from os import listdir
from PIL import Image, ImageOps
import numpy

ROWS = 28 #Height of image
COLS = 28 #Width of image

'''
Reads the MNIST handwriten digits IDx3 and IDx1 files.
Returns a list of lists. every list contains the 784 pixels
(0...255)of an image (row by row) and the label of the image
(0...9).
images is the name of the IDx3 file of the images.
labels is the name of the IDx1 file of the labels.
'''
def idx2list(images,labels):
    lst = []  #initializes the returned list
    fimages = open(images,"rb")#"rb" is "read bytes"
    flabels = open(labels,"rb")
    '''
    Your code will move the cursors in flabels and fimage
    to the first "interesting" byte: a byte that contains data of the image
    '''
    
    x = fimages.read(1)#Reads the first byte of an image
    while x != b"": 
        line = [ord(x)] #Line will acummulate all the image's pixels
        '''
        Your code will read the rest of the pixels in the image
        and add them to line.
        '''
        line += [ord(flabels.read(1))]#Reads the label
        lst += [line]#Adds the image to lst
        x = fimages.read(1)

    fimages.close()
    flabels.close()
    return lst

'''
Transforms a list of data
(compatible with MINST training dataset)
to a grayscale COLSXROWS image.
'''
def data2img(lst):
    lst2d = [] #Transforms the list to a 2D list
    for i in range(ROWS):
        lst2d += [lst[COLS*i : COLS*(i+1)]]
    #Image.fromarray converts a numpy 2D array to an image
    return Image.fromarray(numpy.uint8(numpy.array(lst2d)))

'''
Displays the image representes by ds[i].
If i is too big - does nothing
'''
def display_image(ds, i):
    if i < len(ds):
        im = data2img(ds[i])
        '''
        Your code will display the image im.
        A hint: find the method (in Image) that displays an image.
        '''

'''
Returns the Euclidean distance between a and b
'''
def distance(a, b):
    dis = 0
    '''
    Your code will calculate the euclidean distance between a and b 
    '''
    return dis

'''
Returns the label of the line in dataset
that is closest to instance
'''
def nearest_neighbour(dataset, instance):
    mindis = float("inf")#mindis is the dis to the nearest neighbour
    '''
    Your code will scan the dataset and
    find the nearest neighbour to instance.
    the label of this neighbor will be returned.
    '''

'''
Test the accuracy of classification
using as dataset the first train instances
and using as a test set the last test instances
'''
def test(ds, train, test):
    training_set = ds[:train]
    count = 0
    '''
    Your code will count the number of correctly classified
    images from the last test images in the dataset
    '''
    return count / test

'''
Transforms the image in filename to a list of data
compatible with MINST training dataset.
The image is resized and converted to grayscale.
if neg is True then a negative of the is computed
'''
def img2data(filename, neg = False):
    im = Image.open( filename ) #reads the image
    im = im.convert("L")        #converts to grayscale
    im = im.resize([COLS, ROWS])#resizes the image
    '''
    Your code will produce the negative of im, if neg is True.
    Find how to convert an image into it's negative.
    Hint: ImageOps
    '''
    a = list(im.getdata())      #converts to a list of ints 0-255
    return a

'''
Classifies all the jpg files in path
using the dataset ds
Prints file name and its class.
'''
def classify_images(ds, path):
    for f in listdir(path):
        if len(f)> 4 and f[-4:].upper() == ".JPG":#the file is a jpg?
            im = img2data(path + "\\" + f , neg = True)
            print(f, nearest_neighbour(ds, im))