/*
* Receive a tuple from a query, and send it to the designated shm_mq.
*
* Returns true if successful, false if shm_mq has been detached.
*/
static bool
tqueueReceiveSlot(TupleTableSlot *slot, DestReceiver *self)
{
TQueueDestReceiver *tqueue = (TQueueDestReceiver *) self;
HeapTuple tuple;
shm_mq_result result;
bool should_free;
/* Send the tuple itself. */
tuple = ExecFetchSlotHeapTuple(slot, true, &should_free);
result = shm_mq_send(tqueue->queue, tuple->t_len, tuple->t_data, false);
if (should_free)
heap_freetuple(tuple);
/* Check for failure. */
if (result == SHM_MQ_DETACHED)
return false;
else if (result != SHM_MQ_SUCCESS)
ereport(ERROR,
(errcode(ERRCODE_OBJECT_NOT_IN_PREREQUISITE_STATE),
errmsg("could not send tuple to shared-memory queue")));
return true;
}
/*
* Insert tuple represented in the slot to the relation, update the indexes,
* and execute any constraints and per-row triggers.
*
* Caller is responsible for opening the indexes.
*/
void
ExecSimpleRelationInsert(EState *estate, TupleTableSlot *slot)
{
bool skip_tuple = false;
HeapTuple tuple;
ResultRelInfo *resultRelInfo = estate->es_result_relation_info;
Relation rel = resultRelInfo->ri_RelationDesc;
/* For now we support only tables. */
Assert(rel->rd_rel->relkind == RELKIND_RELATION);
CheckCmdReplicaIdentity(rel, CMD_INSERT);
/* BEFORE ROW INSERT Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->trig_insert_before_row)
{
slot = ExecBRInsertTriggers(estate, resultRelInfo, slot);
if (slot == NULL) /* "do nothing" */
skip_tuple = true;
}
if (!skip_tuple)
{
List *recheckIndexes = NIL;
/* Check the constraints of the tuple */
if (rel->rd_att->constr)
ExecConstraints(resultRelInfo, slot, estate);
if (resultRelInfo->ri_PartitionCheck)
ExecPartitionCheck(resultRelInfo, slot, estate, true);
/* Materialize slot into a tuple that we can scribble upon. */
tuple = ExecFetchSlotHeapTuple(slot, true, NULL);
/* OK, store the tuple and create index entries for it */
simple_heap_insert(rel, tuple);
if (resultRelInfo->ri_NumIndices > 0)
recheckIndexes = ExecInsertIndexTuples(slot, &(tuple->t_self),
estate, false, NULL,
NIL);
/* AFTER ROW INSERT Triggers */
ExecARInsertTriggers(estate, resultRelInfo, tuple,
recheckIndexes, NULL);
/*
* XXX we should in theory pass a TransitionCaptureState object to the
* above to capture transition tuples, but after statement triggers
* don't actually get fired by replication yet anyway
*/
list_free(recheckIndexes);
}
}
/* --------------------------------
* ExecFetchSlotTupleDatum
* Fetch the slot's tuple as a composite-type Datum.
*
* We convert the slot's contents to local physical-tuple form,
* and fill in the Datum header fields. Note that the result
* always points to storage owned by the slot.
* --------------------------------
*/
Datum
ExecFetchSlotTupleDatum(TupleTableSlot *slot)
{
HeapTuple tup;
HeapTupleHeader td;
TupleDesc tupdesc;
/* Make sure we can scribble on the slot contents ...
* GPDB: ExecFetchSlotHeapTuple() replaces the Postgres call to
* ExecMaterializeSlot() due to restructuring of the code around
* memtuple support */
tup = ExecFetchSlotHeapTuple(slot);
/* ... and set up the composite-Datum header fields, in case not done */
td = tup->t_data;
tupdesc = slot->tts_tupleDescriptor;
HeapTupleHeaderSetDatumLength(td, tup->t_len);
HeapTupleHeaderSetTypeId(td, tupdesc->tdtypeid);
HeapTupleHeaderSetTypMod(td, tupdesc->tdtypmod);
return PointerGetDatum(td);
}
/* ----------------------------------------------------------------
* FunctionNext
*
* This is a workhorse for ExecFunctionScan
* ----------------------------------------------------------------
*/
static TupleTableSlot *
FunctionNext(FunctionScanState *node)
{
TupleTableSlot *slot;
EState *estate;
ScanDirection direction;
Tuplestorestate *tuplestorestate;
/*
* get information from the estate and scan state
*/
estate = node->ss.ps.state;
direction = estate->es_direction;
tuplestorestate = node->tuplestorestate;
/*
* If first time through, read all tuples from function and put them in a
* tuplestore. Subsequent calls just fetch tuples from tuplestore.
*/
if (tuplestorestate == NULL)
{
tuplestorestate = ExecMakeTableFunctionResult(
node->funcexpr,
node->ss.ps.ps_ExprContext,
node->tupdesc,
PlanStateOperatorMemKB( (PlanState *) node));
node->tuplestorestate = tuplestorestate;
/* CDB: Offer extra info for EXPLAIN ANALYZE. */
if (node->ss.ps.instrument)
{
/* Let the tuplestore share our Instrumentation object. */
tuplestore_set_instrument(tuplestorestate, node->ss.ps.instrument);
/* Request a callback at end of query. */
node->ss.ps.cdbexplainfun = ExecFunctionScanExplainEnd;
}
}
/*
* Get the next tuple from tuplestore. Return NULL if no more tuples.
