import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from dateutil.parser import parse
from datetime import datetime, date, time


# df = pd.read_csv("data/Data.csv")
df.Date = pd.to_datetime(df.Date,utc='GMT')
#df = df.set_index('GMT')
print df.head()
df.tail()

# create a column which is only the month, day and hour of day. 
# This is used to find the average for each respective hour later
df['hour'] = df['Date'].apply(lambda t: "%d-%d-%d" % (t.month, t.day,t.hour))
#find average for each hour... This makes sense for all the columns except precip
g = df.groupby('hour').mean().reset_index()
# g = df.groupby('hour').agg({'Temp' : np.mean,'H_Pcnt' : np.mean,'C_Pcnt' : np.mean,'Temp_Feels' : np.mean,'Precip' : np.sum,'Radiation' : np.mean,'Wind' : np.mean}).reset_index()
g

# Do a left merge with the original dataset
new_df = pd.merge(left=df, right=g, on='hour', suffixes=('','_avg') )
new_df = new_df.set_index('Date')
new_df.head()

# I don't need the dates before 2009 so I remove them here.
# I kept them till now in order to calculate the averages
test = []

def drop_dates(indexes):
    for index in indexes:
        if index < datetime(2009,5,1):
            test.append(index)
    return test
trimmed = new_df.drop(drop_dates(new_df.index)).sort_index().drop('hour',axis=1)
#trimmed = trimmed.drop('date', 1)
print trimmed.head()

trimmed.Precip.plot()

#pd.write_csv("data/aggedWeather.csv")
trimmed.to_csv("aggedWeather.csv")



# print trimmed.columns
# trimmed.drop()






# grouped = df.groupby("Date")
# #grouped['Temp'].agg([np.sum, np.mean, np.std])
# agged = grouped.agg({'Temp' : np.mean,
#              'Humidity_Pcnt' : np.mean,
#              'Clouds_Pcnt' : np.mean,
#              'Temp_Feels' : np.mean,
#              'Precip' : np.sum,
#              'Radiation' : np.mean,
#              'Wind' : np.mean})

# agged = np.round(agged,1)
# agged['Year'] = "holder"
# agged['Month'] = "holder"
# agged['Day'] = "holder"
# agged['Weekday'] = "holder"

# #slimmed = agged.drop(lambda x: "2004" is x)
# #slimmed

# dates = []
# def dropDates():
#     for dateString in agged.index:
#         date = parse(dateString)
#         if date < parse("2009-05-01"): 
#             dates.append(dateString)
#         else:
#             agged.Year[dateString] = date.year
#             agged.Month[dateString] = date.month
#             agged.Weekday[dateString] = date.weekday()
#             agged.Day[dateString] = date.day
            
            



# dropDates()
# agged = agged.drop(dates)

# agged.head()