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-769 Much higher performance bulk insert

Browse Source 
Use void* pointers instead of boost::any with lots of copies
This commit is contained in:
Andrew Hutchings
2017-06-20 19:08:48 +01:00
parent 44977a8c6b
commit 4cca5fa7fa
8 changed files with 2330 additions and 117 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
mariadb-columnstore-engine.config Normal file
View File

@ -0,0 +1,2 @@
// Add predefined macros for your project here. For example:
// #define THE_ANSWER 42

1
mariadb-columnstore-engine.creator Normal file
View File

@ -0,0 +1 @@
[General]

1333
mariadb-columnstore-engine.files Normal file
View File

File diff suppressed because it is too large Load Diff

81
mariadb-columnstore-engine.includes Normal file
View File

@ -0,0 +1,81 @@
dbcon/ddlpackage
dbcon/ddlpackageproc
dbcon/dmlpackage
dbcon/dmlpackageproc
dbcon/doc
dbcon/execplan
dbcon/joblist
dbcon/mysql
ddlproc
decomsvr
dmlproc
exemgr
oam/oamcpp
oam/replaytxnlog
oamapps/alarmmanager
oamapps/calpontConsole
oamapps/hardwareMonitor
oamapps/mcsadmin
oamapps/postConfigure
oamapps/serverMonitor
primitives/blockcache
primitives/linux-port
primitives/primproc
procmgr
procmon
tools/cleartablelock
tools/dbbuilder
tools/dbloadxml
tools/dmldriver
tools/qfe
tools/vbgen
utils/batchloader
utils/cacheutils
utils/common
utils/compress
utils/configcpp
utils/configcpp/md5
utils/dataconvert
utils/ddlcleanup
utils/funcexp
utils/funcexp/utf8
utils/idbdatafile
utils/idbhdfs/hdfs-12
utils/idbhdfs/hdfs-20
utils/idbhdfs/hdfs-shared
utils/joiner
utils/loggingcpp
utils/messageqcpp
utils/multicast
utils/querystats
utils/querytele
utils/rowgroup
utils/rwlock
utils/startup
utils/testbc
utils/threadpool
utils/thrift/thrift
utils/thrift/thrift/async
utils/thrift/thrift/concurrency
utils/thrift/thrift/processor
utils/thrift/thrift/protocol
utils/thrift/thrift/qt
utils/thrift/thrift/server
utils/thrift/thrift/transport
utils/thrift/thrift/windows
utils/udfsdk
utils/windowfunction
utils/winport
utils/winport/sys
versioning/BRM
writeengine
writeengine/bulk
writeengine/client
writeengine/dictionary
writeengine/index
writeengine/redistribute
writeengine/server
writeengine/shared
writeengine/splitter
writeengine/wrapper
writeengine/xml

