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

# #Graph Analysis of Domains Referring to Windeln.de

# This notebook analyses the domains that are linking to Windeln.de and attempting to help understand the relationship between those websites.

# In[13]:


import os
import re
from pandas import *
import numpy as np
import networkx as nx
import matplotlib.pyplot as plt
from mpld3 import display

get_ipython().run_line_magic('matplotlib', 'inline')


# ###Read Raw Data
# Data produced by backlinks.py

# In[17]:


df = read_csv('C:/workspace/analysis_reporting/joe/milk_china/anchor_urls.csv', sep='\t')
# df.head()


# In[15]:


df_domain = df.ix[:, ['source_domain','alexa_cc_rank']].drop_duplicates()


# Find relationship between source domains

# In[16]:


relation_pairs = []
key = ['source_domain','links']
df2 = df.ix[:,key].assign(num_links=1).dropna().groupby(key, as_index=0)['num_links'].sum()
# df2 = df.ix[:,['source_domain','links']].drop_duplicates().dropna()
df2.index = range(len(df2))
for i in df2.index:
    source_domain = df2.ix[i, 'source_domain']
    num_links = df2.ix[i, 'num_links']
    for link_domain in df2.ix[i, 'links'].split('|'):
        
        pair = (source_domain, link_domain, num_links)
        
        relation_pairs.append(pair)
        
df_pair = DataFrame(list(set(relation_pairs)), columns=['src', 'target', 'num_links'])


# In[20]:


# plotting graph
G = nx.Graph()
pos=nx.spring_layout(G)
for line in df_pair.values:
    src, tgt, wt = line.tolist()
    G.add_edge(src, tgt, weight=wt)

# fig, ax = plt.subplots()
plt.figure(figsize=(10, 10))
nx.draw_networkx(G, with_labels=1)

display()