database/mariadb-columnstore-engine
1
0
Fork 0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-08-08 14:22:09 +03:00
Code Activity

Merge pull request #504 from mariadb-corporation/1.1-merge-up-20180621

Browse Source 
Merge develop-1.1 into develop
This commit is contained in:
Roman Nozdrin
2018-07-04 22:52:04 +03:00
committed by GitHub
parent 0976ec7511 8c90419852
commit bd641301da
54 changed files with 1905 additions and 1010 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
dbcon/execplan/aggregatecolumn.cpp
View File

@@ -107,8 +107,8 @@ AggregateColumn::AggregateColumn(const string& functionName, const string& conte
fData(functionName + "(" + content + ")")
{
// TODO: need to handle distinct
SRCP srcp(new ArithmeticColumn(content));
fAggParms.push_back(srcp);
SRCP srcp(new ArithmeticColumn(content));
fAggParms.push_back(srcp);
}
AggregateColumn::AggregateColumn( const AggregateColumn& rhs, const uint32_t sessionID ):
@@ -144,6 +144,7 @@ const string AggregateColumn::toString() const
{
output << *(fAggParms[i]) << " ";
}
output << endl;
if (fConstCol)
@@ -167,6 +168,7 @@ void AggregateColumn::serialize(messageqcpp::ByteStream& b) const
b << static_cast<uint8_t>(fAggOp);
b << static_cast<uint32_t>(fAggParms.size());
for (uint32_t i = 0; i < fAggParms.size(); ++i)
{
fAggParms[i]->serialize(b);
@@ -208,6 +210,7 @@ void AggregateColumn::unserialize(messageqcpp::ByteStream& b)
b >> fAggOp;
b >> size;
for (i = 0; i < size; i++)
{
rc = dynamic_cast<ReturnedColumn*>(ObjectReader::createTreeNode(b));
@@ -264,9 +267,10 @@ bool AggregateColumn::operator==(const AggregateColumn& t) const
{
return false;
}
for (it = fAggParms.begin(), it2 = t.fAggParms.begin();
it != fAggParms.end();
++it, ++it2)
for (it = fAggParms.begin(), it2 = t.fAggParms.begin();
it != fAggParms.end();
++it, ++it2)
{
if (**it != **it2)
return false;

6
dbcon/execplan/aggregatecolumn.h
View File

@@ -163,7 +163,7 @@ public:
fAggParms = parms;
}
/** return a copy of this pointer
*
* deep copy of this pointer and return the copy
@@ -316,8 +316,8 @@ protected:
uint8_t fAggOp;
/**
* ReturnedColumn objects that are the arguments to this
* function
* ReturnedColumn objects that are the arguments to this
* function
*/
AggParms fAggParms;

1
dbcon/joblist/crossenginestep.cpp
View File

@@ -744,7 +744,6 @@ string CrossEngineStep::makeQuery()
// the string must consist of a single SQL statement without a terminating semicolon ; or \g.
// oss << ";";
return oss.str();
}

62
dbcon/joblist/joblistfactory.cpp
View File

@@ -896,36 +896,44 @@ const JobStepVector doAggProject(const CalpontSelectExecutionPlan* csep, JobInfo
continue;
}
#if 0
// MCOL-1201 Add support for multi-parameter UDAnF
UDAFColumn* udafc = dynamic_cast<UDAFColumn*>(retCols[i].get());
if (udafc != NULL)
{
srcp = udafc->aggParms()[0];
const RowColumn* rcp = dynamic_cast<const RowColumn*>(srcp.get());
UDAFColumn* udafc = dynamic_cast<UDAFColumn*>(retCols[i].get());
const vector<SRCP>& cols = rcp->columnVec();
for (vector<SRCP>::const_iterator j = cols.begin(); j != cols.end(); j++)
{
if (udafc != NULL)
{
srcp = udafc->aggParms()[0];
const RowColumn* rcp = dynamic_cast<const RowColumn*>(srcp.get());
const vector<SRCP>& cols = rcp->columnVec();
for (vector<SRCP>::const_iterator j = cols.begin(); j != cols.end(); j++)
{
srcp = *j;
if (dynamic_cast<const ConstantColumn*>(srcp.get()) == NULL)
retCols.push_back(srcp);
if (dynamic_cast<const ConstantColumn*>(srcp.get()) == NULL)
retCols.push_back(srcp);
// Do we need this?
const ArithmeticColumn* ac = dynamic_cast<const ArithmeticColumn*>(srcp.get());
const FunctionColumn* fc = dynamic_cast<const FunctionColumn*>(srcp.get());
if (ac != NULL || fc != NULL)
{
// bug 3728, make a dummy expression step for each expression.
scoped_ptr<ExpressionStep> es(new ExpressionStep(jobInfo));
es->expression(srcp, jobInfo);
}
}
continue;
}
const ArithmeticColumn* ac = dynamic_cast<const ArithmeticColumn*>(srcp.get());
const FunctionColumn* fc = dynamic_cast<const FunctionColumn*>(srcp.get());
if (ac != NULL || fc != NULL)
{
// bug 3728, make a dummy expression step for each expression.
scoped_ptr<ExpressionStep> es(new ExpressionStep(jobInfo));
es->expression(srcp, jobInfo);
}
}
continue;
}
#endif
srcp = retCols[i];
const AggregateColumn* ag = dynamic_cast<const AggregateColumn*>(retCols[i].get());
// bug 3728 Make a dummy expression for srcp if it is an
// expression. This is needed to fill in some stuff.
// Note that es.expression does nothing if the item is not an expression.
@@ -937,7 +945,7 @@ const JobStepVector doAggProject(const CalpontSelectExecutionPlan* csep, JobInfo
}
else
{
// MCOL-1201 multi-argument aggregate. make a dummy expression
// MCOL-1201 multi-argument aggregate. make a dummy expression
// step for each argument that is an expression.
for (uint32_t i = 0; i < ag->aggParms().size(); ++i)
{
@@ -1017,7 +1025,7 @@ const JobStepVector doAggProject(const CalpontSelectExecutionPlan* csep, JobInfo
// replace the aggregate on constant with a count(*)
SRCP clone;
UDAFColumn* udafc = dynamic_cast<UDAFColumn*>(aggc);
if (udafc)
{
clone.reset(new UDAFColumn(*udafc, aggc->sessionID()));
@@ -1026,14 +1034,15 @@ const JobStepVector doAggProject(const CalpontSelectExecutionPlan* csep, JobInfo
{
clone.reset(new AggregateColumn(*aggc, aggc->sessionID()));
}
jobInfo.constAggregate.insert(make_pair(i, clone));
aggc->aggOp(AggregateColumn::COUNT_ASTERISK);
aggc->distinct(false);
}
srcp = aggParms[parm];
sc = dynamic_cast<const SimpleColumn*>(srcp.get());
if (parm == 0)
{
op = aggc->aggOp();
@@ -1042,7 +1051,9 @@ const JobStepVector doAggProject(const CalpontSelectExecutionPlan* csep, JobInfo
{
op = AggregateColumn::MULTI_PARM;
}
doDistinct = aggc->distinct();
if (aggParms.size() == 1)
{
// Set the col type based on the single parm.
@@ -1050,6 +1061,7 @@ const JobStepVector doAggProject(const CalpontSelectExecutionPlan* csep, JobInfo
// doesn't really make sense.
updateAggregateColType(aggc, srcp, op, jobInfo);
}
aggCt = aggc->resultType();
// As of bug3695, make sure varbinary is not used in aggregation.

83
dbcon/joblist/tupleaggregatestep.cpp
View File

@@ -852,6 +852,7 @@ SJSTEP TupleAggregateStep::prepAggregate(SJSTEP& step, JobInfo& jobInfo)
if (ac->aggOp() == ROWAGG_UDAF)
{
UDAFColumn* udafc = dynamic_cast<UDAFColumn*>(ac);
if (udafc)
{
constAggDataVec.push_back(
@@ -1295,10 +1296,12 @@ void TupleAggregateStep::prep1PhaseAggregate(
if (aggOp == ROWAGG_UDAF)
{
std::vector<SRCP>::iterator it = jobInfo.projectionCols.begin() + projColsUDAFIdx;
for (; it != jobInfo.projectionCols.end(); it++)
{
udafc = dynamic_cast<UDAFColumn*>((*it).get());
projColsUDAFIdx++;
if (udafc)
{
pUDAFFunc = udafc->getContext().getFunction();
@@ -1307,6 +1310,7 @@ void TupleAggregateStep::prep1PhaseAggregate(
break;
}
}
if (it == jobInfo.projectionCols.end())
{
throw logic_error("(1)prep1PhaseAggregate: A UDAF function is called but there\'s not enough UDAFColumns");
@@ -1488,10 +1492,12 @@ void TupleAggregateStep::prep1PhaseAggregate(
// If the first param is const
udafcParamIdx = 0;
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
}
++udafcParamIdx;
break;
}
@@ -1504,10 +1510,12 @@ void TupleAggregateStep::prep1PhaseAggregate(
precisionAgg.push_back(precisionProj[colProj]);
typeAgg.push_back(typeProj[colProj]);
widthAgg.push_back(width[colProj]);
// If the param is const
if (udafc)
{
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
@@ -1517,6 +1525,7 @@ void TupleAggregateStep::prep1PhaseAggregate(
{
throw QueryDataExcept("prep1PhaseAggregate: UDAF multi function with no parms", aggregateFuncErr);
}
++udafcParamIdx;
}
break;
@@ -1892,6 +1901,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
break;
}
}
if (it == jobInfo.projectionCols.end())
{
throw logic_error("(1)prep1PhaseDistinctAggregate: A UDAF function is called but there\'s not enough UDAFColumns");
@@ -2111,10 +2121,12 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
// If the first param is const
udafcParamIdx = 0;
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
}
++udafcParamIdx;
break;
}
@@ -2129,10 +2141,12 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
widthAgg.push_back(widthProj[colProj]);
multiParmIndexes.push_back(colAgg);
++colAgg;
// If the param is const
if (udafc)
{
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
@@ -2142,6 +2156,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
{
throw QueryDataExcept("prep1PhaseDistinctAggregate: UDAF multi function with no parms", aggregateFuncErr);
}
++udafcParamIdx;
}
break;
@@ -2193,9 +2208,10 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
groupByNoDist.push_back(groupby);
aggFuncMap.insert(make_pair(boost::make_tuple(keysAgg[i], 0, pUDAFFunc), i));
}
// locate the return column position in aggregated rowgroup
uint64_t outIdx = 0;
for (uint64_t i = 0; i < returnedColVec.size(); i++)
{
udafc = NULL;
@@ -2240,16 +2256,19 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
if (aggOp == ROWAGG_UDAF)
{
std::vector<SRCP>::iterator it = jobInfo.projectionCols.begin() + projColsUDAFIdx;
for (; it != jobInfo.projectionCols.end(); it++)
{
udafc = dynamic_cast<UDAFColumn*>((*it).get());
projColsUDAFIdx++;
if (udafc)
{
pUDAFFunc = udafc->getContext().getFunction();
break;
}
}
if (it == jobInfo.projectionCols.end())
{
throw logic_error("(1)prep1PhaseDistinctAggregate: A UDAF function is called but there\'s not enough UDAFColumns");
@@ -2546,6 +2565,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
{
// update the aggregate function vector
SP_ROWAGG_FUNC_t funct;
if (aggOp == ROWAGG_UDAF)
{
funct.reset(new RowUDAFFunctionCol(udafc->getContext(), colAgg, outIdx));
@@ -2589,6 +2609,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
else if (returnedColVec[i].second == AggregateColumn::DISTINCT_AVG)
avgDistFuncMap.insert(make_pair(returnedColVec[i].first, funct));
}
++outIdx;
} // for (i
@@ -2839,6 +2860,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
++multiParms;
continue;
}
if (returnedColVec[k].first != distinctColKey)
continue;
@@ -2859,7 +2881,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
f->fStatsFunction,
groupBySub.size() - 1,
f->fOutputColumnIndex,
f->fAuxColumnIndex-multiParms));
f->fAuxColumnIndex - multiParms));
functionSub2.push_back(funct);
}
}
@@ -2887,6 +2909,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
++multiParms;
continue;
}
// search non-distinct functions in functionVec
vector<SP_ROWAGG_FUNC_t>::iterator it = functionVec2.begin();
@@ -2902,7 +2925,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
udafFuncCol->fUDAFContext,
udafFuncCol->fInputColumnIndex,
udafFuncCol->fOutputColumnIndex,
udafFuncCol->fAuxColumnIndex-multiParms));
udafFuncCol->fAuxColumnIndex - multiParms));
functionSub2.push_back(funct);
}
else if ((f->fOutputColumnIndex == k) &&
@@ -2924,7 +2947,7 @@ void TupleAggregateStep::prep1PhaseDistinctAggregate(
f->fStatsFunction,
f->fInputColumnIndex,
f->fOutputColumnIndex,
f->fAuxColumnIndex-multiParms));
f->fAuxColumnIndex - multiParms));
functionSub2.push_back(funct);
}
}
@@ -3160,10 +3183,12 @@ void TupleAggregateStep::prep2PhasesAggregate(
if (aggOp == ROWAGG_UDAF)
{
std::vector<SRCP>::iterator it = jobInfo.projectionCols.begin() + projColsUDAFIdx;
for (; it != jobInfo.projectionCols.end(); it++)
{
udafc = dynamic_cast<UDAFColumn*>((*it).get());
projColsUDAFIdx++;
if (udafc)
{
pUDAFFunc = udafc->getContext().getFunction();
@@ -3172,6 +3197,7 @@ void TupleAggregateStep::prep2PhasesAggregate(
break;
}
}
if (it == jobInfo.projectionCols.end())
{
throw logic_error("(1)prep2PhasesAggregate: A UDAF function is called but there\'s not enough UDAFColumns");
@@ -3394,10 +3420,12 @@ void TupleAggregateStep::prep2PhasesAggregate(
// If the first param is const
udafcParamIdx = 0;
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
}
++udafcParamIdx;
break;
}
@@ -3411,10 +3439,12 @@ void TupleAggregateStep::prep2PhasesAggregate(
typeAggPm.push_back(typeProj[colProj]);
widthAggPm.push_back(width[colProj]);
colAggPm++;
// If the param is const
if (udafc)
{
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
@@ -3424,6 +3454,7 @@ void TupleAggregateStep::prep2PhasesAggregate(
{
throw QueryDataExcept("prep2PhasesAggregate: UDAF multi function with no parms", aggregateFuncErr);
}
++udafcParamIdx;
}
break;
@@ -3451,6 +3482,7 @@ void TupleAggregateStep::prep2PhasesAggregate(
AGG_MAP aggDupFuncMap;
projColsUDAFIdx = 0;
// copy over the groupby vector
// update the outputColumnIndex if returned
for (uint64_t i = 0; i < groupByPm.size(); i++)
@@ -3462,6 +3494,7 @@ void TupleAggregateStep::prep2PhasesAggregate(
// locate the return column position in aggregated rowgroup from PM
// outIdx is i without the multi-columns,
uint64_t outIdx = 0;
for (uint64_t i = 0; i < returnedColVec.size(); i++)
{
uint32_t retKey = returnedColVec[i].first;
@@ -3478,6 +3511,7 @@ void TupleAggregateStep::prep2PhasesAggregate(
// Is this a UDAF? use the function as part of the key.
pUDAFFunc = NULL;
udafc = NULL;
if (aggOp == ROWAGG_UDAF)
{
std::vector<SRCP>::iterator it = jobInfo.projectionCols.begin() + projColsUDAFIdx;
@@ -3486,12 +3520,14 @@ void TupleAggregateStep::prep2PhasesAggregate(
{
udafc = dynamic_cast<UDAFColumn*>((*it).get());
projColsUDAFIdx++;
if (udafc)
{
pUDAFFunc = udafc->getContext().getFunction();
break;
}
}
if (it == jobInfo.projectionCols.end())
{
throw logic_error("(3)prep2PhasesAggregate: A UDAF function is called but there\'s not enough UDAFColumns");
@@ -3644,6 +3680,7 @@ void TupleAggregateStep::prep2PhasesAggregate(
{
// update the aggregate function vector
SP_ROWAGG_FUNC_t funct;
if (aggOp == ROWAGG_UDAF)
{
funct.reset(new RowUDAFFunctionCol(udafc->getContext(), colPm, outIdx));
@@ -3685,6 +3722,7 @@ void TupleAggregateStep::prep2PhasesAggregate(
if (returnedColVec[i].second == AggregateColumn::AVG)
avgFuncMap.insert(make_pair(returnedColVec[i].first, funct));
}
++outIdx;
}
@@ -4029,10 +4067,12 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
if (aggOp == ROWAGG_UDAF)
{
std::vector<SRCP>::iterator it = jobInfo.projectionCols.begin() + projColsUDAFIdx;
for (; it != jobInfo.projectionCols.end(); it++)
{
udafc = dynamic_cast<UDAFColumn*>((*it).get());
projColsUDAFIdx++;
if (udafc)
{
pUDAFFunc = udafc->getContext().getFunction();
@@ -4041,6 +4081,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
break;
}
}
if (it == jobInfo.projectionCols.end())
{
throw logic_error("(1)prep2PhasesDistinctAggregate: A UDAF function is called but there\'s not enough UDAFColumns");
@@ -4259,10 +4300,12 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
// If the first param is const
udafcParamIdx = 0;
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
}
++udafcParamIdx;
break;
}
@@ -4277,10 +4320,12 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
widthAggPm.push_back(width[colProj]);
multiParmIndexes.push_back(colAggPm);
colAggPm++;
// If the param is const
if (udafc)
{
ConstantColumn* cc = dynamic_cast<ConstantColumn*>(udafc->aggParms()[udafcParamIdx].get());
if (cc)
{
funct->fpConstCol = udafc->aggParms()[udafcParamIdx];
@@ -4290,6 +4335,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
{
throw QueryDataExcept("prep2PhasesDistinctAggregate: UDAF multi function with no parms", aggregateFuncErr);
}
++udafcParamIdx;
}
break;
@@ -4332,15 +4378,17 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
if (funcPm->fAggFunction == ROWAGG_UDAF)
{
RowUDAFFunctionCol* udafFuncCol = dynamic_cast<RowUDAFFunctionCol*>(funcPm.get());
if (!udafFuncCol)
{
throw logic_error("(3)prep2PhasesDistinctAggregate: A UDAF function is called but there's no RowUDAFFunctionCol");
throw logic_error("(3)prep2PhasesDistinctAggregate: A UDAF function is called but there's no RowUDAFFunctionCol");
}
funct.reset(new RowUDAFFunctionCol(
udafFuncCol->fUDAFContext,
udafFuncCol->fOutputColumnIndex,
udafFuncCol->fOutputColumnIndex-multiParms,
udafFuncCol->fAuxColumnIndex-multiParms));
udafFuncCol->fOutputColumnIndex - multiParms,
udafFuncCol->fAuxColumnIndex - multiParms));
functionNoDistVec.push_back(funct);
pUDAFFunc = udafFuncCol->fUDAFContext.getFunction();
}
@@ -4350,8 +4398,8 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
funcPm->fAggFunction,
funcPm->fStatsFunction,
funcPm->fOutputColumnIndex,
funcPm->fOutputColumnIndex-multiParms,
funcPm->fAuxColumnIndex-multiParms));
funcPm->fOutputColumnIndex - multiParms,
funcPm->fAuxColumnIndex - multiParms));
functionNoDistVec.push_back(funct);
pUDAFFunc = NULL;
}
@@ -4364,6 +4412,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
{
continue;
}
oidsAggUm.push_back(oidsAggPm[idx]);
keysAggUm.push_back(keysAggPm[idx]);
scaleAggUm.push_back(scaleAggPm[idx]);
@@ -4400,6 +4449,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
// locate the return column position in aggregated rowgroup from PM
// outIdx is i without the multi-columns,
uint64_t outIdx = 0;
for (uint64_t i = 0; i < returnedColVec.size(); i++)
{
pUDAFFunc = NULL;
@@ -4420,16 +4470,19 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
if (aggOp == ROWAGG_UDAF)
{
std::vector<SRCP>::iterator it = jobInfo.projectionCols.begin() + projColsUDAFIdx;
for (; it != jobInfo.projectionCols.end(); it++)
{
udafc = dynamic_cast<UDAFColumn*>((*it).get());
projColsUDAFIdx++;
if (udafc)
{
pUDAFFunc = udafc->getContext().getFunction();
break;
}
}
if (it == jobInfo.projectionCols.end())
{
throw logic_error("(4)prep2PhasesDistinctAggregate: A UDAF function is called but there\'s not enough UDAFColumns");
@@ -4606,6 +4659,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
oidsAggDist.push_back(oidsAggUm[colUm]);
keysAggDist.push_back(retKey);
scaleAggDist.push_back(0);
if (isUnsigned(typeAggUm[colUm]))
{
precisionAggDist.push_back(20);
@@ -4616,6 +4670,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
precisionAggDist.push_back(19);
typeAggDist.push_back(CalpontSystemCatalog::BIGINT);
}
widthAggDist.push_back(bigIntWidth);
}
}
@@ -4702,6 +4757,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
{
// update the aggregate function vector
SP_ROWAGG_FUNC_t funct;
if (aggOp == ROWAGG_UDAF)
{
funct.reset(new RowUDAFFunctionCol(udafc->getContext(), colUm, outIdx));
@@ -4745,6 +4801,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
else if (returnedColVec[i].second == AggregateColumn::DISTINCT_AVG)
avgDistFuncMap.insert(make_pair(returnedColVec[i].first, funct));
}
++outIdx;
} // for (i
@@ -4987,6 +5044,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
++multiParms;
continue;
}
if (returnedColVec[k].first != distinctColKey)
continue;
@@ -5008,7 +5066,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
f->fStatsFunction,
groupBySub.size() - 1,
f->fOutputColumnIndex,
f->fAuxColumnIndex-multiParms));
f->fAuxColumnIndex - multiParms));
functionSub2.push_back(funct);
}
}
@@ -5034,6 +5092,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
++multiParms;
continue;
}
// search non-distinct functions in functionVec
vector<SP_ROWAGG_FUNC_t>::iterator it = functionVecUm.begin();
@@ -5051,7 +5110,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
udafFuncCol->fUDAFContext,
udafFuncCol->fInputColumnIndex,
udafFuncCol->fOutputColumnIndex,
udafFuncCol->fAuxColumnIndex-multiParms));
udafFuncCol->fAuxColumnIndex - multiParms));
functionSub2.push_back(funct);
}
else if (f->fAggFunction == ROWAGG_COUNT_ASTERISK ||
@@ -5072,7 +5131,7 @@ void TupleAggregateStep::prep2PhasesDistinctAggregate(
f->fStatsFunction,
f->fInputColumnIndex,
f->fOutputColumnIndex,
f->fAuxColumnIndex-multiParms));
f->fAuxColumnIndex - multiParms));
functionSub2.push_back(funct);
}
}

1
dbcon/joblist/windowfunctionstep.cpp
View File

@@ -655,6 +655,7 @@ void WindowFunctionStep::initialize(const RowGroup& rg, JobInfo& jobInfo)
// make sure index is in range
else if (fields.size() > 1 && fields[1] >= 0 && static_cast<uint64_t>(fields[1]) < types.size())
ct = types[fields[1]];
// workaround for functions using "within group (order by)" syntax
string fn = boost::to_upper_copy(wc->functionName());

65
dbcon/mysql/columnstore_info.sql
View File

@@ -37,43 +37,56 @@ DROP PROCEDURE IF EXISTS `table_usage` //
CREATE PROCEDURE table_usage (IN t_schema char(64), IN t_name char(64))
`table_usage`: BEGIN
DECLARE done INTEGER DEFAULT 0;
DECLARE dbname VARCHAR(64);
DECLARE tbname VARCHAR(64);
DECLARE object_ids TEXT;
DECLARE dictionary_object_ids TEXT;
DECLARE `locker` TINYINT UNSIGNED DEFAULT IS_USED_LOCK('table_usage');
DECLARE columns_list CURSOR FOR SELECT TABLE_SCHEMA, TABLE_NAME, GROUP_CONCAT(object_id) OBJECT_IDS, GROUP_CONCAT(dictionary_object_id) DICT_OBJECT_IDS FROM INFORMATION_SCHEMA.COLUMNSTORE_COLUMNS WHERE table_name = t_name and table_schema = t_schema GROUP BY table_schema, table_name;
DECLARE columns_list_sc CURSOR FOR SELECT TABLE_SCHEMA, TABLE_NAME, GROUP_CONCAT(object_id) OBJECT_IDS, GROUP_CONCAT(dictionary_object_id) DICT_OBJECT_IDS FROM INFORMATION_SCHEMA.COLUMNSTORE_COLUMNS WHERE table_schema = t_schema GROUP BY table_schema, table_name;
DECLARE columns_list_all CURSOR FOR SELECT TABLE_SCHEMA, TABLE_NAME, GROUP_CONCAT(object_id) OBJECT_IDS, GROUP_CONCAT(dictionary_object_id) DICT_OBJECT_IDS FROM INFORMATION_SCHEMA.COLUMNSTORE_COLUMNS GROUP BY table_schema, table_name;
DECLARE CONTINUE HANDLER FOR NOT FOUND SET done = 1;
IF `locker` IS NOT NULL THEN
SIGNAL SQLSTATE '45000' SET MESSAGE_TEXT = 'Error acquiring table_usage lock';
LEAVE `table_usage`;
END IF;
DO GET_LOCK('table_usage', 0);
DROP TABLE IF EXISTS columnstore_info.columnstore_columns;
DROP TABLE IF EXISTS columnstore_info.columnstore_files;
CREATE TABLE columnstore_info.columnstore_columns engine=myisam as (select * from information_schema.columnstore_columns);
ALTER TABLE columnstore_info.columnstore_columns ADD INDEX `object_id` (`object_id`);
ALTER TABLE columnstore_info.columnstore_columns ADD INDEX `dictionary_object_id` (`dictionary_object_id`);
CREATE TABLE columnstore_info.columnstore_files engine=myisam as (select * from information_schema.columnstore_files);
ALTER TABLE columnstore_info.columnstore_files ADD INDEX `object_id` (`object_id`);
CREATE TEMPORARY TABLE columnstore_info.columnstore_files (TABLE_SCHEMA VARCHAR(64), TABLE_NAME VARCHAR(64), DATA BIGINT, DICT BIGINT);
IF t_name IS NOT NULL THEN
SELECT TABLE_SCHEMA, TABLE_NAME, columnstore_info.format_filesize(data) as DATA_DISK_USAGE, columnstore_info.format_filesize(dict) as DICT_DISK_USAGE, columnstore_info.format_filesize(data + COALESCE(dict, 0)) as TOTAL_USAGE FROM (
SELECT TABLE_SCHEMA, TABLE_NAME, (SELECT sum(cf.file_size) as data FROM columnstore_info.columnstore_columns cc JOIN columnstore_info.columnstore_files cf ON cc.object_id = cf.object_id WHERE table_name = ics.table_name and table_schema = ics.table_schema) as data, (SELECT sum(cf.file_size) as dict FROM columnstore_info.columnstore_columns cc JOIN columnstore_info.columnstore_files cf ON cc.dictionary_object_id = cf.object_id WHERE table_name = ics.table_name and table_schema = ics.table_schema GROUP BY table_schema, table_name) as dict
FROM
columnstore_info.columnstore_columns ics where table_name = t_name and (table_schema = t_schema or t_schema IS NULL)
group by table_schema, table_name
) q;
OPEN columns_list;
ELSEIF t_schema IS NOT NULL THEN
SELECT TABLE_SCHEMA, TABLE_NAME, columnstore_info.format_filesize(data) as DATA_DISK_USAGE, columnstore_info.format_filesize(dict) as DICT_DISK_USAGE, columnstore_info.format_filesize(data + COALESCE(dict, 0)) as TOTAL_USAGE FROM (
SELECT TABLE_SCHEMA, TABLE_NAME, (SELECT sum(cf.file_size) as data FROM columnstore_info.columnstore_columns cc JOIN columnstore_info.columnstore_files cf ON cc.object_id = cf.object_id WHERE table_name = ics.table_name and table_schema = ics.table_schema) as data, (SELECT sum(cf.file_size) as dict FROM columnstore_info.columnstore_columns cc JOIN columnstore_info.columnstore_files cf ON cc.dictionary_object_id = cf.object_id WHERE table_name = ics.table_name and table_schema = ics.table_schema GROUP BY table_schema, table_name) as dict
FROM
columnstore_info.columnstore_columns ics where table_schema = t_schema
group by table_schema, table_name
) q;
OPEN columns_list_sc;
ELSE
SELECT TABLE_SCHEMA, TABLE_NAME, columnstore_info.format_filesize(data) as DATA_DISK_USAGE, columnstore_info.format_filesize(dict) as DICT_DISK_USAGE, columnstore_info.format_filesize(data + COALESCE(dict, 0)) as TOTAL_USAGE FROM (
SELECT TABLE_SCHEMA, TABLE_NAME, (SELECT sum(cf.file_size) as data FROM columnstore_info.columnstore_columns cc JOIN columnstore_info.columnstore_files cf ON cc.object_id = cf.object_id WHERE table_name = ics.table_name and table_schema = ics.table_schema) as data, (SELECT sum(cf.file_size) as dict FROM columnstore_info.columnstore_columns cc JOIN columnstore_info.columnstore_files cf ON cc.dictionary_object_id = cf.object_id WHERE table_name = ics.table_name and table_schema = ics.table_schema GROUP BY table_schema, table_name) as dict
FROM
columnstore_info.columnstore_columns ics
group by table_schema, table_name
) q;
OPEN columns_list_all;
END IF;
DROP TABLE IF EXISTS columnstore_info.columnstore_columns;
files_table: LOOP
IF t_name IS NOT NULL THEN
FETCH columns_list INTO dbname, tbname, object_ids, dictionary_object_ids;
ELSEIF t_schema IS NOT NULL THEN
FETCH columns_list_sc INTO dbname, tbname, object_ids, dictionary_object_ids;
ELSE
FETCH columns_list_all INTO dbname, tbname, object_ids, dictionary_object_ids;
END IF;
IF done = 1 THEN LEAVE files_table;
END IF;
INSERT INTO columnstore_info.columnstore_files (SELECT dbname, tbname, sum(file_size), 0 FROM information_schema.columnstore_files WHERE find_in_set(object_id, object_ids));
IF dictionary_object_ids IS NOT NULL THEN
UPDATE columnstore_info.columnstore_files SET DICT = (SELECT sum(file_size) FROM information_schema.columnstore_files WHERE find_in_set(object_id, dictionary_object_ids)) WHERE TABLE_SCHEMA = dbname AND TABLE_NAME = tbname;
END IF;
END LOOP;
IF t_name IS NOT NULL THEN
CLOSE columns_list;
ELSEIF t_schema IS NOT NULL THEN
CLOSE columns_list_sc;
ELSE
CLOSE columns_list_all;
END IF;
SELECT TABLE_SCHEMA, TABLE_NAME, columnstore_info.format_filesize(DATA) as DATA_DISK_USAGE, columnstore_info.format_filesize(DICT) as DICT_DATA_USAGE, columnstore_info.format_filesize(DATA + COALESCE(DICT, 0)) as TOTAL_USAGE FROM columnstore_info.columnstore_files;
DROP TABLE IF EXISTS columnstore_info.columnstore_files;
DO RELEASE_LOCK('table_usage');
END //

1
dbcon/mysql/ha_calpont_dml.cpp
View File

@@ -899,6 +899,7 @@ int ha_calpont_impl_write_batch_row_(uchar* buf, TABLE* table, cal_impl_if::cal_
{
fprintf(ci.filePtr, "-");
}
if (!ltime.second_part)
{
fprintf(ci.filePtr, "%02d:%02d:%02d%c",

753
dbcon/mysql/ha_calpont_execplan.cpp
View File

@@ -4039,8 +4039,8 @@ ParseTree* buildParseTree(Item_func* item, gp_walk_info& gwi, bool& nonSupport)
ReturnedColumn* buildAggregateColumn(Item* item, gp_walk_info& gwi)
{
// MCOL-1201 For UDAnF multiple parameters
vector<SRCP> selCols;
vector<SRCP> orderCols;
vector<SRCP> selCols;
vector<SRCP> orderCols;
if (!(gwi.thd->infinidb_vtable.cal_conn_info))
gwi.thd->infinidb_vtable.cal_conn_info = (void*)(new cal_connection_info());
@@ -4059,6 +4059,7 @@ ReturnedColumn* buildAggregateColumn(Item* item, gp_walk_info& gwi)
// N.B. argument_count() is the # of formal parms to the agg fcn. InifniDB only supports 1 argument
// TODO: Support more than one parm
#if 0
if (isp->argument_count() != 1 && isp->sum_func() != Item_sum::GROUP_CONCAT_FUNC
&& isp->sum_func() != Item_sum::UDF_SUM_FUNC)
{
@@ -4066,6 +4067,7 @@ ReturnedColumn* buildAggregateColumn(Item* item, gp_walk_info& gwi)
gwi.parseErrorText = IDBErrorInfo::instance()->errorMsg(ERR_MUL_ARG_AGG);
return NULL;
}
#endif
AggregateColumn* ac = NULL;
@@ -4089,446 +4091,459 @@ ReturnedColumn* buildAggregateColumn(Item* item, gp_walk_info& gwi)
{
gwi.fatalParseError = true;
gwi.parseErrorText = "Non supported aggregate type on the select clause";
if (ac)
delete ac;
return NULL;
}
try
{
// special parsing for group_concat
if (isp->sum_func() == Item_sum::GROUP_CONCAT_FUNC)
{
Item_func_group_concat* gc = (Item_func_group_concat*)isp;
vector<SRCP> orderCols;
RowColumn* rowCol = new RowColumn();
vector<SRCP> selCols;
uint32_t select_ctn = gc->count_field();
ReturnedColumn* rc = NULL;
for (uint32_t i = 0; i < select_ctn; i++)
// special parsing for group_concat
if (isp->sum_func() == Item_sum::GROUP_CONCAT_FUNC)
{
rc = buildReturnedColumn(sfitempp[i], gwi, gwi.fatalParseError);
Item_func_group_concat* gc = (Item_func_group_concat*)isp;
vector<SRCP> orderCols;
RowColumn* rowCol = new RowColumn();
vector<SRCP> selCols;
if (!rc || gwi.fatalParseError)
{
if (ac)
delete ac;
return NULL;
}
uint32_t select_ctn = gc->count_field();
ReturnedColumn* rc = NULL;
selCols.push_back(SRCP(rc));
}
ORDER** order_item, **end;
for (order_item = gc->get_order(),
end = order_item + gc->order_field(); order_item < end;
order_item++)
{
Item* ord_col = *(*order_item)->item;
if (ord_col->type() == Item::INT_ITEM)
for (uint32_t i = 0; i < select_ctn; i++)
{
Item_int* id = (Item_int*)ord_col;
if (id->val_int() > (int)selCols.size())
{
gwi.fatalParseError = true;
if (ac)
delete ac;
return NULL;
}
rc = selCols[id->val_int() - 1]->clone();
rc->orderPos(id->val_int() - 1);
}
else
{
rc = buildReturnedColumn(ord_col, gwi, gwi.fatalParseError);
rc = buildReturnedColumn(sfitempp[i], gwi, gwi.fatalParseError);
if (!rc || gwi.fatalParseError)
{
if (ac)
delete ac;
if (ac)
delete ac;
return NULL;
}
selCols.push_back(SRCP(rc));
}
// 10.2 TODO: direction is now a tri-state flag
rc->asc((*order_item)->direction == ORDER::ORDER_ASC ? true : false);
orderCols.push_back(SRCP(rc));
}
ORDER** order_item, **end;
rowCol->columnVec(selCols);
(dynamic_cast<GroupConcatColumn*>(ac))->orderCols(orderCols);
parm.reset(rowCol);
ac->aggParms().push_back(parm);
if (gc->str_separator())
{
string separator;
separator.assign(gc->str_separator()->ptr(), gc->str_separator()->length());
(dynamic_cast<GroupConcatColumn*>(ac))->separator(separator);
}
}
else
{
for (uint32_t i = 0; i < isp->argument_count(); i++)
{
Item* sfitemp = sfitempp[i];
Item::Type sfitype = sfitemp->type();
switch (sfitype)
for (order_item = gc->get_order(),
end = order_item + gc->order_field(); order_item < end;
order_item++)
{
case Item::FIELD_ITEM:
{
Item_field* ifp = reinterpret_cast<Item_field*>(sfitemp);
SimpleColumn* sc = buildSimpleColumn(ifp, gwi);
Item* ord_col = *(*order_item)->item;
if (!sc)
if (ord_col->type() == Item::INT_ITEM)
{
Item_int* id = (Item_int*)ord_col;
if (id->val_int() > (int)selCols.size())
{
gwi.fatalParseError = true;
if (ac)
delete ac;
return NULL;
}
rc = selCols[id->val_int() - 1]->clone();
rc->orderPos(id->val_int() - 1);
}
else
{
rc = buildReturnedColumn(ord_col, gwi, gwi.fatalParseError);
if (!rc || gwi.fatalParseError)
{
if (ac)
delete ac;
return NULL;
}
}
// 10.2 TODO: direction is now a tri-state flag
rc->asc((*order_item)->direction == ORDER::ORDER_ASC ? true : false);
orderCols.push_back(SRCP(rc));
}
rowCol->columnVec(selCols);
(dynamic_cast<GroupConcatColumn*>(ac))->orderCols(orderCols);
parm.reset(rowCol);
ac->aggParms().push_back(parm);
if (gc->str_separator())
{
string separator;
separator.assign(gc->str_separator()->ptr(), gc->str_separator()->length());
(dynamic_cast<GroupConcatColumn*>(ac))->separator(separator);
}
}
else
{
for (uint32_t i = 0; i < isp->argument_count(); i++)
{
Item* sfitemp = sfitempp[i];
Item::Type sfitype = sfitemp->type();
switch (sfitype)
{
case Item::FIELD_ITEM:
{
Item_field* ifp = reinterpret_cast<Item_field*>(sfitemp);
SimpleColumn* sc = buildSimpleColumn(ifp, gwi);
if (!sc)
{
gwi.fatalParseError = true;
break;
}
parm.reset(sc);
gwi.columnMap.insert(CalpontSelectExecutionPlan::ColumnMap::value_type(string(ifp->field_name), parm));
TABLE_LIST* tmp = (ifp->cached_table ? ifp->cached_table : 0);
gwi.tableMap[make_aliastable(sc->schemaName(), sc->tableName(), sc->tableAlias(), sc->isInfiniDB())] = make_pair(1, tmp);
break;
}
parm.reset(sc);
gwi.columnMap.insert(CalpontSelectExecutionPlan::ColumnMap::value_type(string(ifp->field_name), parm));
TABLE_LIST* tmp = (ifp->cached_table ? ifp->cached_table : 0);
gwi.tableMap[make_aliastable(sc->schemaName(), sc->tableName(), sc->tableAlias(), sc->isInfiniDB())] = make_pair(1, tmp);
break;
}
case Item::INT_ITEM:
case Item::STRING_ITEM:
case Item::REAL_ITEM:
case Item::DECIMAL_ITEM:
{
// treat as count(*)
if (ac->aggOp() == AggregateColumn::COUNT)
ac->aggOp(AggregateColumn::COUNT_ASTERISK);
case Item::INT_ITEM:
case Item::STRING_ITEM:
case Item::REAL_ITEM:
case Item::DECIMAL_ITEM:
{
// treat as count(*)
if (ac->aggOp() == AggregateColumn::COUNT)
ac->aggOp(AggregateColumn::COUNT_ASTERISK);
parm.reset(buildReturnedColumn(sfitemp, gwi, gwi.fatalParseError));
ac->constCol(parm);
break;
}
case Item::NULL_ITEM:
{
parm.reset(new ConstantColumn("", ConstantColumn::NULLDATA));
ac->constCol(SRCP(buildReturnedColumn(sfitemp, gwi, gwi.fatalParseError)));
break;
}
case Item::FUNC_ITEM:
{
Item_func* ifp = (Item_func*)sfitemp;
ReturnedColumn* rc = 0;
// check count(1+1) case
vector <Item_field*> tmpVec;
uint16_t parseInfo = 0;
parse_item(ifp, tmpVec, gwi.fatalParseError, parseInfo);
if (parseInfo & SUB_BIT)
{
gwi.fatalParseError = true;
break;
}
else if (!gwi.fatalParseError &&
!(parseInfo & AGG_BIT) &&
!(parseInfo & AF_BIT) &&
tmpVec.size() == 0)
case Item::NULL_ITEM:
{
rc = buildFunctionColumn(ifp, gwi, gwi.fatalParseError);
FunctionColumn* fc = dynamic_cast<FunctionColumn*>(rc);
parm.reset(new ConstantColumn("", ConstantColumn::NULLDATA));
ac->constCol(SRCP(buildReturnedColumn(sfitemp, gwi, gwi.fatalParseError)));
break;
}
if ((fc && fc->functionParms().empty()) || !fc)
case Item::FUNC_ITEM:
{
Item_func* ifp = (Item_func*)sfitemp;
ReturnedColumn* rc = 0;
// check count(1+1) case
vector <Item_field*> tmpVec;
uint16_t parseInfo = 0;
parse_item(ifp, tmpVec, gwi.fatalParseError, parseInfo);
if (parseInfo & SUB_BIT)
{
//ac->aggOp(AggregateColumn::COUNT_ASTERISK);
ReturnedColumn* rc = buildReturnedColumn(sfitemp, gwi, gwi.fatalParseError);
gwi.fatalParseError = true;
break;
}
else if (!gwi.fatalParseError &&
!(parseInfo & AGG_BIT) &&
!(parseInfo & AF_BIT) &&
tmpVec.size() == 0)
{
rc = buildFunctionColumn(ifp, gwi, gwi.fatalParseError);
FunctionColumn* fc = dynamic_cast<FunctionColumn*>(rc);
if (dynamic_cast<ConstantColumn*>(rc))
if ((fc && fc->functionParms().empty()) || !fc)
{
//@bug5229. handle constant function on aggregate argument
ac->constCol(SRCP(rc));
break;
//ac->aggOp(AggregateColumn::COUNT_ASTERISK);
ReturnedColumn* rc = buildReturnedColumn(sfitemp, gwi, gwi.fatalParseError);
if (dynamic_cast<ConstantColumn*>(rc))
{
//@bug5229. handle constant function on aggregate argument
ac->constCol(SRCP(rc));
break;
}
}
}
// MySQL carelessly allows correlated aggregate function on the WHERE clause.
// Here is the work around to deal with that inconsistence.
// e.g., SELECT (SELECT t.c FROM t1 AS t WHERE t.b=MAX(t1.b + 0)) FROM t1;
ClauseType clauseType = gwi.clauseType;
if (gwi.clauseType == WHERE)
gwi.clauseType = HAVING;
// @bug 3603. for cases like max(rand()). try to build function first.
if (!rc)
rc = buildFunctionColumn(ifp, gwi, gwi.fatalParseError);
parm.reset(rc);
gwi.clauseType = clauseType;
if (gwi.fatalParseError)
break;
break;
}
case Item::REF_ITEM:
{
ReturnedColumn* rc = buildReturnedColumn(sfitemp, gwi, gwi.fatalParseError);
if (rc)
{
parm.reset(rc);
break;
}
}
// MySQL carelessly allows correlated aggregate function on the WHERE clause.
// Here is the work around to deal with that inconsistence.
// e.g., SELECT (SELECT t.c FROM t1 AS t WHERE t.b=MAX(t1.b + 0)) FROM t1;
ClauseType clauseType = gwi.clauseType;
if (gwi.clauseType == WHERE)
gwi.clauseType = HAVING;
// @bug 3603. for cases like max(rand()). try to build function first.
if (!rc)
rc = buildFunctionColumn(ifp, gwi, gwi.fatalParseError);
parm.reset(rc);
gwi.clauseType = clauseType;
if (gwi.fatalParseError)
break;
break;
}
case Item::REF_ITEM:
{
ReturnedColumn* rc = buildReturnedColumn(sfitemp, gwi, gwi.fatalParseError);
if (rc)
default:
{
parm.reset(rc);
break;
gwi.fatalParseError = true;
//gwi.parseErrorText = "Non-supported Item in Aggregate function";
}
}
default:
if (gwi.fatalParseError)
{
gwi.fatalParseError = true;
//gwi.parseErrorText = "Non-supported Item in Aggregate function";
}
}
if (gwi.parseErrorText.empty())
{
Message::Args args;
if (gwi.fatalParseError)
{
if (gwi.parseErrorText.empty())
{
Message::Args args;
if (item->name)
args.add(item->name);
else
args.add("");
if (item->name)
args.add(item->name);
else
args.add("");
gwi.parseErrorText = IDBErrorInfo::instance()->errorMsg(ERR_NON_SUPPORT_AGG_ARGS, args);
}
gwi.parseErrorText = IDBErrorInfo::instance()->errorMsg(ERR_NON_SUPPORT_AGG_ARGS, args);
if (ac)
delete ac;
return NULL;
}
if (ac)
delete ac;
return NULL;
}
if (parm)
{
// MCOL-1201 multi-argument aggregate
ac->aggParms().push_back(parm);
if (parm)
{
// MCOL-1201 multi-argument aggregate
ac->aggParms().push_back(parm);
}
}
}
}
// Get result type
// Modified for MCOL-1201 multi-argument aggregate
if (ac->aggParms().size() > 0)
{
{
// These are all one parm functions, so we can safely
// use the first parm for result type.
parm = ac->aggParms()[0];
if (isp->sum_func() == Item_sum::AVG_FUNC ||
isp->sum_func() == Item_sum::AVG_DISTINCT_FUNC)
{
CalpontSystemCatalog::ColType ct = parm->resultType();
switch (ct.colDataType)
if (isp->sum_func() == Item_sum::AVG_FUNC ||
isp->sum_func() == Item_sum::AVG_DISTINCT_FUNC)
{
case CalpontSystemCatalog::TINYINT:
case CalpontSystemCatalog::SMALLINT:
case CalpontSystemCatalog::MEDINT:
case CalpontSystemCatalog::INT:
case CalpontSystemCatalog::BIGINT:
case CalpontSystemCatalog::DECIMAL:
case CalpontSystemCatalog::UDECIMAL:
case CalpontSystemCatalog::UTINYINT:
case CalpontSystemCatalog::USMALLINT:
case CalpontSystemCatalog::UMEDINT:
case CalpontSystemCatalog::UINT:
case CalpontSystemCatalog::UBIGINT:
ct.colDataType = CalpontSystemCatalog::DECIMAL;
ct.colWidth = 8;
ct.scale += 4;
break;
CalpontSystemCatalog::ColType ct = parm->resultType();
switch (ct.colDataType)
{
case CalpontSystemCatalog::TINYINT:
