#!/usr/bin/env python
# coding: utf-8

# Let's try to build a chloropleth map using Bokeh, inspired by [this example](http://bokeh.pydata.org/en/0.10.0/docs/gallery/choropleth.html). Let's get our imports going:

# In[1]:


from bokeh.sampledata import us_states, us_counties, unemployment
from bokeh.plotting import figure, show, output_file, output_notebook


# In[2]:


output_notebook()


# It should be noted we're only importing three things from bokeh.plotting: figure, show and output_file.

# In[3]:


help(figure)


# In[4]:


help(show)


# In[5]:


help(output_file)


# We can now go through the example steps: first, let's load the data.

# In[6]:


us_states = us_states.data.copy()
us_counties = us_counties.data.copy()
unemployment = unemployment.data


# In[7]:


type(us_states)


# In[8]:


type(us_counties)


# In[9]:


type(unemployment)


# We now delete Alaska and Hawai from the data and load the coordinates for each state in separate variables.

# In[10]:


del us_states["HI"]
del us_states["AK"]

state_xs = [us_states[code]["lons"] for code in us_states]
state_ys = [us_states[code]["lats"] for code in us_states]

county_xs=[us_counties[code]["lons"] for code in us_counties if us_counties[code]["state"] not in ["ak", "hi", "pr", "gu", "vi", "mp", "as"]]
county_ys=[us_counties[code]["lats"] for code in us_counties if us_counties[code]["state"] not in ["ak", "hi", "pr", "gu", "vi", "mp", "as"]]


# To have a better look at what we're manipulating, we can just plot the coordinates using matplotlib.

# In[11]:


import matplotlib.pyplot as plt
get_ipython().run_line_magic('matplotlib', 'inline')


# In[12]:


plt.plot(county_xs[0], county_ys[0])


# We now go over the data for each county and assign it a color, depending on the numerical value of the unemployment.

# In[13]:


colors = ["#F1EEF6", "#D4B9DA", "#C994C7", "#DF65B0", "#DD1C77", "#980043"]

county_colors = []
for county_id in us_counties:
    if us_counties[county_id]["state"] in ["ak", "hi", "pr", "gu", "vi", "mp", "as"]:
        continue
    try:
        rate = unemployment[county_id]
        idx = min(int(rate/2), 5)
        county_colors.append(colors[idx])
    except KeyError:
        county_colors.append("black")


# Finally, a figure is created and patches are added to the map. Filled patches are used, as well as simpler patches that only map the contour of the states.

# In[14]:


p = figure(title="US Unemployment 2009", toolbar_location="left",
    plot_width=1100, plot_height=700)

p.patches(county_xs, county_ys, fill_color=county_colors, fill_alpha=0.7,
    line_color="white", line_width=0.5)
p.patches(state_xs, state_ys, fill_alpha=0.0,
    line_color="#884444", line_width=2)

show(p)


# With this knowledge in our hands, can we create a similar looking map of France?
# 
# The raw data can be found here: [http://bl.ocks.org/rveciana/02eb5b83848e0b06fa8e](http://bl.ocks.org/rveciana/02eb5b83848e0b06fa8e).
# 
# The instructions for using geopandas here: [https://michelleful.github.io/code-blog/2015/04/27/osm-data/](https://michelleful.github.io/code-blog/2015/04/27/osm-data/).

# In[15]:


import pandas as pd


# In[16]:


import geopandas as gpd


# In[17]:


import urllib


# In[18]:


json_file = urllib.request.urlopen("https://cdn.rawgit.com/rveciana/5919944/raw/b1f826319231c3e06d6e8548bc947ca2c29dc9e8/france.json").readlines()


# In[20]:


import tempfile


# In[38]:


f = tempfile.NamedTemporaryFile(dir='files/', delete=False)
try:
    f.write(json_file[0])
    f.close()
    df = gpd.read_file(f.name)
finally:
    
    os.remove(f.name)


# In[39]:


df.columns


# In[40]:


df.ix[0]['geometry']


# In[41]:


df['centroids_lon'] = [g.centroid.coords.xy[0][0] for g in df.geometry]
df['centroids_lat'] = [g.centroid.coords.xy[1][0] for g in df.geometry]


# In[42]:


df


# Let's filter the map items based on the centroid coordinates.

# In[46]:


metropolitan_france = (df.centroids_lat > 40) & (df.centroids_lon > -10)
df[metropolitan_france].plot()


# Nice! Let's try to do the same thing using bokeh! What we need is a way to extract the polygon coordinates from the geopandas dataframe and plot them using Bokeh.

# In[65]:


s = df[metropolitan_france].geometry


# In[79]:


from ipywidgets import interact
from IPython.display import display

@interact
def disp_geom(ind=(0, s.shape[0] - 1)):
    display(s.iloc[ind])
    print(s.iloc[ind].type)


# In[80]:


o = s.iloc[43]


# In[81]:


o


# Reading the [shapely documentation](http://toblerity.org/shapely/manual.html#polygons), I found out that the coordinates of a polygon can be accessed in the following way:

# In[84]:


len(o.geoms)


# In[82]:


plt.plot(o.exterior.coords.xy[0], o.exterior.coords.xy[1])


# In[87]:


plt.plot(o.convex_hull.exterior.coords.xy[0], o.convex_hull.exterior.coords.xy[1])


# We are now set: let's build a Bokeh figure using this!

# In[95]:


import numpy as np


# In[97]:


dept_xs = [np.array(row.convex_hull.exterior.coords.xy[0]) for row in df[metropolitan_france].geometry]
dept_ys = [np.array(row.convex_hull.exterior.coords.xy[1]) for row in df[metropolitan_france].geometry]


# In[98]:


p = figure(title="France map from geojson data", toolbar_location="left",
    plot_width=1100, plot_height=700)

p.patches(dept_xs, dept_ys, fill_color="#F1EEF6", fill_alpha=0.7,
    line_color="white", line_width=0.5)

show(p)


# In[ ]: