Custom Aggregate Function Help

[John]

Hello,

I am using the RadGridView version 2012.2.607.40 for WPF and am having a heck of a time trying to determine how to write a custom aggregrate function to sum a column of System.TimeSpan values. My XAML definition of my grid view looks like this:

<telerik:RadGridView x:Name="scanSessionList" Grid.Row="0" Grid.Column="0" ItemsSource="{Binding ElementName=_this, Path=ScanSessions}" Margin="0,0,0,6" ShowGroupPanel="False" ShowGroupFooters="True">
    <telerik:RadGridView.GroupDescriptors>
        <telerik:GroupDescriptor Member="ScanSession.VehicleId" SortDirection="Ascending">
            <telerik:GroupDescriptor.AggregateFunctions>
                <telerik:CountFunction Caption="Missions: "/>
            </telerik:GroupDescriptor.AggregateFunctions>
        </telerik:GroupDescriptor>
    </telerik:RadGridView.GroupDescriptors>
    <telerik:RadGridView.Columns>
        <telerik:GridViewDataColumn Header="Description" DataMemberBinding="{Binding Path=ScanSession.Tag}"/>
        <telerik:GridViewDataColumn Header="Start Time" DataMemberBinding="{Binding Path=ScanSession.StartTimestamp, StringFormat=d MMM yyyy HH:mm}"/>
        <telerik:GridViewDataColumn Header="Elapsed Time" DataMemberBinding="{Binding Path=ScanSession.Duration, Converter={StaticResource TimeSpanConverter}}">
            <telerik:GridViewDataColumn.AggregateFunctions>
                <telerik:SumFunction/>

            </telerik:GridViewDataColumn.AggregateFunctions>
        </telerik:GridViewDataColumn>
        <telerik:GridViewDataColumn Header="Distance" DataMemberBinding="{Binding Path=ScanSession.DistanceTravelled, Converter={StaticResource DistanceConverter}}"/>
        <telerik:GridViewDataColumn Header="Avg Speed" DataMemberBinding="{Binding Path=ScanSession.AverageSpeed, Converter={StaticResource SpeedConverter}}"/>
        <telerik:GridViewDataColumn Header="Max Speed" DataMemberBinding="{Binding Path=ScanSession.MaximumSpeed, Converter={StaticResource SpeedConverter}}"/>
    </telerik:RadGridView.Columns>
</telerik:RadGridView>

Please note that ScanSession.Duration is of type System.TimeSpan.

I first tried the naive approach of just using the <telerik:SumFunction/> as shown above and got an InvalidOperationException with the error:

No generic method 'Sum' on type 'System.Linq.Enumerable' is compatible with the supplied type arguments and arguments. No type arguments should be provided if the method is non-generic.

So I assumed I needed to write a custom aggregrate function and looked at the standard deviation example Telerik has and came up with substituting <local:TimeSpanSumFunction SourceField="ScanSession.Duration"/>  for <telerik:SumFunction/> and added the following:

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using Telerik.Windows.Data;

namespace Exponent.Analyzer.Performance
{
    public class TimeSpanSumFunction : EnumerableSelectorAggregateFunction
    {
        protected override string AggregateMethodName
        {
            get { return "TimeSpanSum"; }
        }

        protected override Type ExtensionMethodsType
        {
            get { return typeof(Aggregates); }
        }
    }

    public static class Aggregates
    {
        public static TimeSpan TimeSpanSum<TSource>(IEnumerable<TSource> source, Func<TSource, TimeSpan> selector)
        {
            TimeSpan sum = TimeSpan.FromSeconds(0);

            foreach (TimeSpan ts in source.Select(s => selector(s)))
            {
                sum = sum.Add(ts);
            }

            return sum;
        }
    }
}

I then got an InvalidOperationException with the error:

No generic method 'TimeSpanSum' on type 'Exponent.Analyzer.Performance.Aggregates' is compatible with the supplied type arguments and arguments. No type arguments should be provided if the method is non-generic.

Finally, I wondered if I could just extend EnumerableAggregateFunction so I changed the XAML to use simply <local:TimeSpanSumFunction/> and came up with:

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using Telerik.Windows.Data;

namespace Exponent.Analyzer.Performance
{
    public class TimeSpanSumFunction : EnumerableAggregateFunction
    {
        protected override string AggregateMethodName
        {
            get { return "TimeSpanSum"; }
        }

        protected override Type ExtensionMethodsType
        {
            get { return typeof(Aggregates); }
        }
    }

    public static class Aggregates
    {
        public static TimeSpan TimeSpanSum<TSource>(IEnumerable<TimeSpan> source)
        {
            TimeSpan sum = TimeSpan.FromSeconds(0);

            foreach (TimeSpan ts in source)
            {
                sum = sum.Add(ts);
            }

            return sum;
        }
    }
}

and I still get an InvalidOperationException with the message:

No generic method 'TimeSpanSum' on type 'Exponent.Analyzer.Performance.Aggregates' is compatible with the supplied type arguments and arguments. No type arguments should be provided if the method is non-generic.