case CalpontSystemCatalog::SMALLINT:
case CalpontSystemCatalog::MEDINT:
case CalpontSystemCatalog::INT:
case CalpontSystemCatalog::BIGINT:
case CalpontSystemCatalog::DECIMAL:
case CalpontSystemCatalog::UDECIMAL:
case CalpontSystemCatalog::UTINYINT:
case CalpontSystemCatalog::USMALLINT:
case CalpontSystemCatalog::UMEDINT:
case CalpontSystemCatalog::UINT:
case CalpontSystemCatalog::UBIGINT:
ct.colDataType = CalpontSystemCatalog::DECIMAL;
ct.colWidth = 8;
ct.scale += 4;
break;
#if PROMOTE_FLOAT_TO_DOUBLE_ON_SUM
case CalpontSystemCatalog::FLOAT:
case CalpontSystemCatalog::UFLOAT:
case CalpontSystemCatalog::DOUBLE:
case CalpontSystemCatalog::UDOUBLE:
ct.colDataType = CalpontSystemCatalog::DOUBLE;
ct.colWidth = 8;
break;
case CalpontSystemCatalog::FLOAT:
case CalpontSystemCatalog::UFLOAT:
case CalpontSystemCatalog::DOUBLE:
case CalpontSystemCatalog::UDOUBLE:
ct.colDataType = CalpontSystemCatalog::DOUBLE;
ct.colWidth = 8;
break;
#endif
default:
break;
default:
break;
}
ac->resultType(ct);
}
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::COUNT_FUNC ||
isp->sum_func() == Item_sum::COUNT_DISTINCT_FUNC)
{
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::BIGINT;
ct.colWidth = 8;
ct.scale = parm->resultType().scale;
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::SUM_FUNC ||
isp->sum_func() == Item_sum::SUM_DISTINCT_FUNC)
{
CalpontSystemCatalog::ColType ct = parm->resultType();
switch (ct.colDataType)
else if (isp->sum_func() == Item_sum::COUNT_FUNC ||
isp->sum_func() == Item_sum::COUNT_DISTINCT_FUNC)
{
case CalpontSystemCatalog::TINYINT:
case CalpontSystemCatalog::SMALLINT:
case CalpontSystemCatalog::MEDINT:
case CalpontSystemCatalog::INT:
case CalpontSystemCatalog::BIGINT:
ct.colDataType = CalpontSystemCatalog::BIGINT;
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::BIGINT;
ct.colWidth = 8;
ct.scale = parm->resultType().scale;
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::SUM_FUNC ||
isp->sum_func() == Item_sum::SUM_DISTINCT_FUNC)
{
CalpontSystemCatalog::ColType ct = parm->resultType();
// no break, let fall through
switch (ct.colDataType)
{
case CalpontSystemCatalog::TINYINT:
case CalpontSystemCatalog::SMALLINT:
case CalpontSystemCatalog::MEDINT:
case CalpontSystemCatalog::INT:
case CalpontSystemCatalog::BIGINT:
ct.colDataType = CalpontSystemCatalog::BIGINT;
case CalpontSystemCatalog::DECIMAL:
case CalpontSystemCatalog::UDECIMAL:
ct.colWidth = 8;
break;
// no break, let fall through
case CalpontSystemCatalog::UTINYINT:
case CalpontSystemCatalog::USMALLINT:
case CalpontSystemCatalog::UMEDINT:
case CalpontSystemCatalog::UINT:
case CalpontSystemCatalog::UBIGINT:
ct.colDataType = CalpontSystemCatalog::UBIGINT;
ct.colWidth = 8;
break;
case CalpontSystemCatalog::DECIMAL:
case CalpontSystemCatalog::UDECIMAL:
ct.colWidth = 8;
break;
case CalpontSystemCatalog::UTINYINT:
case CalpontSystemCatalog::USMALLINT:
case CalpontSystemCatalog::UMEDINT:
case CalpontSystemCatalog::UINT:
case CalpontSystemCatalog::UBIGINT:
ct.colDataType = CalpontSystemCatalog::UBIGINT;
ct.colWidth = 8;
break;
#if PROMOTE_FLOAT_TO_DOUBLE_ON_SUM
case CalpontSystemCatalog::FLOAT:
case CalpontSystemCatalog::UFLOAT:
case CalpontSystemCatalog::DOUBLE:
case CalpontSystemCatalog::UDOUBLE:
ct.colDataType = CalpontSystemCatalog::DOUBLE;
ct.colWidth = 8;
break;
case CalpontSystemCatalog::FLOAT:
case CalpontSystemCatalog::UFLOAT:
case CalpontSystemCatalog::DOUBLE:
case CalpontSystemCatalog::UDOUBLE:
ct.colDataType = CalpontSystemCatalog::DOUBLE;
ct.colWidth = 8;
break;
#endif
default:
break;
}
default:
break;
}
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::STD_FUNC ||
isp->sum_func() == Item_sum::VARIANCE_FUNC)
{
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::DOUBLE;
ct.colWidth = 8;
ct.scale = 0;
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::SUM_BIT_FUNC)
{
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::BIGINT;
ct.colWidth = 8;
ct.scale = 0;
ct.precision = -16; // borrowed to indicate skip null value check on connector
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::GROUP_CONCAT_FUNC)
{
//Item_func_group_concat* gc = (Item_func_group_concat*)isp;
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::VARCHAR;
ct.colWidth = isp->max_length;
ct.precision = 0;
ac->resultType(ct);
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::STD_FUNC ||
isp->sum_func() == Item_sum::VARIANCE_FUNC)
{
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::DOUBLE;
ct.colWidth = 8;
ct.scale = 0;
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::SUM_BIT_FUNC)
{
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::BIGINT;
ct.colWidth = 8;
ct.scale = 0;
ct.precision = -16; // borrowed to indicate skip null value check on connector
ac->resultType(ct);
}
else if (isp->sum_func() == Item_sum::GROUP_CONCAT_FUNC)
{
//Item_func_group_concat* gc = (Item_func_group_concat*)isp;
CalpontSystemCatalog::ColType ct;
ct.colDataType = CalpontSystemCatalog::VARCHAR;
ct.colWidth = isp->max_length;
ct.precision = 0;
ac->resultType(ct);
}
else
{
// UDAF result type will be set below.
ac->resultType(parm->resultType());
}
}
else
{
// UDAF result type will be set below.
ac->resultType(parm->resultType());
ac->resultType(colType_MysqlToIDB(isp));
}
}
else
{
ac->resultType(colType_MysqlToIDB(isp));
}
// adjust decimal result type according to internalDecimalScale
if (gwi.internalDecimalScale >= 0 && ac->resultType().colDataType == CalpontSystemCatalog::DECIMAL)
{
CalpontSystemCatalog::ColType ct = ac->resultType();
ct.scale = gwi.internalDecimalScale;
ac->resultType(ct);
}
// check for same aggregate on the select list
ac->expressionId(ci->expressionId++);
if (gwi.clauseType != SELECT)
{
for (uint32_t i = 0; i < gwi.returnedCols.size(); i++)
// adjust decimal result type according to internalDecimalScale
if (gwi.internalDecimalScale >= 0 && ac->resultType().colDataType == CalpontSystemCatalog::DECIMAL)
{
if (*ac == gwi.returnedCols[i].get())
ac->expressionId(gwi.returnedCols[i]->expressionId());
CalpontSystemCatalog::ColType ct = ac->resultType();
ct.scale = gwi.internalDecimalScale;
ac->resultType(ct);
}
}
// @bug5977 @note Temporary fix to avoid mysqld crash. The permanent fix will
// be applied in ExeMgr. When the ExeMgr fix is available, this checking
// will be taken out.
// check for same aggregate on the select list
ac->expressionId(ci->expressionId++);
if (gwi.clauseType != SELECT)
{
for (uint32_t i = 0; i < gwi.returnedCols.size(); i++)
{
if (*ac == gwi.returnedCols[i].get())
ac->expressionId(gwi.returnedCols[i]->expressionId());
}
}
// @bug5977 @note Temporary fix to avoid mysqld crash. The permanent fix will
// be applied in ExeMgr. When the ExeMgr fix is available, this checking
// will be taken out.
if (isp->sum_func() != Item_sum::UDF_SUM_FUNC)
{
if (ac->constCol() && gwi.tbList.empty() && gwi.derivedTbList.empty())
{
gwi.fatalParseError = true;
gwi.parseErrorText = "No project column found for aggregate function";
if (ac->constCol() && gwi.tbList.empty() && gwi.derivedTbList.empty())
{
gwi.fatalParseError = true;
gwi.parseErrorText = "No project column found for aggregate function";
if (ac)
delete ac;
return NULL;
}
else if (ac->constCol())
{
gwi.count_asterisk_list.push_back(ac);
}
return NULL;
}
else if (ac->constCol())
{
gwi.count_asterisk_list.push_back(ac);
}
}
// For UDAF, populate the context and call the UDAF init() function.
// For UDAF, populate the context and call the UDAF init() function.
// The return type is (should be) set in context by init().
if (isp->sum_func() == Item_sum::UDF_SUM_FUNC)
{
UDAFColumn* udafc = dynamic_cast<UDAFColumn*>(ac);
if (udafc)
if (isp->sum_func() == Item_sum::UDF_SUM_FUNC)
{
mcsv1Context& context = udafc->getContext();
context.setName(isp->func_name());
UDAFColumn* udafc = dynamic_cast<UDAFColumn*>(ac);
// Set up the return type defaults for the call to init()
context.setResultType(udafc->resultType().colDataType);
context.setColWidth(udafc->resultType().colWidth);
context.setScale(udafc->resultType().scale);
context.setPrecision(udafc->resultType().precision);
if (udafc)
{
mcsv1Context& context = udafc->getContext();
context.setName(isp->func_name());
// Set up the return type defaults for the call to init()
context.setResultType(udafc->resultType().colDataType);
context.setColWidth(udafc->resultType().colWidth);
context.setScale(udafc->resultType().scale);
context.setPrecision(udafc->resultType().precision);
context.setParamCount(udafc->aggParms().size());
ColumnDatum colType;
ColumnDatum colTypes[udafc->aggParms().size()];
// Build the column type vector.
// Modified for MCOL-1201 multi-argument aggregate
for (uint32_t i = 0; i < udafc->aggParms().size(); ++i)
{
const execplan::CalpontSystemCatalog::ColType& resultType
const execplan::CalpontSystemCatalog::ColType& resultType
= udafc->aggParms()[i]->resultType();
colType.dataType = resultType.colDataType;
colType.precision = resultType.precision;
@@ -4536,65 +4551,78 @@ ReturnedColumn* buildAggregateColumn(Item* item, gp_walk_info& gwi)
colTypes[i] = colType;
}
// Call the user supplied init()
// Call the user supplied init()
mcsv1sdk::mcsv1_UDAF* udaf = context.getFunction();
if (!udaf)
{
gwi.fatalParseError = true;
gwi.parseErrorText = "Aggregate Function " + context.getName() + " doesn't exist in the ColumnStore engine";
if (ac)
delete ac;
return NULL;
}
if (udaf->init(&context, colTypes) == mcsv1_UDAF::ERROR)
{
gwi.fatalParseError = true;
gwi.parseErrorText = udafc->getContext().getErrorMessage();
if (udaf->init(&context, colTypes) == mcsv1_UDAF::ERROR)
{
gwi.fatalParseError = true;
gwi.parseErrorText = udafc->getContext().getErrorMessage();
if (ac)
delete ac;
return NULL;
}
return NULL;
}
// UDAF_OVER_REQUIRED means that this function is for Window
// Function only. Reject it here in aggregate land.
if (udafc->getContext().getRunFlag(UDAF_OVER_REQUIRED))
{
gwi.fatalParseError = true;
gwi.parseErrorText =
logging::IDBErrorInfo::instance()->errorMsg(logging::ERR_WINDOW_FUNC_ONLY,
context.getName());
if (udafc->getContext().getRunFlag(UDAF_OVER_REQUIRED))
{
gwi.fatalParseError = true;
gwi.parseErrorText =
logging::IDBErrorInfo::instance()->errorMsg(logging::ERR_WINDOW_FUNC_ONLY,
context.getName());
if (ac)
delete ac;
return NULL;
}
// Set the return type as set in init()
CalpontSystemCatalog::ColType ct;
ct.colDataType = context.getResultType();
ct.colWidth = context.getColWidth();
ct.scale = context.getScale();
ct.precision = context.getPrecision();
udafc->resultType(ct);
return NULL;
}
// Set the return type as set in init()
CalpontSystemCatalog::ColType ct;
ct.colDataType = context.getResultType();
ct.colWidth = context.getColWidth();
ct.scale = context.getScale();
ct.precision = context.getPrecision();
udafc->resultType(ct);
}
}
}
}
catch (std::logic_error e)
{
gwi.fatalParseError = true;
gwi.parseErrorText = "error building Aggregate Function: ";
gwi.parseErrorText += e.what();
if (ac)
delete ac;
return NULL;
}
catch (...)
{
gwi.fatalParseError = true;
gwi.parseErrorText = "error building Aggregate Function: Unspecified exception";
if (ac)
delete ac;
return NULL;
}
return ac;
}
@@ -7915,6 +7943,7 @@ int getSelectPlan(gp_walk_info& gwi, SELECT_LEX& select_lex, SCSEP& csep, bool i
setError(gwi.thd, ER_INTERNAL_ERROR, gwi.parseErrorText, gwi);
return ER_CHECK_NOT_IMPLEMENTED;
}
// Replace the last (presumably constant) object with minSc
if ((*coliter)->aggParms().empty())
{

