/* ----------------------------------------------------------------
* ExternalNext
*
* This is a workhorse for ExecExtScan
* ----------------------------------------------------------------
*/
static TupleTableSlot *
ExternalNext(ExternalScanState *node)
{
HeapTuple tuple;
FileScanDesc scandesc;
Index scanrelid;
EState *estate;
ScanDirection direction;
TupleTableSlot *slot;
/*
* get information from the estate and scan state
*/
estate = node->ss.ps.state;
scandesc = node->ess_ScanDesc;
scanrelid = ((ExternalScan *) node->ss.ps.plan)->scan.scanrelid;
direction = estate->es_direction;
slot = node->ss.ss_ScanTupleSlot;
/*
* get the next tuple from the file access methods
*/
tuple = external_getnext(scandesc, direction);
/*
* save the tuple and the buffer returned to us by the access methods in
* our scan tuple slot and return the slot. Note: we pass 'false' because
* tuples returned by heap_getnext() are pointers onto disk pages and were
* not created with palloc() and so should not be pfree()'d. Note also
* that ExecStoreTuple will increment the refcount of the buffer; the
* refcount will not be dropped until the tuple table slot is cleared.
*/
if (tuple)
{
Gpmon_M_Incr_Rows_Out(GpmonPktFromExtScanState(node));
CheckSendPlanStateGpmonPkt(&node->ss.ps);
ExecStoreGenericTuple(tuple, slot, true);
/*
* CDB: Label each row with a synthetic ctid if needed for subquery dedup.
*/
if (node->cdb_want_ctid &&
!TupIsNull(slot))
{
slot_set_ctid_from_fake(slot, &node->cdb_fake_ctid);
}
}
else
{
ExecClearTuple(slot);
if (!node->ss.ps.delayEagerFree)
{
ExecEagerFreeExternalScan(node);
}
}
return slot;
}
/*
* tuplestore_gettupleslot - exported function to fetch a tuple into a slot
*
* If successful, put tuple in slot and return TRUE; else, clear the slot
* and return FALSE.
*
* If copy is TRUE, the slot receives a copied tuple (allocated in current
* memory context) that will stay valid regardless of future manipulations of
* the tuplestore's state. If copy is FALSE, the slot may just receive a
* pointer to a tuple held within the tuplestore. The latter is more
* efficient but the slot contents may be corrupted if additional writes to
* the tuplestore occur. (If using tuplestore_trim, see comments therein.)
*/
bool
tuplestore_gettupleslot(Tuplestorestate *state, bool forward,
bool copy, TupleTableSlot *slot)
{
GenericTuple tuple;
bool should_free;
tuple = tuplestore_gettuple(state, forward, &should_free);
if (tuple)
{
if (copy && !should_free)
{
if (is_memtuple(tuple))
tuple = (GenericTuple) memtuple_copy_to((MemTuple) tuple, NULL, NULL);
else
tuple = (GenericTuple) heap_copytuple((HeapTuple) tuple);
should_free = true;
}
ExecStoreGenericTuple(tuple, slot, should_free);
return true;
}
else
{
ExecClearTuple(slot);
return false;
}
}
static TupleTableSlot*
BitmapTableScanPlanQualTuple(BitmapTableScanState *node)
{
EState *estate = node->ss.ps.state;
Index scanrelid = ((BitmapTableScan *) node->ss.ps.plan)->scan.scanrelid;
ExprContext *econtext = node->ss.ps.ps_ExprContext;
TupleTableSlot *slot = node->ss.ss_ScanTupleSlot;
/*
* Check if we are evaluating PlanQual for tuple of this relation.
* Additional checking is not good, but no other way for now. We could
* introduce new nodes for this case and handle IndexScan --> NewNode
* switching in Init/ReScan plan...
*/
if (estate->es_evTuple != NULL &&
estate->es_evTuple[scanrelid - 1] != NULL)
{
if (estate->es_evTupleNull[scanrelid - 1])
{
return ExecClearTuple(slot);
}
ExecStoreGenericTuple(estate->es_evTuple[scanrelid - 1], slot, false);
/* Does the tuple meet the original qual conditions? */
econtext->ecxt_scantuple = slot;
ResetExprContext(econtext);
if (!ExecQual(node->bitmapqualorig, econtext, false))
{
ExecClearTuple(slot); /* would not be returned by scan */
}
/* Flag for the next call that no more tuples */
estate->es_evTupleNull[scanrelid - 1] = true;
return slot;
}
return ExecClearTuple(slot);
}
/* ----------------------------------------------------------------
* IndexNext
*
* Retrieve a tuple from the IndexScan node's currentRelation
* using the index specified in the IndexScanState information.
