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

Disallow INSERT ... SELECT to pushdown joins on non-partition keys 

With this commit, we disallow JOINs on non-partition keys. Simply,
while instantiating the qual we check whether the qual is on the
partition key.
pull/1243/head
Onder Kalaci 2017-02-20 14:00:25 +02:00
parent b8e4763a1a
commit 681da71251
3 changed files with 727 additions and 9 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

57
src/backend/distributed/planner/multi_router_planner.c
View File

@ -72,6 +72,14 @@ typedef struct WalkerState
bool badCoalesce;
} WalkerState;
typedef struct InstantiateQualContext
{
ShardInterval *targetShardInterval;
Var *relationPartitionColumn;
}InstantiateQualContext;
bool EnableRouterExecution = true;
/* planner functions forward declarations */
@ -115,7 +123,7 @@ static bool MultiRouterPlannableQuery(Query *query,
RelationRestrictionContext *restrictionContext);
static RelationRestrictionContext * CopyRelationRestrictionContext(
RelationRestrictionContext *oldContext);
static Node * InstantiatePartitionQual(Node *node, void *context);
static Node * InstantiatePartitionQualWalker(Node *node, void *context);
static DeferredErrorMessage * InsertSelectQuerySupported(Query *queryTree,
RangeTblEntry *insertRte,
RangeTblEntry *subqueryRte,
@ -377,6 +385,8 @@ RouterModifyTaskForShardInterval(Query *originalQuery, ShardInterval *shardInter
{
RelationRestriction *restriction = lfirst(restrictionCell);
List *originalBaserestrictInfo = restriction->relOptInfo->baserestrictinfo;
InstantiateQualContext instantiateQualWalker;
Var *relationPartitionKey = PartitionKey(restriction->relationId);
/*
* We haven't added the quals if all participating tables are reference
@ -387,9 +397,12 @@ RouterModifyTaskForShardInterval(Query *originalQuery, ShardInterval *shardInter
break;
}
instantiateQualWalker.relationPartitionColumn = relationPartitionKey;
instantiateQualWalker.targetShardInterval = shardInterval;
originalBaserestrictInfo =
(List *) InstantiatePartitionQual((Node *) originalBaserestrictInfo,
shardInterval);
(List *) InstantiatePartitionQualWalker((Node *) originalBaserestrictInfo,
&instantiateQualWalker);
}
/*
@ -2948,9 +2961,13 @@ CopyRelationRestrictionContext(RelationRestrictionContext *oldContext)
* (partCol >= shardMinValue && partCol <= shardMaxValue).
*/
static Node *
InstantiatePartitionQual(Node *node, void *context)
InstantiatePartitionQualWalker(Node *node, void *context)
{
ShardInterval *shardInterval = (ShardInterval *) context;
ShardInterval *shardInterval =
((InstantiateQualContext *) context)->targetShardInterval;
Var *relationPartitionColumn =
((InstantiateQualContext *) context)->relationPartitionColumn;
Assert(shardInterval->minValueExists);
Assert(shardInterval->maxValueExists);
@ -2974,6 +2991,7 @@ InstantiatePartitionQual(Node *node, void *context)
Node *leftop = get_leftop((Expr *) op);
Node *rightop = get_rightop((Expr *) op);
Param *param = NULL;
Var *currentColumn = NULL;
Var *hashedGEColumn = NULL;
OpExpr *hashedGEOpExpr = NULL;
@ -2992,10 +3010,28 @@ InstantiatePartitionQual(Node *node, void *context)
if (IsA(leftop, Param))
{
param = (Param *) leftop;
/*
* Before instantiating the qual, ensure that it is equal to
* the partition key.
*/
if (IsA(rightop, Var))
{
currentColumn = (Var *) rightop;
}
}
else if (IsA(rightop, Param))
{
param = (Param *) rightop;
/*
* Before instantiating the qual, ensure that it is equal to
* the partition key.
