELECT users_table.user_id FROM public.users_table DEBUG: generating subplan 77_1 for subquery SELECT users_table.user_id FROM public.users_table UNION SELECT users_table.user_id FROM public.users_table
DEBUG: Plan 76 query after replacing subqueries and CTEs: SELECT a.user_id, foo.value_1 FROM ((SELECT intermediate_result.user_id FROM read_intermediate_result('76_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) a JOIN (SELECT users_table.value_1 FROM public.users_table) foo ON ((a.user_id OPERATOR(pg_catalog.=) foo.value_1))) DEBUG: Plan 77 query after replacing subqueries and CTEs: SELECT a.user_id, foo.value_1 FROM ((SELECT intermediate_result.user_id FROM read_intermediate_result('77_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) a JOIN (SELECT users_table.value_1 FROM public.users_table) foo ON ((a.user_id OPERATOR(pg_catalog.=) foo.value_1)))
valid valid
------- -------
t t
@ -751,11 +756,11 @@ SELECT true AS valid FROM explain_json_2($$
users_table as foo ON (a.user_id = foo.value_1) users_table as foo ON (a.user_id = foo.value_1)
); );
$$); $$);
DEBUG: generating subplan 81_1 for subquery SELECT user_id FROM public.users_table DEBUG: generating subplan 82_1 for subquery SELECT user_id FROM public.users_table
DEBUG: generating subplan 81_2 for subquery SELECT user_id FROM public.users_table DEBUG: generating subplan 82_2 for subquery SELECT user_id FROM public.users_table
DEBUG: Plan 81 query after replacing subqueries and CTEs: SELECT intermediate_result.user_id FROM read_intermediate_result('81_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer) UNION SELECT intermediate_result.user_id FROM read_intermediate_result('81_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer) DEBUG: Plan 82 query after replacing subqueries and CTEs: SELECT intermediate_result.user_id FROM read_intermediate_result('82_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer) UNION SELECT intermediate_result.user_id FROM read_intermediate_result('82_2'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)
DEBUG: generating subplan 80_1 for subquery SELECT users_table.user_id FROM public.users_table UNION SELECT users_table.user_id FROM public.users_table DEBUG: generating subplan 81_1 for subquery SELECT users_table.user_id FROM public.users_table UNION SELECT users_table.user_id FROM public.users_table
DEBUG: Plan 80 query after replacing subqueries and CTEs: SELECT a.user_id, foo.user_id, foo."time", foo.value_1, foo.value_2, foo.value_3, foo.value_4 FROM ((SELECT intermediate_result.user_id FROM read_intermediate_result('80_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) a JOIN public.users_table foo ON ((a.user_id OPERATOR(pg_catalog.=) foo.value_1))) DEBUG: Plan 81 query after replacing subqueries and CTEs: SELECT a.user_id, foo.user_id, foo."time", foo.value_1, foo.value_2, foo.value_3, foo.value_4 FROM ((SELECT intermediate_result.user_id FROM read_intermediate_result('81_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer)) a JOIN public.users_table foo ON ((a.user_id OPERATOR(pg_catalog.=) foo.value_1)))
valid valid
------- -------
t t
@ -784,8 +789,8 @@ SELECT true AS valid FROM explain_json_2($$
ON(foo.user_id = bar.value_1) ON(foo.user_id = bar.value_1)
); );
$$); $$);
DEBUG: generating subplan 84_1 for subquery SELECT value_1 FROM public.users_table DEBUG: generating subplan 85_1 for subquery SELECT value_1 FROM public.users_table
DEBUG: Plan 84 query after replacing subqueries and CTEs: SELECT foo.user_id, a.user_id, bar.value_1 FROM (((SELECT users_table.user_id FROM public.users_table) foo JOIN (SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])) UNION SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) a ON ((a.user_id OPERATOR(pg_catalog.=) foo.user_id))) JOIN (SELECT intermediate_result.value_1 FROM read_intermediate_result('84_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer)) bar ON ((foo.user_id OPERATOR(pg_catalog.=) bar.value_1))) DEBUG: Plan 85 query after replacing subqueries and CTEs: SELECT foo.user_id, a.user_id, bar.value_1 FROM (((SELECT users_table.user_id FROM public.users_table) foo JOIN (SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])) UNION SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) a ON ((a.user_id OPERATOR(pg_catalog.=) foo.user_id))) JOIN (SELECT intermediate_result.value_1 FROM read_intermediate_result('85_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer)) bar ON ((foo.user_id OPERATOR(pg_catalog.=) bar.value_1)))
valid valid
------- -------
t t
@ -823,13 +828,13 @@ SELECT true AS valid FROM explain_json_2($$
WHERE WHERE
non_colocated_subquery.value_2 != non_colocated_subquery_2.cnt non_colocated_subquery.value_2 != non_colocated_subquery_2.cnt
$$); $$);
DEBUG: generating subplan 86_1 for CTE non_colocated_subquery: SELECT foo.value_2 FROM (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])))) foo, (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8])))) bar WHERE (foo.value_2 OPERATOR(pg_catalog.=) bar.value_2) DEBUG: generating subplan 87_1 for CTE non_colocated_subquery: SELECT foo.value_2 FROM (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])))) foo, (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8])))) bar WHERE (foo.value_2 OPERATOR(pg_catalog.=) bar.value_2)
DEBUG: generating subplan 87_1 for subquery SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) DEBUG: generating subplan 88_1 for subquery SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8])))
DEBUG: Plan 87 query after replacing subqueries and CTEs: SELECT foo.value_2 FROM (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])))) foo, (SELECT intermediate_result.value_2 FROM read_intermediate_result('87_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) bar WHERE (foo.value_2 OPERATOR(pg_catalog.=) bar.value_2) DEBUG: Plan 88 query after replacing subqueries and CTEs: SELECT foo.value_2 FROM (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])))) foo, (SELECT intermediate_result.value_2 FROM read_intermediate_result('88_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) bar WHERE (foo.value_2 OPERATOR(pg_catalog.=) bar.value_2)
DEBUG: generating subplan 86_2 for CTE non_colocated_subquery_2: SELECT count(*) AS cnt FROM public.events_table WHERE (event_type OPERATOR(pg_catalog.=) ANY (SELECT events_table_1.event_type FROM public.events_table events_table_1 WHERE (events_table_1.user_id OPERATOR(pg_catalog.<) 4))) DEBUG: generating subplan 87_2 for CTE non_colocated_subquery_2: SELECT count(*) AS cnt FROM public.events_table WHERE (event_type OPERATOR(pg_catalog.=) ANY (SELECT events_table_1.event_type FROM public.events_table events_table_1 WHERE (events_table_1.user_id OPERATOR(pg_catalog.<) 4)))
DEBUG: generating subplan 89_1 for subquery SELECT event_type FROM public.events_table WHERE (user_id OPERATOR(pg_catalog.<) 4) DEBUG: generating subplan 90_1 for subquery SELECT event_type FROM public.events_table WHERE (user_id OPERATOR(pg_catalog.<) 4)
DEBUG: Plan 89 query after replacing subqueries and CTEs: SELECT count(*) AS cnt FROM public.events_table WHERE (event_type OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.event_type FROM read_intermediate_result('89_1'::text, 'binary'::citus_copy_format) intermediate_result(event_type integer))) DEBUG: Plan 90 query after replacing subqueries and CTEs: SELECT count(*) AS cnt FROM public.events_table WHERE (event_type OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.event_type FROM read_intermediate_result('90_1'::text, 'binary'::citus_copy_format) intermediate_result(event_type integer)))
DEBUG: Plan 86 query after replacing subqueries and CTEs: SELECT non_colocated_subquery.value_2, non_colocated_subquery_2.cnt FROM (SELECT intermediate_result.value_2 FROM read_intermediate_result('86_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) non_colocated_subquery, (SELECT intermediate_result.cnt FROM read_intermediate_result('86_2'::text, 'binary'::citus_copy_format) intermediate_result(cnt bigint)) non_colocated_subquery_2 WHERE (non_colocated_subquery.value_2 OPERATOR(pg_catalog.<>) non_colocated_subquery_2.cnt) DEBUG: Plan 87 query after replacing subqueries and CTEs: SELECT non_colocated_subquery.value_2, non_colocated_subquery_2.cnt FROM (SELECT intermediate_result.value_2 FROM read_intermediate_result('87_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) non_colocated_subquery, (SELECT intermediate_result.cnt FROM read_intermediate_result('87_2'::text, 'binary'::citus_copy_format) intermediate_result(cnt bigint)) non_colocated_subquery_2 WHERE (non_colocated_subquery.value_2 OPERATOR(pg_catalog.<>) non_colocated_subquery_2.cnt)
valid valid
------- -------
t t
@ -848,9 +853,9 @@ SELECT true AS valid FROM explain_json_2($$
AND AND
foo.value_2 = baz.value_2 foo.value_2 = baz.value_2
$$); $$);
DEBUG: generating subplan 91_1 for subquery SELECT users_table_local.value_2 FROM non_colocated_subquery.users_table_local, non_colocated_subquery.events_table_local WHERE ((users_table_local.user_id OPERATOR(pg_catalog.=) events_table_local.user_id) AND (events_table_local.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) DEBUG: generating subplan 92_1 for subquery SELECT users_table_local.value_2 FROM non_colocated_subquery.users_table_local, non_colocated_subquery.events_table_local WHERE ((users_table_local.user_id OPERATOR(pg_catalog.=) events_table_local.user_id) AND (events_table_local.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8])))
DEBUG: generating subplan 91_2 for subquery SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[9, 10, 11, 12]))) DEBUG: generating subplan 92_2 for subquery SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[9, 10, 11, 12])))
DEBUG: Plan 91 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])))) foo, (SELECT intermediate_result.value_2 FROM read_intermediate_result('91_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) bar, (SELECT intermediate_result.value_2 FROM read_intermediate_result('91_2'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) baz WHERE ((foo.value_2 OPERATOR(pg_catalog.=) bar.value_2) AND (foo.value_2 OPERATOR(pg_catalog.=) baz.value_2)) DEBUG: Plan 92 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (SELECT users_table.value_2 FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])))) foo, (SELECT intermediate_result.value_2 FROM read_intermediate_result('92_1'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) bar, (SELECT intermediate_result.value_2 FROM read_intermediate_result('92_2'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer)) baz WHERE ((foo.value_2 OPERATOR(pg_catalog.=) bar.value_2) AND (foo.value_2 OPERATOR(pg_catalog.=) baz.value_2))
valid valid
------- -------
t t
@ -885,10 +890,10 @@ SELECT true AS valid FROM explain_json_2($$
AND AND
foo.user_id IN (SELECT users_table.user_id FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2)) foo.user_id IN (SELECT users_table.user_id FROM users_table, events_table WHERE users_table.user_id = events_table.user_id AND event_type IN (1,2))
$$); $$);
DEBUG: generating subplan 93_1 for subquery SELECT value_1, value_2 FROM public.users_table DEBUG: generating subplan 94_1 for subquery SELECT value_1, value_2 FROM public.users_table
DEBUG: generating subplan 93_2 for subquery SELECT value_1 FROM public.users_table WHERE (value_2 OPERATOR(pg_catalog.<) 1) DEBUG: generating subplan 94_2 for subquery SELECT value_1 FROM public.users_table WHERE (value_2 OPERATOR(pg_catalog.<) 1)
DEBUG: generating subplan 93_3 for subquery SELECT value_2 FROM public.users_table WHERE (value_1 OPERATOR(pg_catalog.<) 2) DEBUG: generating subplan 94_3 for subquery SELECT value_2 FROM public.users_table WHERE (value_1 OPERATOR(pg_catalog.<) 2)
