/* ----------------------------------------------------------------
* ExecHash
*
* build hash table for hashjoin, doing partitioning if more
* than one batch is required.
* ----------------------------------------------------------------
*/
TupleTableSlot *
ExecHash(HashState *node)
{
EState *estate;
PlanState *outerNode;
List *hashkeys;
HashJoinTable hashtable;
TupleTableSlot *slot;
ExprContext *econtext;
int nbatch;
int i;
/*
* get state info from node
*/
estate = node->ps.state;
outerNode = outerPlanState(node);
hashtable = node->hashtable;
nbatch = hashtable->nbatch;
if (nbatch > 0)
{
/*
* Open temp files for inner batches, if needed. Note that file
* buffers are palloc'd in regular executor context.
*/
for (i = 0; i < nbatch; i++)
hashtable->innerBatchFile[i] = BufFileCreateTemp(false);
}
/*
* set expression context
*/
hashkeys = node->hashkeys;
econtext = node->ps.ps_ExprContext;
/*
* get all inner tuples and insert into the hash table (or temp files)
*/
for (;;)
{
slot = ExecProcNode(outerNode);
if (TupIsNull(slot))
break;
hashtable->hashNonEmpty = true;
econtext->ecxt_innertuple = slot;
ExecHashTableInsert(hashtable, econtext, hashkeys);
ExecClearTuple(slot);
}
/*
* Return the slot so that we have the tuple descriptor when we need
* to save/restore them. -Jeff 11 July 1991
*/
return slot;
}
/* ----------------------------------------------------------------
* MultiExecHash
*
* build hash table for hashjoin, doing partitioning if more
* than one batch is required.
* ----------------------------------------------------------------
*/
Node *
MultiExecHash(HashState *node)
{
PlanState *outerNode;
List *hashkeys;
HashJoinTable hashtable;
TupleTableSlot *slot;
ExprContext *econtext;
uint32 hashvalue;
/* must provide our own instrumentation support */
if (node->ps.instrument)
InstrStartNode(node->ps.instrument);
/*
* get state info from node
*/
outerNode = outerPlanState(node);
hashtable = node->hashtable;
/*
* set expression context
*/
hashkeys = node->hashkeys;
econtext = node->ps.ps_ExprContext;
/*
* get all inner tuples and insert into the hash table (or temp files)
*/
for (;;)
{
slot = ExecProcNode(outerNode);
if (TupIsNull(slot))
break;
hashtable->totalTuples += 1;
/* We have to compute the hash value */
econtext->ecxt_innertuple = slot;
hashvalue = ExecHashGetHashValue(hashtable, econtext, hashkeys);
ExecHashTableInsert(hashtable, slot, hashvalue);
}
/* must provide our own instrumentation support */
if (node->ps.instrument)
InstrStopNode(node->ps.instrument, hashtable->totalTuples);
/*
* We do not return the hash table directly because it's not a subtype of
* Node, and so would violate the MultiExecProcNode API. Instead, our
* parent Hashjoin node is expected to know how to fish it out of our node
* state. Ugly but not really worth cleaning up, since Hashjoin knows
* quite a bit more about Hash besides that.
*/
return NULL;
}
/* ----------------------------------------------------------------
* MultiExecHash
*
* build hash table for hashjoin, doing partitioning if more
* than one batch is required.
