database/mariadb-columnstore-engine
1
0
Fork 0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-07-29 08:21:15 +03:00
Code Activity

MCOL-641 1. Templatized convertValueNum() function.

Browse Source 
2. Allocate int128_t buffers in batchprimitiveprocessor if
a query involves wide decimal columns.
This commit is contained in:
Gagan Goel
2020-03-10 19:41:25 -04:00
committed by Roman Nozdrin
parent 9b714274db
commit 62d0c82d75
23 changed files with 235 additions and 105 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
dbcon/execplan/calpontsystemcatalog.h
View File

@ -44,6 +44,7 @@
#include "bytestream.h"
#include "joblisttypes.h"
#include "stdexcept"
#include "widedecimalutils.h"
#undef min
#undef max
@ -1046,6 +1047,11 @@ inline bool isSignedInteger(const execplan::CalpontSystemCatalog::ColDataType ty
}
}
inline bool isNull(int128_t val, const execplan::CalpontSystemCatalog::ColType& ct)
{
return utils::isWideDecimalNullValue(val);
}
inline bool isNull(int64_t val, const execplan::CalpontSystemCatalog::ColType& ct)
{
bool ret = false;

21
dbcon/joblist/batchprimitiveprocessor-jl.cpp
View File

@ -60,6 +60,7 @@ BatchPrimitiveProcessorJL::BatchPrimitiveProcessorJL(const ResourceManager* rm)
baseRid(0),
ridCount(0),
needStrValues(false),
hasWideDecimalType(false),
filterCount(0),
projectCount(0),
needRidsAtDelivery(false),
@ -100,6 +101,8 @@ void BatchPrimitiveProcessorJL::addFilterStep(const pColScanStep& scan, vector<B
filterSteps.push_back(cc);
filterCount++;
_hasScan = true;
if (utils::isWide(cc->getWidth()))
hasWideDecimalType = true;
idbassert(sessionID == scan.sessionId());
}
@ -114,6 +117,9 @@ void BatchPrimitiveProcessorJL::addFilterStep(const PseudoColStep& pcs)
cc->setStepUuid(uuid);
filterSteps.push_back(cc);
filterCount++;
// TODO MCOL-641 How do we get to this execution path?
//if (utils::isWide(cc->getWidth()))
// hasWideDecimalType = true;
idbassert(sessionID == pcs.sessionId());
}
@ -128,6 +134,8 @@ void BatchPrimitiveProcessorJL::addFilterStep(const pColStep& step)
cc->setStepUuid(uuid);
filterSteps.push_back(cc);
filterCount++;
if (utils::isWide(cc->getWidth()))
hasWideDecimalType = true;
idbassert(sessionID == step.sessionId());
}
@ -182,6 +190,9 @@ void BatchPrimitiveProcessorJL::addProjectStep(const PseudoColStep& step)
colWidths.push_back(cc->getWidth());
tupleLength += cc->getWidth();
projectCount++;
// TODO MCOL-641 How do we get to this execution path?
//if (utils::isWide(cc->getWidth()))
// hasWideDecimalType = true;
idbassert(sessionID == step.sessionId());
}
@ -198,6 +209,8 @@ void BatchPrimitiveProcessorJL::addProjectStep(const pColStep& step)
colWidths.push_back(cc->getWidth());
tupleLength += cc->getWidth();
projectCount++;
if (utils::isWide(cc->getWidth()))
hasWideDecimalType = true;
idbassert(sessionID == step.sessionId());
}
@ -215,6 +228,9 @@ void BatchPrimitiveProcessorJL::addProjectStep(const PassThruStep& step)
tupleLength += cc->getWidth();
projectCount++;
if (utils::isWide(cc->getWidth()))
hasWideDecimalType = true;
if (filterCount == 0 && !sendRowGroups)
sendValues = true;
@ -958,7 +974,7 @@ void BatchPrimitiveProcessorJL::createBPP(ByteStream& bs) const
{
ISMPacketHeader ism;
uint32_t i;
uint8_t flags = 0;
uint16_t flags = 0;
ism.Command = BATCH_PRIMITIVE_CREATE;
@ -994,6 +1010,9 @@ void BatchPrimitiveProcessorJL::createBPP(ByteStream& bs) const
if (sendTupleJoinRowGroupData)
flags |= JOIN_ROWGROUP_DATA;
if (hasWideDecimalType)
flags |= HAS_WIDE_DECIMAL;
bs << flags;
bs << bop;

