Porting Excel to .NET

Trevor Misfeldt — Tue, 18 Feb 2014 19:11:24 +0000

In previous blog posts, we demonstrated calling NMath from within Excel (C#, Visual Basic). Another common use case is replacing an Excel spreadsheet with an equivalent .NET application. Today, we are releasing .NET code to make this task much easier.

We have created a library of Excel extensions for NMath that work just like the built-in Excel mathematical and statistical functions. They work with arrays in a similar fashion to a range of cells in Excel. Errors have been replaced with exceptions. Return values match as closely as possible.

For example, if you were using the ZTest function in Excel like so:

This is how you would do the same thing in .NET with NMath:

 
      using CenterSpace.NMath.Excel;

      double[] array = { 4, 5, 6, 7, 8, 3, 4 };
      ExcelStats.ZTest( array, 4.5 );

That’s possible because we have code like this to make NMath work just like Excel:

   
    public static double ZTest( double[] array, double mu, double sigma = Double.NaN )
    {
      if ( array.Length == 0 )
      {
        throw new ExcelNumException( "array must not be empty" );
      }
      sigma = Double.IsNaN( sigma ) ? StatsFunctions.StandardDeviation( array, BiasType.Unbiased ) : sigma;
      var zTest = new OneSampleZTest( array, mu, sigma );
      return zTest.P / 2.0;
    }

We have tested these functions against Excel directly and verified the results.

Here we provide an equivalent function to the MInverse (matrix inverse) function in Excel:

   
using CenterSpace.NMath.Excel;

    public static double[,] MInverse( double[,] array )
    {
      if ( array.GetLength( 0 ) != array.GetLength( 1 ) )
      {
        throw new ExcelValueException( "array must be square" );
      }
      var matrix = new DoubleMatrix( array );
      var factorization = new DoubleLUFact( matrix );
      return factorization.Inverse().ToArray();
    }

In the case where you have a user-defined function that calls other built-in functions, you can replace your code easily. For example, a user-defined function as follows:

Function CV(numbers As Range)
  mean = Application.WorksheetFunction.Average(numbers)
  standarddeviation = Application.WorksheetFunction.StDev(numbers)
  CV = (standarddeviation / mean) * 100
End Function

would look as follows:

    private static double CV( double[] a )
    {
      var mean = ExcelStats.Average( a );
      var stddev = ExcelStats.StDev( a );
      return ( stddev / mean ) * 100.0;
    }

To use this functionality:

Download the free trial versions of our NMath library. Trial versions are distributed in distributed in binary form only for a 30-day evaluation period.
Download the Excel extension library here. It comes with source code, so you can see exactly how Excel functions are implement in NMath. (In a future release, this functionality will be built into NMath.)

Note: The Excel extensions are designed as static methods, allowing a one-to-one mapping from Excel to NMath. In some cases, such as within loops, greater performance can be achieved by maintaining NMath state between calls.

– Trevor