*/
if (IsA(leftop, Var))
{
currentColumn = (Var *) leftop;
}
}
/* not an interesting param for our purpose, so return */
@ -3004,6 +3040,13 @@ InstantiatePartitionQual(Node *node, void *context)
return node;
}
/* if the qual is not on the partition column, do not instantiate */
if (relationPartitionColumn && currentColumn &&
currentColumn->varattno != relationPartitionColumn->varattno)
{
return node;
}
/* get the integer >=, <= operators from the catalog */
integer4GEoperatorId = get_opfamily_member(INTEGER_BTREE_FAM_OID, INT4OID,
INT4OID,
@ -3052,13 +3095,13 @@ InstantiatePartitionQual(Node *node, void *context)
if (IsA(node, RestrictInfo))
{
RestrictInfo *restrictInfo = (RestrictInfo *) node;
restrictInfo->clause = (Expr *) InstantiatePartitionQual(
restrictInfo->clause = (Expr *) InstantiatePartitionQualWalker(
(Node *) restrictInfo->clause, context);
return (Node *) restrictInfo;
}
return expression_tree_mutator(node, InstantiatePartitionQual, context);
return expression_tree_mutator(node, InstantiatePartitionQualWalker, context);
}

482
src/test/regress/expected/multi_insert_select.out
View File

@ -973,7 +973,475 @@ FROM
((SELECT user_id FROM raw_events_first WHERE user_id = 15) EXCEPT
(SELECT user_id FROM raw_events_second where user_id = 17)) as foo;
ERROR: set operations are not allowed in INSERT ... SELECT queries
-- unsupported JOIN
-- some supported LEFT joins
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300004 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_first.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_first.user_id) <= '-1073741825'::integer))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300001 raw_events_first LEFT JOIN public.raw_events_second_13300005 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_first.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_first.user_id) <= '-1'::integer))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300002 raw_events_first LEFT JOIN public.raw_events_second_13300006 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_first.user_id) >= 0) AND (hashint4(raw_events_first.user_id) <= 1073741823))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300003 raw_events_first LEFT JOIN public.raw_events_second_13300007 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_first.user_id) >= 1073741824) AND (hashint4(raw_events_first.user_id) <= 2147483647))
DEBUG: Plan is router executable
INSERT INTO agg_events (user_id)
SELECT
raw_events_second.user_id
FROM
reference_table LEFT JOIN raw_events_second ON reference_table.user_id = raw_events_second.user_id;
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_second.user_id FROM (public.reference_table_13300012 reference_table LEFT JOIN public.raw_events_second_13300004 raw_events_second ON ((reference_table.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_second.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_second.user_id) <= '-1073741825'::integer))
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_second.user_id FROM (public.reference_table_13300012 reference_table LEFT JOIN public.raw_events_second_13300005 raw_events_second ON ((reference_table.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_second.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_second.user_id) <= '-1'::integer))
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_second.user_id FROM (public.reference_table_13300012 reference_table LEFT JOIN public.raw_events_second_13300006 raw_events_second ON ((reference_table.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_second.user_id) >= 0) AND (hashint4(raw_events_second.user_id) <= 1073741823))
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_second.user_id FROM (public.reference_table_13300012 reference_table LEFT JOIN public.raw_events_second_13300007 raw_events_second ON ((reference_table.user_id = raw_events_second.user_id))) WHERE ((hashint4(raw_events_second.user_id) >= 1073741824) AND (hashint4(raw_events_second.user_id) <= 2147483647))
DEBUG: Plan is router executable
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id = 10;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300004 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_first.user_id) <= '-1073741825'::integer)))
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN (SELECT NULL::integer AS user_id, NULL::timestamp without time zone AS "time", NULL::integer AS value_1, NULL::integer AS value_2, NULL::double precision AS value_3, NULL::bigint AS value_4 WHERE false) raw_events_second(user_id, "time", value_1, value_2, value_3, value_4) ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_first.user_id) <= '-1'::integer)))
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN (SELECT NULL::integer AS user_id, NULL::timestamp without time zone AS "time", NULL::integer AS value_1, NULL::integer AS value_2, NULL::double precision AS value_3, NULL::bigint AS value_4 WHERE false) raw_events_second(user_id, "time", value_1, value_2, value_3, value_4) ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= 0) AND (hashint4(raw_events_first.user_id) <= 1073741823)))
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN (SELECT NULL::integer AS user_id, NULL::timestamp without time zone AS "time", NULL::integer AS value_1, NULL::integer AS value_2, NULL::double precision AS value_3, NULL::bigint AS value_4 WHERE false) raw_events_second(user_id, "time", value_1, value_2, value_3, value_4) ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= 1073741824) AND (hashint4(raw_events_first.user_id) <= 2147483647)))
DEBUG: Plan is router executable
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_second.user_id = 10 OR raw_events_second.user_id = 11;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300004 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE (((raw_events_second.user_id = 10) OR (raw_events_second.user_id = 11)) AND ((hashint4(raw_events_first.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_first.user_id) <= '-1073741825'::integer)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300001 raw_events_first LEFT JOIN (SELECT NULL::integer AS user_id, NULL::timestamp without time zone AS "time", NULL::integer AS value_1, NULL::integer AS value_2, NULL::double precision AS value_3, NULL::bigint AS value_4 WHERE false) raw_events_second(user_id, "time", value_1, value_2, value_3, value_4) ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE (((raw_events_second.user_id = 10) OR (raw_events_second.user_id = 11)) AND ((hashint4(raw_events_first.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_first.user_id) <= '-1'::integer)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300002 raw_events_first LEFT JOIN (SELECT NULL::integer AS user_id, NULL::timestamp without time zone AS "time", NULL::integer AS value_1, NULL::integer AS value_2, NULL::double precision AS value_3, NULL::bigint AS value_4 WHERE false) raw_events_second(user_id, "time", value_1, value_2, value_3, value_4) ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE (((raw_events_second.user_id = 10) OR (raw_events_second.user_id = 11)) AND ((hashint4(raw_events_first.user_id) >= 0) AND (hashint4(raw_events_first.user_id) <= 1073741823)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300003 raw_events_first LEFT JOIN public.raw_events_second_13300007 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE (((raw_events_second.user_id = 10) OR (raw_events_second.user_id = 11)) AND ((hashint4(raw_events_first.user_id) >= 1073741824) AND (hashint4(raw_events_first.user_id) <= 2147483647)))
DEBUG: Plan is router executable
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id = 10 AND raw_events_first.user_id = 20;
DEBUG: Skipping target shard interval 13300008 since SELECT query for it pruned away
DEBUG: Skipping target shard interval 13300009 since SELECT query for it pruned away
DEBUG: Skipping target shard interval 13300010 since SELECT query for it pruned away
DEBUG: Skipping target shard interval 13300011 since SELECT query for it pruned away
DEBUG: Plan is router executable
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id = 10 AND raw_events_second.user_id = 20;
DEBUG: Skipping target shard interval 13300008 since SELECT query for it pruned away
DEBUG: Skipping target shard interval 13300009 since SELECT query for it pruned away
DEBUG: Skipping target shard interval 13300010 since SELECT query for it pruned away
DEBUG: Skipping target shard interval 13300011 since SELECT query for it pruned away
DEBUG: Plan is router executable
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id IN (19, 20, 21);
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300004 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_first.user_id) <= '-1073741825'::integer)))
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300001 raw_events_first LEFT JOIN public.raw_events_second_13300005 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_first.user_id) <= '-1'::integer)))
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300002 raw_events_first LEFT JOIN public.raw_events_second_13300006 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= 0) AND (hashint4(raw_events_first.user_id) <= 1073741823)))
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300003 raw_events_first LEFT JOIN public.raw_events_second_13300007 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_first.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= 1073741824) AND (hashint4(raw_events_first.user_id) <= 2147483647)))
DEBUG: Plan is router executable
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_second.user_id IN (19, 20, 21);
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first JOIN public.raw_events_second_13300004 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_second.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_first.user_id) <= '-1073741825'::integer)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300001 raw_events_first JOIN public.raw_events_second_13300005 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_second.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_first.user_id) <= '-1'::integer)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300003
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300002 raw_events_first JOIN public.raw_events_second_13300006 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_second.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= 0) AND (hashint4(raw_events_first.user_id) <= 1073741823)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
NOTICE: cannot use shard pruning with ANY/ALL (array expression)
HINT: Consider rewriting the expression with OR/AND clauses.