5
dbcon/joblist/batchprimitiveprocessor-jl.h
View File

@ -281,10 +281,15 @@ private:
uint16_t relRids[LOGICAL_BLOCK_RIDS];
boost::scoped_array<uint64_t> absRids;
// TODO MCOL-641 Do we need uint128_t buffers here?
// When would sendValues=true, in which case values[]
// is sent to primproc?
uint64_t values[LOGICAL_BLOCK_RIDS];
uint16_t ridCount;
bool needStrValues;
bool hasWideDecimalType;
std::vector<SCommand> filterSteps;
std::vector<SCommand> projectSteps;
//@bug 1136

2
dbcon/joblist/jlf_common.cpp
View File

@ -332,7 +332,7 @@ string extractTableAlias(const SSC& sc)
//------------------------------------------------------------------------------
CalpontSystemCatalog::OID isDictCol(const CalpontSystemCatalog::ColType& colType)
{
if (colType.colDataType == CalpontSystemCatalog::BINARY) return 0;
if (utils::isWideDecimalType(colType)) return 0;
if (colType.colWidth > 8) return colType.ddn.dictOID;

56
dbcon/joblist/jlf_execplantojoblist.cpp
View File

@ -88,7 +88,7 @@ using namespace logging;
#include "jlf_common.h"
#include "jlf_subquery.h"
#include "jlf_tuplejoblist.h"
#include "columnwidth.h"
namespace
{
@ -312,11 +312,17 @@ int64_t valueNullNum(const CalpontSystemCatalog::ColType& ct, const string& time
return n;
}
int64_t convertValueNum(const string& str, const CalpontSystemCatalog::ColType& ct, bool isNull, uint8_t& rf, const string& timeZone)
template <typename T>
void convertValueNum(const string& str, const CalpontSystemCatalog::ColType& ct, bool isNull, uint8_t& rf, const string& timeZone, T& v)
{
if (str.size() == 0 || isNull ) return valueNullNum(ct, timeZone);
if (str.size() == 0 || isNull )
{
v = valueNullNum(ct, timeZone);
return;
}
int64_t v = 0;
v = 0;
rf = 0;
bool pushWarning = false;
boost::any anyVal = DataConvert::convertColumnData(ct, str, pushWarning, timeZone, false, true, false);
@ -450,8 +456,10 @@ int64_t convertValueNum(const string& str, const CalpontSystemCatalog::ColType&
#else
v = boost::any_cast<int32_t>(anyVal);
#endif
else
else if (ct.colWidth == execplan::CalpontSystemCatalog::EIGHT_BYTE)
v = boost::any_cast<long long>(anyVal);
else
v = boost::any_cast<int128_t>(anyVal);
break;
@ -485,8 +493,6 @@ int64_t convertValueNum(const string& str, const CalpontSystemCatalog::ColType&
rf = (data[0] == '-') ? ROUND_NEG : ROUND_POS;
}
return v;
}
//TODO: make this totaly case-insensitive
@ -1840,8 +1846,8 @@ const JobStepVector doSimpleFilter(SimpleFilter* sf, JobInfo& jobInfo)
{
// @bug 1151 string longer than colwidth of char/varchar.
int64_t value = 0;
int128_t value128 = 0;
uint8_t rf = 0;
unsigned __int128 val128 = 0;
#ifdef FAILED_ATOI_IS_ZERO
//if cvn throws (because there's non-digit data in the string, treat that as zero rather than
@ -1849,7 +1855,7 @@ const JobStepVector doSimpleFilter(SimpleFilter* sf, JobInfo& jobInfo)
try
{
bool isNull = ConstantColumn::NULLDATA == cc->type();
value = convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone);
convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone, value);
if (ct.colDataType == CalpontSystemCatalog::FLOAT && !isNull)
{
@ -1887,21 +1893,14 @@ const JobStepVector doSimpleFilter(SimpleFilter* sf, JobInfo& jobInfo)
}
#else
bool isNull = ConstantColumn::NULLDATA == cc->type();
// WIP MCOL-641 width check must be a f() not a literal
// make a template from convertValueNum to avoid extra if
// this condition doesn't support UDECIMAL
if (ct.colDataType == CalpontSystemCatalog::DECIMAL &&
ct.colWidth == 16)
{
bool saturate = false;
val128 = dataconvert::string_to_ll<int128_t>(constval, saturate);
// TODO MCOL-641 check saturate
}
if (utils::isWideDecimalType(ct))
convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone, value128);
else
{
bool isNull = ConstantColumn::NULLDATA == cc->type();
value = convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone);
}
convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone, value);
if (ct.colDataType == CalpontSystemCatalog::FLOAT && !isNull)
{
@ -1935,10 +1934,8 @@ const JobStepVector doSimpleFilter(SimpleFilter* sf, JobInfo& jobInfo)
if (sc->isColumnStore())
{
// WIP MCOL-641
if (ct.colDataType == CalpontSystemCatalog::DECIMAL &&
ct.colWidth == 16)
pcs->addFilter(cop, val128, rf);
if (utils::isWideDecimalType(ct))
pcs->addFilter(cop, value128, rf);
else
pcs->addFilter(cop, value, rf);
}
@ -3008,12 +3005,17 @@ const JobStepVector doConstantFilter(const ConstantFilter* cf, JobInfo& jobInfo)
//add each filter to pColStep
int8_t cop = op2num(sop);
int64_t value = 0;
int128_t value128 = 0;
string constval = cc->constval();
// @bug 1151 string longer than colwidth of char/varchar.
uint8_t rf = 0;
bool isNull = ConstantColumn::NULLDATA == cc->type();
value = convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone);
if (utils::isWideDecimalType(ct))
convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone, value128);
else
convertValueNum(constval, ct, isNull, rf, jobInfo.timeZone, value);
if (ct.colDataType == CalpontSystemCatalog::FLOAT && !isNull)
{
@ -3030,6 +3032,9 @@ const JobStepVector doConstantFilter(const ConstantFilter* cf, JobInfo& jobInfo)
if (ConstantColumn::NULLDATA == cc->type() && (opeq == *sop || opne == *sop))
cop = COMPARE_NIL;
if (utils::isWideDecimalType(ct))
pcs->addFilter(cop, value128, rf);
else
pcs->addFilter(cop, value, rf);
}
}
@ -3453,7 +3458,6 @@ JLF_ExecPlanToJobList::walkTree(execplan::ParseTree* n, JobInfo& jobInfo)
break;
case CONSTANTFILTER:
//cout << "ConstantFilter" << endl;
jsv = doConstantFilter(dynamic_cast<const ConstantFilter*>(tn), jobInfo);
JLF_ExecPlanToJobList::addJobSteps(jsv, jobInfo, false);
break;