* ----------------------------------------------------------------
*/
TupleTableSlot *
IndexNext(IndexScanState *node)
{
EState *estate;
ExprContext *econtext;
ScanDirection direction;
IndexScanDesc scandesc;
Index scanrelid;
HeapTuple tuple;
TupleTableSlot *slot;
/*
* extract necessary information from index scan node
*/
estate = node->ss.ps.state;
direction = estate->es_direction;
initScanDesc(node);
/* flip direction if this is an overall backward scan */
if (ScanDirectionIsBackward(((IndexScan *) node->ss.ps.plan)->indexorderdir))
{
if (ScanDirectionIsForward(direction))
direction = BackwardScanDirection;
else if (ScanDirectionIsBackward(direction))
direction = ForwardScanDirection;
}
scandesc = node->iss_ScanDesc;
econtext = node->ss.ps.ps_ExprContext;
slot = node->ss.ss_ScanTupleSlot;
scanrelid = ((IndexScan *) node->ss.ps.plan)->scan.scanrelid;
/*
* Check if we are evaluating PlanQual for tuple of this relation.
* Additional checking is not good, but no other way for now. We could
* introduce new nodes for this case and handle IndexScan --> NewNode
* switching in Init/ReScan plan...
*/
if (estate->es_evTuple != NULL &&
estate->es_evTuple[scanrelid - 1] != NULL)
{
if (estate->es_evTupleNull[scanrelid - 1])
{
if (!node->ss.ps.delayEagerFree)
{
ExecEagerFreeIndexScan(node);
}
return ExecClearTuple(slot);
}
ExecStoreGenericTuple(estate->es_evTuple[scanrelid - 1], slot, false);
/* Does the tuple meet the indexqual condition? */
econtext->ecxt_scantuple = slot;
ResetExprContext(econtext);
if (!ExecQual(node->indexqualorig, econtext, false))
{
if (!node->ss.ps.delayEagerFree)
{
ExecEagerFreeIndexScan(node);
}
ExecClearTuple(slot); /* would not be returned by scan */
}
/* Flag for the next call that no more tuples */
estate->es_evTupleNull[scanrelid - 1] = true;
Gpmon_M_Incr_Rows_Out(GpmonPktFromIndexScanState(node));
CheckSendPlanStateGpmonPkt(&node->ss.ps);
return slot;
}
/*
* ok, now that we have what we need, fetch the next tuple.
*/
if ((tuple = index_getnext(scandesc, direction)) != NULL)
{
/*
* Store the scanned tuple in the scan tuple slot of the scan state.
* Note: we pass 'false' because tuples returned by amgetnext are
* pointers onto disk pages and must not be pfree()'d.
*/
ExecStoreHeapTuple(tuple, /* tuple to store */
slot, /* slot to store in */
scandesc->xs_cbuf, /* buffer containing tuple */
false); /* don't pfree */
Gpmon_M_Incr_Rows_Out(GpmonPktFromIndexScanState(node));
CheckSendPlanStateGpmonPkt(&node->ss.ps);
return slot;
}
if (!node->ss.ps.delayEagerFree)
{
ExecEagerFreeIndexScan(node);
}
/*
* if we get here it means the index scan failed so we are at the end of
* the scan..
*/
return ExecClearTuple(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);
}
}
TupleTableSlot *
HeapScanNext(ScanState *scanState)
{
Assert(IsA(scanState, TableScanState) ||
IsA(scanState, DynamicTableScanState));
SeqScanState *node = (SeqScanState *)scanState;
Assert(node->opaque != NULL);
HeapTuple tuple;
HeapScanDesc scandesc;
Index scanrelid;
EState *estate;
ScanDirection direction;
TupleTableSlot *slot;
Assert((node->ss.scan_state & SCAN_SCAN) != 0);
/*
* get information from the estate and scan state
*/
estate = node->ss.ps.state;
scandesc = node->opaque->ss_currentScanDesc;
scanrelid = ((SeqScan *) node->ss.ps.plan)->scanrelid;
direction = estate->es_direction;
slot = node->ss.ss_ScanTupleSlot;
/*
* Check if we are evaluating PlanQual for tuple of this relation.
* Additional checking is not good, but no other way for now. We could
* introduce new nodes for this case and handle SeqScan --> NewNode
* switching in Init/ReScan plan...