*/
slot = node->ss.ss_ScanTupleSlot;
if (tuplestore_gettupleslot(tuplestorestate,
ScanDirectionIsForward(direction),
slot))
{
/* CDB: Label each row with a synthetic ctid for subquery dedup. */
if (node->cdb_want_ctid)
{
HeapTuple tuple = ExecFetchSlotHeapTuple(slot);
/* Increment 48-bit row count */
node->cdb_fake_ctid.ip_posid++;
if (node->cdb_fake_ctid.ip_posid == 0)
ItemPointerSetBlockNumber(&node->cdb_fake_ctid,
1 + ItemPointerGetBlockNumber(&node->cdb_fake_ctid));
tuple->t_self = node->cdb_fake_ctid;
}
}
if (!TupIsNull(slot))
{
Gpmon_M_Incr_Rows_Out(GpmonPktFromFuncScanState(node));
CheckSendPlanStateGpmonPkt(&node->ss.ps);
}
else if (!node->ss.ps.delayEagerFree)
{
ExecEagerFreeFunctionScan((FunctionScanState *)(&node->ss.ps));
}
return slot;
}
/* ----------------------------------------------------------------
* ExecUpdate
*
* note: we can't run UPDATE queries with transactions
* off because UPDATEs are actually INSERTs and our
* scan will mistakenly loop forever, updating the tuple
* it just inserted.. This should be fixed but until it
* is, we don't want to get stuck in an infinite loop
* which corrupts your database..
* ----------------------------------------------------------------
*/
void
ExecUpdate(TupleTableSlot *slot,
ItemPointer tupleid,
TupleTableSlot *planSlot,
DestReceiver *dest,
EState *estate)
{
HeapTuple tuple;
ResultRelInfo *resultRelInfo;
Relation resultRelationDesc;
HTSU_Result result;
ItemPointerData update_ctid;
TransactionId update_xmax;
/*
* abort the operation if not running transactions
*/
if (IsBootstrapProcessingMode())
elog(ERROR, "cannot UPDATE during bootstrap");
/*
* get the heap tuple out of the tuple table slot, making sure we have a
* writable copy
*/
tuple = ExecFetchSlotHeapTuple(slot);
/*
* get information on the (current) result relation
*/
resultRelInfo = estate->es_result_relation_info;
resultRelationDesc = resultRelInfo->ri_RelationDesc;
/* see if this update would move the tuple to a different partition */
if (estate->es_result_partitions)
{
AttrNumber max_attr;
Datum *values;
bool *nulls;
Oid targetid;
Assert(estate->es_partition_state != NULL &&
estate->es_partition_state->accessMethods != NULL);
if (!estate->es_partition_state->accessMethods->part_cxt)
estate->es_partition_state->accessMethods->part_cxt =
GetPerTupleExprContext(estate)->ecxt_per_tuple_memory;
Assert(PointerIsValid(estate->es_result_partitions));
max_attr = estate->es_partition_state->max_partition_attr;
slot_getsomeattrs(slot, max_attr);
values = slot_get_values(slot);
nulls = slot_get_isnull(slot);
targetid = selectPartition(estate->es_result_partitions, values,
nulls, slot->tts_tupleDescriptor,
estate->es_partition_state->accessMethods);
if (!OidIsValid(targetid))
ereport(ERROR,
(errcode(ERRCODE_NO_PARTITION_FOR_PARTITIONING_KEY),
errmsg("no partition for partitioning key")));
if (RelationGetRelid(resultRelationDesc) != targetid)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("moving tuple from partition \"%s\" to "
"partition \"%s\" not supported",
get_rel_name(RelationGetRelid(resultRelationDesc)),
get_rel_name(targetid)),
errOmitLocation(true)));
}
}
/* BEFORE ROW UPDATE Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->n_before_row[TRIGGER_EVENT_UPDATE] > 0)
{
HeapTuple newtuple;
newtuple = ExecBRUpdateTriggers(estate, resultRelInfo,
tupleid, tuple,
estate->es_snapshot->curcid);
if (newtuple == NULL) /* "do nothing" */
return;
if (newtuple != tuple) /* modified by Trigger(s) */
{
/*
* Put the modified tuple into a slot for convenience of routines
* below. We assume the tuple was allocated in per-tuple memory
* context, and therefore will go away by itself. The tuple table
* slot should not try to clear it.
*/
TupleTableSlot *newslot = estate->es_trig_tuple_slot;
if (newslot->tts_tupleDescriptor != slot->tts_tupleDescriptor)
ExecSetSlotDescriptor(newslot, slot->tts_tupleDescriptor);
ExecStoreGenericTuple(newtuple, newslot, false);
newslot->tts_tableOid = slot->tts_tableOid; /* for constraints */
slot = newslot;
tuple = newtuple;
}
}
/*
* Check the constraints of the tuple
*
* If we generate a new candidate tuple after EvalPlanQual testing, we
* must loop back here and recheck constraints. (We don't need to redo
* triggers, however. If there are any BEFORE triggers then trigger.c
* will have done heap_lock_tuple to lock the correct tuple, so there's no
* need to do them again.)
*/
lreplace:;
if (resultRelationDesc->rd_att->constr)
ExecConstraints(resultRelInfo, slot, estate);
if (!GpPersistent_IsPersistentRelation(resultRelationDesc->rd_id))
{
/*
* Normal UPDATE path.
*/
/*
* replace the heap tuple
*
* Note: if es_crosscheck_snapshot isn't InvalidSnapshot, we check that
* the row to be updated is visible to that snapshot, and throw a can't-
* serialize error if not. This is a special-case behavior needed for
* referential integrity updates in serializable transactions.
*/
result = heap_update(resultRelationDesc, tupleid, tuple,
&update_ctid, &update_xmax,
estate->es_snapshot->curcid,
estate->es_crosscheck_snapshot,
true /* wait for commit */ );
switch (result)
{
case HeapTupleSelfUpdated:
/* already deleted by self; nothing to do */
return;
case HeapTupleMayBeUpdated:
break;
case HeapTupleUpdated:
if (IsXactIsoLevelSerializable)
ereport(ERROR,
(errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
errmsg("could not serialize access due to concurrent update")));
else if (!ItemPointerEquals(tupleid, &update_ctid))
{
TupleTableSlot *epqslot;
epqslot = EvalPlanQual(estate,
resultRelInfo->ri_RangeTableIndex,
&update_ctid,
update_xmax,
estate->es_snapshot->curcid);
if (!TupIsNull(epqslot))
{
*tupleid = update_ctid;
slot = ExecFilterJunk(estate->es_junkFilter, epqslot);
tuple = ExecFetchSlotHeapTuple(slot);
goto lreplace;
}
}
/* tuple already deleted; nothing to do */
return;
default:
elog(ERROR, "unrecognized heap_update status: %u", result);
return;
}
}
else
{
HeapTuple persistentTuple;
/*
* Persistent metadata path.
