mariadb-columnstore-engine/utils/rowgroup/rowaggregation.h

/*
   Copyright (C) 2014 InfiniDB, Inc.
   Copyright (c) 2019 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.
*/

#pragma once

/** @file rowaggregation.h
 * Classes in this file are used to aggregate Rows in RowGroups.
 * RowAggregation is the class that performs the aggregation.
 * RowAggGroupByCol and RowAggFunctionCol are support classes used to describe
 * the columns involved in the aggregation.
 * @endcode
 */

#include <cstring>
#include <cstdint>
#include <utility>
#include <vector>
#include <unordered.h>

#include <boost/shared_ptr.hpp>

#include <boost/scoped_array.hpp>
#include <boost/scoped_ptr.hpp>

#include "serializeable.h"
#include "bytestream.h"
#include "rowgroup.h"
#include "hasher.h"
#include "stlpoolallocator.h"
#include "returnedcolumn.h"
#include "mcsv1_udaf.h"
#include "constantcolumn.h"

#include "resourcemanager.h"
#include "rowstorage.h"
#include "nullstring.h"

// To do: move code that depends on joblist to a proper subsystem.
namespace joblist
{
class ResourceManager;
}

namespace rowgroup
{
/** @brief Enumerates aggregate functions supported by RowAggregation
 */
enum RowAggFunctionType
{
  ROWAGG_FUNCT_UNDEFINE,  // default
  ROWAGG_COUNT_ASTERISK,  // COUNT(*) counts all rows including nulls
  ROWAGG_COUNT_COL_NAME,  // COUNT(column_name) only counts non-null rows
  ROWAGG_SUM,
  ROWAGG_AVG,
  ROWAGG_MIN,
  ROWAGG_MAX,

  // Statistics Function, ROWAGG_STATS is the generic name.
  ROWAGG_STATS,
  ROWAGG_STDDEV_POP,
  ROWAGG_STDDEV_SAMP,
  ROWAGG_VAR_POP,
  ROWAGG_VAR_SAMP,

  // BIT Function, ROWAGG_BIT_OP is the generic name.
  ROWAGG_BIT_OP,
  ROWAGG_BIT_AND,
  ROWAGG_BIT_OR,
  ROWAGG_BIT_XOR,

  // GROUP_CONCAT
  ROWAGG_GROUP_CONCAT,

  ROWAGG_JSON_ARRAY,

  // DISTINCT: performed on UM only
  ROWAGG_COUNT_DISTINCT_COL_NAME,  // COUNT(distinct column_name) only counts non-null rows
  ROWAGG_DISTINCT_SUM,
  ROWAGG_DISTINCT_AVG,

  // Constant
  ROWAGG_CONSTANT,

  // User Defined Aggregate Function
  ROWAGG_UDAF,

  // If an Aggregate has more than one parameter, this will be used for parameters after the first
  ROWAGG_MULTI_PARM,

  // internal function type to avoid duplicate the work
  // handling ROWAGG_COUNT_NO_OP, ROWAGG_DUP_FUNCT and ROWAGG_DUP_AVG is a little different
  // ROWAGG_COUNT_NO_OP  :  count done by AVG, no need to copy
  // ROWAGG_DUP_FUNCT    :  copy data before AVG calculation, because SUM may share by AVG
  // ROWAGG_DUP_AVG      :  copy data after AVG calculation
  ROWAGG_COUNT_NO_OP,  // COUNT(column_name), but leave count() to AVG
  ROWAGG_DUP_FUNCT,    // duplicate aggregate Function(), except AVG and UDAF, in select
  ROWAGG_DUP_AVG,      // duplicate AVG(column_name) in select
  ROWAGG_DUP_STATS,    // duplicate statistics functions in select
  ROWAGG_DUP_UDAF,     // duplicate UDAF function in select

  // a dummy "select some" aggregate needed for non-group-by values in SELECT's with GROUP BY's
  ROWAGG_SELECT_SOME
};

//------------------------------------------------------------------------------
/** @brief Specifies a column in a RowGroup that is part of the aggregation
 *   "GROUP BY" clause.
 */
//------------------------------------------------------------------------------
struct RowAggGroupByCol
{
  /** @brief RowAggGroupByCol constructor
   *
   * @param inputColIndex(in) column index into input row
   * @param outputColIndex(in) column index into output row
   *    outputColIndex argument should be omitted if this GroupBy
   *    column is not to be included in the output.
   */
  explicit RowAggGroupByCol(int32_t inputColIndex, int32_t outputColIndex = -1)
   : fInputColumnIndex(inputColIndex), fOutputColumnIndex(outputColIndex)
  {
  }
  ~RowAggGroupByCol() = default;

  uint32_t fInputColumnIndex;
  uint32_t fOutputColumnIndex;
};

inline messageqcpp::ByteStream& operator<<(messageqcpp::ByteStream& b, RowAggGroupByCol& o)
{
  return (b << o.fInputColumnIndex << o.fOutputColumnIndex);
}
inline messageqcpp::ByteStream& operator>>(messageqcpp::ByteStream& b, RowAggGroupByCol& o)
{
  return (b >> o.fInputColumnIndex >> o.fOutputColumnIndex);
}

