Fix bug on pushing down wrong queries for INSERT .. SELECT

This commit fixes a bug that allows pushing down JOINs on
non-partition column.

We implement this change with the followings:

 * Consider joininfo along with the baserestrictinfo
   in relOptInfo
 * Only instantiate a qual if it includes the partition
   column of that relation
pull/1207/head
Onder Kalaci 2017-01-26 17:08:44 +02:00
parent e6e5f63d9d
commit aa5d56aaab
3 changed files with 822 additions and 9 deletions

View File

@ -48,6 +48,8 @@
#include "nodes/pg_list.h"
#include "nodes/primnodes.h"
#include "optimizer/clauses.h"
#include "optimizer/paths.h"
#include "optimizer/pathnode.h"
#include "optimizer/predtest.h"
#include "optimizer/restrictinfo.h"
#include "optimizer/var.h"
@ -72,6 +74,13 @@ typedef struct WalkerState
bool badCoalesce;
} WalkerState;
typedef struct InstantiateQualWalker
{
ShardInterval *targetShardInterval;
Var *relationPartitionColumn;
}InstantiateQualWalker;
bool EnableRouterExecution = true;
/* planner functions forward declarations */
@ -87,6 +96,8 @@ static Task * RouterModifyTaskForShardInterval(Query *originalQuery,
RelationRestrictionContext *
restrictionContext,
uint32 taskIdIndex);
static List * GetAllActualClausesForRelation(RelOptInfo *relInfo,
PlannerInfo *plannerInfo);
static bool MasterIrreducibleExpression(Node *expression, bool *varArgument,
bool *badCoalesce);
static bool MasterIrreducibleExpressionWalker(Node *expression, WalkerState *state);
@ -377,6 +388,10 @@ RouterModifyTaskForShardInterval(Query *originalQuery, ShardInterval *shardInter
{
RelationRestriction *restriction = lfirst(restrictionCell);
List *originalBaserestrictInfo = restriction->relOptInfo->baserestrictinfo;
List *originalJoinInfo = restriction->relOptInfo->joininfo;
InstantiateQualWalker *instantiateQualWalker = palloc0(
sizeof(InstantiateQualWalker));
Var *relationPartitionKey = PartitionKey(restriction->relationId);
/*
* We haven't added the quals if all participating tables are reference
@ -387,9 +402,15 @@ RouterModifyTaskForShardInterval(Query *originalQuery, ShardInterval *shardInter
break;
}
instantiateQualWalker->relationPartitionColumn = relationPartitionKey;
instantiateQualWalker->targetShardInterval = shardInterval;
originalBaserestrictInfo =
(List *) InstantiatePartitionQual((Node *) originalBaserestrictInfo,
shardInterval);
instantiateQualWalker);
originalJoinInfo =
(List *) InstantiatePartitionQual((Node *) originalJoinInfo,
instantiateQualWalker);
}
/*
@ -454,7 +475,6 @@ RouterModifyTaskForShardInterval(Query *originalQuery, ShardInterval *shardInter
"for shard %ld", shardId)));
}
/* this is required for correct deparsing of the query */
ReorderInsertSelectTargetLists(copiedQuery, copiedInsertRte, copiedSubqueryRte);
@ -488,6 +508,88 @@ RouterModifyTaskForShardInterval(Query *originalQuery, ShardInterval *shardInter
}
/*
* GetAllActualClausesForRelation returns all the restrictions that a relation
* has including the baserestrictinfo and the related parts of joininfo.
*
* This function is heavily inspired by check_index_predicates() from the PostgreSQL's
* source file src/backend/optimizer/path/indexpath.c.
*/
static List *
GetAllActualClausesForRelation(RelOptInfo *relInfo, PlannerInfo *plannerInfo)
{
List *allRestrictInfos = NULL;
ListCell *joinInfoCell = NULL;
Relids otherRelIds = NULL;
List *allActualClauses = NIL;
/* we should definitely add all the entries in baserestrictinfo */
allRestrictInfos = relInfo->baserestrictinfo;
/* Scan the rel's join clauses and get the necessary ones */
foreach(joinInfoCell, relInfo->joininfo)
{
RestrictInfo *restrictInfo = (RestrictInfo *) lfirst(joinInfoCell);
List *columnList = pull_var_clause_default((Node *) restrictInfo->clause);
/* check if the restrict info already contains a hashed restrict info */
if (list_length(columnList) > 0)
{
Var *firstColumn = linitial(columnList);
Var *hashedColumn = MakeInt4Column();
if (equal(firstColumn, hashedColumn))
{
/*
* This means that we've already instantiated the restrict info
* so this should be included in the final restrict info list.
*/
allRestrictInfos = lappend(allRestrictInfos, restrictInfo);
continue;
}
}
/* check if clause can be moved to this rel */
if (join_clause_is_movable_to(restrictInfo, relInfo))
{
allRestrictInfos = lappend(allRestrictInfos, restrictInfo);
}
}
/*
* Add on any equivalence-derivable join clauses. Computing the correct
* relid sets for generate_join_implied_equalities is slightly tricky
* because the rel could be a child rel rather than a true baserel, and in
* that case we must remove its parents' relid(s) from all_baserels.
*/
if (relInfo->reloptkind == RELOPT_OTHER_MEMBER_REL)
{
otherRelIds = bms_difference(plannerInfo->all_baserels,
find_childrel_parents(plannerInfo, relInfo));
}
else
{
otherRelIds = bms_difference(plannerInfo->all_baserels, relInfo->relids);
}
if (!bms_is_empty(otherRelIds))
{
List *impliedJoinEqualities = generate_join_implied_equalities(plannerInfo,
bms_union(
relInfo->relids,
otherRelIds),
otherRelIds,
relInfo);
allRestrictInfos = list_concat(allRestrictInfos, impliedJoinEqualities);
}
/* finally get the actual clauses from the restrict infos */
allActualClauses = get_all_actual_clauses(allRestrictInfos);
return allActualClauses;
}
/*
* AddShardIntervalRestrictionToSelect adds the following range boundaries
* with the given subquery and shardInterval:
@ -2395,8 +2497,6 @@ TargetShardIntervalsForSelect(Query *query,
Index tableId = relationRestriction->index;
DistTableCacheEntry *cacheEntry = DistributedTableCacheEntry(relationId);
int shardCount = cacheEntry->shardIntervalArrayLength;
List *baseRestrictionList = relationRestriction->relOptInfo->baserestrictinfo;
List *restrictClauseList = get_all_actual_clauses(baseRestrictionList);
List *prunedShardList = NIL;
int shardIndex = 0;
List *joinInfoList = relationRestriction->relOptInfo->joininfo;
@ -2415,6 +2515,7 @@ TargetShardIntervalsForSelect(Query *query,
if (!whereFalseQuery && shardCount > 0)
{
List *shardIntervalList = NIL;
List *allActualClauses = NIL;
for (shardIndex = 0; shardIndex < shardCount; shardIndex++)
{
@ -2423,8 +2524,12 @@ TargetShardIntervalsForSelect(Query *query,
shardIntervalList = lappend(shardIntervalList, shardInterval);
}
allActualClauses =
GetAllActualClausesForRelation(relationRestriction->relOptInfo,
relationRestriction->plannerInfo);
prunedShardList = PruneShardList(relationId, tableId,
restrictClauseList,
allActualClauses,
shardIntervalList);
/*
@ -2950,7 +3055,10 @@ CopyRelationRestrictionContext(RelationRestrictionContext *oldContext)
static Node *
InstantiatePartitionQual(Node *node, void *context)
{
ShardInterval *shardInterval = (ShardInterval *) context;
InstantiateQualWalker *instantiateContext = ((InstantiateQualWalker *) context);
ShardInterval *shardInterval = instantiateContext->targetShardInterval;
Var *relationPartitionColumn = instantiateContext->relationPartitionColumn;
Assert(shardInterval->minValueExists);
Assert(shardInterval->maxValueExists);
@ -2974,6 +3082,7 @@ 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 +3101,20 @@ InstantiatePartitionQual(Node *node, void *context)
if (IsA(leftop, Param))
{
param = (Param *) leftop;
if (IsA(rightop, Var))
{
currentColumn = (Var *) rightop;
}
}
else if (IsA(rightop, Param))
{
param = (Param *) rightop;
if (IsA(leftop, Var))
{
currentColumn = (Var *) leftop;
}
}
/* not an interesting param for our purpose, so return */
@ -3004,6 +3123,13 @@ 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,

