# References
# The Onion article on the Grandito
# 	https://www.theonion.com/taco-bells-five-ingredients-combined-in-totally-new-way-1819564909
# Ingredients of the Taco Bell menu
# 	https://www.flashcardmachine.com/taco-bellingredients.html
# Info on Taco Bell scoop and proportion sizes
# 	https://quizlet.com/119643026/taco-bell-portioning-tool-test-flash-cards/


# Imports
from inspect import getsourcefile # For getting this source file's directory
from os.path import abspath       # For getting this source file's directory
import os
import pandas as pd
import numpy as np


# Construct path and filenames for input and output data
ThisFile = abspath(getsourcefile(lambda:0))
ThisDirectory = ThisFile.replace('EigenGrandito.py', '')
SourceCsv = ThisDirectory + 'MenuIngredients.csv'

def Main():
	#-----------------------------
	# Load menu data
	Source_df = pd.read_csv(SourceCsv, header=0, index_col=0)
	Source_df = Source_df.fillna(0)                  # replace NaN values with 0
	UnitsList = Source_df.loc[:,'Units'].to_numpy()  # save the units list, for later output
	Source_df.drop('Units', axis=1, inplace=True)    # remove unneeded columns


	#-----------------------------
	# Show A and compute the SVD
	A = Source_df
	print("\nA= ")
	print(A.shape)
	print(A)
	U, E, Vt = np.linalg.svd(A,full_matrices=True)


	#-----------------------------
	# negate U and V, because otherwise PC1 is all negative
	# and it doesn't make sense for the most dominant component to 
	# be made up of negative amounts of ingredients.
	U = U*-1
	Vt = Vt*-1


	#-----------------------------
	# Optional: Analyze the singular values to determine how many principle 
	# components would be needed to reconstruct A to a desired accuracy.
	if False:
		print("\nSingular Values = ")
		print(np.round(E, decimals=2))
		print("\nSingular Values as percentages= ")
		E_percents = (E/np.sum(E))*100.0
		print(np.round(E_percents, decimals=2))

		iNumSingularValues = 0;
		TotalPercent = 0.0
		TargetPercent = 80;
		while TotalPercent < TargetPercent and iNumSingularValues < len(E_percents):
			TotalPercent += E_percents[iNumSingularValues]
			iNumSingularValues += 1;

		print("\nNum Principle Components needed to recreate {:.2f} percent of A: {:d}".format(TargetPercent, iNumSingularValues))


	#-----------------------------
	# Print U, E, Vt, and UEVt
	print("\nU = ")
	print(U.shape)
	print(np.round(U, decimals=2))

	print("\nE = ")
	E = np.diag(E)
	print(E.shape)
	print(np.round(E, decimals=2))

	print("\nVt = ")
	print(Vt.shape)
	print(np.round(Vt, decimals=2))

	print("\nUEVt =")
	U = U[:,0:44] # remove the last columns of U
	UEVt = U @ (E @ Vt)
	print(np.round(UEVt, decimals=2))



	#-----------------------------
	# prepare the grandito recipe

	# first principle component * first singular value
	Grandito = U[:,0] * E[0,0]           

	# Assume we want at least a 1.0 amount of the 6.5 in flour tortilla
	# and scale everything accordingly. 
	ScalingFactor = 1.0/Grandito[0]      
	Grandito = Grandito * ScalingFactor  

	# if the ingredient amount is too low, assume it's 0
	for i, val in enumerate(Grandito):   
		if val < 0.1:
			Grandito[i] = 0

	# round to the nearest 10th
	Grandito = np.round(Grandito, decimals=1) 

	# Convert the numerical values into a string, so we 
	# can replace 0 values with a character, for better readibility
	strGrandito = []                      
	for i, val in enumerate(Grandito):
		if Grandito[i] == 0:
			strGrandito.append("-")
		else:
			strGrandito.append(str(Grandito[i]))
	


	#-----------------------------
	# print the Eigen Grandito recipe
	print("\n\nBEEP BOOP BEEP BOOP...Your Eigen Grandito Recipe is:")
	Answer_df = pd.DataFrame(index=Source_df.index.copy())
	Answer_df.insert(0, "", UnitsList, allow_duplicates=True)
	Answer_df.insert(1, "", strGrandito, allow_duplicates=True)
	print(Answer_df)


	#-----------------------------
	# Optional: Export the tables to CSVs
	if False:
		Vt = np.round(Vt, decimals=5)
		U = np.round(U, decimals=5)

		Out_df = pd.DataFrame(U)
		Out_df.to_csv(ThisDirectory + 'MenuIngredients_U.csv', index=False)

		Out_df = pd.DataFrame(Vt)
		Out_df.to_csv(ThisDirectory + 'MenuIngredients_Vt.csv', index=False)



Main()