/*
* Executor state preparation for evaluation of constraint expressions,
* indexes and triggers.
*
* This is based on similar code in copy.c
*/
static EState *
create_estate_for_relation(LogicalRepRelMapEntry *rel)
{
EState *estate;
ResultRelInfo *resultRelInfo;
RangeTblEntry *rte;
estate = CreateExecutorState();
rte = makeNode(RangeTblEntry);
rte->rtekind = RTE_RELATION;
rte->relid = RelationGetRelid(rel->localrel);
rte->relkind = rel->localrel->rd_rel->relkind;
estate->es_range_table = list_make1(rte);
resultRelInfo = makeNode(ResultRelInfo);
InitResultRelInfo(resultRelInfo, rel->localrel, 1, NULL, 0);
estate->es_result_relations = resultRelInfo;
estate->es_num_result_relations = 1;
estate->es_result_relation_info = resultRelInfo;
estate->es_output_cid = GetCurrentCommandId(true);
/* Triggers might need a slot */
if (resultRelInfo->ri_TrigDesc)
estate->es_trig_tuple_slot = ExecInitExtraTupleSlot(estate, NULL);
/* Prepare to catch AFTER triggers. */
AfterTriggerBeginQuery();
return estate;
}
EState *
CreateEState(QueryDesc *query_desc)
{
EState *estate;
estate = CreateExecutorState();
estate->es_param_list_info = query_desc->params;
estate->es_snapshot = RegisterSnapshot(query_desc->snapshot);
estate->es_crosscheck_snapshot =
RegisterSnapshot(query_desc->crosscheck_snapshot);
estate->es_instrument = query_desc->instrument_options;
estate->es_range_table = query_desc->plannedstmt->rtable;
estate->es_continuous = query_desc->plannedstmt->is_continuous;
estate->es_lastoid = InvalidOid;
estate->es_processed = estate->es_filtered = 0;
if (query_desc->plannedstmt->nParamExec > 0)
estate->es_param_exec_vals = (ParamExecData *)
palloc0(query_desc->plannedstmt->nParamExec *
sizeof(ParamExecData));
estate->es_top_eflags |= EXEC_FLAG_SKIP_TRIGGERS;
return estate;
}
void
CheckerInit(Checker *checker, Relation rel, TupleChecker *tchecker)
{
TupleDesc desc;
checker->tchecker = tchecker;
/*
* When specify ENCODING, we check the input data encoding.
* Convert encoding if the client and server encodings are different.
*/
checker->db_encoding = GetDatabaseEncoding();
if (checker->encoding != -1 &&
(checker->encoding != PG_SQL_ASCII ||
checker->db_encoding != PG_SQL_ASCII))
checker->check_encoding = true;
if (!rel)
return;
/* When specify CHECK_CONSTRAINTS, we check the constraints */
desc = RelationGetDescr(rel);
if (desc->constr &&
(checker->check_constraints || desc->constr->has_not_null))
{
if (checker->check_constraints)
checker->has_constraints = true;
if (desc->constr->has_not_null)
checker->has_not_null = true;
checker->resultRelInfo = makeNode(ResultRelInfo);
checker->resultRelInfo->ri_RangeTableIndex = 1; /* dummy */
checker->resultRelInfo->ri_RelationDesc = rel;
checker->resultRelInfo->ri_TrigDesc = NULL; /* TRIGGER is not supported */
checker->resultRelInfo->ri_TrigInstrument = NULL;
}
if (checker->has_constraints)
{
checker->estate = CreateExecutorState();
checker->estate->es_result_relations = checker->resultRelInfo;
checker->estate->es_num_result_relations = 1;
checker->estate->es_result_relation_info = checker->resultRelInfo;
/* Set up a tuple slot too */
checker->slot = MakeSingleTupleTableSlot(desc);
}
if (!checker->has_constraints && checker->has_not_null)
{
int i;
checker->desc = CreateTupleDescCopy(desc);
for (i = 0; i < desc->natts; i++)
checker->desc->attrs[i]->attnotnull = desc->attrs[i]->attnotnull;
}
}
/*
* RouterExecutorStart sets up the executor state and queryDesc for router
* execution.
*/
void
RouterExecutorStart(QueryDesc *queryDesc, int eflags, Task *task)
{
LOCKMODE lockMode = NoLock;
EState *executorState = NULL;
CmdType commandType = queryDesc->operation;
/* ensure that the task is not NULL */
Assert(task != NULL);
/* disallow transactions and triggers during distributed modify commands */
if (commandType != CMD_SELECT)
{
bool topLevel = true;
PreventTransactionChain(topLevel, "distributed commands");
eflags |= EXEC_FLAG_SKIP_TRIGGERS;
}
/* signal that it is a router execution */
eflags |= EXEC_FLAG_CITUS_ROUTER_EXECUTOR;
/* build empty executor state to obtain per-query memory context */
executorState = CreateExecutorState();
executorState->es_top_eflags = eflags;
executorState->es_instrument = queryDesc->instrument_options;
queryDesc->estate = executorState;
/*
* As it's similar to what we're doing, use a MaterialState node to store
* our state. This is used to store our tuplestore, so cursors etc. can
* work.
*/
queryDesc->planstate = (PlanState *) makeNode(MaterialState);
#if (PG_VERSION_NUM < 90500)
/* make sure that upsertQuery is false for versions that UPSERT is not available */
Assert(task->upsertQuery == false);
#endif
lockMode = CommutativityRuleToLockMode(commandType, task->upsertQuery);
if (lockMode != NoLock)
{
AcquireExecutorShardLock(task, lockMode);
}
}
/*
* ExecuteSubPlans executes a list of subplans from a distributed plan
* by sequentially executing each plan from the top.
*/
void
ExecuteSubPlans(DistributedPlan *distributedPlan)
{
uint64 planId = distributedPlan->planId;
List *subPlanList = distributedPlan->subPlanList;
ListCell *subPlanCell = NULL;
List *nodeList = NIL;
bool writeLocalFile = false;
if (subPlanList == NIL)
{
/* no subplans to execute */
return;
}
/*
* Make sure that this transaction has a distributed transaction ID.
*
* Intermediate results of subplans will be stored in a directory that is
* derived from the distributed transaction ID.
*/
BeginOrContinueCoordinatedTransaction();
nodeList = ActiveReadableNodeList();
foreach(subPlanCell, subPlanList)
{
DistributedSubPlan *subPlan = (DistributedSubPlan *) lfirst(subPlanCell);
PlannedStmt *plannedStmt = subPlan->plan;
uint32 subPlanId = subPlan->subPlanId;
DestReceiver *copyDest = NULL;
ParamListInfo params = NULL;
EState *estate = NULL;
char *resultId = GenerateResultId(planId, subPlanId);
SubPlanLevel++;
estate = CreateExecutorState();
copyDest = (DestReceiver *) CreateRemoteFileDestReceiver(resultId, estate,
nodeList,
writeLocalFile);
ExecutePlanIntoDestReceiver(plannedStmt, params, copyDest);
SubPlanLevel--;
FreeExecutorState(estate);
}
void
SpoolerOpen(Spooler *self,
Relation rel,
bool use_wal,
ON_DUPLICATE on_duplicate,
int64 max_dup_errors,
const char *dup_badfile)
{
memset(self, 0, sizeof(Spooler));
self->on_duplicate = on_duplicate;
self->use_wal = use_wal;
self->max_dup_errors = max_dup_errors;
self->dup_old = 0;
self->dup_new = 0;
self->dup_badfile = pstrdup(dup_badfile);
self->dup_fp = NULL;
self->relinfo = makeNode(ResultRelInfo);
self->relinfo->ri_RangeTableIndex = 1; /* dummy */
self->relinfo->ri_RelationDesc = rel;
self->relinfo->ri_TrigDesc = NULL; /* TRIGGER is not supported */
self->relinfo->ri_TrigInstrument = NULL;
ExecOpenIndices(self->relinfo);
self->estate = CreateExecutorState();
self->estate->es_num_result_relations = 1;
self->estate->es_result_relations = self->relinfo;
self->estate->es_result_relation_info = self->relinfo;
self->slot = MakeSingleTupleTableSlot(RelationGetDescr(rel));
self->spools = IndexSpoolBegin(self->relinfo,
max_dup_errors == 0);
}
/*
* Implements the 'EXECUTE' utility statement.
*
* Note: this is one of very few places in the code that needs to deal with
* two query strings at once. The passed-in queryString is that of the
* EXECUTE, which we might need for error reporting while processing the
* parameter expressions. The query_string that we copy from the plan
* source is that of the original PREPARE.
*/
void
ExecuteQuery(ExecuteStmt *stmt, const char *queryString,
ParamListInfo params,
DestReceiver *dest, char *completionTag)
{
PreparedStatement *entry;
List *stmt_list;
MemoryContext qcontext;
ParamListInfo paramLI = NULL;
EState *estate = NULL;
Portal portal;
/* Look it up in the hash table */
entry = FetchPreparedStatement(stmt->name, true);
qcontext = entry->context;
/* Evaluate parameters, if any */
if (entry->argtype_list != NIL)
{
/*
* Need an EState to evaluate parameters; must not delete it till end
* of query, in case parameters are pass-by-reference.
*/
estate = CreateExecutorState();
estate->es_param_list_info = params;
paramLI = EvaluateParams(estate, stmt->params, entry->argtype_list);
}
/* Create a new portal to run the query in */
portal = CreateNewPortal();
/* Don't display the portal in pg_cursors, it is for internal use only */
portal->visible = false;
/* Plan the query. If this is a CTAS, copy the "into" information into
* the query so that we construct the plan correctly. Else the table
* might not be created on the segments. (MPP-8135) */
{
List *query_list = copyObject(entry->query_list); /* planner scribbles on query tree :( */
if ( stmt->into )
{
Query *query = (Query*)linitial(query_list);
Assert(IsA(query, Query) && query->intoClause == NULL);
query->intoClause = copyObject(stmt->into);
}
stmt_list = pg_plan_queries(query_list, paramLI, false, QRL_ONCE);
}
/*
* For CREATE TABLE / AS EXECUTE, make a copy of the stored query so that
* we can modify its destination (yech, but this has always been ugly).
* For regular EXECUTE we can just use the stored query where it sits,
* since the executor is read-only.
*/
if (stmt->into)
{
MemoryContext oldContext;
PlannedStmt *pstmt;
if (list_length(stmt_list) != 1)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("prepared statement is not a SELECT")));
oldContext = MemoryContextSwitchTo(PortalGetHeapMemory(portal));
stmt_list = copyObject(stmt_list);
qcontext = PortalGetHeapMemory(portal);
pstmt = (PlannedStmt *) linitial(stmt_list);
pstmt->qdContext = qcontext;
if (pstmt->commandType != CMD_SELECT)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("prepared statement is not a SELECT"),
errOmitLocation(true)));
pstmt->intoClause = copyObject(stmt->into);
/* XXX Is it legitimate to assign a constant default policy without
* even checking the relation?
*/
pstmt->intoPolicy = palloc0(sizeof(GpPolicy)- sizeof(pstmt->intoPolicy->attrs)
+ 255 * sizeof(pstmt->intoPolicy->attrs[0]));
pstmt->intoPolicy->nattrs = 0;
pstmt->intoPolicy->ptype = POLICYTYPE_PARTITIONED;
pstmt->intoPolicy->bucketnum = GetRelOpt_bucket_num_fromRangeVar(stmt->into->rel, GetRandomDistPartitionNum());
MemoryContextSwitchTo(oldContext);
}
/* Copy the plan's saved query string into the portal's memory */
Assert(entry->query_string != NULL);
char *query_string = MemoryContextStrdup(PortalGetHeapMemory(portal),
entry->query_string);
PortalDefineQuery(portal,
NULL,
query_string,
entry->sourceTag,
entry->commandTag,
stmt_list,
qcontext);
create_filesystem_credentials(portal);
/*
* Run the portal to completion.
*/
PortalStart(portal, paramLI, ActiveSnapshot,
savedSeqServerHost, savedSeqServerPort);
(void) PortalRun(portal,
FETCH_ALL,
true, /* Effectively always top level. */
dest,
dest,
completionTag);
PortalDrop(portal, false);
if (estate)
FreeExecutorState(estate);
/* No need to pfree other memory, MemoryContext will be reset */
}
/*
* unique_key_recheck - trigger function to do a deferred uniqueness check.
*
* This now also does deferred exclusion-constraint checks, so the name is
* somewhat historical.
*
* This is invoked as an AFTER ROW trigger for both INSERT and UPDATE,
* for any rows recorded as potentially violating a deferrable unique
* or exclusion constraint.
*
* This may be an end-of-statement check, a commit-time check, or a
* check triggered by a SET CONSTRAINTS command.
*/
Datum
unique_key_recheck(PG_FUNCTION_ARGS)
{
TriggerData *trigdata = (TriggerData *) fcinfo->context;
const char *funcname = "unique_key_recheck";
HeapTuple new_row;
ItemPointerData tmptid;
Relation indexRel;
IndexInfo *indexInfo;
EState *estate;
ExprContext *econtext;
TupleTableSlot *slot;
Datum values[INDEX_MAX_KEYS];
bool isnull[INDEX_MAX_KEYS];
/*
* Make sure this is being called as an AFTER ROW trigger. Note:
* translatable error strings are shared with ri_triggers.c, so resist the
* temptation to fold the function name into them.
*/
if (!CALLED_AS_TRIGGER(fcinfo))
ereport(ERROR,
(errcode(ERRCODE_E_R_I_E_TRIGGER_PROTOCOL_VIOLATED),
errmsg("function \"%s\" was not called by trigger manager",
funcname)));
if (!TRIGGER_FIRED_AFTER(trigdata->tg_event) ||
!TRIGGER_FIRED_FOR_ROW(trigdata->tg_event))
ereport(ERROR,
(errcode(ERRCODE_E_R_I_E_TRIGGER_PROTOCOL_VIOLATED),
errmsg("function \"%s\" must be fired AFTER ROW",
funcname)));
/*
* Get the new data that was inserted/updated.
*/
if (TRIGGER_FIRED_BY_INSERT(trigdata->tg_event))
new_row = trigdata->tg_trigtuple;
else if (TRIGGER_FIRED_BY_UPDATE(trigdata->tg_event))
new_row = trigdata->tg_newtuple;
else
{
ereport(ERROR,
(errcode(ERRCODE_E_R_I_E_TRIGGER_PROTOCOL_VIOLATED),
errmsg("function \"%s\" must be fired for INSERT or UPDATE",
funcname)));
new_row = NULL; /* keep compiler quiet */
}
/*
* If the new_row is now dead (ie, inserted and then deleted within our
* transaction), we can skip the check. However, we have to be careful,
* because this trigger gets queued only in response to index insertions;
* which means it does not get queued for HOT updates. The row we are
* called for might now be dead, but have a live HOT child, in which case
* we still need to make the check. Therefore we have to use
* heap_hot_search, not just HeapTupleSatisfiesVisibility as is done in
* the comparable test in RI_FKey_check.
*
* This might look like just an optimization, because the index AM will
* make this identical test before throwing an error. But it's actually
* needed for correctness, because the index AM will also throw an error
* if it doesn't find the index entry for the row. If the row's dead then
* it's possible the index entry has also been marked dead, and even
* removed.
*/
tmptid = new_row->t_self;
if (!heap_hot_search(&tmptid, trigdata->tg_relation, SnapshotSelf, NULL))
{
/*
* All rows in the HOT chain are dead, so skip the check.
*/
return PointerGetDatum(NULL);
}
/*
* Open the index, acquiring a RowExclusiveLock, just as if we were going
* to update it. (This protects against possible changes of the index
* schema, not against concurrent updates.)
*/
indexRel = index_open(trigdata->tg_trigger->tgconstrindid,
RowExclusiveLock);
indexInfo = BuildIndexInfo(indexRel);
/*
* The heap tuple must be put into a slot for FormIndexDatum.
*/
slot = MakeSingleTupleTableSlot(RelationGetDescr(trigdata->tg_relation));
ExecStoreTuple(new_row, slot, InvalidBuffer, false);
/*
* Typically the index won't have expressions, but if it does we need an
* EState to evaluate them. We need it for exclusion constraints too,
* even if they are just on simple columns.
*/
if (indexInfo->ii_Expressions != NIL ||
indexInfo->ii_ExclusionOps != NULL)
{
estate = CreateExecutorState();
econtext = GetPerTupleExprContext(estate);
econtext->ecxt_scantuple = slot;
}
else
estate = NULL;
/*
* Form the index values and isnull flags for the index entry that we need
* to check.
*
* Note: if the index uses functions that are not as immutable as they are
* supposed to be, this could produce an index tuple different from the
* original. The index AM can catch such errors by verifying that it
* finds a matching index entry with the tuple's TID. For exclusion
* constraints we check this in check_exclusion_constraint().
*/
FormIndexDatum(indexInfo, slot, estate, values, isnull);
/*
* Now do the appropriate check.
*/
if (indexInfo->ii_ExclusionOps == NULL)
{
/*
* Note: this is not a real insert; it is a check that the index entry
* that has already been inserted is unique.
*/
index_insert(indexRel, values, isnull, &(new_row->t_self),
trigdata->tg_relation, UNIQUE_CHECK_EXISTING);
}
else
{
/*
* For exclusion constraints we just do the normal check, but now it's
* okay to throw error.
*/
check_exclusion_constraint(trigdata->tg_relation, indexRel, indexInfo,
&(new_row->t_self), values, isnull,
estate, false, false);
}
/*
* If that worked, then this index entry is unique or non-excluded, and we
* are done.
*/
if (estate != NULL)
FreeExecutorState(estate);
ExecDropSingleTupleTableSlot(slot);
index_close(indexRel, RowExclusiveLock);
return PointerGetDatum(NULL);
}
/*
* IndexBuildHeapScan - scan the heap relation to find tuples to be indexed
*
* This is called back from an access-method-specific index build procedure
* after the AM has done whatever setup it needs. The parent heap relation
* is scanned to find tuples that should be entered into the index. Each
* such tuple is passed to the AM's callback routine, which does the right
* things to add it to the new index. After we return, the AM's index
* build procedure does whatever cleanup is needed; in particular, it should
* close the heap and index relations.
*
* The total count of heap tuples is returned. This is for updating pg_class
* statistics. (It's annoying not to be able to do that here, but we can't
* do it until after the relation is closed.) Note that the index AM itself
* must keep track of the number of index tuples; we don't do so here because
* the AM might reject some of the tuples for its own reasons, such as being
* unable to store NULLs.
*/
double
IndexBuildHeapScan(Relation heapRelation,
Relation indexRelation,
IndexInfo *indexInfo,
IndexBuildCallback callback,
void *callback_state)
{
HeapScanDesc scan;
HeapTuple heapTuple;
TupleDesc heapDescriptor;
Datum attdata[INDEX_MAX_KEYS];
char nulls[INDEX_MAX_KEYS];
double reltuples;
List *predicate;
TupleTable tupleTable;
TupleTableSlot *slot;
EState *estate;
ExprContext *econtext;
Snapshot snapshot;
TransactionId OldestXmin;
/*
* sanity checks
*/
Assert(OidIsValid(indexRelation->rd_rel->relam));
heapDescriptor = RelationGetDescr(heapRelation);
/*
* Need an EState for evaluation of index expressions and
* partial-index predicates.
*/
estate = CreateExecutorState();
econtext = GetPerTupleExprContext(estate);
/*
* If this is a predicate (partial) index, we will need to evaluate
* the predicate using ExecQual, which requires the current tuple to
* be in a slot of a TupleTable. Likewise if there are any
* expressions.
*/
if (indexInfo->ii_Predicate != NIL || indexInfo->ii_Expressions != NIL)
{
tupleTable = ExecCreateTupleTable(1);
slot = ExecAllocTableSlot(tupleTable);
ExecSetSlotDescriptor(slot, heapDescriptor, false);
/* Arrange for econtext's scan tuple to be the tuple under test */
econtext->ecxt_scantuple = slot;
/* Set up execution state for predicate. */
predicate = (List *)
ExecPrepareExpr((Expr *) indexInfo->ii_Predicate,
estate);
}
else
{
tupleTable = NULL;
slot = NULL;
predicate = NIL;
}
/*
* Ok, begin our scan of the base relation. We use SnapshotAny
* because we must retrieve all tuples and do our own time qual
* checks.
*/
if (IsBootstrapProcessingMode())
{
snapshot = SnapshotNow;
OldestXmin = InvalidTransactionId;
}
else
{
snapshot = SnapshotAny;
OldestXmin = GetOldestXmin(heapRelation->rd_rel->relisshared);
}
scan = heap_beginscan(heapRelation, /* relation */
snapshot, /* seeself */
0, /* number of keys */
(ScanKey) NULL); /* scan key */
reltuples = 0;
/*
* Scan all tuples in the base relation.
*/
while ((heapTuple = heap_getnext(scan, ForwardScanDirection)) != NULL)
{
bool tupleIsAlive;
CHECK_FOR_INTERRUPTS();
if (snapshot == SnapshotAny)
{
/* do our own time qual check */
bool indexIt;
uint16 sv_infomask;
/*
* HeapTupleSatisfiesVacuum may update tuple's hint status
* bits. We could possibly get away with not locking the
* buffer here, since caller should hold ShareLock on the
* relation, but let's be conservative about it.
*/
LockBuffer(scan->rs_cbuf, BUFFER_LOCK_SHARE);
sv_infomask = heapTuple->t_data->t_infomask;
switch (HeapTupleSatisfiesVacuum(heapTuple->t_data, OldestXmin))
{
case HEAPTUPLE_DEAD:
indexIt = false;
tupleIsAlive = false;
break;
case HEAPTUPLE_LIVE:
indexIt = true;
tupleIsAlive = true;
break;
case HEAPTUPLE_RECENTLY_DEAD:
/*
* If tuple is recently deleted then we must index it
* anyway to keep VACUUM from complaining.
*/
indexIt = true;
tupleIsAlive = false;
break;
case HEAPTUPLE_INSERT_IN_PROGRESS:
/*
* Since caller should hold ShareLock or better, we
* should not see any tuples inserted by open
* transactions --- unless it's our own transaction.
* (Consider INSERT followed by CREATE INDEX within a
* transaction.) An exception occurs when reindexing
* a system catalog, because we often release lock on
* system catalogs before committing.
*/
if (!TransactionIdIsCurrentTransactionId(
HeapTupleHeaderGetXmin(heapTuple->t_data))
&& !IsSystemRelation(heapRelation))
elog(ERROR, "concurrent insert in progress");
indexIt = true;
tupleIsAlive = true;
break;
case HEAPTUPLE_DELETE_IN_PROGRESS:
/*
* Since caller should hold ShareLock or better, we
* should not see any tuples deleted by open
* transactions --- unless it's our own transaction.
* (Consider DELETE followed by CREATE INDEX within a
* transaction.) An exception occurs when reindexing
* a system catalog, because we often release lock on
* system catalogs before committing.
*/
if (!TransactionIdIsCurrentTransactionId(
HeapTupleHeaderGetXmax(heapTuple->t_data))
&& !IsSystemRelation(heapRelation))
elog(ERROR, "concurrent delete in progress");
indexIt = true;
tupleIsAlive = false;
break;
default:
elog(ERROR, "unexpected HeapTupleSatisfiesVacuum result");
indexIt = tupleIsAlive = false; /* keep compiler quiet */
break;
}
/* check for hint-bit update by HeapTupleSatisfiesVacuum */
if (sv_infomask != heapTuple->t_data->t_infomask)
SetBufferCommitInfoNeedsSave(scan->rs_cbuf);
LockBuffer(scan->rs_cbuf, BUFFER_LOCK_UNLOCK);
if (!indexIt)
continue;
}
else
{
/* heap_getnext did the time qual check */
tupleIsAlive = true;
}
reltuples += 1;
MemoryContextReset(econtext->ecxt_per_tuple_memory);
/* Set up for predicate or expression evaluation */
if (slot)
ExecStoreTuple(heapTuple, slot, InvalidBuffer, false);
/*
* In a partial index, discard tuples that don't satisfy the
* predicate. We can also discard recently-dead tuples, since
* VACUUM doesn't complain about tuple count mismatch for partial
* indexes.
*/
if (predicate != NIL)
{
if (!tupleIsAlive)
continue;
if (!ExecQual(predicate, econtext, false))
continue;
}
/*
* For the current heap tuple, extract all the attributes we use
* in this index, and note which are null. This also performs
* evaluation of any expressions needed.
*/
FormIndexDatum(indexInfo,
heapTuple,
heapDescriptor,
estate,
attdata,
nulls);
/*
* You'd think we should go ahead and build the index tuple here,
* but some index AMs want to do further processing on the data
* first. So pass the attdata and nulls arrays, instead.
*/
/* Call the AM's callback routine to process the tuple */
callback(indexRelation, heapTuple, attdata, nulls, tupleIsAlive,
callback_state);
}
heap_endscan(scan);
if (tupleTable)
ExecDropTupleTable(tupleTable, true);
FreeExecutorState(estate);
/* These may have been pointing to the now-gone estate */
indexInfo->ii_ExpressionsState = NIL;
indexInfo->ii_PredicateState = NIL;
return reltuples;
}
Datum
plpgsql_inline_handler(PG_FUNCTION_ARGS)
{
InlineCodeBlock *codeblock = (InlineCodeBlock *) DatumGetPointer(PG_GETARG_DATUM(0));
PLpgSQL_function *func;
FunctionCallInfoData fake_fcinfo;
FmgrInfo flinfo;
EState *simple_eval_estate;
Datum retval;
int rc;
Assert(IsA(codeblock, InlineCodeBlock));
/*
* Connect to SPI manager
*/
if ((rc = SPI_connect()) != SPI_OK_CONNECT)
elog(ERROR, "SPI_connect failed: %s", SPI_result_code_string(rc));
/* Compile the anonymous code block */
func = plpgsql_compile_inline(codeblock->source_text);
/* Mark the function as busy, just pro forma */
func->use_count++;
/*
* Set up a fake fcinfo with just enough info to satisfy
* plpgsql_exec_function(). In particular note that this sets things up
* with no arguments passed.
*/
MemSet(&fake_fcinfo, 0, sizeof(fake_fcinfo));
MemSet(&flinfo, 0, sizeof(flinfo));
fake_fcinfo.flinfo = &flinfo;
flinfo.fn_oid = InvalidOid;
flinfo.fn_mcxt = CurrentMemoryContext;
/* Create a private EState for simple-expression execution */
simple_eval_estate = CreateExecutorState();
/* And run the function */
PG_TRY();
{
retval = plpgsql_exec_function(func, &fake_fcinfo, simple_eval_estate);
}
PG_CATCH();
{
/*
* We need to clean up what would otherwise be long-lived resources
* accumulated by the failed DO block, principally cached plans for
* statements (which can be flushed with plpgsql_free_function_memory)
* and execution trees for simple expressions, which are in the
* private EState.
*
* Before releasing the private EState, we must clean up any
* simple_econtext_stack entries pointing into it, which we can do by
* invoking the subxact callback. (It will be called again later if
* some outer control level does a subtransaction abort, but no harm
* is done.) We cheat a bit knowing that plpgsql_subxact_cb does not
* pay attention to its parentSubid argument.
*/
plpgsql_subxact_cb(SUBXACT_EVENT_ABORT_SUB,
GetCurrentSubTransactionId(),
0, NULL);
/* Clean up the private EState */
FreeExecutorState(simple_eval_estate);
/* Function should now have no remaining use-counts ... */
func->use_count--;
Assert(func->use_count == 0);
/* ... so we can free subsidiary storage */
plpgsql_free_function_memory(func);
/* And propagate the error */
PG_RE_THROW();
}
PG_END_TRY();
/* Clean up the private EState */
FreeExecutorState(simple_eval_estate);
/* Function should now have no remaining use-counts ... */
func->use_count--;
Assert(func->use_count == 0);
/* ... so we can free subsidiary storage */
plpgsql_free_function_memory(func);
/*
* Disconnect from SPI manager
*/
if ((rc = SPI_finish()) != SPI_OK_FINISH)
elog(ERROR, "SPI_finish failed: %s", SPI_result_code_string(rc));
return retval;
}
Datum partition_insert_trigger(PG_FUNCTION_ARGS)
{
TriggerData *trigdata = (TriggerData *) fcinfo->context;
char *date_time;
char child_table[sizeof(TABLE)+sizeof("2012_09_10")] = TABLE;
int partition_field;
Relation child_table_id;
Oid child_table_oid;
BulkInsertState bistate = GetBulkInsertState();
TupleTableSlot *slot;
EState *estate = CreateExecutorState();
ResultRelInfo *resultRelInfo = makeNode(ResultRelInfo);
List *recheckIndexes = NIL;
/* make sure it's called as a trigger at all */
if (!CALLED_AS_TRIGGER(fcinfo))
elog(ERROR, "partition_insert_trigger: not called by trigger manager");
/* Sanity checks */
if (!TRIGGER_FIRED_BY_INSERT(trigdata->tg_event) || !TRIGGER_FIRED_BEFORE(trigdata->tg_event))
elog(ERROR, "partition_insert_trigger: not called on insert before");
#ifdef DEBUG
elog(INFO, "Trigger Called for: %s", SPI_getrelname(trigdata->tg_relation));
#endif
// Get the field number for the partition
partition_field = SPI_fnumber(trigdata->tg_relation->rd_att, PARTITION_COLUMN);
// Get the value for the partition_field
date_time = SPI_getvalue(trigdata->tg_trigtuple, trigdata->tg_relation->rd_att, partition_field);
//make sure date was specified
if (!date_time)
elog(ERROR, "You cannot insert data without specifying a value for the column %s", PARTITION_COLUMN);
#ifdef DEBUG
elog(INFO, "Trying to insert date_time=%s", date_time);
#endif
//add date_time, child_table_2012_01_23
strncpy(child_table + sizeof(TABLE) -1 , date_time, 4); //2012
child_table[sizeof(TABLE) + 3] = SEPARATOR; //2012_
strncpy(child_table + sizeof(TABLE) + 4 , date_time + 5, 2); //2012_01
child_table[sizeof(TABLE) + 6] = SEPARATOR; //2012_01_
strncpy(child_table + sizeof(TABLE) + 7 , date_time + 8, 2); //2012_01_23
#ifdef DEBUG
elog(INFO, "New table will be %s", child_table);
#endif
pfree(date_time);
//if you care about triggers on the child tables, call ExecBRInsertTriggers
//don't care, continue
//get the OID of the table we are looking for to insert
child_table_oid = RelnameGetRelid(child_table); //Look for child child_table
if (child_table_oid == InvalidOid){
elog(INFO, "partition_insert_trigger: Invalid child table %s, inserting data to main table %s", child_table, TABLE);
return PointerGetDatum(trigdata->tg_trigtuple);
}
//get the descriptor of the table we are looking for
child_table_id = RelationIdGetRelation(child_table_oid); //Get the child relation descriptor
if (child_table_id == NULL){
elog(ERROR, "partition_insert_trigger: Failed to locate relation for child table %s, inserting data to main table %s", child_table, TABLE);
return PointerGetDatum(trigdata->tg_trigtuple);
}
//set the resultRelInfo
resultRelInfo->ri_RangeTableIndex = 1; /* dummy */
resultRelInfo->ri_RelationDesc = child_table_id;
//setup the estate, not sure why
estate->es_result_relations = resultRelInfo;
estate->es_num_result_relations = 1;
estate->es_result_relation_info = resultRelInfo;
/* Set up a tuple slot not sure why yet */
slot = MakeSingleTupleTableSlot(trigdata->tg_relation->rd_att);
ExecStoreTuple(trigdata->tg_trigtuple, slot, InvalidBuffer, false);
//heap_insert(child_table_id, trigdata->tg_trigtuple, GetCurrentCommandId(true), use_wal, bistate);
simple_heap_insert(child_table_id, trigdata->tg_trigtuple);
if (resultRelInfo->ri_NumIndices > 0)
recheckIndexes = ExecInsertIndexTuples(slot, &(trigdata->tg_trigtuple->t_self), estate);
// not sure if this would work CatalogUpdateIndexes(child_table_id, trigdata->tg_trigtuple)
//free the used memory
list_free(recheckIndexes);
ExecDropSingleTupleTableSlot(slot); //saw this somewhere :P
RelationClose(child_table_id); //Must be called to free the relation memory page
FreeBulkInsertState(bistate); // ? not sure if needed ?
//If return next line will add to the regular table
//return PointerGetDatum(trigdata->tg_trigtuple);
//Return Null data, still have to figure out to return a proper X rows affected
return PointerGetDatum(NULL);
}
/*
* CitusCopyFrom implements the COPY table_name FROM ... for hash-partitioned
* and range-partitioned tables.
*/
void
CitusCopyFrom(CopyStmt *copyStatement, char *completionTag)
{
Oid tableId = RangeVarGetRelid(copyStatement->relation, NoLock, false);
char *relationName = get_rel_name(tableId);
Relation distributedRelation = NULL;
char partitionMethod = '\0';
Var *partitionColumn = NULL;
TupleDesc tupleDescriptor = NULL;
uint32 columnCount = 0;
Datum *columnValues = NULL;
bool *columnNulls = NULL;
TypeCacheEntry *typeEntry = NULL;
FmgrInfo *hashFunction = NULL;
FmgrInfo *compareFunction = NULL;
int shardCount = 0;
List *shardIntervalList = NULL;
ShardInterval **shardIntervalCache = NULL;
bool useBinarySearch = false;
HTAB *shardConnectionHash = NULL;
ShardConnections *shardConnections = NULL;
List *connectionList = NIL;
EState *executorState = NULL;
MemoryContext executorTupleContext = NULL;
ExprContext *executorExpressionContext = NULL;
CopyState copyState = NULL;
CopyOutState copyOutState = NULL;
FmgrInfo *columnOutputFunctions = NULL;
uint64 processedRowCount = 0;
/* disallow COPY to/from file or program except for superusers */
if (copyStatement->filename != NULL && !superuser())
{
if (copyStatement->is_program)
{
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser to COPY to or from an external program"),
errhint("Anyone can COPY to stdout or from stdin. "
"psql's \\copy command also works for anyone.")));
}
else
{
ereport(ERROR,
(errcode(ERRCODE_INSUFFICIENT_PRIVILEGE),
errmsg("must be superuser to COPY to or from a file"),
errhint("Anyone can COPY to stdout or from stdin. "
"psql's \\copy command also works for anyone.")));
}
}
partitionColumn = PartitionColumn(tableId, 0);
partitionMethod = PartitionMethod(tableId);
if (partitionMethod != DISTRIBUTE_BY_RANGE && partitionMethod != DISTRIBUTE_BY_HASH)
{
ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("COPY is only supported for hash- and "
"range-partitioned tables")));
}
/* resolve hash function for partition column */
typeEntry = lookup_type_cache(partitionColumn->vartype, TYPECACHE_HASH_PROC_FINFO);
hashFunction = &(typeEntry->hash_proc_finfo);
/* resolve compare function for shard intervals */
compareFunction = ShardIntervalCompareFunction(partitionColumn, partitionMethod);
/* allocate column values and nulls arrays */
distributedRelation = heap_open(tableId, RowExclusiveLock);
tupleDescriptor = RelationGetDescr(distributedRelation);
columnCount = tupleDescriptor->natts;
columnValues = palloc0(columnCount * sizeof(Datum));
columnNulls = palloc0(columnCount * sizeof(bool));
/* load the list of shards and verify that we have shards to copy into */
shardIntervalList = LoadShardIntervalList(tableId);
if (shardIntervalList == NIL)
{
if (partitionMethod == DISTRIBUTE_BY_HASH)
{
ereport(ERROR, (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("could not find any shards into which to copy"),
errdetail("No shards exist for distributed table \"%s\".",
relationName),
errhint("Run master_create_worker_shards to create shards "
"and try again.")));
}
else
{
ereport(ERROR, (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("could not find any shards into which to copy"),
errdetail("No shards exist for distributed table \"%s\".",
relationName)));
}
}
/* prevent concurrent placement changes and non-commutative DML statements */
LockAllShards(shardIntervalList);
/* initialize the shard interval cache */
shardCount = list_length(shardIntervalList);
shardIntervalCache = SortedShardIntervalArray(shardIntervalList);
/* determine whether to use binary search */
if (partitionMethod != DISTRIBUTE_BY_HASH ||
!IsUniformHashDistribution(shardIntervalCache, shardCount))
{
useBinarySearch = true;
}
/* initialize copy state to read from COPY data source */
copyState = BeginCopyFrom(distributedRelation,
copyStatement->filename,
copyStatement->is_program,
copyStatement->attlist,
copyStatement->options);
executorState = CreateExecutorState();
executorTupleContext = GetPerTupleMemoryContext(executorState);
executorExpressionContext = GetPerTupleExprContext(executorState);
copyOutState = (CopyOutState) palloc0(sizeof(CopyOutStateData));
copyOutState->binary = true;
copyOutState->fe_msgbuf = makeStringInfo();
copyOutState->rowcontext = executorTupleContext;
columnOutputFunctions = ColumnOutputFunctions(tupleDescriptor, copyOutState->binary);
/*
* Create a mapping of shard id to a connection for each of its placements.
* The hash should be initialized before the PG_TRY, since it is used and
* PG_CATCH. Otherwise, it may be undefined in the PG_CATCH (see sigsetjmp
* documentation).
*/
shardConnectionHash = CreateShardConnectionHash();
/* we use a PG_TRY block to roll back on errors (e.g. in NextCopyFrom) */
PG_TRY();
{
ErrorContextCallback errorCallback;
/* set up callback to identify error line number */
errorCallback.callback = CopyFromErrorCallback;
errorCallback.arg = (void *) copyState;
errorCallback.previous = error_context_stack;
error_context_stack = &errorCallback;
/* ensure transactions have unique names on worker nodes */
InitializeDistributedTransaction();
while (true)
{
bool nextRowFound = false;
Datum partitionColumnValue = 0;
ShardInterval *shardInterval = NULL;
int64 shardId = 0;
bool shardConnectionsFound = false;
MemoryContext oldContext = NULL;
ResetPerTupleExprContext(executorState);
oldContext = MemoryContextSwitchTo(executorTupleContext);
/* parse a row from the input */
nextRowFound = NextCopyFrom(copyState, executorExpressionContext,
columnValues, columnNulls, NULL);
if (!nextRowFound)
{
MemoryContextSwitchTo(oldContext);
break;
}
CHECK_FOR_INTERRUPTS();
/* find the partition column value */
if (columnNulls[partitionColumn->varattno - 1])
{
ereport(ERROR, (errcode(ERRCODE_NULL_VALUE_NOT_ALLOWED),
errmsg("cannot copy row with NULL value "
"in partition column")));
}
partitionColumnValue = columnValues[partitionColumn->varattno - 1];
/* find the shard interval and id for the partition column value */
shardInterval = FindShardInterval(partitionColumnValue, shardIntervalCache,
shardCount, partitionMethod,
compareFunction, hashFunction,
useBinarySearch);
if (shardInterval == NULL)
{
ereport(ERROR, (errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("could not find shard for partition column "
"value")));
}
shardId = shardInterval->shardId;
MemoryContextSwitchTo(oldContext);
/* get existing connections to the shard placements, if any */
shardConnections = GetShardConnections(shardConnectionHash,
shardId,
&shardConnectionsFound);
if (!shardConnectionsFound)
{
/* open connections and initiate COPY on shard placements */
OpenCopyTransactions(copyStatement, shardConnections);
/* send binary headers to shard placements */
resetStringInfo(copyOutState->fe_msgbuf);
AppendCopyBinaryHeaders(copyOutState);
SendCopyDataToAll(copyOutState->fe_msgbuf,
shardConnections->connectionList);
}
/* replicate row to shard placements */
resetStringInfo(copyOutState->fe_msgbuf);
AppendCopyRowData(columnValues, columnNulls, tupleDescriptor,
copyOutState, columnOutputFunctions);
SendCopyDataToAll(copyOutState->fe_msgbuf, shardConnections->connectionList);
processedRowCount += 1;
}
connectionList = ConnectionList(shardConnectionHash);
/* send binary footers to all shard placements */
resetStringInfo(copyOutState->fe_msgbuf);
AppendCopyBinaryFooters(copyOutState);
SendCopyDataToAll(copyOutState->fe_msgbuf, connectionList);
/* all lines have been copied, stop showing line number in errors */
error_context_stack = errorCallback.previous;
/* close the COPY input on all shard placements */
EndRemoteCopy(connectionList, true);
if (CopyTransactionManager == TRANSACTION_MANAGER_2PC)
{
PrepareRemoteTransactions(connectionList);
}
EndCopyFrom(copyState);
heap_close(distributedRelation, NoLock);
/* check for cancellation one last time before committing */
CHECK_FOR_INTERRUPTS();
}
PG_CATCH();
{
List *abortConnectionList = NIL;
/* roll back all transactions */
abortConnectionList = ConnectionList(shardConnectionHash);
EndRemoteCopy(abortConnectionList, false);
AbortRemoteTransactions(abortConnectionList);
CloseConnections(abortConnectionList);
PG_RE_THROW();
}
PG_END_TRY();
/*
* Ready to commit the transaction, this code is below the PG_TRY block because
* we do not want any of the transactions rolled back if a failure occurs. Instead,
* they should be rolled forward.
*/
CommitRemoteTransactions(connectionList);
CloseConnections(connectionList);
if (completionTag != NULL)
{
snprintf(completionTag, COMPLETION_TAG_BUFSIZE,
"COPY " UINT64_FORMAT, processedRowCount);
}
}
/*
* Assumes that the segment file lock is already held.
* Assumes that the segment file should be compacted.
*/
static bool
AOCSSegmentFileFullCompaction(Relation aorel,
AOCSInsertDesc insertDesc,
AOCSFileSegInfo *fsinfo,
Snapshot snapshot)
{
const char *relname;
AppendOnlyVisimap visiMap;
AOCSScanDesc scanDesc;
TupleDesc tupDesc;
TupleTableSlot *slot;
int compact_segno;
int64 movedTupleCount = 0;
ResultRelInfo *resultRelInfo;
MemTupleBinding *mt_bind;
EState *estate;
bool *proj;
int i;
AOTupleId *aoTupleId;
int64 tupleCount = 0;
int64 tuplePerPage = INT_MAX;
Assert(Gp_role == GP_ROLE_EXECUTE || Gp_role == GP_ROLE_UTILITY);
Assert(RelationIsAoCols(aorel));
Assert(insertDesc);
compact_segno = fsinfo->segno;
if (fsinfo->varblockcount > 0)
{
tuplePerPage = fsinfo->total_tupcount / fsinfo->varblockcount;
}
relname = RelationGetRelationName(aorel);
AppendOnlyVisimap_Init(&visiMap,
aorel->rd_appendonly->visimaprelid,
aorel->rd_appendonly->visimapidxid,
ShareLock,
snapshot);
elogif(Debug_appendonly_print_compaction,
LOG, "Compact AO segfile %d, relation %sd",
compact_segno, relname);
proj = palloc0(sizeof(bool) * RelationGetNumberOfAttributes(aorel));
for (i = 0; i < RelationGetNumberOfAttributes(aorel); ++i)
{
proj[i] = true;
}
scanDesc = aocs_beginrangescan(aorel,
snapshot, snapshot,
&compact_segno, 1, NULL, proj);
tupDesc = RelationGetDescr(aorel);
slot = MakeSingleTupleTableSlot(tupDesc);
mt_bind = create_memtuple_binding(tupDesc);
/*
* We need a ResultRelInfo and an EState so we can use the regular
* executor's index-entry-making machinery.
*/
estate = CreateExecutorState();
resultRelInfo = makeNode(ResultRelInfo);
resultRelInfo->ri_RangeTableIndex = 1; /* dummy */
resultRelInfo->ri_RelationDesc = aorel;
resultRelInfo->ri_TrigDesc = NULL; /* we don't fire triggers */
ExecOpenIndices(resultRelInfo);
estate->es_result_relations = resultRelInfo;
estate->es_num_result_relations = 1;
estate->es_result_relation_info = resultRelInfo;
while (aocs_getnext(scanDesc, ForwardScanDirection, slot))
{
CHECK_FOR_INTERRUPTS();
aoTupleId = (AOTupleId *) slot_get_ctid(slot);
if (AppendOnlyVisimap_IsVisible(&scanDesc->visibilityMap, aoTupleId))
{
AOCSMoveTuple(slot,
insertDesc,
resultRelInfo,
estate);
movedTupleCount++;
}
else
{
/* Tuple is invisible and needs to be dropped */
AppendOnlyThrowAwayTuple(aorel,
slot,
mt_bind);
}
/*
* Check for vacuum delay point after approximatly a var block
*/
tupleCount++;
if (VacuumCostActive && tupleCount % tuplePerPage == 0)
{
vacuum_delay_point();
}
}
SetAOCSFileSegInfoState(aorel, compact_segno,
AOSEG_STATE_AWAITING_DROP);
AppendOnlyVisimap_DeleteSegmentFile(&visiMap,
compact_segno);
/* Delete all mini pages of the segment files if block directory exists */
if (OidIsValid(aorel->rd_appendonly->blkdirrelid))
{
AppendOnlyBlockDirectory_DeleteSegmentFile(aorel,
snapshot,
compact_segno,
0);
}
elogif(Debug_appendonly_print_compaction, LOG,
"Finished compaction: "
"AO segfile %d, relation %s, moved tuple count " INT64_FORMAT,
compact_segno, relname, movedTupleCount);
AppendOnlyVisimap_Finish(&visiMap, NoLock);
ExecCloseIndices(resultRelInfo);
FreeExecutorState(estate);
ExecDropSingleTupleTableSlot(slot);
destroy_memtuple_binding(mt_bind);
aocs_endscan(scanDesc);
pfree(proj);
return true;
}
/*
* Assumes that the segment file lock is already held.
* Assumes that the segment file should be compacted.
*
*/
static void
AppendOnlySegmentFileFullCompaction(Relation aorel,
AppendOnlyEntry *aoEntry,
AppendOnlyInsertDesc insertDesc,
FileSegInfo* fsinfo)
{
const char* relname;
AppendOnlyVisimap visiMap;
AppendOnlyScanDesc scanDesc;
TupleDesc tupDesc;
MemTuple tuple;
TupleTableSlot *slot;
MemTupleBinding *mt_bind;
int compact_segno;
int64 movedTupleCount = 0;
ResultRelInfo *resultRelInfo;
EState *estate;
AOTupleId *aoTupleId;
int64 tupleCount = 0;
int64 tuplePerPage = INT_MAX;
Assert(Gp_role == GP_ROLE_EXECUTE || Gp_role == GP_ROLE_UTILITY);
Assert(RelationIsAoRows(aorel));
Assert(insertDesc);
compact_segno = fsinfo->segno;
if (fsinfo->varblockcount > 0)
{
tuplePerPage = fsinfo->total_tupcount / fsinfo->varblockcount;
}
relname = RelationGetRelationName(aorel);
AppendOnlyVisimap_Init(&visiMap,
aoEntry->visimaprelid,
aoEntry->visimapidxid,
ShareUpdateExclusiveLock,
SnapshotNow);
elogif(Debug_appendonly_print_compaction,
LOG, "Compact AO segno %d, relation %s, insert segno %d",
compact_segno, relname, insertDesc->storageWrite.segmentFileNum);
/*
* Todo: We need to limit the scan to one file and we need to avoid to
* lock the file again.
*
* We use SnapshotAny to get visible and invisible tuples.
*/
scanDesc = appendonly_beginrangescan(aorel,
SnapshotAny, SnapshotNow,
&compact_segno, 1, 0, NULL);
tupDesc = RelationGetDescr(aorel);
slot = MakeSingleTupleTableSlot(tupDesc);
mt_bind = create_memtuple_binding(tupDesc);
/*
* We need a ResultRelInfo and an EState so we can use the regular
* executor's index-entry-making machinery.
*/
estate = CreateExecutorState();
resultRelInfo = makeNode(ResultRelInfo);
resultRelInfo->ri_RangeTableIndex = 1; /* dummy */
resultRelInfo->ri_RelationDesc = aorel;
resultRelInfo->ri_TrigDesc = NULL; /* we don't fire triggers */
ExecOpenIndices(resultRelInfo);
estate->es_result_relations = resultRelInfo;
estate->es_num_result_relations = 1;
estate->es_result_relation_info = resultRelInfo;
/*
* Go through all visible tuples and move them to a new segfile.
*/
while ((tuple = appendonly_getnext(scanDesc, ForwardScanDirection, slot)) != NULL)
{
/* Check interrupts as this may take time. */
CHECK_FOR_INTERRUPTS();
aoTupleId = (AOTupleId*)slot_get_ctid(slot);
if (AppendOnlyVisimap_IsVisible(&scanDesc->visibilityMap, aoTupleId))
{
AppendOnlyMoveTuple(tuple,
slot,
mt_bind,
insertDesc,
resultRelInfo,
estate);
movedTupleCount++;
}
else
{
/* Tuple is invisible and needs to be dropped */
AppendOnlyThrowAwayTuple(aorel,
tuple,
slot,
mt_bind);
}
/*
* Check for vacuum delay point after approximatly a var block
*/
tupleCount++;
if (VacuumCostActive && tupleCount % tuplePerPage == 0)
{
vacuum_delay_point();
}
}
SetFileSegInfoState(aorel, aoEntry, compact_segno, AOSEG_STATE_AWAITING_DROP);
AppendOnlyVisimap_DeleteSegmentFile(&visiMap, compact_segno);
/* Delete all mini pages of the segment files if block directory exists */
if (OidIsValid(aoEntry->blkdirrelid))
{
AppendOnlyBlockDirectory_DeleteSegmentFile(
aoEntry,
SnapshotNow,
compact_segno,
0);
}
elogif(Debug_appendonly_print_compaction, LOG,
"Finished compaction: "
"AO segfile %d, relation %s, moved tuple count " INT64_FORMAT,
compact_segno, relname, movedTupleCount);
AppendOnlyVisimap_Finish(&visiMap, NoLock);
ExecCloseIndices(resultRelInfo);
FreeExecutorState(estate);
ExecDropSingleTupleTableSlot(slot);
destroy_memtuple_binding(mt_bind);
appendonly_endscan(scanDesc);
}
/* ----------------------------------------------------------------
* ExecInitSubqueryScan
* ----------------------------------------------------------------
*/
SubqueryScanState *
ExecInitSubqueryScan(SubqueryScan *node, EState *estate)
{
SubqueryScanState *subquerystate;
RangeTblEntry *rte;
EState *sp_estate;
MemoryContext oldcontext;
/*
* SubqueryScan should not have any "normal" children.
*/
Assert(outerPlan(node) == NULL);
Assert(innerPlan(node) == NULL);
/*
* create state structure
*/
subquerystate = makeNode(SubqueryScanState);
subquerystate->ss.ps.plan = (Plan *) node;
subquerystate->ss.ps.state = estate;
/*
* Miscellaneous initialization
*
* create expression context for node
*/
ExecAssignExprContext(estate, &subquerystate->ss.ps);
/*
* initialize child expressions
*/
subquerystate->ss.ps.targetlist = (List *)
ExecInitExpr((Expr *) node->scan.plan.targetlist,
(PlanState *) subquerystate);
subquerystate->ss.ps.qual = (List *)
ExecInitExpr((Expr *) node->scan.plan.qual,
(PlanState *) subquerystate);
#define SUBQUERYSCAN_NSLOTS 2
/*
* tuple table initialization
*/
ExecInitResultTupleSlot(estate, &subquerystate->ss.ps);
ExecInitScanTupleSlot(estate, &subquerystate->ss);
/*
* initialize subquery
*
* This should agree with ExecInitSubPlan
*/
rte = rt_fetch(node->scan.scanrelid, estate->es_range_table);
Assert(rte->rtekind == RTE_SUBQUERY);
/*
* Do access checking on the rangetable entries in the subquery.
*/
ExecCheckRTPerms(rte->subquery->rtable);
/*
* The subquery needs its own EState because it has its own rangetable. It
* shares our Param ID space, however. XXX if rangetable access were done
* differently, the subquery could share our EState, which would eliminate
* some thrashing about in this module...
*/
sp_estate = CreateExecutorState();
subquerystate->sss_SubEState = sp_estate;
oldcontext = MemoryContextSwitchTo(sp_estate->es_query_cxt);
sp_estate->es_range_table = rte->subquery->rtable;
sp_estate->es_param_list_info = estate->es_param_list_info;
sp_estate->es_param_exec_vals = estate->es_param_exec_vals;
sp_estate->es_tupleTable =
ExecCreateTupleTable(ExecCountSlotsNode(node->subplan) + 10);
sp_estate->es_snapshot = estate->es_snapshot;
sp_estate->es_crosscheck_snapshot = estate->es_crosscheck_snapshot;
sp_estate->es_instrument = estate->es_instrument;
/*
* Start up the subplan (this is a very cut-down form of InitPlan())
*/
subquerystate->subplan = ExecInitNode(node->subplan, sp_estate);
MemoryContextSwitchTo(oldcontext);
subquerystate->ss.ps.ps_TupFromTlist = false;
/*
* Initialize scan tuple type (needed by ExecAssignScanProjectionInfo)
*/
ExecAssignScanType(&subquerystate->ss,
ExecGetResultType(subquerystate->subplan),
false);
/*
* Initialize result tuple type and projection info.
*/
ExecAssignResultTypeFromTL(&subquerystate->ss.ps);
ExecAssignScanProjectionInfo(&subquerystate->ss);
return subquerystate;
}
/* Checks CdbCheckDispatchResult is called when queryDesc
* is not null (when shouldDispatch is true).
* This test falls in PG_CATCH when SetupInterconnect
* does not allocate queryDesc->estate->interconnect_context.