*/
persistentTuple = heap_copytuple(tuple);
persistentTuple->t_self = *tupleid;
frozen_heap_inplace_update(resultRelationDesc, persistentTuple);
heap_freetuple(persistentTuple);
}
IncrReplaced();
(estate->es_processed)++;
/*
* Note: instead of having to update the old index tuples associated with
* the heap tuple, all we do is form and insert new index tuples. This is
* because UPDATEs are actually DELETEs and INSERTs, and index tuple
* deletion is done later by VACUUM (see notes in ExecDelete). All we do
* here is insert new index tuples. -cim 9/27/89
*/
/*
* insert index entries for tuple
*
* Note: heap_update returns the tid (location) of the new tuple in the
* t_self field.
*/
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(slot, &(tuple->t_self), estate, false);
/* AFTER ROW UPDATE Triggers */
ExecARUpdateTriggers(estate, resultRelInfo, tupleid, tuple);
}
/* ----------------------------------------------------------------
* ExecInsert
*
* INSERTs have to add the tuple into
* the base relation and insert appropriate tuples into the
* index relations.
* Insert can be part of an update operation when
* there is a preceding SplitUpdate node.
* ----------------------------------------------------------------
*/
void
ExecInsert(TupleTableSlot *slot,
DestReceiver *dest,
EState *estate,
PlanGenerator planGen,
bool isUpdate)
{
void *tuple = NULL;
ResultRelInfo *resultRelInfo = NULL;
Relation resultRelationDesc = NULL;
Oid newId = InvalidOid;
TupleTableSlot *partslot = NULL;
AOTupleId aoTupleId = AOTUPLEID_INIT;
bool rel_is_heap = false;
bool rel_is_aorows = false;
bool rel_is_external = false;
bool rel_is_parquet = false;
/*
* get information on the (current) result relation
*/
if (estate->es_result_partitions)
{
resultRelInfo = slot_get_partition(slot, estate);
estate->es_result_relation_info = resultRelInfo;
if (NULL != resultRelInfo->ri_parquetSendBack)
{
/*
* The Parquet part we are about to insert into
* has sendBack information. This means we're inserting into the
* part twice, which is not supported. Error out (GPSQL-2291)
*/
Assert(gp_parquet_insert_sort);
ereport(ERROR, (errcode(ERRCODE_CDB_FEATURE_NOT_YET),
errmsg("Cannot insert out-of-order tuples in parquet partitions"),
errhint("Sort the data on the partitioning key(s) before inserting"),
errOmitLocation(true)));
}
/*
* Check if we need to close the last parquet partition we
* inserted into (GPSQL-2291).
*/
Oid new_part_oid = resultRelInfo->ri_RelationDesc->rd_id;
if (gp_parquet_insert_sort &&
PLANGEN_OPTIMIZER == planGen &&
InvalidOid != estate->es_last_parq_part &&
new_part_oid != estate->es_last_parq_part)
{
Assert(NULL != estate->es_partition_state->result_partition_hash);
ResultPartHashEntry *entry = hash_search(estate->es_partition_state->result_partition_hash,
&estate->es_last_parq_part,
HASH_FIND,
NULL /* found */);
Assert(NULL != entry);
Assert(entry->offset < estate->es_num_result_relations);
ResultRelInfo *oldResultRelInfo = & estate->es_result_relations[entry->offset];
elog(DEBUG1, "Switching from old part oid=%d name=[%s] to new part oid=%d name=[%s]",
estate->es_last_parq_part,
oldResultRelInfo->ri_RelationDesc->rd_rel->relname.data,
new_part_oid,
resultRelInfo->ri_RelationDesc->rd_rel->relname.data);
/*
* We are opening a new partition, and the last partition we
* inserted into was a Parquet part. Let's close the old
* parquet insert descriptor to free the memory before
* opening the new one.
*/
ParquetInsertDescData *oldInsertDesc = oldResultRelInfo->ri_parquetInsertDesc;
/*
* We need to preserve the "sendback" information that needs to be
* sent back to the QD process from this part.
* Compute it here, and store it for later use.
*/
QueryContextDispatchingSendBack sendback =
CreateQueryContextDispatchingSendBack(1);
sendback->relid = RelationGetRelid(oldResultRelInfo->ri_RelationDesc);
oldInsertDesc->sendback = sendback;
parquet_insert_finish(oldInsertDesc);
/* Store the sendback information in the resultRelInfo for this part */
oldResultRelInfo->ri_parquetSendBack = sendback;
/* Record in the resultRelInfo that we closed the parquet insert descriptor */
oldResultRelInfo->ri_parquetInsertDesc = NULL;
/* Reset the last parquet part Oid, it's now closed */
estate->es_last_parq_part = InvalidOid;
}
}
else
{
resultRelInfo = estate->es_result_relation_info;
}
Assert (!resultRelInfo->ri_projectReturning);
resultRelationDesc = resultRelInfo->ri_RelationDesc;
rel_is_heap = RelationIsHeap(resultRelationDesc);
rel_is_aorows = RelationIsAoRows(resultRelationDesc);
rel_is_external = RelationIsExternal(resultRelationDesc);
rel_is_parquet = RelationIsParquet(resultRelationDesc);
/* Validate that insert is not part of an non-allowed update operation. */
if (isUpdate && (rel_is_aorows || rel_is_parquet))
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("Append-only tables are not updatable. Operation not permitted."),
errOmitLocation(true)));
}
partslot = reconstructMatchingTupleSlot(slot, resultRelInfo);
if (rel_is_heap || rel_is_external)
{
tuple = ExecFetchSlotHeapTuple(partslot);
}
else if (rel_is_aorows)
{
tuple = ExecFetchSlotMemTuple(partslot, false);
}
else if (rel_is_parquet)
{
tuple = NULL;
}
Assert( partslot != NULL );
Assert( rel_is_parquet || (tuple != NULL));
/* Execute triggers in Planner-generated plans */
if (planGen == PLANGEN_PLANNER)
{
/* BEFORE ROW INSERT Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->n_before_row[TRIGGER_EVENT_INSERT] > 0)
{
HeapTuple newtuple;
/* NYI */
if(rel_is_parquet)
elog(ERROR, "triggers are not supported on tables that use column-oriented storage");
newtuple = ExecBRInsertTriggers(estate, resultRelInfo, tuple);
if (newtuple == NULL) /* "do nothing" */
{
return;
}
if (newtuple != tuple) /* modified by Trigger(s) */
{
/*
* Put the modified tuple into a slot for convenience of routines
* below. We assume the tuple was allocated in per-tuple memory
* context, and therefore will go away by itself. The tuple table
* slot should not try to clear it.