8
dbcon/joblist/lbidlist.cpp
View File

@ -28,6 +28,7 @@
#include "brm.h"
#include "brmtypes.h"
#include "dataconvert.h"
#include "columnwidth.h"
#define IS_VERBOSE (fDebug >= 4)
#define IS_DETAIL (fDebug >= 3)
@ -808,7 +809,12 @@ bool LBIDList::CasualPartitionPredicate(const BRM::EMCasualPartition_t& cpRange,
// Should we also check for empty here?
// TODO MCOL-641
if (isNull(value, ct)) // This will work even if the data column is unsigned.
if (utils::isWideDecimalType(ct))
{
if (isNull(bigValue, ct))
continue;
}
else if (isNull(value, ct)) // This will work even if the data column is unsigned.
{
continue;
}

4
dbcon/joblist/pcolstep.cpp
View File

@ -635,13 +635,13 @@ void pColStep::addFilter(int8_t COP, int64_t value, uint8_t roundFlag)
}
// WIP MCOL-641
void pColStep::addFilter(int8_t COP, unsigned __int128 value, uint8_t roundFlag)
void pColStep::addFilter(int8_t COP, const int128_t& value, uint8_t roundFlag)
{
fFilterString << (uint8_t) COP;
fFilterString << roundFlag;
// bitwise copies into the filter ByteStream
fFilterString << value;
fFilterString << *reinterpret_cast<const uint128_t*>(&value);
fFilterCount++;
}

17
dbcon/joblist/primitivemsg.h
View File

@ -190,14 +190,15 @@ enum ISMPACKETCOMMAND
#undef PRIM_DELIVERBASE
/* Flags for BPP messages */
const uint8_t NEED_STR_VALUES = 0x01; //1;
const uint8_t GOT_ABS_RIDS = 0x02; //2;
const uint8_t GOT_VALUES = 0x04; //4;
const uint8_t LBID_TRACE = 0x08; //8;
const uint8_t HAS_JOINER = 0x10; //16;
const uint8_t SEND_RIDS_AT_DELIVERY = 0x20; //32;
const uint8_t HAS_ROWGROUP = 0x40; //64;
const uint8_t JOIN_ROWGROUP_DATA = 0x80; //128
const uint16_t NEED_STR_VALUES = 0x01; //1;
const uint16_t GOT_ABS_RIDS = 0x02; //2;
const uint16_t GOT_VALUES = 0x04; //4;
const uint16_t LBID_TRACE = 0x08; //8;
const uint16_t HAS_JOINER = 0x10; //16;
const uint16_t SEND_RIDS_AT_DELIVERY = 0x20; //32;
const uint16_t HAS_ROWGROUP = 0x40; //64;
const uint16_t JOIN_ROWGROUP_DATA = 0x80; //128
const uint16_t HAS_WIDE_DECIMAL = 0x100; //256;
//TODO: put this in a namespace to stop global ns pollution
enum PrimFlags

