1
0
mirror of https://github.com/mariadb-corporation/mariadb-columnstore-engine.git synced 2025-04-18 21:44:02 +03:00

1270 lines
33 KiB
C++

/* Copyright (C) 2014 InfiniDB, Inc.
Copyright (C) 2016-2021 MariaDB Corporation
This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; version 2 of
the License.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston,
MA 02110-1301, USA. */
//
// $Id: columncommand.cpp 2057 2013-02-13 17:00:10Z pleblanc $
// C++ Implementation: columncommand
//
// Description:
//
//
// Author: Patrick LeBlanc <pleblanc@calpont.com>, (C) 2008
// Copyright: See COPYING file that comes with this distribution
//
//
#include <unistd.h>
#include <sstream>
#include <map>
#include <cstdlib>
#include <cmath>
using namespace std;
#include "bpp.h"
#include "errorcodes.h"
#include "exceptclasses.h"
#include "primitiveserver.h"
#include "primproc.h"
#include "stats.h"
#include "datatypes/mcs_int128.h"
using namespace messageqcpp;
using namespace rowgroup;
#include "messageids.h"
using namespace logging;
namespace primitiveprocessor
{
extern int noVB;
ColumnCommand::ColumnCommand() : Command(COLUMN_COMMAND), blockCount(0), loadCount(0), suppressFilter(false)
{
}
ColumnCommand::ColumnCommand(execplan::CalpontSystemCatalog::ColType& aColType) : ColumnCommand()
{
colType = aColType;
}
ColumnCommand::~ColumnCommand()
{
}
void ColumnCommand::_execute()
{
if (_isScan)
makeScanMsg();
else if (bpp->ridCount == 0) // this would cause a scan
{
blockCount += colType.colWidth;
return; // a step with no input rids does nothing
}
else
makeStepMsg();
issuePrimitive();
processResult();
// check if feeding a filtercommand
if (fFilterFeeder != NOT_FEEDER)
copyRidsForFilterCmd();
}
void ColumnCommand::execute()
{
if (fFilterFeeder == LEFT_FEEDER)
{
values = bpp->fFiltCmdValues[0].get();
wide128Values = bpp->fFiltCmdBinaryValues[0].get();
}
else if (fFilterFeeder == RIGHT_FEEDER)
{
values = bpp->fFiltCmdValues[1].get();
wide128Values = bpp->fFiltCmdBinaryValues[1].get();
}
else
{
values = bpp->values;
wide128Values = bpp->wide128Values;
bpp->valuesLBID = lbid;
}
_execute();
}
void ColumnCommand::execute(int64_t* vals)
{
values = vals;
_execute();
}
void ColumnCommand::makeScanMsg()
{
/* Finish the NewColRequestHeader. */
/* XXXPAT: if there is a Command preceeding this one, it's a DictScan feeding tokens
which need to become filter elements. Can we handle that with this design?
Implement that later. */
primMsg->ism.Size = baseMsgLength;
primMsg->NVALS = 0;
primMsg->LBID = lbid;
primMsg->RidFlags = 0xFFFF;
// cout << "scanning lbid " << lbid << " colwidth = " << primMsg->DataSize <<
// " filterCount = " << filterCount << " outputType = " <<
// (int) primMsg->OutputType << endl;
}
void ColumnCommand::makeStepMsg()
{
memcpy(&inputMsg[baseMsgLength], bpp->relRids, bpp->ridCount << 1);
primMsg->RidFlags = bpp->ridMap;
primMsg->ism.Size = baseMsgLength + (bpp->ridCount << 1);
primMsg->NVALS = bpp->ridCount;
primMsg->LBID = lbid;
// cout << "lbid is " << lbid << endl;
}
template <int W>
void ColumnCommand::_loadData()
{
using ByteStreamType = typename messageqcpp::ByteStreamType<W>::type;
uint32_t wasCached;
uint32_t blocksRead;
uint16_t _mask;
uint64_t oidLastLbid = 0;
bool lastBlockReached = false;
oidLastLbid = getLastLbid();
uint32_t blocksToLoad = 0;
// The number of elements allocated equals to the number of
// iteratations of the first loop here.