*/
if (estate->es_evTuple != NULL &&
estate->es_evTuple[scanrelid - 1] != NULL)
{
if (estate->es_evTupleNull[scanrelid - 1])
{
return ExecClearTuple(slot);
}
ExecStoreGenericTuple(estate->es_evTuple[scanrelid - 1], slot, false);
/*
* Note that unlike IndexScan, SeqScan never uses keys in
* heap_beginscan (and this is very bad) - so, here we do not check
* the keys.
*/
/* Flag for the next call that no more tuples */
estate->es_evTupleNull[scanrelid - 1] = true;
return slot;
}
/*
* get the next tuple from the access methods
*/
if (node->opaque->ss_heapTupleData.bot == node->opaque->ss_heapTupleData.top &&
!node->opaque->ss_heapTupleData.seen_EOS)
{
node->opaque->ss_heapTupleData.last = NULL;
node->opaque->ss_heapTupleData.bot = 0;
node->opaque->ss_heapTupleData.top = lengthof(node->opaque->ss_heapTupleData.item);
heap_getnextx(scandesc, direction, node->opaque->ss_heapTupleData.item,
&node->opaque->ss_heapTupleData.top,
&node->opaque->ss_heapTupleData.seen_EOS);
if (scandesc->rs_pageatatime &&
IsA(scanState, TableScanState))
{
Gpmon_M_Incr(GpmonPktFromTableScanState((TableScanState *)scanState), GPMON_TABLESCAN_PAGE);
CheckSendPlanStateGpmonPkt(&node->ss.ps);
}
}
node->opaque->ss_heapTupleData.last = NULL;
if (node->opaque->ss_heapTupleData.bot < node->opaque->ss_heapTupleData.top)
{
node->opaque->ss_heapTupleData.last =
&node->opaque->ss_heapTupleData.item[node->opaque->ss_heapTupleData.bot++];
}
tuple = node->opaque->ss_heapTupleData.last;
/*
* save the tuple and the buffer returned to us by the access methods in
* our scan tuple slot and return the slot. Note: we pass 'false' because
* tuples returned by heap_getnext() are pointers onto disk pages and were
* not created with palloc() and so should not be pfree()'d. Note also
* that ExecStoreTuple will increment the refcount of the buffer; the
* refcount will not be dropped until the tuple table slot is cleared.
*/
if (tuple)
{
ExecStoreHeapTuple(tuple,
slot,
scandesc->rs_cbuf,
false);
}
else
{
ExecClearTuple(slot);
}
return slot;
}
static TupleTableSlot *
execMotionUnsortedReceiver(MotionState * node)
{
/* RECEIVER LOGIC */
TupleTableSlot *slot;
HeapTuple tuple;
Motion *motion = (Motion *) node->ps.plan;
ReceiveReturnCode recvRC;
AssertState(motion->motionType == MOTIONTYPE_HASH ||
(motion->motionType == MOTIONTYPE_EXPLICIT && motion->segidColIdx > 0) ||
(motion->motionType == MOTIONTYPE_FIXED && motion->numOutputSegs <= 1));
Assert(node->ps.state->motionlayer_context);
Assert(node->ps.state->interconnect_context);
if (node->stopRequested)
{
SendStopMessage(node->ps.state->motionlayer_context,
node->ps.state->interconnect_context,
motion->motionID);
return NULL;
}
recvRC = RecvTupleFrom(node->ps.state->motionlayer_context,
node->ps.state->interconnect_context,
motion->motionID, &tuple, ANY_ROUTE);
if (recvRC == END_OF_STREAM)
{
#ifdef CDB_MOTION_DEBUG
if (gp_log_interconnect >= GPVARS_VERBOSITY_DEBUG)
elog(DEBUG4, "motionID=%d saw end of stream", motion->motionID);
#endif
Assert(node->numTuplesFromAMS == node->numTuplesToParent);
Assert(node->numTuplesFromChild == 0);
Assert(node->numTuplesToAMS == 0);
return NULL;
}
node->numTuplesFromAMS++;
node->numTuplesToParent++;
/* store it in our result slot and return this. */
slot = node->ps.ps_ResultTupleSlot;
slot = ExecStoreGenericTuple(tuple, slot, true /* shouldFree */);
#ifdef CDB_MOTION_DEBUG
if (node->numTuplesToParent <= 20)
{
StringInfoData buf;
initStringInfo(&buf);
appendStringInfo(&buf, " motion%-3d rcv %5d.",
motion->motionID,
node->numTuplesToParent);
formatTuple(&buf, tuple, ExecGetResultType(&node->ps),
node->outputFunArray);
elog(DEBUG3, buf.data);
pfree(buf.data);
}
#endif
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);
}
}