119
writeengine/server/we_dmlcommandproc.cpp
View File

@ -1292,7 +1292,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
WriteEngine::ColStructList colStructs;
WriteEngine::DctnryStructList dctnryStructList;
WriteEngine::DctnryValueList dctnryValueList;
WriteEngine::ColValueList colValuesList;
std::vector<uint64_t> colValuesList;
WriteEngine::DictStrList dicStringList ;
CalpontSystemCatalog::TableName tableName;
CalpontSystemCatalog::TableColName tableColName;
@ -1520,12 +1520,12 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
std::string tmpStr("");
uint32_t valuesPerColumn;
bs >> valuesPerColumn;
colValuesList.reserve(columnCount * valuesPerColumn);
try
{
bool pushWarning = false;
for (uint32_t j = 0; j < columnCount; j++)
{
WriteEngine::ColTupleList colTuples;
WriteEngine::DctColTupleList dctColTuples;
tableColName.column = colNames[j];
CalpontSystemCatalog::OID oid = colStructs[j].dataOid;
@ -1533,7 +1533,6 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
CalpontSystemCatalog::ColType colType;
colType = systemCatalogPtr->colType(oid);
boost::any datavalue;
bool isNULL = false;
WriteEngine::dictStr dicStrings;
// token
@ -1569,14 +1568,10 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
if ( !pushWarning )
pushWarning = true;
}
WriteEngine::ColTuple colTuple;
colTuple.data = datavalue;
colTuples.push_back(colTuple);
colValuesList.push_back(0);
//@Bug 2515. Only pass string values to write engine
dicStrings.push_back( tmpStr );
}
colValuesList.push_back(colTuples);
//@Bug 2515. Only pass string values to write engine
dicStringList.push_back( dicStrings );
}
@ -1606,10 +1601,11 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
bs >> tmp8;
isNULL = tmp8;
int8_t val8;
int16_t val16;
int32_t val32;
int64_t val64;
uint8_t val8;
uint16_t val16;
uint32_t val32;
uint64_t val64;
uint64_t colValue;
float valF;
double valD;
std::string valStr;
@ -1617,56 +1613,36 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
switch (colType.colDataType)
{
case execplan::CalpontSystemCatalog::TINYINT:
bs >> val8;
if (val8 == 0)
valZero = true;
datavalue = (char)val8;
break;
case execplan::CalpontSystemCatalog::SMALLINT:
bs >> val16;
if (val16 == 0)
valZero = true;
datavalue = (short)val16;
break;
case execplan::CalpontSystemCatalog::MEDINT:
case execplan::CalpontSystemCatalog::INT:
bs >> val32;
if (val32 == 0)
valZero = true;
datavalue = (int)val32;
break;
case execplan::CalpontSystemCatalog::BIGINT:
bs >> val64;
if (val64 == 0)
valZero = true;
datavalue = (long long)val64;
break;
case execplan::CalpontSystemCatalog::UTINYINT:
bs >> val8;
if (val8 == 0)
valZero = true;
datavalue = (uint8_t)val8;
colValue = val8;
break;
case execplan::CalpontSystemCatalog::SMALLINT:
case execplan::CalpontSystemCatalog::DATE:
case execplan::CalpontSystemCatalog::USMALLINT:
bs >> val16;
if (val16 == 0)
valZero = true;
datavalue = (uint16_t)val16;
colValue = val16;
break;
case execplan::CalpontSystemCatalog::MEDINT:
case execplan::CalpontSystemCatalog::INT:
case execplan::CalpontSystemCatalog::UMEDINT:
case execplan::CalpontSystemCatalog::UINT:
bs >> val32;
if (val32 == 0)
valZero = true;
datavalue = (uint32_t)val32;
colValue = val32;
break;
case execplan::CalpontSystemCatalog::BIGINT:
case execplan::CalpontSystemCatalog::DATETIME:
case execplan::CalpontSystemCatalog::UBIGINT:
bs >> val64;
if (val64 == 0)
valZero = true;
datavalue = (uint64_t)val64;
colValue = val64;
break;
case execplan::CalpontSystemCatalog::DECIMAL:
switch (colType.colWidth)
@ -1674,25 +1650,25 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
case 1:
{
bs >> val8;
datavalue = (char) val8;
colValue = val8;
break;
}
case 2:
{
bs >> val16;
datavalue = (short) val16;
colValue = val16;
break;
}
case 4:
{
bs >> val32;
datavalue = (int) val32;
colValue = val32;
break;
}
default:
{
bs >> val64;
datavalue = (long long) val64;
colValue = val64;
break;
}
}
@ -1710,7 +1686,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
val8 = 0;
pushWarning = true;
}
datavalue = (char)val8;
colValue = val8;
}
else if (colType.colWidth == 2)
{
@ -1722,7 +1698,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
val16 = 0;
pushWarning = true;
}
datavalue = (short)val16;
colValue = val16;
}
else if (colType.colWidth == 4)
{
@ -1734,7 +1710,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
val32 = 0;
pushWarning = true;
}
datavalue = (int)val32;
colValue = val32;
}
else if (colType.colWidth == 8)
{
@ -1746,13 +1722,12 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
val64 = 0;
pushWarning = true;
}
datavalue = (long long)val64;
colValue = val64;
}
break;
case execplan::CalpontSystemCatalog::DOUBLE:
bs >> val64;
memcpy(&valD, &val64, 8);
datavalue = valD;
colValue = val64;
break;
case execplan::CalpontSystemCatalog::UDOUBLE:
bs >> val64;
@ -1763,12 +1738,11 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
pushWarning = true;
}
datavalue = valD;
colValue = val64;
break;
case execplan::CalpontSystemCatalog::FLOAT:
bs >> val32;
memcpy(&valF, &val32, 4);
datavalue = valF;
colValue = val32;
break;
case execplan::CalpontSystemCatalog::UFLOAT:
bs >> val32;
@ -1779,7 +1753,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
pushWarning = true;
}
datavalue = valF;
colValue = val32;
break;
case execplan::CalpontSystemCatalog::CHAR:
@ -1800,7 +1774,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
valStr.resize(colType.colWidth, 0);
}
}
datavalue = valStr;
memcpy(&colValue, valStr.c_str(), valStr.length());
break;
default:
rc = 1;
@ -1830,28 +1804,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
rc = 1;
return rc;
}
switch (colType.colDataType)
{
case execplan::CalpontSystemCatalog::TINYINT:
case execplan::CalpontSystemCatalog::UTINYINT:
datavalue = (uint8_t) nextVal;
break;
case execplan::CalpontSystemCatalog::SMALLINT:
case execplan::CalpontSystemCatalog::USMALLINT:
datavalue = (uint16_t) nextVal;
break;
case execplan::CalpontSystemCatalog::MEDINT:
case execplan::CalpontSystemCatalog::UMEDINT:
case execplan::CalpontSystemCatalog::INT:
case execplan::CalpontSystemCatalog::UINT:
datavalue = (uint32_t) nextVal;
break;
case execplan::CalpontSystemCatalog::BIGINT:
case execplan::CalpontSystemCatalog::UBIGINT:
default:
datavalue = (uint64_t) nextVal;
break;
}
colValue = nextVal;
}
if (colType.constraintType == CalpontSystemCatalog::NOTNULL_CONSTRAINT)
@ -1866,7 +1819,7 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
}
else if (isNULL && !(colType.defaultValue.empty()))
{
datavalue = colType.defaultValue;
memcpy(&colValue, colType.defaultValue.c_str(), colType.defaultValue.length());
isNULL = false;
}
}
@ -1878,14 +1831,11 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
if ( pushWarning && ( rc != dmlpackageprocessor::DMLPackageProcessor::IDBRANGE_WARNING ) )
rc = dmlpackageprocessor::DMLPackageProcessor::IDBRANGE_WARNING;
WriteEngine::ColTuple colTuple;
colTuple.data = datavalue;
colTuples.push_back(colTuple);
colValuesList.push_back(colValue);
//@Bug 2515. Only pass string values to write engine
dicStrings.push_back( valStr );
}
colValuesList.push_back(colTuples);
dicStringList.push_back( dicStrings );
}
@ -1909,11 +1859,9 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
//fWriteEngine.setDebugLevel(WriteEngine::DEBUG_3);
//cout << "Batch inserting a row with transaction id " << txnid.id << endl;
if (colValuesList.size() > 0)
{
if (colValuesList[0].size() > 0)
{
if (NO_ERROR !=
(error = fWEWrapper.insertColumnRecs(txnid.id, colStructs, colValuesList, dctnryStructList, dicStringList,
(error = fWEWrapper.insertColumnRecsBinary(txnid.id, colStructs, colValuesList, dctnryStructList, dicStringList,
dbRootExtTrackerVec, 0, bFirstExtentOnThisPM, isInsertSelect, isAutocommitOn, roPair.objnum, fIsFirstBatchPm)))
{
if (error == ERR_BRM_DEAD_LOCK)
@ -1935,7 +1883,6 @@ uint8_t WE_DMLCommandProc::processBatchInsertBinary(messageqcpp::ByteStream& bs,
}
}
}
}
if (fIsFirstBatchPm && isAutocommitOn)
{
//fWEWrapper.writeVBEnd(txnid.id, rangeList);