*/
TupleTableSlot *newslot = estate->es_trig_tuple_slot;
if (newslot->tts_tupleDescriptor != partslot->tts_tupleDescriptor)
ExecSetSlotDescriptor(newslot, partslot->tts_tupleDescriptor);
ExecStoreGenericTuple(newtuple, newslot, false);
newslot->tts_tableOid = partslot->tts_tableOid; /* for constraints */
tuple = newtuple;
partslot = newslot;
}
}
}
/*
* Check the constraints of the tuple
*/
if (resultRelationDesc->rd_att->constr &&
planGen == PLANGEN_PLANNER)
{
ExecConstraints(resultRelInfo, partslot, estate);
}
/*
* insert the tuple
*
* Note: heap_insert returns the tid (location) of the new tuple in the
* t_self field.
*
* NOTE: for append-only relations we use the append-only access methods.
*/
if (rel_is_aorows)
{
if (resultRelInfo->ri_aoInsertDesc == NULL)
{
ResultRelSegFileInfo *segfileinfo = NULL;
/* Set the pre-assigned fileseg number to insert into */
ResultRelInfoSetSegFileInfo(resultRelInfo, estate->es_result_segfileinfos);
segfileinfo = (ResultRelSegFileInfo *)list_nth(resultRelInfo->ri_aosegfileinfos, GetQEIndex());
resultRelInfo->ri_aoInsertDesc =
appendonly_insert_init(resultRelationDesc,
segfileinfo);
}
appendonly_insert(resultRelInfo->ri_aoInsertDesc, tuple, &newId, &aoTupleId);
}
else if (rel_is_external)
{
/* Writable external table */
if (resultRelInfo->ri_extInsertDesc == NULL)
resultRelInfo->ri_extInsertDesc = external_insert_init(
resultRelationDesc, 0);
newId = external_insert(resultRelInfo->ri_extInsertDesc, tuple);
}
else if(rel_is_parquet)
{
/* If there is no parquet insert descriptor, create it now. */
if (resultRelInfo->ri_parquetInsertDesc == NULL)
{
ResultRelSegFileInfo *segfileinfo = NULL;
ResultRelInfoSetSegFileInfo(resultRelInfo, estate->es_result_segfileinfos);
segfileinfo = (ResultRelSegFileInfo *)list_nth(resultRelInfo->ri_aosegfileinfos, GetQEIndex());
resultRelInfo->ri_parquetInsertDesc = parquet_insert_init(resultRelationDesc, segfileinfo);
/*
* Just opened a new parquet partition for insert. Save the Oid
* in estate, so that we can close it when switching to a
* new partition (GPSQL-2291)
*/
elog(DEBUG1, "Saving es_last_parq_part. Old=%d, new=%d", estate->es_last_parq_part, resultRelationDesc->rd_id);
estate->es_last_parq_part = resultRelationDesc->rd_id;
}
newId = parquet_insert(resultRelInfo->ri_parquetInsertDesc, partslot);
}
else
{
Insist(rel_is_heap);
newId = heap_insert(resultRelationDesc,
tuple,
estate->es_snapshot->curcid,
true, true, GetCurrentTransactionId());
}
IncrAppended();
(estate->es_processed)++;
(resultRelInfo->ri_aoprocessed)++;
estate->es_lastoid = newId;
partslot->tts_tableOid = RelationGetRelid(resultRelationDesc);
if (rel_is_aorows || rel_is_parquet)
{
/* NOTE: Current version does not support index upon parquet table. */
/*
* insert index entries for AO Row-Store tuple
*/
if (resultRelInfo->ri_NumIndices > 0 && !rel_is_parquet)
ExecInsertIndexTuples(partslot, (ItemPointer)&aoTupleId, estate, false);
}
else
{
/* Use parttuple for index update in case this is an indexed heap table. */
TupleTableSlot *xslot = partslot;
void *xtuple = tuple;
setLastTid(&(((HeapTuple) xtuple)->t_self));
/*
* insert index entries for tuple
*/
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(xslot, &(((HeapTuple) xtuple)->t_self), estate, false);
}
if (planGen == PLANGEN_PLANNER)
{
/* AFTER ROW INSERT Triggers */
ExecARInsertTriggers(estate, resultRelInfo, tuple);
}
}
/* ----------------------------------------------------------------
* ValuesNext
*
* This is a workhorse for ExecValuesScan
* ----------------------------------------------------------------
*/
static TupleTableSlot *
ValuesNext(ValuesScanState *node)
{
TupleTableSlot *slot;
EState *estate;
ExprContext *econtext;
ScanDirection direction;
List *exprlist;
/*
* get information from the estate and scan state
*/
estate = node->ss.ps.state;
direction = estate->es_direction;
slot = node->ss.ss_ScanTupleSlot;
econtext = node->rowcontext;
/*
* Get the next tuple. Return NULL if no more tuples.
*/
if (ScanDirectionIsForward(direction))
{
if (node->curr_idx < node->array_len)
node->curr_idx++;
if (node->curr_idx < node->array_len)
exprlist = node->exprlists[node->curr_idx];
else
exprlist = NIL;
}
else
{
if (node->curr_idx >= 0)
node->curr_idx--;
if (node->curr_idx >= 0)
exprlist = node->exprlists[node->curr_idx];
else
exprlist = NIL;
}
/*
* Always clear the result slot; this is appropriate if we are at the end
* of the data, and if we're not, we still need it as the first step of
* the store-virtual-tuple protocol. It seems wise to clear the slot
* before we reset the context it might have pointers into.