3
dbcon/mysql/ha_calpont_impl.cpp
View File

@@ -782,7 +782,7 @@ int fetchNextRow(uchar* buf, cal_table_info& ti, cal_connection_info* ci, bool h
//double double_val = *(double*)(&value);
//f2->store(double_val);
if ((f2->decimals() == DECIMAL_NOT_SPECIFIED && row.getScale(s) > 0)
|| f2->decimals() < row.getScale(s))
|| f2->decimals() < row.getScale(s))
{
f2->dec = row.getScale(s);
}
@@ -5278,6 +5278,7 @@ int ha_calpont_impl_group_by_init(ha_calpont_group_by_handler* group_hand, TABLE
execplan::CalpontSelectExecutionPlan::ColumnMap::iterator colMapIter;
execplan::CalpontSelectExecutionPlan::ColumnMap::iterator condColMapIter;
execplan::ParseTree* ptIt;
for (TABLE_LIST* tl = gi.groupByTables; tl; tl = tl->next_local)
{
mapiter = ci->tableMap.find(tl->table);

2
dbcon/mysql/ha_window_function.cpp
View File

@@ -384,7 +384,7 @@ ReturnedColumn* buildWindowFunctionColumn(Item* item, gp_walk_info& gwi, bool& n
// Modified for MCOL-1201 multi-argument aggregate
for (size_t i = 0; i < funcParms.size(); ++i)
{
const execplan::CalpontSystemCatalog::ColType& resultType
const execplan::CalpontSystemCatalog::ColType& resultType
= funcParms[i]->resultType();
colType.dataType = resultType.colDataType;
colType.precision = resultType.precision;

75
dbcon/mysql/is_columnstore_columns.cpp
View File

@@ -56,10 +56,62 @@ ST_FIELD_INFO is_columnstore_columns_fields[] =
};
static void get_cond_item(Item_func* item, String** table, String** db)
{
char tmp_char[MAX_FIELD_WIDTH];
Item_field* item_field = (Item_field*) item->arguments()[0]->real_item();
if (strcasecmp(item_field->field_name, "table_name") == 0)
{
String str_buf(tmp_char, sizeof(tmp_char), system_charset_info);
*table = item->arguments()[1]->val_str(&str_buf);
return;
}
else if (strcasecmp(item_field->field_name, "table_schema") == 0)
{
String str_buf(tmp_char, sizeof(tmp_char), system_charset_info);
*db = item->arguments()[1]->val_str(&str_buf);
return;
}
}
static void get_cond_items(COND* cond, String** table, String** db)
{
if (cond->type() == Item::FUNC_ITEM)
{
Item_func* fitem = (Item_func*) cond;
if (fitem->arguments()[0]->real_item()->type() == Item::FIELD_ITEM &&
fitem->arguments()[1]->const_item())
{
get_cond_item(fitem, table, db);
}
}
else if ((cond->type() == Item::COND_ITEM) && (((Item_cond*) cond)->functype() == Item_func::COND_AND_FUNC))
{
List_iterator<Item> li(*((Item_cond*) cond)->argument_list());
Item* item;
while ((item = li++))
{
if (item->type() == Item::FUNC_ITEM)
{
get_cond_item((Item_func*)item, table, db);
}
else
{
get_cond_items(item, table, db);
}
}
}
}
static int is_columnstore_columns_fill(THD* thd, TABLE_LIST* tables, COND* cond)
{
CHARSET_INFO* cs = system_charset_info;
TABLE* table = tables->table;
String* table_name = NULL;
String* db_name = NULL;
boost::shared_ptr<execplan::CalpontSystemCatalog> systemCatalogPtr =
execplan::CalpontSystemCatalog::makeCalpontSystemCatalog(execplan::CalpontSystemCatalog::idb_tid2sid(thd->thread_id));
@@ -69,9 +121,30 @@ static int is_columnstore_columns_fill(THD* thd, TABLE_LIST* tables, COND* cond)
systemCatalogPtr->identity(execplan::CalpontSystemCatalog::FE);
if (cond)
{
get_cond_items(cond, &table_name, &db_name);
}
for (std::vector<std::pair<execplan::CalpontSystemCatalog::OID, execplan::CalpontSystemCatalog::TableName> >::const_iterator it = catalog_tables.begin();
it != catalog_tables.end(); ++it)
{
if (db_name)
{
if ((*it).second.schema.compare(db_name->ptr()) != 0)
{
continue;
}
}
if (table_name)
{
if ((*it).second.table.compare(table_name->ptr()) != 0)
{
continue;
}
}
execplan::CalpontSystemCatalog::RIDList column_rid_list;
// Note a table may get dropped as you iterate over the list of tables.
@@ -184,8 +257,6 @@ static int is_columnstore_columns_fill(THD* thd, TABLE_LIST* tables, COND* cond)
}
}
return 0;
}