2
dbcon/joblist/primitivestep.h
View File

@ -196,7 +196,7 @@ public:
void addFilter(int8_t COP, int64_t value, uint8_t roundFlag = 0);
void addFilter(int8_t COP, float value);
// WIP MCOL-641
void addFilter(int8_t COP, unsigned __int128 value, uint8_t roundFlag = 0);
void addFilter(int8_t COP, const int128_t& value, uint8_t roundFlag = 0);
/** @brief Sets the DataList to get RID values from.
*

30
primitives/primproc/batchprimitiveprocessor.cpp
View File

@ -100,6 +100,7 @@ BatchPrimitiveProcessor::BatchPrimitiveProcessor() :
baseRid(0),
ridCount(0),
needStrValues(false),
hasWideDecimalType(false),
filterCount(0),
projectCount(0),
sendRidsAtDelivery(false),
@ -145,6 +146,7 @@ BatchPrimitiveProcessor::BatchPrimitiveProcessor(ByteStream& b, double prefetch,
baseRid(0),
ridCount(0),
needStrValues(false),
hasWideDecimalType(false),
filterCount(0),
projectCount(0),
sendRidsAtDelivery(false),
@ -218,6 +220,7 @@ void BatchPrimitiveProcessor::initBPP(ByteStream& bs)
{
uint32_t i;
uint8_t tmp8;
uint16_t tmp16;
Command::CommandType type;
bs.advance(sizeof(ISMPacketHeader)); // skip the header
@ -229,15 +232,16 @@ void BatchPrimitiveProcessor::initBPP(ByteStream& bs)
bs >> uniqueID;
bs >> versionInfo;
bs >> tmp8;
needStrValues = tmp8 & NEED_STR_VALUES;
gotAbsRids = tmp8 & GOT_ABS_RIDS;
gotValues = tmp8 & GOT_VALUES;
LBIDTrace = tmp8 & LBID_TRACE;
sendRidsAtDelivery = tmp8 & SEND_RIDS_AT_DELIVERY;
doJoin = tmp8 & HAS_JOINER;
hasRowGroup = tmp8 & HAS_ROWGROUP;
getTupleJoinRowGroupData = tmp8 & JOIN_ROWGROUP_DATA;
bs >> tmp16;
needStrValues = tmp16 & NEED_STR_VALUES;
gotAbsRids = tmp16 & GOT_ABS_RIDS;
gotValues = tmp16 & GOT_VALUES;
LBIDTrace = tmp16 & LBID_TRACE;
sendRidsAtDelivery = tmp16 & SEND_RIDS_AT_DELIVERY;
doJoin = tmp16 & HAS_JOINER;
hasRowGroup = tmp16 & HAS_ROWGROUP;
getTupleJoinRowGroupData = tmp16 & JOIN_ROWGROUP_DATA;
hasWideDecimalType = tmp16 & HAS_WIDE_DECIMAL;
// This used to signify that there was input row data from previous jobsteps, and
// it never quite worked right. No need to fix it or update it; all BPP's have started
@ -1019,6 +1023,8 @@ void BatchPrimitiveProcessor::initProcessor()
fFiltRidCount[i] = 0;
fFiltCmdRids[i].reset(new uint16_t[LOGICAL_BLOCK_RIDS]);
fFiltCmdValues[i].reset(new int64_t[LOGICAL_BLOCK_RIDS]);
if (hasWideDecimalType)
fFiltCmdBinaryValues[i].reset(new int128_t[LOGICAL_BLOCK_RIDS]);
if (filtOnString) fFiltStrValues[i].reset(new string[LOGICAL_BLOCK_RIDS]);
}
@ -1539,6 +1545,11 @@ void BatchPrimitiveProcessor::execute()
projectSteps[j]->projectIntoRowGroup(fe1Input, projectForFE1[j]);
for (j = 0; j < ridCount; j++, fe1In.nextRow())
// TODO MCOL-641
// WHERE clause on a numeric and a non-numeric column
// leads to this execution path:
// SELECT a, b from t1 where a!=b
// Here, a is e.g., decimal(38), b is varchar(15)
if (fe1->evaluate(&fe1In))
{
applyMapping(fe1ToProjection, fe1In, &fe1Out);
@ -2339,6 +2350,7 @@ SBPP BatchPrimitiveProcessor::duplicate()
bpp->stepID = stepID;
bpp->uniqueID = uniqueID;
bpp->needStrValues = needStrValues;
bpp->hasWideDecimalType = hasWideDecimalType;
bpp->gotAbsRids = gotAbsRids;
bpp->gotValues = gotValues;
bpp->LBIDTrace = LBIDTrace;