* ----------------------------------------------------------------
*/
Node *
MultiExecHash(HashState *node)
{
PlanState *outerNode;
List *hashkeys;
HashJoinTable hashtable;
TupleTableSlot *slot;
ExprContext *econtext;
uint32 hashvalue = 0;
/* must provide our own instrumentation support */
if (node->ps.instrument)
InstrStartNode(node->ps.instrument);
/*
* get state info from node
*/
outerNode = outerPlanState(node);
hashtable = node->hashtable;
/*
* set expression context
*/
hashkeys = node->hashkeys;
econtext = node->ps.ps_ExprContext;
#ifdef FAULT_INJECTOR
FaultInjector_InjectFaultIfSet(
MultiExecHashLargeVmem,
DDLNotSpecified,
"", // databaseName
""); // tableName
#endif
/*
* get all inner tuples and insert into the hash table (or temp files)
*/
for (;;)
{
slot = ExecProcNode(outerNode);
if (TupIsNull(slot))
break;
Gpmon_M_Incr(GpmonPktFromHashState(node), GPMON_QEXEC_M_ROWSIN);
CheckSendPlanStateGpmonPkt(&node->ps);
/* We have to compute the hash value */
econtext->ecxt_innertuple = slot;
bool hashkeys_null = false;
if (ExecHashGetHashValue(node, hashtable, econtext, hashkeys, false,
node->hs_keepnull, &hashvalue, &hashkeys_null))
{
ExecHashTableInsert(node, hashtable, slot, hashvalue);
}
if (hashkeys_null)
{
node->hs_hashkeys_null = true;
if (node->hs_quit_if_hashkeys_null)
{
ExecSquelchNode(outerNode);
return NULL;
}
}
}
/* Now we have set up all the initial batches & primary overflow batches. */
hashtable->nbatch_outstart = hashtable->nbatch;
/* must provide our own instrumentation support */
if (node->ps.instrument)
InstrStopNode(node->ps.instrument, hashtable->totalTuples);
/*
* We do not return the hash table directly because it's not a subtype of
* Node, and so would violate the MultiExecProcNode API. Instead, our
* parent Hashjoin node is expected to know how to fish it out of our node
* state. Ugly but not really worth cleaning up, since Hashjoin knows
* quite a bit more about Hash besides that.
*/
return NULL;
}
/* ----------------------------------------------------------------
* MultiExecHash
*
* build hash table for hashjoin, doing partitioning if more
* than one batch is required.
* ----------------------------------------------------------------
*/
Node *
MultiExecHash(HashState *node)
{
PlanState *outerNode;
List *hashkeys;
HashJoinTable hashtable;
TupleTableSlot *slot;
ExprContext *econtext;
uint32 hashvalue;
/* must provide our own instrumentation support */
if (node->ps.instrument)
InstrStartNode(node->ps.instrument);
/*
* get state info from node
*/
outerNode = outerPlanState(node);
hashtable = node->hashtable;
/*
* set expression context
*/
hashkeys = node->hashkeys;
econtext = node->ps.ps_ExprContext;
/*
* get all inner tuples and insert into the hash table (or temp files)
*/
for (;;)
{
slot = ExecProcNode(outerNode);
if (TupIsNull(slot))
break;
/* We have to compute the hash value */
econtext->ecxt_innertuple = slot;
if (ExecHashGetHashValue(hashtable, econtext, hashkeys,
false, hashtable->keepNulls,
&hashvalue))
{
int bucketNumber;
bucketNumber = ExecHashGetSkewBucket(hashtable, hashvalue);
if (bucketNumber != INVALID_SKEW_BUCKET_NO)
{
/* It's a skew tuple, so put it into that hash table */
ExecHashSkewTableInsert(hashtable, slot, hashvalue,
bucketNumber);
}
else
{
/* Not subject to skew optimization, so insert normally */
ExecHashTableInsert(hashtable, slot, hashvalue);
}
hashtable->totalTuples += 1;
}
}
/* must provide our own instrumentation support */
if (node->ps.instrument)
InstrStopNode(node->ps.instrument, hashtable->totalTuples);
/*
* We do not return the hash table directly because it's not a subtype of
* Node, and so would violate the MultiExecProcNode API. Instead, our
* parent Hashjoin node is expected to know how to fish it out of our node
* state. Ugly but not really worth cleaning up, since Hashjoin knows
* quite a bit more about Hash besides that.
*/
return NULL;
}
/* ----------------------------------------------------------------
* MultiExecHash
*
* build hash table for hashjoin, doing partitioning if more
* than one batch is required.