//------------------------------------------------------------------------------
/** @brief Specifies a column in a RowGroup that is to be aggregated, and what
 *   aggregation function is to be performed.
 *
 *   If a column is aggregated more than once(ex: SELECT MIN(l_shipdate),
 *   MAX(l_shipdate)...), then 2 RowAggFunctionCol objects should be created
 *   with the same inputColIndex, one for the MIN function, and one for the
 *   MAX function.
 */
//------------------------------------------------------------------------------
struct RowAggFunctionCol
{
  /** @brief RowAggFunctionCol constructor
   *
   * @param aggFunction(in)    aggregation function to be performed
   * @param inputColIndex(in)  column index into input row
   * @param outputColIndex(in) column index into output row
   * @param auxColIndex(in)    auxiliary index into output row for avg/count
   * @param stats(in)          real statistics function where generic name in aggFunction
   */
  RowAggFunctionCol(RowAggFunctionType aggFunction, RowAggFunctionType stats, int32_t inputColIndex,
                    int32_t outputColIndex, int32_t auxColIndex = -1)
   : fAggFunction(aggFunction)
   , fStatsFunction(stats)
   , fInputColumnIndex(inputColIndex)
   , fOutputColumnIndex(outputColIndex)
   , fAuxColumnIndex(auxColIndex)
   , hasMultiParm(false)
  {
  }
  virtual ~RowAggFunctionCol() = default;

  virtual void serialize(messageqcpp::ByteStream& bs) const;
  virtual void deserialize(messageqcpp::ByteStream& bs);

  RowAggFunctionType fAggFunction;  // aggregate function
  // statistics function stores ROWAGG_STATS in fAggFunction and real function in fStatsFunction
  RowAggFunctionType fStatsFunction;

  uint32_t fInputColumnIndex;
  uint32_t fOutputColumnIndex;

  // fAuxColumnIndex is used in 4 cases:
  // 1. for AVG - point to the count column, the fInputColumnIndex is for sum
  // 2. for statistics function - point to sum(x), +1 is sum(x**2)
  // 3. for UDAF - contain the context user data as binary
  // 4. for duplicate - point to the real aggretate column to be copied from
  // Set only on UM, the fAuxColumnIndex is defaulted to fOutputColumnIndex+1 on PM.
  uint32_t fAuxColumnIndex;

  // For UDAF that have more than one parameter and some parameters are constant.
  // There will be a series of RowAggFunctionCol created, one for each parameter.
  // The first will be a RowUDAFFunctionCol. Subsequent ones will be RowAggFunctionCol
  // with fAggFunction == ROWAGG_MULTI_PARM. Order is important.
  // If this parameter is constant, that value is here.
  execplan::SRCP fpConstCol;

  bool hasMultiParm;
};

struct RowUDAFFunctionCol : public RowAggFunctionCol
{
  RowUDAFFunctionCol(mcsv1sdk::mcsv1Context& context, int32_t inputColIndex, int32_t outputColIndex,
                     int32_t auxColIndex = -1)
   : RowAggFunctionCol(ROWAGG_UDAF, ROWAGG_FUNCT_UNDEFINE, inputColIndex, outputColIndex, auxColIndex)
   , fUDAFContext(context)
   , bInterrupted(false)
  {
    fUDAFContext.setInterrupted(&bInterrupted);
  }

  RowUDAFFunctionCol(int32_t inputColIndex, int32_t outputColIndex, int32_t auxColIndex = -1)
   : RowAggFunctionCol(ROWAGG_UDAF, ROWAGG_FUNCT_UNDEFINE, inputColIndex, outputColIndex, auxColIndex)
   , bInterrupted(false)
  {
  }
  RowUDAFFunctionCol(const RowUDAFFunctionCol& rhs)
   : RowAggFunctionCol(ROWAGG_UDAF, ROWAGG_FUNCT_UNDEFINE, rhs.fInputColumnIndex, rhs.fOutputColumnIndex,
                       rhs.fAuxColumnIndex)
   , fUDAFContext(rhs.fUDAFContext)
   , bInterrupted(false)
  {
  }