BRM::LBID_t* lbids = (BRM::LBID_t*)alloca(W * sizeof(BRM::LBID_t));
uint8_t** blockPtrs = (uint8_t**)alloca(W * sizeof(uint8_t*));
int i;
_mask = mask;
// primMsg->RidFlags = 0xffff; // disables selective block loading
// cerr << "::ColumnCommand::_loadData OID " << getOID() << " l:" << primMsg->LBID << " ll: " << oidLastLbid
// << " primMsg->RidFlags " << primMsg->RidFlags << endl;
for (i = 0; i < W; ++i, _mask <<= shift)
{
if ((!lastBlockReached && _isScan) || (!_isScan && primMsg->RidFlags & _mask))
{
lbids[blocksToLoad] = primMsg->LBID + i;
blockPtrs[blocksToLoad] = &bpp->blockData[i * BLOCK_SIZE];
blocksToLoad++;
loadCount++;
}
else if (lastBlockReached && _isScan)
{
// fill remaining blocks with empty values when col scan
uint32_t blockLen = BLOCK_SIZE / W;
auto attrs = datatypes::TypeAttributesStd(W, 0, -1);
const auto* typeHandler = datatypes::TypeHandler::find(colType.colDataType, attrs);
const uint8_t* emptyValue = typeHandler->getEmptyValueForType(attrs);
uint8_t* blockDataPtr = &bpp->blockData[i * BLOCK_SIZE];
idbassert(blockDataPtr);
fillEmptyBlock<ByteStreamType>(blockDataPtr, emptyValue, blockLen);
} // else
if ((primMsg->LBID + i) == oidLastLbid)
lastBlockReached = true;
blockCount++;
} // for
/* Do the load */
wasCached = primitiveprocessor::loadBlocks(lbids, bpp->versionInfo, bpp->txnID, colType.compressionType,
blockPtrs, &blocksRead, bpp->LBIDTrace, bpp->sessionID,
blocksToLoad, &wasVersioned, willPrefetch(), &bpp->vssCache);
bpp->cachedIO += wasCached;
bpp->physIO += blocksRead;
bpp->touchedBlocks += blocksToLoad;
if (hasAuxCol_)
{
BRM::LBID_t* lbidsAux = (BRM::LBID_t*)alloca(1 * sizeof(BRM::LBID_t));
uint8_t** blockPtrsAux = (uint8_t**)alloca(1 * sizeof(uint8_t*));
blockPtrsAux[0] = &bpp->blockDataAux[0];
lbidsAux[0] = lbidAux;
wasCached = primitiveprocessor::loadBlocks(lbidsAux, bpp->versionInfo, bpp->txnID, 2, blockPtrsAux,
&blocksRead, bpp->LBIDTrace, bpp->sessionID, 1, &wasVersioned,
true, &bpp->vssCache);
bpp->cachedIO += wasCached;
bpp->physIO += blocksRead;
bpp->touchedBlocks += 1;
}
}
void ColumnCommand::loadData()
{
switch (colType.colWidth)
{
case 1: _loadData<1>(); break;
case 2: _loadData<2>(); break;
case 4: _loadData<4>(); break;
case 8: _loadData<8>(); break;
case 16: _loadData<16>(); break;
default:
throw NotImplementedExcept(std::string("ColumnCommand::loadData does not support ") +
std::to_string(colType.colWidth) + std::string(" byte width."));
}
}
void ColumnCommand::issuePrimitive()
{
loadData();
if (!suppressFilter)
{
bpp->getPrimitiveProcessor().setParsedColumnFilter(parsedColumnFilter);
}
else
{
bpp->getPrimitiveProcessor().setParsedColumnFilter(emptyFilter);
}
switch (colType.colWidth)
{
case 1: _issuePrimitive<1>(); break;
case 2: _issuePrimitive<2>(); break;
case 4: _issuePrimitive<4>(); break;
case 8: _issuePrimitive<8>(); break;
case 16: _issuePrimitive<16>(); break;
default:
throw NotImplementedExcept(std::string("ColumnCommand::_issuePrimitive does not support ") +
std::to_string(colType.colWidth) + std::string(" byte width."));
}
if (_isScan)
{
if (LIKELY(colType.isNarrow()))
updateCPDataNarrow();
else
updateCPDataWide();
}
}
template <int W>
void ColumnCommand::_issuePrimitiveNarrow()
{
_loadData<W>();
if (!suppressFilter)
bpp->getPrimitiveProcessor().setParsedColumnFilter(parsedColumnFilter);
else
bpp->getPrimitiveProcessor().setParsedColumnFilter(emptyFilter);
_issuePrimitive<W>();
if (_isScan)
updateCPDataNarrow();
}
template <int W>
void ColumnCommand::_issuePrimitive()
{
using IntegralType = typename datatypes::WidthToSIntegralType<W>::type;
primMsg->hasAuxCol = hasAuxCol_;
// Down the call stack the code presumes outMsg buffer has enough space to store
// ColRequestHeader + uint16_t Rids[8192] + IntegralType[8192].
bpp->getPrimitiveProcessor().columnScanAndFilter<IntegralType>(primMsg, outMsg);
} // _issuePrimitive()
void ColumnCommand::updateCPDataNarrow()
{
/* Update CP data, the PseudoColumn code should always be !_isScan. Should be safe
to leave this here for now. */
if (_isScan)
{
bpp->validCPData = (outMsg->ValidMinMax && !wasVersioned);
bpp->cpDataFromDictScan = false;
bpp->lbidForCP = lbid;
bpp->maxVal = static_cast<int64_t>(outMsg->Max);
bpp->minVal = static_cast<int64_t>(outMsg->Min);
}
}
void ColumnCommand::updateCPDataWide()
{
/* Update CP data, the PseudoColumn code should always be !_isScan. Should be safe
to leave this here for now. */
if (_isScan)
{
bpp->validCPData = (outMsg->ValidMinMax && !wasVersioned);
bpp->cpDataFromDictScan = false;
bpp->lbidForCP = lbid;
if (colType.isWideDecimalType())
{
bpp->hasWideColumnOut = true;
// colWidth is int32 and wideColumnWidthOut's
// value is expected to be at most uint8.
