void ExHdfsFastExtractTcb::insertUpQueueEntry(ex_queue::up_status status, ComDiagsArea *diags, NABoolean popDownQueue)
{
ex_queue_entry *upEntry = qParent_.up->getTailEntry();
ex_queue_entry *downEntry = qParent_.down->getHeadEntry();
ExFastExtractPrivateState &privateState = *((ExFastExtractPrivateState *) downEntry->pstate);
// Initialize the up queue entry.
//
// copyAtp() will copy all tuple pointers and the diags area from
// the down queue entry to the up queue entry.
//
// When we return Q_NO_DATA if the match count is > 0:
// * assume down queue diags were returned with the Q_OK_MMORE entries
// * release down queue diags before copyAtp()
//
if (status == ex_queue::Q_NO_DATA && privateState.matchCount_ > 0)
{
downEntry->setDiagsArea(NULL);
upEntry->copyAtp(downEntry);
}
else
{
upEntry->copyAtp(downEntry);
downEntry->setDiagsArea(NULL);
}
upEntry->upState.status = status;
upEntry->upState.parentIndex = downEntry->downState.parentIndex;
upEntry->upState.downIndex = qParent_.down->getHeadIndex();
upEntry->upState.setMatchNo(privateState.matchCount_);
// rows affected code (below) medeled after ex_partn_access.cpp
ExMasterStmtGlobals *g = getGlobals()->castToExExeStmtGlobals()->castToExMasterStmtGlobals();
if (!g)
{
ComDiagsArea *da = upEntry->getDiagsArea();
if (da == NULL)
{
da = ComDiagsArea::allocate(getGlobals()->getDefaultHeap());
upEntry->setDiagsArea(da);
}
da->addRowCount(privateState.matchCount_);
}
else
{
g->setRowsAffected(privateState.matchCount_);
}
//
// Insert into up queue
qParent_.up->insert();
// Optionally remove the head of the down queue
if (popDownQueue)
{
privateState.init();
qParent_.down->removeHead();
}
}
short ExTupleFlowTcb::work()
{
// This is some sort of a hack to fix the problems with the number of rows
// inserted returned to the user for packed tables. For these tables, rows
// are packed (by the Pack node which is the left child of this tuple flow)
// before they are sent off to DP2. DP2 has no idea that it's actually
// inserting multiple logical rows (as perceived by the user). However,
// there is actually a hidden count of logical rows stored as the first 4
// bytes of the packed row. This counter is supposed to keep track of a sum
// of this number in each packed row it gets from the left. When all
// insertions are done, this sum is used to override what's stored by the
// PA node in the executor global area the number of rows inserted. This is
// not a very good place to have this fix, but since this is a low-priority
// problem at this time, here we are.
//
//
// NB: The code introduced for this fix
// could be safely removed if desired. Also, all changes are within
// this file.
//
if (qParent_.down->isEmpty())
return WORK_OK;
ex_queue_entry * pentry_down = qParent_.down->getHeadEntry();
ExTupleFlowPrivateState & pstate =
*((ExTupleFlowPrivateState*) pentry_down->pstate);
if ((tflowTdb().userSidetreeInsert()) &&
(pentry_down->downState.request == ex_queue::GET_EOD) &&
(NOT pstate.parentEOD_))
{
pstate.step_ = MOVE_EOD_TO_TGT_;
}
else if ((pstate.step_ != DONE_) &&
(pstate.step_ != CANCELLED_) &&
(pentry_down->downState.request == ex_queue::GET_NOMORE))
{
if (pstate.step_ == EMPTY_)
pstate.step_ = DONE_;
else
pstate.step_ = CANCELLED_;
}
while (1)
{
switch (pstate.step_)
{
case EMPTY_:
{
if (qSrc_.down->isFull())
return WORK_OK;
ex_queue_entry * src_entry = qSrc_.down->getTailEntry();
src_entry->downState.request = pentry_down->downState.request;
src_entry->downState.requestValue =
pentry_down->downState.requestValue;
if ((tflowTdb().firstNRows() >= 0) &&
(pentry_down->downState.request != ex_queue::GET_N))
{
src_entry->downState.request = ex_queue::GET_N;
src_entry->downState.requestValue = tflowTdb().firstNRows();
}
src_entry->downState.parentIndex =
qParent_.down->getHeadIndex();
src_entry->passAtp(pentry_down);
qSrc_.down->insert();
// just checking to make sure we got a diags area from the CLI if we are
// executing a non-tomic insert. This is done now so that we don't have to do it in multiple
// places later.
