database/mariadb-columnstore-engine
1
0
Fork 0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-08-05 16:15:50 +03:00
Code Activity

MCOL-4505 Fix DISTINCT with sub query gets wrong results

Browse Source
This commit is contained in:
Alexey Antipovsky
2021-01-20 14:36:49 +03:00
parent 704f196a83
commit a9e96acf8b
1 changed files with 43 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

57
dbcon/joblist/tupleannexstep.cpp
View File

@@ -479,8 +479,12 @@ void TupleAnnexStep::executeNoOrderByWithDistinct()
utils::setThreadName("TASwoOrdDist"); utils::setThreadName("TASwoOrdDist");
scoped_ptr<DistinctMap_t> distinctMap(new DistinctMap_t(10, TAHasher(this), TAEq(this))); scoped_ptr<DistinctMap_t> distinctMap(new DistinctMap_t(10, TAHasher(this), TAEq(this)));
vector<RGData> dataVec; vector<RGData> dataVec;
vector<RGData> dataVecSkip;
RGData rgDataIn; RGData rgDataIn;
RGData rgDataOut; RGData rgDataOut;
RGData rgDataSkip;
RowGroup rowGroupSkip;
Row rowSkip;
bool more = false; bool more = false;
rgDataOut.reinit(fRowGroupOut); rgDataOut.reinit(fRowGroupOut);
@@ -491,6 +495,13 @@ void TupleAnnexStep::executeNoOrderByWithDistinct()
fRowGroupOut.initRow(&row1); fRowGroupOut.initRow(&row1);
fRowGroupOut.initRow(&row2); fRowGroupOut.initRow(&row2);
rowGroupSkip = fRowGroupOut;
rgDataSkip.reinit(rowGroupSkip);
rowGroupSkip.setData(&rgDataSkip);
rowGroupSkip.resetRowGroup(0);
rowGroupSkip.initRow(&rowSkip);
rowGroupSkip.getRow(0, &rowSkip);
try try
{ {
more = fInputDL->next(fInputIterator, &rgDataIn); more = fInputDL->next(fInputIterator, &rgDataIn);
@@ -512,19 +523,24 @@ void TupleAnnexStep::executeNoOrderByWithDistinct()
for (uint64_t i = 0; i < fRowGroupIn.getRowCount() && !cancelled() && !fLimitHit; ++i) for (uint64_t i = 0; i < fRowGroupIn.getRowCount() && !cancelled() && !fLimitHit; ++i)
{ {
pair<DistinctMap_t::iterator, bool> inserted; pair<DistinctMap_t::iterator, bool> inserted;
Row* rowPtr;
inserted = distinctMap->insert(fRowIn.getPointer()); if (distinctMap->size() < fLimitStart)
rowPtr = &rowSkip;
else
rowPtr = &fRowOut;
if (fConstant)
fConstant->fillInConstants(fRowIn, *rowPtr);
else
copyRow(fRowIn, rowPtr);
fRowIn.nextRow();
inserted = distinctMap->insert(rowPtr->getPointer());
++fRowsProcessed; ++fRowsProcessed;
if (inserted.second) if (inserted.second)
{ {
// skip first limit-start rows
if (distinctMap->size() <= fLimitStart)
{
fRowIn.nextRow();
continue;
}
if (UNLIKELY(fRowsReturned >= fLimitCount)) if (UNLIKELY(fRowsReturned >= fLimitCount))
{ {
fLimitHit = true; fLimitHit = true;
@@ -532,11 +548,26 @@ void TupleAnnexStep::executeNoOrderByWithDistinct()
break; break;
} }
// skip first limit-start rows
if (distinctMap->size() <= fLimitStart)
{
rowGroupSkip.incRowCount();
rowSkip.nextRow();
if (UNLIKELY(rowGroupSkip.getRowCount() >= rowgroup::rgCommonSize))
{
// allocate new RGData for skipped rows below the fLimitStart
// offset (do not take it into account in RM assuming there
// are few skipped rows
dataVecSkip.push_back(rgDataSkip);
rgDataSkip.reinit(rowGroupSkip);
rowGroupSkip.setData(&rgDataSkip);
rowGroupSkip.resetRowGroup(0);
rowGroupSkip.getRow(0, &rowSkip);
}
continue;
}
++fRowsReturned; ++fRowsReturned;
if (fConstant)
fConstant->fillInConstants(fRowIn, fRowOut);
else
copyRow(fRowIn, &fRowOut);
fRowGroupOut.incRowCount(); fRowGroupOut.incRowCount();
fRowOut.nextRow(); fRowOut.nextRow();
@@ -550,8 +581,6 @@ void TupleAnnexStep::executeNoOrderByWithDistinct()
fRowGroupOut.getRow(0, &fRowOut); fRowGroupOut.getRow(0, &fRowOut);
} }
} }
fRowIn.nextRow();
} }
more = fInputDL->next(fInputIterator, &rgDataIn); more = fInputDL->next(fInputIterator, &rgDataIn);