bpp->wideColumnWidthOut = colType.colWidth;
bpp->max128Val = outMsg->Max;
bpp->min128Val = outMsg->Min;
}
else
{
throw NotImplementedExcept("ColumnCommand::updateCPDataWide(): unsupported dataType " +
std::to_string(primMsg->colType.DataType) + " with width " +
std::to_string(colType.colWidth));
}
}
}
template <int W>
void ColumnCommand::_process_OT_BOTH_wAbsRids()
{
using T = typename datatypes::WidthToSIntegralType<W>::type;
bpp->ridCount = outMsg->NVALS;
bpp->ridMap = outMsg->RidFlags;
uint8_t* outPtr = reinterpret_cast<uint8_t*>(&outMsg[1]);
auto* ridPos = primitives::getRIDArrayPosition(outPtr, 0);
T* valuesPos = primitives::getValuesArrayPosition<T>(primitives::getFirstValueArrayPosition(outMsg), 0);
for (size_t i = 0; i < outMsg->NVALS; ++i)
{
bpp->relRids[i] = ridPos[i];
bpp->absRids[i] = ridPos[i] + bpp->baseRid;
values[i] = valuesPos[i];
}
}
template <>
void ColumnCommand::_process_OT_BOTH_wAbsRids<16>()
{
using T = typename datatypes::WidthToSIntegralType<16>::type;
bpp->ridCount = outMsg->NVALS;
bpp->ridMap = outMsg->RidFlags;
uint8_t* outPtr = reinterpret_cast<uint8_t*>(&outMsg[1]);
auto* ridPos = primitives::getRIDArrayPosition(outPtr, 0);
int128_t* valuesPos =
primitives::getValuesArrayPosition<T>(primitives::getFirstValueArrayPosition(outMsg), 0);
for (size_t i = 0; i < outMsg->NVALS; ++i)
{
bpp->relRids[i] = ridPos[i];
bpp->absRids[i] = ridPos[i] + bpp->baseRid;
datatypes::TSInt128::assignPtrPtr(&wide128Values[i], &valuesPos[i]);
}
}
template <int W>
void ColumnCommand::_process_OT_BOTH()
{
using T = typename datatypes::WidthToSIntegralType<W>::type;
bpp->ridCount = outMsg->NVALS;
bpp->ridMap = outMsg->RidFlags;
uint8_t* outPtr = reinterpret_cast<uint8_t*>(&outMsg[1]);
auto* ridPos = primitives::getRIDArrayPosition(outPtr, 0);
T* valuesPos = primitives::getValuesArrayPosition<T>(primitives::getFirstValueArrayPosition(outMsg), 0);
for (size_t i = 0; i < outMsg->NVALS; ++i)
{
bpp->relRids[i] = ridPos[i];
values[i] = valuesPos[i];
}
}
template <>
void ColumnCommand::_process_OT_BOTH<16>()
{
using T = typename datatypes::WidthToSIntegralType<16>::type;
bpp->ridCount = outMsg->NVALS;
bpp->ridMap = outMsg->RidFlags;
uint8_t* outPtr = reinterpret_cast<uint8_t*>(&outMsg[1]);
auto* ridPos = primitives::getRIDArrayPosition(outPtr, 0);
T* valuesPos = primitives::getValuesArrayPosition<T>(primitives::getFirstValueArrayPosition(outMsg), 0);
for (size_t i = 0; i < outMsg->NVALS; ++i)
{
bpp->relRids[i] = ridPos[i];
datatypes::TSInt128::assignPtrPtr(&wide128Values[i], &valuesPos[i]);
}
}
void ColumnCommand::process_OT_BOTH()
{
if (makeAbsRids)
{
switch (colType.colWidth)
{
case 16: _process_OT_BOTH_wAbsRids<16>(); break;
case 8: _process_OT_BOTH_wAbsRids<8>(); break;
case 4: _process_OT_BOTH_wAbsRids<4>(); break;
case 2: _process_OT_BOTH_wAbsRids<2>(); break;
case 1: _process_OT_BOTH_wAbsRids<1>(); break;
default:
throw NotImplementedExcept(
std::string("ColumnCommand::process_OT_BOTH with ABS RIDs does not support ") +
std::to_string(colType.colWidth) + std::string(" byte width."));
}
}
else
{
switch (colType.colWidth)
{
case 16: _process_OT_BOTH<16>(); break;
case 8: _process_OT_BOTH<8>(); break;
case 4: _process_OT_BOTH<4>(); break;
case 2: _process_OT_BOTH<2>(); break;
case 1: _process_OT_BOTH<1>(); break;
default:
throw NotImplementedExcept(std::string("ColumnCommand::process_OT_BOTH does not support ") +
std::to_string(colType.colWidth) + std::string(" byte width."));
}
}
}
void ColumnCommand::process_OT_RID()
{
memcpy(bpp->relRids, outMsg + 1, outMsg->NVALS << 1);
bpp->ridCount = outMsg->NVALS;
bpp->ridMap = outMsg->RidFlags;
}
// TODO This spec makes an impicit type conversion to fit types with sizeof(type) <= 4
// into 8 bytes. Treat values as a buffer and use memcpy to store the values in one go.
