import pandas as pd
import matplotlib.pyplot as plt

# 'data_original.csv' is the raw data downloaded from
# https://www.kaggle.com/hugodarwood/epirecipes/

data = pd.read_csv('data_original.csv')

# - - - remove rows with missing nutrition values - - -

data = data[~data['calories'].isnull()]
data = data[~data['protein'].isnull()]
data = data[~data['fat'].isnull()]
data = data[~data['sodium'].isnull()]

# - - plotting nutrition values - -

'''

plt.rcParams.update({'font.size': 14})

plt.subplot(2,2,1)
plt.title('calories')
plt.ylabel('value')
plt.xticks([])
plt.plot(data['calories'])

plt.subplot(2,2,2)
plt.title('protein')
plt.ylabel('value')
plt.xticks([])
plt.plot(data['protein'])

plt.subplot(2,2,3)
plt.title('fat')
plt.ylabel('value')
plt.xticks([])
plt.plot(data['fat'])

plt.subplot(2,2,4)
plt.title('sodium')
plt.ylabel('value')
plt.xticks([])
plt.plot(data['sodium'])

# plt.show() # uncomment to show the plot

'''

# - - plot rows with extreme outliers - -

'''

# - calories -

count = []
xticks = []

for i in range(1,10):
    count.append(sum(data['calories'] >= i*1000))
    xticks.append(i*1000)

plt.subplot(2,2,1)
plt.title('recipes with calories >= threshold')
plt.ylabel('count')
plt.xlabel('threshold')
plt.xticks(range(0,9),xticks)
plt.plot(count, '--bo')

ax = plt.gca()
ax.set_ylim([0,None])

# - protein -

count = []
xticks = []

for i in range(1,10):
    count.append(sum(data['protein'] >= i*1000))
    xticks.append(i*1000)

plt.subplot(2,2,2)
plt.title('recipes with protein >= threshold')
plt.ylabel('count')
plt.xlabel('threshold')
plt.xticks(range(0,9),xticks)
plt.plot(count, '--bo')

# - fat -

count = []
xticks = []

for i in range(1,10):
    count.append(sum(data['fat'] >= i*1000))
    xticks.append(i*1000)

plt.subplot(2,2,3)
plt.title('recipes with fat >= threshold')
plt.ylabel('count')
plt.xlabel('threshold')
plt.xticks(range(0,9),xticks)
plt.plot(count, '--bo')

# - sodium -

count = []
xticks = []

for i in range(1,10):
    count.append(sum(data['sodium'] >= i*1000))
    xticks.append(i*1000)

plt.subplot(2,2,4)
plt.title('recipes with sodium >= threshold')
plt.ylabel('count')
plt.xlabel('threshold')
plt.xticks(range(0,9),xticks)
plt.plot(count, '--bo')

plt.show()

ax = plt.gca()
ax.set_ylim([0,None])

'''

# - - remove rows with exreme outliers - -

data = data[data['calories'] < 5000]
data = data[data['sodium'] < 5000]

sum(data['protein']>600) # result = 0
sum(data['fat']>600) # result = 0


# - - trim data based on breakfast, lunch, dinner labels - -

label_sum = (data['breakfast'] + data['lunch'] + data['dinner'])
data = data[(label_sum > 0) & (label_sum < 2)]

len(data) # 2460
sum(data['breakfast']==1) + sum(data['lunch']==1) + sum(data['dinner']==1) #2460


# - - drop columns with zero variation - -

zero_var = []

for i in range(2,len(data.columns)):
    if data.iloc[:,i].var() == 0:
        zero_var.append(data.iloc[:,i].name)

for col in zero_var:
    data = data.drop(columns=col)

# also drop the brunch column
data = data.drop(columns="brunch")

data.to_csv('bld.csv', index=False) # save a copy for next step
# index = False leave out the auto generated "unamed 0" column