View File

@ -719,24 +719,30 @@ DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid:
DEBUG: predicate pruning for shardId 13300000
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.raw_events_second_13300004 AS citus_table_alias (user_id) SELECT user_id FROM public.raw_events_first_13300003 raw_events_first WHERE ((user_id IN (SELECT raw_events_second.user_id FROM public.raw_events_second_13300007 raw_events_second WHERE (raw_events_second.user_id = 2))) AND ((hashint4(user_id) >= '-2147483648'::integer) AND (hashint4(user_id) <= '-1073741825'::integer)))
DEBUG: predicate pruning for shardId 13300007
DEBUG: Skipping target shard interval 13300004 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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.raw_events_second_13300005 AS citus_table_alias (user_id) SELECT user_id FROM public.raw_events_first_13300003 raw_events_first WHERE ((user_id IN (SELECT raw_events_second.user_id FROM public.raw_events_second_13300007 raw_events_second WHERE (raw_events_second.user_id = 2))) AND ((hashint4(user_id) >= '-1073741824'::integer) AND (hashint4(user_id) <= '-1'::integer)))
DEBUG: predicate pruning for shardId 13300007
DEBUG: Skipping target shard interval 13300005 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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.raw_events_second_13300006 AS citus_table_alias (user_id) SELECT user_id FROM public.raw_events_first_13300003 raw_events_first WHERE ((user_id IN (SELECT raw_events_second.user_id FROM public.raw_events_second_13300007 raw_events_second WHERE (raw_events_second.user_id = 2))) AND ((hashint4(user_id) >= 0) AND (hashint4(user_id) <= 1073741823)))
DEBUG: predicate pruning for shardId 13300007
DEBUG: Skipping target shard interval 13300006 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
DEBUG: predicate pruning for shardId 13300001
DEBUG: predicate pruning for shardId 13300002
@ -1175,6 +1181,537 @@ DEBUG: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
ERROR: set operations are not allowed in INSERT ... SELECT queries
-- 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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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 13300000
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: Skipping target shard interval 13300009 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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: Skipping target shard interval 13300010 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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: Skipping target shard interval 13300011 since SELECT query for it pruned away
DEBUG: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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;
DEBUG: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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) 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 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: Skipping target shard interval 13300009 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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: Skipping target shard interval 13300010 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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: Skipping target shard interval 13300011 since SELECT query for it pruned away
DEBUG: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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 IN (19, 20, 21);
DEBUG: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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 = 10) AND (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 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: Skipping target shard interval 13300009 since SELECT query for it pruned away
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 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: Skipping target shard interval 13300010 since SELECT query for it pruned away
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 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: Skipping target shard interval 13300011 since SELECT query for it pruned away
DEBUG: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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_first.user_id = 10 AND raw_events_second.user_id IN (19, 20, 21);
DEBUG: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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_first.user_id = 10) AND (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 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 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: Skipping target shard interval 13300009 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: Skipping target shard interval 13300010 since SELECT query for it pruned away
DEBUG: predicate pruning for shardId 13300000
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 13300004
DEBUG: predicate pruning for shardId 13300005
DEBUG: predicate pruning for shardId 13300006
DEBUG: predicate pruning for shardId 13300007
DEBUG: Skipping target shard interval 13300011 since SELECT query for it pruned away
DEBUG: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
-- 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.
-- Since that shard is on both nodes the query can be pushed down
-- but for instance if there were 3 worker nodes the query would fail
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
-- 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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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;
DEBUG: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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.
-- JOIN on one table with partition column other is not
-- also an equality qual is added on the partition key
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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 13300000
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 ((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_first(user_id, "time", value_1, value_2, value_3, value_4) 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 13300000
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 ((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_first(user_id, "time", value_1, value_2, value_3, value_4) 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 13300000
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 ((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_first(user_id, "time", value_1, value_2, value_3, value_4) 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: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
-- 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
WHERE raw_events_first.user_id = 10;
DEBUG: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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 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 13300000
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 ((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_first(user_id, "time", value_1, value_2, value_3, value_4) 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 13300000
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 ((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_first(user_id, "time", value_1, value_2, value_3, value_4) 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 13300000
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 ((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_first(user_id, "time", value_1, value_2, value_3, value_4) 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: ProcessQuery
DEBUG: Plan is router executable
DEBUG: CommitTransactionCommand
DEBUG: CommitTransaction
DEBUG: name: unnamed; blockState: STARTED; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
-- 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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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.
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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: StartTransactionCommand
DEBUG: StartTransaction
DEBUG: name: unnamed; blockState: DEFAULT; state: INPROGR, xid/subid/cid: 0/1/0, nestlvl: 1, children:
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.
-- unsupported JOIN
INSERT INTO agg_events
(value_4_agg,
@ -2073,3 +2610,10 @@ DROP TABLE table_with_serial;
DROP TABLE text_table;
DROP TABLE char_table;
DROP TABLE table_with_starts_with_defaults;
-- clean any 2PC transactions we've used
SELECT recover_prepared_transactions();
recover_prepared_transactions
-------------------------------
0
(1 row)

View File

@ -496,6 +496,146 @@ 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;
-- 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_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;
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_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_first.user_id = 10 AND 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.
-- Since that shard is on both nodes the query can be pushed down
-- but for instance if there were 3 worker nodes the query would fail
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;
-- JOIN on one table with partition column other is not
-- also an equality qual is added on the partition key
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
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;
-- unsupported JOIN
INSERT INTO agg_events
(value_4_agg,
@ -979,3 +1119,6 @@ DROP TABLE table_with_serial;
DROP TABLE text_table;
DROP TABLE char_table;
DROP TABLE table_with_starts_with_defaults;
-- clean any 2PC transactions we've used
SELECT recover_prepared_transactions();