template <int W>
void ColumnCommand::_process_OT_DATAVALUE()
{
using T = typename datatypes::WidthToSIntegralType<W>::type;
bpp->ridCount = outMsg->NVALS;
T* arr = (T*)(outMsg + 1);
for (size_t i = 0; i < outMsg->NVALS; ++i)
values[i] = arr[i];
}
template <>
void ColumnCommand::_process_OT_DATAVALUE<8>()
{
bpp->ridCount = outMsg->NVALS;
memcpy(values, outMsg + 1, outMsg->NVALS << 3);
}
template <>
void ColumnCommand::_process_OT_DATAVALUE<16>()
{
bpp->ridCount = outMsg->NVALS;
memcpy(wide128Values, outMsg + 1, outMsg->NVALS << 4);
}
void ColumnCommand::process_OT_DATAVALUE()
{
bpp->ridCount = outMsg->NVALS;
switch (colType.colWidth)
{
case 16: _process_OT_DATAVALUE<16>(); break;
case 8: _process_OT_DATAVALUE<8>(); break;
case 4: _process_OT_DATAVALUE<4>(); break;
case 2: _process_OT_DATAVALUE<2>(); break;
case 1: _process_OT_DATAVALUE<2>(); break;
default:
throw NotImplementedExcept(std::string("ColumnCommand::process_OT_DATAVALUE does not support ") +
std::to_string(colType.colWidth) + std::string(" byte width."));
}
}
void ColumnCommand::processResult()
{
/* Switch on output type, turn pCol output into something useful, store it in
the containing BPP */
switch (outMsg->OutputType)
{
case OT_BOTH: process_OT_BOTH(); break;
case OT_RID: process_OT_RID(); break;
case OT_DATAVALUE: process_OT_DATAVALUE(); break;
default:
cout << "outputType = " << outMsg->OutputType << endl;
throw logic_error("ColumnCommand got a bad OutputType");
}
// check if feeding a filtercommand
if (fFilterFeeder == LEFT_FEEDER)
{
bpp->fFiltRidCount[0] = bpp->ridCount;
for (uint64_t i = 0; i < bpp->ridCount; i++)
bpp->fFiltCmdRids[0][i] = bpp->relRids[i];
}
else if (fFilterFeeder == RIGHT_FEEDER)
{
bpp->fFiltRidCount[1] = bpp->ridCount;
for (uint64_t i = 0; i < bpp->ridCount; i++)
bpp->fFiltCmdRids[1][i] = bpp->relRids[i];
}
}
// Used by PseudoCC at least
void ColumnCommand::createCommand(ByteStream& bs)
{
uint8_t tmp8;
bs.advance(1);
colType.unserialize(bs);
bs >> tmp8;
_isScan = tmp8;
bs >> traceFlags;
bs >> filterString;
bs >> BOP;
bs >> filterCount;
bs >> tmp8;
hasAuxCol_ = tmp8;
deserializeInlineVector(bs, lastLbid);
Command::createCommand(bs);
switch (colType.colWidth)
{
case 1: createColumnFilter<datatypes::WidthToSIntegralType<1>::type>(); break;
case 2: createColumnFilter<datatypes::WidthToSIntegralType<2>::type>(); break;
case 4: createColumnFilter<datatypes::WidthToSIntegralType<4>::type>(); break;
case 8: createColumnFilter<datatypes::WidthToSIntegralType<8>::type>(); break;
case 16: createColumnFilter<datatypes::WidthToSIntegralType<16>::type>(); break;
default:
throw NotImplementedExcept(std::string("ColumnCommand::createCommand does not support ") +
std::to_string(colType.colWidth) + std::string(" byte width."));
}
}
void ColumnCommand::createCommand(execplan::CalpontSystemCatalog::ColType& aColType, ByteStream& bs)
{
colType = aColType;
uint8_t tmp8;
bs >> tmp8;
_isScan = tmp8;
bs >> traceFlags;
bs >> filterString;
bs >> BOP;
bs >> filterCount;
bs >> tmp8;
hasAuxCol_ = tmp8;
deserializeInlineVector(bs, lastLbid);
Command::createCommand(bs);
}
void ColumnCommand::resetCommand(ByteStream& bs)
{
bs >> lbid;
if (hasAuxCol_)
bs >> lbidAux;
}
void ColumnCommand::prep(int8_t outputType, bool absRids)
{
fillInPrimitiveMessageHeader(outputType, absRids);
switch (colType.colWidth)
{
case 1:
shift = 16;
mask = 0xFFFF;
break;
case 2:
shift = 8;
mask = 0xFF;
break;
case 4:
shift = 4;
mask = 0x0F;
break;
case 8:
shift = 2;
mask = 0x03;
break;
case 16:
shift = 1;
mask = 0x01;
break;
default:
cout << "CC: colWidth is " << colType.colWidth << endl;
throw logic_error("ColumnCommand: bad column width?");
}
}
void ColumnCommand::fillInPrimitiveMessageHeader(const int8_t outputType, const bool absRids)
{
// WIP Align this structure or move input RIDs away.
baseMsgLength = sizeof(NewColRequestHeader) + (suppressFilter ? 0 : filterString.length());
size_t inputMsgBufSize = baseMsgLength + (LOGICAL_BLOCK_RIDS * sizeof(primitives::RIDType));
if (!inputMsg)
inputMsg.reset(new (std::align_val_t(utils::MAXCOLUMNWIDTH)) uint8_t[inputMsgBufSize]);
primMsg = (NewColRequestHeader*)inputMsg.get();
outMsg = (ColResultHeader*)bpp->outputMsg.get();
makeAbsRids = absRids;
primMsg->ism.Interleave = 0;
primMsg->ism.Flags = 0;
primMsg->ism.Command = COL_BY_SCAN;
primMsg->ism.Size = sizeof(NewColRequestHeader) + (suppressFilter ? 0 : filterString.length());
primMsg->ism.Type = 2;
primMsg->hdr.SessionID = bpp->sessionID;
primMsg->hdr.TransactionID = bpp->txnID;
primMsg->hdr.VerID = bpp->versionInfo.currentScn;
primMsg->hdr.StepID = bpp->stepID;
primMsg->colType = ColRequestHeaderDataType(colType);
primMsg->OutputType = outputType;
primMsg->BOP = BOP;
primMsg->NOPS = (suppressFilter ? 0 : filterCount);
primMsg->sort = 0;
}
/* Assumes OT_DATAVALUE */
void ColumnCommand::projectResult()
{
auto nvals = outMsg->NVALS;
if (primMsg->NVALS != nvals || nvals != bpp->ridCount)
{
ostringstream os;
BRM::DBRM brm;
BRM::OID_t oid;
uint16_t l_dbroot;
uint32_t partNum;
uint16_t segNum;
uint32_t fbo;
brm.lookupLocal(lbid, 0, false, oid, l_dbroot, partNum, segNum, fbo);
os << __FILE__ << " error on projection for oid " << oid << " lbid " << lbid;
if (primMsg->NVALS != outMsg->NVALS)
os << ": input rids " << primMsg->NVALS;
else
os << ": ridcount " << bpp->ridCount;
os << ", output rids " << nvals << endl;
// cout << os.str();
if (bpp->sessionID & 0x80000000)
throw NeedToRestartJob(os.str());
else
throw PrimitiveColumnProjectResultExcept(os.str());
}
idbassert(primMsg->NVALS == nvals);
idbassert(bpp->ridCount == nvals);
uint32_t valuesByteSize = nvals * colType.colWidth;
*bpp->serialized << valuesByteSize;
bpp->serialized->append(primitives::getFirstValueArrayPosition(outMsg), valuesByteSize);
}
void ColumnCommand::removeRowsFromRowGroup(RowGroup& rg)
{
uint32_t gapSize = colType.colWidth + 2;
uint8_t* msg8;
uint16_t rid;
Row oldRow, newRow;
uint32_t oldIdx, newIdx;
rg.initRow(&oldRow);
rg.initRow(&newRow);
rg.getRow(0, &oldRow);
rg.getRow(0, &newRow);
msg8 = (uint8_t*)(outMsg + 1);
for (oldIdx = newIdx = 0; newIdx < outMsg->NVALS; newIdx++, msg8 += gapSize)
{
rid = *((uint16_t*)msg8);
while (rid != bpp->relRids[oldIdx])
{
// need to find rid in relrids, and it is in there
oldIdx++;
oldRow.nextRow();
}
if (oldIdx != newIdx)
{
bpp->relRids[newIdx] = rid;
// we use a memcpy here instead of copyRow() to avoid expanding the string table;
memcpy(newRow.getData(), oldRow.getData(), newRow.getSize());
}
oldIdx++;
oldRow.nextRow();
newRow.nextRow();
}
rg.setRowCount(outMsg->NVALS); // this gets rid of trailing rows, no need to set them to NULL
bpp->ridCount = outMsg->NVALS;
primMsg->NVALS = outMsg->NVALS;
}
template <typename T>
void ColumnCommand::_projectResultRGLoop(rowgroup::Row& r, const T* valuesArray, const uint32_t offset)
{
for (size_t i = 0; i < outMsg->NVALS; ++i)
{
r.setIntField_offset(valuesArray[i], offset);
r.nextRow(rowSize);
}
}
template <>
void ColumnCommand::_projectResultRGLoop<int128_t>(rowgroup::Row& r, const int128_t* valuesArray,
const uint32_t offset)
{
for (size_t i = 0; i < outMsg->NVALS; ++i)
{
r.setBinaryField_offset(&valuesArray[i], colType.colWidth, offset);
r.nextRow(rowSize);
}
}
template <int W>
void ColumnCommand::_projectResultRG(RowGroup& rg, uint32_t pos)
