import numpy as np
import gradio as gr
import pandas as pd


def homework01_solution1(K, X1, X2):

  K = int(K)

  # Verify your solutions by code
  import numpy as np
  X = np.array([[2.0, 5.0], [3.0, 4.0], [3.0, 2.0], [1.0, 3.0], [5.0, 2.0], [7.0, 1.0],
                [6.0, 3.0], [7.0, 4.0]])
  y = np.array([2.9,4.2,5.8,3.2,8.9,9.2,7.4,8.2])

  import pandas as pd
  train_data = pd.DataFrame(X, columns=['X1', 'X2'])
  train_data['Label (Y)'] = y

  from sklearn.neighbors import KNeighborsRegressor
  #(1) predict class for point (3,3) with K = 3

  neigh = KNeighborsRegressor(n_neighbors=K)
  neigh.fit(X, y)
  predicted_label = neigh.predict(np.array([[X1, X2]]))[0]
  predicted_label = np.round(predicted_label, 3)

  #(Q) calculate squared error
  pred = neigh.predict(X)
  squared_error = (pred-y)**2

  #(Q2.3) evaluate mean squared error 
  import sklearn
  y_pred = neigh.predict(X)
  mse = np.round(sklearn.metrics.mean_squared_error(y,y_pred),3)


  train_data['Predicted Label ('+str(K)+'-NN)'] = pred
  train_data['Squaredd Error'] = squared_error
  train_data['Predicted Label ('+str(K)+'-NN)'] = train_data['Predicted Label ('+str(K)+'-NN)'].round(3)
  train_data['Squaredd Error'] = train_data['Squaredd Error'].round(3)

  
    

  (nb_dist, nb_indice) = neigh.kneighbors(np.array([[X1, X2]]), K)

  import pandas as pd

  results = pd.DataFrame(columns=['Rank of closest neighbor', 'Features (X_1,X_2)', 	'Label (Y)', 	'Distance to query data'])

  for i in range(K):
    idx = nb_indice[0][i]
    fea = X[idx].tolist()
    fea = '({})'.format(', '.join(map(str, fea)))

    dist = nb_dist[0][i]
    label = y[idx]

    #print(idx, fea, dist, label)

    # Dictionary to append
    new_data = {'Rank of closest neighbor': i, 'Features (X_1,X_2)': fea, 	'Label (Y)':label , 	'Distance to query data': dist}
    tmp = pd.DataFrame(new_data, index=[0])
    # Append dictionary to DataFrame
    #data = data.append(new_data, ignore_index=True)
    results = pd.concat([results, tmp], ignore_index=True)

  results = results.sort_values(by='Rank of closest neighbor')
  results['Distance to query data'] = results['Distance to query data'].round(3)
  
  return train_data, results, predicted_label, mse




### configure inputs
set_K = gr.Number(value=7)
set_X1 = gr.Number(value=1)
set_X2 = gr.Number(value=2)

### configure outputs
set_output_traindata = gr.Dataframe(type='pandas', label ='Train Dataset')
set_output_q1a = gr.Dataframe(type='pandas', label ='Question 1: KNN-Regressor Search')
set_output_q1b = gr.Textbox(label ='Question 1: KNN-Regressor Prediction')
set_output_q3 = gr.Textbox(label ='Question 3: KNN-Regressor MSE (Training data)')

### configure Gradio
interface = gr.Interface(fn=homework01_solution1, 
                         inputs=[set_K, set_X1, set_X2], 
                         outputs=[set_output_traindata, set_output_q1a, set_output_q1b, set_output_q3],
                         
                         title="CSCI4750/5750(hw01-PartI): Mathematics for KNN (Question 1 & 3: KNN-Regressor Search)", 
                         description= "Click examples below for a quick demo",
                         theme = 'huggingface'
                         )


interface.launch(debug=True)