3
primitives/primproc/batchprimitiveprocessor.h
View File

@ -205,10 +205,12 @@ private:
uint16_t relRids[LOGICAL_BLOCK_RIDS];
int64_t values[LOGICAL_BLOCK_RIDS];
int128_t binaryValues[LOGICAL_BLOCK_RIDS];
boost::scoped_array<uint64_t> absRids;
boost::scoped_array<std::string> strValues;
uint16_t ridCount;
bool needStrValues;
bool hasWideDecimalType;
/* Common space for primitive data */
static const uint32_t BUFFER_SIZE = 131072;
@ -274,6 +276,7 @@ private:
bool filtOnString;
boost::scoped_array<uint16_t> fFiltCmdRids[2];
boost::scoped_array<int64_t> fFiltCmdValues[2];
boost::scoped_array<int128_t> fFiltCmdBinaryValues[2];
boost::scoped_array<std::string> fFiltStrValues[2];
uint64_t fFiltRidCount[2];

36
primitives/primproc/columncommand.cpp
View File

@ -71,7 +71,6 @@ ColumnCommand::~ColumnCommand() { }
void ColumnCommand::_execute()
{
// cout << "CC: executing" << endl;
if (_isScan)
makeScanMsg();
else if (bpp->ridCount == 0) // this would cause a scan
@ -93,11 +92,20 @@ void ColumnCommand::_execute()
void ColumnCommand::execute()
{
if (fFilterFeeder == LEFT_FEEDER)
{
values = bpp->fFiltCmdValues[0].get();
binaryValues = bpp->fFiltCmdBinaryValues[0].get();
}
else if (fFilterFeeder == RIGHT_FEEDER)
{
values = bpp->fFiltCmdValues[1].get();
binaryValues = bpp->fFiltCmdBinaryValues[1].get();
}
else
{
values = bpp->values;
binaryValues = bpp->binaryValues;
}
_execute();
}
@ -258,7 +266,6 @@ void ColumnCommand::issuePrimitive()
loadData();
// cout << "issuing primitive for LBID " << primMsg->LBID << endl;
if (!suppressFilter)
bpp->pp.setParsedColumnFilter(parsedColumnFilter);
else
@ -295,7 +302,6 @@ void ColumnCommand::process_OT_BOTH()
bpp->ridCount = outMsg->NVALS;
bpp->ridMap = outMsg->RidFlags;
// cout << "rid Count is " << bpp->ridCount << endl;
/* this is verbose and repetative to minimize the work per row */
switch (colType.colWidth)
@ -308,24 +314,12 @@ void ColumnCommand::process_OT_BOTH()
bpp->relRids[i] = *((uint16_t*) &bpp->outputMsg[pos]);
pos += 2;
// WIP
// values[i] is 8 Bytes wide so coping the pointer to bpp->outputMsg[pos] and crossing fingers
// I dont know the liveness of bpp->outputMsg but also I dont know if there is other memory area I can use
values[i] = (int64_t) &bpp->outputMsg[pos];
// cout<< "CC: BIN16 " << i << " "
// << hex
// << *((int64_t*)values[i])
// << " "
// << *(((int64_t*)values[i]) +1)
// << endl;
binaryValues[i] = *((int128_t*) &bpp->outputMsg[pos]);
pos += 16;
}
break;
case 8:
for (i = 0, pos = sizeof(NewColResultHeader); i < outMsg->NVALS; ++i)
{
@ -389,28 +383,24 @@ void ColumnCommand::process_OT_RID()
memcpy(bpp->relRids, outMsg + 1, outMsg->NVALS << 1);
bpp->ridCount = outMsg->NVALS;
bpp->ridMap = outMsg->RidFlags;
// cout << "rid Count is " << bpp->ridCount << endl;
}
void ColumnCommand::process_OT_DATAVALUE()
{
bpp->ridCount = outMsg->NVALS;
// cout << "rid Count is " << bpp->ridCount << endl;
switch (colType.colWidth)
{
case 16:
{
memcpy(values, outMsg + 1, outMsg->NVALS << 3);
memcpy(binaryValues, outMsg + 1, outMsg->NVALS << 4);
cout << " CC: first value is " << values[0] << endl;
break;
}
case 8:
{
memcpy(values, outMsg + 1, outMsg->NVALS << 3);
// cout << " CC: first value is " << values[0] << endl;
break;
}
@ -488,8 +478,6 @@ void ColumnCommand::processResult()
for (uint64_t i = 0; i < bpp->ridCount; i++)
bpp->fFiltCmdRids[1][i] = bpp->relRids[i];
}
// cout << "processed " << outMsg->NVALS << " rows" << endl;
}
void ColumnCommand::createCommand(ByteStream& bs)
@ -823,7 +811,7 @@ void ColumnCommand::projectResultRG(RowGroup& rg, uint32_t pos)
cout << __FILE__<< ":" <<__LINE__ << " ColumnCommand::projectResultRG " << endl;
for (i = 0; i < outMsg->NVALS; ++i, msg8 += gapSize)
{
r.setBinaryField_offset(msg8, colType.colWidth, offset);
r.setBinaryField_offset((int128_t*)msg8, colType.colWidth, offset);
r.nextRow(rowSize);
}
break;