{
using T = typename datatypes::WidthToSIntegralType<W>::type;
T* valuesArray = primitives::getValuesArrayPosition<T>(primitives::getFirstValueArrayPosition(outMsg), 0);
uint32_t offset;
auto nvals = outMsg->NVALS;
rg.initRow(&r);
offset = r.getOffset(pos);
rowSize = r.getSize();
if ((primMsg->NVALS != nvals || nvals != bpp->ridCount) && (!noVB || bpp->sessionID & 0x80000000))
{
ostringstream os;
BRM::DBRM brm;
BRM::OID_t oid;
uint16_t dbroot;
uint32_t partNum;
uint16_t segNum;
uint32_t fbo;
brm.lookupLocal(lbid, 0, false, oid, dbroot, partNum, segNum, fbo);
os << __FILE__ << " error on projectResultRG for oid " << oid << " lbid " << lbid;
if (primMsg->NVALS != nvals)
os << ": input rids " << primMsg->NVALS;
else
os << ": ridcount " << bpp->ridCount;
os << ", output rids " << nvals;
os << endl;
if (bpp->sessionID & 0x80000000)
throw NeedToRestartJob(os.str());
else
throw PrimitiveColumnProjectResultExcept(os.str());
}
else if (primMsg->NVALS != nvals || nvals != bpp->ridCount)
removeRowsFromRowGroup(rg);
idbassert(primMsg->NVALS == nvals);
idbassert(bpp->ridCount == nvals);
rg.getRow(0, &r);
_projectResultRGLoop(r, valuesArray, offset);
}
void ColumnCommand::projectResultRG(RowGroup& rg, uint32_t pos)
{
switch (colType.colWidth)
{
case 1: _projectResultRG<1>(rg, pos); break;
case 2: _projectResultRG<2>(rg, pos); break;
case 4: _projectResultRG<4>(rg, pos); break;
case 8: _projectResultRG<8>(rg, pos); break;
case 16: _projectResultRG<16>(rg, pos); break;
}
}
void ColumnCommand::project()
{
/* bpp->ridCount == 0 would signify a scan operation */
if (bpp->ridCount == 0)
{
*bpp->serialized << (uint32_t)0;
blockCount += colType.colWidth;
return;
}
makeStepMsg();
issuePrimitive();
projectResult();
}
void ColumnCommand::projectIntoRowGroup(RowGroup& rg, uint32_t pos)
{
if (bpp->ridCount == 0)
{
blockCount += colType.colWidth;
return;
}
makeStepMsg();
issuePrimitive();
projectResultRG(rg, pos);
}
void ColumnCommand::nextLBID()
{
lbid += colType.colWidth;
if (hasAuxCol_)
lbidAux += execplan::AUX_COL_WIDTH;
}
void ColumnCommand::duplicate(ColumnCommand* cc)
{
cc->_isScan = _isScan;
cc->hasAuxCol_ = hasAuxCol_;
cc->traceFlags = traceFlags;
cc->filterString = filterString;
cc->colType.colDataType = colType.colDataType;
cc->colType.compressionType = colType.compressionType;
cc->colType.colWidth = colType.colWidth;
cc->colType.charsetNumber = colType.charsetNumber;
cc->BOP = BOP;
cc->filterCount = filterCount;
cc->fFilterFeeder = fFilterFeeder;
cc->parsedColumnFilter = parsedColumnFilter;
cc->suppressFilter = suppressFilter;
cc->lastLbid = lastLbid;
cc->r = r;
cc->rowSize = rowSize;
cc->Command::operator=(*this);
}
SCommand ColumnCommand::duplicate()
{
SCommand ret;
ret.reset(new ColumnCommand());
duplicate((ColumnCommand*)ret.get());
return ret;
}
bool ColumnCommand::operator==(const ColumnCommand& cc) const
{
if (_isScan != cc._isScan)
return false;
if (hasAuxCol_ != cc.hasAuxCol_)
return false;
if (BOP != cc.BOP)
return false;
if (filterString != cc.filterString)
return false;
if (filterCount != cc.filterCount)
return false;
if (makeAbsRids != cc.makeAbsRids)
return false;
if (colType.colWidth != cc.colType.colWidth)
return false;
if (colType.colDataType != cc.colType.colDataType)
return false;
return true;
}
bool ColumnCommand::operator!=(const ColumnCommand& cc) const
{
return !(*this == cc);
}
ColumnCommand& ColumnCommand::operator=(const ColumnCommand& c)
{
_isScan = c._isScan;