4
writeengine/wrapper/we_colop.h
View File

@ -313,13 +313,13 @@ public:
*/
int copyVB(IDBDataFile* pSource, const BRM::VER_t txnD, const OID oid, std::vector<uint32_t>& fboList,
std::vector<BRM::LBIDRange>& rangeList);
protected:
/**
* @brief close column file
*/
EXPORT virtual void closeColumnFile(Column& column) const;
protected:
/**
* @brief populate readBuf with data in block #lbid
*/

830
writeengine/wrapper/writeengine.cpp
View File

@ -1457,6 +1457,638 @@ timer.start("writeColumnRec");
return rc;
}
int WriteEngineWrapper::insertColumnRecsBinary(const TxnID& txnid,
ColStructList& colStructList,
std::vector<uint64_t>& colValueList,
DctnryStructList& dctnryStructList,
DictStrList& dictStrList,
std::vector<boost::shared_ptr<DBRootExtentTracker> > & dbRootExtentTrackers,
RBMetaWriter* fRBMetaWriter,
bool bFirstExtentOnThisPM,
bool insertSelect,
bool isAutoCommitOn,
OID tableOid,
bool isFirstBatchPm)
{
int rc;
RID* rowIdArray = NULL;
Column curCol;
ColStruct curColStruct;
ColStructList newColStructList;
DctnryStructList newDctnryStructList;
HWM hwm = 0;
HWM oldHwm = 0;
HWM newHwm = 0;
size_t totalRow;
ColStructList::size_type totalColumns;
uint64_t rowsLeft = 0;
bool newExtent = false;
RIDList ridList;
ColumnOp* colOp = NULL;
// Set tmp file suffix to modify HDFS db file
bool useTmpSuffix = false;
if (idbdatafile::IDBPolicy::useHdfs())
{
if (!bFirstExtentOnThisPM)
useTmpSuffix = true;
}
unsigned i=0;
#ifdef PROFILE
StopWatch timer;
#endif
//Convert data type and column width to write engine specific
for (i = 0; i < colStructList.size(); i++)
Convertor::convertColType(&colStructList[i]);
// rc = checkValid(txnid, colStructList, colValueList, ridList);
// if (rc != NO_ERROR)
// return rc;
setTransId(txnid);
uint16_t dbRoot, segmentNum;
uint32_t partitionNum;
string segFile;
bool newFile;
TableMetaData* tableMetaData= TableMetaData::makeTableMetaData(tableOid);
//populate colStructList with file information
IDBDataFile* pFile = NULL;
std::vector<DBRootExtentInfo> extentInfo;
int currentDBrootIdx = 0;
std::vector<BRM::CreateStripeColumnExtentsArgOut> extents;
//--------------------------------------------------------------------------
// For first batch on this PM:
// o get starting extent from ExtentTracker, and allocate extent if needed
// o construct colStructList and dctnryStructList accordingly
// o save extent information in tableMetaData for future use
// If not first batch on this PM:
// o construct colStructList and dctnryStructList from tableMetaData
//--------------------------------------------------------------------------
if (isFirstBatchPm)
{
currentDBrootIdx = dbRootExtentTrackers[0]->getCurrentDBRootIdx();
extentInfo = dbRootExtentTrackers[0]->getDBRootExtentList();
dbRoot = extentInfo[currentDBrootIdx].fDbRoot;
partitionNum = extentInfo[currentDBrootIdx].fPartition;
//----------------------------------------------------------------------
// check whether this extent is the first on this PM
//----------------------------------------------------------------------
if (bFirstExtentOnThisPM)
{
//cout << "bFirstExtentOnThisPM is " << bFirstExtentOnThisPM << endl;
std::vector<BRM::CreateStripeColumnExtentsArgIn> cols;
BRM::CreateStripeColumnExtentsArgIn createStripeColumnExtentsArgIn;
for (i=0; i < colStructList.size(); i++)
{
createStripeColumnExtentsArgIn.oid = colStructList[i].dataOid;
createStripeColumnExtentsArgIn.width = colStructList[i].colWidth;
createStripeColumnExtentsArgIn.colDataType = colStructList[i].colDataType;
cols.push_back(createStripeColumnExtentsArgIn);
}
rc = BRMWrapper::getInstance()->allocateStripeColExtents(cols, dbRoot, partitionNum, segmentNum, extents);
if (rc != NO_ERROR)
return rc;
//Create column files
BRM::CPInfoList_t cpinfoList;
BRM::CPInfo cpInfo;
if (isUnsigned(colStructList[i].colDataType))
{
cpInfo.max = 0;
cpInfo.min = static_cast<int64_t>(numeric_limits<uint64_t>::max());
}
else
{
cpInfo.max = numeric_limits<int64_t>::min();
cpInfo.min = numeric_limits<int64_t>::max();
}
cpInfo.seqNum = -1;
for ( i=0; i < extents.size(); i++)
{
colOp = m_colOp[op(colStructList[i].fCompressionType)];
colOp->initColumn(curCol);
colOp->setColParam(curCol, 0, colStructList[i].colWidth, colStructList[i].colDataType,
colStructList[i].colType, colStructList[i].dataOid, colStructList[i].fCompressionType,
dbRoot, partitionNum, segmentNum);
rc = colOp->extendColumn(curCol, false, extents[i].startBlkOffset, extents[i].startLbid, extents[i].allocSize, dbRoot,