1
primitives/primproc/columncommand.h
View File

@ -147,6 +147,7 @@ private:
uint16_t filterCount;
bool makeAbsRids;
int64_t* values; // this is usually bpp->values; RTSCommand needs to use a different container
int128_t* binaryValues;
uint8_t mask, shift; // vars for the selective block loader

84
primitives/primproc/filtercommand.cpp
View File

@ -174,7 +174,8 @@ Command* FilterCommand::makeFilterCommand(ByteStream& bs, vector<SCommand>& cmds
}
FilterCommand::FilterCommand() : Command(FILTER_COMMAND), fBOP(0)
FilterCommand::FilterCommand() : Command(FILTER_COMMAND), fBOP(0),
hasWideDecimalType(false)
{
}
@ -247,6 +248,9 @@ void FilterCommand::setColTypes(const execplan::CalpontSystemCatalog::ColType& l
{
leftColType = left;
rightColType = right;
if (utils::isWideDecimalType(left) || utils::isWideDecimalType(right))
hasWideDecimalType = true;
}
@ -255,6 +259,13 @@ void FilterCommand::doFilter()
bpp->ridMap = 0;
bpp->ridCount = 0;
bool (FilterCommand::*compareFunc)(uint64_t, uint64_t);
if (hasWideDecimalType)
compareFunc = &FilterCommand::binaryCompare;
else
compareFunc = &FilterCommand::compare;
// rids in [0] is used for scan [1], so [1] is a subset of [0], and same order.
// -- see makeFilterCommand() above.
for (uint64_t i = 0, j = 0; j < bpp->fFiltRidCount[1]; )
@ -265,9 +276,14 @@ void FilterCommand::doFilter()
}
else
{
if (compare(i, j) == true)
if ((this->*compareFunc)(i, j) == true)
{
bpp->relRids[bpp->ridCount] = bpp->fFiltCmdRids[0][i];
// WIP MCOL-641 How is bpp->(binary)values used given that
// we are setting the relRids?
if (utils::isWideDecimalType(leftColType))
bpp->binaryValues[bpp->ridCount] = bpp->fFiltCmdBinaryValues[0][i];
else
bpp->values[bpp->ridCount] = bpp->fFiltCmdValues[0][i];
bpp->ridMap |= 1 << (bpp->relRids[bpp->ridCount] >> 10);
bpp->ridCount++;
@ -321,6 +337,70 @@ bool FilterCommand::compare(uint64_t i, uint64_t j)
}
}
bool FilterCommand::binaryCompare(uint64_t i, uint64_t j)
{
// We type-promote to int128_t if either of the columns are
// not int128_t
int128_t leftVal, rightVal;
if (utils::isWideDecimalType(leftColType))
{
if (execplan::isNull(bpp->fFiltCmdBinaryValues[0][i], leftColType))
return false;
leftVal = bpp->fFiltCmdBinaryValues[0][i];
}
else
{
if (execplan::isNull(bpp->fFiltCmdValues[0][i], leftColType))
return false;
leftVal = bpp->fFiltCmdValues[0][i];
}
if (utils::isWideDecimalType(rightColType))
{
if (execplan::isNull(bpp->fFiltCmdBinaryValues[1][j], rightColType))
return false;
rightVal = bpp->fFiltCmdBinaryValues[1][j];
}
else
{
if (execplan::isNull(bpp->fFiltCmdValues[1][j], rightColType))
return false;
rightVal = bpp->fFiltCmdValues[1][j];
}
switch (fBOP)
{
case COMPARE_GT:
return leftVal > rightVal;
break;
case COMPARE_LT:
return leftVal < rightVal;
break;
case COMPARE_EQ:
return leftVal == rightVal;
break;
case COMPARE_GE:
return leftVal >= rightVal;
break;
case COMPARE_LE:
return leftVal <= rightVal;
break;
case COMPARE_NE:
return leftVal != rightVal;
break;
default:
return false;
break;
}
}
bool FilterCommand::operator==(const FilterCommand& c) const
{