  ~RowUDAFFunctionCol() override = default;

  void serialize(messageqcpp::ByteStream& bs) const override;
  void deserialize(messageqcpp::ByteStream& bs) override;

  mcsv1sdk::mcsv1Context fUDAFContext;  // The UDAF context
  bool bInterrupted;                    // Shared by all the threads
};

inline void RowAggFunctionCol::serialize(messageqcpp::ByteStream& bs) const
{
  bs << (uint8_t)fAggFunction;
  bs << fInputColumnIndex;
  bs << fOutputColumnIndex;

  if (fpConstCol)
  {
    bs << (uint8_t)1;
    fpConstCol.get()->serialize(bs);
  }
  else
  {
    bs << (uint8_t)0;
  }
}

inline void RowAggFunctionCol::deserialize(messageqcpp::ByteStream& bs)
{
  bs >> (uint8_t&)fAggFunction;
  bs >> fInputColumnIndex;
  bs >> fOutputColumnIndex;
  uint8_t t;
  bs >> t;

  if (t)
  {
    fpConstCol.reset(new execplan::ConstantColumn);
    fpConstCol.get()->unserialize(bs);
  }
}

inline void RowUDAFFunctionCol::serialize(messageqcpp::ByteStream& bs) const
{
  RowAggFunctionCol::serialize(bs);
  fUDAFContext.serialize(bs);
}

inline void RowUDAFFunctionCol::deserialize(messageqcpp::ByteStream& bs)
{
  // This deserialize is called when the function gets to PrimProc.
  // reset is called because we're starting a new sub-evaluate cycle.
  RowAggFunctionCol::deserialize(bs);
  fUDAFContext.unserialize(bs);
  fUDAFContext.setInterrupted(&bInterrupted);
  mcsv1sdk::mcsv1_UDAF::ReturnCode rc;
  rc = fUDAFContext.getFunction()->reset(&fUDAFContext);

  if (rc == mcsv1sdk::mcsv1_UDAF::ERROR)
  {
    bInterrupted = true;
    throw logging::QueryDataExcept(fUDAFContext.getErrorMessage(), logging::aggregateFuncErr);
  }
}

struct ConstantAggData
{
  utils::NullString fConstValue;
  std::string fUDAFName;  // If a UDAF is called with constant.
  RowAggFunctionType fOp;

  ConstantAggData() : fOp(ROWAGG_FUNCT_UNDEFINE)
  {
  }

  ConstantAggData(utils::NullString v, RowAggFunctionType f, bool /*n*/) : fConstValue(v), fOp(f)
  {
  }

  ConstantAggData(utils::NullString v, std::string u, RowAggFunctionType f, bool /*n*/)
   : fConstValue(v), fUDAFName(u), fOp(f)
  {
  }
  bool isNull() const
  {
    return fConstValue.isNull();
  }
};

typedef boost::shared_ptr<RowAggGroupByCol> SP_ROWAGG_GRPBY_t;
typedef boost::shared_ptr<RowAggFunctionCol> SP_ROWAGG_FUNC_t;

struct GroupConcat : public messageqcpp::Serializeable
{
  // GROUP_CONCAT(DISTINCT col1, 'const', col2 ORDER BY col3 desc SEPARATOR 'sep')
  std::vector<std::pair<uint32_t, uint32_t>> fGroupCols;  // columns to concatenate, and position
  std::vector<std::pair<uint32_t, bool>> fOrderCols;      // columns to order by [asc/desc]
  std::string fSeparator;
  std::vector<std::pair<utils::NullString, uint32_t>> fConstCols;  // constant columns in group
  bool fDistinct;
  uint64_t fSize;

  RowGroup fRowGroup;
  std::shared_ptr<int[]> fMapping;
  std::vector<std::pair<int, bool>> fOrderCond;  // position to order by [asc/desc]
  long fTimeZone;
  uint32_t id;

  GroupConcat() = default;
  GroupConcat(joblist::ResourceManager* rm, boost::shared_ptr<int64_t> sessLimit)
   : fRm(rm), fSessionMemLimit(sessLimit)
  {
  }

  void serialize(messageqcpp::ByteStream& bs) const override;
  void deserialize(messageqcpp::ByteStream& bs) override;
  RGDataSizeType getDataSize() const;

  joblist::ResourceManager* fRm{nullptr};
  boost::shared_ptr<int64_t> fSessionMemLimit;
};

typedef boost::shared_ptr<GroupConcat> SP_GroupConcat;

//------------------------------------------------------------------------------
/** @brief Class that aggregates RowGroups.
 */
//------------------------------------------------------------------------------
class RowAggregation : public messageqcpp::Serializeable
{
 public:
  /** @brief RowAggregation default constructor
   *
   * @param rowAggGroupByCols(in) specify GroupBy columns and their
   *    mapping from input to output.  If vector is empty, then all the
   *    rows will be aggregated into a single implied group.  Order is
   *    important here. The primary GroupBy column should be first, the
   *    secondary GroupBy column should be second, etc.
   * @param rowAggFunctionCols(in) specify function columns and their
   *    mapping from input to output.
   */
  RowAggregation();
  RowAggregation(const std::vector<SP_ROWAGG_GRPBY_t>& rowAggGroupByCols,
                 const std::vector<SP_ROWAGG_FUNC_t>& rowAggFunctionCols,
                 joblist::ResourceManager* rm = nullptr, boost::shared_ptr<int64_t> sessMemLimit = {},
                 bool withRollup = false);
  RowAggregation(const RowAggregation& rhs);