if (tflowTdb().isNonFatalErrorTolerated()) {
ComDiagsArea *cliDiagsArea = pentry_down->getDiagsArea();
ex_assert(cliDiagsArea, "In Tupleflow : Non-Atomic insert received no diags area from the CLI");
}
pstate.parentEOD_ = FALSE;
pstate.srcEOD_ = FALSE;
pstate.matchCount_ = 0;
pstate.tgtRequests_ = 0;
pstate.tgtRowsSent_ = FALSE;
pstate.noOfUnPackedRows_ = 0;
pstate.srcRequestCount_ = -1;
pstate.nonFatalErrorSeen_ = FALSE;
// Set startRightIndex_ so that CancelReques doesn't do anything.
pstate.startRightIndex_ = pstate.srcRequestCount_;
pstate.step_ = MOVE_SRC_TO_TGT_;
}
break;
case MOVE_SRC_TO_TGT_:
{
// if there are some rows in source up queue, move them to target.
while ((! qSrc_.up->isEmpty()) && (! qTgt_.down->isFull())
&& (pstate.step_ != HANDLE_ERROR_))
{
ex_queue_entry * src_entry = qSrc_.up->getHeadEntry();
ex_queue_entry * tgt_entry = qTgt_.down->getTailEntry();
switch (src_entry->upState.status)
{
case ex_queue::Q_OK_MMORE:
{
// move this source row to target.
// LCOV_EXCL_START
// BEGIN: - Read note at beginning of work().
//
if (tcbSrc_->getNodeType() == ComTdb::ex_PACKROWS)
{
char* packTuppPtr =
src_entry->getTupp(src_entry->numTuples()-1)
.getDataPointer();
Int32 noOfRows = *((Int32 *)packTuppPtr);
pstate.noOfUnPackedRows_ += (noOfRows - 1);
}
//
// END:- Read note at beginning of work().
// LCOV_EXCL_STOP
pstate.srcRequestCount_++;
tgt_entry->downState.request =
pentry_down->downState.request;
tgt_entry->downState.requestValue =
pentry_down->downState.requestValue;
tgt_entry->downState.parentIndex =
(Lng32) pstate.srcRequestCount_;
tgt_entry->copyAtp(src_entry);
qTgt_.down->insert();
pstate.tgtRequests_++;
pstate.tgtRowsSent_ = TRUE;
qSrc_.up->removeHead();
}
break;
case ex_queue::Q_NO_DATA:
{
if ((tflowTdb().vsbbInsertOn()) &&
(pstate.tgtRowsSent_ == TRUE))
{
if (tflowTdb().userSidetreeInsert())
{
tgt_entry->downState.request =
ex_queue::GET_EOD_NO_ST_COMMIT;
}
else
{
tgt_entry->downState.request =
ex_queue::GET_EOD;
}
tgt_entry->downState.requestValue =
pentry_down->downState.requestValue;
tgt_entry->downState.parentIndex =
(Lng32) pstate.srcRequestCount_;
tgt_entry->copyAtp(src_entry);
qTgt_.down->insert();
pstate.tgtRequests_++;
}
// LCOV_EXCL_START
if ((pstate.tgtRowsSent_ == FALSE) &&
(src_entry->getDiagsArea()))
{
// a warning is returned with EOD and
// nothing else was returned from source.
