from math import sqrt
import matplotlib.pyplot as plt
import csv

DATA_POINT_NULL = 99
GFX_COLOR_1		= 1
GFX_COLOR_2		= 2

def load_data_user_based(path):
	prefs = {}
	i = 0;
	with open(path, 'r') as csvfile:
		spamreader = csv.reader(csvfile, delimiter=' ', quotechar='"')
		for user in spamreader:
			i +=1
			prefs.setdefault(int(i), {})
			array2d = [float(digit.replace(',', '.')) for digit in user]
			for joke in range(1, len(array2d)):
				if(float(array2d[joke]) == DATA_POINT_NULL):
					continue
				prefs[int(i)][int(joke)]=float(array2d[joke])
	return prefs

def load_data_item_based(path):
	prefs = {}
	i = 0;
	with open(path, 'r') as csvfile:
		spamreader = csv.reader(csvfile, delimiter=' ', quotechar='"')
		for user in spamreader:
			i += 1
			array2d = [float(digit.replace(',', '.')) for digit in user]
			for joke in range(1, len(array2d)):
				prefs.setdefault(int(joke) , {})
				if(float(array2d[joke]) == DATA_POINT_NULL):
					continue
				prefs[int(joke)][int(i)]=float(array2d[joke])
	return prefs


def visualize(prefs):
	x = []
	y = []
	z = []
	for user in prefs:
		for joke in prefs[user]:
			if(float(prefs[user][joke]) == DATA_POINT_NULL):
				continue
			x.append(int(user))
			y.append(int(joke))
			if(float(prefs[user][joke]) < 0):
				z.append(GFX_COLOR_1)
			else:
				z.append(GFX_COLOR_2)

	plt.xlim([0,100])
	plt.ylim([0,100])
	plt.scatter(x, y, c = z)
	plt.show()

#homer
def sim_distance_1(prefs, person1, person2):
	X = set(prefs[int(person1)])
	Y = set(prefs[int(person2)])
	X_len = len(X)
	Y_len = len(Y)

	min_len = min(X_len, Y_len)
	Z = list(X & Y)
	if DATA_POINT_NULL in Z:
		Z.remove(DATA_POINT_NULL)
	Z_len = len(Z)

	return (float(Z_len) / float(min_len))

def sim_distance_2(prefs, person1, person2):
	si = {}

	for item in prefs[int(person1)]:
		if item in prefs[int(person2)]:
			si[item] = 1

	if len(si) == 0:
		return 0

	sum_of_squares = sum([pow(prefs[int(person1)][int(item)]-prefs[int(person2)][int(item)],2)
						for item in prefs[int(person1)] if item in prefs[int(person2)]])

	return 1/(1+sum_of_squares)

def topMatches(prefs_learn_data, person, object_id, k=5, similarity=sim_distance_1):

	prefs_with_object_id = {}
	for user in prefs_learn_data:
		if object_id in prefs_learn_data[int(user)]:
			prefs_with_object_id[int(user)] = prefs_learn_data[int(user)]

	prefs_with_object_id[int(person)] = prefs_learn_data[int(person)]

	scores = [(similarity(prefs_with_object_id, person, other), prefs_learn_data[int(other)][int(object_id)], other)
						for other in prefs_with_object_id if other!=person]
	scores.sort()
	scores.reverse()

	result_scores = [score for score in scores if score[0] > 0]
	return result_scores[0:k]

def get_rating(prefs_learn_data, person, object_id, similarity=sim_distance_1):
	scores = topMatches(prefs_learn_data, person, object_id, similarity=similarity)
	if len(scores) == 0: return 0
	sum_sim_score = sum(score[0]*score[1] for score in scores)
	sum_sims = sum(score[0] for score in scores)
	rating = sum_sim_score/sum_sims
	return rating

def calculate_error(rating_real, rating_predict):
	sum = 0
	for i in range(len(rating_real)-1):
		sum += pow(rating_real[i]-rating_predict[i], 2)
	return sqrt(sum)

def user_based_party(visualize = False, sim = sim_distance_1):
	learn_data = load_data_user_based(path='jester-data-small-supper-80.csv')
	test_data  = load_data_user_based(path='jester-data-small-supper-20.csv')

	if visualize == True:
		visualize(learn_data)

	rating_real = []
	rating_predict = []
	i = 0;
	for user in test_data:
		i += 1
		print i
		for joke in test_data[user]:
			rating_real.append(test_data[user][joke])
			rating_predict.append(get_rating(learn_data, int(user), int(joke), similarity = sim))

	print "========== USER BASED PARTY GOES HERE =========="
	print rating_predict
	print rating_real
	print calculate_error(rating_real, rating_predict)

def item_based_party(visualize = False, sim = sim_distance_1):
	learn_data = load_data_item_based(path='jester-data-small-supper-80.csv')
	test_data  = load_data_item_based(path='jester-data-small-supper-20.csv')

	if visualize == True:
		visualize(learn_data)

	rating_real = []
	rating_predict = []

	i = 0

	for user in test_data:
		i += 1
		print i
		for joke in test_data[int(user)]:
			rating_real.append(test_data[user][joke])
			rating_predict.append(get_rating(learn_data, int(joke), int(user), similarity = sim))

	print "========== ITEM BASED PARTY GOES HERE =========="
	print rating_predict
	print rating_real
	print calculate_error(rating_real, rating_predict)

def main():
	user_based_party(visualize = False, sim = sim_distance_1)
	item_based_party(visualize = False, sim = sim_distance_1)


main()