  /** @brief RowAggregation default destructor
   */
  ~RowAggregation() override;

  /** @brief clone this object for multi-thread use
   */
  inline virtual RowAggregation* clone() const
  {
    return new RowAggregation(*this);
  }

  /** @brief Denotes end of data insertion following multiple calls to addRowGroup().
   */
  virtual void endOfInput();

  /** @brief reset RowAggregation outputRowGroup and hashMap
   */
  virtual void aggReset();

  /** @brief Define content of data to be aggregated and its aggregated output.
   *
   * @param pRowGroupIn(in)   contains definition of the input data.
   * @param pRowGroupOut(out) contains definition of the output data.
   */
  virtual void setInputOutput(const RowGroup& pRowGroupIn, RowGroup* pRowGroupOut)
  {
    fRowGroupIn = pRowGroupIn;
    fRowGroupOut = pRowGroupOut;
    initialize();
  }

  void clearRollup()
  {
    fRollupFlag = false;
  }

  bool hasRollup() const
  {
    return fRollupFlag;
  }

  /** @brief Define content of data to be joined
   *
   *    This method must be call after setInputOutput() for PM hashjoin case.
   *
   * @param pSmallSideRG(in) contains definition of the small side data.
   * @param pLargeSideRG(in) contains definition of the large side data.
   */
  void setJoinRowGroups(std::vector<RowGroup>* pSmallSideRG, RowGroup* pLargeSideRG);

  /** @brief Returns group by column vector
   *
   * This function is used to duplicate the RowAggregation object
   *
   * @returns a reference of the group by vector
   */
  std::vector<SP_ROWAGG_GRPBY_t>& getGroupByCols()
  {
    return fGroupByCols;
  }

  /** @brief Returns aggregate function vector
   *
   * This function is used to duplicate the RowAggregation object
   *
   * @returns a reference of the aggregation function vector
   */
  std::vector<SP_ROWAGG_FUNC_t>& getAggFunctions()
  {
    return fFunctionCols;
  }

  /** @brief Add a group of rows to be aggregated.
   *
   * This function can be called to iteratively add RowGroups for aggregation.
   *
   * @parm pRowGroupIn(in) RowGroup to be added to aggregation.
   */
  virtual void addRowGroup(const RowGroup* pRowGroupIn);
  virtual void addRowGroup(const RowGroup* pRowGroupIn,
                           std::vector<std::pair<Row::Pointer, uint64_t>>& inRows);

  /** @brief Serialize RowAggregation object into a ByteStream.
   *
   * @parm bs(out) BytesStream that is to be written to.
   */
  void serialize(messageqcpp::ByteStream& bs) const override;

  /** @brief Unserialize RowAggregation object from a ByteStream.
   *
   * @parm bs(in) BytesStream that is to be read from.
   */
  void deserialize(messageqcpp::ByteStream& bs) override;

  /** @brief load result set into byte stream
   *
   * @parm bs(out) BytesStream that is to be written to.
   */
  void loadResult(messageqcpp::ByteStream& bs);
  void loadEmptySet(messageqcpp::ByteStream& bs);

  /** @brief get output rowgroup
   *
   * @returns a const pointer of the output rowgroup
   */
  const RowGroup* getOutputRowGroup() const
  {
    return fRowGroupOut;
  }
  RowGroup* getOutputRowGroup()
  {
    return fRowGroupOut;
  }

  void append(RowAggregation* other);

  virtual void aggregateRow(Row& row, const uint64_t* hash = nullptr,
                            std::vector<mcsv1sdk::mcsv1Context>* rgContextColl = nullptr);
  inline uint32_t aggMapKeyLength() const
  {
    return fAggMapKeyCount;
  }

  inline void timeZone(long timeZone)
  {
    fTimeZone = timeZone;
  }
  inline long timeZone() const
  {
    return fTimeZone;
  }
  inline std::vector<mcsv1sdk::mcsv1Context>* rgContextColl()
  {
    return &fRGContextColl;
  }

  void finalAggregation()
  {
    return fRowAggStorage->finalize([this](Row& row) { mergeEntries(row); }, fRow);
  }

  std::unique_ptr<RGData> moveCurrentRGData()
  {
    return std::move(fCurRGData);
  }

 protected:
  virtual void initialize(bool hasGroupConcat = false);
  virtual void initMapData(const Row& row);
  virtual void attachGroupConcatAg();

  virtual void updateEntry(const Row& row, std::vector<mcsv1sdk::mcsv1Context>* rgContextColl = nullptr);
  void mergeEntries(const Row& row);
  virtual void doMinMax(const Row&, int64_t, int64_t, int);
  virtual void doSelectSome(const Row& rowIn, int64_t colIn, int64_t colOut);
  virtual void doSum(const Row&, int64_t, int64_t, int);
  virtual void doAvg(const Row&, int64_t, int64_t, int64_t, bool merge = false);
  virtual void doStatistics(const Row&, int64_t, int64_t, int64_t);
  void mergeStatistics(const Row&, uint64_t colOut, uint64_t colAux);
  void mergeGroupConcat(const Row& rowIn, uint64_t colOut);