*/
ExecClearTuple(slot);
if (exprlist)
{
MemoryContext oldContext;
List *exprstatelist;
Datum *values;
bool *isnull;
ListCell *lc;
int resind;
/*
* Get rid of any prior cycle's leftovers. We use ReScanExprContext
* not just ResetExprContext because we want any registered shutdown
* callbacks to be called.
*/
ReScanExprContext(econtext);
/*
* Build the expression eval state in the econtext's per-tuple memory.
* This is a tad unusual, but we want to delete the eval state again
* when we move to the next row, to avoid growth of memory
* requirements over a long values list.
*/
oldContext = MemoryContextSwitchTo(econtext->ecxt_per_tuple_memory);
/*
* Pass NULL, not my plan node, because we don't want anything in this
* transient state linking into permanent state. The only possibility
* is a SubPlan, and there shouldn't be any (any subselects in the
* VALUES list should be InitPlans).
*/
exprstatelist = (List *) ExecInitExpr((Expr *) exprlist, NULL);
/* parser should have checked all sublists are the same length */
Assert(list_length(exprstatelist) == slot->tts_tupleDescriptor->natts);
/*
* Compute the expressions and build a virtual result tuple. We
* already did ExecClearTuple(slot).
*/
ExecClearTuple(slot);
values = slot_get_values(slot);
isnull = slot_get_isnull(slot);
resind = 0;
foreach(lc, exprstatelist)
{
ExprState *estate = (ExprState *) lfirst(lc);
values[resind] = ExecEvalExpr(estate,
econtext,
&isnull[resind],
NULL);
resind++;
}
MemoryContextSwitchTo(oldContext);
/*
* And return the virtual tuple.
*/
ExecStoreVirtualTuple(slot);
/* CDB: Label each row with a synthetic ctid for subquery dedup. */
if (node->cdb_want_ctid)
{
HeapTuple tuple = ExecFetchSlotHeapTuple(slot);
ItemPointerSet(&tuple->t_self, node->curr_idx >> 16,
(OffsetNumber)node->curr_idx);
}
/*
* Find the searchslot tuple and update it with data in the slot,
* update the indexes, and execute any constraints and per-row triggers.
*
* Caller is responsible for opening the indexes.
*/
void
ExecSimpleRelationUpdate(EState *estate, EPQState *epqstate,
TupleTableSlot *searchslot, TupleTableSlot *slot)
{
bool skip_tuple = false;
HeapTuple tuple;
ResultRelInfo *resultRelInfo = estate->es_result_relation_info;
Relation rel = resultRelInfo->ri_RelationDesc;
HeapTupleTableSlot *hsearchslot = (HeapTupleTableSlot *)searchslot;
HeapTupleTableSlot *hslot = (HeapTupleTableSlot *)slot;
/* We expect both searchslot and the slot to contain a heap tuple. */
Assert(TTS_IS_HEAPTUPLE(searchslot) || TTS_IS_BUFFERTUPLE(searchslot));
Assert(TTS_IS_HEAPTUPLE(slot) || TTS_IS_BUFFERTUPLE(slot));
/* For now we support only tables. */
Assert(rel->rd_rel->relkind == RELKIND_RELATION);
CheckCmdReplicaIdentity(rel, CMD_UPDATE);
/* BEFORE ROW UPDATE Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->trig_update_before_row)
{
slot = ExecBRUpdateTriggers(estate, epqstate, resultRelInfo,
&hsearchslot->tuple->t_self, NULL, slot);
if (slot == NULL) /* "do nothing" */
skip_tuple = true;
}
if (!skip_tuple)
{
List *recheckIndexes = NIL;
/* Check the constraints of the tuple */
if (rel->rd_att->constr)
ExecConstraints(resultRelInfo, slot, estate);
if (resultRelInfo->ri_PartitionCheck)
ExecPartitionCheck(resultRelInfo, slot, estate, true);
/* Materialize slot into a tuple that we can scribble upon. */
tuple = ExecFetchSlotHeapTuple(slot, true, NULL);
/* OK, update the tuple and index entries for it */
simple_heap_update(rel, &hsearchslot->tuple->t_self, hslot->tuple);
if (resultRelInfo->ri_NumIndices > 0 &&
!HeapTupleIsHeapOnly(hslot->tuple))
recheckIndexes = ExecInsertIndexTuples(slot, &(tuple->t_self),
estate, false, NULL,
NIL);
/* AFTER ROW UPDATE Triggers */
ExecARUpdateTriggers(estate, resultRelInfo,
&hsearchslot->tuple->t_self, NULL, tuple,
recheckIndexes, NULL);
list_free(recheckIndexes);
}
}
/*
* Executes INSERT and DELETE DML operations. The
* action is specified within the TupleTableSlot at
* plannode->actionColIdx.The ctid of the tuple to delete
* is in position plannode->ctidColIdx in the current slot.
* */
TupleTableSlot*
ExecDML(DMLState *node)
{
PlanState *outerNode = outerPlanState(node);
DML *plannode = (DML *) node->ps.plan;
Assert(outerNode != NULL);
TupleTableSlot *slot = ExecProcNode(outerNode);
if (TupIsNull(slot))
{
return NULL;
}
bool isnull = false;
int action = DatumGetUInt32(slot_getattr(slot, plannode->actionColIdx, &isnull));
Assert(!isnull);
isnull = false;
Datum oid = slot_getattr(slot, plannode->oidColIdx, &isnull);
slot->tts_tableOid = DatumGetUInt32(oid);
bool isUpdate = false;
if (node->ps.state->es_plannedstmt->commandType == CMD_UPDATE)
{
isUpdate = true;
}
Assert(action == DML_INSERT || action == DML_DELETE);
/*
* Reset per-tuple memory context to free any expression evaluation
* storage allocated in the previous tuple cycle.
*/
ExprContext *econtext = node->ps.ps_ExprContext;
ResetExprContext(econtext);
/* Prepare cleaned-up tuple by projecting it and filtering junk columns */
econtext->ecxt_outertuple = slot;
TupleTableSlot *projectedSlot = ExecProject(node->ps.ps_ProjInfo, NULL);
/* remove 'junk' columns from tuple */
node->cleanedUpSlot = ExecFilterJunk(node->junkfilter, projectedSlot);
if (DML_INSERT == action)
{
/* Respect any given tuple Oid when updating a tuple. */
if(isUpdate &&
plannode->tupleoidColIdx != 0)
{
isnull = false;
oid = slot_getattr(slot, plannode->tupleoidColIdx, &isnull);
HeapTuple htuple = ExecFetchSlotHeapTuple(node->cleanedUpSlot);
Assert(htuple == node->cleanedUpSlot->PRIVATE_tts_heaptuple);
HeapTupleSetOid(htuple, oid);
}
/* The plan origin is required since ExecInsert performs different actions
* depending on the type of plan (constraint enforcement and triggers.)
*/
ExecInsert(node->cleanedUpSlot, NULL /* destReceiver */,
node->ps.state, PLANGEN_OPTIMIZER /* Plan origin */,
isUpdate);
}
else /* DML_DELETE */
{
Datum ctid = slot_getattr(slot, plannode->ctidColIdx, &isnull);
Assert(!isnull);
ItemPointer tupleid = (ItemPointer) DatumGetPointer(ctid);
ItemPointerData tuple_ctid = *tupleid;
tupleid = &tuple_ctid;
/* Correct tuple count by ignoring deletes when splitting tuples. */
ExecDelete(tupleid, node->cleanedUpSlot, NULL /* DestReceiver */, node->ps.state,
PLANGEN_OPTIMIZER /* Plan origin */, isUpdate);
}
return slot;
}
/* ----------------------------------------------------------------
* ExecUpdate
*
* note: we can't run UPDATE queries with transactions
* off because UPDATEs are actually INSERTs and our
* scan will mistakenly loop forever, updating the tuple
* it just inserted.. This should be fixed but until it
* is, we don't want to get stuck in an infinite loop
* which corrupts your database..
* ----------------------------------------------------------------
*/
void
ExecUpdate(TupleTableSlot *slot,
ItemPointer tupleid,
TupleTableSlot *planSlot,
DestReceiver *dest,
EState *estate)
{
void* tuple;
ResultRelInfo *resultRelInfo;
Relation resultRelationDesc;
HTSU_Result result;
ItemPointerData update_ctid;
TransactionId update_xmax;
AOTupleId aoTupleId = AOTUPLEID_INIT;
TupleTableSlot *partslot = NULL;
/*
* abort the operation if not running transactions
*/
if (IsBootstrapProcessingMode())
elog(ERROR, "cannot UPDATE during bootstrap");
/*
* get information on the (current) result relation
*/
resultRelInfo = estate->es_result_relation_info;
resultRelationDesc = resultRelInfo->ri_RelationDesc;
bool rel_is_heap = RelationIsHeap(resultRelationDesc);
bool rel_is_aorows = RelationIsAoRows(resultRelationDesc);
bool rel_is_aocols = RelationIsAoCols(resultRelationDesc);
bool rel_is_external = RelationIsExternal(resultRelationDesc);
/*
* get the heap tuple out of the tuple table slot, making sure we have a
* writable copy
*/
if (rel_is_heap)
{
partslot = slot;
tuple = ExecFetchSlotHeapTuple(partslot);
}
else if (rel_is_aorows || rel_is_aocols)
{
/*
* It is necessary to reconstruct a logically compatible tuple to
* a phyiscally compatible tuple. The slot's tuple descriptor comes
* from the projection target list, which doesn't indicate dropped
* columns, and MemTuple cannot deal with cases without converting
* the target list back into the original relation's tuple desc.
*/
partslot = reconstructMatchingTupleSlot(slot, resultRelInfo);
/*
* We directly inline toasted columns here as update with toasted columns
* would create two references to the same toasted value.
*/
tuple = ExecFetchSlotMemTuple(partslot, true);
}
else if (rel_is_external)
{
if (estate->es_result_partitions &&
estate->es_result_partitions->part->parrelid != 0)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("Update external partitions not supported.")));
return;
}
else
{
partslot = slot;
tuple = ExecFetchSlotHeapTuple(partslot);
}
}
else
{
Insist(false);
}
/* see if this update would move the tuple to a different partition */
if (estate->es_result_partitions)
checkPartitionUpdate(estate, partslot, resultRelInfo);
/* BEFORE ROW UPDATE Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->n_before_row[TRIGGER_EVENT_UPDATE] > 0)
{
HeapTuple newtuple;
newtuple = ExecBRUpdateTriggers(estate, resultRelInfo,
tupleid, tuple,
estate->es_snapshot->curcid);
if (newtuple == NULL) /* "do nothing" */
return;
if (newtuple != tuple) /* modified by Trigger(s) */
{
/*
* Put the modified tuple into a slot for convenience of routines
* below. We assume the tuple was allocated in per-tuple memory
* context, and therefore will go away by itself. The tuple table
* slot should not try to clear it.
*/
TupleTableSlot *newslot = estate->es_trig_tuple_slot;
if (newslot->tts_tupleDescriptor != partslot->tts_tupleDescriptor)
ExecSetSlotDescriptor(newslot, partslot->tts_tupleDescriptor);
ExecStoreGenericTuple(newtuple, newslot, false);
newslot->tts_tableOid = partslot->tts_tableOid; /* for constraints */
partslot = newslot;
tuple = newtuple;
}
}
/*
* Check the constraints of the tuple
*
* If we generate a new candidate tuple after EvalPlanQual testing, we
* must loop back here and recheck constraints. (We don't need to redo
* triggers, however. If there are any BEFORE triggers then trigger.c
* will have done heap_lock_tuple to lock the correct tuple, so there's no
* need to do them again.)
*/
lreplace:;
if (resultRelationDesc->rd_att->constr)
ExecConstraints(resultRelInfo, partslot, estate);
if (!GpPersistent_IsPersistentRelation(resultRelationDesc->rd_id))
{
/*
* Normal UPDATE path.
*/
/*
* replace the heap tuple
*
* Note: if es_crosscheck_snapshot isn't InvalidSnapshot, we check that
* the row to be updated is visible to that snapshot, and throw a can't-
* serialize error if not. This is a special-case behavior needed for
* referential integrity updates in serializable transactions.
*/
if (rel_is_heap)
{
result = heap_update(resultRelationDesc, tupleid, tuple,
&update_ctid, &update_xmax,
estate->es_snapshot->curcid,
estate->es_crosscheck_snapshot,
true /* wait for commit */ );
}
else if (rel_is_aorows)
{
if (IsXactIsoLevelSerializable)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("Updates on append-only tables are not supported in serializable transactions.")));
}
if (resultRelInfo->ri_updateDesc == NULL)
{
ResultRelInfoSetSegno(resultRelInfo, estate->es_result_aosegnos);
resultRelInfo->ri_updateDesc = (AppendOnlyUpdateDesc)
appendonly_update_init(resultRelationDesc, ActiveSnapshot, resultRelInfo->ri_aosegno);
}
result = appendonly_update(resultRelInfo->ri_updateDesc,
tuple, (AOTupleId *) tupleid, &aoTupleId);
}
else if (rel_is_aocols)
{
if (IsXactIsoLevelSerializable)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("Updates on append-only tables are not supported in serializable transactions.")));
}
if (resultRelInfo->ri_updateDesc == NULL)
{
ResultRelInfoSetSegno(resultRelInfo, estate->es_result_aosegnos);
resultRelInfo->ri_updateDesc = (AppendOnlyUpdateDesc)
aocs_update_init(resultRelationDesc, resultRelInfo->ri_aosegno);
}
result = aocs_update(resultRelInfo->ri_updateDesc,
partslot, (AOTupleId *) tupleid, &aoTupleId);
}
else
{
Assert(!"We should not be here");
}
switch (result)
{
case HeapTupleSelfUpdated:
/* already deleted by self; nothing to do */
return;
case HeapTupleMayBeUpdated:
break;
case HeapTupleUpdated:
if (IsXactIsoLevelSerializable)
ereport(ERROR,
(errcode(ERRCODE_T_R_SERIALIZATION_FAILURE),
errmsg("could not serialize access due to concurrent update")));
else if (!ItemPointerEquals(tupleid, &update_ctid))
{
TupleTableSlot *epqslot;
epqslot = EvalPlanQual(estate,
resultRelInfo->ri_RangeTableIndex,
&update_ctid,
update_xmax,
estate->es_snapshot->curcid);
if (!TupIsNull(epqslot))
{
*tupleid = update_ctid;
partslot = ExecFilterJunk(estate->es_junkFilter, epqslot);
tuple = ExecFetchSlotHeapTuple(partslot);
goto lreplace;
}
}
/* tuple already deleted; nothing to do */
return;
default:
elog(ERROR, "unrecognized heap_update status: %u", result);
return;
}
}
else
{
HeapTuple persistentTuple;
/*
* Persistent metadata path.
*/
persistentTuple = heap_copytuple(tuple);
persistentTuple->t_self = *tupleid;
frozen_heap_inplace_update(resultRelationDesc, persistentTuple);
heap_freetuple(persistentTuple);
}
IncrReplaced();
(estate->es_processed)++;
(resultRelInfo->ri_aoprocessed)++;
/*
* Note: instead of having to update the old index tuples associated with
* the heap tuple, all we do is form and insert new index tuples. This is
* because UPDATEs are actually DELETEs and INSERTs, and index tuple
* deletion is done later by VACUUM (see notes in ExecDelete). All we do
* here is insert new index tuples. -cim 9/27/89
*/
/*
* insert index entries for tuple
*
* Note: heap_update returns the tid (location) of the new tuple in the
* t_self field.
*/
if (rel_is_aorows || rel_is_aocols)
{
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(partslot, (ItemPointer)&aoTupleId, estate, false);
}
else
{
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(partslot, &(((HeapTuple) tuple)->t_self), estate, false);
}
/* AFTER ROW UPDATE Triggers */
ExecARUpdateTriggers(estate, resultRelInfo, tupleid, tuple);
}
/* ----------------------------------------------------------------
* ExecInsert
*
* INSERTs have to add the tuple into
* the base relation and insert appropriate tuples into the
* index relations.
* Insert can be part of an update operation when
* there is a preceding SplitUpdate node.
* ----------------------------------------------------------------
*/
void
ExecInsert(TupleTableSlot *slot,
DestReceiver *dest,
EState *estate,
PlanGenerator planGen,
bool isUpdate)
{
void *tuple = NULL;
ResultRelInfo *resultRelInfo = NULL;
Relation resultRelationDesc = NULL;
Oid newId = InvalidOid;
TupleTableSlot *partslot = NULL;
AOTupleId aoTupleId = AOTUPLEID_INIT;
bool rel_is_heap = false;
bool rel_is_aorows = false;
bool rel_is_aocols = false;
bool rel_is_external = false;
/*
* get information on the (current) result relation
*/
if (estate->es_result_partitions)
{
resultRelInfo = slot_get_partition(slot, estate);
/* Check whether the user provided the correct leaf part only if required */
if (!dml_ignore_target_partition_check)
{
Assert(NULL != estate->es_result_partitions->part &&
NULL != resultRelInfo->ri_RelationDesc);
List *resultRelations = estate->es_plannedstmt->resultRelations;
/*
* Only inheritance can generate multiple result relations and inheritance
* is not compatible with partitions. As we are in inserting in partitioned
* table, we should not have more than one resultRelation
*/
Assert(list_length(resultRelations) == 1);
/* We only have one resultRelations entry where the user originally intended to insert */
int rteIdxForUserRel = linitial_int(resultRelations);
Assert (rteIdxForUserRel > 0);
Oid userProvidedRel = InvalidOid;