* ----------------------------------------------------------------
*/
node_n*
MultiExecHash(hash_ps* node)
{
plan_state_n* outerNode;
struct list* hashkeys;
struct hash_join_table* hashtable;
struct tupslot* slot;
expr_ctx_n* econtext;
uint32 hashvalue;
/* must provide our own instrumentation support */
if (node->ps.instrument)
instr_start_node(node->ps.instrument);
/*
* get state info from node
*/
outerNode = OUTER_PLAN_STATE(node);
hashtable = node->hashtable;
/*
* set expression context
*/
hashkeys = node->hashkeys;
econtext = node->ps.ps_ExprContext;
/*
* get all inner tuples and insert into the hash table (or temp files)
*/
for (;;) {
slot = exec_proc_node(outerNode);
if (TUPSLOT_NULL(slot))
break;
/* We have to compute the hash value */
econtext->ecxt_innertuple = slot;
if (ExecHashGetHashValue(hashtable,
econtext,
hashkeys,
false,
hashtable->keepNulls,
&hashvalue)) {
int bucketNumber;
bucketNumber = ExecHashGetSkewBucket(hashtable, hashvalue);
if (bucketNumber != INVALID_SKEW_BUCKET_NO) {
/* It's a skew tuple, so put it into that hash table */
ExecHashSkewTableInsert(hashtable, slot, hashvalue, bucketNumber);
} else {
/* Not subject to skew optimization, so insert normally */
ExecHashTableInsert(hashtable, slot, hashvalue);
}
hashtable->totalTuples += 1;
}
}
/* must provide our own instrumentation support */
if (node->ps.instrument)
instr_stop_node(node->ps.instrument, hashtable->totalTuples);
/*
* We do not return the hash table directly because it's not a subtype of
* node_n, and so would violate the multi_exec_proc_node API. Instead, our
* parent Hashjoin node is expected to know how to fish it out of our node
* state. Ugly but not really worth cleaning up, since Hashjoin knows
* quite a bit more about hash_pl besides that.
*/
return NULL;
}
/* ----------------------------------------------------------------
* ExecHash
*
* build hash table for hashjoin, all do partitioning if more
* than one batches are required.
* ----------------------------------------------------------------
*/
TupleTableSlot *
ExecHash(Hash *node)
{
EState *estate;
HashState *hashstate;
Plan *outerNode;
Var *hashkey;
HashJoinTable hashtable;
TupleTableSlot *slot;
ExprContext *econtext;
int nbatch;
File *batches;
RelativeAddr *batchPos;
int *batchSizes;
int i;
RelativeAddr *innerbatchNames;
/* ----------------
* get state info from node
* ----------------
*/
hashstate = node->hashstate;
estate = node->plan.state;
outerNode = outerPlan(node);
hashtable = node->hashtable;
if (hashtable == NULL)
elog(WARN, "ExecHash: hash table is NULL.");
nbatch = hashtable->nbatch;
if (nbatch > 0) { /* if needs hash partition */
innerbatchNames = (RelativeAddr *) ABSADDR(hashtable->innerbatchNames);
/* --------------
* allocate space for the file descriptors of batch files
* then open the batch files in the current processes.
* --------------
*/
batches = (File*)palloc(nbatch * sizeof(File));
for (i=0; i<nbatch; i++) {
batches[i] = FileNameOpenFile(ABSADDR(innerbatchNames[i]),
O_CREAT | O_RDWR, 0600);
}
hashstate->hashBatches = batches;
batchPos = (RelativeAddr*) ABSADDR(hashtable->innerbatchPos);
batchSizes = (int*) ABSADDR(hashtable->innerbatchSizes);
}
/* ----------------
* set expression context
* ----------------
*/
hashkey = node->hashkey;
econtext = hashstate->cstate.cs_ExprContext;
/* ----------------
* get tuple and insert into the hash table
* ----------------
*/
for (;;) {
slot = ExecProcNode(outerNode, (Plan*)node);
if (TupIsNull(slot))
break;
econtext->ecxt_innertuple = slot;
ExecHashTableInsert(hashtable, econtext, hashkey,
hashstate->hashBatches);
ExecClearTuple(slot);
}
/*
* end of build phase, flush all the last pages of the batches.
*/
for (i=0; i<nbatch; i++) {
if (FileSeek(batches[i], 0L, SEEK_END) < 0)
perror("FileSeek");
if (FileWrite(batches[i],ABSADDR(hashtable->batch)+i*BLCKSZ,BLCKSZ) < 0)
perror("FileWrite");
NDirectFileWrite++;
}
/* ---------------------
* Return the slot so that we have the tuple descriptor
* when we need to save/restore them. -Jeff 11 July 1991
* ---------------------
*/
return slot;
}
