using System;

using System.Collections.Generic;

using System.Linq;

using System.Text;

namespace NumSharp.Core.Extensions

{

public static partial class NDArrayExtensions

{

public static NDArray mean(this NDArray np, int axis = -1)

{

var mean = new NDArray(typeof(double));

mean.Storage.SetData(new double[0]);

// axis == -1: DEFAULT; to compute the mean of the flattened array.

if (axis == -1)

{

var data = np.Storage.GetData();

double sum = 0;

for (int idx =0; idx < data.Length;idx++)

sum += Convert.ToDouble(data.GetValue(idx));

mean.Storage.SetData(new double[] { sum / np.size});

}

// to compute mean by compressing row and row

else if (axis == 0)

{

double[] sumVec = new double[np.shape[0]];

for (int d = 0; d < sumVec.Length; d++)

{

for (int p = 0; p < np.shape[1]; p++)

{

sumVec[p] += Convert.ToDouble(np[d,p]);

}

}

var puffer = mean.Storage.CloneData<double>().ToList();

for (int d = 0; d < np.shape[1]; d++)

{

puffer.Add(sumVec[d] / np.shape[0]);

}

mean.Storage.SetData(puffer.ToArray());

mean.Storage.Reshape(mean.Storage.GetData().Length);

}

else if (axis == 1)

{

var puffer = mean.Storage.GetData<double>().ToList();

for (int d = 0; d < np.shape[0]; d++)

{

double rowSum = 0;

for (int p = 0; p < np.shape[1]; p++)

{

rowSum += Convert.ToDouble(np[d,p]);

}

puffer.Add(rowSum / np.shape[1]);

}

mean.Storage.SetData(puffer.ToArray());

mean.Storage.Reshape(mean.Storage.GetData().Length);

}

return mean;

}

}

}