partitionNum, segmentNum, segFile, pFile, newFile);
if (rc != NO_ERROR)
return rc;
//mark the extents to invalid
cpInfo.firstLbid = extents[i].startLbid;
cpinfoList.push_back(cpInfo);
colStructList[i].fColPartition = partitionNum;
colStructList[i].fColSegment = segmentNum;
colStructList[i].fColDbRoot = dbRoot;
dctnryStructList[i].fColPartition = partitionNum;
dctnryStructList[i].fColSegment = segmentNum;
dctnryStructList[i].fColDbRoot = dbRoot;
}
//mark the extents to invalid
rc = BRMWrapper::getInstance()->setExtentsMaxMin(cpinfoList);
if (rc != NO_ERROR)
return rc;
//create corresponding dictionary files
for (i=0; i < dctnryStructList.size(); i++)
{
if (dctnryStructList[i].dctnryOid > 0)
{
rc = createDctnry(txnid, dctnryStructList[i].dctnryOid, dctnryStructList[i].colWidth, dbRoot, partitionNum,
segmentNum, dctnryStructList[i].fCompressionType);
if ( rc != NO_ERROR)
return rc;
}
}
} // if ( bFirstExtentOnThisPM)
else // if (!bFirstExtentOnThisPM)
{
std::vector<DBRootExtentInfo> tmpExtentInfo;
for (i=0; i < dbRootExtentTrackers.size(); i++)
{
tmpExtentInfo = dbRootExtentTrackers[i]->getDBRootExtentList();
colStructList[i].fColPartition = tmpExtentInfo[currentDBrootIdx].fPartition;
colStructList[i].fColSegment = tmpExtentInfo[currentDBrootIdx].fSegment;
colStructList[i].fColDbRoot = tmpExtentInfo[currentDBrootIdx].fDbRoot;
//cout << "Load from dbrootExtenttracker oid:dbroot:part:seg = " <<colStructList[i].dataOid<<":"
//<<colStructList[i].fColDbRoot<<":"<<colStructList[i].fColPartition<<":"<<colStructList[i].fColSegment<<endl;
dctnryStructList[i].fColPartition = tmpExtentInfo[currentDBrootIdx].fPartition;
dctnryStructList[i].fColSegment = tmpExtentInfo[currentDBrootIdx].fSegment;
dctnryStructList[i].fColDbRoot = tmpExtentInfo[currentDBrootIdx].fDbRoot;
}
}
//----------------------------------------------------------------------
// Save the extents info in tableMetaData
//----------------------------------------------------------------------
for (i=0; i < colStructList.size(); i++)
{
ColExtsInfo aColExtsInfo = tableMetaData->getColExtsInfo(colStructList[i].dataOid);
ColExtsInfo::iterator it = aColExtsInfo.begin();
while (it != aColExtsInfo.end())
{
if ((it->dbRoot == colStructList[i].fColDbRoot) && (it->partNum == colStructList[i].fColPartition) && (it->segNum == colStructList[i].fColSegment))
break;
it++;
}
if (it == aColExtsInfo.end()) //add this one to the list
{
ColExtInfo aExt;
aExt.dbRoot = colStructList[i].fColDbRoot;
aExt.partNum = colStructList[i].fColPartition;
aExt.segNum = colStructList[i].fColSegment;
aExt.compType = colStructList[i].fCompressionType;
aExt.isDict = false;
if (bFirstExtentOnThisPM)
{
aExt.hwm = extents[i].startBlkOffset;
aExt.isNewExt = true;
//cout << "adding a ext to metadata" << endl;
}
else
{
std::vector<DBRootExtentInfo> tmpExtentInfo;
tmpExtentInfo = dbRootExtentTrackers[i]->getDBRootExtentList();
aExt.isNewExt = false;
aExt.hwm = tmpExtentInfo[currentDBrootIdx].fLocalHwm;
//cout << "oid " << colStructList[i].dataOid << " gets hwm " << aExt.hwm << endl;
}
aExt.current = true;
aColExtsInfo.push_back(aExt);
//cout << "get from extentinfo oid:hwm = " << colStructList[i].dataOid << ":" << aExt.hwm << endl;
}
tableMetaData->setColExtsInfo(colStructList[i].dataOid, aColExtsInfo);
}
for (i=0; i < dctnryStructList.size(); i++)
{
if (dctnryStructList[i].dctnryOid > 0)
{
ColExtsInfo aColExtsInfo = tableMetaData->getColExtsInfo(dctnryStructList[i].dctnryOid);
ColExtsInfo::iterator it = aColExtsInfo.begin();
while (it != aColExtsInfo.end())
{
if ((it->dbRoot == dctnryStructList[i].fColDbRoot) && (it->partNum == dctnryStructList[i].fColPartition) && (it->segNum == dctnryStructList[i].fColSegment))
break;
it++;
}
if (it == aColExtsInfo.end()) //add this one to the list
{
ColExtInfo aExt;
aExt.dbRoot = dctnryStructList[i].fColDbRoot;
aExt.partNum = dctnryStructList[i].fColPartition;
aExt.segNum = dctnryStructList[i].fColSegment;
aExt.compType = dctnryStructList[i].fCompressionType;
aExt.isDict = true;
aColExtsInfo.push_back(aExt);
}
tableMetaData->setColExtsInfo(dctnryStructList[i].dctnryOid, aColExtsInfo);
}
}
} // if (isFirstBatchPm)
else //get the extent info from tableMetaData
{
ColExtsInfo aColExtsInfo = tableMetaData->getColExtsInfo(colStructList[0].dataOid);
ColExtsInfo::iterator it = aColExtsInfo.begin();
while (it != aColExtsInfo.end())