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300003 raw_events_first JOIN public.raw_events_second_13300007 raw_events_second ON ((raw_events_first.user_id = raw_events_second.user_id))) WHERE ((raw_events_second.user_id = ANY (ARRAY[19, 20, 21])) AND ((hashint4(raw_events_first.user_id) >= 1073741824) AND (hashint4(raw_events_first.user_id) <= 2147483647)))
DEBUG: Plan is router executable
-- the following is a very tricky query for Citus
-- although we do not support pushing down JOINs on non-partition
-- columns here it is safe to push it down given that we're looking for
-- a specific value (i.e., value_1 = 12) on the joining column.
-- Note that the query always hits the same shard on raw_events_second
INSERT INTO agg_events
(user_id)
SELECT raw_events_first.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_second.user_id = raw_events_first.value_1
AND raw_events_first.value_1 = 12;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM public.raw_events_first_13300000 raw_events_first, public.raw_events_second_13300007 raw_events_second WHERE (((raw_events_second.user_id = raw_events_first.value_1) AND (raw_events_first.value_1 = 12)) AND ((hashint4(raw_events_first.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_first.user_id) <= '-1073741825'::integer)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM public.raw_events_first_13300001 raw_events_first, public.raw_events_second_13300007 raw_events_second WHERE (((raw_events_second.user_id = raw_events_first.value_1) AND (raw_events_first.value_1 = 12)) AND ((hashint4(raw_events_first.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_first.user_id) <= '-1'::integer)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM public.raw_events_first_13300002 raw_events_first, public.raw_events_second_13300007 raw_events_second WHERE (((raw_events_second.user_id = raw_events_first.value_1) AND (raw_events_first.value_1 = 12)) AND ((hashint4(raw_events_first.user_id) >= 0) AND (hashint4(raw_events_first.user_id) <= 1073741823)))
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM public.raw_events_first_13300003 raw_events_first, public.raw_events_second_13300007 raw_events_second WHERE (((raw_events_second.user_id = raw_events_first.value_1) AND (raw_events_first.value_1 = 12)) AND ((hashint4(raw_events_first.user_id) >= 1073741824) AND (hashint4(raw_events_first.user_id) <= 2147483647)))
DEBUG: Plan is router executable
-- some unsupported LEFT/INNER JOINs
-- JOIN on one table with partition column other is not
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- same as the above with INNER JOIN
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- a not meaningful query
INSERT INTO agg_events
(user_id)
SELECT raw_events_second.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_first.user_id = raw_events_first.value_1;
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- both tables joined on non-partition columns
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.value_1 = raw_events_second.value_1;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- same as the above with INNER JOIN
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- although we do not support pushing down JOINs on non-partition
-- columns here it is safe to push it down given that we're looking for
-- a specific value (i.e., user_id = 10) on the joining column.