// Move warning to parent's up queue.
if (qParent_.up->isFull())
return WORK_OK;
ex_queue_entry * up_entry =
qParent_.up->getTailEntry();
up_entry->setDiagsArea(src_entry->getDiagsArea());
}
// LCOV_EXCL_STOP
qSrc_.up->removeHead();
pstate.srcEOD_ = TRUE;
// LCOV_EXCL_START
if (tflowTdb().sendEODtoTgt())
pstate.step_ = MOVE_EOD_TO_TGT_;
// LCOV_EXCL_STOP
}
break;
case ex_queue::Q_SQLERROR:
{
if (qParent_.up->isFull())
return WORK_OK;
ex_queue_entry * pentry = qParent_.up->getTailEntry();
ComDiagsArea * da = src_entry->getDiagsArea();
ex_assert(da, "We have a Q_SQLERROR in Tupleflow but no diags area");
if (tflowTdb().isNonFatalErrorTolerated() &&
(da->getNextRowNumber(ComCondition::NONFATAL_ERROR) ==
ComCondition::NONFATAL_ERROR))
{
pstate.nonFatalErrorSeen_ = TRUE;
}
else
{
pstate.step_ = HANDLE_ERROR_;
pstate.nonFatalErrorSeen_ = FALSE;
}
pstate.srcRequestCount_++;
if(tflowTdb().isRowsetIterator())
da->setAllRowNumber((Lng32) pstate.srcRequestCount_);
ComDiagsArea *accumulatedDiagsArea = pentry->getDiagsArea();
if (accumulatedDiagsArea)
{
accumulatedDiagsArea->mergeAfter(*da);
if (!(accumulatedDiagsArea->canAcceptMoreErrors()) &&
tflowTdb().isNonFatalErrorTolerated())
{
pstate.nonFatalErrorSeen_ = FALSE;
pstate.step_ = HANDLE_ERROR_;
}
}
else
{
pentry->setDiagsArea(da);
da->incrRefCount();
accumulatedDiagsArea = da ;
if (tflowTdb().isNonFatalErrorTolerated())
{
ComDiagsArea *cliDiagsArea = pentry_down->getDiagsArea();
da->setLengthLimit(cliDiagsArea->getLengthLimit());
}
}
// For Non-Fatal errors we will remove this Q_SQLERROR reply from the
// left child right below as we will continue to stay in this state (MOVE_SRC_TO_TGT_).
// For fatal errors this Q_SQLERROR reply is removed in HANDLE_ERROR step to which
// we will transition immediately.
if (pstate.nonFatalErrorSeen_ == TRUE)
qSrc_.up->removeHead();
}
break;
case ex_queue::Q_REC_SKIPPED:
{
pstate.srcRequestCount_++;
ComDiagsArea * da = src_entry->getDiagsArea();
if (da)
pstate.nonFatalErrorSeen_ = TRUE;
qSrc_.up->removeHead();
}
break;
default:
{
ex_assert(0, "ExTupleFlowTcb::work() Error returned from src"); // LCOV_EXCL_LINE
}
break;
} // switch
} // while
// if the child reply is not an Q_SQLERROR, then process target
if ((pstate.step_ != HANDLE_ERROR_) &&
(pstate.step_ != MOVE_EOD_TO_TGT_))
pstate.step_ = PROCESS_TGT_;
} // MOVE_SRC_TO_TGT
break;
case MOVE_EOD_TO_TGT_:
{
pstate.parentEOD_ = TRUE;
if (qTgt_.down->isFull())
return WORK_OK;
ex_queue_entry * tgt_entry = qTgt_.down->getTailEntry();
tgt_entry->downState.request = ex_queue::GET_EOD;
tgt_entry->downState.requestValue =
pentry_down->downState.requestValue;
tgt_entry->downState.parentIndex =
qParent_.down->getHeadIndex();
//tgt_entry->passAtp(pentry_down);
qTgt_.down->insert();
pstate.tgtRequests_++;
// LCOV_EXCL_START
if (tflowTdb().sendEODtoTgt())
pstate.srcEOD_ = TRUE;
// LCOV_EXCL_STOP
else
pstate.srcEOD_ = FALSE;
pstate.step_ = PROCESS_TGT_;
}
break;
case PROCESS_TGT_:
{
while (! qTgt_.up->isEmpty() && pstate.step_ != HANDLE_ERROR_)
{
ex_queue_entry * tgt_entry = qTgt_.up->getHeadEntry();
switch (tgt_entry->upState.status)
{
case ex_queue::Q_OK_MMORE:
{
