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

Merge from master branch into feature/citusdb-to-citus

pull/337/head
Murat Tuncer 2016-02-17 14:49:01 +02:00
parent 55c44b48dd 39b333996b
commit 3528d7ce85
63 changed files with 1599 additions and 1427 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
.gitattributes vendored
View File

@ -20,3 +20,23 @@ src/test/regress/output/*.source -whitespace
# These files are maintained or generated elsewhere. We take them as is.
configure -whitespace
# all C files (implementation and header) use our style...
*.[ch] citus-style
# except these exceptions...
src/backend/distributed/utils/citus_outfuncs.c -citus-style
src/backend/distributed/utils/citus_read.c -citus-style
src/backend/distributed/utils/citus_readfuncs_94.c -citus-style
src/backend/distributed/utils/citus_readfuncs_95.c -citus-style
src/backend/distributed/utils/ruleutils_94.c -citus-style
src/backend/distributed/utils/ruleutils_95.c -citus-style
src/include/distributed/citus_nodes.h -citus-style
src/include/dumputils.h -citus-style
# all csql files use PostgreSQL style...
src/bin/csql/*.[ch] -citus-style
# except these exceptions
src/bin/csql/copy_options.c citus-style
src/bin/csql/stage.[ch] citus-style

7
Makefile
View File

@ -42,6 +42,13 @@ clean-csql:
install: install-csql
clean: clean-csql
# apply or check style
reindent:
cd ${citusdb_abs_top_srcdir} && citus_indent --quiet
check-style:
cd ${citusdb_abs_top_srcdir} && citus_indent --quiet --check
.PHONY: reindent check-style
# depend on install for now
check: all install
$(MAKE) -C src/test/regress check-full

14
src/backend/distributed/commands/transmit.c
View File

@ -271,17 +271,27 @@ ReceiveCopyData(StringInfo copyData)
switch (messageType)
{
case 'd': /* CopyData */
{
copyDone = false;
break;
}
case 'c': /* CopyDone */
{
copyDone = true;
break;
}
case 'f': /* CopyFail */
{
ereport(ERROR, (errcode(ERRCODE_QUERY_CANCELED),
errmsg("COPY data failed: %s", pq_getmsgstring(copyData))));
break;
}
case 'H': /* Flush */
case 'S': /* Sync */
{
/*
* Ignore Flush/Sync for the convenience of client libraries (such
* as libpq) that may send those without noticing that the command
@ -289,12 +299,16 @@ ReceiveCopyData(StringInfo copyData)
*/
copyDone = false;
break;
}
default:
{
ereport(ERROR, (errcode(ERRCODE_PROTOCOL_VIOLATION),
errmsg("unexpected message type 0x%02X during COPY data",
messageType)));
break;
}
}
return copyDone;
}

1
src/backend/distributed/executor/multi_executor.c
View File

@ -157,7 +157,6 @@ multi_ExecutorStart(QueryDesc *queryDesc, int eflags)
queryDesc->plannedstmt = masterSelectPlan;
eflags |= EXEC_FLAG_CITUS_MASTER_SELECT;
}
}
/* if the execution is not done for router executor, drop into standard executor */

17
src/backend/distributed/executor/multi_real_time_executor.c
View File

@ -287,8 +287,9 @@ ManageTaskExecution(Task *task, TaskExecution *taskExecution)
uint32 currentCount = taskExecution->connectPollCount;
if (currentCount >= maxCount)
{
ereport(WARNING, (errmsg("could not establish asynchronous connection "
"after %u ms", REMOTE_NODE_CONNECT_TIMEOUT)));
ereport(WARNING, (errmsg("could not establish asynchronous "
"connection after %u ms",
REMOTE_NODE_CONNECT_TIMEOUT)));
taskStatusArray[currentIndex] = EXEC_TASK_FAILED;
}
@ -342,7 +343,8 @@ ManageTaskExecution(Task *task, TaskExecution *taskExecution)
{
List *dataFetchTaskList = task->dependedTaskList;
int32 dataFetchTaskIndex = taskExecution->dataFetchTaskIndex;
Task *dataFetchTask = (Task *) list_nth(dataFetchTaskList, dataFetchTaskIndex);
Task *dataFetchTask = (Task *) list_nth(dataFetchTaskList,
dataFetchTaskIndex);
char *dataFetchQuery = dataFetchTask->queryString;
int32 connectionId = connectionIdArray[currentIndex];
@ -411,11 +413,13 @@ ManageTaskExecution(Task *task, TaskExecution *taskExecution)
StringInfo computeTaskQuery = makeStringInfo();
if (BinaryMasterCopyFormat)
{
appendStringInfo(computeTaskQuery, COPY_QUERY_TO_STDOUT_BINARY, queryString);
appendStringInfo(computeTaskQuery, COPY_QUERY_TO_STDOUT_BINARY,
queryString);
}
else
{
appendStringInfo(computeTaskQuery, COPY_QUERY_TO_STDOUT_TEXT, queryString);
appendStringInfo(computeTaskQuery, COPY_QUERY_TO_STDOUT_TEXT,
queryString);
}
querySent = MultiClientSendQuery(connectionId, computeTaskQuery->data);
@ -475,7 +479,8 @@ ManageTaskExecution(Task *task, TaskExecution *taskExecution)
else
{
ereport(WARNING, (errcode_for_file_access(),
errmsg("could not open file \"%s\": %m", filename)));
errmsg("could not open file \"%s\": %m",
filename)));
taskStatusArray[currentIndex] = EXEC_TASK_FAILED;
}

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

@ -80,6 +80,7 @@ RouterExecutorStart(QueryDesc *queryDesc, int eflags, Task *task)
queryDesc->estate = executorState;
#if (PG_VERSION_NUM < 90500)
/* make sure that upsertQuery is false for versions that UPSERT is not available */
Assert(task->upsertQuery == false);
#endif
@ -153,7 +154,7 @@ CommutativityRuleToLockMode(CmdType commandType, bool upsertQuery)
static void
AcquireExecutorShardLock(Task *task, LOCKMODE lockMode)
{
int64 shardId = task->shardId;
int64 shardId = task->anchorShardId;
LockShardResource(shardId, lockMode);
}
@ -219,9 +220,9 @@ RouterExecutorRun(QueryDesc *queryDesc, ScanDirection direction, long count, Tas
}
MemoryContextSwitchTo(oldcontext);
}
/*
* ExecuteDistributedModify is the main entry point for modifying distributed
* tables. A distributed modification is successful if any placement of the
@ -532,6 +533,7 @@ StoreQueryResult(PGconn *connection, TupleDesc tupleDescriptor,
return true;
}
/*
* RouterExecutorFinish cleans up after a distributed execution.
*/

25
src/backend/distributed/executor/multi_task_tracker_executor.c
View File

@ -41,7 +41,6 @@ typedef struct TaskMapKey
TaskType taskType;
uint64 jobId;
uint32 taskId;
} TaskMapKey;
@ -53,7 +52,6 @@ typedef struct TaskMapEntry
{
TaskMapKey key;
Task *task;
} TaskMapEntry;
@ -83,7 +81,8 @@ static TaskTracker * TrackerHashLookup(HTAB *trackerHash, const char *nodeName,
static TaskExecStatus ManageTaskExecution(TaskTracker *taskTracker,
TaskTracker *sourceTaskTracker,
Task *task, TaskExecution *taskExecution);
static TransmitExecStatus ManageTransmitExecution(TaskTracker *transmitTracker, Task *task,
static TransmitExecStatus ManageTransmitExecution(TaskTracker *transmitTracker,
Task *task,
TaskExecution *taskExecution);
static bool TaskExecutionsCompleted(List *taskList);
static StringInfo MapFetchTaskQueryString(Task *mapFetchTask, Task *mapTask);
@ -826,7 +825,8 @@ TrackerConnectPoll(TaskTracker *taskTracker)
uint32 nodePort = taskTracker->workerPort;
char *nodeDatabase = get_database_name(MyDatabaseId);
int32 connectionId = MultiClientConnectStart(nodeName, nodePort, nodeDatabase);
int32 connectionId = MultiClientConnectStart(nodeName, nodePort,
nodeDatabase);
if (connectionId != INVALID_CONNECTION_ID)
{
taskTracker->connectionId = connectionId;
@ -869,8 +869,9 @@ TrackerConnectPoll(TaskTracker *taskTracker)
uint32 currentCount = taskTracker->connectPollCount;
if (currentCount >= maxCount)
{
ereport(WARNING, (errmsg("could not establish asynchronous connection "
"after %u ms", REMOTE_NODE_CONNECT_TIMEOUT)));
ereport(WARNING, (errmsg("could not establish asynchronous "
"connection after %u ms",
REMOTE_NODE_CONNECT_TIMEOUT)));
taskTracker->trackerStatus = TRACKER_CONNECTION_FAILED;
@ -1212,7 +1213,8 @@ ManageTaskExecution(TaskTracker *taskTracker, TaskTracker *sourceTaskTracker,
default:
{
/* we fatal here to avoid leaking client-side resources */
ereport(FATAL, (errmsg("invalid execution status: %d", currentExecutionStatus)));
ereport(FATAL, (errmsg("invalid execution status: %d",
currentExecutionStatus)));
break;
}
}
@ -1327,7 +1329,8 @@ ManageTransmitExecution(TaskTracker *transmitTracker,
else
{
ereport(WARNING, (errcode_for_file_access(),
errmsg("could not open file \"%s\": %m", filename)));
errmsg("could not open file \"%s\": %m",
filename)));
nextTransmitStatus = EXEC_TRANSMIT_TRACKER_RETRY;
}
@ -1463,7 +1466,8 @@ ManageTransmitExecution(TaskTracker *transmitTracker,
default:
{
/* we fatal here to avoid leaking client-side resources */
ereport(FATAL, (errmsg("invalid transmit status: %d", currentTransmitStatus)));
ereport(FATAL, (errmsg("invalid transmit status: %d",
currentTransmitStatus)));
break;
}
}
@ -2833,7 +2837,8 @@ TrackerHashCleanupJob(HTAB *taskTrackerHash, Task *jobCleanupTask)
if (queryStatus == CLIENT_QUERY_DONE)
{
ereport(DEBUG4, (errmsg("completed cleanup query for job " UINT64_FORMAT
" on node \"%s:%u\"", jobId, nodeName, nodePort)));
" on node \"%s:%u\"", jobId, nodeName,
nodePort)));
/* clear connection for future cleanup queries */
taskTracker->connectionBusy = false;

9
src/backend/distributed/executor/multi_utility.c
View File

@ -905,8 +905,9 @@ ExecuteCommandOnWorkerShards(Oid relationId, const char *commandString,
}
else
{
ereport(DEBUG2, (errmsg("applied command on shard " UINT64_FORMAT " on "
"node %s:%d", shardId, workerName, workerPort)));
ereport(DEBUG2, (errmsg("applied command on shard " UINT64_FORMAT
" on node %s:%d", shardId, workerName,
workerPort)));
}
isFirstPlacement = false;
@ -988,6 +989,7 @@ AllFinalizedPlacementsAccessible(Oid relationId)
static void
RangeVarCallbackForDropIndex(const RangeVar *rel, Oid relOid, Oid oldRelOid, void *arg)
{
/* *INDENT-OFF* */
HeapTuple tuple;
struct DropRelationCallbackState *state;
char relkind;
@ -1022,10 +1024,8 @@ RangeVarCallbackForDropIndex(const RangeVar *rel, Oid relOid, Oid oldRelOid, voi
classform = (Form_pg_class) GETSTRUCT(tuple);
if (classform->relkind != relkind)
{
ereport(ERROR, (errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("\"%s\" is not an index", rel->relname)));
}
/* Allow DROP to either table owner or schema owner */
if (!pg_class_ownercheck(relOid, GetUserId()) &&
@ -1054,4 +1054,5 @@ RangeVarCallbackForDropIndex(const RangeVar *rel, Oid relOid, Oid oldRelOid, voi
if (OidIsValid(state->heapOid))
LockRelationOid(state->heapOid, heap_lockmode);
}
/* *INDENT-ON* */
}

15
src/backend/distributed/master/master_delete_protocol.c
View File

@ -167,7 +167,8 @@ master_apply_delete_command(PG_FUNCTION_ARGS)
shardPlacementList = ShardPlacementList(shardId);
foreach(shardPlacementCell, shardPlacementList)
{
ShardPlacement *shardPlacement = (ShardPlacement *) lfirst(shardPlacementCell);
ShardPlacement *shardPlacement =
(ShardPlacement *) lfirst(shardPlacementCell);
char *workerName = shardPlacement->nodeName;
uint32 workerPort = shardPlacement->nodePort;
bool dropSuccessful = false;
@ -176,14 +177,17 @@ master_apply_delete_command(PG_FUNCTION_ARGS)
char tableType = get_rel_relkind(relationId);
if (tableType == RELKIND_RELATION)
{
appendStringInfo(workerDropQuery, DROP_REGULAR_TABLE_COMMAND, quotedShardName);
appendStringInfo(workerDropQuery, DROP_REGULAR_TABLE_COMMAND,
quotedShardName);
}
else if (tableType == RELKIND_FOREIGN_TABLE)
{
appendStringInfo(workerDropQuery, DROP_FOREIGN_TABLE_COMMAND, quotedShardName);
appendStringInfo(workerDropQuery, DROP_FOREIGN_TABLE_COMMAND,
quotedShardName);
}
dropSuccessful = ExecuteRemoteCommand(workerName, workerPort, workerDropQuery);
dropSuccessful = ExecuteRemoteCommand(workerName, workerPort,
workerDropQuery);
if (dropSuccessful)
{
droppedPlacementList = lappend(droppedPlacementList, shardPlacement);
@ -227,7 +231,8 @@ master_apply_delete_command(PG_FUNCTION_ARGS)
if (QueryCancelPending)
{
ereport(WARNING, (errmsg("cancel requests are ignored during shard deletion")));
ereport(WARNING, (errmsg("cancel requests are ignored during shard "
"deletion")));
QueryCancelPending = false;
}

5
src/backend/distributed/master/master_node_protocol.c
View File

@ -709,11 +709,16 @@ hostname_client_addr(void)
#ifdef HAVE_IPV6
case AF_INET6:
#endif
{
break;
}
default:
{
ereport(ERROR, (errmsg("invalid address family in connection")));
break;
}
}
remoteHost = palloc0(remoteHostLen);

32
src/backend/distributed/master/master_stage_protocol.c
View File

@ -45,7 +45,8 @@ static bool WorkerCreateShard(char *nodeName, uint32 nodePort,
static bool WorkerShardStats(char *nodeName, uint32 nodePort, Oid relationId,
char *shardName, uint64 *shardLength,
text **shardMinValue, text **shardMaxValue);
static uint64 WorkerTableSize(char *nodeName, uint32 nodePort, char *tableName);
static uint64 WorkerTableSize(char *nodeName, uint32 nodePort, Oid relationId,
char *tableName);
static StringInfo WorkerPartitionValue(char *nodeName, uint32 nodePort, Oid relationId,
char *shardName, char *selectQuery);
@ -77,16 +78,15 @@ master_create_empty_shard(PG_FUNCTION_ARGS)
List *candidateNodeList = NIL;
text *nullMinValue = NULL;
text *nullMaxValue = NULL;
char tableType = 0;
char partitionMethod = 0;
char storageType = SHARD_STORAGE_TABLE;
Oid relationId = ResolveRelationId(relationNameText);
CheckDistributedTable(relationId);
tableType = get_rel_relkind(relationId);
if (tableType != RELKIND_RELATION)
if (CStoreTable(relationId))
{
ereport(ERROR, (errmsg("relation \"%s\" is not a regular table", relationName)));
storageType = SHARD_STORAGE_COLUMNAR;
}
partitionMethod = PartitionMethod(relationId);
@ -130,7 +130,7 @@ master_create_empty_shard(PG_FUNCTION_ARGS)
CreateShardPlacements(shardId, ddlEventList, candidateNodeList, 0,
ShardReplicationFactor);
InsertShardRow(relationId, shardId, SHARD_STORAGE_TABLE, nullMinValue, nullMaxValue);
InsertShardRow(relationId, shardId, storageType, nullMinValue, nullMaxValue);
PG_RETURN_INT64(shardId);
}
@ -171,9 +171,10 @@ master_append_table_to_shard(PG_FUNCTION_ARGS)
ShardInterval *shardInterval = LoadShardInterval(shardId);
Oid relationId = shardInterval->relationId;
bool cstoreTable = CStoreTable(relationId);
char storageType = shardInterval->storageType;
if (storageType != SHARD_STORAGE_TABLE)
if (storageType != SHARD_STORAGE_TABLE && !cstoreTable)
{
ereport(ERROR, (errmsg("cannot append to shardId " UINT64_FORMAT, shardId),
errdetail("The underlying shard is not a regular table")));
@ -457,7 +458,7 @@ WorkerShardStats(char *nodeName, uint32 nodePort, Oid relationId, char *shardNam
PG_TRY();
{
uint64 tableSize = WorkerTableSize(nodeName, nodePort, shardName);
uint64 tableSize = WorkerTableSize(nodeName, nodePort, relationId, shardName);
StringInfo minValue = WorkerPartitionValue(nodeName, nodePort, relationId,
shardName, SHARD_MIN_VALUE_QUERY);
StringInfo maxValue = WorkerPartitionValue(nodeName, nodePort, relationId,
@ -479,18 +480,27 @@ WorkerShardStats(char *nodeName, uint32 nodePort, Oid relationId, char *shardNam
/*
* WorkerTableSize queries the worker node to extract the disk space used by the
* given relation. The function assumes the relation represents a regular table.
* given relation. The function assumes the relation represents a regular table or
* a cstore_fdw table.
*/
static uint64
WorkerTableSize(char *nodeName, uint32 nodePort, char *tableName)
WorkerTableSize(char *nodeName, uint32 nodePort, Oid relationId, char *tableName)
{
uint64 tableSize = 0;
List *queryResultList = NIL;
StringInfo tableSizeString = NULL;
char *tableSizeStringEnd = NULL;
bool cstoreTable = CStoreTable(relationId);
StringInfo tableSizeQuery = makeStringInfo();
if (cstoreTable)
{
appendStringInfo(tableSizeQuery, SHARD_CSTORE_TABLE_SIZE_QUERY, tableName);
}
else
{
appendStringInfo(tableSizeQuery, SHARD_TABLE_SIZE_QUERY, tableName);
}
queryResultList = ExecuteRemoteQuery(nodeName, nodePort, tableSizeQuery);
if (queryResultList == NIL)

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

@ -393,6 +393,7 @@ DistributedModifyTask(Query *query)
query->onConflict = RebuildOnConflict(relationId, query->onConflict);
}
#else
/* always set to false for PG_VERSION_NUM < 90500 */
upsertQuery = false;
#endif
@ -414,6 +415,7 @@ DistributedModifyTask(Query *query)
#if (PG_VERSION_NUM >= 90500)
/*
* RebuildOnConflict rebuilds OnConflictExpr for correct deparsing. The function
* makes WHERE clause elements explicit and filters dropped columns
@ -468,6 +470,8 @@ RebuildOnConflict(Oid relationId, OnConflictExpr *originalOnConflict)
return updatedOnConflict;
}
#endif

39
src/backend/distributed/planner/multi_join_order.c
View File

@ -54,7 +54,8 @@ static bool JoinExprListWalker(Node *node, List **joinList);
static bool ExtractLeftMostRangeTableIndex(Node *node, int *rangeTableIndex);
static List * MergeShardIntervals(List *leftShardIntervalList,
List *rightShardIntervalList, JoinType joinType);
static bool ShardIntervalsMatch(List *leftShardIntervalList, List *rightShardIntervalList);
static bool ShardIntervalsMatch(List *leftShardIntervalList,
List *rightShardIntervalList);
static List * LoadSortedShardIntervalList(Oid relationId);
static List * JoinOrderForTable(TableEntry *firstTable, List *tableEntryList,
List *joinClauseList);
@ -68,31 +69,41 @@ static List * TableEntryListDifference(List *lhsTableList, List *rhsTableList);
static TableEntry * FindTableEntry(List *tableEntryList, uint32 tableId);
/* Local functions forward declarations for join evaluations */
static JoinOrderNode * EvaluateJoinRules(List *joinedTableList, JoinOrderNode *currentJoinNode,
TableEntry *candidateTable, List *candidateShardList,
static JoinOrderNode * EvaluateJoinRules(List *joinedTableList,
JoinOrderNode *currentJoinNode,
TableEntry *candidateTable,
List *candidateShardList,
List *joinClauseList, JoinType joinType);
static List * RangeTableIdList(List *tableList);
static RuleEvalFunction JoinRuleEvalFunction(JoinRuleType ruleType);
static char * JoinRuleName(JoinRuleType ruleType);
static JoinOrderNode * BroadcastJoin(JoinOrderNode *joinNode, TableEntry *candidateTable,
List *candidateShardList, List *applicableJoinClauses,
List *candidateShardList,
List *applicableJoinClauses,
JoinType joinType);
static JoinOrderNode * LocalJoin(JoinOrderNode *joinNode, TableEntry *candidateTable,
List *candidateShardList, List *applicableJoinClauses,
JoinType joinType);
static bool JoinOnColumns(Var *currentPartitioncolumn, Var *candidatePartitionColumn,
List *joinClauseList);
static JoinOrderNode * SinglePartitionJoin(JoinOrderNode *joinNode, TableEntry *candidateTable,
List *candidateShardList, List *applicableJoinClauses,
static JoinOrderNode * SinglePartitionJoin(JoinOrderNode *joinNode,
TableEntry *candidateTable,
List *candidateShardList,
List *applicableJoinClauses,
JoinType joinType);
static JoinOrderNode * DualPartitionJoin(JoinOrderNode *joinNode, TableEntry *candidateTable,
List *candidateShardList, List *applicableJoinClauses,
static JoinOrderNode * DualPartitionJoin(JoinOrderNode *joinNode,
TableEntry *candidateTable,
List *candidateShardList,
List *applicableJoinClauses,
JoinType joinType);
static JoinOrderNode * CartesianProduct(JoinOrderNode *joinNode, TableEntry *candidateTable,
List *candidateShardList, List *applicableJoinClauses,
static JoinOrderNode * CartesianProduct(JoinOrderNode *joinNode,
TableEntry *candidateTable,
List *candidateShardList,
List *applicableJoinClauses,
JoinType joinType);
static JoinOrderNode * MakeJoinOrderNode(TableEntry *tableEntry, JoinRuleType joinRuleType,
Var *partitionColumn, char partitionMethod);
static JoinOrderNode * MakeJoinOrderNode(TableEntry *tableEntry, JoinRuleType
joinRuleType, Var *partitionColumn,
char partitionMethod);
/*
@ -199,7 +210,6 @@ FixedJoinOrderList(FromExpr *fromExpr, List *tableEntryList)
"query"),
errdetail("Shards of relations in outer join queries "
"must have 1-to-1 shard partitioning")));
}
}
else
@ -586,7 +596,8 @@ ShardIntervalsMatch(List *leftShardIntervalList, List *rightShardIntervalList)
nextRightIntervalCell = lnext(rightShardIntervalCell);
if (nextRightIntervalCell != NULL)
{
ShardInterval *nextRightInterval = (ShardInterval *) lfirst(nextRightIntervalCell);
ShardInterval *nextRightInterval =
(ShardInterval *) lfirst(nextRightIntervalCell);
shardIntervalsIntersect = ShardIntervalsOverlap(leftInterval,
nextRightInterval);
if (shardIntervalsIntersect)

34
src/backend/distributed/planner/multi_logical_optimizer.c
View File

@ -91,7 +91,8 @@ static void ParentSetNewChild(MultiNode *parentNode, MultiNode *oldChildNode,
/* Local functions forward declarations for aggregate expressions */
static void ApplyExtendedOpNodes(MultiExtendedOp *originalNode,
MultiExtendedOp *masterNode, MultiExtendedOp *workerNode);
MultiExtendedOp *masterNode,
MultiExtendedOp *workerNode);
static void TransformSubqueryNode(MultiTable *subqueryNode);
static MultiExtendedOp * MasterExtendedOpNode(MultiExtendedOp *originalOpNode);
static Node * MasterAggregateMutator(Node *originalNode, AttrNumber *columnId);
@ -117,7 +118,8 @@ static void ErrorIfUnsupportedArrayAggregate(Aggref *arrayAggregateExpression);
static void ErrorIfUnsupportedAggregateDistinct(Aggref *aggregateExpression,
MultiNode *logicalPlanNode);
static Var * AggregateDistinctColumn(Aggref *aggregateExpression);
static bool TablePartitioningSupportsDistinct(List *tableNodeList, MultiExtendedOp *opNode,
static bool TablePartitioningSupportsDistinct(List *tableNodeList,
MultiExtendedOp *opNode,
Var *distinctColumn);
static bool GroupedByColumn(List *groupClauseList, List *targetList, Var *column);
@ -257,6 +259,7 @@ MultiLogicalPlanOptimize(MultiTreeRoot *multiLogicalPlan)
MultiTable *tableNode = (MultiTable *) lfirst(tableNodeCell);
if (tableNode->relationId == SUBQUERY_RELATION_ID)
{
ErrorIfContainsUnsupportedAggregate((MultiNode *) tableNode);
TransformSubqueryNode(tableNode);
}
}
@ -1980,23 +1983,31 @@ CountDistinctHashFunctionName(Oid argumentType)
switch (argumentType)
{
case INT4OID:
{
hashFunctionName = pstrdup(HLL_HASH_INTEGER_FUNC_NAME);
break;
}
case INT8OID:
{
hashFunctionName = pstrdup(HLL_HASH_BIGINT_FUNC_NAME);
break;
}
case TEXTOID:
case BPCHAROID:
case VARCHAROID:
{
hashFunctionName = pstrdup(HLL_HASH_TEXT_FUNC_NAME);
break;
}
default:
{
hashFunctionName = pstrdup(HLL_HASH_ANY_FUNC_NAME);
break;
}
}
return hashFunctionName;
}
@ -2135,8 +2146,9 @@ ErrorIfUnsupportedAggregateDistinct(Aggref *aggregateExpression,
bool distinctSupported = true;
List *repartitionNodeList = NIL;
Var *distinctColumn = NULL;
AggregateType aggregateType = GetAggregateType(aggregateExpression->aggfnoid);
List *multiTableNodeList = NIL;
ListCell *multiTableNodeCell = NULL;
AggregateType aggregateType = AGGREGATE_INVALID_FIRST;
/* check if logical plan includes a subquery */
List *subqueryMultiTableList = SubqueryMultiTableList(logicalPlanNode);
@ -2147,7 +2159,20 @@ ErrorIfUnsupportedAggregateDistinct(Aggref *aggregateExpression,
errdetail("distinct in the outermost query is unsupported")));
}
multiTableNodeList = FindNodesOfType(logicalPlanNode, T_MultiTable);
foreach(multiTableNodeCell, multiTableNodeList)
{
MultiTable *multiTable = (MultiTable *) lfirst(multiTableNodeCell);
if (multiTable->relationId == SUBQUERY_RELATION_ID)
{
ereport(ERROR, (errmsg("cannot compute count (distinct)"),
errdetail("Subqueries with aggregate (distinct) are "
"not supported yet")));
}
}
/* if we have a count(distinct), and distinct approximation is enabled */
aggregateType = GetAggregateType(aggregateExpression->aggfnoid);
if (aggregateType == AGGREGATE_COUNT &&
CountDistinctErrorRate != DISABLE_DISTINCT_APPROXIMATION)
{
@ -3229,6 +3254,7 @@ SupportedLateralQuery(Query *parentQuery, Query *lateralQuery)
CompositeFieldRecursive(outerQueryExpression, parentQuery);
FieldSelect *localCompositeField =
CompositeFieldRecursive(localQueryExpression, lateralQuery);
/*
* If partition colums are composite fields, add them to list to
* check later if all composite fields are used.

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

@ -285,6 +285,7 @@ MultiPlanTree(Query *queryTree)
else
{
bool hasOuterJoin = false;
/*
* We calculate the join order using the list of tables in the query and
* the join clauses between them. Note that this function owns the table
@ -465,6 +466,7 @@ ErrorIfQueryNotSupported(Query *queryTree)
#if (PG_VERSION_NUM >= 90500)
/* HasTablesample returns tree if the query contains tablesample */
static bool
HasTablesample(Query *queryTree)
@ -485,6 +487,8 @@ HasTablesample(Query *queryTree)
return hasTablesample;
}
#endif
@ -529,7 +533,8 @@ HasUnsupportedJoinWalker(Node *node, void *context)
* ErrorIfSubqueryNotSupported checks that we can perform distributed planning for
* the given subquery.
*/
static void ErrorIfSubqueryNotSupported(Query *subqueryTree)
static void
ErrorIfSubqueryNotSupported(Query *subqueryTree)
{
char *errorDetail = NULL;
bool preconditionsSatisfied = true;
@ -587,7 +592,6 @@ HasOuterJoin(Query *queryTree)
static bool
HasOuterJoinWalker(Node *node, void *context)
{
bool hasOuterJoin = false;
if (node == NULL)
{

16
src/backend/distributed/planner/multi_physical_planner.c
View File

@ -155,7 +155,8 @@ static StringInfo DatumArrayString(Datum *datumArray, uint32 datumCount, Oid dat
static Task * CreateBasicTask(uint64 jobId, uint32 taskId, TaskType taskType,
char *queryString);
static void UpdateRangeTableAlias(List *rangeTableList, List *fragmentList);
static Alias * FragmentAlias(RangeTblEntry *rangeTableEntry, RangeTableFragment *fragment);
static Alias * FragmentAlias(RangeTblEntry *rangeTableEntry,
RangeTableFragment *fragment);
static uint64 AnchorShardId(List *fragmentList, uint32 anchorRangeTableId);
static List * PruneSqlTaskDependencies(List *sqlTaskList);
static List * AssignTaskList(List *sqlTaskList);
@ -309,6 +310,7 @@ BuildJobTree(MultiTreeRoot *multiTree)
partitionKey, partitionType,
baseRelationId,
JOIN_MAP_MERGE_JOB);
/* reset depended job list */
loopDependedJobList = NIL;
loopDependedJobList = list_make1(mapMergeJob);
@ -1093,7 +1095,8 @@ QueryJoinTree(MultiNode *multiNode, List *dependedJobList, List **rangeTableList
uint32 columnCount = (uint32) list_length(dependedTargetList);
List *columnNameList = DerivedColumnNameList(columnCount, dependedJob->jobId);
RangeTblEntry *rangeTableEntry = DerivedRangeTableEntry(multiNode, columnNameList,
RangeTblEntry *rangeTableEntry = DerivedRangeTableEntry(multiNode,
columnNameList,
tableIdList);
RangeTblRef *rangeTableRef = makeNode(RangeTblRef);
@ -1864,6 +1867,7 @@ SplitPointObject(ShardInterval **shardIntervalArray, uint32 shardIntervalCount)
return splitPointObject;
}
/* ------------------------------------------------------------
* Functions that relate to building and assigning tasks follow
* ------------------------------------------------------------
@ -4046,6 +4050,7 @@ FragmentAlias(RangeTblEntry *rangeTableEntry, RangeTableFragment *fragment)
return alias;
}
/*
* AnchorShardId walks over each fragment in the given fragment list, finds the
* fragment that corresponds to the given anchor range tableId, and returns this
@ -4984,6 +4989,7 @@ RoundRobinAssignTaskList(List *taskList)
return taskList;
}
/*
* RoundRobinReorder implements the core of the round-robin assignment policy.
* It takes a task and placement list and rotates a copy of the placement list
@ -5116,7 +5122,8 @@ ActiveShardPlacementLists(List *taskList)
List *activeShardPlacementList = ActivePlacementList(shardPlacementList);
/* sort shard placements by their insertion time */
activeShardPlacementList = SortList(activeShardPlacementList, CompareShardPlacements);
activeShardPlacementList = SortList(activeShardPlacementList,
CompareShardPlacements);
shardPlacementLists = lappend(shardPlacementLists, activeShardPlacementList);
}
@ -5257,7 +5264,8 @@ AssignDualHashTaskList(List *taskList)
uint32 replicaIndex = 0;
for (replicaIndex = 0; replicaIndex < ShardReplicationFactor; replicaIndex++)
{
uint32 assignmentOffset = beginningNodeIndex + assignedTaskIndex + replicaIndex;
uint32 assignmentOffset = beginningNodeIndex + assignedTaskIndex +
replicaIndex;
uint32 assignmentIndex = assignmentOffset % workerNodeCount;
WorkerNode *workerNode = list_nth(workerNodeList, assignmentIndex);

11
src/backend/distributed/relay/relay_event_utility.c
View File

@ -205,20 +205,31 @@ RelayEventExtendNames(Node *parseTree, uint64 shardId)
switch (relationNameListLength)
{
case 1:
{
relationNameValue = linitial(relationNameList);
break;
}
case 2:
{
relationNameValue = lsecond(relationNameList);
break;
}
case 3:
{
relationNameValue = lthird(relationNameList);
break;
}
default:
{
ereport(ERROR, (errcode(ERRCODE_SYNTAX_ERROR),
errmsg("improper relation name: \"%s\"",
NameListToString(relationNameList))));
break;
}
}
relationName = &(relationNameValue->val.str);
AppendShardIdToName(relationName, shardId);

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

@ -206,9 +206,10 @@ RegisterCitusConfigVariables(void)
DefineCustomBoolVariable(
"citus.expire_cached_shards",
gettext_noop("Enables shard cache expiration if a shard's size on disk has changed. "),
gettext_noop("When appending to an existing shard, old data may still be cached on "
"other workers. This configuration entry activates automatic "
gettext_noop("Enables shard cache expiration if a shard's size on disk has "
"changed."),
gettext_noop("When appending to an existing shard, old data may still be cached "
"on other workers. This configuration entry activates automatic "
"expiration, but should not be used with manual updates to shards."),
&ExpireCachedShards,
false,
@ -488,8 +489,6 @@ RegisterCitusConfigVariables(void)
/* warn about config items in the citus namespace that are not registered above */
EmitWarningsOnPlaceholders("citus");
/* Also warn about citus namespace, as that's a very likely misspelling */
EmitWarningsOnPlaceholders("citus");
}
@ -515,8 +514,10 @@ NormalizeWorkerListPath(void)
{
absoluteFileName = malloc(strlen(DataDir) + strlen(WORKER_LIST_FILENAME) + 2);
if (absoluteFileName == NULL)
{
ereport(FATAL, (errcode(ERRCODE_OUT_OF_MEMORY),
errmsg("out of memory")));
}
sprintf(absoluteFileName, "%s/%s", DataDir, WORKER_LIST_FILENAME);
}
@ -530,6 +531,7 @@ NormalizeWorkerListPath(void)
"environment variable.\n", progname, ConfigFileName)));
}
SetConfigOption("citus.worker_list_file", absoluteFileName, PGC_POSTMASTER, PGC_S_OVERRIDE);
SetConfigOption("citus.worker_list_file", absoluteFileName, PGC_POSTMASTER,
PGC_S_OVERRIDE);
free(absoluteFileName);
}

5
src/backend/distributed/utils/citus_nodefuncs.c
View File

@ -273,9 +273,13 @@ GetRangeTblKind(RangeTblEntry *rte)
case RTE_JOIN:
case RTE_VALUES:
case RTE_CTE:
{
rteKind = (CitusRTEKind) rte->rtekind;
break;
}
case RTE_FUNCTION:
{
/*
* Extract extra data - correct even if a plain RTE_FUNCTION, not
* an extended one, ExtractRangeTblExtraData handles that case
@ -284,6 +288,7 @@ GetRangeTblKind(RangeTblEntry *rte)
ExtractRangeTblExtraData(rte, &rteKind, NULL, NULL, NULL);
break;
}
}
return rteKind;
}

16
src/backend/distributed/utils/citus_ruleutils.c
View File

@ -102,6 +102,7 @@ pg_get_extensiondef_string(Oid tableRelationId)
static Oid
get_extension_schema(Oid ext_oid)
{
/* *INDENT-OFF* */
Oid result;
Relation rel;
SysScanDesc scandesc;
@ -131,6 +132,7 @@ get_extension_schema(Oid ext_oid)
heap_close(rel, AccessShareLock);
return result;
/* *INDENT-ON* */
}
@ -447,22 +449,36 @@ pg_get_tablecolumnoptionsdef_string(Oid tableRelationId)
switch (attributeForm->attstorage)
{
case 'p':
{
storageName = "PLAIN";
break;
}
case 'e':
{
storageName = "EXTERNAL";
break;
}
case 'm':
{
storageName = "MAIN";
break;
}
case 'x':
{
storageName = "EXTENDED";
break;
}
default:
{
ereport(ERROR, (errmsg("unrecognized storage type: %c",
attributeForm->attstorage)));
break;
}
}
appendStringInfo(&statement, "ALTER COLUMN %s ",
quote_identifier(attributeName));

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

@ -87,6 +87,7 @@ IsDistributedTable(Oid relationId)
return cacheEntry->isDistributedTable;
}
/*
* LoadShardInterval reads shard metadata for given shardId from pg_dist_shard,
* and converts min/max values in these metadata to their properly typed datum
@ -139,6 +140,7 @@ LoadShardInterval(uint64 shardId)
return shardInterval;
}
/*
* DistributedTableCacheEntry looks up a pg_dist_partition entry for a
* relation.

16
src/backend/distributed/utils/multi_resowner.c
View File

@ -22,7 +22,8 @@
#include "distributed/multi_resowner.h"
typedef struct JobDirectoryEntry {
typedef struct JobDirectoryEntry
{
ResourceOwner owner;
uint64 jobId;
} JobDirectoryEntry;
@ -91,15 +92,17 @@ ResourceOwnerEnlargeJobDirectories(ResourceOwner owner)
if (RegisteredJobDirectories == NULL)
{
newMax = 16;
RegisteredJobDirectories = (JobDirectoryEntry *)
MemoryContextAlloc(TopMemoryContext, newMax * sizeof(JobDirectoryEntry));
RegisteredJobDirectories =
(JobDirectoryEntry *) MemoryContextAlloc(TopMemoryContext,
newMax * sizeof(JobDirectoryEntry));
NumAllocatedJobDirectories = newMax;
}
else if (NumRegisteredJobDirectories + 1 > NumAllocatedJobDirectories)
{
newMax = NumAllocatedJobDirectories * 2;
RegisteredJobDirectories = (JobDirectoryEntry *)
repalloc(RegisteredJobDirectories, newMax * sizeof(JobDirectoryEntry));
RegisteredJobDirectories =
(JobDirectoryEntry *) repalloc(RegisteredJobDirectories,
newMax * sizeof(JobDirectoryEntry));
NumAllocatedJobDirectories = newMax;
}
}
@ -135,7 +138,8 @@ ResourceOwnerForgetJobDirectory(ResourceOwner owner, uint64 jobId)
/* move all later entries one up */
while (jobIndex < lastJobIndex)
{
RegisteredJobDirectories[jobIndex] = RegisteredJobDirectories[jobIndex + 1];
RegisteredJobDirectories[jobIndex] =
RegisteredJobDirectories[jobIndex + 1];
jobIndex++;
}
NumRegisteredJobDirectories = lastJobIndex;

5
src/backend/distributed/utils/resource_lock.c
View File

@ -14,9 +14,10 @@
*/
#include "postgres.h"
#include "c.h"
#include "miscadmin.h"
#include "distributed/relay_utility.h"
#include "distributed/resource_lock.h"
#include "storage/lmgr.h"
@ -68,6 +69,8 @@ LockShardResource(uint64 shardId, LOCKMODE lockmode)
const bool sessionLock = false;
const bool dontWait = false;
AssertArg(shardId != INVALID_SHARD_ID);
SET_LOCKTAG_SHARD_RESOURCE(tag, MyDatabaseId, shardId);
(void) LockAcquire(&tag, lockmode, sessionLock, dontWait);

6
src/backend/distributed/worker/task_tracker.c
View File

@ -494,6 +494,7 @@ TrackerDelayLoop(void)
}
}
/* ------------------------------------------------------------
* Signal handling and shared hash initialization functions follow
* ------------------------------------------------------------
@ -579,8 +580,8 @@ TaskTrackerShmemInit(void)
LWLockAcquire(AddinShmemInitLock, LW_EXCLUSIVE);
/* allocate struct containing task tracker related shared state */
WorkerTasksSharedState = (WorkerTasksSharedStateData *)
ShmemInitStruct("Worker Task Control",
WorkerTasksSharedState =
(WorkerTasksSharedStateData *) ShmemInitStruct("Worker Task Control",
sizeof(WorkerTasksSharedStateData),
&alreadyInitialized);
@ -607,6 +608,7 @@ TaskTrackerShmemInit(void)
}
}
/* ------------------------------------------------------------
* Task scheduling and management functions follow
* ------------------------------------------------------------

2
src/backend/distributed/worker/task_tracker_protocol.c
View File

@ -331,7 +331,7 @@ UpdateTask(WorkerTask *workerTask, char *taskCallString)
if (taskStatus == TASK_SUCCEEDED || taskStatus == TASK_CANCEL_REQUESTED ||
taskStatus == TASK_CANCELED)
{
; /* nothing to do */
/* nothing to do */
}
else if (taskStatus == TASK_PERMANENTLY_FAILED)
{

23
src/backend/distributed/worker/worker_data_fetch_protocol.c
View File

@ -53,11 +53,14 @@ static void ReceiveResourceCleanup(int32 connectionId, const char *filename,
static void DeleteFile(const char *filename);
static void FetchTableCommon(text *tableName, uint64 remoteTableSize,
ArrayType *nodeNameObject, ArrayType *nodePortObject,
bool (*FetchTableFunction) (const char *, uint32, StringInfo));
bool (*FetchTableFunction)(const char *, uint32,
StringInfo));
static uint64 LocalTableSize(Oid relationId);
static uint64 ExtractShardId(StringInfo tableName);
static bool FetchRegularTable(const char *nodeName, uint32 nodePort, StringInfo tableName);
static bool FetchForeignTable(const char *nodeName, uint32 nodePort, StringInfo tableName);
static bool FetchRegularTable(const char *nodeName, uint32 nodePort,
StringInfo tableName);
static bool FetchForeignTable(const char *nodeName, uint32 nodePort,
StringInfo tableName);
static List * TableDDLCommandList(const char *nodeName, uint32 nodePort,
StringInfo tableName);
static StringInfo ForeignFilePath(const char *nodeName, uint32 nodePort,
@ -309,7 +312,7 @@ ReceiveRegularFile(const char *nodeName, uint32 nodePort,
}
else if (copyStatus == CLIENT_COPY_MORE)
{
; /* remote node will continue to send more data */
/* remote node will continue to send more data */
}
else
{
@ -994,11 +997,10 @@ worker_append_table_to_shard(PG_FUNCTION_ARGS)
StringInfo remoteCopyCommand = NULL;
CopyStmt *localCopyCommand = NULL;
RangeVar *localTable = NULL;
uint64 copiedRowCount = 0;
uint64 shardId = INVALID_SHARD_ID;
bool received = false;
char *quotedTableName = NULL;
const char *queryString = NULL;
StringInfo queryString = NULL;
const char *schemaName = NULL;
/* copy remote table's data to this node */
@ -1032,8 +1034,13 @@ worker_append_table_to_shard(PG_FUNCTION_ARGS)
localTable = makeRangeVar((char *) schemaName, shardNameString->data, -1);
localCopyCommand = CopyStatement(localTable, localFilePath->data);
DoCopy(localCopyCommand, queryString, &copiedRowCount);
(void) copiedRowCount;
quotedTableName = quote_qualified_identifier(schemaName, shardNameString->data);
queryString = makeStringInfo();
appendStringInfo(queryString, COPY_IN_COMMAND, quotedTableName, localFilePath->data);
ProcessUtility((Node *) localCopyCommand, queryString->data,
PROCESS_UTILITY_TOPLEVEL, NULL, None_Receiver, NULL);
/* finally delete the temporary file we created */
DeleteFile(localFilePath->data);

4
src/backend/distributed/worker/worker_merge_protocol.c
View File

@ -256,8 +256,8 @@ JobSchemaName(uint64 jobId)
*/
#ifdef HAVE_INTTYPES_H
StringInfo jobSchemaName = makeStringInfo();
appendStringInfo(jobSchemaName, "%s%0*"PRIu64,
JOB_SCHEMA_PREFIX, MIN_JOB_DIRNAME_WIDTH, jobId);
appendStringInfo(jobSchemaName, "%s%0*" PRIu64, JOB_SCHEMA_PREFIX,
MIN_JOB_DIRNAME_WIDTH, jobId);
#else
StringInfo jobSchemaName = makeStringInfo();
appendStringInfo(jobSchemaName, "%s%0*llu",

2
src/backend/distributed/worker/worker_partition_protocol.c
View File

@ -1143,6 +1143,7 @@ OutputBinaryFooters(FileOutputStream *partitionFileArray, uint32 fileCount)
}
/* *INDENT-OFF* */
/* Append data to the copy buffer in outputState */
static void
CopySendData(PartialCopyState outputState, const void *databuf, int datasize)
@ -1282,6 +1283,7 @@ CopyAttributeOutText(PartialCopyState cstate, char *string)
}
/* *INDENT-ON* */
/* Helper function to send pending copy output */
static inline void
CopyFlushOutput(PartialCopyState cstate, char *start, char *pointer)

31
src/bin/csql/copy_options.c
View File

@ -16,7 +16,22 @@
#include "stringutils.h"
/* Concatenates "more" onto "var", and frees the original value of *var. */
/* *INDENT-OFF* */
void
free_copy_options(copy_options * ptr)
{
if (!ptr)
return;
free(ptr->before_tofrom);
free(ptr->after_tofrom);
free(ptr->file);
free(ptr->tableName);
free(ptr->columnList);
free(ptr);
}
/* concatenate "more" onto "var", freeing the original value of *var */
static void
xstrcat(char **var, const char *more)
{
@ -210,21 +225,9 @@ error:
return NULL;
}
/* *INDENT-ON* */
/* Frees copy options. */
void
free_copy_options(copy_options * ptr)
{
if (!ptr)
return;
free(ptr->before_tofrom);
free(ptr->after_tofrom);
free(ptr->file);
free(ptr->tableName);
free(ptr->columnList);
free(ptr);
}
/*
* ParseStageOptions takes the given copy options, parses the additional options

10
src/bin/csql/stage.c
View File

@ -26,7 +26,8 @@
static bool FileSize(char *filename, uint64 *fileSize);
static PGconn * ConnectToWorkerNode(const char *nodeName, uint32 nodePort,
const char *nodeDatabase);
static PGresult * ExecuteRemoteCommand(PGconn *remoteConnection, const char *remoteCommand,
static PGresult * ExecuteRemoteCommand(PGconn *remoteConnection,
const char *remoteCommand,
const char **parameterValues, int parameterCount);
static TableMetadata * InitTableMetadata(const char *tableName);
static ShardMetadata * InitShardMetadata(int shardPlacementPolicy);
@ -41,7 +42,8 @@ static uint64 GetValueUint64(const PGresult *result, int rowNumber, int columnNu
static bool MasterGetTableMetadata(const char *tableName, TableMetadata *tableMetadata);
static bool MasterGetTableDDLEvents(const char *tableName, TableMetadata *tableMetadata);
static bool MasterGetNewShardId(ShardMetadata *shardMetadata);
static bool MasterGetCandidateNodes(ShardMetadata *shardMetadata, int shardPlacementPolicy);
static bool MasterGetCandidateNodes(ShardMetadata *shardMetadata,
int shardPlacementPolicy);
static bool MasterInsertShardRow(uint32 logicalRelid, char storageType,
const ShardMetadata *shardMetadata);
static bool MasterInsertPlacementRows(const ShardMetadata *shardMetadata);
@ -62,7 +64,8 @@ static bool ApplyShardDDLCommand(PGconn *workerNode, uint64 shardId, const char
static bool TransmitTableData(PGconn *workerNode, uint64 shardId,
uint64 shardMaxSize, copy_options *stageOptions,
uint64 currentFileOffset, uint64 *nextFileOffset);
static bool TransmitFile(PGconn *workerNode, const char *localPath, const char *remotePath);
static bool TransmitFile(PGconn *workerNode, const char *localPath,
const char *remotePath);
static bool FileStreamOK(const copy_options *stageOptions);
static PQExpBuffer CreateCopyQueryString(const char *tableName, const char *columnList,
const char *afterToFrom);
@ -341,7 +344,6 @@ DoStageData(const char *stageCommand)
/* update current file offset */
currentFileOffset = nextFileOffset;
} /* while more file data left for sharding */
/*

34
src/bin/csql/stage.h
View File

@ -33,17 +33,19 @@
#define MASTER_GET_TABLE_METADATA "SELECT * FROM master_get_table_metadata($1::text)"
#define MASTER_GET_TABLE_DDL_EVENTS "SELECT * FROM master_get_table_ddl_events($1::text)"
#define MASTER_GET_NEW_SHARDID "SELECT * FROM master_get_new_shardid()"
#define MASTER_GET_LOCAL_FIRST_CANDIDATE_NODES "SELECT * FROM \
master_get_local_first_candidate_nodes()"
#define MASTER_GET_ROUND_ROBIN_CANDIDATE_NODES "SELECT * FROM \
master_get_round_robin_candidate_nodes($1::int8)"
#define MASTER_GET_LOCAL_FIRST_CANDIDATE_NODES \
"SELECT * FROM master_get_local_first_candidate_nodes()"
#define MASTER_GET_ROUND_ROBIN_CANDIDATE_NODES \
"SELECT * FROM master_get_round_robin_candidate_nodes($1::int8)"
#define MASTER_INSERT_SHARD_ROW "INSERT INTO pg_dist_shard \
(logicalrelid, shardid, shardstorage, shardminvalue, shardmaxvalue) VALUES \
($1::oid, $2::int8, $3::char, $4::text, $5::text)"
#define MASTER_INSERT_PLACEMENT_ROW "INSERT INTO pg_dist_shard_placement \
(shardid, shardstate, shardlength, nodename, nodeport) VALUES \
($1::int8, $2::int4, $3::int8, $4::text, $5::int4)"
#define MASTER_INSERT_SHARD_ROW \
"INSERT INTO pg_dist_shard " \
"(logicalrelid, shardid, shardstorage, shardminvalue, shardmaxvalue) VALUES " \
"($1::oid, $2::int8, $3::char, $4::text, $5::text)"
#define MASTER_INSERT_PLACEMENT_ROW \
"INSERT INTO pg_dist_shard_placement " \
"(shardid, shardstate, shardlength, nodename, nodeport) VALUES " \
"($1::int8, $2::int4, $3::int8, $4::text, $5::int4)"
/* Column names used to identify response fields as returned from the master. */
#define LOGICAL_RELID_FIELD "logical_relid"
@ -61,11 +63,11 @@
#define SHARD_MIN_MAX_COMMAND "SELECT min(%s), max(%s) FROM %s"
#define SHARD_TABLE_SIZE_COMMAND "SELECT pg_table_size('%s')"
#define SET_FOREIGN_TABLE_FILENAME "ALTER FOREIGN TABLE %s OPTIONS (SET filename '%s')"
#define GET_COLUMNAR_TABLE_FILENAME_OPTION "SELECT * FROM \
(SELECT (pg_options_to_table(ftoptions)).* FROM pg_foreign_table \
WHERE ftrelid = %u) AS Q WHERE option_name = 'filename';"
#define APPLY_SHARD_DDL_COMMAND "SELECT * FROM worker_apply_shard_ddl_command \
($1::int8, $2::text)"
#define GET_COLUMNAR_TABLE_FILENAME_OPTION \
"SELECT * FROM (SELECT (pg_options_to_table(ftoptions)).* FROM pg_foreign_table " \
"WHERE ftrelid = %u) AS Q WHERE option_name = 'filename';"
#define APPLY_SHARD_DDL_COMMAND \
"SELECT * FROM worker_apply_shard_ddl_command ($1::int8, $2::text)"
#define REMOTE_FILE_SIZE_COMMAND "SELECT size FROM pg_stat_file('%s')"
#define SHARD_COLUMNAR_TABLE_SIZE_COMMAND "SELECT cstore_table_size('%s')"
@ -100,7 +102,6 @@ typedef struct TableMetadata
char **ddlEventList; /* DDL statements used for creating new shard */
uint32 ddlEventCount; /* DDL statement count; statement list size */
} TableMetadata;
@ -122,7 +123,6 @@ typedef struct ShardMetadata
char *shardMinValue; /* partition key's minimum value in shard */
char *shardMaxValue; /* partition key's maximum value in shard */
uint64 shardSize; /* shard size; updated during staging */
} ShardMetadata;