-- Note that the query always hits the same shard on raw_events_second
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1
WHERE
raw_events_first.user_id = 10;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300008 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300004 raw_events_second ON ((raw_events_first.user_id = raw_events_second.value_1))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= '-2147483648'::integer) AND (hashint4(raw_events_first.user_id) <= '-1073741825'::integer)))
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300009 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300005 raw_events_second ON ((raw_events_first.user_id = raw_events_second.value_1))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= '-1073741824'::integer) AND (hashint4(raw_events_first.user_id) <= '-1'::integer)))
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300007
DEBUG: distributed statement: INSERT INTO public.agg_events_13300010 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300006 raw_events_second ON ((raw_events_first.user_id = raw_events_second.value_1))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= 0) AND (hashint4(raw_events_first.user_id) <= 1073741823)))
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: distributed statement: INSERT INTO public.agg_events_13300011 AS citus_table_alias (user_id) SELECT raw_events_first.user_id FROM (public.raw_events_first_13300000 raw_events_first LEFT JOIN public.raw_events_second_13300007 raw_events_second ON ((raw_events_first.user_id = raw_events_second.value_1))) WHERE ((raw_events_first.user_id = 10) AND ((hashint4(raw_events_first.user_id) >= 1073741824) AND (hashint4(raw_events_first.user_id) <= 2147483647)))
DEBUG: Plan is router executable
-- same as the above with INNER JOIN
-- however this time query is not pushed down
-- to the worker. This is related to how we process
-- restriction infos, which we're considering to
-- improve
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1
WHERE raw_events_first.user_id = 10;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- make things a bit more complicate with IN clauses
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1
WHERE raw_events_first.value_1 IN (10, 11,12) OR raw_events_second.user_id IN (1,2,3,4);
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- implicit join on non partition column should also not be pushed down
INSERT INTO agg_events
(user_id)
SELECT raw_events_first.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_second.user_id = raw_events_first.value_1;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- the following is again a very tricky query for Citus
-- if the given filter was on value_1 as shown in the above, Citus could
-- push it down. But here the query is refused
INSERT INTO agg_events
(user_id)
SELECT raw_events_first.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_second.user_id = raw_events_first.value_1
AND raw_events_first.value_2 = 12;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- foo is not joined on the partition key so the query is not
-- pushed down
INSERT INTO agg_events
(user_id, value_4_agg)
SELECT
outer_most.id, max(outer_most.value)
FROM
(
SELECT f2.id as id, f2.v4 as value FROM
(SELECT
id
FROM (SELECT reference_table.user_id AS id
FROM raw_events_first LEFT JOIN
reference_table
ON (raw_events_first.value_1 = reference_table.user_id)) AS foo) as f
INNER JOIN
(SELECT v4,
v1,
id
FROM (SELECT SUM(raw_events_second.value_4) AS v4,
SUM(raw_events_first.value_1) AS v1,
raw_events_second.user_id AS id
FROM raw_events_first,
raw_events_second
WHERE raw_events_first.user_id = raw_events_second.user_id
GROUP BY raw_events_second.user_id
HAVING SUM(raw_events_second.value_4) > 10) AS foo2 ) as f2
ON (f.id = f2.id)) as outer_most
GROUP BY
outer_most.id;
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- not equals on the partition column cannot be pushed down
INSERT INTO agg_events
(value_4_agg,
value_1_agg,
@ -1170,6 +1638,18 @@ DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- cannot pushdown since subquery returns another column than partition key
INSERT INTO raw_events_second
(user_id)
SELECT user_id
FROM raw_events_first
WHERE user_id IN (SELECT value_2
FROM raw_events_second);
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
DEBUG: predicate pruning for shardId 13300003
ERROR: cannot perform distributed planning for the given modification
DETAIL: Select query cannot be pushed down to the worker.
-- we currently not support grouping sets
INSERT INTO agg_events
(user_id,

197
src/test/regress/sql/multi_insert_select.sql
View File

@ -496,7 +496,194 @@ FROM
((SELECT user_id FROM raw_events_first WHERE user_id = 15) EXCEPT
(SELECT user_id FROM raw_events_second where user_id = 17)) as foo;
-- unsupported JOIN
-- some supported LEFT joins
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id;
INSERT INTO agg_events (user_id)
SELECT
raw_events_second.user_id
FROM
reference_table LEFT JOIN raw_events_second ON reference_table.user_id = raw_events_second.user_id;
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id = 10;
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_second.user_id = 10 OR raw_events_second.user_id = 11;
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id = 10 AND raw_events_first.user_id = 20;
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id = 10 AND raw_events_second.user_id = 20;
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_first.user_id IN (19, 20, 21);
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.user_id
WHERE raw_events_second.user_id IN (19, 20, 21);
-- the following is a very tricky query for Citus
-- although we do not support pushing down JOINs on non-partition
-- columns here it is safe to push it down given that we're looking for
-- a specific value (i.e., value_1 = 12) on the joining column.
