/* ----------------------------------------------------------------
* 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;
}
/*
* BufFileWrite
*
* Like fwrite() except we assume 1-byte element size.
*/
size_t
BufFileWrite(BufFile *file, void *ptr, size_t size)
{
size_t nwritten = 0;
size_t nthistime;
while (size > 0)
{
if (file->pos >= BLCKSZ)
{
/* Buffer full, dump it out */
if (file->dirty)
{
/* by cywang */
InstrStartNode(file->instr_filedump);
BufFileDumpBuffer(file);
InstrStopNode(file->instr_filedump, 0);
file->fileDumpNum++;
if (file->dirty)
break; /* I/O error */
}
else
{
/* Hmm, went directly from reading to writing? */
file->curOffset += file->pos;
file->pos = 0;
file->nbytes = 0;
}
}
nthistime = BLCKSZ - file->pos;
if (nthistime > size)
nthistime = size;
Assert(nthistime > 0);
memcpy(file->buffer + file->pos, ptr, nthistime);
file->dirty = true;
file->pos += nthistime;
if (file->nbytes < file->pos)
file->nbytes = file->pos;
ptr = (void *) ((char *) ptr + nthistime);
size -= nthistime;
nwritten += nthistime;
}
return nwritten;
}
/*
* BufFileLoadBuffer
*
* Load some data into buffer, if possible, starting from curOffset.
* At call, must have dirty = false, pos and nbytes = 0.
* On exit, nbytes is number of bytes loaded.
*/
static void
BufFileLoadBuffer(BufFile *file)
{
File thisfile;
/*
* Advance to next component file if necessary and possible.
*
* This path can only be taken if there is more than one component, so it
* won't interfere with reading a non-temp file that is over
* MAX_PHYSICAL_FILESIZE.
*/
if (file->curOffset >= MAX_PHYSICAL_FILESIZE &&
file->curFile + 1 < file->numFiles)
{
file->curFile++;
file->curOffset = 0L;
}
/*
* May need to reposition physical file.
*/
thisfile = file->files[file->curFile];
if (file->curOffset != file->offsets[file->curFile])
{
if (FileSeek(thisfile, file->curOffset, SEEK_SET) != file->curOffset)
return; /* seek failed, read nothing */
file->offsets[file->curFile] = file->curOffset;
}
/* add by cywang */
InstrStartNode(file->instr_fileload);
/*
* Read whatever we can get, up to a full bufferload.
*/
file->nbytes = FileRead(thisfile, file->buffer, sizeof(file->buffer));
InstrStopNode(file->instr_fileload, 0);
file->fileLoadNum++;
if (file->nbytes < 0)
file->nbytes = 0;
file->offsets[file->curFile] += file->nbytes;
/* we choose not to advance curOffset here */
pgBufferUsage.temp_blks_read++;
}
/*
* BufFileFlush
*
* Like fflush()
*/
static int
BufFileFlush(BufFile *file)
{
if (file->dirty)
{
/* by cywang */
InstrStartNode(file->instr_filedump);
BufFileDumpBuffer(file);
InstrStopNode(file->instr_filedump, 0);
file->fileDumpNum++;
if (file->dirty)
return EOF;
}
return 0;
}
/* ----------------------------------------------------------------
* MultiExecBitmapIndexScan(node)
* ----------------------------------------------------------------
*/
Node *
MultiExecBitmapIndexScan(BitmapIndexScanState *node)
{
TIDBitmap *tbm;
IndexScanDesc scandesc;
double nTuples = 0;
bool doscan;
/* must provide our own instrumentation support */
if (node->ss.ps.instrument)
InstrStartNode(node->ss.ps.instrument);
/*
* extract necessary information from index scan node
*/
scandesc = node->biss_ScanDesc;
/*
* If we have runtime keys and they've not already been set up, do it now.
* Array keys are also treated as runtime keys; note that if ExecReScan
* returns with biss_RuntimeKeysReady still false, then there is an empty
* array key so we should do nothing.
*/
if (!node->biss_RuntimeKeysReady &&
(node->biss_NumRuntimeKeys != 0 || node->biss_NumArrayKeys != 0))
{
ExecReScan((PlanState *) node);
doscan = node->biss_RuntimeKeysReady;
}
else
doscan = true;
/*
* Prepare the result bitmap. Normally we just create a new one to pass
* back; however, our parent node is allowed to store a pre-made one into
* node->biss_result, in which case we just OR our tuple IDs into the
* existing bitmap. (This saves needing explicit UNION steps.)
