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

Drop postgres 11 support

pull/4799/head
SaitTalhaNisanci 2021-03-25 11:29:12 +03:00 committed by Sait Talha Nisanci
parent 7081690480
commit 03832f353c
149 changed files with 1135 additions and 17423 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

73
.circleci/config.yml
View File

@ -199,7 +199,7 @@ jobs:
make -C src/test/regress \
check-citus-upgrade \
bindir=/usr/lib/postgresql/${PG_MAJOR}/bin \
citus-pre-tar=/install-pg11-citus${citus_version}.tar \
citus-pre-tar=/install-pg${PG_MAJOR}-citus${citus_version}.tar \
citus-post-tar=/home/circleci/project/install-$PG_MAJOR.tar; \
done;
@ -210,7 +210,7 @@ jobs:
make -C src/test/regress \
check-citus-upgrade-mixed \
bindir=/usr/lib/postgresql/${PG_MAJOR}/bin \
citus-pre-tar=/install-pg11-citus${citus_version}.tar \
citus-pre-tar=/install-pg${PG_MAJOR}-citus${citus_version}.tar \
citus-post-tar=/home/circleci/project/install-$PG_MAJOR.tar; \
done;
no_output_timeout: 2m
@ -426,10 +426,6 @@ workflows:
ignore:
- /release-[0-9]+\.[0-9]+.*/ # match with releaseX.Y.*
- build:
name: build-11
pg_major: 11
image_tag: '11.9'
- build:
name: build-12
pg_major: 12
@ -442,56 +438,6 @@ workflows:
- check-style
- check-sql-snapshots
- test-citus:
name: 'test-11_check-multi'
pg_major: 11
image_tag: '11.9'
make: check-multi
requires: [build-11]
- test-citus:
name: 'test-11_check-mx'
pg_major: 11
image_tag: '11.9'
make: check-multi-mx
requires: [build-11]
- test-citus:
name: 'test-11_check-vanilla'
pg_major: 11
image_tag: '11.9'
make: check-vanilla
requires: [build-11]
- test-citus:
name: 'test-11_check-isolation'
pg_major: 11
image_tag: '11.9'
make: check-isolation
requires: [build-11]
- test-citus:
name: 'test-11_check-worker'
pg_major: 11
image_tag: '11.9'
make: check-worker
requires: [build-11]
- test-citus:
name: 'test-11_check-operations'
pg_major: 11
image_tag: '11.9'
make: check-operations
requires: [build-11]
- test-citus:
name: 'test-11_check-follower-cluster'
pg_major: 11
image_tag: '11.9'
make: check-follower-cluster
requires: [build-11]
- test-citus:
name: 'test-11_check-failure'
pg_major: 11
image: citus/failtester
image_tag: '11.9'
make: check-failure
requires: [build-11]
- test-citus:
name: 'test-12_check-multi'
pg_major: 12
@ -628,13 +574,6 @@ workflows:
make: check-failure
requires: [build-13]
- test-pg-upgrade:
name: 'test-11-12_check-pg-upgrade'
old_pg_major: 11
new_pg_major: 12
image_tag: 11-12-13
requires: [build-11,build-12]
- test-pg-upgrade:
name: 'test-12-13_check-pg-upgrade'
old_pg_major: 12
@ -643,10 +582,10 @@ workflows:
requires: [build-12,build-13]
- test-citus-upgrade:
name: test-11_check-citus-upgrade
pg_major: 11
image_tag: '11.9'
requires: [build-11]
name: test-12_check-citus-upgrade
pg_major: 12
image_tag: '12.4'
requires: [build-12]
- ch_benchmark:
requires: [build-13]

11
configure vendored
View File

@ -2556,7 +2556,7 @@ if test -z "$version_num"; then
as_fn_error $? "Could not detect PostgreSQL version from pg_config." "$LINENO" 5
fi
if test "$version_num" != '11' -a "$version_num" != '12' -a "$version_num" != '13'; then
if test "$version_num" != '12' -a "$version_num" != '13'; then
as_fn_error $? "Citus is not compatible with the detected PostgreSQL version ${version_num}." "$LINENO" 5
else
{ $as_echo "$as_me:${as_lineno-$LINENO}: building against PostgreSQL $version_num" >&5
@ -4533,16 +4533,9 @@ cat >>confdefs.h <<_ACEOF
_ACEOF
if test "$version_num" != '11'; then
HAS_TABLEAM=yes
HAS_TABLEAM=yes
$as_echo "#define HAS_TABLEAM 1" >>confdefs.h
else
{ $as_echo "$as_me:${as_lineno-$LINENO}: postgres version does not support table access methods" >&5
$as_echo "$as_me: postgres version does not support table access methods" >&6;}
fi;
# Require lz4 & zstd only if we are compiling columnar
if test "$HAS_TABLEAM" == 'yes'; then
#

10
configure.in
View File

@ -74,7 +74,7 @@ if test -z "$version_num"; then
AC_MSG_ERROR([Could not detect PostgreSQL version from pg_config.])
fi
if test "$version_num" != '11' -a "$version_num" != '12' -a "$version_num" != '13'; then
if test "$version_num" != '12' -a "$version_num" != '13'; then
AC_MSG_ERROR([Citus is not compatible with the detected PostgreSQL version ${version_num}.])
else
AC_MSG_NOTICE([building against PostgreSQL $version_num])
@ -216,12 +216,8 @@ PGAC_ARG_REQ(with, reports-hostname, [HOSTNAME],
AC_DEFINE_UNQUOTED(REPORTS_BASE_URL, "$REPORTS_BASE_URL",
[Base URL for statistics collection and update checks])
if test "$version_num" != '11'; then
HAS_TABLEAM=yes
AC_DEFINE([HAS_TABLEAM], 1, [Define to 1 to build with table access method support, pg12 and up])
else
AC_MSG_NOTICE([postgres version does not support table access methods])
fi;
HAS_TABLEAM=yes
AC_DEFINE([HAS_TABLEAM], 1, [Define to 1 to build with table access method support, pg12 and up])
# Require lz4 & zstd only if we are compiling columnar
if test "$HAS_TABLEAM" == 'yes'; then

20
src/backend/columnar/columnar_compression.c
View File

@ -226,10 +226,8 @@ DecompressBuffer(StringInfo buffer,
case COMPRESSION_PG_LZ:
{
StringInfo decompressedBuffer = NULL;
uint32 compressedDataSize = VARSIZE(buffer->data) - COLUMNAR_COMPRESS_HDRSZ;
uint32 decompressedDataSize = COLUMNAR_COMPRESS_RAWSIZE(buffer->data);
int32 decompressedByteCount = 0;
if (compressedDataSize + COLUMNAR_COMPRESS_HDRSZ != buffer->len)
{
@ -240,17 +238,11 @@ DecompressBuffer(StringInfo buffer,
char *decompressedData = palloc0(decompressedDataSize);
#if PG_VERSION_NUM >= 120000
decompressedByteCount = pglz_decompress(COLUMNAR_COMPRESS_RAWDATA(
buffer->data),
compressedDataSize, decompressedData,
decompressedDataSize, true);
#else
decompressedByteCount = pglz_decompress(COLUMNAR_COMPRESS_RAWDATA(
buffer->data),
compressedDataSize, decompressedData,
decompressedDataSize);
#endif
int32 decompressedByteCount = pglz_decompress(COLUMNAR_COMPRESS_RAWDATA(
buffer->data),
compressedDataSize,
decompressedData,
decompressedDataSize, true);
if (decompressedByteCount < 0)
{
@ -258,7 +250,7 @@ DecompressBuffer(StringInfo buffer,
errdetail("compressed data is corrupted")));
}
decompressedBuffer = palloc0(sizeof(StringInfoData));
StringInfo decompressedBuffer = palloc0(sizeof(StringInfoData));
decompressedBuffer->data = decompressedData;
decompressedBuffer->len = decompressedDataSize;
decompressedBuffer->maxlen = decompressedDataSize;

7
src/backend/columnar/columnar_debug.c
View File

@ -39,14 +39,9 @@ PG_FUNCTION_INFO_V1(column_store_memory_stats);
Datum
column_store_memory_stats(PG_FUNCTION_ARGS)
{
TupleDesc tupleDescriptor = NULL;
const int resultColumnCount = 3;
#if PG_VERSION_NUM >= PG_VERSION_12
tupleDescriptor = CreateTemplateTupleDesc(resultColumnCount);
#else
tupleDescriptor = CreateTemplateTupleDesc(resultColumnCount, false);
#endif
TupleDesc tupleDescriptor = CreateTemplateTupleDesc(resultColumnCount);
TupleDescInitEntry(tupleDescriptor, (AttrNumber) 1, "TopMemoryContext",
INT8OID, -1, 0);

20
src/backend/columnar/columnar_metadata.c
View File

@ -1067,15 +1067,10 @@ InsertTupleAndEnforceConstraints(ModifyState *state, Datum *values, bool *nulls)
TupleDesc tupleDescriptor = RelationGetDescr(state->rel);
HeapTuple tuple = heap_form_tuple(tupleDescriptor, values, nulls);
#if PG_VERSION_NUM >= 120000
TupleTableSlot *slot = ExecInitExtraTupleSlot(state->estate, tupleDescriptor,
&TTSOpsHeapTuple);
ExecStoreHeapTuple(tuple, slot, false);
#else
TupleTableSlot *slot = ExecInitExtraTupleSlot(state->estate, tupleDescriptor);
ExecStoreTuple(tuple, slot, InvalidBuffer, false);
#endif
/* use ExecSimpleRelationInsert to enforce constraints */
ExecSimpleRelationInsert(state->estate, slot);
@ -1127,20 +1122,16 @@ FinishModifyRelation(ModifyState *state)
static EState *
create_estate_for_relation(Relation rel)
{
ResultRelInfo *resultRelInfo;
EState *estate = CreateExecutorState();
RangeTblEntry *rte = makeNode(RangeTblEntry);
rte->rtekind = RTE_RELATION;
rte->relid = RelationGetRelid(rel);
rte->relkind = rel->rd_rel->relkind;
#if PG_VERSION_NUM >= 120000
rte->rellockmode = AccessShareLock;
ExecInitRangeTable(estate, list_make1(rte));
#endif
resultRelInfo = makeNode(ResultRelInfo);
ResultRelInfo *resultRelInfo = makeNode(ResultRelInfo);
InitResultRelInfo(resultRelInfo, rel, 1, NULL, 0);
estate->es_result_relations = resultRelInfo;
@ -1149,15 +1140,6 @@ create_estate_for_relation(Relation rel)
estate->es_output_cid = GetCurrentCommandId(true);
#if PG_VERSION_NUM < 120000
/* Triggers might need a slot */
if (resultRelInfo->ri_TrigDesc)
{
estate->es_trig_tuple_slot = ExecInitExtraTupleSlot(estate, NULL);
}
#endif
/* Prepare to catch AFTER triggers. */
AfterTriggerBeginQuery();

5
src/backend/columnar/columnar_reader.c
View File

@ -23,13 +23,10 @@
#include "commands/defrem.h"
#include "distributed/listutils.h"
#include "nodes/makefuncs.h"
#if PG_VERSION_NUM >= 120000
#include "nodes/nodeFuncs.h"
#include "optimizer/optimizer.h"
#else
#include "optimizer/clauses.h"
#include "optimizer/predtest.h"
#endif
#include "optimizer/restrictinfo.h"
#include "storage/fd.h"
#include "utils/guc.h"
#include "utils/memutils.h"

9
src/backend/distributed/commands/alter_table.c
View File

@ -435,11 +435,6 @@ AlterDistributedTable(TableConversionParameters *params)
TableConversionReturn *
AlterTableSetAccessMethod(TableConversionParameters *params)
{
#if PG_VERSION_NUM < PG_VERSION_12
ereport(ERROR, (errmsg("table access methods are not supported "
"for Postgres versions earlier than 12")));
#endif
EnsureRelationExists(params->relationId);
EnsureTableOwner(params->relationId);
@ -963,7 +958,6 @@ CreateTableConversion(TableConversionParameters *params)
BuildDistributionKeyFromColumnName(relation, con->distributionColumn);
con->originalAccessMethod = NULL;
#if PG_VERSION_NUM >= PG_VERSION_12
if (!PartitionedTable(con->relationId))
{
HeapTuple amTuple = SearchSysCache1(AMOID, ObjectIdGetDatum(
@ -977,7 +971,6 @@ CreateTableConversion(TableConversionParameters *params)
con->originalAccessMethod = NameStr(amForm->amname);
ReleaseSysCache(amTuple);
}
#endif
con->colocatedTableList = NIL;
@ -1296,12 +1289,10 @@ GetNonGeneratedStoredColumnNameList(Oid relationId)
continue;
}
#if PG_VERSION_NUM >= 120000
if (currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED)
{
continue;
}
#endif
const char *quotedColumnName = quote_identifier(NameStr(currentColumn->attname));
nonStoredColumnNameList = lappend(nonStoredColumnNameList,

3
src/backend/distributed/commands/cascade_table_operation_for_connected_relations.c
View File

@ -14,9 +14,6 @@
#include "distributed/pg_version_constants.h"
#if (PG_VERSION_NUM < PG_VERSION_12)
#include "access/htup_details.h"
#endif
#include "access/xact.h"
#include "catalog/pg_constraint.h"
#include "distributed/commands/utility_hook.h"

36
src/backend/distributed/commands/collation.c
View File

@ -50,35 +50,21 @@ static void EnsureSequentialModeForCollationDDL(void);
static char *
CreateCollationDDLInternal(Oid collationId, Oid *collowner, char **quotedCollationName)
{
char *schemaName = NULL;
StringInfoData collationNameDef;
const char *providerString = NULL;
HeapTuple heapTuple = NULL;
Form_pg_collation collationForm = NULL;
char collprovider;
const char *collcollate;
const char *collctype;
const char *collname;
Oid collnamespace;
#if PG_VERSION_NUM >= PG_VERSION_12
bool collisdeterministic;
#endif
heapTuple = SearchSysCache1(COLLOID, ObjectIdGetDatum(collationId));
HeapTuple heapTuple = SearchSysCache1(COLLOID, ObjectIdGetDatum(collationId));
if (!HeapTupleIsValid(heapTuple))
{
elog(ERROR, "citus cache lookup failed for collation %u", collationId);
}
collationForm = (Form_pg_collation) GETSTRUCT(heapTuple);
collprovider = collationForm->collprovider;
collcollate = NameStr(collationForm->collcollate);
collctype = NameStr(collationForm->collctype);
collnamespace = collationForm->collnamespace;
collname = NameStr(collationForm->collname);
#if PG_VERSION_NUM >= PG_VERSION_12
collisdeterministic = collationForm->collisdeterministic;
#endif
Form_pg_collation collationForm = (Form_pg_collation) GETSTRUCT(heapTuple);
char collprovider = collationForm->collprovider;
const char *collcollate = NameStr(collationForm->collcollate);
const char *collctype = NameStr(collationForm->collctype);
Oid collnamespace = collationForm->collnamespace;
const char *collname = NameStr(collationForm->collname);
bool collisdeterministic = collationForm->collisdeterministic;
if (collowner != NULL)
{
@ -86,9 +72,9 @@ CreateCollationDDLInternal(Oid collationId, Oid *collowner, char **quotedCollati
}
ReleaseSysCache(heapTuple);
schemaName = get_namespace_name(collnamespace);
char *schemaName = get_namespace_name(collnamespace);
*quotedCollationName = quote_qualified_identifier(schemaName, collname);
providerString =
const char *providerString =
collprovider == COLLPROVIDER_DEFAULT ? "default" :
collprovider == COLLPROVIDER_ICU ? "icu" :
collprovider == COLLPROVIDER_LIBC ? "libc" : NULL;
@ -117,12 +103,10 @@ CreateCollationDDLInternal(Oid collationId, Oid *collowner, char **quotedCollati
quote_literal_cstr(collctype));
}
#if PG_VERSION_NUM >= PG_VERSION_12
if (!collisdeterministic)
{
appendStringInfoString(&collationNameDef, ", deterministic = false");
}
#endif
appendStringInfoChar(&collationNameDef, ')');

32
src/backend/distributed/commands/create_distributed_table.c
View File

@ -29,9 +29,7 @@
#include "catalog/pg_extension.h"
#include "catalog/pg_namespace.h"
#include "catalog/pg_opclass.h"
#if PG_VERSION_NUM >= 12000
#include "catalog/pg_proc.h"
#endif
#include "catalog/pg_trigger.h"
#include "commands/defrem.h"
#include "commands/extension.h"
@ -819,18 +817,6 @@ EnsureRelationCanBeDistributed(Oid relationId, Var *distributionColumn,
ErrorIfTableIsACatalogTable(relation);
#if PG_VERSION_NUM < PG_VERSION_12
/* verify target relation does not use WITH (OIDS) PostgreSQL feature */
if (relationDesc->tdhasoid)
{
ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("cannot distribute relation: %s", relationName),
errdetail("Distributed relations must not specify the WITH "
"(OIDS) option in their definitions.")));
}
#endif
/* verify target relation does not use identity columns */
if (RelationUsesIdentityColumns(relationDesc))
{
@ -866,7 +852,6 @@ EnsureRelationCanBeDistributed(Oid relationId, Var *distributionColumn,
"defined to use hash partitioning.")));
}
#if PG_VERSION_NUM >= PG_VERSION_12
if (distributionColumn->varcollid != InvalidOid &&
!get_collation_isdeterministic(distributionColumn->varcollid))
{
@ -874,7 +859,6 @@ EnsureRelationCanBeDistributed(Oid relationId, Var *distributionColumn,
errmsg("Hash distributed partition columns may not use "
"a non deterministic collation")));
}
#endif
}
else if (distributionMethod == DISTRIBUTE_BY_RANGE)
{
@ -1539,12 +1523,8 @@ DoCopyFromLocalTableIntoShards(Relation distributedRelation,
EState *estate)
{
/* begin reading from local table */
#if PG_VERSION_NUM >= PG_VERSION_12
TableScanDesc scan = table_beginscan(distributedRelation, GetActiveSnapshot(), 0,
NULL);
#else
HeapScanDesc scan = heap_beginscan(distributedRelation, GetActiveSnapshot(), 0, NULL);
#endif
MemoryContext oldContext = MemoryContextSwitchTo(GetPerTupleMemoryContext(estate));
@ -1593,11 +1573,7 @@ DoCopyFromLocalTableIntoShards(Relation distributedRelation,
MemoryContextSwitchTo(oldContext);
/* finish reading from the local table */
#if PG_VERSION_NUM >= PG_VERSION_12
table_endscan(scan);
#else
heap_endscan(scan);
#endif
}
@ -1615,10 +1591,8 @@ TupleDescColumnNameList(TupleDesc tupleDescriptor)
Form_pg_attribute currentColumn = TupleDescAttr(tupleDescriptor, columnIndex);
char *columnName = NameStr(currentColumn->attname);
if (currentColumn->attisdropped
#if PG_VERSION_NUM >= PG_VERSION_12
|| currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
#endif
if (currentColumn->attisdropped ||
currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
)
{
continue;
@ -1660,7 +1634,6 @@ static bool
DistributionColumnUsesGeneratedStoredColumn(TupleDesc relationDesc,
Var *distributionColumn)
{
#if PG_VERSION_NUM >= PG_VERSION_12
Form_pg_attribute attributeForm = TupleDescAttr(relationDesc,
distributionColumn->varattno - 1);
@ -1668,7 +1641,6 @@ DistributionColumnUsesGeneratedStoredColumn(TupleDesc relationDesc,
{
return true;
}
#endif
return false;
}

2
src/backend/distributed/commands/distribute_object_ops.c
View File

@ -62,7 +62,7 @@ static DistributeObjectOps Any_AlterEnum = {
.deparse = DeparseAlterEnumStmt,
.qualify = QualifyAlterEnumStmt,
.preprocess = PreprocessAlterEnumStmt,
.postprocess = PostprocessAlterEnumStmt,
.postprocess = NULL,
.address = AlterEnumStmtObjectAddress,
};
static DistributeObjectOps Any_AlterExtension = {

29
src/backend/distributed/commands/foreign_constraint.c
View File

@ -19,9 +19,7 @@
#include "access/xact.h"
#include "catalog/namespace.h"
#include "catalog/pg_constraint.h"
#if (PG_VERSION_NUM >= PG_VERSION_12)
#include "access/genam.h"
#endif
#include "catalog/pg_type.h"
#include "distributed/colocation_utils.h"
#include "distributed/commands.h"
@ -715,14 +713,9 @@ get_relation_constraint_oid_compat(HeapTuple heapTuple)
{
Assert(heapTuple != NULL);
Oid constraintOid = InvalidOid;
#if PG_VERSION_NUM >= PG_VERSION_12
Form_pg_constraint constraintForm = (Form_pg_constraint) GETSTRUCT(heapTuple);
constraintOid = constraintForm->oid;
#else
constraintOid = HeapTupleGetOid(heapTuple);
#endif
Oid constraintOid = constraintForm->oid;
return constraintOid;
}
@ -1284,8 +1277,6 @@ GetForeignConstraintCommandsToReferenceTable(ShardInterval *shardInterval)
static void
UpdateConstraintIsValid(Oid constraintId, bool isValid)
{
HeapTuple heapTuple = NULL;
SysScanDesc scanDescriptor;
ScanKeyData scankey[1];
Relation pgConstraint = table_open(ConstraintRelationId, AccessShareLock);
TupleDesc tupleDescriptor = RelationGetDescr(pgConstraint);
@ -1294,21 +1285,17 @@ UpdateConstraintIsValid(Oid constraintId, bool isValid)
bool replace[Natts_pg_constraint];
ScanKeyInit(&scankey[0],
#if PG_VERSION_NUM >= 120000
Anum_pg_constraint_oid,
#else
ObjectIdAttributeNumber,
#endif
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(constraintId));
scanDescriptor = systable_beginscan(pgConstraint,
ConstraintOidIndexId,
true,
NULL,
1,
scankey);
heapTuple = systable_getnext(scanDescriptor);
SysScanDesc scanDescriptor = systable_beginscan(pgConstraint,
ConstraintOidIndexId,
true,
NULL,
1,
scankey);
HeapTuple heapTuple = systable_getnext(scanDescriptor);
if (!HeapTupleIsValid(heapTuple))
{
elog(ERROR, "could not find tuple for constraint %u", constraintId);

22
src/backend/distributed/commands/function.c
View File

@ -21,9 +21,7 @@
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/genam.h"
#endif
#include "access/htup_details.h"
#include "access/xact.h"
#include "catalog/pg_aggregate.h"
@ -736,9 +734,6 @@ static char *
GetAggregateDDLCommand(const RegProcedure funcOid, bool useCreateOrReplace)
{
StringInfoData buf = { 0 };
HeapTuple aggtup = NULL;
Form_pg_aggregate agg = NULL;
int numargs = 0;
int i = 0;
Oid *argtypes = NULL;
char **argnames = NULL;
@ -762,7 +757,6 @@ GetAggregateDDLCommand(const RegProcedure funcOid, bool useCreateOrReplace)
const char *name = NameStr(proc->proname);
const char *nsp = get_namespace_name(proc->pronamespace);
#if PG_VERSION_NUM >= PG_VERSION_12
if (useCreateOrReplace)
{
appendStringInfo(&buf, "CREATE OR REPLACE AGGREGATE %s(",
@ -773,20 +767,16 @@ GetAggregateDDLCommand(const RegProcedure funcOid, bool useCreateOrReplace)
appendStringInfo(&buf, "CREATE AGGREGATE %s(",
quote_qualified_identifier(nsp, name));
}
#else
appendStringInfo(&buf, "CREATE AGGREGATE %s(",
quote_qualified_identifier(nsp, name));
#endif
/* Parameters, borrows heavily from print_function_arguments in postgres */
numargs = get_func_arg_info(proctup, &argtypes, &argnames, &argmodes);
int numargs = get_func_arg_info(proctup, &argtypes, &argnames, &argmodes);
aggtup = SearchSysCache1(AGGFNOID, funcOid);
HeapTuple aggtup = SearchSysCache1(AGGFNOID, funcOid);
if (!HeapTupleIsValid(aggtup))
{
elog(ERROR, "cache lookup failed for %d", funcOid);
}
agg = (Form_pg_aggregate) GETSTRUCT(aggtup);
Form_pg_aggregate agg = (Form_pg_aggregate) GETSTRUCT(aggtup);
if (AGGKIND_IS_ORDERED_SET(agg->aggkind))
{
@ -1066,12 +1056,6 @@ GetAggregateDDLCommand(const RegProcedure funcOid, bool useCreateOrReplace)
ReleaseSysCache(aggtup);
ReleaseSysCache(proctup);
#if PG_VERSION_NUM < PG_VERSION_12
if (useCreateOrReplace)
{
return WrapCreateOrReplace(buf.data);
}
#endif
return buf.data;
}

30
src/backend/distributed/commands/index.c
View File

@ -11,9 +11,7 @@
#include "postgres.h"
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/genam.h"
#endif
#include "access/htup_details.h"
#include "access/xact.h"
#include "catalog/catalog.h"
@ -93,9 +91,7 @@ struct DropRelationCallbackState
*/
struct ReindexIndexCallbackState
{
#if PG_VERSION_NUM >= PG_VERSION_12
bool concurrent;
#endif
Oid locked_table_oid;
};
@ -544,13 +540,8 @@ PreprocessReindexStmt(Node *node, const char *reindexCommand,
{
Relation relation = NULL;
Oid relationId = InvalidOid;
bool isCitusRelation = false;
#if PG_VERSION_NUM >= PG_VERSION_12
LOCKMODE lockmode = reindexStatement->concurrent ? ShareUpdateExclusiveLock :
AccessExclusiveLock;
#else
LOCKMODE lockmode = AccessExclusiveLock;
#endif
MemoryContext relationContext = NULL;
Assert(reindexStatement->kind == REINDEX_OBJECT_INDEX ||
@ -558,17 +549,14 @@ PreprocessReindexStmt(Node *node, const char *reindexCommand,
if (reindexStatement->kind == REINDEX_OBJECT_INDEX)
{
Oid indOid;
struct ReindexIndexCallbackState state;
#if PG_VERSION_NUM >= PG_VERSION_12
state.concurrent = reindexStatement->concurrent;
#endif
state.locked_table_oid = InvalidOid;
indOid = RangeVarGetRelidExtended(reindexStatement->relation,
lockmode, 0,
RangeVarCallbackForReindexIndex,
&state);
Oid indOid = RangeVarGetRelidExtended(reindexStatement->relation,
lockmode, 0,
RangeVarCallbackForReindexIndex,
&state);
relation = index_open(indOid, NoLock);
relationId = IndexGetRelation(indOid, false);
}
@ -581,7 +569,7 @@ PreprocessReindexStmt(Node *node, const char *reindexCommand,
relationId = RelationGetRelid(relation);
}
isCitusRelation = IsCitusTable(relationId);
bool isCitusRelation = IsCitusTable(relationId);
if (reindexStatement->relation->schemaname == NULL)
{
@ -613,12 +601,8 @@ PreprocessReindexStmt(Node *node, const char *reindexCommand,
{
DDLJob *ddlJob = palloc0(sizeof(DDLJob));
ddlJob->targetRelationId = relationId;
#if PG_VERSION_NUM >= PG_VERSION_12
ddlJob->concurrentIndexCmd = reindexStatement->concurrent;
ddlJob->startNewTransaction = reindexStatement->concurrent;
#else
ddlJob->concurrentIndexCmd = false;
#endif
ddlJob->commandString = reindexCommand;
ddlJob->taskList = CreateReindexTaskList(relationId, reindexStatement);
@ -1053,11 +1037,7 @@ RangeVarCallbackForReindexIndex(const RangeVar *relation, Oid relId, Oid oldRelI
* non-concurrent case and table locks used by index_concurrently_*() for
* concurrent case.
*/
#if PG_VERSION_NUM >= PG_VERSION_12
table_lockmode = state->concurrent ? ShareUpdateExclusiveLock : ShareLock;
#else
table_lockmode = ShareLock;
#endif
/*
* If we previously locked some other index's heap, and the name we're

86
src/backend/distributed/commands/multi_copy.c
View File

@ -423,21 +423,13 @@ CopyToExistingShards(CopyStmt *copyStatement, QueryCompletionCompat *completionT
{
Oid tableId = RangeVarGetRelid(copyStatement->relation, NoLock, false);
CitusCopyDestReceiver *copyDest = NULL;
DestReceiver *dest = NULL;
Relation copiedDistributedRelation = NULL;
Form_pg_class copiedDistributedRelationTuple = NULL;
List *columnNameList = NIL;
int partitionColumnIndex = INVALID_PARTITION_COLUMN_INDEX;
EState *executorState = NULL;
MemoryContext executorTupleContext = NULL;
ExprContext *executorExpressionContext = NULL;
bool stopOnFailure = false;
CopyState copyState = NULL;
uint64 processedRowCount = 0;
ErrorContextCallback errorCallback;
@ -469,10 +461,8 @@ CopyToExistingShards(CopyStmt *copyStatement, QueryCompletionCompat *completionT
Form_pg_attribute currentColumn = TupleDescAttr(tupleDescriptor, columnIndex);
char *columnName = NameStr(currentColumn->attname);
if (currentColumn->attisdropped
#if PG_VERSION_NUM >= PG_VERSION_12
|| currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
#endif
if (currentColumn->attisdropped ||
currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
)
{
continue;
@ -481,9 +471,9 @@ CopyToExistingShards(CopyStmt *copyStatement, QueryCompletionCompat *completionT
columnNameList = lappend(columnNameList, columnName);
}
executorState = CreateExecutorState();
executorTupleContext = GetPerTupleMemoryContext(executorState);
executorExpressionContext = GetPerTupleExprContext(executorState);
EState *executorState = CreateExecutorState();
MemoryContext executorTupleContext = GetPerTupleMemoryContext(executorState);
ExprContext *executorExpressionContext = GetPerTupleExprContext(executorState);
if (IsCitusTableType(tableId, REFERENCE_TABLE))
{
@ -491,9 +481,11 @@ CopyToExistingShards(CopyStmt *copyStatement, QueryCompletionCompat *completionT
}
/* set up the destination for the COPY */
copyDest = CreateCitusCopyDestReceiver(tableId, columnNameList, partitionColumnIndex,
executorState, stopOnFailure, NULL);
dest = (DestReceiver *) copyDest;
CitusCopyDestReceiver *copyDest = CreateCitusCopyDestReceiver(tableId, columnNameList,
partitionColumnIndex,
executorState,
stopOnFailure, NULL);
DestReceiver *dest = (DestReceiver *) copyDest;
dest->rStartup(dest, 0, tupleDescriptor);
/*
@ -501,8 +493,9 @@ CopyToExistingShards(CopyStmt *copyStatement, QueryCompletionCompat *completionT
* of BeginCopyFrom. However, we obviously should not do this in relcache
* and therefore make a copy of the Relation.
*/
copiedDistributedRelation = (Relation) palloc(sizeof(RelationData));
copiedDistributedRelationTuple = (Form_pg_class) palloc(CLASS_TUPLE_SIZE);
Relation copiedDistributedRelation = (Relation) palloc(sizeof(RelationData));
Form_pg_class copiedDistributedRelationTuple =
(Form_pg_class) palloc(CLASS_TUPLE_SIZE);
/*
* There is no need to deep copy everything. We will just deep copy of the fields
@ -527,13 +520,13 @@ CopyToExistingShards(CopyStmt *copyStatement, QueryCompletionCompat *completionT
}
/* initialize copy state to read from COPY data source */
copyState = BeginCopyFrom(NULL,
copiedDistributedRelation,
copyStatement->filename,
copyStatement->is_program,
NULL,
copyStatement->attlist,
copyStatement->options);
CopyState copyState = BeginCopyFrom(NULL,
copiedDistributedRelation,
copyStatement->filename,
copyStatement->is_program,
NULL,
copyStatement->attlist,
copyStatement->options);
/* set up callback to identify error line number */
errorCallback.callback = CopyFromErrorCallback;
@ -945,18 +938,15 @@ CanUseBinaryCopyFormat(TupleDesc tupleDescription)
for (int columnIndex = 0; columnIndex < totalColumnCount; columnIndex++)
{
Form_pg_attribute currentColumn = TupleDescAttr(tupleDescription, columnIndex);
Oid typeId = InvalidOid;
if (currentColumn->attisdropped
#if PG_VERSION_NUM >= PG_VERSION_12
|| currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
#endif
if (currentColumn->attisdropped ||
currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
)
{
continue;
}
typeId = currentColumn->atttypid;
Oid typeId = currentColumn->atttypid;
if (!CanUseBinaryCopyFormatForType(typeId))
{
useBinaryCopyFormat = false;
@ -1417,10 +1407,8 @@ TypeArrayFromTupleDescriptor(TupleDesc tupleDescriptor)
for (int columnIndex = 0; columnIndex < columnCount; columnIndex++)
{
Form_pg_attribute attr = TupleDescAttr(tupleDescriptor, columnIndex);
if (attr->attisdropped
#if PG_VERSION_NUM >= PG_VERSION_12
|| attr->attgenerated == ATTRIBUTE_GENERATED_STORED
#endif
if (attr->attisdropped ||
attr->attgenerated == ATTRIBUTE_GENERATED_STORED
)
{
typeArray[columnIndex] = InvalidOid;
@ -1589,10 +1577,8 @@ AppendCopyRowData(Datum *valueArray, bool *isNullArray, TupleDesc rowDescriptor,
value = CoerceColumnValue(value, &columnCoercionPaths[columnIndex]);
}
if (currentColumn->attisdropped
#if PG_VERSION_NUM >= PG_VERSION_12
|| currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
#endif
if (currentColumn->attisdropped ||
currentColumn->attgenerated == ATTRIBUTE_GENERATED_STORED
)
{
continue;
@ -1712,10 +1698,8 @@ AvailableColumnCount(TupleDesc tupleDescriptor)
{
Form_pg_attribute currentColumn = TupleDescAttr(tupleDescriptor, columnIndex);
if (!currentColumn->attisdropped
#if PG_VERSION_NUM >= PG_VERSION_12
&& currentColumn->attgenerated != ATTRIBUTE_GENERATED_STORED
#endif
if (!currentColumn->attisdropped &&
currentColumn->attgenerated != ATTRIBUTE_GENERATED_STORED
)
{
columnCount++;
@ -2993,13 +2977,11 @@ ProcessCopyStmt(CopyStmt *copyStatement, QueryCompletionCompat *completionTag, c
{
if (copyStatement->is_from)
{
#if PG_VERSION_NUM >= PG_VERSION_12
if (copyStatement->whereClause)
{
ereport(ERROR, (errmsg(
"Citus does not support COPY FROM with WHERE")));
}
#endif
/* check permissions, we're bypassing postgres' normal checks */
CheckCopyPermissions(copyStatement);
@ -3054,10 +3036,8 @@ CitusCopySelect(CopyStmt *copyStatement)
{
Form_pg_attribute attr = &tupleDescriptor->attrs[i];
if (attr->attisdropped
#if PG_VERSION_NUM >= PG_VERSION_12
|| attr->attgenerated
#endif
if (attr->attisdropped ||
attr->attgenerated
)
{
continue;
@ -3312,10 +3292,8 @@ CopyGetAttnums(TupleDesc tupDesc, Relation rel, List *attnamelist)
{
if (TupleDescAttr(tupDesc, i)->attisdropped)
continue;
#if PG_VERSION_NUM >= PG_VERSION_12
if (TupleDescAttr(tupDesc, i)->attgenerated)
continue;
#endif
attnums = lappend_int(attnums, i + 1);
}
}
@ -3340,14 +3318,12 @@ CopyGetAttnums(TupleDesc tupDesc, Relation rel, List *attnamelist)
continue;
if (namestrcmp(&(att->attname), name) == 0)
{
#if PG_VERSION_NUM >= PG_VERSION_12
if (att->attgenerated)
ereport(ERROR,
(errcode(ERRCODE_INVALID_COLUMN_REFERENCE),
errmsg("column \"%s\" is a generated column",
name),
errdetail("Generated columns cannot be used in COPY.")));
#endif
attnum = att->attnum;
break;
}