if (!tflowTdb().isNonFatalErrorTolerated())
{
// ex_assert(0, "ExTupleFlowTcb::work() OK_MMORE from tgt");
if (qParent_.up->isFull())
return WORK_OK;
ex_queue_entry * pentry = qParent_.up->getTailEntry();
pentry->upState.status = ex_queue::Q_OK_MMORE;
pentry->upState.downIndex = qParent_.down->getHeadIndex();
pentry->upState.parentIndex = pentry_down->downState.parentIndex;
pentry->upState.setMatchNo(pstate.matchCount_);
// copy input tupps from parent request
pentry->copyAtp(pentry_down);
// copy child's atp to
// the output atp (to parent's up queue)
pentry->copyAtp(tgt_entry);
// insert into parent up queue
qParent_.up->insert();
}
else
{
ComDiagsArea * da = tgt_entry->getDiagsArea();
ex_assert(da, "We have a Q_OK_MMORE in Tupleflow but no diags area");
if (da->mainSQLCODE() != 0) {
// Non-atomic Rowsets sends OK_MMORE with non-empty diags from child
// empty diags (mainsqlcode == 0) implies OK_MMORE sent by ignoreDupKey code
// when NAR is on, for -8102 error. Just consume the OK_MMORE.
if(tflowTdb().isRowsetIterator())
{
da->setAllRowNumber(Lng32 (tgt_entry->upState.parentIndex));
}
pstate.nonFatalErrorSeen_ = TRUE;
ex_queue_entry * pentry = qParent_.up->getTailEntry();
ComDiagsArea *accumulatedDiagsArea = pentry->getDiagsArea();
if (accumulatedDiagsArea)
{
accumulatedDiagsArea->mergeAfter(*da);
if (!(accumulatedDiagsArea->canAcceptMoreErrors()) &&
tflowTdb().isNonFatalErrorTolerated())
{
pstate.nonFatalErrorSeen_ = FALSE;
pstate.step_ = HANDLE_ERROR_;
}
}
else
{
pentry->setDiagsArea(da);
da->incrRefCount();
if (tflowTdb().isNonFatalErrorTolerated()) {
ComDiagsArea *cliDiagsArea = pentry_down->getDiagsArea();
da->setLengthLimit(cliDiagsArea->getLengthLimit());
}
}
}
}
qTgt_.up->removeHead();
}
break;
case ex_queue::Q_NO_DATA:
{
ComDiagsArea * da = tgt_entry->getDiagsArea();
if (da)
{
ex_queue_entry * pentry = qParent_.up->getTailEntry();
ComDiagsArea *accumulatedDiagsArea = pentry->getDiagsArea();
if (accumulatedDiagsArea)
accumulatedDiagsArea->mergeAfter(*da);
else
{
pentry->setDiagsArea(da);
da->incrRefCount();
if (tflowTdb().isNonFatalErrorTolerated()) {
ComDiagsArea *cliDiagsArea = pentry_down->getDiagsArea();
da->setLengthLimit(cliDiagsArea->getLengthLimit());
}
}
}
pstate.matchCount_ += tgt_entry->upState.getMatchNo();
qTgt_.up->removeHead();
pstate.tgtRequests_--;
pstate.startRightIndex_++;
}
break;
case ex_queue::Q_SQLERROR:
{
if (qParent_.up->isFull())
return WORK_OK;
ex_queue_entry * pentry = qParent_.up->getTailEntry();
pentry->copyAtp(tgt_entry);
pstate.nonFatalErrorSeen_ = FALSE;
pstate.step_ = HANDLE_ERROR_;
if(tflowTdb().isRowsetIterator())
{
ex_queue_entry * pentry = qParent_.up->getTailEntry();
ComDiagsArea *da = pentry->getDiagsArea();
ex_assert(da, "To set RowNumber, an error condition must be present in the diags area");
da->setAllRowNumber(Lng32 (tgt_entry->upState.parentIndex));
}
}
break;
default:
{
ex_assert(0, "ExTupleFlowTcb::work() Error returned from tgt"); // LCOV_EXCL_LINE
}
break;
} // switch
} // while
if (pstate.step_ == HANDLE_ERROR_)
break;
// if source has returned EOD,
// and there are no pending requests in target's down
// queue, then we are done with this parent request.
