def build_friends_edgelist():
"""
Build friends graph
from full habrauser DB and
store it as edgelist (easy readable from igraph).
"""
users_dir = ha.get_users_dir()
user_files_list = os.listdir(users_dir)
user_files_list.sort()
login2id = {}
for userid, filename in enumerate(user_files_list):
login = filename.split(".")[0].replace("@", "")
login2id[login] = userid
edgefile = open(os.path.join(ha.get_graph_dir(), "friends.edgelist"), "w")
for userid, filename in enumerate(user_files_list):
user = so.SmartObject(ut.pickle2data(os.path.join(users_dir, filename)))
login = filename.split(".")[0].replace("@", "")
if user.friends:
for friend in user.friends:
if friend in login2id:
friend_id = login2id[friend]
edgefile.write("%d %d\n" % (userid, friend_id))
edgefile.close()
def analyze_friends():
# pylint: disable=E1101
sourcefile = os.path.join(ha.get_graph_dir(), 'friends.edgelist')
thegraph = Graph.Read_Edgelist(sourcefile, directed=False)
print "Graph loaded"
print thegraph
# dd = thegraph.degree_distribution() /??
degs = thegraph.vs.degree()
print numpy.median(degs)
print numpy.mean(degs)
hist = Histogram(bin_width=2)
hist << degs
plot(hist, os.path.join(ha.get_reports_dir(), "habrafriends-degrees-distribution.svg"))
def cut_thread(postpath):
"""
Build a graph matrix, set MAX-CUT problem
"""
thepostsdir = os.path.join(ha.get_data_dir(), 'posts')
filename = os.path.join(thepostsdir, postpath) + '.dat'
thread_tree = ut.pickle2data(filename)
def walk_tree_for_login2id(subtree, login2id):
for login in subtree:
if login not in login2id:
login2id[login] = len(login2id)
walk_tree_for_login2id(subtree[login], login2id)
login2id = {}
walk_tree_for_login2id(thread_tree, login2id)
N = len(login2id)
weights = np.zeros((N, N), dtype=np.int16)
def walk_tree_for_weights(root_login, subtree, weights):
u = login2id[root_login]
for login in subtree:
v = login2id[login]
if u != v:
weights[u, v] += 1
weights[v, u] += 1
walk_tree_for_weights(login, subtree[login], weights)
for root_login in thread_tree:
walk_tree_for_weights(root_login, thread_tree[root_login], weights)
id2login = {}
for login in login2id:
id2login[login2id[login]] = login
y = greedy_max_cut(weights)
def print_habrauser(uid):
login = id2login[uid]
print '* [http://' + login + '.habrahabr.ru ' + login + ']'
print "Analysis of http://habrahabr.ru/" + postpath
print "----"
print "Party 1"
for i in xrange(N):
if y[i] > 0:
print_habrauser(i)
print "----"
print "Party 2"
for i in xrange(N):
if y[i] < 0:
print_habrauser(i)
print "----"
def get_comment_tree(postpath):
url_to_parse = 'http://habrahabr.ru/' + postpath
root = ut.doc4url(url_to_parse)
if not root:
return None
author = root.xpath('//div[@class="author"]/a')[0].text
print author
comment_root_tree = {}
## Словарь вложенных словарей
## автор
## |→ автор_комментарария
## |→ автор подкомментария
def dfs_process(node, tree):
"""
Рекурсивно идет вглубь от node
и набивает словарь-дерево tree
"""
print node.get('id')
comments = node.xpath('.//div[@id="comments" or @class="reply_comments"]')[0]
for comment in comments.xpath('./div[@class="comment_item"]'):
author = comment.xpath('.//a[@class="username"]')[0].text
print author
child_tree = {}
dfs_process(comment, child_tree)
tree[author] = deepcopy(child_tree)
dfs_process(root, comment_root_tree)
comment_tree = {author: comment_root_tree}
print 'tree:', comment_tree
thepostsdir = os.path.join(ha.get_data_dir(), 'posts')
filename = os.path.join(thepostsdir, postpath) + '.dat'
ut.createdir(os.path.split(filename)[0])
ut.data2pickle(comment_tree, filename)