{
if (it->current)
break;
it++;
}
if (it == aColExtsInfo.end())
return 1;
for (i=0; i < colStructList.size(); i++)
{
colStructList[i].fColPartition = it->partNum;
colStructList[i].fColSegment = it->segNum;
colStructList[i].fColDbRoot = it->dbRoot;
dctnryStructList[i].fColPartition = it->partNum;
dctnryStructList[i].fColSegment = it->segNum;
dctnryStructList[i].fColDbRoot = it->dbRoot;
}
}
totalColumns = colStructList.size();
totalRow = colValueList.size() / totalColumns;
rowIdArray = new RID[totalRow];
// use scoped_array to ensure ptr deletion regardless of where we return
boost::scoped_array<RID> rowIdArrayPtr(rowIdArray);
memset(rowIdArray, 0, (sizeof(RID)*totalRow));
//--------------------------------------------------------------------------
// allocate row id(s)
//--------------------------------------------------------------------------
curColStruct = colStructList[0];
colOp = m_colOp[op(curColStruct.fCompressionType)];
colOp->initColumn(curCol);
//Get the correct segment, partition, column file
vector<ExtentInfo> colExtentInfo; //Save those empty extents in case of failure to rollback
vector<ExtentInfo> dictExtentInfo; //Save those empty extents in case of failure to rollback
vector<ExtentInfo> fileInfo;
dbRoot = curColStruct.fColDbRoot;
//use the first column to calculate row id
ColExtsInfo aColExtsInfo = tableMetaData->getColExtsInfo(colStructList[0].dataOid);
ColExtsInfo::iterator it = aColExtsInfo.begin();
while (it != aColExtsInfo.end())
{
if ((it->dbRoot == colStructList[0].fColDbRoot) && (it->partNum == colStructList[0].fColPartition) && (it->segNum == colStructList[0].fColSegment) && it->current )
break;
it++;
}
if (it != aColExtsInfo.end())
{
hwm = it->hwm;
//cout << "Got from colextinfo hwm for oid " << colStructList[0].dataOid << " is " << hwm << " and seg is " << colStructList[0].fColSegment << endl;
}
oldHwm = hwm; //Save this info for rollback
//need to pass real dbRoot, partition, and segment to setColParam
colOp->setColParam(curCol, 0, curColStruct.colWidth, curColStruct.colDataType,
curColStruct.colType, curColStruct.dataOid, curColStruct.fCompressionType,
curColStruct.fColDbRoot, curColStruct.fColPartition, curColStruct.fColSegment);
rc = colOp->openColumnFile(curCol, segFile, useTmpSuffix); // @bug 5572 HDFS tmp file
if (rc != NO_ERROR) {
return rc;
}
//get hwm first
// @bug 286 : fix for bug 286 - correct the typo in getHWM
//RETURN_ON_ERROR(BRMWrapper::getInstance()->getHWM(curColStruct.dataOid, hwm));
Column newCol;
#ifdef PROFILE
timer.start("allocRowId");
#endif
newColStructList = colStructList;
newDctnryStructList = dctnryStructList;
bool bUseStartExtent = true;
if (idbdatafile::IDBPolicy::useHdfs())
insertSelect = true;
rc = colOp->allocRowId(txnid, bUseStartExtent,
curCol, (uint64_t)totalRow, rowIdArray, hwm, newExtent, rowsLeft, newHwm, newFile,
newColStructList, newDctnryStructList, dbRootExtentTrackers, insertSelect, true, tableOid, isFirstBatchPm);
//cout << "after allocrowid, total row = " <<totalRow << " newExtent is " << newExtent << endl;
// cout << "column oid " << curColStruct.dataOid << " has hwm:newHwm = " << hwm <<":" << newHwm<< endl;
if (rc != NO_ERROR) //Clean up is already done
return rc;
#ifdef PROFILE
timer.stop("allocRowId");
#endif
//--------------------------------------------------------------------------
// Expand initial abbreviated extent if any RID in 1st extent is > 256K.
// if totalRow == rowsLeft, then not adding rows to 1st extent, so skip it.
//--------------------------------------------------------------------------
// DMC-SHARED_NOTHING_NOTE: Is it safe to assume only part0 seg0 is abbreviated?
if ((curCol.dataFile.fPartition == 0) &&
(curCol.dataFile.fSegment == 0) &&
((totalRow-rowsLeft) > 0) &&
(rowIdArray[totalRow-rowsLeft-1] >= (RID)INITIAL_EXTENT_ROWS_TO_DISK))
{
for (unsigned k=1; k<colStructList.size(); k++)
{
Column expandCol;
colOp = m_colOp[op(colStructList[k].fCompressionType)];
colOp->setColParam(expandCol, 0,
colStructList[k].colWidth,
colStructList[k].colDataType,
colStructList[k].colType,
colStructList[k].dataOid,
colStructList[k].fCompressionType,
colStructList[k].fColDbRoot,
colStructList[k].fColPartition,
colStructList[k].fColSegment);
rc = colOp->openColumnFile(expandCol, segFile, true); // @bug 5572 HDFS tmp file
if (rc == NO_ERROR)
{
if (colOp->abbreviatedExtent(expandCol.dataFile.pFile, colStructList[k].colWidth))
{