6
src/backend/distributed/commands/role.c
View File

@ -14,9 +14,7 @@
#include "access/heapam.h"
#include "access/htup_details.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/table.h"
#endif
#include "catalog/catalog.h"
#include "catalog/pg_auth_members.h"
#include "catalog/pg_authid.h"
@ -543,11 +541,7 @@ GenerateAlterRoleSetCommandForRole(Oid roleid)
List *commands = NIL;
#if PG_VERSION_NUM >= PG_VERSION_12
TableScanDesc scan = table_beginscan_catalog(DbRoleSetting, 0, NULL);
#else
HeapScanDesc scan = heap_beginscan_catalog(DbRoleSetting, 0, NULL);
#endif
while ((tuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
{

2
src/backend/distributed/commands/sequence.c
View File

@ -178,13 +178,11 @@ ExtractDefaultColumnsAndOwnedSequences(Oid relationId, List **columnNameList,
continue;
}
#if PG_VERSION_NUM >= PG_VERSION_12
if (attributeForm->attgenerated == ATTRIBUTE_GENERATED_STORED)
{
/* skip columns with GENERATED AS ALWAYS expressions */
continue;
}
#endif
char *columnName = NameStr(attributeForm->attname);
*columnNameList = lappend(*columnNameList, columnName);

7
src/backend/distributed/commands/statistics.c
View File

@ -594,14 +594,9 @@ GetExplicitStatisticsIdList(Oid relationId)
HeapTuple heapTuple = systable_getnext(scanDescriptor);
while (HeapTupleIsValid(heapTuple))
{
Oid statisticsId = InvalidOid;
#if PG_VERSION_NUM >= PG_VERSION_12
FormData_pg_statistic_ext *statisticsForm =
(FormData_pg_statistic_ext *) GETSTRUCT(heapTuple);
statisticsId = statisticsForm->oid;
#else
statisticsId = HeapTupleGetOid(heapTuple);
#endif
Oid statisticsId = statisticsForm->oid;
statisticsIdList = lappend_oid(statisticsIdList, statisticsId);
heapTuple = systable_getnext(scanDescriptor);

2
src/backend/distributed/commands/table.c
View File

@ -10,9 +10,7 @@
#include "postgres.h"
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/genam.h"
#endif
#include "access/htup_details.h"
#include "access/xact.h"
#include "catalog/index.h"

17
src/backend/distributed/commands/trigger.c
View File

@ -12,13 +12,7 @@
#include "distributed/pg_version_constants.h"
#include "access/genam.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/table.h"
#else
#include "access/heapam.h"
#include "access/htup_details.h"
#include "access/sysattr.h"
#endif
#include "catalog/indexing.h"
#include "catalog/namespace.h"
#include "catalog/pg_trigger.h"
@ -101,11 +95,7 @@ GetTriggerTupleById(Oid triggerId, bool missingOk)
int scanKeyCount = 1;
ScanKeyData scanKey[1];
#if PG_VERSION_NUM >= PG_VERSION_12
AttrNumber attrNumber = Anum_pg_trigger_oid;
#else
AttrNumber attrNumber = ObjectIdAttributeNumber;
#endif
ScanKeyInit(&scanKey[0], attrNumber, BTEqualStrategyNumber,
F_OIDEQ, ObjectIdGetDatum(triggerId));
@ -198,14 +188,9 @@ get_relation_trigger_oid_compat(HeapTuple heapTuple)
{
Assert(HeapTupleIsValid(heapTuple));
Oid triggerOid = InvalidOid;
#if PG_VERSION_NUM >= PG_VERSION_12
Form_pg_trigger triggerForm = (Form_pg_trigger) GETSTRUCT(heapTuple);
triggerOid = triggerForm->oid;
#else
triggerOid = HeapTupleGetOid(heapTuple);
#endif
Oid triggerOid = triggerForm->oid;
return triggerOid;
}

91
src/backend/distributed/commands/type.c
View File

@ -322,8 +322,6 @@ List *
PreprocessAlterEnumStmt(Node *node, const char *queryString,
ProcessUtilityContext processUtilityContext)
{
List *commands = NIL;
ObjectAddress typeAddress = GetObjectAddressFromParseTree(node, false);
if (!ShouldPropagateObject(&typeAddress))
{
@ -352,98 +350,15 @@ PreprocessAlterEnumStmt(Node *node, const char *queryString,
* creating a DDLTaksList we won't return anything here. During the processing phase
* we directly connect to workers and execute the commands remotely.
*/
#if PG_VERSION_NUM < PG_VERSION_12
if (AlterEnumIsAddValue(castNode(AlterEnumStmt, node)))
{
/*
* a plan cannot be made as it will be committed via 2PC when ran through the
* executor, instead we directly distributed during processing phase
*/
return NIL;
}
#endif
commands = list_make3(DISABLE_DDL_PROPAGATION,
(void *) alterEnumStmtSql,
ENABLE_DDL_PROPAGATION);
List *commands = list_make3(DISABLE_DDL_PROPAGATION,
(void *) alterEnumStmtSql,
ENABLE_DDL_PROPAGATION);
return NodeDDLTaskList(NON_COORDINATOR_NODES, commands);
}
/*
* PostprocessAlterEnumStmt is called after the AlterEnumStmt has been applied locally.
*
* This function is used for ALTER ENUM ... ADD VALUE for postgres versions lower than 12
* to distribute the call. Before pg12 these statements could not be called in a
* transaction. If we would plan the distirbution of these statements the same as we do
* with the other statements they would get executed in a transaction to perform 2PC, that
* would error out.
*
* If it would error on some workers we provide a warning to the user that the statement
* failed to distributed with some detail on what to call after the cluster has been
* repaired.
*
* For pg12 the statements can be called in a transaction but will only become visible
* when the transaction commits. This is behaviour that is ok to perform in a 2PC.
*/
List *
PostprocessAlterEnumStmt(Node *node, const char *queryString)
{
/*
* Before pg12 ALTER ENUM ... ADD VALUE could not be within a xact block. Normally we
* would propagate the statements in a xact block to perform 2pc on changes via ddl.
* Instead we need to connect directly to the workers here and execute the command.
*
* From pg12 and up we use the normal infrastructure and create the ddl jobs during
* planning.
*/
#if PG_VERSION_NUM < PG_VERSION_12
AlterEnumStmt *stmt = castNode(AlterEnumStmt, node);
ObjectAddress typeAddress = GetObjectAddressFromParseTree((Node *) stmt, false);
if (!ShouldPropagateObject(&typeAddress))
{
return NIL;
}
if (AlterEnumIsAddValue(stmt))
{
/*
* ADD VALUE can't be executed in a transaction, we will execute optimistically
* and on an error we will advise to fix the issue with the worker and rerun the
* query with the IF NOT EXTISTS modifier. The modifier is needed as the value
* might already be added to some nodes, but not all.
*/
/* qualification of the stmt happened during planning */
const char *alterEnumStmtSql = DeparseTreeNode((Node *) stmt);
List *commands = list_make2(DISABLE_DDL_PROPAGATION, (void *) alterEnumStmtSql);
int result = SendBareOptionalCommandListToAllWorkersAsUser(commands, NULL);
if (result != RESPONSE_OKAY)
{
bool oldSkipIfNewValueExists = stmt->skipIfNewValExists;
/* deparse the query with IF NOT EXISTS */
stmt->skipIfNewValExists = true;
const char *alterEnumStmtIfNotExistsSql = DeparseTreeNode((Node *) stmt);
stmt->skipIfNewValExists = oldSkipIfNewValueExists;
ereport(WARNING, (errmsg("not all workers applied change to enum"),
errdetail("retry with: %s", alterEnumStmtIfNotExistsSql),
errhint("make sure the coordinators can communicate with "
"all workers")));
}
}
#endif
return NIL;
}
/*
* PreprocessDropTypeStmt is called for all DROP TYPE statements. For all types in the list that
* citus has distributed to the workers it will drop the type on the workers as well. If

16
src/backend/distributed/commands/utility_hook.c
View File

@ -568,22 +568,6 @@ ProcessUtilityInternal(PlannedStmt *pstmt,
PostprocessAlterExtensionCitusUpdateStmt(parsetree);
}
/*
* Postgres added the following CommandCounterIncrement as a patch in:
* - 10.7 -> 10.8
* - 11.2 -> 11.3
* The patch was a response to bug #15631.
*
* CommandCounterIncrement is used to make changes to the catalog visible for post
* processing of create commands (eg. create type). It is safe to call
* CommandCounterIncrement twice, as the call is a no-op if the command id is not
* used yet.
*
* Once versions older than above are not deemed important anymore this patch can
* be remove from citus.
*/
CommandCounterIncrement();
PostStandardProcessUtility(parsetree);
}
PG_CATCH();

30
src/backend/distributed/commands/vacuum.c
View File

@ -12,9 +12,7 @@
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "commands/defrem.h"
#endif
#include "commands/vacuum.h"
#include "distributed/adaptive_executor.h"
#include "distributed/commands.h"
@ -41,14 +39,12 @@
typedef struct CitusVacuumParams
{
int options;
#if PG_VERSION_NUM >= PG_VERSION_12
VacOptTernaryValue truncate;
VacOptTernaryValue index_cleanup;
#endif
#if PG_VERSION_NUM >= PG_VERSION_13
#if PG_VERSION_NUM >= PG_VERSION_13
int nworkers;
#endif
#endif
} CitusVacuumParams;
/* Local functions forward declarations for processing distributed table commands */
@ -349,11 +345,9 @@ DeparseVacuumStmtPrefix(CitusVacuumParams vacuumParams)
}
/* if no flags remain, exit early */
if (vacuumFlags == 0
#if PG_VERSION_NUM >= PG_VERSION_12
&& vacuumParams.truncate == VACOPT_TERNARY_DEFAULT &&
if (vacuumFlags == 0 &&
vacuumParams.truncate == VACOPT_TERNARY_DEFAULT &&
vacuumParams.index_cleanup == VACOPT_TERNARY_DEFAULT
#endif
#if PG_VERSION_NUM >= PG_VERSION_13
&& vacuumParams.nworkers == VACUUM_PARALLEL_NOTSET
#endif
@ -390,7 +384,6 @@ DeparseVacuumStmtPrefix(CitusVacuumParams vacuumParams)
appendStringInfoString(vacuumPrefix, "VERBOSE,");
}
#if PG_VERSION_NUM >= PG_VERSION_12
if (vacuumFlags & VACOPT_SKIP_LOCKED)
{
appendStringInfoString(vacuumPrefix, "SKIP_LOCKED,");
@ -411,7 +404,6 @@ DeparseVacuumStmtPrefix(CitusVacuumParams vacuumParams)
"INDEX_CLEANUP," : "INDEX_CLEANUP false,"
);
}
#endif
#if PG_VERSION_NUM >= PG_VERSION_13
if (vacuumParams.nworkers != VACUUM_PARALLEL_NOTSET)
@ -496,7 +488,6 @@ ExtractVacuumTargetRels(VacuumStmt *vacuumStmt)
/*
* VacuumStmtParams returns a CitusVacuumParams based on the supplied VacuumStmt.
*/
#if PG_VERSION_NUM >= PG_VERSION_12
/*
* This is mostly ExecVacuum from Postgres's commands/vacuum.c
@ -611,16 +602,3 @@ VacuumStmtParams(VacuumStmt *vacstmt)
(disable_page_skipping ? VACOPT_DISABLE_PAGE_SKIPPING : 0);
return params;
}
#else
static CitusVacuumParams
VacuumStmtParams(VacuumStmt *vacuumStmt)
{
CitusVacuumParams params;
params.options = vacuumStmt->options;
return params;
}
#endif

21
src/backend/distributed/deparser/citus_ruleutils.c
View File

@ -125,22 +125,17 @@ get_extension_schema(Oid ext_oid)
/* *INDENT-OFF* */
Oid result;
Relation rel;
SysScanDesc scandesc;
HeapTuple tuple;
ScanKeyData entry[1];
rel = table_open(ExtensionRelationId, AccessShareLock);
ScanKeyInit(&entry[0],
#if PG_VERSION_NUM >= PG_VERSION_12
Anum_pg_extension_oid,
#else
ObjectIdAttributeNumber,
#endif
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(ext_oid));
scandesc = systable_beginscan(rel, ExtensionOidIndexId, true,
SysScanDesc scandesc = systable_beginscan(rel, ExtensionOidIndexId, true,
NULL, 1, entry);
tuple = systable_getnext(scandesc);
@ -251,7 +246,6 @@ char *
pg_get_tableschemadef_string(Oid tableRelationId, bool includeSequenceDefaults,
char *accessMethod)
{
char relationKind = 0;
bool firstAttributePrinted = false;
AttrNumber defaultValueIndex = 0;
AttrNumber constraintIndex = 0;
@ -361,7 +355,6 @@ pg_get_tableschemadef_string(Oid tableRelationId, bool includeSequenceDefaults,
defaultString = deparse_expression(defaultNode, defaultContext,
false, false);
#if PG_VERSION_NUM >= PG_VERSION_12
if (attributeForm->attgenerated == ATTRIBUTE_GENERATED_STORED)
{
appendStringInfo(&buffer, " GENERATED ALWAYS AS (%s) STORED",
@ -371,9 +364,6 @@ pg_get_tableschemadef_string(Oid tableRelationId, bool includeSequenceDefaults,
{
appendStringInfo(&buffer, " DEFAULT %s", defaultString);
}
#else
appendStringInfo(&buffer, " DEFAULT %s", defaultString);
#endif
}
}
@ -436,7 +426,7 @@ pg_get_tableschemadef_string(Oid tableRelationId, bool includeSequenceDefaults,
* If the relation is a foreign table, append the server name and options to
* the create table statement.
*/
relationKind = relation->rd_rel->relkind;
char relationKind = relation->rd_rel->relkind;
if (relationKind == RELKIND_FOREIGN_TABLE)
{
ForeignTable *foreignTable = GetForeignTable(tableRelationId);
@ -452,8 +442,6 @@ pg_get_tableschemadef_string(Oid tableRelationId, bool includeSequenceDefaults,
appendStringInfo(&buffer, " PARTITION BY %s ", partitioningInformation);
}
#if PG_VERSION_NUM >= 120000
/*
* Add table access methods for pg12 and higher when the table is configured with an
* access method
@ -475,7 +463,6 @@ pg_get_tableschemadef_string(Oid tableRelationId, bool includeSequenceDefaults,
appendStringInfo(&buffer, " USING %s", quote_identifier(NameStr(amForm->amname)));
ReleaseSysCache(amTup);
}
#endif
/*
* Add any reloptions (storage parameters) defined on the table in a WITH
@ -745,11 +732,7 @@ deparse_shard_reindex_statement(ReindexStmt *origStmt, Oid distrelid, int64 shar
{
ReindexStmt *reindexStmt = copyObject(origStmt); /* copy to avoid modifications */
char *relationName = NULL;
#if PG_VERSION_NUM >= PG_VERSION_12
const char *concurrentlyString = reindexStmt->concurrent ? "CONCURRENTLY " : "";
#else
const char *concurrentlyString = "";
#endif
if (reindexStmt->kind == REINDEX_OBJECT_INDEX ||

8436
src/backend/distributed/deparser/ruleutils_11.c
View File

File diff suppressed because it is too large Load Diff

6
src/backend/distributed/executor/citus_custom_scan.c
View File

@ -36,11 +36,7 @@
#include "distributed/worker_protocol.h"
#include "executor/executor.h"
#include "nodes/makefuncs.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/planner.h"
#endif
#include "optimizer/clauses.h"
#include "utils/memutils.h"
#include "utils/rel.h"
@ -157,7 +153,6 @@ CitusBeginScan(CustomScanState *node, EState *estate, int eflags)
*/
EnableWorkerMessagePropagation();
#if PG_VERSION_NUM >= PG_VERSION_12
/*
* Since we are using a tuplestore we cannot use the virtual tuples postgres had
@ -176,7 +171,6 @@ CitusBeginScan(CustomScanState *node, EState *estate, int eflags)
ExecAssignScanProjectionInfoWithVarno(&node->ss, INDEX_VAR);
node->ss.ps.qual = ExecInitQual(node->ss.ps.plan->qual, (PlanState *) node);
#endif
DistributedPlan *distributedPlan = scanState->distributedPlan;
if (distributedPlan->insertSelectQuery != NULL)

14
src/backend/distributed/executor/distributed_intermediate_results.c
View File

@ -272,14 +272,9 @@ WrapTasksForPartitioning(const char *resultIdPrefix, List *selectTaskList,
static PartitioningTupleDest *
CreatePartitioningTupleDest(CitusTableCacheEntry *targetRelation)
{
TupleDesc tupleDescriptor = NULL;
int resultColumnCount = 3;
#if PG_VERSION_NUM >= PG_VERSION_12
tupleDescriptor = CreateTemplateTupleDesc(resultColumnCount);
#else
tupleDescriptor = CreateTemplateTupleDesc(resultColumnCount, false);
#endif
TupleDesc tupleDescriptor = CreateTemplateTupleDesc(resultColumnCount);
TupleDescInitEntry(tupleDescriptor, (AttrNumber) 1, "partition_index",
INT4OID, -1, 0);
@ -686,14 +681,9 @@ QueryStringForFragmentsTransfer(NodeToNodeFragmentsTransfer *fragmentsTransfer)
static void
ExecuteFetchTaskList(List *taskList)
{
TupleDesc resultDescriptor = NULL;
int resultColumnCount = 1;
#if PG_VERSION_NUM >= PG_VERSION_12
resultDescriptor = CreateTemplateTupleDesc(resultColumnCount);
#else
resultDescriptor = CreateTemplateTupleDesc(resultColumnCount, false);
#endif
TupleDesc resultDescriptor = CreateTemplateTupleDesc(resultColumnCount);
TupleDescInitEntry(resultDescriptor, (AttrNumber) 1, "byte_count", INT8OID, -1, 0);

4
src/backend/distributed/executor/local_executor.c
View File

@ -100,11 +100,7 @@
#include "distributed/worker_protocol.h"
#include "executor/tstoreReceiver.h"
#include "executor/tuptable.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/planner.h"
#endif
#include "nodes/params.h"
#include "utils/snapmgr.h"

3
src/backend/distributed/executor/multi_executor.c
View File

@ -181,9 +181,6 @@ CitusExecutorRun(QueryDesc *queryDesc,
EState *estate = queryDesc->estate;
estate->es_processed = 0;
#if PG_VERSION_NUM < PG_VERSION_12
estate->es_lastoid = InvalidOid;
#endif
/* start and shutdown tuple receiver to simulate empty result */
dest->rStartup(queryDesc->dest, CMD_SELECT, queryDesc->tupDesc);

4
src/backend/distributed/executor/tuple_destination.c
View File

@ -280,11 +280,7 @@ TupleDestDestReceiverReceive(TupleTableSlot *slot,
Assert(task->queryCount == 1);
int queryNumber = 0;
#if PG_VERSION_NUM >= PG_VERSION_12
HeapTuple heapTuple = ExecFetchSlotHeapTuple(slot, true, NULL);
#else
HeapTuple heapTuple = ExecFetchSlotTuple(slot);
#endif
uint64 tupleLibpqSize = 0;

9
src/backend/distributed/metadata/dependency.c
View File

@ -548,7 +548,6 @@ SupportedDependencyByCitus(const ObjectAddress *address)
*/
switch (getObjectClass(address))
{
#if PG_VERSION_NUM >= 120000
case OCLASS_AM:
{
/*
@ -559,7 +558,6 @@ SupportedDependencyByCitus(const ObjectAddress *address)
*/
return IsObjectAddressOwnedByExtension(address, NULL);
}
#endif
case OCLASS_COLLATION:
case OCLASS_SCHEMA:
@ -1188,17 +1186,10 @@ GetDependingView(Form_pg_depend pg_depend)
Relation rewriteRel = table_open(RewriteRelationId, AccessShareLock);
ScanKeyData rkey[1];
#if PG_VERSION_NUM >= PG_VERSION_12
ScanKeyInit(&rkey[0],
Anum_pg_rewrite_oid,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(pg_depend->objid));
#else
ScanKeyInit(&rkey[0],
ObjectIdAttributeNumber,
BTEqualStrategyNumber, F_OIDEQ,
ObjectIdGetDatum(pg_depend->objid));
#endif
SysScanDesc rscan = systable_beginscan(rewriteRel, RewriteOidIndexId,
true, NULL, 1, rkey);

10
src/backend/distributed/metadata/distobject.c
View File

@ -114,15 +114,11 @@ ObjectExists(const ObjectAddress *address)
if (is_objectclass_supported(address->classId))
{
HeapTuple objtup;
Relation catalog = table_open(address->classId, AccessShareLock);
#if PG_VERSION_NUM >= PG_VERSION_12
objtup = get_catalog_object_by_oid(catalog, get_object_attnum_oid(
address->classId), address->objectId);
#else
objtup = get_catalog_object_by_oid(catalog, address->objectId);
#endif
HeapTuple objtup = get_catalog_object_by_oid(catalog, get_object_attnum_oid(
address->classId),
address->objectId);
table_close(catalog, AccessShareLock);
if (objtup != NULL)
{

7
src/backend/distributed/metadata/metadata_cache.c
View File

@ -2552,7 +2552,6 @@ LookupTypeOid(char *schemaNameSting, char *typeNameString)
List *qualifiedName = list_make2(schemaName, typeName);
TypeName *enumTypeName = makeTypeNameFromNameList(qualifiedName);
Oid nodeRoleTypId;
/* typenameTypeId but instead of raising an error return InvalidOid */
Type tup = LookupTypeName(NULL, enumTypeName, NULL, false);
@ -2561,11 +2560,7 @@ LookupTypeOid(char *schemaNameSting, char *typeNameString)
return InvalidOid;
}
#if PG_VERSION_NUM >= PG_VERSION_12
nodeRoleTypId = ((Form_pg_type) GETSTRUCT(tup))->oid;
#else
nodeRoleTypId = HeapTupleGetOid(tup);
#endif
Oid nodeRoleTypId = ((Form_pg_type) GETSTRUCT(tup))->oid;
ReleaseSysCache(tup);
return nodeRoleTypId;

2
src/backend/distributed/metadata/metadata_utility.c
View File

@ -17,9 +17,7 @@
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/genam.h"
#endif
#include "access/htup_details.h"
#include "access/sysattr.h"
#include "access/xact.h"

7
src/backend/distributed/operations/delete_protocol.c
View File

@ -48,21 +48,14 @@
#include "distributed/worker_protocol.h"
#include "distributed/worker_transaction.h"
#include "lib/stringinfo.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/nodeFuncs.h"
#endif
#include "nodes/nodes.h"
#include "nodes/parsenodes.h"
#include "nodes/pg_list.h"
#include "nodes/primnodes.h"
#include "optimizer/clauses.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#include "optimizer/optimizer.h"
#else
#include "nodes/relation.h"
#include "optimizer/predtest.h"
#endif
#include "optimizer/restrictinfo.h"
#include "storage/lock.h"
#include "storage/lmgr.h"

5
src/backend/distributed/operations/modify_multiple_shards.c
View File

@ -48,12 +48,7 @@
#include "distributed/worker_protocol.h"
#include "distributed/worker_transaction.h"
#include "optimizer/clauses.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/predtest.h"
#include "optimizer/var.h"
#endif
#include "optimizer/restrictinfo.h"
#include "nodes/makefuncs.h"
#include "tcop/tcopprot.h"

13
src/backend/distributed/operations/node_protocol.c
View File

@ -451,15 +451,12 @@ Datum
citus_get_active_worker_nodes(PG_FUNCTION_ARGS)
{
FuncCallContext *functionContext = NULL;
uint32 workerNodeIndex = 0;
uint32 workerNodeCount = 0;
CheckCitusVersion(ERROR);
if (SRF_IS_FIRSTCALL())
{
TupleDesc tupleDescriptor = NULL;
/* create a function context for cross-call persistence */
functionContext = SRF_FIRSTCALL_INIT();
@ -477,11 +474,7 @@ citus_get_active_worker_nodes(PG_FUNCTION_ARGS)
* This tuple descriptor must match the output parameters declared for
* the function in pg_proc.
*/
#if PG_VERSION_NUM < PG_VERSION_12
tupleDescriptor = CreateTemplateTupleDesc(WORKER_NODE_FIELDS, false);
#else
tupleDescriptor = CreateTemplateTupleDesc(WORKER_NODE_FIELDS);
#endif
TupleDesc tupleDescriptor = CreateTemplateTupleDesc(WORKER_NODE_FIELDS);
TupleDescInitEntry(tupleDescriptor, (AttrNumber) 1, "node_name",
TEXTOID, -1, 0);
TupleDescInitEntry(tupleDescriptor, (AttrNumber) 2, "node_port",
@ -493,7 +486,7 @@ citus_get_active_worker_nodes(PG_FUNCTION_ARGS)
}
functionContext = SRF_PERCALL_SETUP();
workerNodeIndex = functionContext->call_cntr;
uint32 workerNodeIndex = functionContext->call_cntr;
workerNodeCount = functionContext->max_calls;
if (workerNodeIndex < workerNodeCount)
@ -668,7 +661,6 @@ GetPreLoadTableCreationCommands(Oid relationId, bool includeSequenceDefaults,
tableColumnOptionsDef));
}
#if PG_VERSION_NUM >= 120000
/* add columnar options for cstore tables */
if (accessMethod == NULL && IsColumnarTableAmTable(relationId))
@ -679,7 +671,6 @@ GetPreLoadTableCreationCommands(Oid relationId, bool includeSequenceDefaults,
tableDDLEventList = lappend(tableDDLEventList, cstoreOptionsDDL);
}
}
#endif
char *tableOwnerDef = TableOwnerResetCommand(relationId);
if (tableOwnerDef != NULL)

4
src/backend/distributed/planner/combine_query_planner.c
View File

@ -346,11 +346,7 @@ FindCitusExtradataContainerRTE(Node *node, RangeTblEntry **result)
}
else if (IsA(node, Query))
{
#if PG_VERSION_NUM >= PG_VERSION_12
const int flags = QTW_EXAMINE_RTES_BEFORE;
#else
const int flags = QTW_EXAMINE_RTES;
#endif
return query_tree_walker((Query *) node, FindCitusExtradataContainerRTE, result,
flags);
}

160
src/backend/distributed/planner/cte_inline.c
View File

@ -7,8 +7,6 @@
* planning, the query pushdown planning, kicks in and the CTEs can actually
* be pushed down as long as it is safe to pushdown as a subquery.
*
* Most of the logic in this function is inspired (and some is copy & pasted)
* from PostgreSQL 12's CTE inlining feature.
*
* Copyright (c) Citus Data, Inc.
*-------------------------------------------------------------------------
@ -18,29 +16,9 @@
#include "distributed/cte_inline.h"
#include "nodes/nodeFuncs.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/cost.h"
#include "optimizer/clauses.h"
#endif
#include "rewrite/rewriteManip.h"
#if PG_VERSION_NUM < PG_VERSION_12
/* copy & paste from PG 12 */
#define PG_12_QTW_EXAMINE_RTES_BEFORE 0x10
#define PG_12_QTW_EXAMINE_RTES_AFTER 0x20
bool pg_12_query_tree_walker(Query *query,
bool (*walker)(),
void *context,
int flags);
bool pg_12_range_table_walker(List *rtable,
bool (*walker)(),
void *context,
int flags);
#endif
typedef struct inline_cte_walker_context
{
const char *ctename; /* name and relative level of target CTE */
@ -233,18 +211,9 @@ PostgreSQLCTEInlineCondition(CommonTableExpr *cte, CmdType cmdType)
* will be inlined even if multiply referenced.
*/
if (
#if PG_VERSION_NUM >= PG_VERSION_12
(cte->ctematerialized == CTEMaterializeNever ||
(cte->ctematerialized == CTEMaterializeDefault &&
cte->cterefcount == 1)) &&
#else
/*
* If referenced only once inlining would probably perform
* better, so for pg < 12, try inlining
*/
cte->cterefcount == 1 &&
#endif
!cte->cterecursive &&
cmdType == CMD_SELECT &&
!contain_dml(cte->ctequery) &&
@ -294,18 +263,8 @@ inline_cte_walker(Node *node, inline_cte_walker_context *context)
context->levelsup++;
/*
* Visit the query's RTE nodes after their contents; otherwise
* query_tree_walker would descend into the newly inlined CTE query,
* which we don't want.
*/
#if PG_VERSION_NUM < PG_VERSION_12
(void) pg_12_query_tree_walker(query, inline_cte_walker, context,
PG_12_QTW_EXAMINE_RTES_AFTER);
#else
(void) query_tree_walker(query, inline_cte_walker, context,
QTW_EXAMINE_RTES_AFTER);
#endif
context->levelsup--;
return false;
@ -411,123 +370,4 @@ contain_dml_walker(Node *node, void *context)
return expression_tree_walker(node, contain_dml_walker, context);
}
#if PG_VERSION_NUM < PG_VERSION_12
/*
* pg_12_query_tree_walker is copied from Postgres 12's source
* code. The only difference between query_tree_walker the new
* two flags added in range_table_walker: QTW_EXAMINE_RTES_AFTER
* and QTW_EXAMINE_RTES_BEFORE.
*/
bool
pg_12_query_tree_walker(Query *query,
bool (*walker) (),
void *context,
int flags)
{
Assert(query != NULL && IsA(query, Query));
if (walker((Node *) query->targetList, context))
return true;
if (walker((Node *) query->withCheckOptions, context))
return true;
if (walker((Node *) query->onConflict, context))
return true;
if (walker((Node *) query->returningList, context))
return true;
if (walker((Node *) query->jointree, context))
return true;
if (walker(query->setOperations, context))
return true;
if (walker(query->havingQual, context))
return true;
if (walker(query->limitOffset, context))
return true;
if (walker(query->limitCount, context))
return true;
if (!(flags & QTW_IGNORE_CTE_SUBQUERIES))
{
if (walker((Node *) query->cteList, context))
return true;
}
if (!(flags & QTW_IGNORE_RANGE_TABLE))
{
if (pg_12_range_table_walker(query->rtable, walker, context, flags))
return true;
}
return false;
}
/*
* pg_12_range_table_walker is copied from Postgres 12's source
* code. The only difference between range_table_walker the new
* two flags added in range_table_walker: QTW_EXAMINE_RTES_AFTER
* and QTW_EXAMINE_RTES_BEFORE.
*/
bool
pg_12_range_table_walker(List *rtable,
bool (*walker) (),
void *context,
int flags)
{
ListCell *rt;
foreach(rt, rtable)
{
RangeTblEntry *rte = (RangeTblEntry *) lfirst(rt);
/*
* Walkers might need to examine the RTE node itself either before or
* after visiting its contents (or, conceivably, both). Note that if
* you specify neither flag, the walker won't visit the RTE at all.
*/
if (flags & PG_12_QTW_EXAMINE_RTES_BEFORE)
if (walker(rte, context))
return true;
switch (rte->rtekind)
{
case RTE_RELATION:
if (walker(rte->tablesample, context))
return true;
break;
case RTE_CTE:
case RTE_NAMEDTUPLESTORE:
/* nothing to do */
break;
case RTE_SUBQUERY:
if (!(flags & QTW_IGNORE_RT_SUBQUERIES))
if (walker(rte->subquery, context))
return true;
break;
case RTE_JOIN:
if (!(flags & QTW_IGNORE_JOINALIASES))
if (walker(rte->joinaliasvars, context))
return true;
break;
case RTE_FUNCTION:
if (walker(rte->functions, context))
return true;
break;
case RTE_TABLEFUNC:
if (walker(rte->tablefunc, context))
return true;
break;
case RTE_VALUES:
if (walker(rte->values_lists, context))
return true;
break;
}
if (walker(rte->securityQuals, context))
return true;
if (flags & PG_12_QTW_EXAMINE_RTES_AFTER)
if (walker(rte, context))
return true;
}
return false;
}
#endif
/* *INDENT-ON* */

55
src/backend/distributed/planner/distributed_planner.c
View File

@ -52,12 +52,8 @@
#include "nodes/pg_list.h"
#include "parser/parsetree.h"
#include "parser/parse_type.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#include "optimizer/plancat.h"
#else
#include "optimizer/cost.h"
#endif
#include "optimizer/pathnode.h"
#include "optimizer/planner.h"
#include "optimizer/planmain.h"
@ -751,24 +747,6 @@ InlineCtesAndCreateDistributedPlannedStmt(uint64 planId,
/* after inlining, we shouldn't have any inlinable CTEs */
Assert(!QueryTreeContainsInlinableCTE(copyOfOriginalQuery));
#if PG_VERSION_NUM < PG_VERSION_12
Query *query = planContext->query;
/*
* We had to implement this hack because on Postgres11 and below, the originalQuery
* and the query would have significant differences in terms of CTEs where CTEs
* would not be inlined on the query (as standard_planner() wouldn't inline CTEs
* on PG 11 and below).
*
* Instead, we prefer to pass the inlined query to the distributed planning. We rely
* on the fact that the query includes subqueries, and it'd definitely go through
* query pushdown planning. During query pushdown planning, the only relevant query
* tree is the original query.
*/
planContext->query = copyObject(copyOfOriginalQuery);
#endif
/* simply recurse into CreateDistributedPlannedStmt() in a PG_TRY() block */
PlannedStmt *result = TryCreateDistributedPlannedStmt(planContext->plan,
copyOfOriginalQuery,
@ -777,15 +755,6 @@ InlineCtesAndCreateDistributedPlannedStmt(uint64 planId,
planContext->
plannerRestrictionContext);
#if PG_VERSION_NUM < PG_VERSION_12
/*
* Set back the original query, in case the planning failed and we need to go
* into distributed planning again.
*/
planContext->query = query;
#endif
return result;
}
@ -1509,15 +1478,10 @@ BlessRecordExpression(Expr *expr)
* Handle row expressions, e.g. SELECT (1,2);
*/
RowExpr *rowExpr = (RowExpr *) expr;
TupleDesc rowTupleDesc = NULL;
ListCell *argCell = NULL;
int currentResno = 1;
#if PG_VERSION_NUM >= PG_VERSION_12
rowTupleDesc = CreateTemplateTupleDesc(list_length(rowExpr->args));
#else
rowTupleDesc = CreateTemplateTupleDesc(list_length(rowExpr->args), false);
#endif
TupleDesc rowTupleDesc = CreateTemplateTupleDesc(list_length(rowExpr->args));
foreach(argCell, rowExpr->args)
{
@ -2015,16 +1979,11 @@ AdjustReadIntermediateResultsCostInternal(RelOptInfo *relOptInfo, List *columnTy
{
PathTarget *reltarget = relOptInfo->reltarget;
List *pathList = relOptInfo->pathlist;
Path *path = NULL;
double rowCost = 0.;
double rowSizeEstimate = 0;
double rowCountEstimate = 0.;
double ioCost = 0.;
#if PG_VERSION_NUM >= PG_VERSION_12
QualCost funcCost = { 0., 0. };
#else
double funcCost = 0.;
#endif
int64 totalResultSize = 0;
ListCell *typeCell = NULL;
@ -2083,17 +2042,9 @@ AdjustReadIntermediateResultsCostInternal(RelOptInfo *relOptInfo, List *columnTy
/* add the cost of parsing a column */
#if PG_VERSION_NUM >= PG_VERSION_12
add_function_cost(NULL, inputFunctionId, NULL, &funcCost);
#else
funcCost += get_func_cost(inputFunctionId);
#endif
}
#if PG_VERSION_NUM >= PG_VERSION_12
rowCost += funcCost.per_tuple;
#else
rowCost += funcCost * cpu_operator_cost;
#endif
/* estimate the number of rows based on the file size and estimated row size */
rowCountEstimate = Max(1, (double) totalResultSize / rowSizeEstimate);
@ -2104,13 +2055,11 @@ AdjustReadIntermediateResultsCostInternal(RelOptInfo *relOptInfo, List *columnTy
Assert(pathList != NIL);
/* tell the planner about the cost and row count of the function */
path = (Path *) linitial(pathList);
Path *path = (Path *) linitial(pathList);
path->rows = rowCountEstimate;
path->total_cost = rowCountEstimate * rowCost + ioCost;
#if PG_VERSION_NUM >= PG_VERSION_12
path->startup_cost = funcCost.startup + relOptInfo->baserestrictcost.startup;
#endif
}

4
src/backend/distributed/planner/extended_op_node_utils.c
View File

@ -16,11 +16,7 @@
#include "distributed/metadata_cache.h"
#include "distributed/multi_logical_optimizer.h"
#include "distributed/pg_dist_partition.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#endif
#include "optimizer/restrictinfo.h"
#include "nodes/nodeFuncs.h"
#include "nodes/pg_list.h"

6
src/backend/distributed/planner/fast_path_router_planner.c
View File

@ -44,17 +44,11 @@
#include "distributed/pg_dist_partition.h"
#include "distributed/shardinterval_utils.h"
#include "distributed/shard_pruning.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/makefuncs.h"
#endif
#include "nodes/nodeFuncs.h"
#include "nodes/parsenodes.h"
#include "nodes/pg_list.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/clauses.h"
#endif
#include "tcop/pquery.h"
bool EnableFastPathRouterPlanner = true;

40
src/backend/distributed/planner/function_call_delegation.c
View File

@ -43,9 +43,7 @@
#include "nodes/primnodes.h"
#include "optimizer/clauses.h"
#include "parser/parse_coerce.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "parser/parsetree.h"
#endif
#include "miscadmin.h"
#include "tcop/dest.h"
#include "utils/lsyscache.h"
@ -102,18 +100,8 @@ contain_param_walker(Node *node, void *context)
PlannedStmt *
TryToDelegateFunctionCall(DistributedPlanningContext *planContext)
{
List *targetList = NIL;
TargetEntry *targetEntry = NULL;
FuncExpr *funcExpr = NULL;
DistObjectCacheEntry *procedure = NULL;
Oid colocatedRelationId = InvalidOid;
bool colocatedWithReferenceTable = false;
CitusTableCacheEntry *distTable = NULL;
Var *partitionColumn = NULL;
ShardPlacement *placement = NULL;
WorkerNode *workerNode = NULL;
Task *task = NULL;
Job *job = NULL;
DistributedPlan *distributedPlan = CitusMakeNode(DistributedPlan);
struct ParamWalkerContext walkerParamContext = { 0 };
@ -157,8 +145,6 @@ TryToDelegateFunctionCall(DistributedPlanningContext *planContext)
if (joinTree->fromlist != NIL)
{
#if PG_VERSION_NUM >= PG_VERSION_12
/*
* In pg12's planning phase empty FROMs are represented with an RTE_RESULT.
* When we arrive here, standard_planner has already been called which calls
@ -194,29 +180,25 @@ TryToDelegateFunctionCall(DistributedPlanningContext *planContext)
Assert(list_length(joinTree->fromlist) > 1);
return NULL;
}
#else
/* query has a FROM section */
return NULL;
#endif
}
targetList = planContext->query->targetList;
List *targetList = planContext->query->targetList;
if (list_length(planContext->query->targetList) != 1)
{
/* multiple target list items */
return NULL;
}
targetEntry = (TargetEntry *) linitial(targetList);
TargetEntry *targetEntry = (TargetEntry *) linitial(targetList);
if (!IsA(targetEntry->expr, FuncExpr))
{
/* target list item is not a function call */
return NULL;
}
funcExpr = (FuncExpr *) targetEntry->expr;
procedure = LookupDistObjectCacheEntry(ProcedureRelationId, funcExpr->funcid, 0);
FuncExpr *funcExpr = (FuncExpr *) targetEntry->expr;
DistObjectCacheEntry *procedure = LookupDistObjectCacheEntry(ProcedureRelationId,
funcExpr->funcid, 0);
if (procedure == NULL || !procedure->isDistributed)
{
/* not a distributed function call */
@ -252,15 +234,15 @@ TryToDelegateFunctionCall(DistributedPlanningContext *planContext)
return NULL;
}
colocatedRelationId = ColocatedTableId(procedure->colocationId);
Oid colocatedRelationId = ColocatedTableId(procedure->colocationId);
if (colocatedRelationId == InvalidOid)
{
ereport(DEBUG1, (errmsg("function does not have co-located tables")));
return NULL;
}
distTable = GetCitusTableCacheEntry(colocatedRelationId);
partitionColumn = distTable->partitionColumn;
CitusTableCacheEntry *distTable = GetCitusTableCacheEntry(colocatedRelationId);
Var *partitionColumn = distTable->partitionColumn;
if (partitionColumn == NULL)
{
colocatedWithReferenceTable = true;
@ -295,7 +277,7 @@ TryToDelegateFunctionCall(DistributedPlanningContext *planContext)
return false;
}
workerNode = FindWorkerNode(placement->nodeName, placement->nodePort);
WorkerNode *workerNode = FindWorkerNode(placement->nodeName, placement->nodePort);
if (workerNode == NULL || !workerNode->hasMetadata || !workerNode->metadataSynced)
{
@ -334,14 +316,14 @@ TryToDelegateFunctionCall(DistributedPlanningContext *planContext)
ereport(DEBUG1, (errmsg("pushing down the function call")));
task = CitusMakeNode(Task);
Task *task = CitusMakeNode(Task);
task->taskType = READ_TASK;
task->taskPlacementList = list_make1(placement);
SetTaskQueryIfShouldLazyDeparse(task, planContext->query);
task->anchorShardId = placement->shardId;
task->replicationModel = distTable->replicationModel;
job = CitusMakeNode(Job);
Job *job = CitusMakeNode(Job);
job->jobId = UniqueJobId();
job->jobQuery = planContext->query;
job->taskList = list_make1(task);

4
src/backend/distributed/planner/insert_select_planner.c
View File

@ -40,11 +40,7 @@
#include "optimizer/planner.h"
#include "optimizer/restrictinfo.h"
#include "optimizer/tlist.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#endif
#include "parser/parsetree.h"
#include "parser/parse_coerce.h"
#include "parser/parse_relation.h"

8
src/backend/distributed/planner/local_distributed_join_planner.c
View File

@ -103,11 +103,7 @@
#include "distributed/version_compat.h"
#include "lib/stringinfo.h"
#include "optimizer/clauses.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#endif
#include "optimizer/planner.h"
#include "optimizer/prep.h"
#include "parser/parsetree.h"
@ -117,11 +113,7 @@
#include "nodes/nodeFuncs.h"
#include "nodes/pg_list.h"
#include "nodes/primnodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#else
#include "nodes/relation.h"
#endif
#include "utils/builtins.h"
#include "utils/guc.h"
#include "utils/lsyscache.h"

4
src/backend/distributed/planner/local_plan_cache.c
View File

@ -18,11 +18,7 @@
#include "distributed/citus_ruleutils.h"
#include "distributed/metadata_cache.h"
#include "distributed/version_compat.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/planner.h"
#endif
#include "optimizer/clauses.h"

4
src/backend/distributed/planner/multi_explain.c
View File

@ -1273,11 +1273,7 @@ CreateExplainAnlyzeDestination(Task *task, TupleDestination *taskDest)
tupleDestination->originalTask = task;
tupleDestination->originalTaskDestination = taskDest;
#if PG_VERSION_NUM >= PG_VERSION_12
TupleDesc lastSavedExplainAnalyzeTupDesc = CreateTemplateTupleDesc(2);
#else
TupleDesc lastSavedExplainAnalyzeTupDesc = CreateTemplateTupleDesc(2, false);
#endif
TupleDescInitEntry(lastSavedExplainAnalyzeTupDesc, 1, "explain analyze", TEXTOID, 0,
0);

4
src/backend/distributed/planner/multi_join_order.c
View File

@ -28,11 +28,7 @@
#include "distributed/pg_dist_partition.h"
#include "distributed/worker_protocol.h"
#include "lib/stringinfo.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#endif
#include "utils/builtins.h"
#include "nodes/nodeFuncs.h"
#include "utils/builtins.h"

8
src/backend/distributed/planner/multi_logical_optimizer.c
View File

@ -47,11 +47,7 @@
#include "nodes/nodeFuncs.h"
#include "optimizer/clauses.h"
#include "optimizer/tlist.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#endif
#include "parser/parse_agg.h"
#include "parser/parse_coerce.h"
#include "parser/parse_oper.h"
@ -3608,11 +3604,7 @@ AggregateFunctionOid(const char *functionName, Oid inputType)
if (procForm->proargtypes.values[0] == inputType ||
procForm->proargtypes.values[0] == ANYELEMENTOID)
{
#if PG_VERSION_NUM < PG_VERSION_12
functionOid = HeapTupleGetOid(heapTuple);
#else
functionOid = procForm->oid;
#endif
break;
}
}

5
src/backend/distributed/planner/multi_logical_planner.c
View File

@ -38,13 +38,8 @@
#include "distributed/version_compat.h"
#include "nodes/makefuncs.h"
#include "nodes/nodeFuncs.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#include "optimizer/optimizer.h"
#else
#include "nodes/relation.h"
#include "optimizer/var.h"
#endif
#include "optimizer/clauses.h"
#include "optimizer/prep.h"
#include "optimizer/tlist.h"

5
src/backend/distributed/planner/multi_physical_planner.c
View File

@ -61,13 +61,8 @@
#include "nodes/makefuncs.h"
#include "nodes/nodeFuncs.h"
#include "optimizer/clauses.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#include "optimizer/optimizer.h"
#else
#include "nodes/relation.h"
#include "optimizer/var.h"
#endif
#include "optimizer/restrictinfo.h"
#include "optimizer/tlist.h"
#include "parser/parse_relation.h"

11
src/backend/distributed/planner/multi_router_planner.c
View File

@ -65,12 +65,7 @@
#include "optimizer/joininfo.h"
#include "optimizer/pathnode.h"
#include "optimizer/paths.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#include "optimizer/predtest.h"
#endif
#include "optimizer/restrictinfo.h"
#include "parser/parsetree.h"
#include "parser/parse_oper.h"
@ -967,10 +962,8 @@ ModifyQuerySupported(Query *queryTree, Query *originalQuery, bool multiShardQuer
queryTableCount++;
}
else if (rangeTableEntry->rtekind == RTE_VALUES
#if PG_VERSION_NUM >= PG_VERSION_12
|| rangeTableEntry->rtekind == RTE_RESULT
#endif
else if (rangeTableEntry->rtekind == RTE_VALUES ||
rangeTableEntry->rtekind == RTE_RESULT
)
{
/* do nothing, this type is supported */

4
src/backend/distributed/planner/query_colocation_checker.c
View File

@ -23,11 +23,7 @@
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/relation.h"
#else
#include "access/heapam.h"
#endif
#include "distributed/multi_logical_planner.h"
#include "distributed/query_colocation_checker.h"
#include "distributed/pg_dist_partition.h"

15
src/backend/distributed/planner/query_pushdown_planning.c
View File

@ -38,12 +38,8 @@
#include "distributed/relation_restriction_equivalence.h"
#include "distributed/version_compat.h"
#include "nodes/nodeFuncs.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/makefuncs.h"
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#endif
#include "nodes/pg_list.h"
#include "optimizer/clauses.h"
#include "parser/parsetree.h"
@ -235,8 +231,6 @@ HasEmptyJoinTree(Query *query)
{
return true;
}
#if PG_VERSION_NUM >= PG_VERSION_12
else if (list_length(query->rtable) == 1)
{
RangeTblEntry *rte = (RangeTblEntry *) linitial(query->rtable);
@ -245,7 +239,6 @@ HasEmptyJoinTree(Query *query)
return true;
}
}
#endif
return false;
}
@ -1153,10 +1146,8 @@ DeferErrorIfUnsupportedTableCombination(Query *queryTree)
* subquery, or immutable function.
*/
if (rangeTableEntry->rtekind == RTE_RELATION ||
rangeTableEntry->rtekind == RTE_SUBQUERY
#if PG_VERSION_NUM >= PG_VERSION_12
|| rangeTableEntry->rtekind == RTE_RESULT
#endif
rangeTableEntry->rtekind == RTE_SUBQUERY ||
rangeTableEntry->rtekind == RTE_RESULT
)
{
/* accepted */
@ -1472,13 +1463,11 @@ HasRecurringTuples(Node *node, RecurringTuplesType *recurType)
*/
return true;
}
#if PG_VERSION_NUM >= PG_VERSION_12
else if (rangeTableEntry->rtekind == RTE_RESULT)
{
*recurType = RECURRING_TUPLES_EMPTY_JOIN_TREE;
return true;
}
#endif
return false;
}

8
src/backend/distributed/planner/recursive_planning.c
View File

@ -77,11 +77,7 @@
#include "distributed/version_compat.h"
#include "lib/stringinfo.h"
#include "optimizer/clauses.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#else
#include "optimizer/var.h"
#endif
#include "optimizer/planner.h"
#include "optimizer/prep.h"
#include "parser/parsetree.h"
@ -91,11 +87,7 @@
#include "nodes/nodeFuncs.h"
#include "nodes/pg_list.h"
#include "nodes/primnodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#else
#include "nodes/relation.h"
#endif
#include "utils/builtins.h"
#include "utils/guc.h"
#include "utils/lsyscache.h"

6
src/backend/distributed/planner/relation_restriction_equivalence.c
View File

@ -28,14 +28,8 @@
#include "nodes/nodeFuncs.h"
#include "nodes/pg_list.h"
#include "nodes/primnodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#include "optimizer/optimizer.h"
#else
#include "optimizer/cost.h"
#include "nodes/relation.h"
#include "optimizer/var.h"
#endif
#include "nodes/makefuncs.h"
#include "optimizer/paths.h"
#include "parser/parsetree.h"

4
src/backend/distributed/planner/shard_pruning.c
View File

@ -188,16 +188,12 @@ typedef struct PendingPruningInstance
PruningTreeNode *continueAt;
} PendingPruningInstance;
#if PG_VERSION_NUM >= PG_VERSION_12
typedef union \
{ \
FunctionCallInfoBaseData fcinfo; \
/* ensure enough space for nargs args is available */ \
char fcinfo_data[SizeForFunctionCallInfo(2)]; \
} FunctionCall2InfoData;
#else
typedef FunctionCallInfoData FunctionCall2InfoData;
#endif
/*
* We also ignore this warning in ./configure, but that's not always enough.

14
src/backend/distributed/shared_library_init.c
View File

@ -125,7 +125,6 @@ static void CitusAuthHook(Port *port, int status);
static ClientAuthentication_hook_type original_client_auth_hook = NULL;
/* *INDENT-OFF* */
/* GUC enum definitions */
static const struct config_enum_entry propagate_set_commands_options[] = {
@ -1076,12 +1075,13 @@ RegisterCitusConfigVariables(void)
*/
DefineCustomBoolVariable(
"citus.enable_cte_inlining",
gettext_noop("When set to false, CTE inlining feature is disabled"),
gettext_noop("This feature is not intended for users. It is developed "
"to get consistent regression test outputs between Postgres 11"
"and Postgres 12. In Postgres 12+, the user can control the behaviour"
"by [NOT] MATERIALIZED keyword on CTEs. However, in PG 11, we cannot do "
"that."),
gettext_noop("When set to false, CTE inlining feature is disabled."),
gettext_noop(
"This feature is not intended for users and it is deprecated. It is developed "
"to get consistent regression test outputs between Postgres 11"
"and Postgres 12. In Postgres 12+, the user can control the behaviour"
"by [NOT] MATERIALIZED keyword on CTEs. However, in PG 11, we cannot do "
"that."),
&EnableCTEInlining,
true,
PGC_SUSET,

4
src/backend/distributed/test/fake_am.c
View File

@ -21,7 +21,6 @@
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/amapi.h"
#include "access/heapam.h"
@ -586,6 +585,3 @@ fake_am_handler(PG_FUNCTION_ARGS)
{
PG_RETURN_POINTER(&fake_methods);
}
#endif

4
src/backend/distributed/test/fake_fdw.c
View File

@ -25,11 +25,7 @@
#include "nodes/nodes.h"
#include "nodes/pg_list.h"
#include "nodes/plannodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#else
#include "nodes/relation.h"
#endif
#include "optimizer/pathnode.h"
#include "optimizer/planmain.h"
#include "optimizer/restrictinfo.h"

2
src/backend/distributed/test/prune_shard_list.c
View File

@ -28,10 +28,8 @@
#include "distributed/multi_physical_planner.h"
#include "distributed/resource_lock.h"
#include "distributed/shard_pruning.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/makefuncs.h"
#include "nodes/nodeFuncs.h"
#endif
#include "nodes/nodes.h"
#include "nodes/pg_list.h"
#include "nodes/primnodes.h"

4
src/backend/distributed/transaction/backend_data.c
View File

@ -32,9 +32,7 @@
#include "distributed/tuplestore.h"
#include "nodes/execnodes.h"
#include "postmaster/autovacuum.h" /* to access autovacuum_max_workers */
#if PG_VERSION_NUM >= PG_VERSION_12
#include "replication/walsender.h"
#endif
#include "storage/ipc.h"
#include "storage/lmgr.h"
#include "storage/lwlock.h"
@ -598,9 +596,7 @@ TotalProcCount(void)
*/
totalProcs = maxBackends + NUM_AUXILIARY_PROCS + max_prepared_xacts;
#if PG_VERSION_NUM >= PG_VERSION_12
totalProcs += max_wal_senders;
#endif
return totalProcs;
}

2
src/backend/distributed/transaction/transaction_recovery.c
View File

@ -22,9 +22,7 @@
#include <sys/stat.h>
#include <unistd.h>
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/genam.h"
#endif
#include "access/heapam.h"
#include "access/htup_details.h"
#include "access/relscan.h"

2
src/backend/distributed/utils/citus_clauses.c
View File

@ -21,9 +21,7 @@
#include "nodes/nodes.h"
#include "nodes/primnodes.h"
#include "optimizer/clauses.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "optimizer/optimizer.h"
#endif
#include "optimizer/planmain.h"
#include "utils/datum.h"
#include "utils/lsyscache.h"

10
src/backend/distributed/utils/citus_nodefuncs.c
View File

@ -310,13 +310,11 @@ GetRangeTblKind(RangeTblEntry *rte)
case RTE_JOIN:
case RTE_VALUES:
case RTE_CTE:
#if PG_VERSION_NUM >= PG_VERSION_12
case RTE_RESULT:
#endif
{
rteKind = (CitusRTEKind) rte->rtekind;
break;
}
{
rteKind = (CitusRTEKind) rte->rtekind;
break;
}
case RTE_FUNCTION:
{

4
src/backend/distributed/utils/citus_outfuncs.c
View File

@ -34,11 +34,7 @@
#include "distributed/metadata_utility.h"
#include "lib/stringinfo.h"
#include "nodes/plannodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#else
#include "nodes/relation.h"
#endif
#include "utils/datum.h"

10
src/backend/distributed/utils/citus_safe_lib.c
View File

@ -23,17 +23,7 @@
#include "distributed/citus_safe_lib.h"
#include "lib/stringinfo.h"
/*
* In PG 11 pg_vsnprintf is not exported and compiled in most cases, in that
* case use the copied one from pg11_snprintf.c
* NOTE: Whenever removing this section also remove pg11_snprintf.c
*/
#if PG_VERSION_NUM < PG_VERSION_12
extern int pg11_vsnprintf(char *str, size_t count, const char *fmt, va_list args);
#define citus_vsnprintf pg11_vsnprintf
#else
#define citus_vsnprintf pg_vsnprintf
#endif
/*

7
src/backend/distributed/utils/foreign_key_relationship.c
View File

@ -14,14 +14,10 @@
#include "distributed/pg_version_constants.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/genam.h"
#endif
#include "access/htup_details.h"
#include "access/stratnum.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/table.h"
#endif
#include "catalog/pg_constraint.h"
#include "distributed/commands.h"
#include "distributed/foreign_key_relationship.h"
@ -37,9 +33,6 @@
#include "common/hashfn.h"
#endif
#include "utils/memutils.h"
#if PG_VERSION_NUM < PG_VERSION_12
#include "utils/rel.h"
#endif
/*

2
src/backend/distributed/utils/multi_partitioning_utils.c
View File

@ -35,9 +35,7 @@
#include "lib/stringinfo.h"
#include "nodes/pg_list.h"
#include "pgstat.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "partitioning/partdesc.h"
#endif
#include "utils/builtins.h"
#include "utils/fmgroids.h"
#include "utils/lsyscache.h"

1184
src/backend/distributed/utils/pg11_snprintf.c
View File

File diff suppressed because it is too large Load Diff

15
src/backend/distributed/worker/worker_data_fetch_protocol.c
View File

@ -81,7 +81,6 @@ PG_FUNCTION_INFO_V1(worker_append_table_to_shard);
* Following UDFs are stub functions, you can check their comments for more
* detail.
*/
PG_FUNCTION_INFO_V1(worker_fetch_query_results_file);
PG_FUNCTION_INFO_V1(worker_fetch_regular_table);
PG_FUNCTION_INFO_V1(worker_fetch_foreign_file);
PG_FUNCTION_INFO_V1(master_expire_table_cache);
@ -814,20 +813,6 @@ SetDefElemArg(AlterSeqStmt *statement, const char *name, Node *arg)
}
/*
* worker_fetch_query_results_file is a stub UDF to allow the function object
* to be re-created during upgrades. We should keep this around until we drop
* support for Postgres 11, since Postgres 11 is the highest version for which
* this object may have been created.
*/
Datum
worker_fetch_query_results_file(PG_FUNCTION_ARGS)
{
ereport(DEBUG2, (errmsg("this function is deprecated and no longer is used")));
PG_RETURN_VOID();
}
/*
* worker_fetch_regular_table UDF is a stub UDF to install Citus flawlessly.
* Otherwise we need to delete them from our sql files, which is confusing

20
src/backend/distributed/worker/worker_merge_protocol.c
View File

@ -20,10 +20,8 @@
#include "funcapi.h"
#include "miscadmin.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "access/genam.h"
#include "access/table.h"
#endif
#include "access/htup_details.h"
#include "access/xact.h"
#include "catalog/dependency.h"
@ -266,26 +264,16 @@ worker_merge_files_and_run_query(PG_FUNCTION_ARGS)
Datum
worker_cleanup_job_schema_cache(PG_FUNCTION_ARGS)
{
Relation pgNamespace = NULL;
#if PG_VERSION_NUM >= PG_VERSION_12
TableScanDesc scanDescriptor = NULL;
#else
HeapScanDesc scanDescriptor = NULL;
#endif
ScanKey scanKey = NULL;
int scanKeyCount = 0;
HeapTuple heapTuple = NULL;
CheckCitusVersion(ERROR);
pgNamespace = table_open(NamespaceRelationId, AccessExclusiveLock);
#if PG_VERSION_NUM >= PG_VERSION_12
scanDescriptor = table_beginscan_catalog(pgNamespace, scanKeyCount, scanKey);
#else
scanDescriptor = heap_beginscan_catalog(pgNamespace, scanKeyCount, scanKey);
#endif
Relation pgNamespace = table_open(NamespaceRelationId, AccessExclusiveLock);
TableScanDesc scanDescriptor = table_beginscan_catalog(pgNamespace, scanKeyCount,
scanKey);
heapTuple = heap_getnext(scanDescriptor, ForwardScanDirection);
HeapTuple heapTuple = heap_getnext(scanDescriptor, ForwardScanDirection);
while (HeapTupleIsValid(heapTuple))
{
Form_pg_namespace schemaForm = (Form_pg_namespace) GETSTRUCT(heapTuple);

11
src/include/columnar/columnar_version_compat.h
View File

@ -17,17 +17,6 @@
#define ExplainPropertyLong(qlabel, value, es) \
ExplainPropertyInteger(qlabel, NULL, value, es)
#if PG_VERSION_NUM < 120000
#define TTS_EMPTY(slot) ((slot)->tts_isempty)
#define ExecForceStoreHeapTuple(tuple, slot, shouldFree) \
ExecStoreTuple(newTuple, tupleSlot, InvalidBuffer, shouldFree);
#define table_open(r, l) heap_open(r, l)
#define table_close(r, l) heap_close(r, l)
#define TableScanDesc HeapScanDesc
#define table_beginscan heap_beginscan
#define table_endscan heap_endscan
#endif
#if PG_VERSION_NUM < 130000
#define detoast_attr(X) heap_tuple_untoast_attr(X)
#endif

4
src/include/distributed/combine_query_planner.h
View File

@ -16,11 +16,7 @@
#include "nodes/parsenodes.h"
#include "nodes/plannodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#else
#include "nodes/relation.h"
#endif
/* Function declarations for building local plans on the coordinator node */

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

@ -408,7 +408,6 @@ extern List * PreprocessCreateEnumStmt(Node *stmt, const char *queryString,
extern List * PostprocessCreateEnumStmt(Node *stmt, const char *queryString);
extern List * PreprocessAlterEnumStmt(Node *stmt, const char *queryString,
ProcessUtilityContext processUtilityContext);
extern List * PostprocessAlterEnumStmt(Node *stmt, const char *queryString);
extern List * PreprocessDropTypeStmt(Node *stmt, const char *queryString,
ProcessUtilityContext processUtilityContext);
extern List * PreprocessRenameTypeStmt(Node *stmt, const char *queryString,

4
src/include/distributed/distributed_planner.h
View File

@ -16,11 +16,7 @@
#include "nodes/plannodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#else
#include "nodes/relation.h"
#endif
#include "distributed/citus_nodes.h"
#include "distributed/errormessage.h"

20
src/include/distributed/hash_helpers.h
View File

@ -15,26 +15,6 @@
#include "utils/hsearch.h"
/* pg12 includes this exact implementation of hash_combine */
#if PG_VERSION_NUM < PG_VERSION_12
/*
* Combine two hash values, resulting in another hash value, with decent bit
* mixing.
*
* Similar to boost's hash_combine().
*/
static inline uint32
hash_combine(uint32 a, uint32 b)
{
a ^= b + 0x9e3779b9 + (a << 6) + (a >> 2);
return a;
}
#endif
extern void hash_delete_all(HTAB *htab);
/*

2
src/include/distributed/multi_physical_planner.h
View File

@ -58,9 +58,7 @@ typedef enum CitusRTEKind
CITUS_RTE_VALUES = RTE_VALUES, /* VALUES (<exprlist>), (<exprlist>), ... */
CITUS_RTE_CTE = RTE_CTE, /* common table expr (WITH list element) */
CITUS_RTE_NAMEDTUPLESTORE = RTE_NAMEDTUPLESTORE, /* tuplestore, e.g. for triggers */
#if (PG_VERSION_NUM >= PG_VERSION_12)
CITUS_RTE_RESULT = RTE_RESULT, /* RTE represents an empty FROM clause */
#endif
CITUS_RTE_SHARD,
CITUS_RTE_REMOTE_QUERY
} CitusRTEKind;

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

@ -11,7 +11,6 @@
#ifndef PG_VERSION_CONSTANTS
#define PG_VERSION_CONSTANTS
#define PG_VERSION_11 110000
#define PG_VERSION_12 120000
#define PG_VERSION_13 130000
#define PG_VERSION_14 140000

4
src/include/distributed/recursive_planning.h
View File

@ -16,11 +16,7 @@
#include "distributed/relation_restriction_equivalence.h"
#include "nodes/pg_list.h"
#include "nodes/primnodes.h"
#if PG_VERSION_NUM >= PG_VERSION_12
#include "nodes/pathnodes.h"
#else
#include "nodes/relation.h"
#endif
typedef struct RecursivePlanningContextInternal RecursivePlanningContext;

83
src/include/distributed/version_compat.h
View File

@ -24,9 +24,7 @@
#include "executor/tuptable.h"
#include "nodes/parsenodes.h"
#include "parser/parse_func.h"
#if (PG_VERSION_NUM >= PG_VERSION_12)
#include "optimizer/optimizer.h"
#endif
#if (PG_VERSION_NUM >= PG_VERSION_13)
#include "tcop/tcopprot.h"
@ -125,87 +123,6 @@ FileCompatFromFileStart(File fileDesc)
}
#else /* pre PG12 */
#define CreateTableSlotForRel(rel) MakeSingleTupleTableSlot(RelationGetDescr(rel))
#define table_open(r, l) heap_open(r, l)
#define table_openrv(r, l) heap_openrv(r, l)
#define table_openrv_extended(r, l, m) heap_openrv_extended(r, l, m)
#define table_close(r, l) heap_close(r, l)
#define QTW_EXAMINE_RTES_BEFORE QTW_EXAMINE_RTES
#define MakeSingleTupleTableSlotCompat(tupleDesc, tts_opts) \
MakeSingleTupleTableSlot(tupleDesc)
#define NextCopyFromCompat(cstate, econtext, values, nulls) \
NextCopyFrom(cstate, econtext, values, nulls, NULL)
/*
* In PG12 GetSysCacheOid requires an oid column,
* whereas beforehand the oid column was implicit with WITH OIDS
*/
#define GetSysCacheOid1Compat(cacheId, oidcol, key1) \
GetSysCacheOid1(cacheId, key1)
#define GetSysCacheOid2Compat(cacheId, oidcol, key1, key2) \
GetSysCacheOid2(cacheId, key1, key2)
#define GetSysCacheOid3Compat(cacheId, oidcol, key1, key2, key3) \
GetSysCacheOid3(cacheId, key1, key2, key3)
#define GetSysCacheOid4Compat(cacheId, oidcol, key1, key2, key3, key4) \
GetSysCacheOid4(cacheId, key1, key2, key3, key4)
#define LOCAL_FCINFO(name, nargs) \
FunctionCallInfoData name ## data; \
FunctionCallInfoData *name = &name ## data
#define fcGetArgValue(fc, n) ((fc)->arg[n])
#define fcGetArgNull(fc, n) ((fc)->argnull[n])
#define fcSetArgExt(fc, n, val, is_null) \
(((fc)->argnull[n] = (is_null)), ((fc)->arg[n] = (val)))
typedef struct
{
File fd;
} FileCompat;
static inline int
FileWriteCompat(FileCompat *file, char *buffer, int amount, uint32 wait_event_info)
{
return FileWrite(file->fd, buffer, amount, wait_event_info);
}
static inline int
FileReadCompat(FileCompat *file, char *buffer, int amount, uint32 wait_event_info)
{
return FileRead(file->fd, buffer, amount, wait_event_info);
}
static inline FileCompat
FileCompatFromFileStart(File fileDesc)
{
FileCompat fc = {
.fd = fileDesc,
};
return fc;
}
/*
* postgres 11 equivalent for a function with the same name in postgres 12+.
*/
static inline bool
table_scan_getnextslot(HeapScanDesc scan, ScanDirection dir, TupleTableSlot *slot)
{
HeapTuple tuple = heap_getnext(scan, ForwardScanDirection);
if (tuple == NULL)
{
return false;
}
ExecStoreTuple(tuple, slot, InvalidBuffer, false);
return true;
}
#endif /* PG12 */
#define fcSetArg(fc, n, value) fcSetArgExt(fc, n, value, false)

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

@ -144,7 +144,6 @@ extern Node * ParseTreeRawStmt(const char *ddlCommand);
/* Function declarations for applying distributed execution primitives */
extern Datum worker_fetch_partition_file(PG_FUNCTION_ARGS);
extern Datum worker_fetch_query_results_file(PG_FUNCTION_ARGS);
extern Datum worker_apply_shard_ddl_command(PG_FUNCTION_ARGS);
extern Datum worker_range_partition_table(PG_FUNCTION_ARGS);
extern Datum worker_hash_partition_table(PG_FUNCTION_ARGS);

13
src/test/regress/bin/normalize.sed
View File

@ -43,12 +43,6 @@ s/"citus_local_table_([0-9]+)_[0-9]+"/"citus_local_table_\1_xxxxxxx"/g
# normalize relation oid suffix for the truncate triggers created by citus
s/truncate_trigger_[0-9]+/truncate_trigger_xxxxxxx/g
# (citus_table_triggers.sql)
# postgres generates create trigger commands for triggers with:
# "EXECUTE FUNCTION" in pg12
# "EXECUTE PROCEDURE" in pg11
s/FOR EACH (ROW|STATEMENT)(.*)EXECUTE PROCEDURE/FOR EACH \1\2EXECUTE FUNCTION/g
# In foreign_key_restriction_enforcement, normalize shard names
s/"(on_update_fkey_table_|fkey_)[0-9]+"/"\1xxxxxxx"/g
@ -116,9 +110,6 @@ s/partition ".*" would be violated by some row/partition would be violated by so
/.*Peak Memory Usage:.*$/d
s/of relation ".*" contains null values/contains null values/g
s/of relation "t1" is violated by some row/is violated by some row/g
# can be removed when we remove PG_VERSION_NUM >= 120000
s/(.*)Output:.*$/\1Output: xxxxxx/g
# intermediate_results
s/(ERROR.*)pgsql_job_cache\/([0-9]+_[0-9]+_[0-9]+)\/(.*).data/\1pgsql_job_cache\/xx_x_xxx\/\3.data/g
@ -145,10 +136,6 @@ s/repartitioned_results_[0-9]+/repartitioned_results_xxxxx/g
# ignore job id in worker_hash_partition_table
s/worker_hash_partition_table \([0-9]+/worker_hash_partition_table \(xxxxxxx/g
# ignore first parameter for citus_extradata_container due to differences between pg11 and pg12
# can be removed when we remove PG_VERSION_NUM >= 120000
s/pg_catalog.citus_extradata_container\([0-9]+/pg_catalog.citus_extradata_container\(XXX/g
# ignore referene table replication messages
/replicating reference table.*$/d

900
src/test/regress/expected/alter_distributed_table_0.out
View File

@ -1,900 +0,0 @@
SHOW server_version \gset
SELECT substring(:'server_version', '\d+')::int > 11 AS server_version_above_eleven;
server_version_above_eleven
---------------------------------------------------------------------
f
(1 row)
\gset
CREATE SCHEMA alter_distributed_table;
SET search_path TO alter_distributed_table;
SET citus.shard_count TO 4;
SET citus.shard_replication_factor TO 1;
CREATE TABLE dist_table (a INT, b INT);
SELECT create_distributed_table ('dist_table', 'a', colocate_with := 'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO dist_table VALUES (1, 1), (2, 2), (3, 3);
CREATE TABLE colocation_table (a INT, b INT);
SELECT create_distributed_table ('colocation_table', 'a', colocate_with := 'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE colocation_table_2 (a INT, b INT);
SELECT create_distributed_table ('colocation_table_2', 'a', colocate_with := 'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT table_name, citus_table_type, distribution_column, shard_count FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2');
table_name | citus_table_type | distribution_column | shard_count
---------------------------------------------------------------------
colocation_table | distributed | a | 4
colocation_table_2 | distributed | a | 4
dist_table | distributed | a | 4
(3 rows)
SELECT STRING_AGG(table_name::text, ', ' ORDER BY 1) AS "Colocation Groups" FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2') GROUP BY colocation_id ORDER BY 1;
Colocation Groups
---------------------------------------------------------------------
colocation_table
colocation_table_2
dist_table
(3 rows)
-- test altering distribution column
SELECT alter_distributed_table('dist_table', distribution_column := 'b');
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT table_name, citus_table_type, distribution_column, shard_count FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2');
table_name | citus_table_type | distribution_column | shard_count
---------------------------------------------------------------------
colocation_table | distributed | a | 4
colocation_table_2 | distributed | a | 4
dist_table | distributed | b | 4
(3 rows)
SELECT STRING_AGG(table_name::text, ', ' ORDER BY 1) AS "Colocation Groups" FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2') GROUP BY colocation_id ORDER BY 1;
Colocation Groups
---------------------------------------------------------------------
colocation_table
colocation_table_2
dist_table
(3 rows)
-- test altering shard count
SELECT alter_distributed_table('dist_table', shard_count := 6);
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT table_name, citus_table_type, distribution_column, shard_count FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2');
table_name | citus_table_type | distribution_column | shard_count
---------------------------------------------------------------------
colocation_table | distributed | a | 4
colocation_table_2 | distributed | a | 4
dist_table | distributed | b | 6
(3 rows)
SELECT STRING_AGG(table_name::text, ', ' ORDER BY 1) AS "Colocation Groups" FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2') GROUP BY colocation_id ORDER BY 1;
Colocation Groups
---------------------------------------------------------------------
colocation_table
colocation_table_2
dist_table
(3 rows)
-- test altering colocation, note that shard count will also change
SELECT alter_distributed_table('dist_table', colocate_with := 'alter_distributed_table.colocation_table');
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT table_name, citus_table_type, distribution_column, shard_count FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2');
table_name | citus_table_type | distribution_column | shard_count
---------------------------------------------------------------------
colocation_table | distributed | a | 4
colocation_table_2 | distributed | a | 4
dist_table | distributed | b | 4
(3 rows)
SELECT STRING_AGG(table_name::text, ', ' ORDER BY 1) AS "Colocation Groups" FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2') GROUP BY colocation_id ORDER BY 1;
Colocation Groups
---------------------------------------------------------------------
colocation_table, dist_table
colocation_table_2
(2 rows)
-- test altering shard count with cascading, note that the colocation will be kept
SELECT alter_distributed_table('dist_table', shard_count := 8, cascade_to_colocated := true);
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
NOTICE: creating a new table for alter_distributed_table.colocation_table
NOTICE: moving the data of alter_distributed_table.colocation_table
NOTICE: dropping the old alter_distributed_table.colocation_table
NOTICE: renaming the new table to alter_distributed_table.colocation_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT table_name, citus_table_type, distribution_column, shard_count FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2');
table_name | citus_table_type | distribution_column | shard_count
---------------------------------------------------------------------
colocation_table | distributed | a | 8
colocation_table_2 | distributed | a | 4
dist_table | distributed | b | 8
(3 rows)
SELECT STRING_AGG(table_name::text, ', ' ORDER BY 1) AS "Colocation Groups" FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2') GROUP BY colocation_id ORDER BY 1;
Colocation Groups
---------------------------------------------------------------------
colocation_table, dist_table
colocation_table_2
(2 rows)
-- test altering shard count without cascading, note that the colocation will be broken
SELECT alter_distributed_table('dist_table', shard_count := 10, cascade_to_colocated := false);
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT table_name, citus_table_type, distribution_column, shard_count FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2');
table_name | citus_table_type | distribution_column | shard_count
---------------------------------------------------------------------
colocation_table | distributed | a | 8
colocation_table_2 | distributed | a | 4
dist_table | distributed | b | 10
(3 rows)
SELECT STRING_AGG(table_name::text, ', ' ORDER BY 1) AS "Colocation Groups" FROM public.citus_tables
WHERE table_name IN ('dist_table', 'colocation_table', 'colocation_table_2') GROUP BY colocation_id ORDER BY 1;
Colocation Groups
---------------------------------------------------------------------
colocation_table
colocation_table_2
dist_table
(3 rows)
-- test partitions
CREATE TABLE partitioned_table (id INT, a INT) PARTITION BY RANGE (id);
SELECT create_distributed_table('partitioned_table', 'id', colocate_with := 'none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE partitioned_table_1_5 PARTITION OF partitioned_table FOR VALUES FROM (1) TO (5);
CREATE TABLE partitioned_table_6_10 PARTITION OF partitioned_table FOR VALUES FROM (6) TO (10);
INSERT INTO partitioned_table VALUES (2, 12), (7, 2);
SELECT logicalrelid::text FROM pg_dist_partition WHERE logicalrelid::regclass::text LIKE 'partitioned\_table%' ORDER BY 1;
logicalrelid
---------------------------------------------------------------------
partitioned_table
partitioned_table_1_5
partitioned_table_6_10
(3 rows)
SELECT run_command_on_workers($$SELECT COUNT(*) FROM pg_catalog.pg_class WHERE relname LIKE 'partitioned\_table%'$$);
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,6)
(localhost,57638,t,6)
(2 rows)
SELECT inhrelid::regclass::text FROM pg_catalog.pg_inherits WHERE inhparent = 'partitioned_table'::regclass ORDER BY 1;
inhrelid
---------------------------------------------------------------------
partitioned_table_1_5
partitioned_table_6_10
(2 rows)
SELECT table_name::text, distribution_column, shard_count FROM public.citus_tables WHERE table_name::text LIKE 'partitioned\_table%' ORDER BY 1;
table_name | distribution_column | shard_count
---------------------------------------------------------------------
partitioned_table | id | 4
partitioned_table_1_5 | id | 4
partitioned_table_6_10 | id | 4
(3 rows)
SELECT * FROM partitioned_table ORDER BY 1, 2;
id | a
---------------------------------------------------------------------
2 | 12
7 | 2
(2 rows)
SELECT * FROM partitioned_table_1_5 ORDER BY 1, 2;
id | a
---------------------------------------------------------------------
2 | 12
(1 row)
SELECT * FROM partitioned_table_6_10 ORDER BY 1, 2;
id | a
---------------------------------------------------------------------
7 | 2
(1 row)
-- test altering the parent table
SELECT alter_distributed_table('partitioned_table', shard_count := 10, distribution_column := 'a');
NOTICE: converting the partitions of alter_distributed_table.partitioned_table
NOTICE: creating a new table for alter_distributed_table.partitioned_table_1_5
NOTICE: moving the data of alter_distributed_table.partitioned_table_1_5
NOTICE: dropping the old alter_distributed_table.partitioned_table_1_5
NOTICE: renaming the new table to alter_distributed_table.partitioned_table_1_5
NOTICE: creating a new table for alter_distributed_table.partitioned_table_6_10
NOTICE: moving the data of alter_distributed_table.partitioned_table_6_10
NOTICE: dropping the old alter_distributed_table.partitioned_table_6_10
NOTICE: renaming the new table to alter_distributed_table.partitioned_table_6_10
NOTICE: creating a new table for alter_distributed_table.partitioned_table
NOTICE: dropping the old alter_distributed_table.partitioned_table
NOTICE: renaming the new table to alter_distributed_table.partitioned_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- test altering the partition
SELECT alter_distributed_table('partitioned_table_1_5', shard_count := 10, distribution_column := 'a');
ERROR: cannot complete operation because table is a partition
HINT: the parent table is "partitioned_table"
SELECT logicalrelid::text FROM pg_dist_partition WHERE logicalrelid::regclass::text LIKE 'partitioned\_table%' ORDER BY 1;
logicalrelid
---------------------------------------------------------------------
partitioned_table
partitioned_table_1_5
partitioned_table_6_10
(3 rows)
SELECT run_command_on_workers($$SELECT COUNT(*) FROM pg_catalog.pg_class WHERE relname LIKE 'partitioned\_table%'$$);
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,15)
(localhost,57638,t,15)
(2 rows)
SELECT inhrelid::regclass::text FROM pg_catalog.pg_inherits WHERE inhparent = 'partitioned_table'::regclass ORDER BY 1;
inhrelid
---------------------------------------------------------------------
partitioned_table_1_5
partitioned_table_6_10
(2 rows)
SELECT table_name::text, distribution_column, shard_count FROM public.citus_tables WHERE table_name::text LIKE 'partitioned\_table%' ORDER BY 1;
table_name | distribution_column | shard_count
---------------------------------------------------------------------
partitioned_table | a | 10
partitioned_table_1_5 | a | 10
partitioned_table_6_10 | a | 10
(3 rows)
SELECT * FROM partitioned_table ORDER BY 1, 2;
id | a
---------------------------------------------------------------------
2 | 12
7 | 2
(2 rows)
SELECT * FROM partitioned_table_1_5 ORDER BY 1, 2;
id | a
---------------------------------------------------------------------
2 | 12
(1 row)
SELECT * FROM partitioned_table_6_10 ORDER BY 1, 2;
id | a
---------------------------------------------------------------------
7 | 2
(1 row)
-- test references
CREATE TABLE referenced_dist_table (a INT UNIQUE);
CREATE TABLE referenced_ref_table (a INT UNIQUE);
CREATE TABLE table_with_references (a1 INT UNIQUE REFERENCES referenced_dist_table(a), a2 INT REFERENCES referenced_ref_table(a));
CREATE TABLE referencing_dist_table (a INT REFERENCES table_with_references(a1));
SELECT create_distributed_table('referenced_dist_table', 'a', colocate_with:='none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_reference_table('referenced_ref_table');
create_reference_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('table_with_references', 'a1', colocate_with:='referenced_dist_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('referencing_dist_table', 'a', colocate_with:='referenced_dist_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SET client_min_messages TO WARNING;
SELECT conrelid::regclass::text AS "Referencing Table", pg_get_constraintdef(oid, true) AS "Definition" FROM pg_constraint
WHERE (conrelid::regclass::text = 'table_with_references' OR confrelid::regclass::text = 'table_with_references') AND contype = 'f' ORDER BY 1;
Referencing Table | Definition
---------------------------------------------------------------------
referencing_dist_table | FOREIGN KEY (a) REFERENCES table_with_references(a1)
table_with_references | FOREIGN KEY (a1) REFERENCES referenced_dist_table(a)
table_with_references | FOREIGN KEY (a2) REFERENCES referenced_ref_table(a)
(3 rows)
SELECT alter_distributed_table('table_with_references', shard_count := 12, cascade_to_colocated := true);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT conrelid::regclass::text AS "Referencing Table", pg_get_constraintdef(oid, true) AS "Definition" FROM pg_constraint
WHERE (conrelid::regclass::text = 'table_with_references' OR confrelid::regclass::text = 'table_with_references') AND contype = 'f' ORDER BY 1;
Referencing Table | Definition
---------------------------------------------------------------------
referencing_dist_table | FOREIGN KEY (a) REFERENCES table_with_references(a1)
table_with_references | FOREIGN KEY (a2) REFERENCES referenced_ref_table(a)
table_with_references | FOREIGN KEY (a1) REFERENCES referenced_dist_table(a)
(3 rows)
SELECT alter_distributed_table('table_with_references', shard_count := 10, cascade_to_colocated := false);
WARNING: foreign key table_with_references_a1_fkey will be dropped
WARNING: foreign key referencing_dist_table_a_fkey will be dropped
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT conrelid::regclass::text AS "Referencing Table", pg_get_constraintdef(oid, true) AS "Definition" FROM pg_constraint
WHERE (conrelid::regclass::text = 'table_with_references' OR confrelid::regclass::text = 'table_with_references') AND contype = 'f' ORDER BY 1;
Referencing Table | Definition
---------------------------------------------------------------------
table_with_references | FOREIGN KEY (a2) REFERENCES referenced_ref_table(a)
(1 row)
-- check when multi shard modify mode is set to sequential
SELECT alter_distributed_table('referenced_dist_table', colocate_with:='none');
alter_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE ref_to_dist_table(a INT REFERENCES referenced_dist_table(a));
CREATE TABLE ref_to_ref_table(a INT REFERENCES referenced_ref_table(a));
SELECT create_distributed_table('ref_to_dist_table', 'a', colocate_with:='referenced_dist_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('ref_to_ref_table', 'a', colocate_with:='none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- alter a table referencing a reference table
SELECT alter_distributed_table('ref_to_ref_table', shard_count:=6);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- let's create a table that is not colocated with a table that references a reference table
CREATE TABLE col_with_ref_to_dist (a INT);
SELECT create_distributed_table('col_with_ref_to_dist', 'a', colocate_with:='ref_to_dist_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- and create a table colocated with a table that references a reference table
CREATE TABLE col_with_ref_to_ref (a INT);
SELECT alter_distributed_table('ref_to_ref_table', colocate_with:='none');
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('col_with_ref_to_ref', 'a', colocate_with:='ref_to_ref_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
-- alter a table colocated with a table referencing a reference table with cascading
SELECT alter_distributed_table('col_with_ref_to_ref', shard_count:=8, cascade_to_colocated:=true);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- alter a table colocated with a table referencing a reference table without cascading
SELECT alter_distributed_table('col_with_ref_to_ref', shard_count:=10, cascade_to_colocated:=false);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- alter a table not colocated with a table referencing a reference table with cascading
SELECT alter_distributed_table('col_with_ref_to_dist', shard_count:=6, cascade_to_colocated:=true);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
\if :server_version_above_eleven
-- test altering columnar table
CREATE TABLE columnar_table (a INT) USING columnar;
SELECT create_distributed_table('columnar_table', 'a', colocate_with:='none');
SELECT table_name::text, shard_count, access_method FROM public.citus_tables WHERE table_name::text = 'columnar_table';
SELECT alter_distributed_table('columnar_table', shard_count:=6);
SELECT table_name::text, shard_count, access_method FROM public.citus_tables WHERE table_name::text = 'columnar_table';
\endif
-- test with metadata sync
SET citus.replication_model TO 'streaming';
SELECT start_metadata_sync_to_node('localhost', :worker_1_port);
start_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
CREATE TABLE metadata_sync_table (a BIGSERIAL);
SELECT create_distributed_table('metadata_sync_table', 'a', colocate_with:='none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('metadata_sync_table', shard_count:=6);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('metadata_sync_table', shard_count:=8);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT table_name, shard_count FROM public.citus_tables WHERE table_name::text = 'metadata_sync_table';
table_name | shard_count
---------------------------------------------------------------------
metadata_sync_table | 8
(1 row)
SET citus.replication_model TO DEFAULT;
SELECT stop_metadata_sync_to_node('localhost', :worker_1_port);
stop_metadata_sync_to_node
---------------------------------------------------------------------
(1 row)
-- test complex cascade operations
CREATE TABLE cas_1 (a INT UNIQUE);
CREATE TABLE cas_2 (a INT UNIQUE);
CREATE TABLE cas_3 (a INT UNIQUE);
CREATE TABLE cas_4 (a INT UNIQUE);
CREATE TABLE cas_par (a INT UNIQUE) PARTITION BY RANGE(a);
CREATE TABLE cas_par_1 PARTITION OF cas_par FOR VALUES FROM (1) TO (4);
CREATE TABLE cas_par_2 PARTITION OF cas_par FOR VALUES FROM (5) TO (8);
CREATE TABLE cas_col (a INT UNIQUE);
-- add foreign keys from and to partitions
ALTER TABLE cas_par_1 ADD CONSTRAINT fkey_from_par_1 FOREIGN KEY (a) REFERENCES cas_1(a);
ALTER TABLE cas_2 ADD CONSTRAINT fkey_to_par_1 FOREIGN KEY (a) REFERENCES cas_par_1(a);
ALTER TABLE cas_par ADD CONSTRAINT fkey_from_par FOREIGN KEY (a) REFERENCES cas_3(a);
ALTER TABLE cas_4 ADD CONSTRAINT fkey_to_par FOREIGN KEY (a) REFERENCES cas_par(a);
ERROR: cannot reference partitioned table "cas_par"
-- distribute all the tables
SELECT create_distributed_table('cas_1', 'a', colocate_with:='none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('cas_3', 'a', colocate_with:='cas_1');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('cas_par', 'a', colocate_with:='cas_1');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('cas_2', 'a', colocate_with:='cas_1');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('cas_4', 'a', colocate_with:='cas_1');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('cas_col', 'a', colocate_with:='cas_1');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT conrelid::regclass::text AS "Referencing Table", pg_get_constraintdef(oid, true) AS "Definition" FROM pg_constraint
WHERE (conrelid::regclass::text = 'cas_par_1' OR confrelid::regclass::text = 'cas_par_1') ORDER BY 1, 2;
Referencing Table | Definition
---------------------------------------------------------------------
cas_2 | FOREIGN KEY (a) REFERENCES cas_par_1(a)
cas_par_1 | FOREIGN KEY (a) REFERENCES cas_1(a)
cas_par_1 | FOREIGN KEY (a) REFERENCES cas_3(a)
cas_par_1 | UNIQUE (a)
(4 rows)
SELECT inhrelid::regclass::text FROM pg_catalog.pg_inherits WHERE inhparent = 'cas_par'::regclass ORDER BY 1;
inhrelid
---------------------------------------------------------------------
cas_par_1
cas_par_2
(2 rows)
-- alter the cas_col and cascade the change
SELECT alter_distributed_table('cas_col', shard_count:=6, cascade_to_colocated:=true);
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT conrelid::regclass::text AS "Referencing Table", pg_get_constraintdef(oid, true) AS "Definition" FROM pg_constraint
WHERE (conrelid::regclass::text = 'cas_par_1' OR confrelid::regclass::text = 'cas_par_1') ORDER BY 1, 2;
Referencing Table | Definition
---------------------------------------------------------------------
cas_2 | FOREIGN KEY (a) REFERENCES cas_par_1(a)
cas_par_1 | FOREIGN KEY (a) REFERENCES cas_1(a)
cas_par_1 | FOREIGN KEY (a) REFERENCES cas_3(a)
cas_par_1 | UNIQUE (a)
(4 rows)
SELECT inhrelid::regclass::text FROM pg_catalog.pg_inherits WHERE inhparent = 'cas_par'::regclass ORDER BY 1;
inhrelid
---------------------------------------------------------------------
cas_par_1
cas_par_2
(2 rows)
SET client_min_messages TO DEFAULT;
-- test changing dist column and colocating partitioned table without changing shard count
CREATE TABLE col_table (a INT);
SELECT create_distributed_table('col_table', 'a', colocate_with:='none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE par_table (a BIGINT, b INT) PARTITION BY RANGE (a);
SELECT create_distributed_table('par_table', 'a', colocate_with:='none');
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE par_table_1 (a BIGINT, b INT);
SELECT create_distributed_table('par_table_1', 'a', colocate_with:='par_table');
create_distributed_table
---------------------------------------------------------------------
(1 row)
ALTER TABLE par_table ATTACH PARTITION par_table_1 FOR VALUES FROM (1) TO (5);
SELECT alter_distributed_table('par_table', distribution_column:='b', colocate_with:='col_table');
NOTICE: converting the partitions of alter_distributed_table.par_table
NOTICE: creating a new table for alter_distributed_table.par_table_1
NOTICE: moving the data of alter_distributed_table.par_table_1
NOTICE: dropping the old alter_distributed_table.par_table_1
NOTICE: renaming the new table to alter_distributed_table.par_table_1
NOTICE: creating a new table for alter_distributed_table.par_table
NOTICE: dropping the old alter_distributed_table.par_table
NOTICE: renaming the new table to alter_distributed_table.par_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- test messages
-- test nothing to change
SELECT alter_distributed_table('dist_table');
ERROR: you have to specify at least one of the distribution_column, shard_count or colocate_with parameters
SELECT alter_distributed_table('dist_table', cascade_to_colocated := false);
ERROR: you have to specify at least one of the distribution_column, shard_count or colocate_with parameters
-- no operation UDF calls
SELECT alter_distributed_table('dist_table', distribution_column := 'b');
ERROR: this call doesn't change any properties of the table
HINT: check citus_tables view to see current properties of the table
SELECT alter_distributed_table('dist_table', shard_count := 10);
ERROR: this call doesn't change any properties of the table
HINT: check citus_tables view to see current properties of the table
-- first colocate the tables, then try to re-colococate
SELECT alter_distributed_table('dist_table', colocate_with := 'colocation_table');
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('dist_table', colocate_with := 'colocation_table');
ERROR: this call doesn't change any properties of the table
HINT: check citus_tables view to see current properties of the table
-- test some changes while keeping others same
-- shouldn't error but should have notices about no-change parameters
SELECT alter_distributed_table('dist_table', distribution_column:='b', shard_count:=4, cascade_to_colocated:=false);
NOTICE: table is already distributed by b
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('dist_table', shard_count:=4, colocate_with:='colocation_table_2');
NOTICE: shard count of the table is already 4
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('dist_table', colocate_with:='colocation_table_2', distribution_column:='a');
NOTICE: table is already colocated with colocation_table_2
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- test cascading distribution column, should error
SELECT alter_distributed_table('dist_table', distribution_column := 'b', cascade_to_colocated := true);
ERROR: distribution_column cannot be cascaded to colocated tables
SELECT alter_distributed_table('dist_table', distribution_column := 'b', shard_count:=12, colocate_with:='colocation_table_2', cascade_to_colocated := true);
ERROR: distribution_column cannot be cascaded to colocated tables
-- test nothing to cascade
SELECT alter_distributed_table('dist_table', cascade_to_colocated := true);
ERROR: shard_count or colocate_with is necessary for cascading to colocated tables
-- test cascading colocate_with := 'none'
SELECT alter_distributed_table('dist_table', colocate_with := 'none', cascade_to_colocated := true);
ERROR: colocate_with := 'none' cannot be cascaded to colocated tables
-- test changing shard count of a colocated table without cascade_to_colocated, should error
SELECT alter_distributed_table('dist_table', shard_count := 14);
ERROR: cascade_to_colocated parameter is necessary
DETAIL: this table is colocated with some other tables
HINT: cascade_to_colocated := false will break the current colocation, cascade_to_colocated := true will change the shard count of colocated tables too.
-- test changing shard count of a non-colocated table without cascade_to_colocated, shouldn't error
SELECT alter_distributed_table('dist_table', colocate_with := 'none');
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('dist_table', shard_count := 14);
NOTICE: creating a new table for alter_distributed_table.dist_table
NOTICE: moving the data of alter_distributed_table.dist_table
NOTICE: dropping the old alter_distributed_table.dist_table
NOTICE: renaming the new table to alter_distributed_table.dist_table
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- test altering a table into colocating with a table but giving a different shard count
SELECT alter_distributed_table('dist_table', colocate_with := 'colocation_table', shard_count := 16);
ERROR: shard_count cannot be different than the shard count of the table in colocate_with
HINT: if no shard_count is specified shard count will be same with colocate_with table's
-- test colocation with distribution columns with different data types
CREATE TABLE different_type_table (a TEXT);
SELECT create_distributed_table('different_type_table', 'a');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('dist_table', colocate_with := 'different_type_table');
ERROR: cannot colocate with different_type_table because data type of its distribution column is different than dist_table
SELECT alter_distributed_table('dist_table', distribution_column := 'a', colocate_with := 'different_type_table');
ERROR: cannot colocate with different_type_table and change distribution column to a because data type of column a is different then the distribution column of the different_type_table
-- test shard_count := 0
SELECT alter_distributed_table('dist_table', shard_count := 0);
ERROR: shard_count cannot be 0
HINT: if you no longer want this to be a distributed table you can try undistribute_table() function
-- test colocating with non-distributed table
CREATE TABLE reference_table (a INT);
SELECT create_reference_table('reference_table');
create_reference_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('dist_table', colocate_with:='reference_table');
ERROR: cannot colocate with reference_table because it is not a distributed table
-- test append table
CREATE TABLE append_table (a INT);
SELECT create_distributed_table('append_table', 'a', 'append');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('append_table', shard_count:=6);
ERROR: relation append_table should be a hash distributed table
-- test keeping dependent materialized views
CREATE TABLE mat_view_test (a int, b int);
SELECT create_distributed_table('mat_view_test', 'a');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO mat_view_test VALUES (1,1), (2,2);
CREATE MATERIALIZED VIEW mat_view AS SELECT * FROM mat_view_test;
SELECT alter_distributed_table('mat_view_test', shard_count := 5, cascade_to_colocated := false);
NOTICE: creating a new table for alter_distributed_table.mat_view_test
NOTICE: moving the data of alter_distributed_table.mat_view_test
NOTICE: dropping the old alter_distributed_table.mat_view_test
NOTICE: drop cascades to materialized view mat_view
CONTEXT: SQL statement "DROP TABLE alter_distributed_table.mat_view_test CASCADE"
NOTICE: renaming the new table to alter_distributed_table.mat_view_test
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT * FROM mat_view ORDER BY a;
a | b
---------------------------------------------------------------------
1 | 1
2 | 2
(2 rows)
-- test long table names
SET client_min_messages TO DEBUG1;
CREATE TABLE abcde_0123456789012345678901234567890123456789012345678901234567890123456789 (x int, y int);
NOTICE: identifier "abcde_0123456789012345678901234567890123456789012345678901234567890123456789" will be truncated to "abcde_012345678901234567890123456789012345678901234567890123456"
SELECT create_distributed_table('abcde_0123456789012345678901234567890123456789012345678901234567890123456789', 'x');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT alter_distributed_table('abcde_0123456789012345678901234567890123456789012345678901234567890123456789', distribution_column := 'y');
DEBUG: the name of the shard (abcde_01234567890123456789012345678901234567890_f7ff6612_xxxxxx) for relation (abcde_012345678901234567890123456789012345678901234567890123456) is too long, switching to sequential and local execution mode to prevent self deadlocks
NOTICE: creating a new table for alter_distributed_table.abcde_012345678901234567890123456789012345678901234567890123456
NOTICE: moving the data of alter_distributed_table.abcde_012345678901234567890123456789012345678901234567890123456
DEBUG: cannot perform distributed INSERT INTO ... SELECT because the partition columns in the source table and subquery do not match
DETAIL: The target table's partition column should correspond to a partition column in the subquery.
CONTEXT: SQL statement "INSERT INTO alter_distributed_table.abcde_0123456789012345678901234567890123456_f7ff6612_4160710162 (x,y) SELECT x,y FROM alter_distributed_table.abcde_012345678901234567890123456789012345678901234567890123456"
DEBUG: performing repartitioned INSERT ... SELECT
CONTEXT: SQL statement "INSERT INTO alter_distributed_table.abcde_0123456789012345678901234567890123456_f7ff6612_4160710162 (x,y) SELECT x,y FROM alter_distributed_table.abcde_012345678901234567890123456789012345678901234567890123456"
NOTICE: dropping the old alter_distributed_table.abcde_012345678901234567890123456789012345678901234567890123456
CONTEXT: SQL statement "DROP TABLE alter_distributed_table.abcde_012345678901234567890123456789012345678901234567890123456 CASCADE"
NOTICE: renaming the new table to alter_distributed_table.abcde_012345678901234567890123456789012345678901234567890123456
DEBUG: the name of the shard (abcde_01234567890123456789012345678901234567890_f7ff6612_xxxxxx) for relation (abcde_012345678901234567890123456789012345678901234567890123456) is too long, switching to sequential and local execution mode to prevent self deadlocks
CONTEXT: SQL statement "ALTER TABLE alter_distributed_table.abcde_0123456789012345678901234567890123456_f7ff6612_4160710162 RENAME TO abcde_012345678901234567890123456789012345678901234567890123456"
alter_distributed_table
---------------------------------------------------------------------
(1 row)
RESET client_min_messages;
-- test long partitioned table names
CREATE TABLE partition_lengths
(
tenant_id integer NOT NULL,
timeperiod timestamp without time zone NOT NULL,
inserted_utc timestamp without time zone NOT NULL DEFAULT now()
) PARTITION BY RANGE (timeperiod);
SELECT create_distributed_table('partition_lengths', 'tenant_id');
create_distributed_table
---------------------------------------------------------------------
(1 row)
CREATE TABLE partition_lengths_p2020_09_28_12345678901234567890123456789012345678901234567890 PARTITION OF partition_lengths FOR VALUES FROM ('2020-09-28 00:00:00') TO ('2020-09-29 00:00:00');
NOTICE: identifier "partition_lengths_p2020_09_28_12345678901234567890123456789012345678901234567890" will be truncated to "partition_lengths_p2020_09_28_123456789012345678901234567890123"
-- verify alter_distributed_table works with long partition names
SELECT alter_distributed_table('partition_lengths', shard_count := 29, cascade_to_colocated := false);
NOTICE: converting the partitions of alter_distributed_table.partition_lengths
NOTICE: creating a new table for alter_distributed_table.partition_lengths_p2020_09_28_123456789012345678901234567890123
NOTICE: moving the data of alter_distributed_table.partition_lengths_p2020_09_28_123456789012345678901234567890123
NOTICE: dropping the old alter_distributed_table.partition_lengths_p2020_09_28_123456789012345678901234567890123
NOTICE: renaming the new table to alter_distributed_table.partition_lengths_p2020_09_28_123456789012345678901234567890123
NOTICE: creating a new table for alter_distributed_table.partition_lengths
NOTICE: dropping the old alter_distributed_table.partition_lengths
NOTICE: renaming the new table to alter_distributed_table.partition_lengths
alter_distributed_table
---------------------------------------------------------------------
(1 row)
-- test long partition table names
ALTER TABLE partition_lengths_p2020_09_28_12345678901234567890123456789012345678901234567890 RENAME TO partition_lengths_p2020_09_28;
NOTICE: identifier "partition_lengths_p2020_09_28_12345678901234567890123456789012345678901234567890" will be truncated to "partition_lengths_p2020_09_28_123456789012345678901234567890123"
ALTER TABLE partition_lengths RENAME TO partition_lengths_12345678901234567890123456789012345678901234567890;
NOTICE: identifier "partition_lengths_12345678901234567890123456789012345678901234567890" will be truncated to "partition_lengths_123456789012345678901234567890123456789012345"
-- verify alter_distributed_table works with long partitioned table names
SELECT alter_distributed_table('partition_lengths_12345678901234567890123456789012345678901234567890', shard_count := 17, cascade_to_colocated := false);
NOTICE: converting the partitions of alter_distributed_table.partition_lengths_123456789012345678901234567890123456789012345
NOTICE: creating a new table for alter_distributed_table.partition_lengths_p2020_09_28
NOTICE: moving the data of alter_distributed_table.partition_lengths_p2020_09_28
NOTICE: dropping the old alter_distributed_table.partition_lengths_p2020_09_28
NOTICE: renaming the new table to alter_distributed_table.partition_lengths_p2020_09_28
NOTICE: creating a new table for alter_distributed_table.partition_lengths_123456789012345678901234567890123456789012345
NOTICE: dropping the old alter_distributed_table.partition_lengths_123456789012345678901234567890123456789012345
NOTICE: renaming the new table to alter_distributed_table.partition_lengths_123456789012345678901234567890123456789012345
alter_distributed_table
---------------------------------------------------------------------
(1 row)
SET client_min_messages TO WARNING;
DROP SCHEMA alter_distributed_table CASCADE;

12
src/test/regress/expected/alter_table_set_access_method_0.out
View File

@ -1,12 +0,0 @@
-- test for Postgres version
-- should error before PG12
CREATE TABLE alter_am_pg_version_table (a INT);
SELECT alter_table_set_access_method('alter_am_pg_version_table', 'columnar');
ERROR: table access methods are not supported for Postgres versions earlier than 12
DROP TABLE alter_am_pg_version_table;
SHOW server_version \gset
SELECT substring(:'server_version', '\d+')::int > 11 AS server_version_above_eleven
\gset
\if :server_version_above_eleven
\else
\q

36
src/test/regress/expected/coordinator_shouldhaveshards.out
View File

@ -335,7 +335,6 @@ NOTICE: executing the command locally: SELECT ref.a, ref.b, local.x, local.y FR
(1 row)
ROLLBACK;
set citus.enable_cte_inlining to off;
BEGIN;
SELECT count(*) FROM test;
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503000 test WHERE true
@ -348,11 +347,11 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinato
-- we wont see the modifying cte in this query because we will use local execution and
-- in postgres we wouldn't see this modifying cte, so it is consistent with postgres.
WITH a AS (SELECT count(*) FROM test), b AS (INSERT INTO local VALUES (3,2) RETURNING *), c AS (INSERT INTO ref VALUES (3,2) RETURNING *), d AS (SELECT count(*) FROM ref JOIN local ON (a = x)) SELECT * FROM a, b, c, d ORDER BY x,y,a,b;
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503000 test WHERE true
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503003 test WHERE true
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN coordinator_shouldhaveshards.local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN (SELECT local_1.x, NULL::integer AS y FROM (SELECT intermediate_result.x FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(x integer)) local_1) local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
count | x | y | a | b | count
---------------------------------------------------------------------
100 | 3 | 2 | 3 | 2 | 0
@ -369,11 +368,11 @@ NOTICE: executing the command locally: SELECT ref.a, ref.b, local.x, local.y FR
-- we wont see the modifying cte in this query because we will use local execution and
-- in postgres we wouldn't see this modifying cte, so it is consistent with postgres.
WITH a AS (SELECT count(*) FROM test), b AS (INSERT INTO local VALUES (3,2) RETURNING *), c AS (INSERT INTO ref VALUES (3,2) RETURNING *), d AS (SELECT count(*) FROM ref JOIN local ON (a = x)) SELECT * FROM a, b, c, d ORDER BY x,y,a,b;
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503000 test WHERE true
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503003 test WHERE true
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN coordinator_shouldhaveshards.local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN (SELECT local_1.x, NULL::integer AS y FROM (SELECT intermediate_result.x FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(x integer)) local_1) local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
count | x | y | a | b | count
---------------------------------------------------------------------
100 | 3 | 2 | 3 | 2 | 0
@ -384,11 +383,11 @@ BEGIN;
-- we wont see the modifying cte in this query because we will use local execution and
-- in postgres we wouldn't see this modifying cte, so it is consistent with postgres.
WITH a AS (SELECT count(*) FROM test), b AS (INSERT INTO local VALUES (3,2) RETURNING *), c AS (INSERT INTO ref VALUES (3,2) RETURNING *), d AS (SELECT count(*) FROM ref JOIN local ON (a = x)) SELECT * FROM a, b, c, d ORDER BY x,y,a,b;
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503000 test WHERE true
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503003 test WHERE true
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN coordinator_shouldhaveshards.local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN (SELECT local_1.x, NULL::integer AS y FROM (SELECT intermediate_result.x FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(x integer)) local_1) local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
count | x | y | a | b | count
---------------------------------------------------------------------
100 | 3 | 2 | 3 | 2 | 0
@ -399,12 +398,12 @@ BEGIN;
-- we wont see the modifying cte in this query because we will use local execution and
-- in postgres we wouldn't see this modifying cte, so it is consistent with postgres.
WITH a AS (SELECT count(*) FROM test), b AS (INSERT INTO local VALUES (3,2) RETURNING *), c AS (INSERT INTO ref SELECT *,* FROM generate_series(1,10) RETURNING *), d AS (SELECT count(*) FROM ref JOIN local ON (a = x)) SELECT * FROM a, b, c, d ORDER BY x,y,a,b;
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503000 test WHERE true
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503003 test WHERE true
NOTICE: executing the copy locally for colocated file with shard xxxxx
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 AS citus_table_alias (a, b) SELECT a, b FROM read_intermediate_result('insert_select_XXX_1503020'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer) RETURNING citus_table_alias.a, citus_table_alias.b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN coordinator_shouldhaveshards.local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503000 test WHERE true
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503003 test WHERE true
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN (SELECT local_1.x, NULL::integer AS y FROM (SELECT intermediate_result.x FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(x integer)) local_1) local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
count | x | y | a | b | count
---------------------------------------------------------------------
100 | 3 | 2 | 1 | 1 | 0
@ -434,11 +433,11 @@ NOTICE: executing the command locally: SELECT ref.a, ref.b, local.x, local.y FR
-- we wont see the modifying cte in this query because we will use local execution and
-- in postgres we wouldn't see this modifying cte, so it is consistent with postgres.
WITH a AS (SELECT count(*) FROM test), b AS (INSERT INTO local VALUES (3,2) RETURNING *), c AS (INSERT INTO ref VALUES (3,2) RETURNING *), d AS (SELECT count(*) FROM ref JOIN local ON (a = x)) SELECT * FROM a, b, c, d ORDER BY x,y,a,b;
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503000 test WHERE true
NOTICE: executing the command locally: SELECT count(*) AS count FROM coordinator_shouldhaveshards.test_1503003 test WHERE true
NOTICE: executing the command locally: INSERT INTO coordinator_shouldhaveshards.ref_1503020 (a, b) VALUES (3, 2) RETURNING a, b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN coordinator_shouldhaveshards.local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
NOTICE: executing the command locally: SELECT count(*) AS count FROM (coordinator_shouldhaveshards.ref_1503020 ref JOIN (SELECT local_1.x, NULL::integer AS y FROM (SELECT intermediate_result.x FROM read_intermediate_result('XXX_4'::text, 'binary'::citus_copy_format) intermediate_result(x integer)) local_1) local ON ((ref.a OPERATOR(pg_catalog.=) local.x)))
NOTICE: executing the command locally: SELECT a.count, b.x, b.y, c.a, c.b, d.count FROM (SELECT intermediate_result.count FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) a, (SELECT intermediate_result.x, intermediate_result.y FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(x integer, y integer)) b, (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) c, (SELECT intermediate_result.count FROM read_intermediate_result('XXX_5'::text, 'binary'::citus_copy_format) intermediate_result(count bigint)) d ORDER BY b.x, b.y, c.a, c.b
count | x | y | a | b | count
---------------------------------------------------------------------
100 | 3 | 2 | 3 | 2 | 1
@ -489,7 +488,7 @@ NOTICE: executing the command locally: SELECT ref.a, ref.b, local.x, local.y, c
-- full router query with CTE and local
WITH cte_1 AS (SELECT * FROM ref LIMIT 1)
SELECT * FROM ref JOIN local ON (a = x) JOIN cte_1 ON (local.x = cte_1.a);
NOTICE: executing the command locally: WITH cte_1 AS (SELECT ref_1.a, ref_1.b FROM coordinator_shouldhaveshards.ref_1503020 ref_1 LIMIT 1) SELECT ref.a, ref.b, local.x, local.y, cte_1.a, cte_1.b FROM ((coordinator_shouldhaveshards.ref_1503020 ref JOIN coordinator_shouldhaveshards.local ON ((ref.a OPERATOR(pg_catalog.=) local.x))) JOIN cte_1 ON ((local.x OPERATOR(pg_catalog.=) cte_1.a)))
NOTICE: executing the command locally: SELECT ref.a, ref.b, local.x, local.y, cte_1.a, cte_1.b FROM ((coordinator_shouldhaveshards.ref_1503020 ref JOIN coordinator_shouldhaveshards.local ON ((ref.a OPERATOR(pg_catalog.=) local.x))) JOIN (SELECT ref_1.a, ref_1.b FROM coordinator_shouldhaveshards.ref_1503020 ref_1 LIMIT 1) cte_1 ON ((local.x OPERATOR(pg_catalog.=) cte_1.a)))
a | b | x | y | a | b
---------------------------------------------------------------------
1 | 2 | 1 | 2 | 1 | 2
@ -531,7 +530,6 @@ NOTICE: executing the command locally: SELECT worker_apply_shard_ddl_command (1
(1 row)
ROLLBACK;
RESET citus.enable_cte_inlining;
CREATE table ref_table(x int PRIMARY KEY, y int);
-- this will be replicated to the coordinator because of add_coordinator test
SELECT create_reference_table('ref_table');

2
src/test/regress/expected/create_ref_dist_from_citus_local.out
View File

@ -342,7 +342,7 @@ BEGIN;
$update_value$ LANGUAGE plpgsql;
CREATE TRIGGER update_value_dist
AFTER INSERT ON citus_local_table_6
FOR EACH ROW EXECUTE FUNCTION update_value();
FOR EACH ROW EXECUTE PROCEDURE update_value();
-- show that we error out as we don't supprt triggers on distributed tables
SELECT create_distributed_table('citus_local_table_6', 'col_1');
ERROR: cannot distribute relation "citus_local_table_6" because it has triggers

1396
src/test/regress/expected/cte_inline_0.out
View File

File diff suppressed because it is too large Load Diff

646
src/test/regress/expected/custom_aggregate_support_1.out
View File

@ -1,646 +0,0 @@
--
-- CUSTOM_AGGREGATE_SUPPORT
--
-- Create HLL extension if present, print false result otherwise
SELECT CASE WHEN COUNT(*) > 0 THEN
'CREATE EXTENSION HLL'
ELSE 'SELECT false AS hll_present' END
AS create_cmd FROM pg_available_extensions()
WHERE name = 'hll'
\gset
:create_cmd;
SET citus.shard_count TO 4;
set citus.coordinator_aggregation_strategy to 'disabled';
CREATE TABLE raw_table (day date, user_id int);
CREATE TABLE daily_uniques(day date, unique_users hll);
SELECT create_distributed_table('raw_table', 'user_id');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('daily_uniques', 'day');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO raw_table
SELECT day, user_id % 19
FROM generate_series('2018-05-24'::timestamp, '2018-06-24'::timestamp, '1 day'::interval) as f(day),
generate_series(1,100) as g(user_id);
INSERT INTO raw_table
SELECT day, user_id % 13
FROM generate_series('2018-06-10'::timestamp, '2018-07-10'::timestamp, '1 day'::interval) as f(day),
generate_series(1,100) as g(user_id);
-- Run hll on raw data
SELECT hll_cardinality(hll_union_agg(agg))
FROM (
SELECT hll_add_agg(hll_hash_integer(user_id)) AS agg
FROM raw_table)a;
hll_cardinality
---------------------------------------------------------------------
19
(1 row)
-- Aggregate the data into daily_uniques
INSERT INTO daily_uniques
SELECT day, hll_add_agg(hll_hash_integer(user_id))
FROM raw_table
GROUP BY 1;
-- Basic hll_cardinality check on aggregated data
SELECT day, hll_cardinality(unique_users)
FROM daily_uniques
WHERE day >= '2018-06-20' and day <= '2018-06-30'
ORDER BY 2 DESC,1
LIMIT 10;
day | hll_cardinality
---------------------------------------------------------------------
06-20-2018 | 19
06-21-2018 | 19
06-22-2018 | 19
06-23-2018 | 19
06-24-2018 | 19
06-25-2018 | 13
06-26-2018 | 13
06-27-2018 | 13
06-28-2018 | 13
06-29-2018 | 13
(10 rows)
-- Union aggregated data for one week
SELECT hll_cardinality(hll_union_agg(unique_users))
FROM daily_uniques
WHERE day >= '2018-05-24'::date AND day <= '2018-05-31'::date;
hll_cardinality
---------------------------------------------------------------------
19
(1 row)
SELECT EXTRACT(MONTH FROM day) AS month, hll_cardinality(hll_union_agg(unique_users))
FROM daily_uniques
WHERE day >= '2018-06-23' AND day <= '2018-07-01'
GROUP BY 1
ORDER BY 1;
month | hll_cardinality
---------------------------------------------------------------------
6 | 19
7 | 13
(2 rows)
SELECT day, hll_cardinality(hll_union_agg(unique_users) OVER seven_days)
FROM daily_uniques
WINDOW seven_days AS (ORDER BY day ASC ROWS 6 PRECEDING)
ORDER BY 1;
day | hll_cardinality
---------------------------------------------------------------------
05-24-2018 | 19
05-25-2018 | 19
05-26-2018 | 19
05-27-2018 | 19
05-28-2018 | 19
05-29-2018 | 19
05-30-2018 | 19
05-31-2018 | 19
06-01-2018 | 19
06-02-2018 | 19
06-03-2018 | 19
06-04-2018 | 19
06-05-2018 | 19
06-06-2018 | 19
06-07-2018 | 19
06-08-2018 | 19
06-09-2018 | 19
06-10-2018 | 19
06-11-2018 | 19
06-12-2018 | 19
06-13-2018 | 19
06-14-2018 | 19
06-15-2018 | 19
06-16-2018 | 19
06-17-2018 | 19
06-18-2018 | 19
06-19-2018 | 19
06-20-2018 | 19
06-21-2018 | 19
06-22-2018 | 19
06-23-2018 | 19
06-24-2018 | 19
06-25-2018 | 19
06-26-2018 | 19
06-27-2018 | 19
06-28-2018 | 19
06-29-2018 | 19
06-30-2018 | 19
07-01-2018 | 13
07-02-2018 | 13
07-03-2018 | 13
07-04-2018 | 13
07-05-2018 | 13
07-06-2018 | 13
07-07-2018 | 13
07-08-2018 | 13
07-09-2018 | 13
07-10-2018 | 13
(48 rows)
SELECT day, (hll_cardinality(hll_union_agg(unique_users) OVER two_days)) - hll_cardinality(unique_users) AS lost_uniques
FROM daily_uniques
WINDOW two_days AS (ORDER BY day ASC ROWS 1 PRECEDING)
ORDER BY 1;
day | lost_uniques
---------------------------------------------------------------------
05-24-2018 | 0
05-25-2018 | 0
05-26-2018 | 0
05-27-2018 | 0
05-28-2018 | 0
05-29-2018 | 0
05-30-2018 | 0
05-31-2018 | 0
06-01-2018 | 0
06-02-2018 | 0
06-03-2018 | 0
06-04-2018 | 0
06-05-2018 | 0
06-06-2018 | 0
06-07-2018 | 0
06-08-2018 | 0
06-09-2018 | 0
06-10-2018 | 0
06-11-2018 | 0
06-12-2018 | 0
06-13-2018 | 0
06-14-2018 | 0
06-15-2018 | 0
06-16-2018 | 0
06-17-2018 | 0
06-18-2018 | 0
06-19-2018 | 0
06-20-2018 | 0
06-21-2018 | 0
06-22-2018 | 0
06-23-2018 | 0
06-24-2018 | 0
06-25-2018 | 6
06-26-2018 | 0
06-27-2018 | 0
06-28-2018 | 0
06-29-2018 | 0
06-30-2018 | 0
07-01-2018 | 0
07-02-2018 | 0
07-03-2018 | 0
07-04-2018 | 0
07-05-2018 | 0
07-06-2018 | 0
07-07-2018 | 0
07-08-2018 | 0
07-09-2018 | 0
07-10-2018 | 0
(48 rows)
-- Test disabling hash_agg on coordinator query
SET citus.explain_all_tasks to true;
SET hll.force_groupagg to OFF;
EXPLAIN(COSTS OFF)
SELECT
day, hll_union_agg(unique_users)
FROM
daily_uniques
GROUP BY(1);
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(23 rows)
SET hll.force_groupagg to ON;
EXPLAIN(COSTS OFF)
SELECT
day, hll_union_agg(unique_users)
FROM
daily_uniques
GROUP BY(1);
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(23 rows)
-- Test disabling hash_agg with operator on coordinator query
SET hll.force_groupagg to OFF;
EXPLAIN(COSTS OFF)
SELECT
day, hll_union_agg(unique_users) || hll_union_agg(unique_users)
FROM
daily_uniques
GROUP BY(1);
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(23 rows)
SET hll.force_groupagg to ON;
EXPLAIN(COSTS OFF)
SELECT
day, hll_union_agg(unique_users) || hll_union_agg(unique_users)
FROM
daily_uniques
GROUP BY(1);
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(23 rows)
-- Test disabling hash_agg with expression on coordinator query
SET hll.force_groupagg to OFF;
EXPLAIN(COSTS OFF)
SELECT
day, hll_cardinality(hll_union_agg(unique_users))
FROM
daily_uniques
GROUP BY(1);
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(23 rows)
SET hll.force_groupagg to ON;
EXPLAIN(COSTS OFF)
SELECT
day, hll_cardinality(hll_union_agg(unique_users))
FROM
daily_uniques
GROUP BY(1);
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(23 rows)
-- Test disabling hash_agg with having
SET hll.force_groupagg to OFF;
EXPLAIN(COSTS OFF)
SELECT
day, hll_cardinality(hll_union_agg(unique_users))
FROM
daily_uniques
GROUP BY(1);
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(23 rows)
SET hll.force_groupagg to ON;
EXPLAIN(COSTS OFF)
SELECT
day, hll_cardinality(hll_union_agg(unique_users))
FROM
daily_uniques
GROUP BY(1)
HAVING hll_cardinality(hll_union_agg(unique_users)) > 1;
QUERY PLAN
---------------------------------------------------------------------
Custom Scan (Citus Adaptive)
Task Count: 4
Tasks Shown: All
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
Filter: (hll_cardinality(hll_union_agg(unique_users)) > '1'::double precision)
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
Filter: (hll_cardinality(hll_union_agg(unique_users)) > '1'::double precision)
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
Filter: (hll_cardinality(hll_union_agg(unique_users)) > '1'::double precision)
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
-> Task
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Group Key: day
Filter: (hll_cardinality(hll_union_agg(unique_users)) > '1'::double precision)
-> Seq Scan on daily_uniques_xxxxxxx daily_uniques
(27 rows)
DROP TABLE raw_table;
DROP TABLE daily_uniques;
-- Check if TopN aggregates work as expected
-- Create TopN extension if present, print false result otherwise
SELECT CASE WHEN COUNT(*) > 0 THEN
'CREATE EXTENSION TOPN'
ELSE 'SELECT false AS topn_present' END
AS create_topn FROM pg_available_extensions()
WHERE name = 'topn'
\gset
:create_topn;
CREATE TABLE customer_reviews (day date, user_id int, review int);
CREATE TABLE popular_reviewer(day date, reviewers jsonb);
SELECT create_distributed_table('customer_reviews', 'user_id');
create_distributed_table
---------------------------------------------------------------------
(1 row)
SELECT create_distributed_table('popular_reviewer', 'day');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO customer_reviews
SELECT day, user_id % 7, review % 5
FROM generate_series('2018-05-24'::timestamp, '2018-06-24'::timestamp, '1 day'::interval) as f(day),
generate_series(1,30) as g(user_id), generate_series(0,30) AS r(review);
INSERT INTO customer_reviews
SELECT day, user_id % 13, review % 3
FROM generate_series('2018-06-10'::timestamp, '2018-07-10'::timestamp, '1 day'::interval) as f(day),
generate_series(1,30) as g(user_id), generate_series(0,30) AS r(review);
-- Run topn on raw data
SELECT (topn(agg, 10)).*
FROM (
SELECT topn_add_agg(user_id::text) AS agg
FROM customer_reviews
)a
ORDER BY 2 DESC, 1;
item | frequency
---------------------------------------------------------------------
1 | 7843
2 | 7843
3 | 6851
4 | 6851
0 | 5890
5 | 5890
6 | 5890
7 | 1922
8 | 1922
9 | 1922
(10 rows)
-- Aggregate the data into popular_reviewer
INSERT INTO popular_reviewer
SELECT day, topn_add_agg(user_id::text)
FROM customer_reviews
GROUP BY 1;
-- Basic topn check on aggregated data
SELECT day, (topn(reviewers, 10)).*
FROM popular_reviewer
WHERE day >= '2018-06-20' and day <= '2018-06-30'
ORDER BY 3 DESC, 1, 2
LIMIT 10;
day | item | frequency
---------------------------------------------------------------------
06-20-2018 | 1 | 248
06-20-2018 | 2 | 248
06-21-2018 | 1 | 248
06-21-2018 | 2 | 248
06-22-2018 | 1 | 248
06-22-2018 | 2 | 248
06-23-2018 | 1 | 248
06-23-2018 | 2 | 248
06-24-2018 | 1 | 248
06-24-2018 | 2 | 248
(10 rows)
-- Union aggregated data for one week
SELECT (topn(agg, 10)).*
FROM (
SELECT topn_union_agg(reviewers) AS agg
FROM popular_reviewer
WHERE day >= '2018-05-24'::date AND day <= '2018-05-31'::date
)a
ORDER BY 2 DESC, 1;
item | frequency
---------------------------------------------------------------------
1 | 1240
2 | 1240
0 | 992
3 | 992
4 | 992
5 | 992
6 | 992
(7 rows)
SELECT month, (topn(agg, 5)).*
FROM (
SELECT EXTRACT(MONTH FROM day) AS month, topn_union_agg(reviewers) AS agg
FROM popular_reviewer
WHERE day >= '2018-06-23' AND day <= '2018-07-01'
GROUP BY 1
ORDER BY 1
)a
ORDER BY 1, 3 DESC, 2;
month | item | frequency
---------------------------------------------------------------------
6 | 1 | 1054
6 | 2 | 1054
6 | 3 | 992
6 | 4 | 992
6 | 0 | 744
7 | 1 | 93
7 | 2 | 93
7 | 3 | 93
7 | 4 | 93
7 | 8 | 62
(10 rows)
-- TODO the following queries will be supported after we fix #2265
-- They work for PG9.6 but not for PG10
SELECT (topn(topn_union_agg(reviewers), 10)).*
FROM popular_reviewer
WHERE day >= '2018-05-24'::date AND day <= '2018-05-31'::date
ORDER BY 2 DESC, 1;
ERROR: set-valued function called in context that cannot accept a set
SELECT (topn(topn_add_agg(user_id::text), 10)).*
FROM customer_reviews
ORDER BY 2 DESC, 1;
ERROR: set-valued function called in context that cannot accept a set
SELECT day, (topn(agg, 10)).*
FROM (
SELECT day, topn_union_agg(reviewers) OVER seven_days AS agg
FROM popular_reviewer
WINDOW seven_days AS (ORDER BY day ASC ROWS 6 PRECEDING)
)a
ORDER BY 3 DESC, 1, 2
LIMIT 10;
day | item | frequency
---------------------------------------------------------------------
06-16-2018 | 1 | 1736
06-16-2018 | 2 | 1736
06-17-2018 | 1 | 1736
06-17-2018 | 2 | 1736
06-18-2018 | 1 | 1736
06-18-2018 | 2 | 1736
06-19-2018 | 1 | 1736
06-19-2018 | 2 | 1736
06-20-2018 | 1 | 1736
06-20-2018 | 2 | 1736
(10 rows)
SELECT day, (topn(topn_add_agg(user_id::text) OVER seven_days, 10)).*
FROM customer_reviews
WINDOW seven_days AS (ORDER BY day ASC ROWS 6 PRECEDING)
ORDER BY 3 DESC, 1, 2
LIMIT 10;
ERROR: set-valued function called in context that cannot accept a set
DROP TABLE customer_reviews;
DROP TABLE popular_reviewer;

75
src/test/regress/expected/distributed_types_xact_add_enum_value_0.out
View File

@ -1,75 +0,0 @@
SHOW server_version \gset
SELECT substring(:'server_version', '\d+')::int > 11 AS version_above_eleven;
version_above_eleven
---------------------------------------------------------------------
f
(1 row)
SET citus.next_shard_id TO 20040000;
CREATE SCHEMA xact_enum_type;
SET search_path TO xact_enum_type;
SET citus.shard_count TO 4;
-- transaction block with simple type
BEGIN;
CREATE TYPE xact_enum_edit AS ENUM ('yes', 'no');
CREATE TABLE t1 (a int PRIMARY KEY, b xact_enum_edit);
SELECT create_distributed_table('t1','a');
create_distributed_table
---------------------------------------------------------------------
(1 row)
INSERT INTO t1 VALUES (1, 'yes');
SELECT * FROM t1;
a | b
---------------------------------------------------------------------
1 | yes
(1 row)
COMMIT;
BEGIN;
ALTER TYPE xact_enum_edit ADD VALUE 'maybe';
ERROR: ALTER TYPE ... ADD cannot run inside a transaction block
ABORT;
-- maybe should not be on the workers
SELECT string_agg(enumlabel, ',' ORDER BY enumsortorder ASC) FROM pg_enum WHERE enumtypid = 'xact_enum_type.xact_enum_edit'::regtype;
string_agg
---------------------------------------------------------------------
yes,no
(1 row)
SELECT run_command_on_workers($$SELECT string_agg(enumlabel, ',' ORDER BY enumsortorder ASC) FROM pg_enum WHERE enumtypid = 'xact_enum_type.xact_enum_edit'::regtype;$$);
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"yes,no")
(localhost,57638,t,"yes,no")
(2 rows)
BEGIN;
ALTER TYPE xact_enum_edit ADD VALUE 'maybe';
ERROR: ALTER TYPE ... ADD cannot run inside a transaction block
COMMIT;
-- maybe should be on the workers (pg12 and above)
SELECT string_agg(enumlabel, ',' ORDER BY enumsortorder ASC) FROM pg_enum WHERE enumtypid = 'xact_enum_type.xact_enum_edit'::regtype;
string_agg
---------------------------------------------------------------------
yes,no
(1 row)
SELECT run_command_on_workers($$SELECT string_agg(enumlabel, ',' ORDER BY enumsortorder ASC) FROM pg_enum WHERE enumtypid = 'xact_enum_type.xact_enum_edit'::regtype;$$);
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"yes,no")
(localhost,57638,t,"yes,no")
(2 rows)
-- clear objects
SET client_min_messages TO error; -- suppress cascading objects dropping
DROP SCHEMA xact_enum_type CASCADE;
SELECT run_command_on_workers($$DROP SCHEMA xact_enum_type CASCADE;$$);
run_command_on_workers
---------------------------------------------------------------------
(localhost,57637,t,"DROP SCHEMA")
(localhost,57638,t,"DROP SCHEMA")
(2 rows)

50
src/test/regress/expected/failure_cte_subquery.out
View File

@ -5,7 +5,7 @@ SET citus.shard_replication_factor to 1;
SET citus.next_shard_id TO 16000000;
-- CTE inlining should not happen because
-- the tests rely on intermediate results
SET citus.enable_cte_inlining TO false;
-- That's why we use MATERIALIZED CTEs in the test file
-- prevent using locally executing the intermediate results
SET citus.task_assignment_policy TO "round-robin";
SELECT pg_backend_pid() as pid \gset
@ -31,11 +31,11 @@ SELECT citus.mitmproxy('conn.onQuery(query="^COPY").kill()');
(1 row)
WITH cte AS (
WITH local_cte AS (
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS (
dist_cte AS MATERIALIZED (
SELECT user_id FROM events_table
)
SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id
@ -65,11 +65,11 @@ SELECT citus.mitmproxy('conn.onQuery(query="SELECT user_id FROM cte_failure.even
(1 row)
WITH cte AS (
WITH local_cte AS (
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS (
dist_cte AS MATERIALIZED (
SELECT user_id FROM events_table
)
SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id
@ -98,11 +98,11 @@ SELECT citus.mitmproxy('conn.onQuery(query="SELECT DISTINCT users_table.user").k
(1 row)
WITH cte AS (
WITH local_cte AS (
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS (
dist_cte AS MATERIALIZED (
SELECT user_id FROM events_table
)
SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id
@ -131,11 +131,11 @@ SELECT citus.mitmproxy('conn.onQuery(query="^COPY").cancel(' || :pid || ')');
(1 row)
WITH cte AS (
WITH local_cte AS (
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS (
dist_cte AS MATERIALIZED (
SELECT user_id FROM events_table
)
SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id
@ -162,11 +162,11 @@ SELECT citus.mitmproxy('conn.onQuery(query="SELECT user_id FROM").cancel(' || :p
(1 row)
WITH cte AS (
WITH local_cte AS (
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS (
dist_cte AS MATERIALIZED (
SELECT user_id FROM events_table
)
SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id
@ -193,11 +193,11 @@ SELECT citus.mitmproxy('conn.onQuery(query="SELECT DISTINCT users_table.user").c
(1 row)
WITH cte AS (
WITH local_cte AS (
WITH cte AS MATERIALIZED (
WITH local_cte AS MATERIALIZED (
SELECT * FROM users_table_local
),
dist_cte AS (
dist_cte AS MATERIALIZED (
SELECT user_id FROM events_table
)
SELECT dist_cte.user_id FROM local_cte join dist_cte on dist_cte.user_id=local_cte.user_id
@ -238,7 +238,7 @@ SELECT * FROM users_table ORDER BY 1, 2;
(5 rows)
-- following will delete and insert the same rows
WITH cte_delete as (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
WITH cte_delete AS MATERIALIZED (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
INSERT INTO users_table SELECT * FROM cte_delete;
-- verify contents are the same
SELECT * FROM users_table ORDER BY 1, 2;
@ -258,7 +258,7 @@ SELECT citus.mitmproxy('conn.onQuery(query="^DELETE FROM").kill()');
(1 row)
WITH cte_delete as (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
WITH cte_delete AS MATERIALIZED (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
INSERT INTO users_table SELECT * FROM cte_delete;
ERROR: connection to the remote node localhost:xxxxx failed with the following error: server closed the connection unexpectedly
This probably means the server terminated abnormally
@ -287,7 +287,7 @@ SELECT citus.mitmproxy('conn.onQuery(query="^COPY").kill()');
(1 row)
WITH cte_delete as (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
WITH cte_delete AS MATERIALIZED (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
INSERT INTO users_table SELECT * FROM cte_delete;
ERROR: server closed the connection unexpectedly
This probably means the server terminated abnormally
@ -317,7 +317,7 @@ SELECT citus.mitmproxy('conn.onQuery(query="^DELETE FROM").cancel(' || :pid || '
(1 row)
WITH cte_delete as (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
WITH cte_delete AS MATERIALIZED (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
INSERT INTO users_table SELECT * FROM cte_delete;
ERROR: canceling statement due to user request
-- verify contents are the same
@ -344,7 +344,7 @@ SELECT citus.mitmproxy('conn.onQuery(query="^COPY").cancel(' || :pid || ')');
(1 row)
WITH cte_delete as (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
WITH cte_delete AS MATERIALIZED (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
INSERT INTO users_table SELECT * FROM cte_delete;
ERROR: canceling statement due to user request
-- verify contents are the same
@ -373,7 +373,7 @@ SELECT citus.mitmproxy('conn.onQuery(query="^DELETE FROM").kill()');
BEGIN;
SET LOCAL citus.multi_shard_modify_mode = 'sequential';
WITH cte_delete as (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
WITH cte_delete AS MATERIALIZED (DELETE FROM users_table WHERE user_name in ('A', 'D') RETURNING *)
INSERT INTO users_table SELECT * FROM cte_delete;
ERROR: connection to the remote node localhost:xxxxx failed with the following error: server closed the connection unexpectedly
This probably means the server terminated abnormally

129
src/test/regress/expected/intermediate_result_pruning.out
View File

@ -32,8 +32,6 @@ SELECT create_reference_table('ref_table');
(1 row)
-- prevent PG 11 - PG 12 outputs to diverge
SET citus.enable_cte_inlining TO false;
-- load some data
INSERT INTO table_1 VALUES (1, '1'), (2, '2'), (3, '3'), (4, '4');
INSERT INTO table_2 VALUES (3, '3'), (4, '4'), (5, '5'), (6, '6');
@ -43,7 +41,7 @@ INSERT INTO ref_table VALUES (1, '1'), (2, '2'), (3, '3'), (4, '4'), (5, '5'),
SET client_min_messages TO DEBUG1;
-- a very basic case, where the intermediate result
-- should go to both workers
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key FROM table_1 WHERE value IN ('3', '4'))
SELECT
count(*)
@ -61,7 +59,7 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
-- a very basic case, where the intermediate result
-- should only go to one worker because the final query is a router
-- we use random() to prevent postgres inline the CTE(s)
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4'))
SELECT
count(*)
@ -78,7 +76,7 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
-- a similar query, but with a reference table now
-- given that reference tables are replicated to all nodes
-- we have to broadcast to all nodes
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4'))
SELECT
count(*)
@ -95,9 +93,9 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
-- a similar query as above, but this time use the CTE inside
-- another CTE
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1)
SELECT
count(*)
@ -115,9 +113,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- the second CTE does a join with a distributed table
-- and the final query is a router query
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key))
SELECT
count(*)
@ -137,9 +135,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- the first CTE is used both within second CTE and the final query
-- the second CTE does a join with a distributed table
-- and the final query is a router query
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key))
SELECT
count(*)
@ -159,9 +157,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- the first CTE is used both within second CTE and the final query
-- the second CTE does a join with a distributed table but a router query on a worker
-- and the final query is another router query on another worker
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key) WHERE table_2.key = 1)
SELECT
count(*)
@ -182,9 +180,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- the second CTE does a join with a distributed table but a router query on a worker
-- and the final query is a router query on the same worker, so the first result is only
-- broadcasted to a single node
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key) WHERE table_2.key = 1)
SELECT
count(*)
@ -201,9 +199,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
(1 row)
-- the same query with the above, but the final query is hitting all shards
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key))
SELECT
count(*)
@ -224,9 +222,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- even if we add a filter on the first query and make it a router query,
-- the first intermediate result still hits all workers because of the final
-- join is hitting all workers
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key) WHERE table_2.key = 3)
SELECT
count(*)
@ -247,7 +245,7 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- the reference table is joined with a distributed table and an intermediate
-- result, but the distributed table hits all shards, so the intermediate
-- result is sent to all nodes
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM ref_table WHERE value IN ('3', '4'))
SELECT
count(*)
@ -264,7 +262,7 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
-- similar query as above, but this time the whole query is a router
-- query, so no intermediate results
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM ref_table WHERE value IN ('3', '4'))
SELECT
count(*)
@ -279,9 +277,9 @@ FROM
-- so the first CTE should only be broadcasted to that node
-- since the final query doesn't have a join, it should simply be broadcasted
-- to one node
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key) WHERE key = 1)
SELECT
count(*)
@ -299,10 +297,10 @@ DEBUG: Subplan XXX_2 will be written to local file
-- the same query inlined inside a CTE, and the final query has a
-- join with a distributed table
WITH top_cte as (
WITH some_values_1 AS
WITH top_cte as MATERIALIZED (
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key) WHERE key = 1)
SELECT
DISTINCT key
@ -313,7 +311,7 @@ SELECT
count(*)
FROM
top_cte JOIN table_2 USING (key);
DEBUG: generating subplan XXX_1 for CTE top_cte: WITH some_values_1 AS (SELECT table_1.key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (table_1.value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))), some_values_2 AS (SELECT some_values_1.key, random() AS random FROM (some_values_1 JOIN intermediate_result_pruning.table_2 USING (key)) WHERE (some_values_1.key OPERATOR(pg_catalog.=) 1)) SELECT DISTINCT key FROM some_values_2
DEBUG: generating subplan XXX_1 for CTE top_cte: WITH some_values_1 AS MATERIALIZED (SELECT table_1.key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (table_1.value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))), some_values_2 AS MATERIALIZED (SELECT some_values_1.key, random() AS random FROM (some_values_1 JOIN intermediate_result_pruning.table_2 USING (key)) WHERE (some_values_1.key OPERATOR(pg_catalog.=) 1)) SELECT DISTINCT key FROM some_values_2
DEBUG: generating subplan XXX_1 for CTE some_values_1: SELECT key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))
DEBUG: generating subplan XXX_2 for CTE some_values_2: SELECT some_values_1.key, random() AS random FROM ((SELECT intermediate_result.key, intermediate_result.random FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, random double precision)) some_values_1 JOIN intermediate_result_pruning.table_2 USING (key)) WHERE (some_values_1.key OPERATOR(pg_catalog.=) 1)
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT DISTINCT key FROM (SELECT intermediate_result.key, intermediate_result.random FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, random double precision)) some_values_2
@ -329,10 +327,10 @@ DEBUG: Subplan XXX_2 will be written to local file
-- very much the same query, but this time the top query is also a router query
-- on a single worker, so all intermediate results only hit a single node
WITH top_cte as (
WITH some_values_1 AS
WITH top_cte as MATERIALIZED (
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key) WHERE key = 1)
SELECT
DISTINCT key
@ -343,7 +341,7 @@ SELECT
count(*)
FROM
top_cte JOIN table_2 USING (key) WHERE table_2.key = 2;
DEBUG: generating subplan XXX_1 for CTE top_cte: WITH some_values_1 AS (SELECT table_1.key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (table_1.value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))), some_values_2 AS (SELECT some_values_1.key, random() AS random FROM (some_values_1 JOIN intermediate_result_pruning.table_2 USING (key)) WHERE (some_values_1.key OPERATOR(pg_catalog.=) 1)) SELECT DISTINCT key FROM some_values_2
DEBUG: generating subplan XXX_1 for CTE top_cte: WITH some_values_1 AS MATERIALIZED (SELECT table_1.key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (table_1.value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))), some_values_2 AS MATERIALIZED (SELECT some_values_1.key, random() AS random FROM (some_values_1 JOIN intermediate_result_pruning.table_2 USING (key)) WHERE (some_values_1.key OPERATOR(pg_catalog.=) 1)) SELECT DISTINCT key FROM some_values_2
DEBUG: generating subplan XXX_1 for CTE some_values_1: SELECT key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))
DEBUG: generating subplan XXX_2 for CTE some_values_2: SELECT some_values_1.key, random() AS random FROM ((SELECT intermediate_result.key, intermediate_result.random FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, random double precision)) some_values_1 JOIN intermediate_result_pruning.table_2 USING (key)) WHERE (some_values_1.key OPERATOR(pg_catalog.=) 1)
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT DISTINCT key FROM (SELECT intermediate_result.key, intermediate_result.random FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, random double precision)) some_values_2
@ -358,11 +356,11 @@ DEBUG: Subplan XXX_2 will be written to local file
-- some_values_1 is first used by a single shard-query, and than with a multi-shard
-- CTE, finally a cartesian product join
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1 JOIN table_2 USING (key) WHERE key = 1),
some_values_3 AS
some_values_3 AS MATERIALIZED
(SELECT key FROM (some_values_2 JOIN table_2 USING (key)) JOIN some_values_1 USING (key))
SELECT * FROM some_values_3 JOIN ref_table ON (true);
DEBUG: generating subplan XXX_1 for CTE some_values_1: SELECT key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))
@ -381,9 +379,9 @@ DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
-- join on intermediate results, so should only
-- go to a single node
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM table_2 WHERE value IN ('3', '4'))
SELECT count(*) FROM some_values_2 JOIN some_values_1 USING (key);
DEBUG: generating subplan XXX_1 for CTE some_values_1: SELECT key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))
@ -398,9 +396,9 @@ DEBUG: Subplan XXX_2 will be written to local file
-- same query with WHERE false make sure that we're not broken
-- for such edge cases
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM table_2 WHERE value IN ('3', '4'))
SELECT count(*) FROM some_values_2 JOIN some_values_1 USING (key) WHERE false;
DEBUG: generating subplan XXX_1 for CTE some_values_1: SELECT key, random() AS random FROM intermediate_result_pruning.table_1 WHERE (value OPERATOR(pg_catalog.=) ANY (ARRAY['3'::text, '4'::text]))
@ -415,11 +413,11 @@ DEBUG: Subplan XXX_2 will be written to local file
-- do not use some_values_2 at all, so only 2 intermediate results are
-- broadcasted
WITH some_values_1 AS
WITH some_values_1 AS MATERIALIZED
(SELECT key, random() FROM table_1 WHERE value IN ('3', '4')),
some_values_2 AS
some_values_2 AS MATERIALIZED
(SELECT key, random() FROM some_values_1),
some_values_3 AS
some_values_3 AS MATERIALIZED
(SELECT key, random() FROM some_values_1)
SELECT
count(*)
@ -566,13 +564,13 @@ DEBUG: Subplan XXX_2 will be written to local file
(0 rows)
-- the intermediate results should just hit a single worker
WITH cte_1 AS
WITH cte_1 AS MATERIALIZED
(
(SELECT key FROM table_1 WHERE key = 1)
INTERSECT
(SELECT key FROM table_1 WHERE key = 2)
),
cte_2 AS
cte_2 AS MATERIALIZED
(
(SELECT key FROM table_1 WHERE key = 3)
INTERSECT
@ -599,13 +597,13 @@ DEBUG: Subplan XXX_2 will be written to local file
-- we join the results with distributed tables
-- so cte_1 should hit all workers, but still the
-- others should hit single worker each
WITH cte_1 AS
WITH cte_1 AS MATERIALIZED
(
(SELECT key FROM table_1 WHERE key = 1)
INTERSECT
(SELECT key FROM table_1 WHERE key = 2)
),
cte_2 AS
cte_2 AS MATERIALIZED
(
SELECT count(*) FROM table_1 JOIN cte_1 USING (key)
)
@ -668,10 +666,10 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
-- however, the subquery in WHERE clause of the DELETE query is broadcasted to all
-- nodes
BEGIN;
WITH select_data AS (
WITH select_data AS MATERIALIZED (
SELECT * FROM table_1
),
raw_data AS (
raw_data AS MATERIALIZED (
DELETE FROM table_2 WHERE key >= (SELECT min(key) FROM select_data WHERE key > 1) RETURNING *
)
SELECT * FROM raw_data;
@ -698,10 +696,10 @@ ROLLBACK;
-- however, the subquery in WHERE clause of the DELETE query is broadcasted to all
-- nodes
BEGIN;
WITH select_data AS (
WITH select_data AS MATERIALIZED (
SELECT * FROM table_1
),
raw_data AS (
raw_data AS MATERIALIZED (
DELETE FROM table_2 WHERE value::int >= (SELECT min(key) FROM select_data WHERE key > 1 + random()) RETURNING *
)
SELECT * FROM raw_data;
@ -726,10 +724,10 @@ ROLLBACK;
-- now, we need only two intermediate results as the subquery in WHERE clause is
-- router plannable
BEGIN;
WITH select_data AS (
WITH select_data AS MATERIALIZED (
SELECT * FROM table_1
),
raw_data AS (
raw_data AS MATERIALIZED (
DELETE FROM table_2 WHERE value::int >= (SELECT min(key) FROM table_1 WHERE key > random()) AND key = 6 RETURNING *
)
SELECT * FROM raw_data;
@ -766,13 +764,13 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
INSERT INTO table_1
SELECT * FROM table_2 where key = 1 AND
value::int IN
(WITH cte_1 AS
(WITH cte_1 AS MATERIALIZED
(
(SELECT key FROM table_1 WHERE key = 1)
INTERSECT
(SELECT key FROM table_1 WHERE key = 2)
),
cte_2 AS
cte_2 AS MATERIALIZED
(
(SELECT key FROM table_1 WHERE key = 3)
INTERSECT
@ -800,13 +798,13 @@ DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
-- hits all the shards because table_2.key != 1
INSERT INTO table_1
SELECT table_2.* FROM table_2,
(WITH cte_1 AS
(WITH cte_1 AS MATERIALIZED
(
(SELECT key FROM table_1 WHERE key = 1)
INTERSECT
(SELECT key FROM table_1 WHERE key = 2)
),
cte_2 AS
cte_2 AS MATERIALIZED
(
(SELECT key FROM table_1 WHERE key = 3)
INTERSECT
@ -924,8 +922,9 @@ FROM
WHERE
range_column IN ('A', 'E') AND
range_partitioned.data IN (SELECT data FROM some_data);
DEBUG: generating subplan XXX_1 for CTE some_data: SELECT data FROM intermediate_result_pruning.range_partitioned
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM intermediate_result_pruning.range_partitioned WHERE ((range_column OPERATOR(pg_catalog.=) ANY (ARRAY['A'::text, 'E'::text])) AND (data OPERATOR(pg_catalog.=) ANY (SELECT some_data.data FROM (SELECT intermediate_result.data FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(data integer)) some_data)))
DEBUG: CTE some_data is going to be inlined via distributed planning
DEBUG: generating subplan XXX_1 for subquery SELECT data FROM (SELECT range_partitioned.data FROM intermediate_result_pruning.range_partitioned) some_data
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM intermediate_result_pruning.range_partitioned WHERE ((range_column OPERATOR(pg_catalog.=) ANY (ARRAY['A'::text, 'E'::text])) AND (data OPERATOR(pg_catalog.=) ANY (SELECT intermediate_result.data FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(data integer))))
DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
count
@ -955,16 +954,16 @@ INSERT INTO stats (account_id, spent) VALUES ('foo', 100);
SELECT *
FROM
(
WITH accounts_cte AS (
WITH accounts_cte AS MATERIALIZED (
SELECT id AS account_id
FROM accounts
),
joined_stats_cte_1 AS (
joined_stats_cte_1 AS MATERIALIZED (
SELECT spent, account_id
FROM stats
INNER JOIN accounts_cte USING (account_id)
),
joined_stats_cte_2 AS (
joined_stats_cte_2 AS MATERIALIZED (
SELECT spent, account_id
FROM joined_stats_cte_1
INNER JOIN accounts_cte USING (account_id)
@ -992,16 +991,16 @@ SET citus.task_assignment_policy to 'round-robin';
SELECT *
FROM
(
WITH accounts_cte AS (
WITH accounts_cte AS MATERIALIZED (
SELECT id AS account_id
FROM accounts
),
joined_stats_cte_1 AS (
joined_stats_cte_1 AS MATERIALIZED (
SELECT spent, account_id
FROM stats
INNER JOIN accounts_cte USING (account_id)
),
joined_stats_cte_2 AS (
joined_stats_cte_2 AS MATERIALIZED (
SELECT spent, account_id
FROM joined_stats_cte_1
INNER JOIN accounts_cte USING (account_id)
@ -1027,10 +1026,10 @@ RESET citus.task_assignment_policy;
-- Insert..select is planned differently, make sure we have results everywhere.
-- We put the insert..select in a CTE here to prevent the CTE from being moved
-- into the select, which would follow the regular code path for select.
WITH stats AS (
WITH stats AS MATERIALIZED (
SELECT count(key) m FROM table_3
),
inserts AS (
inserts AS MATERIALIZED (
INSERT INTO table_2
SELECT key, count(*)
FROM table_1