321
dbcon/mysql/is_columnstore_extents.cpp
View File

@@ -52,14 +52,142 @@ ST_FIELD_INFO is_columnstore_extents_fields[] =
{0, 0, MYSQL_TYPE_NULL, 0, 0, 0, 0}
};
static int is_columnstore_extents_fill(THD* thd, TABLE_LIST* tables, COND* cond)
static int generate_result(BRM::OID_t oid, BRM::DBRM* emp, TABLE* table, THD* thd)
{
CHARSET_INFO* cs = system_charset_info;
TABLE* table = tables->table;
std::vector<struct BRM::EMEntry> entries;
std::vector<struct BRM::EMEntry>::iterator iter;
std::vector<struct BRM::EMEntry>::iterator end;
emp->getExtents(oid, entries, false, false, true);
if (entries.size() == 0)
return 0;
iter = entries.begin();
end = entries.end();
while (iter != end)
{
table->field[0]->store(oid);
if (iter->colWid > 0)
{
table->field[1]->store("Column", strlen("Column"), cs);
if (iter->partition.cprange.lo_val == std::numeric_limits<int64_t>::max() ||
iter->partition.cprange.lo_val <= (std::numeric_limits<int64_t>::min() + 2))
{
table->field[4]->set_null();
}
else
{
table->field[4]->set_notnull();
table->field[4]->store(iter->partition.cprange.lo_val);
}
if (iter->partition.cprange.hi_val == std::numeric_limits<int64_t>::max() ||
iter->partition.cprange.hi_val <= (std::numeric_limits<int64_t>::min() + 2))
{
table->field[5]->set_null();
}
else
{
table->field[5]->set_notnull();
table->field[5]->store(iter->partition.cprange.hi_val);
}
table->field[6]->store(iter->colWid);
}
else
{
table->field[1]->store("Dictionary", strlen("Dictionary"), cs);
table->field[4]->set_null();
table->field[5]->set_null();
table->field[6]->store(8192);
}
table->field[2]->store(iter->range.start);
table->field[3]->store(iter->range.start + (iter->range.size * 1024) - 1);
table->field[7]->store(iter->dbRoot);
table->field[8]->store(iter->partitionNum);
table->field[9]->store(iter->segmentNum);
table->field[10]->store(iter->blockOffset);
table->field[11]->store(iter->range.size * 1024);
table->field[12]->store(iter->HWM);
switch (iter->partition.cprange.isValid)
{
case 0:
table->field[13]->store("Invalid", strlen("Invalid"), cs);
break;
case 1:
table->field[13]->store("Updating", strlen("Updating"), cs);
break;
case 2:
table->field[13]->store("Valid", strlen("Valid"), cs);
break;
default:
table->field[13]->store("Unknown", strlen("Unknown"), cs);
break;
}
switch (iter->status)
{
case BRM::EXTENTAVAILABLE:
table->field[14]->store("Available", strlen("Available"), cs);
break;
case BRM::EXTENTUNAVAILABLE:
table->field[14]->store("Unavailable", strlen("Unavailable"), cs);
break;
case BRM::EXTENTOUTOFSERVICE:
table->field[14]->store("Out of service", strlen("Out of service"), cs);
break;
default:
table->field[14]->store("Unknown", strlen("Unknown"), cs);
}
// MCOL-1016: on multiple segments HWM is set to 0 on the lower
// segments, we don't want these to show as 8KB. The down side is
// if the column has less than 1 block it will show as 0 bytes.
// We have no lookahead without it getting messy so this is the
// best compromise.
if (iter->HWM == 0)
{
table->field[15]->store(0);
}
else
{
table->field[15]->store((iter->HWM + 1) * 8192);
}
if (schema_table_store_record(thd, table))
{
delete emp;
return 1;
}
iter++;
}
return 0;
}
static int is_columnstore_extents_fill(THD* thd, TABLE_LIST* tables, COND* cond)
{
BRM::OID_t cond_oid = 0;
TABLE* table = tables->table;
BRM::DBRM* emp = new BRM::DBRM();
if (!emp || !emp->isDBRMReady())
@@ -67,130 +195,83 @@ static int is_columnstore_extents_fill(THD* thd, TABLE_LIST* tables, COND* cond)
return 1;
}
if (cond && cond->type() == Item::FUNC_ITEM)
{
Item_func* fitem = (Item_func*) cond;
if ((fitem->functype() == Item_func::EQ_FUNC) && (fitem->argument_count() == 2))
{
if (fitem->arguments()[0]->real_item()->type() == Item::FIELD_ITEM &&
fitem->arguments()[1]->const_item())
{
// WHERE object_id = value
Item_field* item_field = (Item_field*) fitem->arguments()[0]->real_item();
if (strcasecmp(item_field->field_name, "object_id") == 0)
{
cond_oid = fitem->arguments()[1]->val_int();
return generate_result(cond_oid, emp, table, thd);
}
}
else if (fitem->arguments()[1]->real_item()->type() == Item::FIELD_ITEM &&
fitem->arguments()[0]->const_item())
{
// WHERE value = object_id
Item_field* item_field = (Item_field*) fitem->arguments()[1]->real_item();
if (strcasecmp(item_field->field_name, "object_id") == 0)
{
cond_oid = fitem->arguments()[0]->val_int();
return generate_result(cond_oid, emp, table, thd);
}
}
}
else if (fitem->functype() == Item_func::IN_FUNC)
{
// WHERE object_id in (value1, value2)
Item_field* item_field = (Item_field*) fitem->arguments()[0]->real_item();
if (strcasecmp(item_field->field_name, "object_id") == 0)
{
for (unsigned int i = 1; i < fitem->argument_count(); i++)
{
cond_oid = fitem->arguments()[i]->val_int();
int result = generate_result(cond_oid, emp, table, thd);
if (result)
return 1;
}
}
}
else if (fitem->functype() == Item_func::UNKNOWN_FUNC &&
strcasecmp(fitem->func_name(), "find_in_set") == 0)
{
// WHERE FIND_IN_SET(object_id, values)
String* tmp_var = fitem->arguments()[1]->val_str();
std::stringstream ss(tmp_var->ptr());
while (ss >> cond_oid)
{
int ret = generate_result(cond_oid, emp, table, thd);
if (ret)
return 1;
if (ss.peek() == ',')
ss.ignore();
}
}
}
execplan::ObjectIDManager oidm;
BRM::OID_t MaxOID = oidm.size();
for (BRM::OID_t oid = 3000; oid <= MaxOID; oid++)
{
emp->getExtents(oid, entries, false, false, true);
int result = generate_result(oid, emp, table, thd);
if (entries.size() == 0)
continue;
iter = entries.begin();
end = entries.end();
while (iter != end)
{
table->field[0]->store(oid);
if (iter->colWid > 0)
{
table->field[1]->store("Column", strlen("Column"), cs);
if (iter->partition.cprange.lo_val == std::numeric_limits<int64_t>::max() ||
iter->partition.cprange.lo_val <= (std::numeric_limits<int64_t>::min() + 2))
{
table->field[4]->set_null();
}
else
{
table->field[4]->set_notnull();
table->field[4]->store(iter->partition.cprange.lo_val);
}
if (iter->partition.cprange.hi_val == std::numeric_limits<int64_t>::max() ||
iter->partition.cprange.hi_val <= (std::numeric_limits<int64_t>::min() + 2))
{
table->field[5]->set_null();
}
else
{
table->field[5]->set_notnull();
table->field[5]->store(iter->partition.cprange.hi_val);
}
table->field[6]->store(iter->colWid);
}
else
{
table->field[1]->store("Dictionary", strlen("Dictionary"), cs);
table->field[4]->set_null();
table->field[5]->set_null();
table->field[6]->store(8192);
}
table->field[2]->store(iter->range.start);
table->field[3]->store(iter->range.start + (iter->range.size * 1024) - 1);
table->field[7]->store(iter->dbRoot);
table->field[8]->store(iter->partitionNum);
table->field[9]->store(iter->segmentNum);
table->field[10]->store(iter->blockOffset);
table->field[11]->store(iter->range.size * 1024);
table->field[12]->store(iter->HWM);
switch (iter->partition.cprange.isValid)
{
case 0:
table->field[13]->store("Invalid", strlen("Invalid"), cs);
break;
case 1:
table->field[13]->store("Updating", strlen("Updating"), cs);
break;
case 2:
table->field[13]->store("Valid", strlen("Valid"), cs);
break;
default:
table->field[13]->store("Unknown", strlen("Unknown"), cs);
break;
}
switch (iter->status)
{
case BRM::EXTENTAVAILABLE:
table->field[14]->store("Available", strlen("Available"), cs);
break;
case BRM::EXTENTUNAVAILABLE:
table->field[14]->store("Unavailable", strlen("Unavailable"), cs);
break;
case BRM::EXTENTOUTOFSERVICE:
table->field[14]->store("Out of service", strlen("Out of service"), cs);
break;
default:
table->field[14]->store("Unknown", strlen("Unknown"), cs);
}
// MCOL-1016: on multiple segments HWM is set to 0 on the lower
// segments, we don't want these to show as 8KB. The down side is
// if the column has less than 1 block it will show as 0 bytes.
// We have no lookahead without it getting messy so this is the
// best compromise.
if (iter->HWM == 0)
{
table->field[15]->store(0);
}
else
{
table->field[15]->store((iter->HWM + 1) * 8192);
}
if (schema_table_store_record(thd, table))
{
delete emp;
return 1;
}
iter++;
}
if (result)
return 1;
}
delete emp;