if (((pstate.srcEOD_ == TRUE) ||
(pstate.parentEOD_ == TRUE)) &&
(qTgt_.down->isEmpty()))
pstate.step_ = DONE_;
else
{
if (NOT pstate.parentEOD_)
pstate.step_ = MOVE_SRC_TO_TGT_;
if (qSrc_.up->isEmpty() || qTgt_.down->isFull())
return WORK_OK;
else
return WORK_CALL_AGAIN;
}
}
break;
case HANDLE_ERROR_:
{
ex_queue_entry * pentry = qParent_.up->getTailEntry();
pentry->upState.status = ex_queue::Q_SQLERROR;
pentry->upState.downIndex = qParent_.down->getHeadIndex();
pentry->upState.parentIndex = pentry_down->downState.parentIndex;
pentry->upState.setMatchNo(pstate.matchCount_);
ComDiagsArea *da = pentry->getDiagsArea();
if (tflowTdb().isNonFatalErrorTolerated() &&
!(da->canAcceptMoreErrors())) {
ComDiagsArea *cliDiagsArea = pentry_down->getDiagsArea();
da->removeLastErrorCondition();
*da << DgSqlCode(-EXE_NONATOMIC_FAILURE_LIMIT_EXCEEDED)
<< DgInt0(cliDiagsArea->getLengthLimit());
}
// insert into parent up queue
qParent_.up->insert();
pstate.step_ = CANCELLED_;
}
break;
case CANCELLED_:
{
qSrc_.down->cancelRequestWithParentIndex(qParent_.down->getHeadIndex());
// Cancel all the outstanding requests that have been sent to the target.
// Cancel all requests within given range (inclusive)
qTgt_.down->cancelRequestWithParentIndexRange((queue_index)pstate.startRightIndex_+1,
(queue_index)pstate.srcRequestCount_);
pstate.startRightIndex_ = pstate.srcRequestCount_;
//ignore all rows from source child, till Q_NO_DATA is reached
while ((pstate.srcEOD_ != TRUE) && (!qSrc_.up->isEmpty()))
{
ex_queue_entry * src_entry = qSrc_.up->getHeadEntry();
switch(src_entry->upState.status)
{
case ex_queue::Q_OK_MMORE:
case ex_queue::Q_SQLERROR:
case ex_queue::Q_REC_SKIPPED:
{
qSrc_.up->removeHead();
}
break;
case ex_queue::Q_NO_DATA:
{
pstate.srcEOD_ = TRUE;
qSrc_.up->removeHead();
}
break;
default:
{
ex_assert(0, "ExTupleFlowTcb::work() Error returned from src"); // LCOV_EXCL_LINE
}
break;
}
}
//ignore all rows from target child, till Q_NO_DATA is reached
while (pstate.tgtRequests_ && !qTgt_.up->isEmpty())
{
ex_queue_entry * tgt_entry = qTgt_.up->getHeadEntry();
switch(tgt_entry->upState.status)
{
case ex_queue::Q_OK_MMORE:
case ex_queue::Q_SQLERROR:
{
qTgt_.up->removeHead();
}
break;
case ex_queue::Q_NO_DATA:
{
qTgt_.up->removeHead();
pstate.tgtRequests_--;
}
break;
default:
{
ex_assert(0, "ExTupleFlowTcb::work() Error returned from tgt"); // LCOV_EXCL_LINE
}
break;
}
}
// if both source and target returned all the rows,
// insert Q_SQLERROR into the parent up queue
if ((pstate.srcEOD_ == TRUE) && !pstate.tgtRequests_)
