using System;
using System.Data;
using System.Linq;
using System.Collections.Generic;
using System.Reflection;
using System.ComponentModel;

namespace PricePrediction.MachineLearning
{
    /// <summary>
    /// Standardize features by removing the mean and scaling to unit variance.
    /// more : https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html
    /// </summary>
    public class StandardScaler<T> where T : new()
    {
        private List<double> _mean;
        private List<double> _standardDeviation;

        /// <summary>
        /// fit then transform
        /// </summary>
        /// <param name="dt"></param>
        /// <returns></returns>
        public List<T> FitTransform(List<T> listOfObjects)
        {
            return Fit(listOfObjects).Transform(listOfObjects);
        }

        /// <summary>
        /// Reset then 
        /// </summary>
        /// <param name="dt"></param>
        /// <returns></returns>
        public StandardScaler<T> Fit(List<T> listOfObjects)
        {
            _mean = new();
            _standardDeviation = new();

            if (listOfObjects.Count < 1)
                throw new Exception("no data");

            var dt = listOfObjects.ToArraysOfColumns<T>();

            for (int i = 0; i < dt.Length; i++)
            {
                _mean.Add(dt[i].Average());
                _standardDeviation.Add(Calculations.StandardDeviation(dt[i]));
            }

            return this;
        }

        /// <summary>
        /// Get 
        /// </summary>
        /// <param name="dt"></param>
        /// <returns></returns>
        public List<T> Transform(List<T> listOfObjects)
        {
            if (_mean == null)
                throw new Exception("This StandardScaler instance is not fitted yet. Call 'Fit' with appropriate arguments before using this estimator.");

            //if (dt.Columns.Count != _mean.Count)
            //    throw new Exception("number of fitted columns not same as current one");

            var dt = listOfObjects.ToArraysOfColumns<T>();

            for (int c = 0; c < dt.Length; c++)
                for (int r = 0; r < dt[c].Length; r++)
                    dt[c][r] = (dt[c][r] - _mean[c]) / _standardDeviation[c];

            return ToListOfObject(dt);
        }

        private static List<T> ToListOfObject(double[][] arr)
        {
            var res = new List<T>();

            PropertyDescriptorCollection properties = TypeDescriptor.GetProperties(typeof(T));
            var ObjectsCount = arr[0].Length;

            for (int i = 0; i < ObjectsCount; i++)
            {
                T o = new();
                for (int j = 0; j < properties.Count; j++)
                    properties[j].SetValue(o, arr[j][i]);
                res.Add(o);
            }
            return res;

        }

    }
}