*/
if (node->biss_result)
{
tbm = node->biss_result;
node->biss_result = NULL; /* reset for next time */
}
else
{
/* XXX should we use less than work_mem for this? */
tbm = tbm_create(work_mem * 1024L);
}
/*
* Get TIDs from index and insert into bitmap
*/
while (doscan)
{
nTuples += (double) index_getbitmap(scandesc, tbm);
CHECK_FOR_INTERRUPTS();
doscan = ExecIndexAdvanceArrayKeys(node->biss_ArrayKeys,
node->biss_NumArrayKeys);
if (doscan) /* reset index scan */
index_rescan(node->biss_ScanDesc,
node->biss_ScanKeys, node->biss_NumScanKeys,
NULL, 0);
}
/* must provide our own instrumentation support */
if (node->ss.ps.instrument)
InstrStopNode(node->ss.ps.instrument, nTuples);
return (Node *) tbm;
}
/* ----------------------------------------------------------------
* 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;
}
/* ----------------------------------------------------------------
* MultiExecBitmapIndexScan(node)
* ----------------------------------------------------------------
*/
Node *
MultiExecBitmapIndexScan(BitmapIndexScanState *node)
{
IndexScanState *scanState = (IndexScanState*)node;
Node *bitmap = NULL;
/* must provide our own instrumentation support */
if (scanState->ss.ps.instrument)
{
InstrStartNode(scanState->ss.ps.instrument);
}
bool partitionIsReady = DynamicScan_BeginIndexPartition(scanState, false /* initQual */,
false /* initTargetList */, true /* supportsArrayKeys */,
true /* isMultiScan */);
Assert(partitionIsReady);
if (!partitionIsReady)
{
DynamicScan_EndIndexPartition(scanState);
return NULL;
}
bool doscan = node->indexScanState.iss_RuntimeKeysReady;
IndexScanDesc scandesc = scanState->iss_ScanDesc;
/* Get bitmap from index */
while (doscan)
{
bitmap = index_getmulti(scandesc, node->bitmap);
if ((NULL != bitmap) &&
!(IsA(bitmap, HashBitmap) || IsA(bitmap, StreamBitmap)))
{
elog(ERROR, "unrecognized result from bitmap index scan");
}
CHECK_FOR_INTERRUPTS();
/* CDB: If EXPLAIN ANALYZE, let bitmap share our Instrumentation. */
if (scanState->ss.ps.instrument)
{
tbm_bitmap_set_instrument(bitmap, scanState->ss.ps.instrument);
}
if(node->bitmap == NULL)
{
node->bitmap = (Node *)bitmap;
}
doscan = ExecIndexAdvanceArrayKeys(scanState->iss_ArrayKeys,
scanState->iss_NumArrayKeys);
if (doscan)
{
/* reset index scan */
index_rescan(scanState->iss_ScanDesc, scanState->iss_ScanKeys);
}
}
DynamicScan_EndIndexPartition(scanState);
/* must provide our own instrumentation support */
if (scanState->ss.ps.instrument)
{
InstrStopNode(scanState->ss.ps.instrument, 1 /* nTuples */);
}
return (Node *) bitmap;
}
/*
* RouterExecutorRun actually executes a single task on a worker.
*/
void
RouterExecutorRun(QueryDesc *queryDesc, ScanDirection direction, long count, Task *task)
{
EState *estate = queryDesc->estate;
CmdType operation = queryDesc->operation;
MemoryContext oldcontext = NULL;
DestReceiver *destination = queryDesc->dest;
MaterialState *routerState = (MaterialState *) queryDesc->planstate;
bool sendTuples = operation == CMD_SELECT || queryDesc->plannedstmt->hasReturning;
Assert(estate != NULL);
Assert(!(estate->es_top_eflags & EXEC_FLAG_EXPLAIN_ONLY));
Assert(task != NULL);
/* we only support default scan direction and row fetch count */
if (!ScanDirectionIsForward(direction))
{
ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED),
errmsg("scan directions other than forward scans "
"are unsupported")));
}
oldcontext = MemoryContextSwitchTo(estate->es_query_cxt);
if (queryDesc->totaltime != NULL)
{
InstrStartNode(queryDesc->totaltime);
}
estate->es_processed = 0;
/* startup the tuple receiver */
if (sendTuples)
{
(*destination->rStartup)(destination, operation, queryDesc->tupDesc);
}
/*
* If query has not yet been executed, do so now. The main reason why the
* query might already have been executed is cursors.
*/
if (!routerState->eof_underlying)
{
bool resultsOK = false;
bool isModificationQuery = false;
if (operation == CMD_INSERT || operation == CMD_UPDATE ||
operation == CMD_DELETE)
{
isModificationQuery = true;
}
else if (operation != CMD_SELECT)
{
ereport(ERROR, (errmsg("unrecognized operation code: %d",
(int) operation)));
}
resultsOK = ExecuteTaskAndStoreResults(queryDesc, task,
isModificationQuery,
sendTuples);
if (!resultsOK)
{
ereport(ERROR, (errmsg("could not receive query results")));
}
/* mark underlying query as having executed */
routerState->eof_underlying = true;
}
/* if the underlying query produced output, return it */
if (routerState->tuplestorestate != NULL)
{
TupleDesc resultTupleDescriptor = queryDesc->tupDesc;
int64 returnedRows = 0;
/* return rows from the tuplestore */
returnedRows = ReturnRowsFromTuplestore(count, resultTupleDescriptor,
destination,
routerState->tuplestorestate);
/*
* Count tuples processed, if this is a SELECT. (For modifications
* it'll already have been increased, as we want the number of
* modified tuples, not the number of RETURNed tuples.)
*/
if (operation == CMD_SELECT)
{
estate->es_processed += returnedRows;
}
}
/* shutdown tuple receiver, if we started it */
if (sendTuples)
{
(*destination->rShutdown)(destination);
}
if (queryDesc->totaltime != NULL)
{
InstrStopNode(queryDesc->totaltime, estate->es_processed);
}
MemoryContextSwitchTo(oldcontext);
}