import pandas as pd
import numpy as np
import matplotlib.pyplot as plt


# Make the graphs a bit prettier, and bigger
pd.set_option('display.mpl_style', 'default') 
pd.set_option('display.line_width', 5000) 
pd.set_option('display.max_columns', 60) 

## Read sample data set and convert string dates to datetimes
bydatetime_df = pd.read_csv('data/bydate_shortstay_csv.csv',parse_dates=['datetime'])

bydatetime_df.head()

meanocc_df = bydatetime_df.groupby(['category','dayofweek'])['occupancy'].mean()

meanocc_df

meanoccdate = bydatetime_df.groupby(['category',bydatetime_df['datetime'].map(lambda x: x.date())])['occupancy'].mean()

meanoccdate.head()

meanoccdate.index[0]

meanoccdate.index[0][1]

ts = pd.Timestamp('6/26/2014')

ts.strftime('%a')

meanoccdate_df = pd.DataFrame(meanoccdate)

meanoccdate_df['dayofweek'] = meanoccdate_df.index.map(lambda x: pd.Timestamp(x[1]).strftime('%a'))

meanoccdate_df.head()

meanoccdate_df.tail()

meanoccdate_df['occupancy']['Total']

# Load the extension
%load_ext rmagic

%matplotlib inline

# Create some data in Python and scatter it

import pylab
X = np.array([0,1,2,3,4])
Y = np.array([3,5,4,6,7])
pylab.scatter(X, Y)

# "Push" these two numpy arrays into the R "space"
%Rpush X Y

%%R
linmodel <- lm(Y~X)
print(summary(linmodel))

# normed=1 plots probs instead of counts, alpha in [0,1] is transparency level (RGBA colors)
plt.hist(meanoccdate_df['occupancy']['Total'], 20, normed=1, facecolor='green', alpha=0.75) 
plt.xlabel('Occupancy')
plt.ylabel('Probability')
plt.title(r'Histogram of Short Stay Occupancy')
plt.grid(True)
plt.show()

%Rpush meanoccdate_df
%R str(meanoccdate_df)

meanoccdate_df['patient_type'] = meanoccdate_df.index.map(lambda x: x[0])
meanoccdate_df['date'] = meanoccdate_df.index.map(lambda x: x[1])

%Rpush meanoccdate_df
%R str(meanoccdate_df)

%R library(ggplot2)

%%R
g <- ggplot(data=meanoccdate_df[meanoccdate_df$patient_type == 'Total',]) + geom_histogram(aes(x=occupancy, y=..density..), fill="#FF9999", colour="black")
print(g)

%%R
g2 <- ggplot(data=meanoccdate_df) + geom_histogram(aes(x=occupancy, y=..density..), binwidth=2, fill="#FF9999", colour="black")
print(g2 + facet_grid(patient_type ~ dayofweek))

%R str(meanoccdate_df)

%%R
# Create vector with DOWs ordered as you wish
DOW_order <- c("Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat")

# Change DOW from factor to ordered factor using vector you just made
meanoccdate_df$dayofweek <- factor(meanoccdate_df$dayofweek,levels=DOW_order,ordered=TRUE)

%R str(meanoccdate_df)

%%R
g2 <- ggplot(data=meanoccdate_df) + geom_histogram(aes(x=occupancy, y=..density..), binwidth=2, fill="#FF9999", colour="black")
print(g2 + facet_grid(patient_type ~ dayofweek))