if (1 == rteIdxForUserRel)
{
/* Optimization for typical case */
userProvidedRel = ((RangeTblEntry *) estate->es_plannedstmt->rtable->head->data.ptr_value)->relid;
}
else
{
userProvidedRel = getrelid(rteIdxForUserRel, estate->es_plannedstmt->rtable);
}
/* Error out if user provides a leaf partition that does not match with our calculated partition */
if (userProvidedRel != estate->es_result_partitions->part->parrelid &&
userProvidedRel != resultRelInfo->ri_RelationDesc->rd_id)
{
ereport(ERROR,
(errcode(ERRCODE_CHECK_VIOLATION),
errmsg("Trying to insert row into wrong partition"),
errdetail("Expected partition: %s, provided partition: %s",
resultRelInfo->ri_RelationDesc->rd_rel->relname.data,
estate->es_result_relation_info->ri_RelationDesc->rd_rel->relname.data)));
}
}
estate->es_result_relation_info = resultRelInfo;
}
else
{
resultRelInfo = estate->es_result_relation_info;
}
Assert (!resultRelInfo->ri_projectReturning);
resultRelationDesc = resultRelInfo->ri_RelationDesc;
rel_is_heap = RelationIsHeap(resultRelationDesc);
rel_is_aocols = RelationIsAoCols(resultRelationDesc);
rel_is_aorows = RelationIsAoRows(resultRelationDesc);
rel_is_external = RelationIsExternal(resultRelationDesc);
partslot = reconstructMatchingTupleSlot(slot, resultRelInfo);
if (rel_is_heap)
{
tuple = ExecFetchSlotHeapTuple(partslot);
}
else if (rel_is_aorows)
{
tuple = ExecFetchSlotMemTuple(partslot, false);
}
else if (rel_is_external)
{
if (estate->es_result_partitions &&
estate->es_result_partitions->part->parrelid != 0)
{
ereport(ERROR,
(errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("Insert into external partitions not supported.")));
return;
}
else
{
tuple = ExecFetchSlotHeapTuple(partslot);
}
}
else
{
Assert(rel_is_aocols);
tuple = ExecFetchSlotMemTuple(partslot, true);
}
Assert(partslot != NULL && tuple != NULL);
/* Execute triggers in Planner-generated plans */
if (planGen == PLANGEN_PLANNER)
{
/* BEFORE ROW INSERT Triggers */
if (resultRelInfo->ri_TrigDesc &&
resultRelInfo->ri_TrigDesc->n_before_row[TRIGGER_EVENT_INSERT] > 0)
{
HeapTuple newtuple;
/* NYI */
if(rel_is_aocols)
elog(ERROR, "triggers are not supported on tables that use column-oriented storage");
newtuple = ExecBRInsertTriggers(estate, resultRelInfo, tuple);
if (newtuple == NULL) /* "do nothing" */
{
return;
}
if (newtuple != tuple) /* modified by Trigger(s) */
{
/*
* Put the modified tuple into a slot for convenience of routines
* below. We assume the tuple was allocated in per-tuple memory
* context, and therefore will go away by itself. The tuple table
* slot should not try to clear it.
*/
TupleTableSlot *newslot = estate->es_trig_tuple_slot;
if (newslot->tts_tupleDescriptor != partslot->tts_tupleDescriptor)
ExecSetSlotDescriptor(newslot, partslot->tts_tupleDescriptor);
ExecStoreGenericTuple(newtuple, newslot, false);
newslot->tts_tableOid = partslot->tts_tableOid; /* for constraints */
tuple = newtuple;
partslot = newslot;
}
}
}
/*
* Check the constraints of the tuple
*/
if (resultRelationDesc->rd_att->constr &&
planGen == PLANGEN_PLANNER)
{
ExecConstraints(resultRelInfo, partslot, estate);
}
/*
* insert the tuple
*
* Note: heap_insert returns the tid (location) of the new tuple in the
* t_self field.
*
* NOTE: for append-only relations we use the append-only access methods.
*/
if (rel_is_aorows)
{
if (resultRelInfo->ri_aoInsertDesc == NULL)
{
/* Set the pre-assigned fileseg number to insert into */
ResultRelInfoSetSegno(resultRelInfo, estate->es_result_aosegnos);
resultRelInfo->ri_aoInsertDesc =
appendonly_insert_init(resultRelationDesc,
ActiveSnapshot,
resultRelInfo->ri_aosegno,
false);
}
appendonly_insert(resultRelInfo->ri_aoInsertDesc, tuple, &newId, &aoTupleId);
}
else if (rel_is_aocols)
{
if (resultRelInfo->ri_aocsInsertDesc == NULL)
{
ResultRelInfoSetSegno(resultRelInfo, estate->es_result_aosegnos);
resultRelInfo->ri_aocsInsertDesc = aocs_insert_init(resultRelationDesc,
resultRelInfo->ri_aosegno, false);
}
newId = aocs_insert(resultRelInfo->ri_aocsInsertDesc, partslot);
aoTupleId = *((AOTupleId*)slot_get_ctid(partslot));
}
else if (rel_is_external)
{
/* Writable external table */
if (resultRelInfo->ri_extInsertDesc == NULL)
resultRelInfo->ri_extInsertDesc = external_insert_init(resultRelationDesc);
newId = external_insert(resultRelInfo->ri_extInsertDesc, tuple);
}
else
{
Insist(rel_is_heap);
newId = heap_insert(resultRelationDesc,
tuple,
estate->es_snapshot->curcid,
true, true, GetCurrentTransactionId());
}
IncrAppended();
(estate->es_processed)++;
(resultRelInfo->ri_aoprocessed)++;
estate->es_lastoid = newId;
partslot->tts_tableOid = RelationGetRelid(resultRelationDesc);
if (rel_is_aorows || rel_is_aocols)
{
/*
* insert index entries for AO Row-Store tuple
*/
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(partslot, (ItemPointer)&aoTupleId, estate, false);
}
else
{
/* Use parttuple for index update in case this is an indexed heap table. */
TupleTableSlot *xslot = partslot;
void *xtuple = tuple;
setLastTid(&(((HeapTuple) xtuple)->t_self));
/*
* insert index entries for tuple
*/
if (resultRelInfo->ri_NumIndices > 0)
ExecInsertIndexTuples(xslot, &(((HeapTuple) xtuple)->t_self), estate, false);
}
if (planGen == PLANGEN_PLANNER)
{
/* AFTER ROW INSERT Triggers */
ExecARInsertTriggers(estate, resultRelInfo, tuple);
}
}