{
pstate.step_ = DONE_;
}
else
return WORK_OK;
}
break;
case DONE_:
{
if (qParent_.up->isFull())
return WORK_OK;
ex_queue_entry * pentry = qParent_.up->getTailEntry();
if (pstate.nonFatalErrorSeen_) {
ComDiagsArea *da = pentry->getDiagsArea();
ComDiagsArea *cliDiagsArea = pentry_down->getDiagsArea();
ex_assert((da || cliDiagsArea), "We have non-fatal errors in Tupleflow but no diags area");
if (cliDiagsArea) {
if (da)
da->mergeAfter(*cliDiagsArea);
else
{
pentry->setDiagsArea(cliDiagsArea);
cliDiagsArea->incrRefCount();
}
}
if (cliDiagsArea->canAcceptMoreErrors()) {
ComDiagsArea *mergedDiagsArea = pentry->getDiagsArea();
// used to make mainSQLCODE() return 30022 or 30035.
mergedDiagsArea->setNonFatalErrorSeen(TRUE);
NABoolean anyRowsAffected = FALSE;
// This tupleflow should be in the master for
// non-atomic rowsets.
ExMasterStmtGlobals *g = getGlobals()->
castToExExeStmtGlobals()->castToExMasterStmtGlobals();
ex_assert(g, "Rowset insert has a flow node that is not in the master executor");
if (g->getRowsAffected() > 0)
anyRowsAffected = TRUE;
if (anyRowsAffected)
*mergedDiagsArea << DgSqlCode(EXE_NONFATAL_ERROR_SEEN);
else
*mergedDiagsArea << DgSqlCode(EXE_NONFATAL_ERROR_ON_ALL_ROWS);
} // we exceeded the Nonfatal error limit when merging with the CLI diags area
else {
pstate.step_ = HANDLE_ERROR_;
// will prevent us from merging the diags areas again
pstate.nonFatalErrorSeen_ = FALSE ;
break ;
}
}
pentry->upState.status = ex_queue::Q_NO_DATA;
pentry->upState.downIndex = qParent_.down->getHeadIndex();
pentry->upState.parentIndex = pentry_down->downState.parentIndex;
pentry->upState.setMatchNo(pstate.matchCount_);
// LCOV_EXCL_START
// BEGIN: Read note at beginning of work().
//
if(pstate.noOfUnPackedRows_ != 0)
{
ComDiagsArea *da = pentry->getDiagsArea();
if (da == NULL)
{
da = ComDiagsArea::allocate(getGlobals()->getDefaultHeap());
pentry->setDiagsArea(da);
}
da->addRowCount(pstate.noOfUnPackedRows_);
pstate.noOfUnPackedRows_ = 0;
}
//
// END: - Read note at beginning of work().
// LCOV_EXCL_STOP
// if stats are to be collected, collect them.
if (getStatsEntry())
{
// nothing yet returned from right child or returned
// to parent.
getStatsEntry()->setActualRowsReturned(0);
}
// insert into parent up queue
qParent_.up->insert();
pstate.step_ = EMPTY_;
qParent_.down->removeHead();
return WORK_CALL_AGAIN; // check for more down requests
}
break;
} // switch pstate.step_
} // while
#pragma nowarn(203) // warning elimination
return 0;
#pragma warn(203) // warning elimination
}