8
src/include/distributed/citus_ruleutils.h
View File

@ -1,8 +1,13 @@
/*-------------------------------------------------------------------------
*
* citus_ruleutils.h
<<<<<<< HEAD
* Citus ruleutils wrapper functions and exported PostgreSQL ruleutils
* functions.
=======
* CitusDB ruleutils wrapper functions and exported PostgreSQL ruleutils
* functions.
>>>>>>> origin
*
* Copyright (c) 2012-2015, Citus Data, Inc.
*-------------------------------------------------------------------------
@ -24,7 +29,8 @@ extern char *pg_get_indexclusterdef_string(Oid indexRelationId);
/* Function declarations for version dependent PostgreSQL ruleutils functions */
extern void pg_get_query_def(Query *query, StringInfo buffer);
extern void deparse_shard_query(Query *query, Oid distrelid, int64 shardid, StringInfo buffer);
extern void deparse_shard_query(Query *query, Oid distrelid, int64 shardid, StringInfo
buffer);
extern char * generate_relation_name(Oid relid, List *namespaces);

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

@ -38,7 +38,6 @@ typedef struct ShardInterval
Datum minValue; /* a shard's typed min value datum */
Datum maxValue; /* a shard's typed max value datum */
uint64 shardId;
} ShardInterval;
@ -52,7 +51,6 @@ typedef struct ShardPlacement
RelayFileState shardState;
char *nodeName;
uint32 nodePort;
} ShardPlacement;

10
src/include/distributed/master_protocol.h
View File

@ -49,13 +49,14 @@
#define SHARDID_SEQUENCE_NAME "pg_dist_shardid_seq"
/* Remote call definitions to help with data staging and deletion */
#define WORKER_APPLY_SHARD_DDL_COMMAND "SELECT worker_apply_shard_ddl_command \
("UINT64_FORMAT", %s)"
#define WORKER_APPEND_TABLE_TO_SHARD "SELECT worker_append_table_to_shard \
(%s, %s, %s, %u)"
#define WORKER_APPLY_SHARD_DDL_COMMAND \
"SELECT worker_apply_shard_ddl_command (" UINT64_FORMAT ", %s)"
#define WORKER_APPEND_TABLE_TO_SHARD \
"SELECT worker_append_table_to_shard (%s, %s, %s, %u)"
#define SHARD_MIN_VALUE_QUERY "SELECT min(%s) FROM %s"
#define SHARD_MAX_VALUE_QUERY "SELECT max(%s) FROM %s"
#define SHARD_TABLE_SIZE_QUERY "SELECT pg_table_size('%s')"
#define SHARD_CSTORE_TABLE_SIZE_QUERY "SELECT cstore_table_size('%s')"
#define DROP_REGULAR_TABLE_COMMAND "DROP TABLE IF EXISTS %s"
#define DROP_FOREIGN_TABLE_COMMAND "DROP FOREIGN TABLE IF EXISTS %s"
#define CREATE_SCHEMA_COMMAND "CREATE SCHEMA IF NOT EXISTS %s"
@ -67,7 +68,6 @@ typedef enum
SHARD_PLACEMENT_INVALID_FIRST = 0,
SHARD_PLACEMENT_LOCAL_NODE_FIRST = 1,
SHARD_PLACEMENT_ROUND_ROBIN = 2
} ShardPlacementPolicyType;

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

@ -24,6 +24,7 @@
#define INVALID_TASK_ID 0
#if (PG_VERSION_NUM >= 90500)
/* reserved alias name for UPSERTs */
#define UPSERT_ALIAS "citus_table_alias"
#endif

5
src/include/distributed/multi_client_executor.h
View File

@ -29,7 +29,6 @@ typedef enum
CLIENT_CONNECTION_BAD = 1,
CLIENT_CONNECTION_BUSY = 2,
CLIENT_CONNECTION_READY = 3
} ConnectStatus;
@ -40,7 +39,6 @@ typedef enum
CLIENT_RESULT_UNAVAILABLE = 1,
CLIENT_RESULT_BUSY = 2,
CLIENT_RESULT_READY = 3
} ResultStatus;
@ -51,7 +49,6 @@ typedef enum
CLIENT_QUERY_FAILED = 1,
CLIENT_QUERY_DONE = 2,
CLIENT_QUERY_COPY = 3
} QueryStatus;
@ -62,7 +59,6 @@ typedef enum
CLIENT_COPY_MORE = 1,
CLIENT_COPY_FAILED = 2,
CLIENT_COPY_DONE = 3
} CopyStatus;
@ -73,7 +69,6 @@ typedef enum
CLIENT_BATCH_QUERY_FAILED = 1,
CLIENT_BATCH_QUERY_CONTINUE = 2,
CLIENT_BATCH_QUERY_DONE = 3
} BatchQueryStatus;

3
src/include/distributed/multi_join_order.h
View File

@ -40,7 +40,6 @@ typedef enum JoinRuleType
* RuleNameArray.
*/
JOIN_RULE_LAST
} JoinRuleType;
@ -53,7 +52,6 @@ typedef struct TableEntry
{
Oid relationId;
uint32 rangeTableId;
} TableEntry;
@ -72,7 +70,6 @@ typedef struct JoinOrderNode
char partitionMethod;
List *joinClauseList; /* not relevant for the first table */
List *shardIntervalList;
} JoinOrderNode;

9
src/include/distributed/multi_logical_optimizer.h
View File

@ -55,7 +55,6 @@ typedef enum
AGGREGATE_SUM = 4,
AGGREGATE_COUNT = 5,
AGGREGATE_ARRAY_AGG = 6
} AggregateType;
@ -69,7 +68,6 @@ typedef enum
PUSH_DOWN_VALID = 1,
PUSH_DOWN_NOT_VALID = 2,
PUSH_DOWN_SPECIAL_CONDITIONS = 3
} PushDownStatus;
@ -82,7 +80,6 @@ typedef enum
PULL_UP_INVALID_FIRST = 0,
PULL_UP_VALID = 1,
PULL_UP_NOT_VALID = 2
} PullUpStatus;
@ -97,8 +94,10 @@ typedef enum
* Please note that the order of elements in this array is tied to the order of
* values in the preceding AggregateType enum. This order needs to be preserved.
*/
static const char * const AggregateNames[] = { "invalid", "avg", "min", "max",
"sum", "count", "array_agg" };
static const char *const AggregateNames[] = {
"invalid", "avg", "min", "max", "sum",
"count", "array_agg"
};
/* Config variable managed via guc.c */

13
src/include/distributed/multi_logical_planner.h
View File

@ -40,8 +40,8 @@ typedef struct MultiNode
CitusNodeTag type;
struct MultiNode *parentNode;
/* child node(s) are defined in unary and binary nodes */
/* child node(s) are defined in unary and binary nodes */
} MultiNode;
@ -51,7 +51,6 @@ typedef struct MultiUnaryNode
MultiNode node;
struct MultiNode *childNode;
} MultiUnaryNode;
@ -62,7 +61,6 @@ typedef struct MultiBinaryNode
struct MultiNode *leftChildNode;
struct MultiNode *rightChildNode;
} MultiBinaryNode;
@ -73,7 +71,6 @@ typedef struct MultiBinaryNode
typedef struct MultiTreeRoot
{
MultiUnaryNode unaryNode;
} MultiTreeRoot;
@ -91,7 +88,6 @@ typedef struct MultiTable
Alias *alias;
Alias *referenceNames;
Query *subquery; /* this field is only valid for non-relation subquery types */
} MultiTable;
@ -100,7 +96,6 @@ typedef struct MultiProject
{
MultiUnaryNode unaryNode;
List *columnList;
} MultiProject;
@ -112,7 +107,6 @@ typedef struct MultiProject
typedef struct MultiCollect
{
MultiUnaryNode unaryNode;
} MultiCollect;
@ -125,7 +119,6 @@ typedef struct MultiSelect
{
MultiUnaryNode unaryNode;
List *selectClauseList;
} MultiSelect;
@ -140,7 +133,6 @@ typedef struct MultiJoin
List *joinClauseList;
JoinRuleType joinRuleType;
JoinType joinType;
} MultiJoin;
@ -150,7 +142,6 @@ typedef struct MultiPartition
MultiUnaryNode unaryNode;
Var *partitionColumn;
uint32 splitPointTableId;
} MultiPartition;
@ -158,7 +149,6 @@ typedef struct MultiPartition
typedef struct MultiCartesianProduct
{
MultiBinaryNode binaryNode;
} MultiCartesianProduct;
@ -183,7 +173,6 @@ typedef struct MultiExtendedOp
List *sortClauseList;
Node *limitCount;
Node *limitOffset;
} MultiExtendedOp;

12
src/include/distributed/multi_physical_planner.h
View File

@ -40,7 +40,8 @@
(" UINT64_FORMAT ", %d, %s, '%s', %d, %d)"
#define MERGE_FILES_INTO_TABLE_COMMAND "SELECT worker_merge_files_into_table \
(" UINT64_FORMAT ", %d, '%s', '%s')"
#define MERGE_FILES_AND_RUN_QUERY_COMMAND "SELECT worker_merge_files_and_run_query(" UINT64_FORMAT ", %d, '%s', '%s')"
#define MERGE_FILES_AND_RUN_QUERY_COMMAND \
"SELECT worker_merge_files_and_run_query(" UINT64_FORMAT ", %d, '%s', '%s')"
typedef enum CitusRTEKind
@ -62,7 +63,6 @@ typedef enum
PARTITION_INVALID_FIRST = 0,
RANGE_PARTITION_TYPE = 1,
HASH_PARTITION_TYPE = 2
} PartitionType;
@ -77,7 +77,6 @@ typedef enum
MAP_OUTPUT_FETCH_TASK = 5,
MERGE_FETCH_TASK = 6,
MODIFY_TASK = 7
} TaskType;
@ -88,7 +87,6 @@ typedef enum
TASK_ASSIGNMENT_GREEDY = 1,
TASK_ASSIGNMENT_ROUND_ROBIN = 2,
TASK_ASSIGNMENT_FIRST_REPLICA = 3
} TaskAssignmentPolicyType;
@ -99,7 +97,6 @@ typedef enum
JOIN_MAP_MERGE_JOB = 1,
SUBQUERY_MAP_MERGE_JOB = 2,
TOP_LEVEL_WORKER_JOB = 3
} BoundaryNodeJobType;
@ -133,7 +130,6 @@ typedef struct MapMergeJob
ShardInterval **sortedShardIntervalArray; /* only applies to range partitioning */
List *mapTaskList;
List *mergeTaskList;
} MapMergeJob;
@ -164,7 +160,6 @@ typedef struct Task
uint64 shardId; /* only applies to shard fetch tasks */
TaskExecution *taskExecution; /* used by task tracker executor */
bool upsertQuery; /* only applies to modify tasks */
} Task;
@ -177,7 +172,6 @@ typedef struct RangeTableFragment
CitusRTEKind fragmentType;
void *fragmentReference;
uint32 rangeTableId;
} RangeTableFragment;
@ -190,7 +184,6 @@ typedef struct JoinSequenceNode
{
uint32 rangeTableId;
int32 joiningRangeTableId;
} JoinSequenceNode;
@ -203,7 +196,6 @@ typedef struct MultiPlan
Job *workerJob;
Query *masterQuery;
char *masterTableName;
} MultiPlan;

9
src/include/distributed/multi_server_executor.h
View File

@ -60,7 +60,6 @@ typedef enum
EXEC_TASK_TRACKER_FAILED = 14,
EXEC_SOURCE_TASK_TRACKER_RETRY = 15,
EXEC_SOURCE_TASK_TRACKER_FAILED = 16
} TaskExecStatus;
@ -74,7 +73,6 @@ typedef enum
EXEC_TRANSMIT_TRACKER_RETRY = 4,
EXEC_TRANSMIT_TRACKER_FAILED = 5,
EXEC_TRANSMIT_DONE = 6
} TransmitExecStatus;
@ -86,7 +84,6 @@ typedef enum
TRACKER_CONNECT_POLL = 2,
TRACKER_CONNECTED = 3,
TRACKER_CONNECTION_FAILED = 4
} TrackerStatus;
@ -97,7 +94,6 @@ typedef enum
MULTI_EXECUTOR_REAL_TIME = 1,
MULTI_EXECUTOR_TASK_TRACKER = 2,
MULTI_EXECUTOR_ROUTER = 3
} MultiExecutorType;
@ -107,7 +103,6 @@ typedef enum
CONNECT_ACTION_NONE = 0,
CONNECT_ACTION_OPENED = 1,
CONNECT_ACTION_CLOSED = 2
} ConnectAction;
@ -132,7 +127,6 @@ struct TaskExecution
uint32 querySourceNodeIndex; /* only applies to map fetch tasks */
int32 dataFetchTaskIndex;
uint32 failureCount;
};
@ -147,7 +141,6 @@ typedef struct TrackerTaskState
uint32 taskId;
TaskStatus status;
StringInfo taskAssignmentQuery;
} TrackerTaskState;
@ -171,7 +164,6 @@ typedef struct TaskTracker
int32 currentTaskIndex;
bool connectionBusy;
TrackerTaskState *connectionBusyOnTask;
} TaskTracker;
@ -184,7 +176,6 @@ typedef struct WorkerNodeState
uint32 workerPort;
char workerName[WORKER_LENGTH];
uint32 openConnectionCount;
} WorkerNodeState;

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

@ -35,7 +35,6 @@ typedef enum
FILE_CACHED = 2,
FILE_INACTIVE = 3,
FILE_TO_DELETE = 4
} RelayFileState;

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

@ -29,6 +29,7 @@ typedef enum AdvisoryLocktagClass
/* values defined in postgres' lockfuncs.c */
ADV_LOCKTAG_CLASS_INT64 = 1,
ADV_LOCKTAG_CLASS_INT32 = 2,
/* Citus lock types */
ADV_LOCKTAG_CLASS_CITUS_SHARD_METADATA = 4,
ADV_LOCKTAG_CLASS_CITUS_SHARD = 5,

3
src/include/distributed/task_tracker.h
View File

@ -63,7 +63,6 @@ typedef enum
* TASK_STATUS_LAST, should never have their numbers changed.
*/
TASK_STATUS_LAST
} TaskStatus;
@ -85,7 +84,6 @@ typedef struct WorkerTask
char databaseName[NAMEDATALEN]; /* name to use for local backend connection */
int32 connectionId; /* connection id to local backend */
uint32 failureCount; /* number of task failures */
} WorkerTask;
@ -97,6 +95,7 @@ typedef struct WorkerTasksSharedStateData
{
/* Hash table shared by the task tracker and task tracker protocol functions */
HTAB *taskHash;
/* Lock protecting workerNodesHash */
LWLock *taskHashLock;
} WorkerTasksSharedStateData;

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

@ -48,7 +48,6 @@ typedef struct WorkerNode
char workerRack[WORKER_LENGTH]; /* node's network location */
bool inWorkerFile; /* is node in current membership file? */
} WorkerNode;

3
src/include/distributed/worker_protocol.h
View File

@ -65,7 +65,6 @@ typedef struct RangePartitionContext
FmgrInfo *comparisonFunction;
Datum *splitPointArray;
int32 splitPointCount;
} RangePartitionContext;
@ -77,7 +76,6 @@ typedef struct HashPartitionContext
{
FmgrInfo *hashFunction;
uint32 partitionCount;
} HashPartitionContext;
@ -114,7 +112,6 @@ typedef struct FileOutputStream
File fileDescriptor;
StringInfo fileBuffer;
StringInfo filePath;
} FileOutputStream;

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

@ -171,6 +171,18 @@ from
l_tax) as distributed_table;
ERROR: cannot perform distributed planning on this query
DETAIL: Subqueries without aggregates are not supported yet
-- Check that we don't support subqueries with count(distinct).
select
different_shipment_days
from
(select
count(distinct l_shipdate) as different_shipment_days
from
lineitem
group by
l_partkey) as distributed_table;
ERROR: cannot compute count (distinct)
DETAIL: Subqueries with aggregate (distinct) are not supported yet
-- Check that if subquery is pulled, we don't error and run query properly.
SELECT max(l_suppkey) FROM
(

12
src/test/regress/sql/multi_single_relation_subquery.sql
View File

@ -125,6 +125,18 @@ from
group by
l_tax) as distributed_table;
-- Check that we don't support subqueries with count(distinct).
select
different_shipment_days
from
(select
count(distinct l_shipdate) as different_shipment_days
from
lineitem
group by
l_partkey) as distributed_table;
-- Check that if subquery is pulled, we don't error and run query properly.
SELECT max(l_suppkey) FROM