78
src/test/regress/expected/limit_intermediate_size.out
View File

@ -1,13 +1,11 @@
SET citus.enable_repartition_joins to ON;
-- prevent PG 11 - PG 12 outputs to diverge
SET citus.enable_cte_inlining TO false;
SET citus.max_intermediate_result_size TO 2;
-- should fail because the copy size is ~4kB for each cte
WITH cte AS
WITH cte AS MATERIALIZED
(
SELECT * FROM users_table
),
cte2 AS (
cte2 AS MATERIALIZED (
SELECT * FROM events_table
)
SELECT cte.user_id, cte.value_2 FROM cte,cte2 ORDER BY 1,2 LIMIT 10;
@ -15,7 +13,7 @@ ERROR: the intermediate result size exceeds citus.max_intermediate_result_size
DETAIL: Citus restricts the size of intermediate results of complex subqueries and CTEs to avoid accidentally pulling large result sets into once place.
HINT: To run the current query, set citus.max_intermediate_result_size to a higher value or -1 to disable.
SET citus.max_intermediate_result_size TO 9;
WITH cte AS
WITH cte AS MATERIALIZED
(
SELECT
users_table.user_id, users_table.value_1, users_table.value_2
@ -26,7 +24,7 @@ WITH cte AS
on
(users_table.value_3=events_table.value_3)
),
cte2 AS (
cte2 AS MATERIALIZED(
SELECT * FROM events_table
)
SELECT
@ -53,7 +51,7 @@ LIMIT 10;
-- router queries should be able to get limitted too
SET citus.max_intermediate_result_size TO 2;
-- this should pass, since we fetch small portions in each subplan
with cte as (select * from users_table where user_id=1),
with cte as MATERIALIZED (select * from users_table where user_id=1),
cte2 as (select * from users_table where user_id=2),
cte3 as (select * from users_table where user_id=3),
cte4 as (select * from users_table where user_id=4),
@ -75,18 +73,18 @@ UNION
(1 row)
-- if we fetch the same amount of data at once, it should fail
WITH cte AS (SELECT * FROM users_table WHERE user_id IN (1,2,3,4,5))
WITH cte AS MATERIALIZED (SELECT * FROM users_table WHERE user_id IN (1,2,3,4,5))
SELECT * FROM cte ORDER BY 1,2,3,4,5 LIMIT 10;
ERROR: the intermediate result size exceeds citus.max_intermediate_result_size (currently 2 kB)
DETAIL: Citus restricts the size of intermediate results of complex subqueries and CTEs to avoid accidentally pulling large result sets into once place.
HINT: To run the current query, set citus.max_intermediate_result_size to a higher value or -1 to disable.
SET citus.max_intermediate_result_size TO 0;
-- this should fail
WITH cte AS (SELECT * FROM users_table WHERE user_id=1),
cte2 AS (SELECT * FROM users_table WHERE user_id=2),
cte3 AS (SELECT * FROM users_table WHERE user_id=3),
cte4 AS (SELECT * FROM users_table WHERE user_id=4),
cte5 AS (SELECT * FROM users_table WHERE user_id=5)
WITH cte AS MATERIALIZED (SELECT * FROM users_table WHERE user_id=1),
cte2 AS MATERIALIZED (SELECT * FROM users_table WHERE user_id=2),
cte3 AS MATERIALIZED (SELECT * FROM users_table WHERE user_id=3),
cte4 AS MATERIALIZED (SELECT * FROM users_table WHERE user_id=4),
cte5 AS MATERIALIZED (SELECT * FROM users_table WHERE user_id=5)
SELECT * FROM (
(SELECT * FROM cte)
UNION
@ -104,11 +102,11 @@ HINT: To run the current query, set citus.max_intermediate_result_size to a hig
-- this fails since cte-subplan exceeds limit even if cte2 and cte3 don't
-- WHERE EXISTS forces materialization in pg12
SET citus.max_intermediate_result_size TO 4;
WITH cte AS (
WITH cte2 AS (
WITH cte AS MATERIALIZED (
WITH cte2 AS MATERIALIZED (
SELECT * FROM users_table
),
cte3 AS (
cte3 AS MATERIALIZED(
SELECT * FROM events_table
)
SELECT * FROM cte2, cte3 WHERE cte2.user_id = cte3.user_id AND cte2.user_id = 1
@ -123,11 +121,11 @@ SELECT count(*) FROM cte WHERE EXISTS (select * from cte);
SET citus.max_intermediate_result_size TO 3;
-- this should fail since the cte-subplan exceeds the limit even if the
-- cte2 and cte3 does not
WITH cte AS (
WITH cte2 AS (
WITH cte AS MATERIALIZED (
WITH cte2 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id IN (3,4,5,6)
),
cte3 AS (
cte3 AS MATERIALIZED(
SELECT * FROM events_table WHERE event_type = 1
)
SELECT * FROM cte2, cte3 WHERE cte2.value_1 IN (SELECT value_2 FROM cte3)
@ -138,16 +136,16 @@ DETAIL: Citus restricts the size of intermediate results of complex subqueries
HINT: To run the current query, set citus.max_intermediate_result_size to a higher value or -1 to disable.
-- this will fail in remote execution
SET citus.max_intermediate_result_size TO 2;
WITH cte AS (
WITH cte2 AS (
WITH cte AS MATERIALIZED (
WITH cte2 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id IN (1, 2)
),
cte3 AS (
cte3 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id = 3
)
SELECT * FROM cte2 UNION (SELECT * FROM cte3)
),
cte4 AS (
cte4 AS MATERIALIZED (
SELECT * FROM events_table
)
SELECT * FROM cte UNION ALL
@ -157,16 +155,16 @@ DETAIL: Citus restricts the size of intermediate results of complex subqueries
HINT: To run the current query, set citus.max_intermediate_result_size to a higher value or -1 to disable.
SET citus.max_intermediate_result_size TO 1;
-- this will fail in router_executor
WITH cte AS (
WITH cte2 AS (
WITH cte AS MATERIALIZED (
WITH cte2 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id IN (1, 2)
),
cte3 AS (
cte3 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id = 3
)
SELECT * FROM cte2 UNION (SELECT * FROM cte3)
),
cte4 AS (
cte4 AS MATERIALIZED (
SELECT * FROM events_table
)
SELECT * FROM cte UNION ALL
@ -174,19 +172,19 @@ SELECT * FROM cte4 ORDER BY 1,2,3,4,5 LIMIT 5;
ERROR: the intermediate result size exceeds citus.max_intermediate_result_size (currently 1 kB)
DETAIL: Citus restricts the size of intermediate results of complex subqueries and CTEs to avoid accidentally pulling large result sets into once place.
HINT: To run the current query, set citus.max_intermediate_result_size to a higher value or -1 to disable.
-- Below that, all should pass since -1 disables the limit
-- Below that, all should pAS MATERIALIZEDs since -1 disables the limit
SET citus.max_intermediate_result_size TO -1;
-- real_time_executor + router_executor + real_time_executor will pass
WITH cte AS (
WITH cte2 AS (
WITH cte AS MATERIALIZED (
WITH cte2 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id IN (1, 2)
),
cte3 AS (
cte3 AS MATERIALIZED (
SELECT * FROM users_table WHERE user_id = 3
)
SELECT * FROM cte2 UNION (SELECT * FROM cte3)
),
cte4 AS (
cte4 AS MATERIALIZED (
SELECT * FROM events_table
)
SELECT * FROM cte UNION ALL
@ -201,7 +199,7 @@ SELECT * FROM cte4 ORDER BY 1,2,3,4,5 LIMIT 5;
(5 rows)
-- regular adaptive executor CTE, should work since -1 disables the limit
WITH cte AS
WITH cte AS MATERIALIZED
(
SELECT
users_table.user_id, users_table.value_1, users_table.value_2
@ -212,7 +210,7 @@ WITH cte AS
on
(users_table.value_2=events_table.value_2)
),
cte2 AS (
cte2 AS MATERIALIZED (
SELECT * FROM events_table
)
SELECT
@ -237,11 +235,11 @@ LIMIT 10;
(10 rows)
-- regular real-time CTE fetches around ~4kb data in each subplan
WITH cte AS
WITH cte AS MATERIALIZED
(
SELECT * FROM users_table
),
cte2 AS (
cte2 AS MATERIALIZED (
SELECT * FROM events_table
)
SELECT cte.user_id, cte.value_2 FROM cte,cte2 ORDER BY 1,2 LIMIT 10;
@ -260,7 +258,7 @@ SELECT cte.user_id, cte.value_2 FROM cte,cte2 ORDER BY 1,2 LIMIT 10;
(10 rows)
-- regular real-time query fetches ~4kB
WITH cte AS
WITH cte AS MATERIALIZED
(
SELECT * FROM users_table WHERE user_id IN (1,2,3,4,5)
)
@ -280,11 +278,11 @@ SELECT * FROM cte ORDER BY 1,2,3,4,5 LIMIT 10;
(10 rows)
-- nested CTEs
WITH cte AS (
WITH cte2 AS (
WITH cte AS MATERIALIZED (
WITH cte2 AS MATERIALIZED (
SELECT * FROM users_table
),
cte3 AS (
cte3 AS MATERIALIZED (
SELECT * FROM events_table
)
SELECT

348
src/test/regress/expected/locally_execute_intermediate_results.out
View File

@ -34,14 +34,12 @@ INSERT INTO table_2 VALUES (3, '3'), (4, '4'), (5, '5'),
INSERT INTO ref_table VALUES (1, '1'), (2, '2'), (3, '3'), (4, '4'), (5, '5'), (6, '6');
NOTICE: executing the command locally: INSERT INTO locally_execute_intermediate_results.ref_table_1580008 AS citus_table_alias (key, value) VALUES (1,'1'::text), (2,'2'::text), (3,'3'::text), (4,'4'::text), (5,'5'::text), (6,'6'::text)
INSERT INTO local_table VALUES (3, '3'), (4, '4'), (5, '5'), (6, '6');
-- prevent PG 11 - PG 12 outputs to diverge
-- and have a lot more CTEs recursively planned for the
-- sake of increasing the test coverage
SET citus.enable_cte_inlining TO false;
SET client_min_messages TO DEBUG1;
-- the query cannot be executed locally, but still because of
-- HAVING the intermediate result is written to local file as well
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -58,7 +56,7 @@ DEBUG: Subplan XXX_1 will be written to local file
(2 rows)
-- in this case, the HAVING Is also pushed down
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -76,8 +74,8 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
(2 rows)
-- subquery in the WHERE part of the query can be executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -104,8 +102,8 @@ NOTICE: executing the command locally: SELECT key FROM (SELECT intermediate_res
-- subquery in the WHERE part of the query should not be executed locally
-- because it can be pushed down with the jointree
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(key) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(key) FROM table_2)
SELECT
count(*)
FROM
@ -126,9 +124,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
(0 rows)
-- now all the intermediate results are safe to be in local files
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(key) FROM table_2),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(key) FROM table_2),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -151,8 +149,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- multiple CTEs are joined inside HAVING, so written to file
-- locally, but nothing executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -171,8 +169,8 @@ DEBUG: Subplan XXX_2 will be written to local file
(2 rows)
-- same as above, but HAVING pushed down to workers
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -195,9 +193,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- multiple CTEs are joined inside HAVING, so written to file
-- locally, also the join tree contains only another CTE, so should be
-- executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -221,7 +219,7 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- now, the CTE is going to be written locally,
-- plus that is going to be read locally because
-- of the aggragate over the cte in HAVING
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -241,7 +239,7 @@ NOTICE: executing the command locally: SELECT max(max) AS max FROM (SELECT inte
(2 rows)
-- same as above, but with HAVING pushed down
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -262,8 +260,8 @@ NOTICE: executing the command locally: SELECT max(max) AS max FROM (SELECT inte
(2 rows)
-- two ctes are going to be written locally and executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT * FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM table_1)
SELECT
count(*)
FROM
@ -288,7 +286,7 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- this time the same CTE is both joined with a distributed
-- table and used in HAVING
WITH a AS (SELECT * FROM table_1 ORDER BY 1,2 DESC LIMIT 1)
WITH a AS MATERIALIZED (SELECT * FROM table_1 ORDER BY 1,2 DESC LIMIT 1)
SELECT count(*),
key
FROM a JOIN table_2 USING (key)
@ -307,7 +305,7 @@ DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
-- this time the same CTE is both joined with a distributed
-- table and used in HAVING -- but used in another subquery/aggregate
-- so one more level of recursive planning
WITH a AS (SELECT * FROM table_1)
WITH a AS MATERIALIZED (SELECT * FROM table_1)
SELECT count(*),
key
FROM a JOIN table_2 USING (key)
@ -327,13 +325,13 @@ NOTICE: executing the command locally: SELECT max(value) AS max FROM (SELECT in
(1 row)
-- same query as the above, without the aggragate
WITH a AS (SELECT max(key) as key, max(value) as value FROM ref_table)
WITH a AS MATERIALIZED (SELECT max(key) as key, max(value) as value FROM ref_table)
SELECT count(*),
key
FROM a JOIN ref_table USING (key)
GROUP BY key
HAVING (max(ref_table.value) <= (SELECT value FROM a));
NOTICE: executing the command locally: WITH a AS (SELECT max(ref_table_1.key) AS key, max(ref_table_1.value) AS value FROM locally_execute_intermediate_results.ref_table_1580008 ref_table_1) SELECT count(*) AS count, a.key FROM (a JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key)) GROUP BY a.key HAVING (max(ref_table.value) OPERATOR(pg_catalog.<=) (SELECT a_1.value FROM a a_1))
NOTICE: executing the command locally: WITH a AS MATERIALIZED (SELECT max(ref_table_1.key) AS key, max(ref_table_1.value) AS value FROM locally_execute_intermediate_results.ref_table_1580008 ref_table_1) SELECT count(*) AS count, a.key FROM (a JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key)) GROUP BY a.key HAVING (max(ref_table.value) OPERATOR(pg_catalog.<=) (SELECT a_1.value FROM a a_1))
count | key
---------------------------------------------------------------------
1 | 6
@ -341,9 +339,9 @@ NOTICE: executing the command locally: WITH a AS (SELECT max(ref_table_1.key) A
-- some edge cases around CTEs used inside other CTEs
-- everything can be executed locally
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) FROM cte_2)
WITH cte_1 as MATERIALIZED (SELECT * FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) FROM cte_2)
SELECT * FROM cte_3;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
@ -361,9 +359,9 @@ NOTICE: executing the command locally: SELECT max FROM (SELECT intermediate_res
(1 row)
-- the join between cte_3 and table_2 has to happen remotely
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
WITH cte_1 as MATERIALIZED (SELECT * FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN table_2 USING (key) WHERE table_2.key = 1;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
@ -379,9 +377,9 @@ NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT inte
(0 rows)
-- the join between cte_3 and table_2 has to happen remotely
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
WITH cte_1 as MATERIALIZED (SELECT * FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN ref_table USING (key);
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
@ -402,9 +400,9 @@ NOTICE: executing the command locally: SELECT cte_3.key, ref_table.value FROM (
-- some cases around router queries
-- a router query, but the having has two cte joins
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2
WHERE KEY = 3
@ -422,9 +420,9 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- a router query, but the having has two cte joins
-- and the jointree has a join with another cte
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2 JOIN cte_3 USING(key)
WHERE KEY = 3
@ -444,9 +442,9 @@ DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
-- a router query, but the having has two cte joins
-- and the jointree has a join with the same CTEs
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2 JOIN cte_3 USING(key) JOIN cte_2 ON (key = MAX::int) JOIN cte_1 USING(MAX)
WHERE KEY = 3
@ -465,8 +463,8 @@ DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
(0 rows)
-- subPlans needed remotely as the subquery is pushed down
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT key FROM table_1 GROUP BY key HAVING max(value) > (SELECT * FROM cte_1)) as foo,
(SELECT key FROM table_2 GROUP BY key HAVING max(value) > (SELECT * FROM cte_2)) as bar
@ -484,8 +482,8 @@ DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
-- the second subquery needs to be recursively planned due to non-colocated subquery join
-- so cte_2 becomes part of master query of that recursive subquery planning
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT value AS key FROM table_1 GROUP BY value HAVING max(value) > (SELECT * FROM cte_1)) as foo,
(SELECT value AS key FROM table_2 GROUP BY value HAVING max(value) > (SELECT * FROM cte_2)) as bar
@ -505,8 +503,8 @@ NOTICE: executing the command locally: SELECT foo.key, bar.key FROM (SELECT int
(0 rows)
-- similar to above, but having pushed down
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT key FROM table_1 GROUP BY key HAVING max(value) > (SELECT * FROM cte_1)) as foo,
(SELECT key FROM table_2 GROUP BY key HAVING max(value) > (SELECT * FROM cte_2)) as bar
@ -526,8 +524,8 @@ DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
(0 rows)
-- now, forcing all subqueries to be on the local node
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT value AS key FROM table_1 GROUP BY value HAVING max(value) > (SELECT * FROM cte_1) LIMIT 1) as foo,
(SELECT value AS key FROM table_2 GROUP BY value HAVING max(value) > (SELECT * FROM cte_2) LIMIT 1) as bar
@ -547,7 +545,7 @@ NOTICE: executing the command locally: SELECT foo.key, bar.key FROM (SELECT int
(0 rows)
-- queries in which the last step has only CTEs can use local tables
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -564,8 +562,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM locally_ex
1
(2 rows)
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -595,13 +593,9 @@ SET search_path TO locally_execute_intermediate_results;
SET citus.log_intermediate_results TO TRUE;
SET citus.log_local_commands TO TRUE;
SET client_min_messages TO DEBUG1;
-- prevent PG 11 - PG 12 outputs to diverge
-- and have a lot more CTEs recursively planned for the
-- sake of increasing the test coverage
SET citus.enable_cte_inlining TO false;
-- the query cannot be executed locally, but still because of
-- HAVING the intermediate result is written to local file as well
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -623,8 +617,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- On non-mx case the subquery in the WHERE part of the query can be executed locally
-- however, on Citus MX we have this limitation where the query cannot be executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -656,8 +650,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- subquery in the WHERE part of the query should not be executed locally
-- because it can be pushed down with the jointree
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(key) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(key) FROM table_2)
SELECT
count(*)
FROM
@ -685,9 +679,9 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- although all the intermediate results are safe to be in local files
-- we currently do not support it on Citus MX
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(key) FROM table_2),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(key) FROM table_2),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -716,8 +710,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- multiple CTEs are joined inside HAVING, so written to file
-- locally, but nothing executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -744,9 +738,9 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- multiple CTEs are joined inside HAVING, so written to file
-- locally, also the join tree contains only another CTE, so should be
-- executed locally, but not on an Citus MX worker
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -777,7 +771,7 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- plus that could have been read locally on the coordinator
-- because of the aggragate over the cte in HAVING
-- but not on Citus MX
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -802,8 +796,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- two could have been written locally and executed locally
-- on the coordinator, but not on the workers
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT * FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM table_1)
SELECT
count(*)
FROM
@ -832,7 +826,7 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- this time the same CTE is both joined with a distributed
-- table and used in HAVING
WITH a AS (SELECT * FROM table_1 ORDER BY 1,2 DESC LIMIT 1)
WITH a AS MATERIALIZED (SELECT * FROM table_1 ORDER BY 1,2 DESC LIMIT 1)
SELECT count(*),
key
FROM a JOIN table_2 USING (key)
@ -854,7 +848,7 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- this time the same CTE is both joined with a distributed
-- table and used in HAVING -- but used in another subquery/aggregate
-- so one more level of recursive planning
WITH a AS (SELECT * FROM table_1)
WITH a AS MATERIALIZED (SELECT * FROM table_1)
SELECT count(*),
key
FROM a JOIN table_2 USING (key)
@ -877,13 +871,13 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
(1 row)
-- same query as the above, without the aggragate
WITH a AS (SELECT max(key) as key, max(value) as value FROM ref_table)
WITH a AS MATERIALIZED (SELECT max(key) as key, max(value) as value FROM ref_table)
SELECT count(*),
key
FROM a JOIN ref_table USING (key)
GROUP BY key
HAVING (max(ref_table.value) <= (SELECT value FROM a));
NOTICE: executing the command locally: WITH a AS (SELECT max(ref_table_1.key) AS key, max(ref_table_1.value) AS value FROM locally_execute_intermediate_results.ref_table_1580008 ref_table_1) SELECT count(*) AS count, a.key FROM (a JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key)) GROUP BY a.key HAVING (max(ref_table.value) OPERATOR(pg_catalog.<=) (SELECT a_1.value FROM a a_1))
NOTICE: executing the command locally: WITH a AS MATERIALIZED (SELECT max(ref_table_1.key) AS key, max(ref_table_1.value) AS value FROM locally_execute_intermediate_results.ref_table_1580008 ref_table_1) SELECT count(*) AS count, a.key FROM (a JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key)) GROUP BY a.key HAVING (max(ref_table.value) OPERATOR(pg_catalog.<=) (SELECT a_1.value FROM a a_1))
count | key
---------------------------------------------------------------------
1 | 6
@ -892,9 +886,9 @@ NOTICE: executing the command locally: WITH a AS (SELECT max(ref_table_1.key) A
-- some edge cases around CTEs used inside other CTEs
-- everything could be executed locally on the coordinator,
-- but not on the worker
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) FROM cte_2)
WITH cte_1 as MATERIALIZED (SELECT * FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) FROM cte_2)
SELECT * FROM cte_3;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
@ -916,9 +910,9 @@ NOTICE: executing the command locally: SELECT max FROM (SELECT intermediate_res
-- the join between cte_3 and table_2 has to could have happened
-- locally since the key = 1 resides on this node
-- but because of the current implementation limitations we can't
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
WITH cte_1 as MATERIALIZED (SELECT * FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN table_2 USING (key) WHERE table_2.key = 1;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
@ -938,9 +932,9 @@ NOTICE: executing the command locally: SELECT cte_3.key, table_2.value FROM ((S
-- the join between cte_3 and table_2 has to cannot happen
-- locally because the key = 2 resides on a remote node
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
WITH cte_1 as MATERIALIZED (SELECT * FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN table_2 USING (key) WHERE table_2.key = 2;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
@ -959,9 +953,9 @@ NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT inte
-- the join between cte_3 and ref can could have happened locally
-- but because of the current implementation limitations we can't
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
WITH cte_1 as MATERIALIZED (SELECT * FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN ref_table USING (key);
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
@ -984,9 +978,9 @@ NOTICE: executing the command locally: SELECT cte_3.key, ref_table.value FROM (
-- some cases around router queries
-- a router query, but the having has two cte joins
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2
WHERE KEY = 3
@ -1008,9 +1002,9 @@ NOTICE: executing the command locally: SELECT max(value) AS max FROM locally_ex
-- a router query, but the having has two cte joins
-- and the jointree has a join with another cte
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2 JOIN cte_3 USING(key)
WHERE KEY = 3
@ -1036,9 +1030,9 @@ NOTICE: executing the command locally: SELECT key, value FROM locally_execute_i
-- a router query, but the having has two cte joins
-- and the jointree has a join with the same CTEs
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2 JOIN cte_3 USING(key) JOIN cte_2 ON (key = MAX::int) JOIN cte_1 USING(MAX)
WHERE KEY = 3
@ -1062,8 +1056,8 @@ NOTICE: executing the command locally: SELECT key, value FROM locally_execute_i
(0 rows)
-- subPlans needed remotely as the subquery is pushed down
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT key FROM table_1 GROUP BY key HAVING max(value) > (SELECT * FROM cte_1)) as foo,
(SELECT key FROM table_2 GROUP BY key HAVING max(value) > (SELECT * FROM cte_2)) as bar
@ -1087,8 +1081,8 @@ NOTICE: executing the command locally: SELECT worker_column_1 AS key, worker_co
-- the second subquery needs to be recursively planned due to non-colocated subquery join
-- so cte_2 becomes part of master query of that recursive subquery planning
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT value AS key FROM table_1 GROUP BY value HAVING max(value) > (SELECT * FROM cte_1)) as foo,
(SELECT value AS key FROM table_2 GROUP BY value HAVING max(value) > (SELECT * FROM cte_2)) as bar
@ -1116,8 +1110,8 @@ NOTICE: executing the command locally: SELECT foo.key, bar.key FROM (SELECT int
(0 rows)
-- now, forcing all subqueries to be on the local node
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT value AS key FROM table_1 GROUP BY value HAVING max(value) > (SELECT * FROM cte_1) LIMIT 1) as foo,
(SELECT value AS key FROM table_2 GROUP BY value HAVING max(value) > (SELECT * FROM cte_2) LIMIT 1) as bar
@ -1148,7 +1142,7 @@ NOTICE: executing the command locally: SELECT foo.key, bar.key FROM (SELECT int
set citus.task_assignment_policy TO "round-robin" ;
-- the query cannot be executed locally, but still because of
-- HAVING the intermediate result is written to local file as well
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -1170,8 +1164,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- On non-mx case the subquery in the WHERE part of the query can be executed locally
-- however, on Citus MX we have this limitation where the query cannot be executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -1203,8 +1197,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- subquery in the WHERE part of the query should not be executed locally
-- because it can be pushed down with the jointree
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(key) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(key) FROM table_2)
SELECT
count(*)
FROM
@ -1232,9 +1226,9 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- although all the intermediate results are safe to be in local files
-- we currently do not support it on Citus MX
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(key) FROM table_2),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(key) FROM table_2),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -1263,8 +1257,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- multiple CTEs are joined inside HAVING, so written to file
-- locally, but nothing executed locally
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -1293,9 +1287,9 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- multiple CTEs are joined inside HAVING, so written to file
-- locally, also the join tree contains only another CTE, so should be
-- executed locally, but not on an Citus MX worker
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT
count(*)
FROM
@ -1325,7 +1319,7 @@ NOTICE: executing the command locally: SELECT key, value FROM locally_execute_i
-- plus that could have been read locally on the coordinator
-- because of the aggragate over the cte in HAVING
-- but not on Citus MX
WITH cte_1 AS (SELECT max(value) FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1)
SELECT
count(*)
FROM
@ -1350,8 +1344,8 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- two could have been written locally and executed locally
-- on the coordinator, but not on the workers
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT * FROM table_1)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT * FROM table_1)
SELECT
count(*)
FROM
@ -1379,7 +1373,7 @@ NOTICE: executing the command locally: SELECT count(*) AS count FROM (SELECT in
-- this time the same CTE is both joined with a distributed
-- table and used in HAVING
WITH a AS (SELECT * FROM table_1 ORDER BY 1,2 DESC LIMIT 1)
WITH a AS MATERIALIZED (SELECT * FROM table_1 ORDER BY 1,2 DESC LIMIT 1)
SELECT count(*),
key
FROM a JOIN table_2 USING (key)
@ -1401,7 +1395,7 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
-- this time the same CTE is both joined with a distributed
-- table and used in HAVING -- but used in another subquery/aggregate
-- so one more level of recursive planning
WITH a AS (SELECT * FROM table_1)
WITH a AS MATERIALIZED (SELECT * FROM table_1)
SELECT count(*),
key
FROM a JOIN table_2 USING (key)
@ -1423,13 +1417,13 @@ NOTICE: executing the command locally: SELECT count(*) AS count, worker_column_
(1 row)
-- same query as the above, without the aggragate
WITH a AS (SELECT max(key) as key, max(value) as value FROM ref_table)
WITH a AS MATERIALIZED (SELECT max(key) as key, max(value) as value FROM ref_table)
SELECT count(*),
key
FROM a JOIN ref_table USING (key)
GROUP BY key
HAVING (max(ref_table.value) <= (SELECT value FROM a));
NOTICE: executing the command locally: WITH a AS (SELECT max(ref_table_1.key) AS key, max(ref_table_1.value) AS value FROM locally_execute_intermediate_results.ref_table_1580008 ref_table_1) SELECT count(*) AS count, a.key FROM (a JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key)) GROUP BY a.key HAVING (max(ref_table.value) OPERATOR(pg_catalog.<=) (SELECT a_1.value FROM a a_1))
NOTICE: executing the command locally: WITH a AS MATERIALIZED (SELECT max(ref_table_1.key) AS key, max(ref_table_1.value) AS value FROM locally_execute_intermediate_results.ref_table_1580008 ref_table_1) SELECT count(*) AS count, a.key FROM (a JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key)) GROUP BY a.key HAVING (max(ref_table.value) OPERATOR(pg_catalog.<=) (SELECT a_1.value FROM a a_1))
count | key
---------------------------------------------------------------------
1 | 6
@ -1439,20 +1433,18 @@ NOTICE: executing the command locally: WITH a AS (SELECT max(ref_table_1.key) A
-- everything could be executed locally on the coordinator,
-- but not on the worker
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) FROM cte_2)
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) FROM cte_2)
SELECT * FROM cte_3;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
DEBUG: generating subplan XXX_3 for CTE cte_3: SELECT max(key) AS max FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT max FROM (SELECT intermediate_result.max FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(max integer)) cte_3
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: generating subplan XXX_1 for CTE cte_2: SELECT key, value FROM (SELECT table_1.key, table_1.value FROM locally_execute_intermediate_results.table_1) cte_1
DEBUG: generating subplan XXX_2 for CTE cte_3: SELECT max(key) AS max FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT max FROM (SELECT intermediate_result.max FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(max integer)) cte_3
DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580000 table_1 WHERE true
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580002 table_1 WHERE true
DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT max FROM (SELECT intermediate_result.max FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(max integer)) cte_3
NOTICE: executing the command locally: SELECT max(key) AS max FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
max
---------------------------------------------------------------------
4
@ -1462,20 +1454,19 @@ NOTICE: executing the command locally: SELECT max FROM (SELECT intermediate_res
-- locally since the key = 1 resides on this node
-- but because of the current implementation limitations we can't
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN table_2 USING (key) WHERE table_2.key = 1;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
DEBUG: generating subplan XXX_3 for CTE cte_3: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cte_3.key, table_2.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.table_2 USING (key)) WHERE (table_2.key OPERATOR(pg_catalog.=) 1)
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: generating subplan XXX_1 for CTE cte_2: SELECT key, value FROM (SELECT table_1.key, table_1.value FROM locally_execute_intermediate_results.table_1) cte_1
DEBUG: generating subplan XXX_2 for CTE cte_3: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cte_3.key, table_2.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.table_2 USING (key)) WHERE (table_2.key OPERATOR(pg_catalog.=) 1)
DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580000 table_1 WHERE true
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580002 table_1 WHERE true
DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
NOTICE: executing the command locally: SELECT cte_3.key, table_2.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.table_2_1580004 table_2(key, value) USING (key)) WHERE (table_2.key OPERATOR(pg_catalog.=) 1)
NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
NOTICE: executing the command locally: SELECT cte_3.key, table_2.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.table_2_1580004 table_2(key, value) USING (key)) WHERE (table_2.key OPERATOR(pg_catalog.=) 1)
key | value
---------------------------------------------------------------------
(0 rows)
@ -1483,19 +1474,17 @@ NOTICE: executing the command locally: SELECT cte_3.key, table_2.value FROM ((S
-- the join between cte_3 and table_2 has to cannot happen
-- locally because the key = 2 resides on a remote node
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN table_2 USING (key) WHERE table_2.key = 2;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
DEBUG: generating subplan XXX_3 for CTE cte_3: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cte_3.key, table_2.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.table_2 USING (key)) WHERE (table_2.key OPERATOR(pg_catalog.=) 2)
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: generating subplan XXX_1 for CTE cte_2: SELECT key, value FROM (SELECT table_1.key, table_1.value FROM locally_execute_intermediate_results.table_1) cte_1
DEBUG: generating subplan XXX_2 for CTE cte_3: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cte_3.key, table_2.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.table_2 USING (key)) WHERE (table_2.key OPERATOR(pg_catalog.=) 2)
DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580000 table_1 WHERE true
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580002 table_1 WHERE true
DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
key | value
---------------------------------------------------------------------
(0 rows)
@ -1503,21 +1492,20 @@ NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT inte
-- the join between cte_3 and ref can could have happened locally
-- but because of the current implementation limitations we can't
WITH cte_1 as (SELECT * FROM table_1),
cte_2 AS (SELECT * FROM cte_1),
cte_3 AS (SELECT max(key) as key FROM cte_2)
cte_2 AS MATERIALIZED (SELECT * FROM cte_1),
cte_3 AS MATERIALIZED (SELECT max(key) as key FROM cte_2)
SELECT * FROM cte_3 JOIN ref_table USING (key);
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT key, value FROM locally_execute_intermediate_results.table_1
DEBUG: generating subplan XXX_2 for CTE cte_2: SELECT key, value FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_1
DEBUG: generating subplan XXX_3 for CTE cte_3: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cte_3.key, ref_table.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.ref_table USING (key))
DEBUG: CTE cte_1 is going to be inlined via distributed planning
DEBUG: generating subplan XXX_1 for CTE cte_2: SELECT key, value FROM (SELECT table_1.key, table_1.value FROM locally_execute_intermediate_results.table_1) cte_1
DEBUG: generating subplan XXX_2 for CTE cte_3: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT cte_3.key, ref_table.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.ref_table USING (key))
DEBUG: Subplan XXX_1 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580000 table_1 WHERE true
NOTICE: executing the command locally: SELECT key, value FROM locally_execute_intermediate_results.table_1_1580002 table_1 WHERE true
DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
DEBUG: Subplan XXX_3 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
NOTICE: executing the command locally: SELECT cte_3.key, ref_table.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_3'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key))
DEBUG: Subplan XXX_2 will be sent to localhost:xxxxx
NOTICE: executing the command locally: SELECT max(key) AS key FROM (SELECT intermediate_result.key, intermediate_result.value FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(key integer, value text)) cte_2
NOTICE: executing the command locally: SELECT cte_3.key, ref_table.value FROM ((SELECT intermediate_result.key FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(key integer)) cte_3 JOIN locally_execute_intermediate_results.ref_table_1580008 ref_table(key, value) USING (key))
key | value
---------------------------------------------------------------------
4 | 4
@ -1525,9 +1513,9 @@ NOTICE: executing the command locally: SELECT cte_3.key, ref_table.value FROM (
-- some cases around router queries
-- a router query, but the having has two cte joins
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2
WHERE KEY = 3
@ -1549,9 +1537,9 @@ NOTICE: executing the command locally: SELECT max(value) AS max FROM locally_ex
-- a router query, but the having has two cte joins
-- and the jointree has a join with another cte
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2 JOIN cte_3 USING(key)
WHERE KEY = 3
@ -1576,9 +1564,9 @@ NOTICE: executing the command locally: SELECT key, value FROM locally_execute_i
(0 rows)
-- the same query as above, try to hit local node with either of the queries
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2 JOIN cte_3 USING(key)
WHERE KEY = 3
@ -1604,9 +1592,9 @@ NOTICE: executing the command locally: SELECT key, value FROM locally_execute_i
-- a router query, but the having has two cte joins
-- and the jointree has a join with the same CTEs
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_1),
cte_3 AS (SELECT * FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_3 AS MATERIALIZED (SELECT * FROM table_2)
SELECT count(*)
FROM table_2 JOIN cte_3 USING(key) JOIN cte_2 ON (key = MAX::int) JOIN cte_1 USING(MAX)
WHERE KEY = 3
@ -1631,8 +1619,8 @@ NOTICE: executing the command locally: SELECT key, value FROM locally_execute_i
(0 rows)
- subPlans needed remotely as the subquery is pushed down
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT key FROM table_1 GROUP BY key HAVING max(value) > (SELECT * FROM cte_1)) as foo,
(SELECT key FROM table_2 GROUP BY key HAVING max(value) > (SELECT * FROM cte_2)) as bar
@ -1640,8 +1628,8 @@ SELECT * FROM
ERROR: syntax error at or near "-"
-- the second subquery needs to be recursively planned due to non-colocated subquery join
-- so cte_2 becomes part of master query of that recursive subquery planning
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT value AS key FROM table_1 GROUP BY value HAVING max(value) > (SELECT * FROM cte_1)) as foo,
(SELECT value AS key FROM table_2 GROUP BY value HAVING max(value) > (SELECT * FROM cte_2)) as bar
@ -1668,8 +1656,8 @@ NOTICE: executing the command locally: SELECT worker_column_1 AS key, max(worke
(0 rows)
-- now, forcing all subqueries to be on the local node
WITH cte_1 AS (SELECT max(value) FROM table_1),
cte_2 AS (SELECT max(value) FROM table_2)
WITH cte_1 AS MATERIALIZED (SELECT max(value) FROM table_1),
cte_2 AS MATERIALIZED (SELECT max(value) FROM table_2)
SELECT * FROM
(SELECT value AS key FROM table_1 GROUP BY value HAVING max(value) > (SELECT * FROM cte_1) LIMIT 1) as foo,
(SELECT value AS key FROM table_2 GROUP BY value HAVING max(value) > (SELECT * FROM cte_2) LIMIT 1) as bar