rc = colOp->expandAbbrevExtent(expandCol);
}
}
if (rc != NO_ERROR)
{
return rc;
}
colOp->closeColumnFile(expandCol);
}
}
//--------------------------------------------------------------------------
// Tokenize data if needed
//--------------------------------------------------------------------------
if (insertSelect && isAutoCommitOn)
BRMWrapper::setUseVb( false );
else
BRMWrapper::setUseVb( true );
dictStr::iterator dctStr_iter;
uint64_t *colValPtr;
size_t rowsPerColumn = colValueList.size() / colStructList.size();
for (i = 0; i < colStructList.size(); i++)
{
if (colStructList[i].tokenFlag)
{
dctStr_iter = dictStrList[i].begin();
Dctnry* dctnry = m_dctnry[op(dctnryStructList[i].fCompressionType)];
rc = dctnry->openDctnry(dctnryStructList[i].dctnryOid,
dctnryStructList[i].fColDbRoot, dctnryStructList[i].fColPartition,
dctnryStructList[i].fColSegment,
useTmpSuffix); // @bug 5572 HDFS tmp file
if (rc !=NO_ERROR)
{
cout << "Error opening dctnry file " << dctnryStructList[i].dctnryOid<< endl;
return rc;
}
for (uint32_t rows = 0; rows < (totalRow - rowsLeft); rows++)
{
colValPtr = &colValueList[(i*rowsPerColumn) + rows];
if (dctStr_iter->length() == 0)
{
Token nullToken;
memcpy(colValPtr, &nullToken, 8);
}
else
{
#ifdef PROFILE
timer.start("tokenize");
#endif
DctnryTuple dctTuple;
dctTuple.sigValue = (unsigned char*)dctStr_iter->c_str();
dctTuple.sigSize = dctStr_iter->length();
dctTuple.isNull = false;
rc = tokenize(txnid, dctTuple, dctnryStructList[i].fCompressionType);
if (rc != NO_ERROR)
{
dctnry->closeDctnry();
return rc;
}
#ifdef PROFILE
timer.stop("tokenize");
#endif
memcpy(colValPtr, &dctTuple.token, 8);
}
dctStr_iter++;
}
//close dictionary files
rc = dctnry->closeDctnry(false);
if (rc != NO_ERROR)
return rc;
if (newExtent)
{
//@Bug 4854 back up hwm chunk for the file to be modified
if (fRBMetaWriter)
fRBMetaWriter->backupDctnryHWMChunk(newDctnryStructList[i].dctnryOid, newDctnryStructList[i].fColDbRoot, newDctnryStructList[i].fColPartition, newDctnryStructList[i].fColSegment);
rc = dctnry->openDctnry(newDctnryStructList[i].dctnryOid,
newDctnryStructList[i].fColDbRoot, newDctnryStructList[i].fColPartition,
newDctnryStructList[i].fColSegment,
false); // @bug 5572 HDFS tmp file
if (rc !=NO_ERROR)
return rc;
for (uint32_t rows = 0; rows < rowsLeft; rows++)
{
colValPtr = &colValueList[(i*rowsPerColumn) + rows];
if (dctStr_iter->length() == 0)
{
Token nullToken;
memcpy(colValPtr, &nullToken, 8);
}
else
{
#ifdef PROFILE
timer.start("tokenize");
#endif
DctnryTuple dctTuple;
dctTuple.sigValue = (unsigned char*)dctStr_iter->c_str();
dctTuple.sigSize = dctStr_iter->length();
dctTuple.isNull = false;
rc = tokenize(txnid, dctTuple, newDctnryStructList[i].fCompressionType);
if (rc != NO_ERROR)
{
dctnry->closeDctnry();
return rc;
}
#ifdef PROFILE
timer.stop("tokenize");
#endif
memcpy(colValPtr, &dctTuple.token, 8);
}
dctStr_iter++;
}
//close dictionary files
rc = dctnry->closeDctnry(false);
if (rc != NO_ERROR)
return rc;
}
}
}
if (insertSelect && isAutoCommitOn)
BRMWrapper::setUseVb( false );
else
BRMWrapper::setUseVb( true );
//--------------------------------------------------------------------------
// Update column info structure @Bug 1862 set hwm, and
// Prepare ValueList for new extent (if applicable)
//--------------------------------------------------------------------------
//@Bug 2205 Check whether all rows go to the new extent
RID lastRid = 0;
RID lastRidNew = 0;
if (totalRow-rowsLeft > 0)
{
lastRid = rowIdArray[totalRow-rowsLeft-1];
lastRidNew = rowIdArray[totalRow-1];
}
else
{
lastRid = 0;
lastRidNew = rowIdArray[totalRow-1];
}
//cout << "rowid allocated is " << lastRid << endl;
//if a new extent is created, all the columns in this table should have their own new extent
//First column already processed
//@Bug 1701. Close the file (if uncompressed)
m_colOp[op(curCol.compressionType)]->closeColumnFile(curCol);
//cout << "Saving hwm info for new ext batch" << endl;
//Update hwm to set them in the end
bool succFlag = false;
unsigned colWidth = 0;
int curFbo = 0, curBio;
for (i=0; i < totalColumns; i++)
{
//shoud be obtained from saved hwm
aColExtsInfo = tableMetaData->getColExtsInfo(colStructList[i].dataOid);
it = aColExtsInfo.begin();
while (it != aColExtsInfo.end())
{
if ((it->dbRoot == colStructList[i].fColDbRoot) && (it->partNum == colStructList[i].fColPartition)
&& (it->segNum == colStructList[i].fColSegment) && it->current)
break;