Obviously, I do not know how to write a custom aggregate function and I am finding what little documentation there is very unhelpful. I have more custom aggregate functions I need to write using other non-simple types.

Could somebody please give me some advise on how to write these custom aggregate functions?

And one other thing, can you also suggest how I can invoke my TimeSpanConverter (as seen in the XAML for the "Elapsed TIme" grid column") when displaying the results of the aggregte function?

Thank you,

-- john

[Dimitrina]

Hello John,

As the custom aggregates do not work in your case, then I would recommend you to try the approach suggested in this forum thread.

Kind regards,
Didie
the Telerik team

[John]

Hello Didie,

Thank you for the quick reply.

Are you saying the custom aggregate mechanism will not work in my case? In my original post, I did not mean to imply that they do not work; I meant to imply that I do not understand how to use them properly. I would hope that the custom aggregate mechanism would let me "sum" (or perform some other function) on all values of a column regardless of the data type of those values. Is this not possible?

Thank you for any more guidance you can provide on how to properly extend EnumerableAggregateFunction and EnumerableSelectorAggregateFunction.

-- john

[John]

Hi Didie,

I've had more time to play with this and found one solution that works. I had to change my custom aggregate code to the following:

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using Telerik.Windows.Data;
using Exponent.Analyzer.ViewModels;

namespace Exponent.Analyzer.Performance
{
    public class TimeSpanSumFunction : EnumerableAggregateFunction
    {
        protected override string AggregateMethodName
        {
            get { return "TimeSpanSum"; }
        }

        protected override Type ExtensionMethodsType
        {
            get { return typeof(Aggregates); }
        }
    }

    public static class Aggregates
    {
        public static TimeSpan TimeSpanSum<TSource>(IEnumerable<TSource> source)
        {
            TimeSpan sum = TimeSpan.FromSeconds(0);

            foreach (TSource ts in source)
            {
                sum = sum.Add((ts as ScanSessionViewModel).ScanSession.Duration);
            }

            return sum;
        }
    }
}

I assumed that my aggregate code would be receiving an IEnumerable of the type of the data being displayed in the particular grid column. That was a bad assumption. Instead, the IEnumerable contains data of the type of the ItemsSource for the grid.

Now I am going to see if I can figure out how to use the EnumerableSelectorAggregateFunction.

-- john

[Dimitrina]

Hi,

 I am glad to hear that you have resolve the problem. You can use the EnumerableSelectorAggregateFunction as demonstrated on Custom Aggregate Function Demo.

All the best,
Didie
the Telerik team

[Tomáš]

@John



To use the EnumerableSelectorAggregateFunction, you have to create function that takes two parameters (Similar to LINQ .Sum function - items and selector).

public class CheckedCountFunction : SumFunction //EnumerableSelectorAggregateFunction
{
    // inheriting SumFunction, so I don't have to create my own Merge function

    protected override string AggregateMethodName => nameof(CheckedCount);

    protected override Type ExtensionMethodsType => typeof(CheckedCountFunction);

    public override object Merge(object item, AggregateResultMergeAction mergeAction,
        AggregateResult originalAggregateResult, QueryableCollectionViewGroup group)
    {
        /*
         * DECOMPILED

        object obj = originalAggregateResult.Value;
        if (obj == null)
          return (object) null;
        Type type1 = obj.GetType();
        object aggregationValue = this.GetAggregationValue(item);
        if (aggregationValue == null)
          return obj;
        Type type2 = aggregationValue.GetType();
        Func<object, object, object> func = (Func<object, object, object>) null;
        switch (mergeAction)
        {
          case AggregateResultMergeAction.Add:
            func = ArithmeticFunctionsCache.GetAdditionFunction(type1, type2);
            break;
          case AggregateResultMergeAction.Remove:
            func = ArithmeticFunctionsCache.GetSubtractionFunction(type1, type2);
            break;
        }
        return func(obj, aggregationValue);

         */

        return base.Merge(item, mergeAction, originalAggregateResult, group);
    }

    public static int CheckedCount<TItem, TProperty>(IEnumerable<TItem> items, Func<TItem, TProperty> selector)
    {
        if (typeof(TProperty) != typeof(bool))
            throw new ArgumentException($@"You must select bool property, to use {nameof(CheckedCountFunction)}.");

        var selectedProperties = items.Where(s => Convert.ToBoolean(selector(s)));
        var count = selectedProperties.Count();
        return count;
    }
}

This function of mine for example does Count of `true` values.

I suppose, you have your problem resolved more than 10 years ago, but maybe someone might find this usefull.

[Martin Ivanov]

Hey Tomáš, thanks for sharing your solution.