5
primitives/primproc/filtercommand.h
View File

@ -76,9 +76,14 @@ protected:
// compare method, take the indices to the values array
virtual bool compare(uint64_t, uint64_t);
// compare method, take the indices to the values array
virtual bool binaryCompare(uint64_t, uint64_t);
// binary operator
uint8_t fBOP;
bool hasWideDecimalType;
// column type for null check
execplan::CalpontSystemCatalog::ColType leftColType;
execplan::CalpontSystemCatalog::ColType rightColType;

15
primitives/primproc/passthrucommand.cpp
View File

@ -79,6 +79,8 @@ void PassThruCommand::project()
{
case 16:
cout << __FILE__<< ":" <<__LINE__ << " Fix for 16 Bytes ?" << endl;
bpp->serialized->append((uint8_t*) bpp->binaryValues, bpp->ridCount << 4);
break;
case 8:
bpp->serialized->append((uint8_t*) bpp->values, bpp->ridCount << 3);
@ -121,7 +123,6 @@ void PassThruCommand::projectIntoRowGroup(RowGroup& rg, uint32_t col)
case 1:
for (i = 0; i < bpp->ridCount; i++)
{
// cout << "PTC: " << bpp->values[i] << endl;
r.setUintField_offset<1>(bpp->values[i], offset);
r.nextRow(rowSize);
}
@ -131,7 +132,6 @@ void PassThruCommand::projectIntoRowGroup(RowGroup& rg, uint32_t col)
case 2:
for (i = 0; i < bpp->ridCount; i++)
{
// cout << "PTC: " << bpp->values[i] << endl;
r.setUintField_offset<2>(bpp->values[i], offset);
r.nextRow(rowSize);
}
@ -150,7 +150,6 @@ void PassThruCommand::projectIntoRowGroup(RowGroup& rg, uint32_t col)
case 8:
for (i = 0; i < bpp->ridCount; i++)
{
// cout << "PTC: " << bpp->values[i] << endl;
r.setUintField_offset<8>(bpp->values[i], offset);
r.nextRow(rowSize);
}
@ -160,15 +159,7 @@ void PassThruCommand::projectIntoRowGroup(RowGroup& rg, uint32_t col)
cout << __FILE__ << ":" << __LINE__ << " PassThruCommand::projectIntoRowGroup" << " Addition for 16 Bytes" << endl;
for (i = 0; i < bpp->ridCount; i++)
{
cout << "PTC: " << "BIN16 " << i << " "
<< hex
<< *((int64_t*) bpp->values[i])
<< " "
<< *(((int64_t*) bpp->values[i]) +1)
<< endl;
// values[i] is 8 bytes so it contains the pointer to bpp->outputMsg set by ColumnCommand::process_OT_BOTH()
r.setBinaryField_offset((uint128_t*)bpp->values[i], 16, offset);
r.setBinaryField_offset(&bpp->binaryValues[i], 16, offset);
r.nextRow(rowSize);
}
}

4
primitives/primproc/primproc.cpp
View File

@ -513,7 +513,9 @@ int ServicePrimProc::Child()
// do not allow to read beyond the end of an extent
const int MaxReadAheadSz = (extentRows) / BLOCK_SIZE;
//defaultBufferSize = 512 * 1024; // @bug 2627 - changed default dict buffer from 256K to 512K, allows for cols w/ length of 61.
defaultBufferSize = 100 * 1024; // 1/17/12 - made the dict buffer dynamic, max size for a numeric col is 80k + ovrhd
// WIP MCOL-641 Check with Patrick on this. Changed it from 100*1024 to 128*1024
// to match with BatchPrimitiveProcessor::BUFFER_SIZE
defaultBufferSize = 128 * 1024; // 1/17/12 - made the dict buffer dynamic, max size for a numeric col is 80k + ovrhd
// This parm controls whether we rotate through the output sockets