it++;
}
if (it != aColExtsInfo.end()) //update hwm info
{
oldHwm = it->hwm;
}
// save hwm for the old extent
colWidth = colStructList[i].colWidth;
succFlag = colOp->calculateRowId(lastRid, BYTE_PER_BLOCK/colWidth, colWidth, curFbo, curBio);
//cout << "insertcolumnrec oid:rid:fbo:oldhwm = " << colStructList[i].dataOid << ":" << lastRid << ":" << curFbo << ":" << oldHwm << endl;
if (succFlag)
{
if ((HWM)curFbo >= oldHwm)
{
it->hwm = (HWM)curFbo;
}
//@Bug 4947. set current to false for old extent.
if (newExtent)
{
it->current = false;
}
//cout << "updated old ext info for oid " << colStructList[i].dataOid << " dbroot:part:seg:hwm:current = "
//<< it->dbRoot<<":"<<it->partNum<<":"<<it->segNum<<":"<<it->hwm<<":"<< it->current<< " and newExtent is " << newExtent << endl;
}
else
return ERR_INVALID_PARAM;
//update hwm for the new extent
if (newExtent)
{
it = aColExtsInfo.begin();
while (it != aColExtsInfo.end())
{
if ((it->dbRoot == newColStructList[i].fColDbRoot) && (it->partNum == newColStructList[i].fColPartition)
&& (it->segNum == newColStructList[i].fColSegment) && it->current)
break;
it++;
}
succFlag = colOp->calculateRowId(lastRidNew, BYTE_PER_BLOCK/colWidth, colWidth, curFbo, curBio);
if (succFlag)
{
if (it != aColExtsInfo.end())
{
it->hwm = (HWM)curFbo;
//cout << "setting hwm to " << (int)curFbo <<" for seg " <<it->segNum << endl;
it->current = true;
}
}
else
return ERR_INVALID_PARAM;
}
tableMetaData->setColExtsInfo(colStructList[i].dataOid, aColExtsInfo);
}
// end of allocate row id
#ifdef PROFILE
timer.start("writeColumnRec");
#endif
//cout << "Writing column record" << endl;
if (rc == NO_ERROR)
{
//----------------------------------------------------------------------
//Mark extents invalid
//----------------------------------------------------------------------
vector<BRM::LBID_t> lbids;
vector<CalpontSystemCatalog::ColDataType> colDataTypes;
bool successFlag = true;
unsigned width = 0;
int curFbo = 0, curBio, lastFbo = -1;
if (isFirstBatchPm && (totalRow == rowsLeft))
{}
else {
for (unsigned i = 0; i < colStructList.size(); i++)
{
colOp = m_colOp[op(colStructList[i].fCompressionType)];
width = colStructList[i].colWidth;
successFlag = colOp->calculateRowId(lastRid , BYTE_PER_BLOCK/width, width, curFbo, curBio);
if (successFlag) {
if (curFbo != lastFbo) {
RETURN_ON_ERROR(AddLBIDtoList(txnid,
lbids,
colDataTypes,
colStructList[i],
curFbo));
}
}
}
}
if (lbids.size() > 0)
rc = BRMWrapper::getInstance()->markExtentsInvalid(lbids, colDataTypes);
//----------------------------------------------------------------------
// Write row(s) to database file(s)
//----------------------------------------------------------------------
rc = writeColumnRecBinary(txnid, colStructList, colValueList, rowIdArray, newColStructList, tableOid, useTmpSuffix); // @bug 5572 HDFS tmp file
}
return rc;
}
int WriteEngineWrapper::insertColumnRec_SYS(const TxnID& txnid,
ColStructList& colStructList,
ColValueList& colValueList,
@ -3909,6 +4541,204 @@ timer.finish();
return rc;
}
int WriteEngineWrapper::writeColumnRecBinary(const TxnID& txnid,
const ColStructList& colStructList,
std::vector<uint64_t>& colValueList,
RID* rowIdArray,
const ColStructList& newColStructList,
const int32_t tableOid,
bool useTmpSuffix,
bool versioning)
{
bool bExcp;
int rc = 0;
void* valArray;
string segFile;
Column curCol;
ColStructList::size_type totalColumn;
ColStructList::size_type i;
size_t totalRow;
setTransId(txnid);
totalColumn = colStructList.size();
#ifdef PROFILE
StopWatch timer;
#endif
totalRow = colValueList.size() / totalColumn;
valArray = malloc(sizeof(uint64_t) * totalRow);
if (totalRow == 0)
return rc;
TableMetaData* aTbaleMetaData = TableMetaData::makeTableMetaData(tableOid);
for (i = 0; i < totalColumn; i++)
{
RID * secondPart = rowIdArray + totalRow;
//@Bug 2205 Check if all rows go to the new extent
//Write the first batch
RID * firstPart = rowIdArray;
ColumnOp* colOp = m_colOp[op(colStructList[i].fCompressionType)];
// set params
colOp->initColumn(curCol);
// need to pass real dbRoot, partition, and segment to setColParam
colOp->setColParam(curCol, 0, colStructList[i].colWidth,
colStructList[i].colDataType, colStructList[i].colType, colStructList[i].dataOid,
colStructList[i].fCompressionType, colStructList[i].fColDbRoot,
colStructList[i].fColPartition, colStructList[i].fColSegment);
ColExtsInfo aColExtsInfo = aTbaleMetaData->getColExtsInfo(colStructList[i].dataOid);