DEBUG: Plan 93 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (((SELECT users_table.user_id FROM public.users_table) foo JOIN (SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])) UNION SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) a ON ((a.user_id OPERATOR(pg_catalog.=) foo.user_id))) JOIN (SELECT intermediate_result.value_1, intermediate_result.value_2 FROM read_intermediate_result('93_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer, value_2 integer)) bar ON ((foo.user_id OPERATOR(pg_catalog.=) bar.value_1))) WHERE ((bar.value_2 OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.value_1 FROM read_intermediate_result('93_2'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer))) AND (bar.value_1 OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.value_2 FROM read_intermediate_result('93_3'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer))) AND (foo.user_id OPERATOR(pg_catalog.=) ANY (SELECT users_table.user_id FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2])))))) DEBUG: Plan 94 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (((SELECT users_table.user_id FROM public.users_table) foo JOIN (SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2, 3, 4])) UNION SELECT users_table.user_id FROM public.users_table WHERE (users_table.user_id OPERATOR(pg_catalog.=) ANY (ARRAY[5, 6, 7, 8]))) a ON ((a.user_id OPERATOR(pg_catalog.=) foo.user_id))) JOIN (SELECT intermediate_result.value_1, intermediate_result.value_2 FROM read_intermediate_result('94_1'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer, value_2 integer)) bar ON ((foo.user_id OPERATOR(pg_catalog.=) bar.value_1))) WHERE ((bar.value_2 OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.value_1 FROM read_intermediate_result('94_2'::text, 'binary'::citus_copy_format) intermediate_result(value_1 integer))) AND (bar.value_1 OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.value_2 FROM read_intermediate_result('94_3'::text, 'binary'::citus_copy_format) intermediate_result(value_2 integer))) AND (foo.user_id OPERATOR(pg_catalog.=) ANY (SELECT users_table.user_id FROM public.users_table, public.events_table WHERE ((users_table.user_id OPERATOR(pg_catalog.=) events_table.user_id) AND (events_table.event_type OPERATOR(pg_catalog.=) ANY (ARRAY[1, 2]))))))
ERROR: cannot pushdown the subquery ERROR: cannot pushdown the subquery
-- make sure that we don't pick the refeence table as -- make sure that we don't pick the refeence table as
-- the anchor -- the anchor
@ -903,8 +908,8 @@ SELECT true AS valid FROM explain_json_2($$
users_table_ref.user_id = foo.user_id users_table_ref.user_id = foo.user_id
AND foo.user_id = bar.value_2; AND foo.user_id = bar.value_2;
$$); $$);
DEBUG: generating subplan 97_1 for subquery SELECT user_id, value_2 FROM public.events_table DEBUG: generating subplan 98_1 for subquery SELECT user_id, value_2 FROM public.events_table
DEBUG: Plan 97 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM public.users_reference_table users_table_ref, (SELECT users_table.user_id FROM public.users_table) foo, (SELECT intermediate_result.user_id, intermediate_result.value_2 FROM read_intermediate_result('97_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, value_2 integer)) bar WHERE ((users_table_ref.user_id OPERATOR(pg_catalog.=) foo.user_id) AND (foo.user_id OPERATOR(pg_catalog.=) bar.value_2)) DEBUG: Plan 98 query after replacing subqueries and CTEs: SELECT count(*) AS count FROM public.users_reference_table users_table_ref, (SELECT users_table.user_id FROM public.users_table) foo, (SELECT intermediate_result.user_id, intermediate_result.value_2 FROM read_intermediate_result('98_1'::text, 'binary'::citus_copy_format) intermediate_result(user_id integer, value_2 integer)) bar WHERE ((users_table_ref.user_id OPERATOR(pg_catalog.=) foo.user_id) AND (foo.user_id OPERATOR(pg_catalog.=) bar.value_2))
valid valid
------- -------
t t