Files
spark/src/DataStreamSuite/DataStream/Primitives/GroupBy/MaxGroupByEvaluator.cs
T

43 lines
2.0 KiB
C#

using System;
using System.Collections;
using BCG.Utilities;
namespace BCG {
/// <summary>
/// Computes the max for a given column
/// </summary>
public class MaxGroupByEvaluator : GroupByEvaluator {
public MaxGroupByEvaluator(string outputCol, string[] groupByCols) : base(outputCol, groupByCols) { }
public MaxGroupByEvaluator(string colMapEpression, string outputCol, string[] groupByCols)
: base(colMapEpression, outputCol, groupByCols) { }
public override IBlueRow evaluate(DataPacket dp) {
myOutputTable = myOutputTable == null ? new GroupByOutputTable(dp.schema) :
myOutputTable;
string hashString = BlueDataUtility.rowToString(GroupByStream.getRepRow(dp), myGroupByCols);
IBlueRow currMaxRow = myOutputTable.getRow(hashString);
IBlueRow maxRow = BlueDataUtility.computeMaxRow(dp.deletedRows, myOutputColumn);
if (maxRow != null) {
bts.Log.assert(currMaxRow != null);
bts.Log.assert(BlueDataUtility.compare_rows(maxRow, currMaxRow, myOutputColumn) <= 0);
if (BlueDataUtility.compare_rows(maxRow, currMaxRow, myOutputColumn) == 0) {
maxRow = BlueDataUtility.computeMaxRow(dp.newRows, myOutputColumn);
if (maxRow == null)
myOutputTable.removeRow(hashString);
else
myOutputTable.addRow(maxRow, hashString);
}
}
maxRow = BlueDataUtility.computeMaxRow(dp.addedRows, myOutputColumn);
// if the row was replaced in the above lines, then currMaxRow < maxRow, so the if below would never return a true,
// otherwise if maxRow > currMaxRow then its replaced
if (maxRow != null && (currMaxRow == null || BlueDataUtility.compare_rows(maxRow, currMaxRow, myOutputColumn) > 0))
myOutputTable.addRow(maxRow, hashString);
return myProject(myOutputTable.getRow(hashString));
}
}
}