8
utils/common/columnwidth.h
View File

@ -18,6 +18,7 @@
#ifndef UTILS_COLWIDTH_H
#define UTILS_COLWIDTH_H
#include "calpontsystemcatalog.h"
#include "branchpred.h"
namespace utils
@ -35,6 +36,13 @@ namespace utils
return width <= MAXLEGACYWIDTH;
}
inline bool isWideDecimalType(const execplan::CalpontSystemCatalog::ColType& ct)
{
return ((ct.colDataType == execplan::CalpontSystemCatalog::DECIMAL ||
ct.colDataType == execplan::CalpontSystemCatalog::UDECIMAL) &&
ct.colWidth == MAXCOLUMNWIDTH);
}
/** @brief Map a DECIMAL precision to data width in bytes */
inline uint8_t widthByPrecision(unsigned p)
{

2
utils/joiner/tuplejoiner.cpp
View File

@ -1147,7 +1147,7 @@ void TupleJoiner::updateCPData(const Row& r)
}
}
}
else if (utils::isWide(r.getColumnWidth(colIdx))
else if (r.getColumnWidth(colIdx) == utils::MAXCOLUMNWIDTH
&& (r.getColType(colIdx) == CalpontSystemCatalog::DECIMAL
|| r.getColType(colIdx) == CalpontSystemCatalog::UDECIMAL))
{

2
utils/messageqcpp/bytestream.cpp
View File

@ -236,7 +236,7 @@ ByteStream& ByteStream::operator<<(const uint64_t o)
}
// WIP MCOL-641
ByteStream& ByteStream::operator<<(const uint128_t o)
ByteStream& ByteStream::operator<<(const uint128_t& o)
{
if (fBuf == 0 || (fCurInPtr - fBuf + 16U > fMaxLen + ISSOverhead))
growBuf(fMaxLen + BlockSize);

8
utils/messageqcpp/bytestream.h
View File

@ -148,9 +148,9 @@ public:
EXPORT ByteStream& operator<<(const uint64_t o);
// WIP MCOL-641
/**
* push an unsigned __int128 onto the end of the stream. The byte order is whatever the native byte order is.
* push an uint128_t onto the end of the stream. The byte order is whatever the native byte order is.
*/
EXPORT ByteStream& operator<<(const uint128_t o);
EXPORT ByteStream& operator<<(const uint128_t& o);
/**
* push a float onto the end of the stream. The byte order is
* whatever the native byte order is.
@ -217,7 +217,7 @@ public:
EXPORT ByteStream& operator>>(uint64_t& o);
// WIP MCOL-641
/**
* extract an unsigned __int128 from the front of the stream. The byte order is whatever the native byte order is.
* extract an uint128_t from the front of the stream. The byte order is whatever the native byte order is.
*/
EXPORT ByteStream& operator>>(uint128_t& o);
/**
@ -292,7 +292,7 @@ public:
EXPORT void peek(uint64_t& o) const;
// WIP MCOL-641
/**
* Peek at an unsigned __int128 from the front of the stream. The byte order is whatever the native byte order is.
* Peek at an uint128_t from the front of the stream. The byte order is whatever the native byte order is.
*/
EXPORT void peek(uint128_t& o) const;
/**

2
utils/rowgroup/rowgroup.cpp
View File

@ -639,7 +639,7 @@ string Row::toString() const
break;
case CalpontSystemCatalog::DECIMAL:
case CalpontSystemCatalog::UDECIMAL:
if (utils::isWide(colWidths[i]))
if (colWidths[i] == utils::MAXCOLUMNWIDTH)
{
unsigned int buflen = precision[i] + 3;
char *buf = (char*)alloca(buflen);

7
writeengine/wrapper/writeengine.cpp
View File

@ -70,6 +70,9 @@ namespace WriteEngine
{
StopWatch timer;
using dataconvert::int128_t;
using dataconvert::uint128_t;
/**@brief WriteEngineWrapper Constructor
*/
WriteEngineWrapper::WriteEngineWrapper() : m_opType(NOOP)
@ -219,10 +222,6 @@ void WriteEngineWrapper::findSmallestColumn(uint32_t& colId, ColStructList colSt
}
}
// MCOL-641 WIP
using int128_t = __int128;
using uint128_t = unsigned __int128;
/*@convertValArray - Convert interface values to internal values
*/
/***********************************************************