* The test is successful if the */
void
test__ExecSetParamPlan__Check_Dispatch_Results(void **state)
{
/*Set plan to explain.*/
SubPlanState *plan = makeNode(SubPlanState);
plan->xprstate.expr = makeNode(SubPlanState);
plan->planstate = makeNode(SubPlanState);
plan->planstate->instrument = (Instrumentation *)palloc(sizeof(Instrumentation));
plan->planstate->plan = makeNode(SubPlanState);
EState *estate = CreateExecutorState();
/*Assign mocked estate to plan.*/
((PlanState *)(plan->planstate))->state= estate;
/*Re-use estate mocked object. Needed as input parameter for
tested function */
ExprContext *econtext = GetPerTupleExprContext(estate);
/*Set QueryDescriptor input parameter for tested function */
PlannedStmt *plannedstmt = (PlannedStmt *)palloc(sizeof(PlannedStmt));
QueryDesc *queryDesc = (QueryDesc *)palloc(sizeof(QueryDesc));
queryDesc->plannedstmt = plannedstmt;
queryDesc->estate = (EState *)palloc(sizeof(EState));
queryDesc->estate->es_sliceTable = (SliceTable *) palloc(sizeof(SliceTable));
/*QueryDescriptor generated when shouldDispatch is true.*/
QueryDesc *internalQueryDesc = (QueryDesc *)palloc(sizeof(QueryDesc));
internalQueryDesc->estate = (EState *)palloc(sizeof(EState));
/* Added to force assertion on queryDesc->estate->interconnect_context;
to fail */
internalQueryDesc->estate->interconnect_context=NULL;
internalQueryDesc->estate->es_sliceTable = (SliceTable *) palloc(sizeof(SliceTable));
expect_any(CreateQueryDesc,plannedstmt);
expect_any(CreateQueryDesc,sourceText);
expect_any(CreateQueryDesc,snapshot);
expect_any(CreateQueryDesc,crosscheck_snapshot);
expect_any(CreateQueryDesc,dest);
expect_any(CreateQueryDesc,params);
expect_any(CreateQueryDesc,doInstrument);
will_return(CreateQueryDesc,internalQueryDesc);
Gp_role = GP_ROLE_DISPATCH;
plan->planstate->plan->dispatch=DISPATCH_PARALLEL;
will_be_called(isCurrentDtxTwoPhase);
expect_any(cdbdisp_dispatchPlan,queryDesc);
expect_any(cdbdisp_dispatchPlan,planRequiresTxn);
expect_any(cdbdisp_dispatchPlan,cancelOnError);
expect_any(cdbdisp_dispatchPlan,ds);
will_be_called(cdbdisp_dispatchPlan);
expect_any(SetupInterconnect,estate);
/* Force SetupInterconnect to fail */
will_be_called_with_sideeffect(SetupInterconnect,&_RETHROW,NULL);
expect_any(cdbexplain_localExecStats,planstate);
expect_any(cdbexplain_localExecStats,showstatctx);
will_be_called(cdbexplain_localExecStats);
expect_any(CdbCheckDispatchResult,ds);
expect_any(CdbCheckDispatchResult,waitMode);
will_be_called(CdbCheckDispatchResult);
expect_any(cdbexplain_recvExecStats,planstate);
expect_any(cdbexplain_recvExecStats,dispatchResults);
expect_any(cdbexplain_recvExecStats,sliceIndex);
expect_any(cdbexplain_recvExecStats,showstatctx);
will_be_called(cdbexplain_recvExecStats);
will_be_called(TeardownSequenceServer);
expect_any(TeardownInterconnect,transportStates);
expect_any(TeardownInterconnect,mlStates);
expect_any(TeardownInterconnect,forceEOS);
will_be_called(TeardownInterconnect);
/* Catch PG_RE_THROW(); after cleaning with CdbCheckDispatchResult */
PG_TRY();
ExecSetParamPlan(plan,econtext,queryDesc);
PG_CATCH();
assert_true(true);
PG_END_TRY();
}
/*
* Implements the 'EXECUTE' utility statement.
*/
void
ExecuteQuery(ExecuteStmt *stmt, ParamListInfo params,
DestReceiver *dest, char *completionTag)
{
PreparedStatement *entry;
char *query_string;
List *query_list,
*plan_list;
MemoryContext qcontext;
ParamListInfo paramLI = NULL;
EState *estate = NULL;
Portal portal;
/* Look it up in the hash table */
entry = FetchPreparedStatement(stmt->name, true);
query_string = entry->query_string;
query_list = entry->query_list;
plan_list = entry->plan_list;
qcontext = entry->context;
Assert(list_length(query_list) == list_length(plan_list));
/* Evaluate parameters, if any */
if (entry->argtype_list != NIL)
{
/*
* Need an EState to evaluate parameters; must not delete it till end
* of query, in case parameters are pass-by-reference.
*/
estate = CreateExecutorState();
estate->es_param_list_info = params;
paramLI = EvaluateParams(estate, stmt->params, entry->argtype_list);
}
/* Create a new portal to run the query in */
portal = CreateNewPortal();
/* Don't display the portal in pg_cursors, it is for internal use only */
portal->visible = false;
/*
* For CREATE TABLE / AS EXECUTE, make a copy of the stored query so that
* we can modify its destination (yech, but this has always been ugly).
* For regular EXECUTE we can just use the stored query where it sits,
* since the executor is read-only.
*/
if (stmt->into)
{
MemoryContext oldContext;
Query *query;
oldContext = MemoryContextSwitchTo(PortalGetHeapMemory(portal));
if (query_string)
query_string = pstrdup(query_string);
query_list = copyObject(query_list);
plan_list = copyObject(plan_list);
qcontext = PortalGetHeapMemory(portal);
if (list_length(query_list) != 1)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("prepared statement is not a SELECT")));
query = (Query *) linitial(query_list);
if (query->commandType != CMD_SELECT)
ereport(ERROR,
(errcode(ERRCODE_WRONG_OBJECT_TYPE),
errmsg("prepared statement is not a SELECT")));
query->into = copyObject(stmt->into);
query->intoOptions = copyObject(stmt->intoOptions);
query->intoOnCommit = stmt->into_on_commit;
if (stmt->into_tbl_space)
query->intoTableSpaceName = pstrdup(stmt->into_tbl_space);
MemoryContextSwitchTo(oldContext);
}
PortalDefineQuery(portal,
NULL,
query_string,
entry->commandTag,
query_list,
plan_list,
qcontext);
/*
* Run the portal to completion.
*/
PortalStart(portal, paramLI, ActiveSnapshot);
(void) PortalRun(portal, FETCH_ALL, dest, dest, completionTag);
PortalDrop(portal, false);
if (estate)
FreeExecutorState(estate);
/* No need to pfree other memory, MemoryContext will be reset */
}
/*
* Copied from Postgres' src/backend/optimizer/util/clauses.c
*
* evaluate_expr: pre-evaluate a constant expression
*
* We use the executor's routine ExecEvalExpr() to avoid duplication of
* code and ensure we get the same result as the executor would get.
*/
Expr *
evaluate_expr(Expr *expr, Oid result_type, int32 result_typmod,
Oid result_collation)
{
EState *estate;
ExprState *exprstate;
MemoryContext oldcontext;
Datum const_val;
bool const_is_null;
int16 resultTypLen;
bool resultTypByVal;
/*
* To use the executor, we need an EState.
*/
estate = CreateExecutorState();
/* We can use the estate's working context to avoid memory leaks. */
oldcontext = MemoryContextSwitchTo(estate->es_query_cxt);
/* Make sure any opfuncids are filled in. */
fix_opfuncids((Node *) expr);
/*
* Prepare expr for execution. (Note: we can't use ExecPrepareExpr
* because it'd result in recursively invoking eval_const_expressions.)
*/
exprstate = ExecInitExpr(expr, NULL);
/*
* And evaluate it.
*
* It is OK to use a default econtext because none of the ExecEvalExpr()
* code used in this situation will use econtext. That might seem
* fortuitous, but it's not so unreasonable --- a constant expression does
* not depend on context, by definition, n'est ce pas?
*/
const_val = ExecEvalExprSwitchContext(exprstate,
GetPerTupleExprContext(estate),
&const_is_null, NULL);
/* Get info needed about result datatype */
get_typlenbyval(result_type, &resultTypLen, &resultTypByVal);
/* Get back to outer memory context */
MemoryContextSwitchTo(oldcontext);
/*
* Must copy result out of sub-context used by expression eval.
*
* Also, if it's varlena, forcibly detoast it. This protects us against
* storing TOAST pointers into plans that might outlive the referenced
* data. (makeConst would handle detoasting anyway, but it's worth a few
* extra lines here so that we can do the copy and detoast in one step.)
*/
if (!const_is_null)
{
if (resultTypLen == -1)
const_val = PointerGetDatum(PG_DETOAST_DATUM_COPY(const_val));
else
const_val = datumCopy(const_val, resultTypByVal, resultTypLen);
}
/* Release all the junk we just created */
FreeExecutorState(estate);
/*
* Make the constant result node.
*/
return (Expr *) makeConst(result_type, result_typmod, result_collation,
resultTypLen,
const_val, const_is_null,
resultTypByVal);
}
