def get(self):
page = requests.get(SITE + self.url,headers=HEAD,proxies=proxies)
dom = etree.HTML(page.content)
session = DB_Session()
# ugly code
people = {}
people['name'] = self.url
people['bio'] = (dom.xpath("//span[@class='bio']/@title") or " ")[0].encode("utf-8")
people['location'] = (dom.xpath("//span[@class='location item']/@title") or " ")[0].encode("utf-8")
people['business'] = (dom.xpath("//span[@class='business item']/@title") or " ")[0].encode("utf-8")
people['education'] = (dom.xpath("//span[@class='education item']/@title") or " ")[0].encode("utf-8")
session.execute(User.__table__.insert(), people)
session.commit()
session.close()
print page.status_code ,"got url %s !" %self.url
return set(re.findall(PEOPLE, page.content)+re.findall(QUESTION, page.content))
def partition(pages):
user_dct=defaultdict(list)
for article , ukey_lst in pages:
session = DB_Session()
session.execute(
Comment.__table__.insert(),
[{'article':article, 'user':ukey} for ukey in ukey_lst]
)
session.commit()
session.close()
for ukey in ukey_lst:
try:
user_dct[ukey].append(article)
except :
user_dct[ukey]=[article]
finally:pass
return user_dct
def main():
urls = json.load(open("urls.txt",'r'))
page_set = pool.imap(foo_map, urls)|where(lambda x:x)
ukeys = partition(page_set)
'''
with open("ukeys.txt",'w') as fuk:
json.dump(ukeys,fuk,indent=4)
'''
print "finish articles !"
users = pool.map(foo_reduce,ukeys.iterkeys())| where(lambda x :x)|as_tuple
users_lst = chunks(users,1000)
session = DB_Session()
for us in users_lst:
session.execute(
User.__table__.insert(),
us
#pool.map(foo_reduce,ukeys.iterkeys())
)
session.commit()
session.close()
#
from bottle import *
from redis import StrictRedis
import sys
sys.path.append("..")
from model import DB_Session, User
Session = DB_Session()
query = Session.query(User)
urlprefix =""
rdb = StrictRedis(db=2)
@route("%s/" %urlprefix)
def index():
return template("index.html")
@route("%s/name"%urlprefix)
def find_people():
key = request.GET.get('keyword')
#return str(request.GET.get
if key:
userlst = rdb.smembers(key)
users = { user : Session.execute("SELECT bio FROM peoples WHERE name='{0}'".format(user)).fetchall()[0][0].encode("utf-8") for user in userlst}
return template("result.tpl",peoples= users)
else:
redirect("%s/"%urlprefix)