  virtual void doBitOp(const Row&, int64_t, int64_t, int);
  virtual void doUDAF(const Row&, int64_t, int64_t, int64_t, uint64_t& funcColsIdx,
                      std::vector<mcsv1sdk::mcsv1Context>* rgContextColl = nullptr);
  virtual bool countSpecial(const RowGroup* pRG)
  {
    fRow.setUintField<8>(fRow.getUintField<8>(0) + pRG->getRowCount(), 0);
    return true;
  }

  void resetUDAF(RowUDAFFunctionCol* rowUDAF);
  void resetUDAF(RowUDAFFunctionCol* rowUDAF, uint64_t funcColIdx);

  inline bool isNull(const RowGroup* pRowGroup, const Row& row, int64_t col);
  inline void makeAggFieldsNull(Row& row);
  inline void copyNullRow(Row& row)
  {
    copyRow(fNullRow, &row);
  }

  inline void updateIntMinMax(int128_t val1, int128_t val2, int64_t col, int func);
  inline void updateIntMinMax(int64_t val1, int64_t val2, int64_t col, int func);
  inline void updateUintMinMax(uint64_t val1, uint64_t val2, int64_t col, int func);
  inline void updateCharMinMax(uint64_t val1, uint64_t val2, int64_t col, int func);
  inline void updateDoubleMinMax(double val1, double val2, int64_t col, int func);
  inline void updateLongDoubleMinMax(long double val1, long double val2, int64_t col, int func);
  inline void updateFloatMinMax(float val1, float val2, int64_t col, int func);
  inline void updateStringMinMax(utils::NullString val1, utils::NullString val2, int64_t col, int func);
  std::vector<SP_ROWAGG_GRPBY_t> fGroupByCols;
  std::vector<SP_ROWAGG_FUNC_t> fFunctionCols;
  uint32_t fAggMapKeyCount;  // the number of columns that make up the key
  RowGroup fRowGroupIn;
  RowGroup* fRowGroupOut;

  // for when the group by & distinct keys are not stored in the output rows
  rowgroup::RowGroup fKeyRG;

  Row fRow;
  Row fNullRow;
  Row* tmpRow;  // used by the hashers & eq functors
  boost::scoped_array<uint8_t> fNullRowData;
  rowgroup::RGData fNullRowRGData;
  rowgroup::RowGroup fNullRowGroup;

  std::unique_ptr<RowAggStorage> fRowAggStorage;

  // for support PM aggregation after PM hashjoin
  std::vector<RowGroup>* fSmallSideRGs;
  RowGroup* fLargeSideRG;
  std::shared_ptr<std::shared_ptr<int[]>[]> fSmallMappings;
  std::shared_ptr<int[]> fLargeMapping;
  uint32_t fSmallSideCount;
  boost::scoped_array<Row> rowSmalls;

  // for 8k poc
  RowGroup fEmptyRowGroup;
  RGData fEmptyRowData;
  Row fEmptyRow;

  bool fKeyOnHeap = false;

  long fTimeZone;

  // We need a separate copy for each thread.
  mcsv1sdk::mcsv1Context fRGContext;
  std::vector<mcsv1sdk::mcsv1Context> fRGContextColl;

  // These are handy for testing the actual type of static_any for UDAF
  static const static_any::any& charTypeId;
  static const static_any::any& scharTypeId;
  static const static_any::any& shortTypeId;
  static const static_any::any& intTypeId;
  static const static_any::any& longTypeId;
  static const static_any::any& llTypeId;
  static const static_any::any& int128TypeId;
  static const static_any::any& ucharTypeId;
  static const static_any::any& ushortTypeId;
  static const static_any::any& uintTypeId;
  static const static_any::any& ulongTypeId;
  static const static_any::any& ullTypeId;
  static const static_any::any& floatTypeId;
  static const static_any::any& doubleTypeId;
  static const static_any::any& longdoubleTypeId;
  static const static_any::any& strTypeId;

  // For UDAF along with with multiple distinct columns
  std::vector<SP_ROWAGG_FUNC_t>* fOrigFunctionCols;

  joblist::ResourceManager* fRm = nullptr;
  boost::shared_ptr<int64_t> fSessionMemLimit;
  std::unique_ptr<RGData> fCurRGData;
  bool fRollupFlag = false;

  std::string fTmpDir =
      config::Config::makeConfig()->getTempFileDir(config::Config::TempDirPurpose::Aggregates);
  std::string fCompStr = config::Config::makeConfig()->getConfig("RowAggregation", "Compression");

  std::vector<SP_GroupConcat> fGroupConcat;
};