-- Note that the query always hits the same shard on raw_events_second
INSERT INTO agg_events
(user_id)
SELECT raw_events_first.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_second.user_id = raw_events_first.value_1
AND raw_events_first.value_1 = 12;
-- some unsupported LEFT/INNER JOINs
-- JOIN on one table with partition column other is not
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1;
-- same as the above with INNER JOIN
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1;
-- a not meaningful query
INSERT INTO agg_events
(user_id)
SELECT raw_events_second.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_first.user_id = raw_events_first.value_1;
-- both tables joined on non-partition columns
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.value_1 = raw_events_second.value_1;
-- same as the above with INNER JOIN
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1;
-- although we do not support pushing down JOINs on non-partition
-- columns here it is safe to push it down given that we're looking for
-- a specific value (i.e., user_id = 10) on the joining column.
-- Note that the query always hits the same shard on raw_events_second
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first LEFT JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1
WHERE
raw_events_first.user_id = 10;
-- same as the above with INNER JOIN
-- however this time query is not pushed down
-- to the worker. This is related to how we process
-- restriction infos, which we're considering to
-- improve
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1
WHERE raw_events_first.user_id = 10;
-- make things a bit more complicate with IN clauses
INSERT INTO agg_events (user_id)
SELECT
raw_events_first.user_id
FROM
raw_events_first INNER JOIN raw_events_second ON raw_events_first.user_id = raw_events_second.value_1
WHERE raw_events_first.value_1 IN (10, 11,12) OR raw_events_second.user_id IN (1,2,3,4);
-- implicit join on non partition column should also not be pushed down
INSERT INTO agg_events
(user_id)
SELECT raw_events_first.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_second.user_id = raw_events_first.value_1;
-- the following is again a very tricky query for Citus
-- if the given filter was on value_1 as shown in the above, Citus could
-- push it down. But here the query is refused
INSERT INTO agg_events
(user_id)
SELECT raw_events_first.user_id
FROM raw_events_first,
raw_events_second
WHERE raw_events_second.user_id = raw_events_first.value_1
AND raw_events_first.value_2 = 12;
-- foo is not joined on the partition key so the query is not
-- pushed down
INSERT INTO agg_events
(user_id, value_4_agg)
SELECT
outer_most.id, max(outer_most.value)
FROM
(
SELECT f2.id as id, f2.v4 as value FROM
(SELECT
id
FROM (SELECT reference_table.user_id AS id
FROM raw_events_first LEFT JOIN
reference_table
ON (raw_events_first.value_1 = reference_table.user_id)) AS foo) as f
INNER JOIN
(SELECT v4,
v1,
id
FROM (SELECT SUM(raw_events_second.value_4) AS v4,
SUM(raw_events_first.value_1) AS v1,
raw_events_second.user_id AS id
FROM raw_events_first,
raw_events_second
WHERE raw_events_first.user_id = raw_events_second.user_id
GROUP BY raw_events_second.user_id
HAVING SUM(raw_events_second.value_4) > 10) AS foo2 ) as f2
ON (f.id = f2.id)) as outer_most
GROUP BY
outer_most.id;
-- not equals on the partition column cannot be pushed down
INSERT INTO agg_events
(value_4_agg,
value_1_agg,
@ -672,6 +859,14 @@ outer_most.id, max(outer_most.value)
ON (f.id != f2.id)) as outer_most
GROUP BY outer_most.id;
-- cannot pushdown since subquery returns another column than partition key
INSERT INTO raw_events_second
(user_id)
SELECT user_id
FROM raw_events_first
WHERE user_id IN (SELECT value_2
FROM raw_events_second);
-- we currently not support grouping sets
INSERT INTO agg_events
(user_id,