251
dbcon/mysql/is_columnstore_files.cpp
View File

@@ -84,12 +84,10 @@ static bool get_file_sizes(messageqcpp::MessageQueueClient* msgQueueClient, cons
}
}
static int is_columnstore_files_fill(THD* thd, TABLE_LIST* tables, COND* cond)
static int generate_result(BRM::OID_t oid, BRM::DBRM* emp, TABLE* table, THD* thd)
{
BRM::DBRM* emp = new BRM::DBRM();
std::vector<struct BRM::EMEntry> entries;
CHARSET_INFO* cs = system_charset_info;
TABLE* table = tables->table;
char oidDirName[WriteEngine::FILE_NAME_SIZE];
char fullFileName[WriteEngine::FILE_NAME_SIZE];
@@ -103,99 +101,184 @@ static int is_columnstore_files_fill(THD* thd, TABLE_LIST* tables, COND* cond)
oam::Oam oam_instance;
int pmId = 0;
emp->getExtents(oid, entries, false, false, true);
if (entries.size() == 0)
return 0;
std::vector<struct BRM::EMEntry>::const_iterator iter = entries.begin();
while ( iter != entries.end() ) //organize extents into files
{
// Don't include files more than once at different block offsets
if (iter->blockOffset > 0)
{
iter++;
return 0;
}
try
{
oam_instance.getDbrootPmConfig(iter->dbRoot, pmId);
}
catch (std::runtime_error)
{
// MCOL-1116: If we are here a DBRoot is offline/missing
iter++;
return 0;
}
table->field[0]->store(oid);
table->field[1]->store(iter->segmentNum);
table->field[2]->store(iter->partitionNum);
WriteEngine::Convertor::oid2FileName(oid, oidDirName, dbDir, iter->partitionNum, iter->segmentNum);
std::stringstream DbRootName;
DbRootName << "DBRoot" << iter->dbRoot;
std::string DbRootPath = config->getConfig("SystemConfig", DbRootName.str());
fileSize = compressedFileSize = 0;
snprintf(fullFileName, WriteEngine::FILE_NAME_SIZE, "%s/%s", DbRootPath.c_str(), oidDirName);
std::ostringstream oss;
oss << "pm" << pmId << "_WriteEngineServer";
std::string client = oss.str();
msgQueueClient = messageqcpp::MessageQueueClientPool::getInstance(oss.str());
if (!get_file_sizes(msgQueueClient, fullFileName, &fileSize, &compressedFileSize))
{
messageqcpp::MessageQueueClientPool::releaseInstance(msgQueueClient);
delete emp;
return 1;
}
table->field[3]->store(fullFileName, strlen(fullFileName), cs);
if (fileSize > 0)
{
table->field[4]->set_notnull();
table->field[4]->store(fileSize);
if (compressedFileSize > 0)
{
table->field[5]->set_notnull();
table->field[5]->store(compressedFileSize);
}
else
{
table->field[5]->set_null();
}
}
else
{
table->field[4]->set_null();
table->field[5]->set_null();
}
if (schema_table_store_record(thd, table))
{
messageqcpp::MessageQueueClientPool::releaseInstance(msgQueueClient);
delete emp;
return 1;
}
iter++;
messageqcpp::MessageQueueClientPool::releaseInstance(msgQueueClient);
msgQueueClient = NULL;
}
return 0;
}
static int is_columnstore_files_fill(THD* thd, TABLE_LIST* tables, COND* cond)
{
BRM::DBRM* emp = new BRM::DBRM();
BRM::OID_t cond_oid = 0;
TABLE* table = tables->table;
if (!emp || !emp->isDBRMReady())
{
return 1;
}
if (cond && cond->type() == Item::FUNC_ITEM)
{
Item_func* fitem = (Item_func*) cond;
if ((fitem->functype() == Item_func::EQ_FUNC) && (fitem->argument_count() == 2))
{
if (fitem->arguments()[0]->real_item()->type() == Item::FIELD_ITEM &&
fitem->arguments()[1]->const_item())
{
// WHERE object_id = value
Item_field* item_field = (Item_field*) fitem->arguments()[0]->real_item();
if (strcasecmp(item_field->field_name, "object_id") == 0)
{
cond_oid = fitem->arguments()[1]->val_int();
return generate_result(cond_oid, emp, table, thd);
}
}
else if (fitem->arguments()[1]->real_item()->type() == Item::FIELD_ITEM &&
fitem->arguments()[0]->const_item())
{
// WHERE value = object_id
Item_field* item_field = (Item_field*) fitem->arguments()[1]->real_item();
if (strcasecmp(item_field->field_name, "object_id") == 0)
{
cond_oid = fitem->arguments()[0]->val_int();
return generate_result(cond_oid, emp, table, thd);
}
}
}
else if (fitem->functype() == Item_func::IN_FUNC)
{
// WHERE object_id in (value1, value2)
Item_field* item_field = (Item_field*) fitem->arguments()[0]->real_item();
if (strcasecmp(item_field->field_name, "object_id") == 0)
{
for (unsigned int i = 1; i < fitem->argument_count(); i++)
{
cond_oid = fitem->arguments()[i]->val_int();
int result = generate_result(cond_oid, emp, table, thd);
if (result)
return 1;
}
}
}
else if (fitem->functype() == Item_func::UNKNOWN_FUNC &&
strcasecmp(fitem->func_name(), "find_in_set") == 0)
{
// WHERE FIND_IN_SET(object_id, values)
String* tmp_var = fitem->arguments()[1]->val_str();
std::stringstream ss(tmp_var->ptr());
while (ss >> cond_oid)
{
int ret = generate_result(cond_oid, emp, table, thd);
if (ret)
return 1;
if (ss.peek() == ',')
ss.ignore();
}
}
}
execplan::ObjectIDManager oidm;
BRM::OID_t MaxOID = oidm.size();
for (BRM::OID_t oid = 3000; oid <= MaxOID; oid++)
if (!cond_oid)
{
emp->getExtents(oid, entries, false, false, true);
if (entries.size() == 0)
continue;
std::vector<struct BRM::EMEntry>::const_iterator iter = entries.begin();
while ( iter != entries.end() ) //organize extents into files
for (BRM::OID_t oid = 3000; oid <= MaxOID; oid++)
{
// Don't include files more than once at different block offsets
if (iter->blockOffset > 0)
{
iter++;
continue;
}
int result = generate_result(oid, emp, table, thd);
try
{
oam_instance.getDbrootPmConfig(iter->dbRoot, pmId);
}
catch (std::runtime_error)
{
// MCOL-1116: If we are here a DBRoot is offline/missing
iter++;
continue;
}
table->field[0]->store(oid);
table->field[1]->store(iter->segmentNum);
table->field[2]->store(iter->partitionNum);
WriteEngine::Convertor::oid2FileName(oid, oidDirName, dbDir, iter->partitionNum, iter->segmentNum);
std::stringstream DbRootName;
DbRootName << "DBRoot" << iter->dbRoot;
std::string DbRootPath = config->getConfig("SystemConfig", DbRootName.str());
fileSize = compressedFileSize = 0;
snprintf(fullFileName, WriteEngine::FILE_NAME_SIZE, "%s/%s", DbRootPath.c_str(), oidDirName);
std::ostringstream oss;
oss << "pm" << pmId << "_WriteEngineServer";
std::string client = oss.str();
msgQueueClient = messageqcpp::MessageQueueClientPool::getInstance(oss.str());
if (!get_file_sizes(msgQueueClient, fullFileName, &fileSize, &compressedFileSize))
{
messageqcpp::MessageQueueClientPool::releaseInstance(msgQueueClient);
delete emp;
if (result)
return 1;
}
table->field[3]->store(fullFileName, strlen(fullFileName), cs);
if (fileSize > 0)
{
table->field[4]->set_notnull();
table->field[4]->store(fileSize);
if (compressedFileSize > 0)
{
table->field[5]->set_notnull();
table->field[5]->store(compressedFileSize);
}
else
{
table->field[5]->set_null();
}
}
else
{
table->field[4]->set_null();
table->field[5]->set_null();
}
if (schema_table_store_record(thd, table))
{
messageqcpp::MessageQueueClientPool::releaseInstance(msgQueueClient);
delete emp;
return 1;
}
iter++;
messageqcpp::MessageQueueClientPool::releaseInstance(msgQueueClient);
msgQueueClient = NULL;
}
}

73
dbcon/mysql/is_columnstore_tables.cpp
View File

@@ -42,22 +42,95 @@ ST_FIELD_INFO is_columnstore_tables_fields[] =
{0, 0, MYSQL_TYPE_NULL, 0, 0, 0, 0}
};
static void get_cond_item(Item_func* item, String** table, String** db)
{
char tmp_char[MAX_FIELD_WIDTH];
Item_field* item_field = (Item_field*) item->arguments()[0]->real_item();
if (strcasecmp(item_field->field_name, "table_name") == 0)
{
String str_buf(tmp_char, sizeof(tmp_char), system_charset_info);
*table = item->arguments()[1]->val_str(&str_buf);
return;
}
else if (strcasecmp(item_field->field_name, "table_schema") == 0)
{
String str_buf(tmp_char, sizeof(tmp_char), system_charset_info);
*db = item->arguments()[1]->val_str(&str_buf);
return;
}
}
static void get_cond_items(COND* cond, String** table, String** db)
{
if (cond->type() == Item::FUNC_ITEM)
{
Item_func* fitem = (Item_func*) cond;
if (fitem->arguments()[0]->real_item()->type() == Item::FIELD_ITEM &&
fitem->arguments()[1]->const_item())
{
get_cond_item(fitem, table, db);
}
}
else if ((cond->type() == Item::COND_ITEM) && (((Item_cond*) cond)->functype() == Item_func::COND_AND_FUNC))
{
List_iterator<Item> li(*((Item_cond*) cond)->argument_list());
Item* item;
while ((item = li++))
{
if (item->type() == Item::FUNC_ITEM)
{
get_cond_item((Item_func*)item, table, db);
}
else
{
get_cond_items(item, table, db);
}
}
}
}
static int is_columnstore_tables_fill(THD* thd, TABLE_LIST* tables, COND* cond)
{
CHARSET_INFO* cs = system_charset_info;
TABLE* table = tables->table;
String* table_name = NULL;
String* db_name = NULL;
boost::shared_ptr<execplan::CalpontSystemCatalog> systemCatalogPtr =
execplan::CalpontSystemCatalog::makeCalpontSystemCatalog(execplan::CalpontSystemCatalog::idb_tid2sid(thd->thread_id));
systemCatalogPtr->identity(execplan::CalpontSystemCatalog::FE);
if (cond)
{
get_cond_items(cond, &table_name, &db_name);
}
const std::vector< std::pair<execplan::CalpontSystemCatalog::OID, execplan::CalpontSystemCatalog::TableName> > catalog_tables
= systemCatalogPtr->getTables();
for (std::vector<std::pair<execplan::CalpontSystemCatalog::OID, execplan::CalpontSystemCatalog::TableName> >::const_iterator it = catalog_tables.begin();
it != catalog_tables.end(); ++it)
{
if (db_name)
{
if ((*it).second.schema.compare(db_name->ptr()) != 0)
{
continue;
}
}
if (table_name)
{
if ((*it).second.table.compare(table_name->ptr()) != 0)
{
continue;
}
}
execplan::CalpontSystemCatalog::TableInfo tb_info = systemCatalogPtr->tableInfo((*it).second);
std::string create_date = dataconvert::DataConvert::dateToString((*it).second.create_date);
table->field[0]->store((*it).second.schema.c_str(), (*it).second.schema.length(), cs);