//------------------------------------------------------------------------------
/** @brief derived Class that aggregates multi-rowgroups on UM
 *    One-phase case: aggregate from projected RG to final aggregated RG.
 */
//------------------------------------------------------------------------------
class RowAggregationUM : public RowAggregation
{
 public:
  /** @brief RowAggregationUM constructor
   */
  RowAggregationUM() = default;
  RowAggregationUM(const std::vector<SP_ROWAGG_GRPBY_t>& rowAggGroupByCols,
                   const std::vector<SP_ROWAGG_FUNC_t>& rowAggFunctionCols, joblist::ResourceManager*,
                   boost::shared_ptr<int64_t> sessionMemLimit, bool withRollup);
  RowAggregationUM(const RowAggregationUM& rhs);

  /** @brief RowAggregationUM default destructor
   */
  ~RowAggregationUM() override;

  /** @brief Denotes end of data insertion following multiple calls to addRowGroup().
   */
  void endOfInput() override;

  /** @brief Finializes the result set before sending back to the front end.
   */
  void finalize();

  /** @brief Returns aggregated rows in a RowGroup.
   *
   * This function should be called repeatedly until false is returned (meaning end of data).
   *
   * @returns true if more data, else false if no more data.
   */
  bool nextRowGroup();

  /** @brief Returns aggregated rows in a RowGroup as long as there are still not returned result RowGroups.
   *
   * This function should be called repeatedly until false is returned (meaning end of data).
   * Returns data from in-memory storage, as well as spilled data from disk. If disk-based aggregation is
   * happening, finalAggregation() should be called before returning result RowGroups to finalize the used
   * RowAggStorages, merge different spilled generations and obtain correct aggregation results.
   *
   * @returns True if there are more result RowGroups, else false if all results have been returned.
   */
  bool nextOutputRowGroup();

  /** @brief Add an aggregator for DISTINCT aggregation
   */
  void distinctAggregator(const boost::shared_ptr<RowAggregation>& da)
  {
    fDistinctAggregator = da;
  }

  /** @brief expressions to be evaluated after aggregation
   */
  void expression(const std::vector<execplan::SRCP>& exp)
  {
    fExpression = exp;
  }
  const std::vector<execplan::SRCP>& expression()
  {
    return fExpression;
  }

  // for multi threaded
  joblist::ResourceManager* getRm()
  {
    return fRm;
  }
  inline RowAggregationUM* clone() const override
  {
    return new RowAggregationUM(*this);
  }

  /** @brief access the aggregate(constant) columns
   */
  void constantAggregate(const std::vector<ConstantAggData>& v)
  {
    fConstantAggregate = v;
  }
  const std::vector<ConstantAggData>& constantAggregate() const
  {
    return fConstantAggregate;
  }

  /** @brief access the group_concat
   */
  void groupConcat(const std::vector<SP_GroupConcat>& v)
  {
    fGroupConcat = v;
  }
  const std::vector<SP_GroupConcat>& groupConcat() const
  {
    return fGroupConcat;
  }

  void aggReset() override;

  void setInputOutput(const RowGroup& pRowGroupIn, RowGroup* pRowGroupOut) override;

 protected:
  // virtual methods from base
  void initialize(bool hasGroupConcat = false) override;

  void attachGroupConcatAg() override;
  void updateEntry(const Row& row, std::vector<mcsv1sdk::mcsv1Context>* rgContextColl = nullptr) override;
  bool countSpecial(const RowGroup* pRG) override
  {
    fRow.setIntField<8>(fRow.getIntField<8>(fFunctionCols[0]->fOutputColumnIndex) + pRG->getRowCount(),
                        fFunctionCols[0]->fOutputColumnIndex);
    return true;
  }

  // calculate the average after all rows received. UM only function.
  void calculateAvgColumns();

  // calculate the statistics function all rows received. UM only function.
  void calculateStatisticsFunctions();

  // Sets the value from valOut into column colOut, performing any conversions.
  void SetUDAFValue(static_any::any& valOut, int64_t colOut);

  // If the datatype returned by evaluate isn't what we expect, convert.
  void SetUDAFAnyValue(static_any::any& valOut, int64_t colOut);

  // calculate the UDAF function all rows received. UM only function.
  void calculateUDAFColumns();

  // fix duplicates. UM only function.
  void fixDuplicates(RowAggFunctionType funct);

  // evaluate expressions
  virtual void evaluateExpression();

  // fix the aggregate(constant)
  virtual void fixConstantAggregate();
  virtual void doNullConstantAggregate(const ConstantAggData&, uint64_t);
  virtual void doNotNullConstantAggregate(const ConstantAggData&, uint64_t);

  // @bug3362, group_concat
  virtual void doGroupConcat(const Row&, int64_t, int64_t);
  virtual void setGroupConcatString();

  bool fHasAvg;
  bool fHasStatsFunc;
  bool fHasUDAF;

  boost::shared_ptr<RowAggregation> fDistinctAggregator;