hasAuxCol_ = c.hasAuxCol_;
traceFlags = c.traceFlags;
filterString = c.filterString;
colType.colDataType = c.colType.colDataType;
colType.compressionType = c.colType.compressionType;
colType.colWidth = c.colType.colWidth;
BOP = c.BOP;
filterCount = c.filterCount;
fFilterFeeder = c.fFilterFeeder;
parsedColumnFilter = c.parsedColumnFilter;
suppressFilter = c.suppressFilter;
lastLbid = c.lastLbid;
return *this;
}
bool ColumnCommand::willPrefetch()
{
return (blockCount == 0 || ((double)loadCount) / ((double)blockCount) > bpp->prefetchThreshold);
}
void ColumnCommand::disableFilters()
{
suppressFilter = true;
prep(primMsg->OutputType, makeAbsRids);
}
void ColumnCommand::enableFilters()
{
suppressFilter = false;
prep(primMsg->OutputType, makeAbsRids);
}
void ColumnCommand::getLBIDList(uint32_t loopCount, vector<int64_t>* lbids)
{
int64_t firstLBID = lbid, lastLBID = firstLBID + (loopCount * colType.colWidth) - 1, i;
for (i = firstLBID; i <= lastLBID; i++)
lbids->push_back(i);
if (hasAuxCol_)
{
firstLBID = lbidAux;
lastLBID = firstLBID + (loopCount * execplan::AUX_COL_WIDTH) - 1;
for (i = firstLBID; i <= lastLBID; i++)
lbids->push_back(i);
}
}
int64_t ColumnCommand::getLastLbid()
{
if (!_isScan)
return 0;
return lastLbid[bpp->dbRoot - 1];
}
ColumnCommand* ColumnCommandFabric::createCommand(messageqcpp::ByteStream& bs)
{
bs.advance(1); // The higher dispatcher Command::makeCommand calls BS::peek so this increments BS ptr
execplan::CalpontSystemCatalog::ColType colType;
colType.unserialize(bs);
switch (colType.colWidth)
{
case 1: return new ColumnCommandInt8(colType, bs); break;
case 2: return new ColumnCommandInt16(colType, bs); break;
case 4: return new ColumnCommandInt32(colType, bs); break;
case 8: return new ColumnCommandInt64(colType, bs); break;
case 16: return new ColumnCommandInt128(colType, bs); break;
default:
throw NotImplementedExcept(std::string("ColumnCommandFabric::createCommand: unsupported width ") +
std::to_string(colType.colWidth));
}
return nullptr;
}
template <typename T>
void ColumnCommand::createColumnFilter()
{
parsedColumnFilter =
primitives::_parseColumnFilter<T>(filterString.buf(), colType.colDataType, filterCount, BOP);
/* OR hack */
emptyFilter = primitives::_parseColumnFilter<T>(filterString.buf(), colType.colDataType, 0, BOP);
}
ColumnCommand* ColumnCommandFabric::duplicate(const ColumnCommandUniquePtr& rhs)
{
auto& command = *rhs;
if (LIKELY(typeid(command) == typeid(ColumnCommandInt64)))
{
ColumnCommandInt64* ret = new ColumnCommandInt64();
*ret = *dynamic_cast<ColumnCommandInt64*>(rhs.get());
return ret;
}
else if (typeid(command) == typeid(ColumnCommandInt128))
{
ColumnCommandInt128* ret = new ColumnCommandInt128();
*ret = *dynamic_cast<ColumnCommandInt128*>(rhs.get());
return ret;
}
else if (typeid(command) == typeid(ColumnCommandInt8))
{
ColumnCommandInt8* ret = new ColumnCommandInt8();
*ret = *dynamic_cast<ColumnCommandInt8*>(rhs.get());
return ret;
}
else if (typeid(command) == typeid(ColumnCommandInt16))
{
ColumnCommandInt16* ret = new ColumnCommandInt16();
*ret = *dynamic_cast<ColumnCommandInt16*>(rhs.get());
return ret;
}
else if (typeid(command) == typeid(ColumnCommandInt32))
{
ColumnCommandInt32* ret = new ColumnCommandInt32();
*ret = *dynamic_cast<ColumnCommandInt32*>(rhs.get());
return ret;
}
else
throw NotImplementedExcept("ColumnCommandFabric::duplicate: Can not detect ColumnCommand child class");
return nullptr;
}
// Code duplication here for the future patch.