13
src/test/regress/expected/mixed_relkind_tests.out
View File

@ -14,7 +14,6 @@ SELECT 1 FROM master_add_node('localhost', :master_port, groupId => 0);
RESET client_min_messages;
-- make results consistent
SET citus.enable_cte_inlining TO OFF;
-- create test tables
CREATE TABLE postgres_local_table (a int);
CREATE TABLE partitioned_postgres_local_table(a int) PARTITION BY RANGE(a);
@ -470,7 +469,7 @@ SELECT COUNT(*) FROM unlogged_distributed_table u1 JOIN partitioned_distributed_
RESET citus.enable_repartition_joins;
-- joins with cte's
WITH cte_1 AS (SELECT * FROM partitioned_distributed_table)
WITH cte_1 AS MATERIALIZED (SELECT * FROM partitioned_distributed_table)
SELECT COUNT(*) FROM cte_1;
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT a, b FROM mixed_relkind_tests.partitioned_distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM (SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte_1
@ -479,7 +478,7 @@ DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS c
78
(1 row)
WITH cte_1 AS (SELECT * FROM partitioned_distributed_table)
WITH cte_1 AS MATERIALIZED (SELECT * FROM partitioned_distributed_table)
SELECT COUNT(*) FROM cte_1 JOIN partitioned_distributed_table USING (a);
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT a, b FROM mixed_relkind_tests.partitioned_distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte_1 JOIN mixed_relkind_tests.partitioned_distributed_table USING (a))
@ -488,7 +487,7 @@ DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS c
1014
(1 row)
WITH cte_1 AS (SELECT * FROM foreign_distributed_table)
WITH cte_1 AS MATERIALIZED (SELECT * FROM foreign_distributed_table)
SELECT COUNT(*) FROM cte_1 JOIN foreign_distributed_table USING (a);
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT a, b FROM mixed_relkind_tests.foreign_distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte_1 JOIN mixed_relkind_tests.foreign_distributed_table USING (a))
@ -497,7 +496,7 @@ DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS c
0
(1 row)
WITH cte_1 AS (SELECT * FROM partitioned_distributed_table)
WITH cte_1 AS MATERIALIZED (SELECT * FROM partitioned_distributed_table)
SELECT COUNT(*) FROM cte_1 JOIN partitioned_distributed_table USING (b);
DEBUG: generating subplan XXX_1 for CTE cte_1: SELECT a, b FROM mixed_relkind_tests.partitioned_distributed_table
DEBUG: Plan XXX query after replacing subqueries and CTEs: SELECT count(*) AS count FROM ((SELECT intermediate_result.a, intermediate_result.b FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(a integer, b integer)) cte_1 JOIN mixed_relkind_tests.partitioned_distributed_table USING (b))
@ -518,8 +517,8 @@ FROM (SELECT * FROM unlogged_distributed_table WHERE b = 1) AS sub1,
(SELECT * FROM unlogged_distributed_table WHERE b = 2) AS sub2
WHERE sub1.a = sub2.a AND sub1.a = dt.a AND dt.a > 1;
-- multi shard non-colocated update
WITH cte1 AS (SELECT * FROM partitioned_distributed_table WHERE b = 1),
cte2 AS (SELECT * FROM partitioned_distributed_table WHERE b = 2)
WITH cte1 AS MATERIALIZED (SELECT * FROM partitioned_distributed_table WHERE b = 1),
cte2 AS MATERIALIZED (SELECT * FROM partitioned_distributed_table WHERE b = 2)
UPDATE partitioned_distributed_table dt SET b = cte1.a + cte2.a
FROM cte1, cte2 WHERE cte1.a != cte2.a AND cte1.a = dt.a AND dt.a > 1;
DEBUG: generating subplan XXX_1 for CTE cte1: SELECT a, b FROM mixed_relkind_tests.partitioned_distributed_table WHERE (b OPERATOR(pg_catalog.=) 1)

555
src/test/regress/expected/multi_complex_expressions_0.out
View File

@ -1,555 +0,0 @@
--
-- MULTI_COMPLEX_EXPRESSIONS
--
-- Check that we can correctly handle complex expressions and aggregates.
SELECT sum(l_quantity) / avg(l_quantity) FROM lineitem;
?column?
---------------------------------------------------------------------
12000.0000000000000000
(1 row)
SELECT sum(l_quantity) / (10 * avg(l_quantity)) FROM lineitem;
?column?
---------------------------------------------------------------------
1200.0000000000000000
(1 row)
SELECT (sum(l_quantity) / (10 * avg(l_quantity))) + 11 FROM lineitem;
?column?
---------------------------------------------------------------------
1211.0000000000000000
(1 row)
SELECT avg(l_quantity) as average FROM lineitem;
average
---------------------------------------------------------------------
25.4462500000000000
(1 row)
SELECT 100 * avg(l_quantity) as average_times_hundred FROM lineitem;
average_times_hundred
---------------------------------------------------------------------
2544.6250000000000000
(1 row)
SELECT 100 * avg(l_quantity) / 10 as average_times_ten FROM lineitem;
average_times_ten
---------------------------------------------------------------------
254.4625000000000000
(1 row)
SELECT l_quantity, 10 * count(*) count_quantity FROM lineitem
GROUP BY l_quantity ORDER BY count_quantity, l_quantity;
l_quantity | count_quantity
---------------------------------------------------------------------
44.00 | 2150
38.00 | 2160
45.00 | 2180
13.00 | 2190
47.00 | 2200
29.00 | 2220
36.00 | 2230
49.00 | 2230
3.00 | 2270
35.00 | 2280
18.00 | 2290
31.00 | 2290
43.00 | 2290
14.00 | 2300
16.00 | 2300
17.00 | 2300
26.00 | 2300
7.00 | 2320
10.00 | 2340
34.00 | 2340
15.00 | 2350
25.00 | 2360
33.00 | 2360
42.00 | 2360
2.00 | 2370
12.00 | 2410
37.00 | 2410
6.00 | 2420
22.00 | 2420
1.00 | 2430
19.00 | 2430
4.00 | 2440
20.00 | 2460
48.00 | 2460
41.00 | 2470
24.00 | 2490
27.00 | 2490
8.00 | 2500
11.00 | 2500
5.00 | 2540
21.00 | 2550
32.00 | 2550
9.00 | 2580
39.00 | 2600
46.00 | 2600
50.00 | 2600
23.00 | 2610
30.00 | 2640
40.00 | 2690
28.00 | 2730
(50 rows)
-- Check that we can handle complex select clause expressions.
SELECT count(*) FROM lineitem
WHERE octet_length(l_comment || l_comment) > 40;
count
---------------------------------------------------------------------
8148
(1 row)
SELECT count(*) FROM lineitem
WHERE octet_length(concat(l_comment, l_comment)) > 40;
count
---------------------------------------------------------------------
8148
(1 row)
SELECT count(*) FROM lineitem
WHERE octet_length(l_comment) + octet_length('randomtext'::text) > 40;
count
---------------------------------------------------------------------
4611
(1 row)
SELECT count(*) FROM lineitem
WHERE octet_length(l_comment) + 10 > 40;
count
---------------------------------------------------------------------
4611
(1 row)
SELECT count(*) FROM lineitem
WHERE (l_receiptdate::timestamp - l_shipdate::timestamp) > interval '5 days';
count
---------------------------------------------------------------------
10008
(1 row)
-- can push down queries where no columns present on the WHERE clause
SELECT count(*) FROM lineitem WHERE random() = -0.1;
count
---------------------------------------------------------------------
0
(1 row)
-- boolean tests can be pushed down
SELECT count(*) FROM lineitem
WHERE (l_partkey > 10000) is true;
count
---------------------------------------------------------------------
11423
(1 row)
-- scalar array operator expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE l_partkey = ANY(ARRAY[19353, 19354, 19355]);
count
---------------------------------------------------------------------
1
(1 row)
-- some more scalar array operator expressions
SELECT count(*) FROM lineitem
WHERE l_partkey = ALL(ARRAY[19353]);
count
---------------------------------------------------------------------
1
(1 row)
-- operator expressions involving arrays
SELECT count(*) FROM lineitem
WHERE ARRAY[19353, 19354, 19355] @> ARRAY[l_partkey];
count
---------------------------------------------------------------------
1
(1 row)
-- coerced via io expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE (l_quantity/100)::int::bool::text::bool;
count
---------------------------------------------------------------------
260
(1 row)
-- case expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE (CASE WHEN l_orderkey > 4000 THEN l_partkey / 100 > 1 ELSE false END);
count
---------------------------------------------------------------------
7948
(1 row)
-- coalesce expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE COALESCE((l_partkey/50000)::bool, false);
count
---------------------------------------------------------------------
9122
(1 row)
-- nullif expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE NULLIF((l_partkey/50000)::bool, false);
count
---------------------------------------------------------------------
9122
(1 row)
-- null test expressions can be pushed down
SELECT count(*) FROM orders
WHERE o_comment IS NOT null;
count
---------------------------------------------------------------------
2985
(1 row)
-- functions can be pushed down
SELECT count(*) FROM lineitem
WHERE isfinite(l_shipdate);
count
---------------------------------------------------------------------
12000
(1 row)
-- constant expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE 0 != 0;
count
---------------------------------------------------------------------
0
(1 row)
-- distinct expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE l_partkey IS DISTINCT FROM 50040;
count
---------------------------------------------------------------------
11999
(1 row)
-- row compare expression can be pushed down
SELECT count(*) FROM lineitem
WHERE row(l_partkey, 2, 3) > row(2000, 2, 3);
count
---------------------------------------------------------------------
11882
(1 row)
-- combination of different expressions can be pushed down
SELECT count(*) FROM lineitem
WHERE
(l_quantity/100)::int::bool::text::bool AND
CASE WHEN l_orderkey > 4000 THEN l_partkey / 100 > 1 ELSE false END AND
COALESCE((l_partkey/50000)::bool, false) AND
NULLIF((l_partkey/50000)::bool, false) AND
isfinite(l_shipdate) AND
l_partkey IS DISTINCT FROM 50040 AND
row(l_partkey, 2, 3) > row(2000, 2, 3);
count
---------------------------------------------------------------------
137
(1 row)
-- constant expression in the WHERE clause with a column in the target list
SELECT l_linenumber FROM lineitem
WHERE
1!=0
ORDER BY
l_linenumber
LIMIT 1;
l_linenumber
---------------------------------------------------------------------
1
(1 row)
-- constant expression in the WHERE clause with expressions and a column the target list
SELECT count(*) * l_discount as total_discount, count(*), sum(l_tax), l_discount FROM lineitem
WHERE
1!=0
GROUP BY
l_discount
ORDER BY
total_discount DESC, sum(l_tax) DESC;
total_discount | count | sum | l_discount
---------------------------------------------------------------------
104.80 | 1048 | 41.08 | 0.10
98.55 | 1095 | 44.15 | 0.09
90.64 | 1133 | 45.94 | 0.08
71.05 | 1015 | 41.19 | 0.07
69.42 | 1157 | 45.75 | 0.06
53.60 | 1072 | 42.82 | 0.05
43.64 | 1091 | 44.40 | 0.04
32.55 | 1085 | 43.30 | 0.03
22.22 | 1111 | 45.07 | 0.02
11.22 | 1122 | 44.54 | 0.01
0.00 | 1071 | 44.00 | 0.00
(11 rows)
-- distinct expressions in the WHERE clause with a column in the target list
SELECT l_linenumber FROM lineitem
WHERE
l_linenumber IS DISTINCT FROM 1 AND
l_orderkey IS DISTINCT FROM 8997
ORDER BY
l_linenumber
LIMIT 1;
l_linenumber
---------------------------------------------------------------------
2
(1 row)
-- distinct expressions in the WHERE clause with expressions and a column the target list
SELECT max(l_linenumber), min(l_discount), l_receiptdate FROM lineitem
WHERE
l_linenumber IS DISTINCT FROM 1 AND
l_orderkey IS DISTINCT FROM 8997
GROUP BY
l_receiptdate
ORDER BY
l_receiptdate
LIMIT 1;
max | min | l_receiptdate
---------------------------------------------------------------------
3 | 0.07 | 01-09-1992
(1 row)
-- Check that we can handle implicit and explicit join clause definitions.
SELECT count(*) FROM lineitem, orders
WHERE l_orderkey = o_orderkey AND l_quantity < 5;
count
---------------------------------------------------------------------
951
(1 row)
SELECT count(*) FROM lineitem
JOIN orders ON l_orderkey = o_orderkey AND l_quantity < 5;
count
---------------------------------------------------------------------
951
(1 row)
SELECT count(*) FROM lineitem JOIN orders ON l_orderkey = o_orderkey
WHERE l_quantity < 5;
count
---------------------------------------------------------------------
951
(1 row)
-- Check that we make sure local joins are between columns only.
SELECT count(*) FROM lineitem, orders WHERE l_orderkey + 1 = o_orderkey;
ERROR: complex joins are only supported when all distributed tables are joined on their distribution columns with equal operator
-- Check that we can issue limit/offset queries
-- the subquery is recursively planned since it contains OFFSET, which is not pushdownable
SELECT * FROM (SELECT o_custkey FROM orders GROUP BY o_custkey ORDER BY o_custkey OFFSET 20) sq ORDER BY 1 LIMIT 5;
ERROR: cannot perform distributed planning on this query
DETAIL: Subqueries with offset are not supported yet
-- the subquery is recursively planned since it contains OFFSET, which is not pushdownable
SELECT * FROM (SELECT o_orderkey FROM orders ORDER BY o_orderkey OFFSET 20) sq ORDER BY 1 LIMIT 5;
ERROR: cannot perform distributed planning on this query
DETAIL: Subqueries with offset are not supported yet
-- Simple LIMIT/OFFSET with ORDER BY
SELECT o_orderkey FROM orders ORDER BY o_orderkey LIMIT 10 OFFSET 20;
o_orderkey
---------------------------------------------------------------------
69
70
71
96
97
98
99
100
101
102
(10 rows)
-- LIMIT/OFFSET with a subquery
SELECT
customer_keys.o_custkey,
SUM(order_count) AS total_order_count
FROM
(SELECT o_custkey, o_orderstatus, COUNT(*) AS order_count
FROM orders GROUP BY o_custkey, o_orderstatus ) customer_keys
GROUP BY
customer_keys.o_custkey
ORDER BY
customer_keys.o_custkey DESC
LIMIT 10 OFFSET 20;
o_custkey | total_order_count
---------------------------------------------------------------------
1466 | 1
1465 | 2
1463 | 4
1462 | 10
1460 | 1
1459 | 6
1457 | 1
1456 | 3
1454 | 2
1453 | 5
(10 rows)
SELECT
customer_keys.o_custkey,
SUM(order_count) AS total_order_count
FROM
(SELECT o_custkey, o_orderstatus, COUNT(*) over (partition by o_orderstatus) AS order_count
FROM orders GROUP BY o_custkey, o_orderstatus ) customer_keys
GROUP BY
customer_keys.o_custkey
ORDER BY
customer_keys.o_custkey DESC
LIMIT 10 OFFSET 20;
o_custkey | total_order_count
---------------------------------------------------------------------
1466 | 759
1465 | 759
1463 | 1499
1462 | 1499
1460 | 759
1459 | 1499
1457 | 740
1456 | 830
1454 | 1499
1453 | 1499
(10 rows)
SELECT
customer_keys.o_custkey,
SUM(order_count1 + order_count) AS total_order_count
FROM
(SELECT o_custkey, o_orderstatus, count(*) order_count1, COUNT(*) over (partition by o_orderstatus) AS order_count
FROM orders GROUP BY o_custkey, o_orderstatus ) customer_keys
GROUP BY
customer_keys.o_custkey
ORDER BY
customer_keys.o_custkey DESC
LIMIT 10 OFFSET 20;
o_custkey | total_order_count
---------------------------------------------------------------------
1466 | 760
1465 | 761
1463 | 1503
1462 | 1509
1460 | 760
1459 | 1505
1457 | 741
1456 | 833
1454 | 1501
1453 | 1504
(10 rows)
SET client_min_messages TO DEBUG1;
-- Ensure that we push down LIMIT and OFFSET properly
-- No Group-By -> Push Down
CREATE TEMP TABLE temp_limit_test_1 AS
SELECT o_custkey FROM orders LIMIT 10 OFFSET 15;
DEBUG: push down of limit count: 25
-- GROUP BY without ORDER BY -> No push-down
CREATE TEMP TABLE temp_limit_test_2 AS
SELECT o_custkey FROM orders GROUP BY o_custkey LIMIT 10 OFFSET 15;
-- GROUP BY and ORDER BY non-aggregate -> push-down
CREATE TEMP TABLE temp_limit_test_3 AS
SELECT o_custkey FROM orders GROUP BY o_custkey ORDER BY o_custkey LIMIT 10 OFFSET 15;
DEBUG: push down of limit count: 25
-- GROUP BY and ORDER BY aggregate -> No push-down
CREATE TEMP TABLE temp_limit_test_4 AS
SELECT o_custkey, COUNT(*) AS ccnt FROM orders GROUP BY o_custkey ORDER BY ccnt DESC LIMIT 10 OFFSET 15;
-- OFFSET without LIMIT
SELECT o_custkey FROM orders ORDER BY o_custkey OFFSET 2980;
o_custkey
---------------------------------------------------------------------
1498
1498
1499
1499
1499
(5 rows)
-- LIMIT/OFFSET with Joins
SELECT
li.l_partkey,
o.o_custkey,
li.l_quantity
FROM
lineitem li JOIN orders o ON li.l_orderkey = o.o_orderkey
WHERE
li.l_quantity > 25
ORDER BY 1, 2, 3
LIMIT 10 OFFSET 20;
DEBUG: push down of limit count: 30
l_partkey | o_custkey | l_quantity
---------------------------------------------------------------------
655 | 58 | 50.00
669 | 319 | 34.00
699 | 1255 | 50.00
716 | 61 | 45.00
723 | 14 | 36.00
802 | 754 | 50.00
831 | 589 | 32.00
835 | 67 | 33.00
864 | 439 | 32.00
875 | 13 | 43.00
(10 rows)
RESET client_min_messages;
-- FILTERs
SELECT
l_orderkey,
sum(l_extendedprice),
sum(l_extendedprice) FILTER (WHERE l_shipmode = 'AIR'),
count(*),
count(*) FILTER (WHERE l_shipmode = 'AIR'),
max(l_extendedprice),
max(l_extendedprice) FILTER (WHERE l_quantity < 30)
FROM lineitem
GROUP BY l_orderkey
ORDER BY 2 DESC, 1 DESC
LIMIT 10;
l_orderkey | sum | sum | count | count | max | max
---------------------------------------------------------------------
12804 | 440012.71 | 45788.16 | 7 | 1 | 94398.00 | 45788.16
9863 | 412560.63 | 175647.63 | 7 | 3 | 85723.77 | 50769.14
2567 | 412076.77 | 59722.26 | 7 | 1 | 94894.00 | 9784.02
11142 | 410502.38 | 44965.95 | 7 | 1 | 83989.44 | 44965.95
12039 | 407048.94 | 76406.30 | 7 | 2 | 94471.02 | 19679.30
2306 | 405629.96 | 28032.60 | 7 | 1 | 92838.00 | 44384.50
5606 | 403595.91 | 36531.51 | 7 | 2 | 94890.18 | 30582.75
11296 | 399079.89 | | 6 | 0 | 102449.00 | 33122.93
11046 | 391163.26 | 31436.34 | 7 | 2 | 94506.24 | 47519.76
4421 | 387313.12 | | 7 | 0 | 67301.52 | 23783.40
(10 rows)
SELECT
l_orderkey,
sum(l_extendedprice),
sum(l_extendedprice) FILTER (WHERE l_shipmode = 'AIR'),
count(*),
count(*) FILTER (WHERE l_shipmode = 'AIR'),
max(l_extendedprice),
max(l_extendedprice) FILTER (WHERE l_quantity < 30)
FROM lineitem
GROUP BY l_orderkey
HAVING count(*) FILTER (WHERE l_shipmode = 'AIR') > 1
ORDER BY 2 DESC, 1 DESC
LIMIT 10;
l_orderkey | sum | sum | count | count | max | max
---------------------------------------------------------------------
9863 | 412560.63 | 175647.63 | 7 | 3 | 85723.77 | 50769.14
12039 | 407048.94 | 76406.30 | 7 | 2 | 94471.02 | 19679.30
5606 | 403595.91 | 36531.51 | 7 | 2 | 94890.18 | 30582.75
11046 | 391163.26 | 31436.34 | 7 | 2 | 94506.24 | 47519.76
14499 | 384140.30 | 67867.08 | 7 | 2 | 84335.36 | 46169.75
11623 | 380598.48 | 133709.82 | 7 | 2 | 93701.54 | 21487.65
10787 | 375688.09 | 99424.78 | 7 | 2 | 76732.67 | 50946.91
12902 | 358191.24 | 76891.00 | 7 | 2 | 82008.08 | 35602.08
3747 | 353701.23 | 68592.23 | 7 | 2 | 67181.10 | 46252.77
5158 | 349889.05 | 159753.19 | 7 | 3 | 78714.67 | 29729.20
(10 rows)

4
src/test/regress/expected/multi_data_types.out
View File

@ -174,7 +174,7 @@ INSERT INTO composite_type_partitioned_table VALUES (123, '(123, 456)'::other_co
Node: host=localhost port=xxxxx dbname=regression
-> Insert on public.composite_type_partitioned_table_530003 (actual rows=0 loops=1)
-> Result (actual rows=1 loops=1)
Output: xxxxxx
Output: 123, '(123,456)'::test_composite_type
(9 rows)
SELECT run_command_on_coordinator_and_workers($cf$
@ -218,7 +218,7 @@ INSERT INTO composite_type_partitioned_table VALUES (123, '(456, 678)'::other_co
Node: host=localhost port=xxxxx dbname=regression
-> Insert on public.composite_type_partitioned_table_530000 (actual rows=0 loops=1)
-> Result (actual rows=1 loops=1)
Output: xxxxxx
Output: 123, '(456,678)'::test_composite_type
(9 rows)
-- create and distribute a table on enum type column

75
src/test/regress/expected/multi_explain.out
View File

@ -345,14 +345,14 @@ EXPLAIN (COSTS FALSE, ANALYZE TRUE, TIMING FALSE, SUMMARY FALSE, VERBOSE TRUE)
SELECT l_quantity, count(*) count_quantity FROM lineitem
GROUP BY l_quantity ORDER BY count_quantity, l_quantity;
Sort (actual rows=50 loops=1)
Output: xxxxxx
Output: remote_scan.l_quantity, (COALESCE((pg_catalog.sum(remote_scan.count_quantity))::bigint, '0'::bigint))
Sort Key: (COALESCE((pg_catalog.sum(remote_scan.count_quantity))::bigint, '0'::bigint)), remote_scan.l_quantity
Sort Method: quicksort Memory: 27kB
-> HashAggregate (actual rows=50 loops=1)
Output: xxxxxx
Output: remote_scan.l_quantity, COALESCE((pg_catalog.sum(remote_scan.count_quantity))::bigint, '0'::bigint)
Group Key: remote_scan.l_quantity
-> Custom Scan (Citus Adaptive) (actual rows=100 loops=1)
Output: xxxxxx
Output: remote_scan.l_quantity, remote_scan.count_quantity
Task Count: 2
Tuple data received from nodes: 780 bytes
Tasks Shown: One of 2
@ -361,48 +361,48 @@ Sort (actual rows=50 loops=1)
Tuple data received from node: 390 bytes
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate (actual rows=50 loops=1)
Output: xxxxxx
Output: l_quantity, count(*)
Group Key: lineitem.l_quantity
-> Seq Scan on public.lineitem_290000 lineitem (actual rows=6000 loops=1)
Output: xxxxxx
Output: l_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment
-- Test query text output, with ANALYZE OFF
EXPLAIN (COSTS FALSE, ANALYZE FALSE, TIMING FALSE, SUMMARY FALSE, VERBOSE TRUE)
SELECT l_quantity, count(*) count_quantity FROM lineitem
GROUP BY l_quantity ORDER BY count_quantity, l_quantity;
Sort
Output: xxxxxx
Output: remote_scan.l_quantity, (COALESCE((pg_catalog.sum(remote_scan.count_quantity))::bigint, '0'::bigint))
Sort Key: (COALESCE((pg_catalog.sum(remote_scan.count_quantity))::bigint, '0'::bigint)), remote_scan.l_quantity
-> HashAggregate
Output: xxxxxx
Output: remote_scan.l_quantity, COALESCE((pg_catalog.sum(remote_scan.count_quantity))::bigint, '0'::bigint)
Group Key: remote_scan.l_quantity
-> Custom Scan (Citus Adaptive)
Output: xxxxxx
Output: remote_scan.l_quantity, remote_scan.count_quantity
Task Count: 2
Tasks Shown: One of 2
-> Task
Query: SELECT l_quantity, count(*) AS count_quantity FROM lineitem_290000 lineitem WHERE true GROUP BY l_quantity
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: xxxxxx
Output: l_quantity, count(*)
Group Key: lineitem.l_quantity
-> Seq Scan on public.lineitem_290000 lineitem
Output: xxxxxx
Output: l_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment
-- Test verbose
EXPLAIN (COSTS FALSE, VERBOSE TRUE)
SELECT sum(l_quantity) / avg(l_quantity) FROM lineitem;
Aggregate
Output: xxxxxx
Output: (sum(remote_scan."?column?") / (sum(remote_scan."?column?_1") / pg_catalog.sum(remote_scan."?column?_2")))
-> Custom Scan (Citus Adaptive)
Output: xxxxxx
Output: remote_scan."?column?", remote_scan."?column?_1", remote_scan."?column?_2"
Task Count: 2
Tasks Shown: One of 2
-> Task
Query: SELECT sum(l_quantity), sum(l_quantity), count(l_quantity) FROM lineitem_290000 lineitem WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: xxxxxx
Output: sum(l_quantity), sum(l_quantity), count(l_quantity)
-> Seq Scan on public.lineitem_290000 lineitem
Output: xxxxxx
Output: l_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment
-- Test join
EXPLAIN (COSTS FALSE)
SELECT * FROM lineitem
@ -525,40 +525,40 @@ EXPLAIN (COSTS FALSE, VERBOSE TRUE)
SELECT sum(l_quantity) / avg(l_quantity) FROM lineitem
HAVING sum(l_quantity) > 100;
Aggregate
Output: xxxxxx
Output: (sum(remote_scan."?column?") / (sum(remote_scan."?column?_1") / pg_catalog.sum(remote_scan."?column?_2")))
Filter: (sum(remote_scan.worker_column_4) > '100'::numeric)
-> Custom Scan (Citus Adaptive)
Output: xxxxxx
Output: remote_scan."?column?", remote_scan."?column?_1", remote_scan."?column?_2", remote_scan.worker_column_4
Task Count: 2
Tasks Shown: One of 2
-> Task
Query: SELECT sum(l_quantity), sum(l_quantity), count(l_quantity), sum(l_quantity) AS worker_column_4 FROM lineitem_290000 lineitem WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> Aggregate
Output: xxxxxx
Output: sum(l_quantity), sum(l_quantity), count(l_quantity), sum(l_quantity)
-> Seq Scan on public.lineitem_290000 lineitem
Output: xxxxxx
Output: l_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment
-- Test having without aggregate
EXPLAIN (COSTS FALSE, VERBOSE TRUE)
SELECT l_quantity FROM lineitem
GROUP BY l_quantity
HAVING l_quantity > (100 * random());
HashAggregate
Output: xxxxxx
Output: remote_scan.l_quantity
Group Key: remote_scan.l_quantity
Filter: ((remote_scan.worker_column_2)::double precision > ('100'::double precision * random()))
-> Custom Scan (Citus Adaptive)
Output: xxxxxx
Output: remote_scan.l_quantity, remote_scan.worker_column_2
Task Count: 2
Tasks Shown: One of 2
-> Task
Query: SELECT l_quantity, l_quantity AS worker_column_2 FROM lineitem_290000 lineitem WHERE true GROUP BY l_quantity
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: xxxxxx
Output: l_quantity, l_quantity
Group Key: lineitem.l_quantity
-> Seq Scan on public.lineitem_290000 lineitem
Output: xxxxxx
Output: l_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment
-- Subquery pushdown tests with explain
EXPLAIN (COSTS OFF)
SELECT
@ -1385,38 +1385,36 @@ Custom Scan (Citus INSERT ... SELECT)
-> Function Scan on generate_series s
-> Function Scan on generate_series s_1
-- explain with recursive planning
-- prevent PG 11 - PG 12 outputs to diverge
SET citus.enable_cte_inlining TO false;
EXPLAIN (COSTS OFF, VERBOSE true)
WITH keys AS (
WITH keys AS MATERIALIZED (
SELECT DISTINCT l_orderkey FROM lineitem_hash_part
),
series AS (
series AS MATERIALIZED (
SELECT s FROM generate_series(1,10) s
)
SELECT l_orderkey FROM series JOIN keys ON (s = l_orderkey)
ORDER BY s;
Custom Scan (Citus Adaptive)
Output: xxxxxx
Output: remote_scan.l_orderkey
-> Distributed Subplan XXX_1
-> HashAggregate
Output: xxxxxx
Output: remote_scan.l_orderkey
Group Key: remote_scan.l_orderkey
-> Custom Scan (Citus Adaptive)
Output: xxxxxx
Output: remote_scan.l_orderkey
Task Count: 4
Tasks Shown: One of 4
-> Task
Query: SELECT DISTINCT l_orderkey FROM lineitem_hash_part_360041 lineitem_hash_part WHERE true
Node: host=localhost port=xxxxx dbname=regression
-> HashAggregate
Output: xxxxxx
Output: l_orderkey
Group Key: lineitem_hash_part.l_orderkey
-> Seq Scan on public.lineitem_hash_part_360041 lineitem_hash_part
Output: xxxxxx
Output: l_orderkey, l_partkey, l_suppkey, l_linenumber, l_quantity, l_extendedprice, l_discount, l_tax, l_returnflag, l_linestatus, l_shipdate, l_commitdate, l_receiptdate, l_shipinstruct, l_shipmode, l_comment
-> Distributed Subplan XXX_2
-> Function Scan on pg_catalog.generate_series s
Output: xxxxxx
Output: s
Function Call: generate_series(1, 10)
Task Count: 1
Tasks Shown: All
@ -1424,21 +1422,20 @@ Custom Scan (Citus Adaptive)
Query: SELECT keys.l_orderkey FROM ((SELECT intermediate_result.s FROM read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format) intermediate_result(s integer)) series JOIN (SELECT intermediate_result.l_orderkey FROM read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format) intermediate_result(l_orderkey bigint)) keys ON ((series.s OPERATOR(pg_catalog.=) keys.l_orderkey))) ORDER BY series.s
Node: host=localhost port=xxxxx dbname=regression
-> Merge Join
Output: xxxxxx
Output: intermediate_result_1.l_orderkey, intermediate_result.s
Merge Cond: (intermediate_result.s = intermediate_result_1.l_orderkey)
-> Sort
Output: xxxxxx
Output: intermediate_result.s
Sort Key: intermediate_result.s
-> Function Scan on pg_catalog.read_intermediate_result intermediate_result
Output: xxxxxx
Output: intermediate_result.s
Function Call: read_intermediate_result('XXX_2'::text, 'binary'::citus_copy_format)
-> Sort
Output: xxxxxx
Output: intermediate_result_1.l_orderkey
Sort Key: intermediate_result_1.l_orderkey
-> Function Scan on pg_catalog.read_intermediate_result intermediate_result_1
Output: xxxxxx
Output: intermediate_result_1.l_orderkey
Function Call: read_intermediate_result('XXX_1'::text, 'binary'::citus_copy_format)
SET citus.enable_cte_inlining TO true;
SELECT true AS valid FROM explain_json($$
WITH result AS (
SELECT l_quantity, count(*) count_quantity FROM lineitem
@ -1757,7 +1754,7 @@ SELECT * FROM worker_save_query_explain_analyze('SELECT * FROM explain_analyze_t
4
(4 rows)
SELECT explain_analyze_output ~ 'Output: xxxxxx
SELECT explain_analyze_output ~ 'Output: a, b' FROM worker_last_saved_explain_analyze();
?column?
---------------------------------------------------------------------
t