  // for function on aggregation
  std::vector<execplan::SRCP> fExpression;

  /* Derived classes that use a lot of memory need to update totalMemUsage and request
   * the memory from rm in that order. */
  uint64_t fTotalMemUsage;

  // @bug3475, aggregate(constant), sum(0), count(null), etc
  std::vector<ConstantAggData> fConstantAggregate;

  // @bug3362, group_concat
  std::vector<SP_ROWAGG_FUNC_t> fFunctionColGc;

 private:
  uint64_t fLastMemUsage;
};

//------------------------------------------------------------------------------
/** @brief derived Class that aggregates PM partially aggregated RowGroups on UM
 *    Two-phase case:
 *      phase 1 - aggregate from projected RG to partial aggregated RG on PM
 *                The base RowAggregation handles the 1st phase.
 *      phase 2 - aggregate from partially aggregated RG to final RG on UM
 *                This class handles the 2nd phase.
 */
//------------------------------------------------------------------------------
class RowAggregationUMP2 : public RowAggregationUM
{
 public:
  /** @brief RowAggregationUM constructor
   */
  RowAggregationUMP2() = default;
  RowAggregationUMP2(const std::vector<SP_ROWAGG_GRPBY_t>& rowAggGroupByCols,
                     const std::vector<SP_ROWAGG_FUNC_t>& rowAggFunctionCols, joblist::ResourceManager*,
                     boost::shared_ptr<int64_t> sessionMemLimit, bool withRollup);
  RowAggregationUMP2(const RowAggregationUMP2& rhs);

  /** @brief RowAggregationUMP2 default destructor
   */
  ~RowAggregationUMP2() override;
  inline RowAggregationUMP2* clone() const override
  {
    return new RowAggregationUMP2(*this);
  }

 protected:
  // virtual methods from base
  void updateEntry(const Row& row, std::vector<mcsv1sdk::mcsv1Context>* rgContextColl = nullptr) override;
  void doAvg(const Row&, int64_t, int64_t, int64_t, bool merge = false) override;
  void doStatistics(const Row&, int64_t, int64_t, int64_t) override;
  void doGroupConcat(const Row&, int64_t, int64_t) override;
  void doBitOp(const Row&, int64_t, int64_t, int) override;
  void doUDAF(const Row&, int64_t, int64_t, int64_t, uint64_t& funcColsIdx,
              std::vector<mcsv1sdk::mcsv1Context>* rgContextColl = nullptr) override;
  bool countSpecial(const RowGroup* /*pRG*/) override
  {
    return false;
  }
};

//------------------------------------------------------------------------------
/** @brief derived Class that aggregates on distinct columns on UM
 *    The internal aggregator will handle one or two phases aggregation
 */
//------------------------------------------------------------------------------
class RowAggregationDistinct : public RowAggregationUMP2
{
 public:
  /** @brief RowAggregationDistinct constructor
   */
  RowAggregationDistinct() = default;
  RowAggregationDistinct(const std::vector<SP_ROWAGG_GRPBY_t>& rowAggGroupByCols,
                         const std::vector<SP_ROWAGG_FUNC_t>& rowAggFunctionCols, joblist::ResourceManager*,
                         boost::shared_ptr<int64_t> sessionMemLimit);

  /** @brief Copy Constructor for multi-threaded aggregation
   */
  RowAggregationDistinct(const RowAggregationDistinct& rhs);

  /** @brief RowAggregationDistinct default destructor
   */
  ~RowAggregationDistinct() override;

  /** @brief Add an aggregator for pre-DISTINCT aggregation
   */
  void addAggregator(const boost::shared_ptr<RowAggregation>& agg, const RowGroup& rg);

  void setInputOutput(const RowGroup& pRowGroupIn, RowGroup* pRowGroupOut) override;

  virtual void doDistinctAggregation();
  virtual void doDistinctAggregation_rowVec(std::vector<std::pair<Row::Pointer, uint64_t>>& inRows);
  void addRowGroup(const RowGroup* pRowGroupIn) override;
  void addRowGroup(const RowGroup* pRowGroupIn,
                   std::vector<std::pair<Row::Pointer, uint64_t>>& inRows) override;

  // multi-threade debug
  boost::shared_ptr<RowAggregation>& aggregator()
  {
    return fAggregator;
  }
  void aggregator(boost::shared_ptr<RowAggregation> aggregator)
  {
    fAggregator = std::move(aggregator);
  }
  RowGroup& rowGroupDist()
  {
    return fRowGroupDist;
  }
  void rowGroupDist(RowGroup& rowGroupDist)
  {
    fRowGroupDist = rowGroupDist;
  }
  inline RowAggregationDistinct* clone() const override
  {
    return new RowAggregationDistinct(*this);
  }

 protected:
  void updateEntry(const Row& row, std::vector<mcsv1sdk::mcsv1Context>* rgContextColl = nullptr) override;

  boost::shared_ptr<RowAggregation> fAggregator;
  RowGroup fRowGroupDist;
  RGData fDataForDist;
};