ColumnCommandInt8::ColumnCommandInt8(execplan::CalpontSystemCatalog::ColType& aColType,
messageqcpp::ByteStream& bs)
{
ColumnCommand::createCommand(aColType, bs);
createColumnFilter<IntegralType>();
}
void ColumnCommandInt8::prep(int8_t outputType, bool absRids)
{
shift = 16;
mask = 0xFFFF;
ColumnCommand::fillInPrimitiveMessageHeader(outputType, absRids);
}
void ColumnCommandInt8::loadData()
{
_loadData<size>();
}
void ColumnCommandInt8::process_OT_BOTH()
{
if (makeAbsRids)
_process_OT_BOTH_wAbsRids<size>();
else
_process_OT_BOTH<size>();
}
void ColumnCommandInt8::process_OT_DATAVALUE()
{
_process_OT_DATAVALUE<size>();
}
void ColumnCommandInt8::projectResultRG(RowGroup& rg, uint32_t pos)
{
_projectResultRG<size>(rg, pos);
}
void ColumnCommandInt8::issuePrimitive()
{
_issuePrimitiveNarrow<size>();
}
ColumnCommandInt16::ColumnCommandInt16(execplan::CalpontSystemCatalog::ColType& aColType,
messageqcpp::ByteStream& bs)
{
ColumnCommand::createCommand(aColType, bs);
createColumnFilter<IntegralType>();
}
void ColumnCommandInt16::prep(int8_t outputType, bool absRids)
{
shift = 8;
mask = 0xFF;
ColumnCommand::fillInPrimitiveMessageHeader(outputType, absRids);
}
void ColumnCommandInt16::loadData()
{
_loadData<size>();
}
void ColumnCommandInt16::process_OT_BOTH()
{
if (makeAbsRids)
_process_OT_BOTH_wAbsRids<size>();
else
_process_OT_BOTH<size>();
}
void ColumnCommandInt16::process_OT_DATAVALUE()
{
_process_OT_DATAVALUE<size>();
}
void ColumnCommandInt16::projectResultRG(RowGroup& rg, uint32_t pos)
{
_projectResultRG<size>(rg, pos);
}
void ColumnCommandInt16::issuePrimitive()
{
_issuePrimitiveNarrow<size>();
}
ColumnCommandInt32::ColumnCommandInt32(execplan::CalpontSystemCatalog::ColType& aColType,
messageqcpp::ByteStream& bs)
{
ColumnCommand::createCommand(aColType, bs);
createColumnFilter<IntegralType>();
}
void ColumnCommandInt32::prep(int8_t outputType, bool absRids)
{
shift = 4;
mask = 0x0F;
ColumnCommand::fillInPrimitiveMessageHeader(outputType, absRids);
}
void ColumnCommandInt32::loadData()
{
_loadData<size>();
}
void ColumnCommandInt32::process_OT_BOTH()
{
if (makeAbsRids)
_process_OT_BOTH_wAbsRids<size>();
else
_process_OT_BOTH<size>();
}
void ColumnCommandInt32::process_OT_DATAVALUE()
{
_process_OT_DATAVALUE<size>();
}
void ColumnCommandInt32::projectResultRG(RowGroup& rg, uint32_t pos)
{
_projectResultRG<size>(rg, pos);
}
void ColumnCommandInt32::issuePrimitive()
{
_issuePrimitiveNarrow<size>();
}
ColumnCommandInt64::ColumnCommandInt64(execplan::CalpontSystemCatalog::ColType& aColType,
messageqcpp::ByteStream& bs)
{
ColumnCommand::createCommand(aColType, bs);
createColumnFilter<IntegralType>();
}
void ColumnCommandInt64::prep(int8_t outputType, bool absRids)
{
shift = 2;
mask = 0x03;
ColumnCommand::fillInPrimitiveMessageHeader(outputType, absRids);
}
void ColumnCommandInt64::loadData()
{
_loadData<size>();
}
void ColumnCommandInt64::process_OT_BOTH()
{
if (makeAbsRids)
_process_OT_BOTH_wAbsRids<size>();
else
_process_OT_BOTH<size>();
}
void ColumnCommandInt64::process_OT_DATAVALUE()
{
_process_OT_DATAVALUE<size>();
}
void ColumnCommandInt64::projectResultRG(RowGroup& rg, uint32_t pos)
{
_projectResultRG<size>(rg, pos);
}
void ColumnCommandInt64::issuePrimitive()
{
_issuePrimitiveNarrow<size>();
}
ColumnCommandInt128::ColumnCommandInt128(execplan::CalpontSystemCatalog::ColType& aColType,
messageqcpp::ByteStream& bs)
{
ColumnCommand::createCommand(aColType, bs);
createColumnFilter<IntegralType>();
}
void ColumnCommandInt128::prep(int8_t outputType, bool absRids)
{
shift = 1;
mask = 0x01;
ColumnCommand::fillInPrimitiveMessageHeader(outputType, absRids);
}
void ColumnCommandInt128::loadData()
{
_loadData<size>();
}
void ColumnCommandInt128::process_OT_BOTH()
{
if (makeAbsRids)
_process_OT_BOTH_wAbsRids<size>();
else
_process_OT_BOTH<size>();
}
void ColumnCommandInt128::process_OT_DATAVALUE()
{
_process_OT_DATAVALUE<size>();
}
void ColumnCommandInt128::projectResultRG(RowGroup& rg, uint32_t pos)
{
_projectResultRG<size>(rg, pos);
}
void ColumnCommandInt128::issuePrimitive()
{
loadData();
if (!suppressFilter)
bpp->getPrimitiveProcessor().setParsedColumnFilter(parsedColumnFilter);
else
bpp->getPrimitiveProcessor().setParsedColumnFilter(emptyFilter);
_issuePrimitive<size>();
if (_isScan)
updateCPDataWide();
}
} // namespace primitiveprocessor