905
src/test/regress/expected/multi_extension_0.out
View File

@ -1,905 +0,0 @@
--
-- MULTI_EXTENSION
--
-- Tests around extension creation / upgrades
--
-- It'd be nice to script generation of this file, but alas, that's
-- not done yet.
-- differentiate the output file for pg11 and versions above, with regards to objects
-- created per citus version depending on the postgres version. Upgrade tests verify the
-- objects are added in citus_finish_pg_upgrade()
SHOW server_version \gset
SELECT substring(:'server_version', '\d+')::int > 11 AS version_above_eleven;
version_above_eleven
---------------------------------------------------------------------
f
(1 row)
SET citus.next_shard_id TO 580000;
SELECT $definition$
CREATE OR REPLACE FUNCTION test.maintenance_worker()
RETURNS pg_stat_activity
LANGUAGE plpgsql
AS $$
DECLARE
activity record;
BEGIN
DO 'BEGIN END'; -- Force maintenance daemon to start
-- we don't want to wait forever; loop will exit after 20 seconds
FOR i IN 1 .. 200 LOOP
PERFORM pg_stat_clear_snapshot();
SELECT * INTO activity FROM pg_stat_activity
WHERE application_name = 'Citus Maintenance Daemon' AND datname = current_database();
IF activity.pid IS NOT NULL THEN
RETURN activity;
ELSE
PERFORM pg_sleep(0.1);
END IF ;
END LOOP;
-- fail if we reach the end of this loop
raise 'Waited too long for maintenance daemon to start';
END;
$$;
$definition$ create_function_test_maintenance_worker
\gset
CREATE TABLE prev_objects(description text);
CREATE TABLE extension_diff(previous_object text COLLATE "C",
current_object text COLLATE "C");
CREATE FUNCTION print_extension_changes()
RETURNS TABLE(previous_object text, current_object text)
AS $func$
BEGIN
TRUNCATE TABLE extension_diff;
CREATE TABLE current_objects AS
SELECT pg_catalog.pg_describe_object(classid, objid, 0) AS description
FROM pg_catalog.pg_depend, pg_catalog.pg_extension e
WHERE refclassid = 'pg_catalog.pg_extension'::pg_catalog.regclass
AND refobjid = e.oid
AND deptype = 'e'
AND e.extname='citus';
INSERT INTO extension_diff
SELECT p.description previous_object, c.description current_object
FROM current_objects c FULL JOIN prev_objects p
ON p.description = c.description
WHERE p.description is null OR c.description is null;
DROP TABLE prev_objects;
ALTER TABLE current_objects RENAME TO prev_objects;
RETURN QUERY SELECT * FROM extension_diff ORDER BY 1, 2;
END
$func$ LANGUAGE plpgsql;
CREATE SCHEMA test;
:create_function_test_maintenance_worker
-- check maintenance daemon is started
SELECT datname, current_database(),
usename, (SELECT extowner::regrole::text FROM pg_extension WHERE extname = 'citus')
FROM test.maintenance_worker();
datname | current_database | usename | extowner
---------------------------------------------------------------------
regression | regression | postgres | postgres
(1 row)
-- ensure no unexpected objects were created outside pg_catalog
SELECT pgio.type, pgio.identity
FROM pg_depend AS pgd,
pg_extension AS pge,
LATERAL pg_identify_object(pgd.classid, pgd.objid, pgd.objsubid) AS pgio
WHERE pgd.refclassid = 'pg_extension'::regclass AND
pgd.refobjid = pge.oid AND
pge.extname = 'citus' AND
pgio.schema NOT IN ('pg_catalog', 'citus', 'citus_internal', 'test', 'columnar')
ORDER BY 1, 2;
type | identity
---------------------------------------------------------------------
view | public.citus_tables
(1 row)
-- DROP EXTENSION pre-created by the regression suite
DROP EXTENSION citus;
\c
-- these tests switch between citus versions and call ddl's that require pg_dist_object to be created
SET citus.enable_object_propagation TO 'false';
SET citus.enable_version_checks TO 'false';
CREATE EXTENSION citus VERSION '8.0-1';
ALTER EXTENSION citus UPDATE TO '8.0-2';
ALTER EXTENSION citus UPDATE TO '8.0-3';
ALTER EXTENSION citus UPDATE TO '8.0-4';
ALTER EXTENSION citus UPDATE TO '8.0-5';
ALTER EXTENSION citus UPDATE TO '8.0-6';
ALTER EXTENSION citus UPDATE TO '8.0-7';
ALTER EXTENSION citus UPDATE TO '8.0-8';
ALTER EXTENSION citus UPDATE TO '8.0-9';
ALTER EXTENSION citus UPDATE TO '8.0-10';
ALTER EXTENSION citus UPDATE TO '8.0-11';
ALTER EXTENSION citus UPDATE TO '8.0-12';
ALTER EXTENSION citus UPDATE TO '8.0-13';
ALTER EXTENSION citus UPDATE TO '8.1-1';
ALTER EXTENSION citus UPDATE TO '8.2-1';
ALTER EXTENSION citus UPDATE TO '8.2-2';
ALTER EXTENSION citus UPDATE TO '8.2-3';
ALTER EXTENSION citus UPDATE TO '8.2-4';
ALTER EXTENSION citus UPDATE TO '8.3-1';
ALTER EXTENSION citus UPDATE TO '9.0-1';
ALTER EXTENSION citus UPDATE TO '9.0-2';
ALTER EXTENSION citus UPDATE TO '9.1-1';
ALTER EXTENSION citus UPDATE TO '9.2-1';
ALTER EXTENSION citus UPDATE TO '9.2-2';
-- Snapshot of state at 9.2-2
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
| event trigger citus_cascade_to_partition
| function alter_role_if_exists(text,text)
| function any_value(anyelement)
| function any_value_agg(anyelement,anyelement)
| function array_cat_agg(anyarray)
| function assign_distributed_transaction_id(integer,bigint,timestamp with time zone)
| function authinfo_valid(text)
| function broadcast_intermediate_result(text,text)
| function check_distributed_deadlocks()
| function citus_add_rebalance_strategy(name,regproc,regproc,regproc,real,real)
| function citus_blocking_pids(integer)
| function citus_create_restore_point(text)
| function citus_dist_stat_activity()
| function citus_drop_trigger()
| function citus_executor_name(integer)
| function citus_extradata_container(internal)
| function citus_finish_pg_upgrade()
| function citus_internal.find_groupid_for_node(text,integer)
| function citus_internal.pg_dist_node_trigger_func()
| function citus_internal.pg_dist_rebalance_strategy_enterprise_check()
| function citus_internal.pg_dist_rebalance_strategy_trigger_func()
| function citus_internal.pg_dist_shard_placement_trigger_func()
| function citus_internal.refresh_isolation_tester_prepared_statement()
| function citus_internal.replace_isolation_tester_func()
| function citus_internal.restore_isolation_tester_func()
| function citus_isolation_test_session_is_blocked(integer,integer[])
| function citus_json_concatenate(json,json)
| function citus_json_concatenate_final(json)
| function citus_jsonb_concatenate(jsonb,jsonb)
| function citus_jsonb_concatenate_final(jsonb)
| function citus_node_capacity_1(integer)
| function citus_prepare_pg_upgrade()
| function citus_query_stats()
| function citus_relation_size(regclass)
| function citus_server_id()
| function citus_set_default_rebalance_strategy(text)
| function citus_shard_allowed_on_node_true(bigint,integer)
| function citus_shard_cost_1(bigint)
| function citus_shard_cost_by_disk_size(bigint)
| function citus_stat_statements()
| function citus_stat_statements_reset()
| function citus_table_is_visible(oid)
| function citus_table_size(regclass)
| function citus_text_send_as_jsonb(text)
| function citus_total_relation_size(regclass)
| function citus_truncate_trigger()
| function citus_validate_rebalance_strategy_functions(regproc,regproc,regproc)
| function citus_version()
| function citus_worker_stat_activity()
| function column_name_to_column(regclass,text)
| function column_to_column_name(regclass,text)
| function coord_combine_agg(oid,cstring,anyelement)
| function coord_combine_agg_ffunc(internal,oid,cstring,anyelement)
| function coord_combine_agg_sfunc(internal,oid,cstring,anyelement)
| function create_distributed_function(regprocedure,text,text)
| function create_distributed_table(regclass,text,citus.distribution_type,text)
| function create_intermediate_result(text,text)
| function create_reference_table(regclass)
| function distributed_tables_colocated(regclass,regclass)
| function dump_global_wait_edges()
| function dump_local_wait_edges()
| function fetch_intermediate_results(text[],text,integer)
| function get_all_active_transactions()
| function get_colocated_shard_array(bigint)
| function get_colocated_table_array(regclass)
| function get_current_transaction_id()
| function get_global_active_transactions()
| function get_rebalance_progress()
| function get_rebalance_table_shards_plan(regclass,real,integer,bigint[],boolean,name)
| function get_shard_id_for_distribution_column(regclass,"any")
| function isolate_tenant_to_new_shard(regclass,"any",text)
| function json_cat_agg(json)
| function jsonb_cat_agg(jsonb)
| function lock_relation_if_exists(text,text)
| function lock_shard_metadata(integer,bigint[])
| function lock_shard_resources(integer,bigint[])
| function mark_tables_colocated(regclass,regclass[])
| function master_activate_node(text,integer)
| function master_add_inactive_node(text,integer,integer,noderole,name)
| function master_add_node(text,integer,integer,noderole,name)
| function master_add_secondary_node(text,integer,text,integer,name)
| function master_append_table_to_shard(bigint,text,text,integer)
| function master_apply_delete_command(text)
| function master_conninfo_cache_invalidate()
| function master_copy_shard_placement(bigint,text,integer,text,integer,boolean,citus.shard_transfer_mode)
| function master_create_distributed_table(regclass,text,citus.distribution_type)
| function master_create_empty_shard(text)
| function master_create_worker_shards(text,integer,integer)
| function master_disable_node(text,integer)
| function master_dist_local_group_cache_invalidate()
| function master_dist_node_cache_invalidate()
| function master_dist_object_cache_invalidate()
| function master_dist_partition_cache_invalidate()
| function master_dist_placement_cache_invalidate()
| function master_dist_shard_cache_invalidate()
| function master_drain_node(text,integer,citus.shard_transfer_mode,name)
| function master_drop_all_shards(regclass,text,text)
| function master_drop_sequences(text[])
| function master_get_active_worker_nodes()
| function master_get_new_placementid()
| function master_get_new_shardid()
| function master_get_table_ddl_events(text)
| function master_get_table_metadata(text)
| function master_modify_multiple_shards(text)
| function master_move_shard_placement(bigint,text,integer,text,integer,citus.shard_transfer_mode)
| function master_remove_distributed_table_metadata_from_workers(regclass,text,text)
| function master_remove_node(text,integer)
| function master_remove_partition_metadata(regclass,text,text)
| function master_run_on_worker(text[],integer[],text[],boolean)
| function master_set_node_property(text,integer,text,boolean)
| function master_unmark_object_distributed(oid,oid,integer)
| function master_update_node(integer,text,integer,boolean,integer)
| function master_update_shard_statistics(bigint)
| function master_update_table_statistics(regclass)
| function poolinfo_valid(text)
| function read_intermediate_result(text,citus_copy_format)
| function read_intermediate_results(text[],citus_copy_format)
| function rebalance_table_shards(regclass,real,integer,bigint[],citus.shard_transfer_mode,boolean,name)
| function recover_prepared_transactions()
| function relation_is_a_known_shard(regclass)
| function replicate_table_shards(regclass,integer,integer,bigint[],citus.shard_transfer_mode)
| function role_exists(name)
| function run_command_on_colocated_placements(regclass,regclass,text,boolean)
| function run_command_on_placements(regclass,text,boolean)
| function run_command_on_shards(regclass,text,boolean)
| function run_command_on_workers(text,boolean)
| function shard_name(regclass,bigint)
| function start_metadata_sync_to_node(text,integer)
| function stop_metadata_sync_to_node(text,integer)
| function task_tracker_assign_task(bigint,integer,text)
| function task_tracker_cleanup_job(bigint)
| function task_tracker_conninfo_cache_invalidate()
| function task_tracker_task_status(bigint,integer)
| function upgrade_to_reference_table(regclass)
| function worker_append_table_to_shard(text,text,text,integer)
| function worker_apply_inter_shard_ddl_command(bigint,text,bigint,text,text)
| function worker_apply_sequence_command(text)
| function worker_apply_sequence_command(text,regtype)
| function worker_apply_shard_ddl_command(bigint,text)
| function worker_apply_shard_ddl_command(bigint,text,text)
| function worker_cleanup_job_schema_cache()
| function worker_create_or_replace_object(text)
| function worker_create_schema(bigint,text)
| function worker_create_truncate_trigger(regclass)
| function worker_drop_distributed_table(text)
| function worker_execute_sql_task(bigint,integer,text,boolean)
| function worker_fetch_foreign_file(text,text,bigint,text[],integer[])
| function worker_fetch_partition_file(bigint,integer,integer,integer,text,integer)
| function worker_hash("any")
| function worker_hash_partition_table(bigint,integer,text,text,oid,anyarray)
| function worker_merge_files_and_run_query(bigint,integer,text,text)
| function worker_merge_files_into_table(bigint,integer,text[],text[])
| function worker_partial_agg(oid,anyelement)
| function worker_partial_agg_ffunc(internal)
| function worker_partial_agg_sfunc(internal,oid,anyelement)
| function worker_partition_query_result(text,text,integer,citus.distribution_type,text[],text[],boolean)
| function worker_range_partition_table(bigint,integer,text,text,oid,anyarray)
| function worker_repartition_cleanup(bigint)
| schema citus
| schema citus_internal
| sequence pg_dist_colocationid_seq
| sequence pg_dist_groupid_seq
| sequence pg_dist_node_nodeid_seq
| sequence pg_dist_placement_placementid_seq
| sequence pg_dist_shardid_seq
| table citus.pg_dist_object
| table pg_dist_authinfo
| table pg_dist_colocation
| table pg_dist_local_group
| table pg_dist_node
| table pg_dist_node_metadata
| table pg_dist_partition
| table pg_dist_placement
| table pg_dist_poolinfo
| table pg_dist_rebalance_strategy
| table pg_dist_shard
| table pg_dist_transaction
| type citus.distribution_type
| type citus.shard_transfer_mode
| type citus_copy_format
| type noderole
| view citus_dist_stat_activity
| view citus_lock_waits
| view citus_shard_indexes_on_worker
| view citus_shards_on_worker
| view citus_stat_statements
| view citus_worker_stat_activity
| view pg_dist_shard_placement
(188 rows)
-- Test downgrade to 9.2-2 from 9.2-4
ALTER EXTENSION citus UPDATE TO '9.2-4';
ALTER EXTENSION citus UPDATE TO '9.2-2';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
/*
* As we mistakenly bumped schema version to 9.3-1 in a bad release, we support
* updating citus schema from 9.3-1 to 9.2-4, but we do not support updates to 9.3-1.
*
* Hence the query below should fail.
*/
ALTER EXTENSION citus UPDATE TO '9.3-1';
ERROR: extension "citus" has no update path from version "9.2-2" to version "9.3-1"
ALTER EXTENSION citus UPDATE TO '9.2-4';
-- Snapshot of state at 9.2-4
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Test downgrade to 9.2-4 from 9.3-2
ALTER EXTENSION citus UPDATE TO '9.3-2';
ALTER EXTENSION citus UPDATE TO '9.2-4';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Snapshot of state at 9.3-2
ALTER EXTENSION citus UPDATE TO '9.3-2';
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
| function citus_remote_connection_stats()
| function replicate_reference_tables()
| function truncate_local_data_after_distributing_table(regclass)
| function update_distributed_table_colocation(regclass,text)
| function worker_create_or_alter_role(text,text,text)
(5 rows)
-- Test downgrade to 9.3-2 from 9.4-1
ALTER EXTENSION citus UPDATE TO '9.4-1';
ALTER EXTENSION citus UPDATE TO '9.3-2';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Snapshot of state at 9.4-1
ALTER EXTENSION citus UPDATE TO '9.4-1';
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
| function worker_last_saved_explain_analyze()
| function worker_save_query_explain_analyze(text,jsonb)
(2 rows)
-- Test downgrade to 9.4-1 from 9.5-1
ALTER EXTENSION citus UPDATE TO '9.5-1';
BEGIN;
SELECT master_add_node('localhost', :master_port, groupId=>0);
master_add_node
---------------------------------------------------------------------
1
(1 row)
CREATE TABLE citus_local_table (a int);
SELECT create_citus_local_table('citus_local_table');
NOTICE: create_citus_local_table is deprecated in favour of citus_add_local_table_to_metadata
create_citus_local_table
---------------------------------------------------------------------
(1 row)
-- downgrade from 9.5-1 to 9.4-1 should fail as we have a citus local table
ALTER EXTENSION citus UPDATE TO '9.4-1';
ERROR: citus local tables are introduced in Citus 9.5
HINT: To downgrade Citus to an older version, you should first convert each citus local table to a postgres table by executing SELECT undistribute_table("%s")
CONTEXT: PL/pgSQL function inline_code_block line 11 at RAISE
ROLLBACK;
-- now we can downgrade as there is no citus local table
ALTER EXTENSION citus UPDATE TO '9.4-1';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Snapshot of state at 9.5-1
ALTER EXTENSION citus UPDATE TO '9.5-1';
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
function master_drop_sequences(text[]) |
function task_tracker_assign_task(bigint,integer,text) |
function task_tracker_cleanup_job(bigint) |
function task_tracker_conninfo_cache_invalidate() |
function task_tracker_task_status(bigint,integer) |
function worker_execute_sql_task(bigint,integer,text,boolean) |
function worker_merge_files_and_run_query(bigint,integer,text,text) |
| function create_citus_local_table(regclass)
| function undistribute_table(regclass)
| function worker_record_sequence_dependency(regclass,regclass,name)
(10 rows)
-- Test downgrade to 9.5-1 from 10.0-1
ALTER EXTENSION citus UPDATE TO '10.0-1';
ALTER EXTENSION citus UPDATE TO '9.5-1';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Snapshot of state at 10.0-1
ALTER EXTENSION citus UPDATE TO '10.0-1';
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
function citus_total_relation_size(regclass) |
function create_citus_local_table(regclass) |
function mark_tables_colocated(regclass,regclass[]) |
function master_conninfo_cache_invalidate() |
function master_create_distributed_table(regclass,text,citus.distribution_type) |
function master_create_worker_shards(text,integer,integer) |
function master_dist_local_group_cache_invalidate() |
function master_dist_node_cache_invalidate() |
function master_dist_object_cache_invalidate() |
function master_dist_partition_cache_invalidate() |
function master_dist_placement_cache_invalidate() |
function master_dist_shard_cache_invalidate() |
function master_drop_all_shards(regclass,text,text) |
function master_modify_multiple_shards(text) |
function undistribute_table(regclass) |
function upgrade_to_reference_table(regclass) |
| function alter_distributed_table(regclass,text,integer,text,boolean)
| function alter_old_partitions_set_access_method(regclass,timestamp with time zone,name)
| function alter_table_set_access_method(regclass,text)
| function citus_activate_node(text,integer)
| function citus_add_inactive_node(text,integer,integer,noderole,name)
| function citus_add_local_table_to_metadata(regclass,boolean)
| function citus_add_node(text,integer,integer,noderole,name)
| function citus_add_secondary_node(text,integer,text,integer,name)
| function citus_conninfo_cache_invalidate()
| function citus_copy_shard_placement(bigint,text,integer,text,integer,boolean,citus.shard_transfer_mode)
| function citus_disable_node(text,integer)
| function citus_dist_local_group_cache_invalidate()
| function citus_dist_node_cache_invalidate()
| function citus_dist_object_cache_invalidate()
| function citus_dist_partition_cache_invalidate()
| function citus_dist_placement_cache_invalidate()
| function citus_dist_shard_cache_invalidate()
| function citus_drain_node(text,integer,citus.shard_transfer_mode,name)
| function citus_drop_all_shards(regclass,text,text)
| function citus_internal.columnar_ensure_objects_exist()
| function citus_move_shard_placement(bigint,text,integer,text,integer,citus.shard_transfer_mode)
| function citus_remove_node(text,integer)
| function citus_set_coordinator_host(text,integer,noderole,name)
| function citus_set_node_property(text,integer,text,boolean)
| function citus_shard_sizes()
| function citus_total_relation_size(regclass,boolean)
| function citus_unmark_object_distributed(oid,oid,integer)
| function citus_update_node(integer,text,integer,boolean,integer)
| function citus_update_shard_statistics(bigint)
| function citus_update_table_statistics(regclass)
| function fix_pre_citus10_partitioned_table_constraint_names()
| function fix_pre_citus10_partitioned_table_constraint_names(regclass)
| function notify_constraint_dropped()
| function remove_local_tables_from_metadata()
| function time_partition_range(regclass)
| function undistribute_table(regclass,boolean)
| function worker_change_sequence_dependency(regclass,regclass,regclass)
| function worker_fix_pre_citus10_partitioned_table_constraint_names(regclass,bigint,text)
| schema columnar
| sequence columnar.storageid_seq
| table columnar.chunk
| table columnar.chunk_group
| table columnar.options
| table columnar.stripe
| view citus_shards
| view citus_tables
| view time_partitions
(63 rows)
-- Test downgrade to 10.0-1 from 10.0-2
ALTER EXTENSION citus UPDATE TO '10.0-2';
ALTER EXTENSION citus UPDATE TO '10.0-1';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Snapshot of state at 10.0-2
ALTER EXTENSION citus UPDATE TO '10.0-2';
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Test downgrade to 10.0-2 from 10.0-3
ALTER EXTENSION citus UPDATE TO '10.0-3';
ALTER EXTENSION citus UPDATE TO '10.0-2';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Snapshot of state at 10.0-3
ALTER EXTENSION citus UPDATE TO '10.0-3';
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
| function citus_get_active_worker_nodes()
(1 row)
-- Test downgrade to 10.0-3 from 10.1-1
ALTER EXTENSION citus UPDATE TO '10.1-1';
ALTER EXTENSION citus UPDATE TO '10.0-3';
-- Should be empty result since upgrade+downgrade should be a no-op
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
-- Snapshot of state at 10.1-1
ALTER EXTENSION citus UPDATE TO '10.1-1';
SELECT * FROM print_extension_changes();
previous_object | current_object
---------------------------------------------------------------------
(0 rows)
DROP TABLE prev_objects, extension_diff;
-- show running version
SHOW citus.version;
citus.version
---------------------------------------------------------------------
10.1devel
(1 row)
-- ensure no unexpected objects were created outside pg_catalog
SELECT pgio.type, pgio.identity
FROM pg_depend AS pgd,
pg_extension AS pge,
LATERAL pg_identify_object(pgd.classid, pgd.objid, pgd.objsubid) AS pgio
WHERE pgd.refclassid = 'pg_extension'::regclass AND
pgd.refobjid = pge.oid AND
pge.extname = 'citus' AND
pgio.schema NOT IN ('pg_catalog', 'citus', 'citus_internal', 'test', 'columnar')
ORDER BY 1, 2;
type | identity
---------------------------------------------------------------------
view | public.citus_tables
(1 row)
-- see incompatible version errors out
RESET citus.enable_version_checks;
DROP EXTENSION citus;
CREATE EXTENSION citus VERSION '8.0-1';
ERROR: specified version incompatible with loaded Citus library
DETAIL: Loaded library requires 10.1, but 8.0-1 was specified.
HINT: If a newer library is present, restart the database and try the command again.
-- Test non-distributed queries work even in version mismatch
SET citus.enable_version_checks TO 'false';
CREATE EXTENSION citus VERSION '8.1-1';
SET citus.enable_version_checks TO 'true';
-- Test CREATE TABLE
CREATE TABLE version_mismatch_table(column1 int);
-- Test COPY
\copy version_mismatch_table FROM STDIN;
-- Test INSERT
INSERT INTO version_mismatch_table(column1) VALUES(5);
-- Test SELECT
SELECT * FROM version_mismatch_table ORDER BY column1;
column1
---------------------------------------------------------------------
0
1
2
3
4
5
(6 rows)
-- Test SELECT from pg_catalog
SELECT d.datname as "Name",
pg_catalog.pg_get_userbyid(d.datdba) as "Owner",
pg_catalog.array_to_string(d.datacl, E'\n') AS "Access privileges"
FROM pg_catalog.pg_database d
ORDER BY 1;
Name | Owner | Access privileges
---------------------------------------------------------------------
postgres | postgres |
regression | postgres |
template0 | postgres | =c/postgres +
| | postgres=CTc/postgres
template1 | postgres | =c/postgres +
| | postgres=CTc/postgres
(4 rows)
-- We should not distribute table in version mistmatch
SELECT create_distributed_table('version_mismatch_table', 'column1');
ERROR: loaded Citus library version differs from installed extension version
DETAIL: Loaded library requires 10.1, but the installed extension version is 8.1-1.
HINT: Run ALTER EXTENSION citus UPDATE and try again.
-- This function will cause fail in next ALTER EXTENSION
CREATE OR REPLACE FUNCTION pg_catalog.relation_is_a_known_shard(regclass)
RETURNS void LANGUAGE plpgsql
AS $function$
BEGIN
END;
$function$;
ERROR: cannot change return type of existing function
HINT: Use DROP FUNCTION relation_is_a_known_shard(regclass) first.
SET citus.enable_version_checks TO 'false';
-- This will fail because of previous function declaration
ALTER EXTENSION citus UPDATE TO '8.1-1';
NOTICE: version "8.1-1" of extension "citus" is already installed
-- We can DROP problematic function and continue ALTER EXTENSION even when version checks are on
SET citus.enable_version_checks TO 'true';
DROP FUNCTION pg_catalog.relation_is_a_known_shard(regclass);
ERROR: cannot drop function relation_is_a_known_shard(regclass) because extension citus requires it
HINT: You can drop extension citus instead.
SET citus.enable_version_checks TO 'false';
ALTER EXTENSION citus UPDATE TO '8.1-1';
NOTICE: version "8.1-1" of extension "citus" is already installed
-- Test updating to the latest version without specifying the version number
ALTER EXTENSION citus UPDATE;
-- re-create in newest version
DROP EXTENSION citus;
\c
CREATE EXTENSION citus;
-- test cache invalidation in workers
\c - - - :worker_1_port
DROP EXTENSION citus;
SET citus.enable_version_checks TO 'false';
CREATE EXTENSION citus VERSION '8.0-1';
SET citus.enable_version_checks TO 'true';
-- during ALTER EXTENSION, we should invalidate the cache
ALTER EXTENSION citus UPDATE;
-- if cache is invalidated succesfull, this \d should work without any problem
\d
List of relations
Schema | Name | Type | Owner
---------------------------------------------------------------------
public | citus_tables | view | postgres
(1 row)
\c - - - :master_port
-- test https://github.com/citusdata/citus/issues/3409
CREATE USER testuser2 SUPERUSER;
NOTICE: not propagating CREATE ROLE/USER commands to worker nodes
HINT: Connect to worker nodes directly to manually create all necessary users and roles.
SET ROLE testuser2;
DROP EXTENSION Citus;
-- Loop until we see there's no maintenance daemon running
DO $$begin
for i in 0 .. 100 loop
if i = 100 then raise 'Waited too long'; end if;
PERFORM pg_stat_clear_snapshot();
perform * from pg_stat_activity where application_name = 'Citus Maintenance Daemon';
if not found then exit; end if;
perform pg_sleep(0.1);
end loop;
end$$;
SELECT datid, datname, usename FROM pg_stat_activity WHERE application_name = 'Citus Maintenance Daemon';
datid | datname | usename
---------------------------------------------------------------------
(0 rows)
CREATE EXTENSION Citus;
-- Loop until we there's a maintenance daemon running
DO $$begin
for i in 0 .. 100 loop
if i = 100 then raise 'Waited too long'; end if;
PERFORM pg_stat_clear_snapshot();
perform * from pg_stat_activity where application_name = 'Citus Maintenance Daemon';
if found then exit; end if;
perform pg_sleep(0.1);
end loop;
end$$;
SELECT datid, datname, usename FROM pg_stat_activity WHERE application_name = 'Citus Maintenance Daemon';
datid | datname | usename
---------------------------------------------------------------------
16384 | regression | testuser2
(1 row)
RESET ROLE;
-- check that maintenance daemon gets (re-)started for the right user
DROP EXTENSION citus;
CREATE USER testuser SUPERUSER;
SET ROLE testuser;
CREATE EXTENSION citus;
SELECT datname, current_database(),
usename, (SELECT extowner::regrole::text FROM pg_extension WHERE extname = 'citus')
FROM test.maintenance_worker();
datname | current_database | usename | extowner
---------------------------------------------------------------------
regression | regression | testuser | testuser
(1 row)
-- and recreate as the right owner
RESET ROLE;
DROP EXTENSION citus;
CREATE EXTENSION citus;
-- Check that maintenance daemon can also be started in another database
CREATE DATABASE another;
NOTICE: Citus partially supports CREATE DATABASE for distributed databases
DETAIL: Citus does not propagate CREATE DATABASE command to workers
HINT: You can manually create a database and its extensions on workers.
\c another
CREATE EXTENSION citus;
CREATE SCHEMA test;
:create_function_test_maintenance_worker
-- see that the daemon started
SELECT datname, current_database(),
usename, (SELECT extowner::regrole::text FROM pg_extension WHERE extname = 'citus')
FROM test.maintenance_worker();
datname | current_database | usename | extowner
---------------------------------------------------------------------
another | another | postgres | postgres
(1 row)
-- Test that database with active worker can be dropped.
\c regression
CREATE SCHEMA test_daemon;
-- we create a similar function on the regression database
-- note that this function checks for the existence of the daemon
-- when not found, returns true else tries for 5 times and
-- returns false
CREATE OR REPLACE FUNCTION test_daemon.maintenance_daemon_died(p_dbname text)
RETURNS boolean
LANGUAGE plpgsql
AS $$
DECLARE
activity record;
BEGIN
PERFORM pg_stat_clear_snapshot();
SELECT * INTO activity FROM pg_stat_activity
WHERE application_name = 'Citus Maintenance Daemon' AND datname = p_dbname;
IF activity.pid IS NULL THEN
RETURN true;
ELSE
RETURN false;
END IF;
END;
$$;
-- drop the database and see that the daemon is dead
DROP DATABASE another;
SELECT
*
FROM
test_daemon.maintenance_daemon_died('another');
maintenance_daemon_died
---------------------------------------------------------------------
t
(1 row)
-- we don't need the schema and the function anymore
DROP SCHEMA test_daemon CASCADE;
NOTICE: drop cascades to function test_daemon.maintenance_daemon_died(text)
-- verify citus does not crash while creating a table when run against an older worker
-- create_distributed_table piggybacks multiple commands into single one, if one worker
-- did not have the required UDF it should fail instead of crash.
-- create a test database, configure citus with single node
CREATE DATABASE another;
NOTICE: Citus partially supports CREATE DATABASE for distributed databases
DETAIL: Citus does not propagate CREATE DATABASE command to workers
HINT: You can manually create a database and its extensions on workers.
\c - - - :worker_1_port
CREATE DATABASE another;
NOTICE: Citus partially supports CREATE DATABASE for distributed databases
DETAIL: Citus does not propagate CREATE DATABASE command to workers
HINT: You can manually create a database and its extensions on workers.
\c - - - :master_port
\c another
CREATE EXTENSION citus;
SET citus.enable_object_propagation TO off; -- prevent distributed transactions during add node
SELECT FROM master_add_node('localhost', :worker_1_port);
WARNING: citus.enable_object_propagation is off, not creating distributed objects on worker
DETAIL: distributed objects are only kept in sync when citus.enable_object_propagation is set to on. Newly activated nodes will not get these objects created
--
(1 row)
\c - - - :worker_1_port
CREATE EXTENSION citus;
ALTER FUNCTION assign_distributed_transaction_id(initiator_node_identifier integer, transaction_number bigint, transaction_stamp timestamp with time zone)
RENAME TO dummy_assign_function;
\c - - - :master_port
SET citus.shard_replication_factor to 1;
-- create_distributed_table command should fail
CREATE TABLE t1(a int, b int);
SET client_min_messages TO ERROR;
DO $$
BEGIN
BEGIN
SELECT create_distributed_table('t1', 'a');
EXCEPTION WHEN OTHERS THEN
RAISE 'create distributed table failed';
END;
END;
$$;
ERROR: create distributed table failed
CONTEXT: PL/pgSQL function inline_code_block line 6 at RAISE
\c regression
\c - - - :master_port
DROP DATABASE another;
\c - - - :worker_1_port
DROP DATABASE another;
\c - - - :master_port
-- only the regression database should have a maintenance daemon
SELECT count(*) FROM pg_stat_activity WHERE application_name = 'Citus Maintenance Daemon';
count
---------------------------------------------------------------------
1
(1 row)
-- recreate the extension immediately after the maintenancae daemon errors
SELECT pg_cancel_backend(pid) FROM pg_stat_activity WHERE application_name = 'Citus Maintenance Daemon';
pg_cancel_backend
---------------------------------------------------------------------
t
(1 row)
DROP EXTENSION citus;
CREATE EXTENSION citus;
-- wait for maintenance daemon restart
SELECT datname, current_database(),
usename, (SELECT extowner::regrole::text FROM pg_extension WHERE extname = 'citus')
FROM test.maintenance_worker();
datname | current_database | usename | extowner
---------------------------------------------------------------------
regression | regression | postgres | postgres
(1 row)
-- confirm that there is only one maintenance daemon
SELECT count(*) FROM pg_stat_activity WHERE application_name = 'Citus Maintenance Daemon';
count
---------------------------------------------------------------------
1
(1 row)
-- kill the maintenance daemon
SELECT pg_cancel_backend(pid) FROM pg_stat_activity WHERE application_name = 'Citus Maintenance Daemon';
pg_cancel_backend
---------------------------------------------------------------------
t
(1 row)
-- reconnect
\c - - - :master_port
-- run something that goes through planner hook and therefore kicks of maintenance daemon
SELECT 1;
?column?
---------------------------------------------------------------------
1
(1 row)
-- wait for maintenance daemon restart
SELECT datname, current_database(),
usename, (SELECT extowner::regrole::text FROM pg_extension WHERE extname = 'citus')
FROM test.maintenance_worker();
datname | current_database | usename | extowner
---------------------------------------------------------------------
regression | regression | postgres | postgres
(1 row)
-- confirm that there is only one maintenance daemon
SELECT count(*) FROM pg_stat_activity WHERE application_name = 'Citus Maintenance Daemon';
count
---------------------------------------------------------------------
1
(1 row)
DROP TABLE version_mismatch_table;

Some files were not shown because too many files have changed in this diff Show More