//------------------------------------------------------------------------------
/** @brief derived Class for aggregates multiple columns with distinct key word
 *    Get distinct values of the column per group by entry
 */
//------------------------------------------------------------------------------
class RowAggregationSubDistinct : public RowAggregationUM
{
 public:
  /** @brief RowAggregationSubDistinct constructor
   */
  RowAggregationSubDistinct() = default;
  RowAggregationSubDistinct(const std::vector<SP_ROWAGG_GRPBY_t>& rowAggGroupByCols,
                            const std::vector<SP_ROWAGG_FUNC_t>& rowAggFunctionCols,
                            joblist::ResourceManager*, boost::shared_ptr<int64_t> sessionMemLimit);
  RowAggregationSubDistinct(const RowAggregationSubDistinct& rhs);

  /** @brief RowAggregationSubDistinct default destructor
   */
  ~RowAggregationSubDistinct() override;

  void setInputOutput(const RowGroup& pRowGroupIn, RowGroup* pRowGroupOut) override;
  void addRowGroup(const RowGroup* pRowGroupIn) override;
  inline RowAggregationSubDistinct* clone() const override
  {
    return new RowAggregationSubDistinct(*this);
  }

  void addRowGroup(const RowGroup* pRowGroupIn,
                   std::vector<std::pair<Row::Pointer, uint64_t>>& inRow) override;

 protected:
  // virtual methods from RowAggregationUM
  void doGroupConcat(const Row&, int64_t, int64_t) override;
  // for groupby columns and the aggregated distinct column
  Row fDistRow;
  boost::scoped_array<uint8_t> fDistRowData;
};

//------------------------------------------------------------------------------
/** @brief derived Class that aggregates multiple columns with distinct key word
 *    Each distinct column will have its own aggregator
 */
//------------------------------------------------------------------------------
class RowAggregationMultiDistinct : public RowAggregationDistinct
{
 public:
  /** @brief RowAggregationMultiDistinct constructor
   */
  RowAggregationMultiDistinct() = default;
  RowAggregationMultiDistinct(const std::vector<SP_ROWAGG_GRPBY_t>& rowAggGroupByCols,
                              const std::vector<SP_ROWAGG_FUNC_t>& rowAggFunctionCols,
                              joblist::ResourceManager*, boost::shared_ptr<int64_t> sessionMemLimit);
  RowAggregationMultiDistinct(const RowAggregationMultiDistinct& rhs);

  /** @brief RowAggregationMultiDistinct default destructor
   */
  ~RowAggregationMultiDistinct() override;

  /** @brief Add sub aggregators
   */
  void addSubAggregator(const boost::shared_ptr<RowAggregationUM>& agg, const RowGroup& rg,
                        const std::vector<SP_ROWAGG_FUNC_t>& funct);

  void setInputOutput(const RowGroup& pRowGroupIn, RowGroup* pRowGroupOut) override;
  using RowAggregationDistinct::addRowGroup;
  void addRowGroup(const RowGroup* pRowGroupIn) override;

  using RowAggregationDistinct::doDistinctAggregation;
  void doDistinctAggregation() override;
  using RowAggregationDistinct::doDistinctAggregation_rowVec;
  virtual void doDistinctAggregation_rowVec(
      std::vector<std::vector<std::pair<Row::Pointer, uint64_t>>>& inRows);

  inline RowAggregationMultiDistinct* clone() const override
  {
    return new RowAggregationMultiDistinct(*this);
  }

  void addRowGroup(const RowGroup* pRowGroupIn,
                   std::vector<std::vector<std::pair<Row::Pointer, uint64_t>>>& inRows);

  std::vector<boost::shared_ptr<RowAggregationUM>>& subAggregators()
  {
    return fSubAggregators;
  }

  void subAggregators(std::vector<boost::shared_ptr<RowAggregationUM>>& subAggregators)
  {
    fSubAggregators = subAggregators;
  }

 protected:
  // virtual methods from base
  std::vector<boost::shared_ptr<RowAggregationUM>> fSubAggregators;
  std::vector<RowGroup> fSubRowGroups;
  std::vector<boost::shared_ptr<RGData>> fSubRowData;
  std::vector<std::vector<SP_ROWAGG_FUNC_t>> fSubFunctions;
};

typedef boost::shared_ptr<RowAggregation> SP_ROWAGG_t;
typedef boost::shared_ptr<RowAggregation> SP_ROWAGG_PM_t;
typedef boost::shared_ptr<RowAggregationUM> SP_ROWAGG_UM_t;
typedef boost::shared_ptr<RowAggregationDistinct> SP_ROWAGG_DIST;

}  // namespace rowgroup