ColExtsInfo::iterator it = aColExtsInfo.begin();
while (it != aColExtsInfo.end())
{
if ((it->dbRoot == colStructList[i].fColDbRoot) && (it->partNum == colStructList[i].fColPartition) && (it->segNum == colStructList[i].fColSegment))
break;
it++;
}
if (it == aColExtsInfo.end()) //add this one to the list
{
ColExtInfo aExt;
aExt.dbRoot =colStructList[i].fColDbRoot;
aExt.partNum = colStructList[i].fColPartition;
aExt.segNum = colStructList[i].fColSegment;
aExt.compType = colStructList[i].fCompressionType;
aColExtsInfo.push_back(aExt);
aTbaleMetaData->setColExtsInfo(colStructList[i].dataOid, aColExtsInfo);
}
rc = colOp->openColumnFile(curCol, segFile, useTmpSuffix, IO_BUFF_SIZE); // @bug 5572 HDFS tmp file
if (rc != NO_ERROR)
break;
// handling versioning
vector<LBIDRange> rangeList;
if (versioning)
{
rc = processVersionBuffer(curCol.dataFile.pFile, txnid, colStructList[i],
colStructList[i].colWidth, totalRow, firstPart, rangeList);
if (rc != NO_ERROR) {
if (colStructList[i].fCompressionType == 0)
{
curCol.dataFile.pFile->flush();
}
BRMWrapper::getInstance()->writeVBEnd(txnid, rangeList);
break;
}
}
//totalRow1 -= totalRow2;
// have to init the size here
// nullArray = (bool*) malloc(sizeof(bool) * totalRow);
uint8_t tmp8;
uint16_t tmp16;
uint32_t tmp32;
float tmpF;
double tmpD;
for (size_t j = 0; j < totalRow; j++)
{
uint64_t curValue = colValueList[(totalRow*i) + j];
switch (colStructList[i].colType)
{
case WriteEngine::WR_INT:
tmp32 = curValue;
((int*)valArray)[j] = tmp32;
break;
case WriteEngine::WR_UINT:
tmp32 = curValue;
((uint32_t*)valArray)[j] = tmp32;
break;
case WriteEngine::WR_VARBINARY : // treat same as char for now
case WriteEngine::WR_CHAR:
case WriteEngine::WR_BLOB:
case WriteEngine::WR_TEXT:
switch (colStructList[i].colWidth)
{
case 1:
tmp8 = curValue;
((uint8_t*)valArray)[j] = tmp8;
break;
case 2:
tmp16 = curValue;
((uint16_t*)valArray)[j] = tmp16;
break;
case 3:
tmp32 = curValue;
((uint32_t*)valArray)[j] = tmp32;
break;
case 4:
((uint32_t*)valArray)[j] = curValue;
break;
}
break;
case WriteEngine::WR_FLOAT:
tmp32 = curValue;
memcpy(&((float*)valArray)[j], &tmp32, 4);
break;
case WriteEngine::WR_DOUBLE:
memcpy(&((double*)valArray)[j], &curValue, 8);
break;
case WriteEngine::WR_BYTE:
tmp8 = curValue;
((char*)valArray)[j] = tmp8;
break;
case WriteEngine::WR_UBYTE:
tmp8 = curValue;
((uint8_t*)valArray)[j] = tmp8;
break;
case WriteEngine::WR_SHORT:
tmp16 = curValue;
((short*)valArray)[j] = tmp16;
break;
case WriteEngine::WR_USHORT:
tmp32 = curValue;
((uint16_t*)valArray)[j] = tmp16;
break;
case WriteEngine::WR_LONGLONG:
tmp32 = curValue;
((long long*)valArray)[j] = tmp32;
break;
case WriteEngine::WR_ULONGLONG:
((uint64_t*)valArray)[j] = curValue;
break;
case WriteEngine::WR_TOKEN:
memcpy(&((Token*)valArray)[j], &curValue, 8);
break;
}
}
#ifdef PROFILE
timer.start("writeRow ");
#endif
rc = colOp->writeRow(curCol, totalRow, firstPart, valArray);
#ifdef PROFILE
timer.stop("writeRow ");
#endif
colOp->closeColumnFile(curCol);
if (versioning)
BRMWrapper::getInstance()->writeVBEnd(txnid, rangeList);
// check error
if (rc != NO_ERROR)
break;
} // end of for (i = 0
if (valArray != NULL)
free(valArray);
#ifdef PROFILE
timer.finish();
#endif
return rc;
}
int WriteEngineWrapper::writeColumnRec(const TxnID& txnid,
const ColStructList& colStructList,
const ColValueList& colValueList,

19
writeengine/wrapper/writeengine.h
View File

@ -304,6 +304,20 @@ public:
OID tableOid = 0,
bool isFirstBatchPm = false);
EXPORT int insertColumnRecsBinary(const TxnID& txnid,
ColStructList& colStructList,
std::vector<uint64_t>& colValueList,
DctnryStructList& dctnryStructList,
DictStrList& dictStrList,
std::vector<boost::shared_ptr<DBRootExtentTracker> > & dbRootExtentTrackers,
RBMetaWriter* fRBMetaWriter,
bool bFirstExtentOnThisPM,
bool insertSelect = false,
bool isAutoCommitOn = false,
OID tableOid = 0,
bool isFirstBatchPm = false);
/**
* @brief Insert values into systables
* @param colStructList column structure list
@ -646,6 +660,11 @@ private:
ColValueList& newColValueList, const int32_t tableOid,
bool useTmpSuffix, bool versioning = true);
int writeColumnRecBinary(const TxnID& txnid, const ColStructList& colStructList,
std::vector<uint64_t>& colValueList,
RID* rowIdArray, const ColStructList& newColStructList,
const int32_t tableOid,
bool useTmpSuffix, bool versioning = true);
//@Bug 1886,2870 pass the address of ridList vector