beautifulsoup4
html5lib
image
requests
redis
pymysql
pip安装所有依赖包：
pip install \ image \ requests \ beautifulsoup4 \ html5lib \ redis \ pymysql
运行环境需要支持中文
测试运行环境python3.5，不保证其他运行环境能完美运行
需要安装mysql和redis
配置
config.ini
文件，设置好mysql和redis，并且填写你的知乎帐号
向数据库导入
init.sql
run
开始抓取数据:
python get_user.py
查看抓取数量:
python check_redis.py
效果
总体思路
1.首先是模拟登陆知乎，利用保存登陆的cookie信息
2.抓取知乎页面的html代码，留待下一步继续进行分析提取信息
3.分析提取页面中用户的个性化url，放入redis（这里特别说明一下redis的思路用法，将提取到的用户的个性化url放入redis的一个名为already_get_user的hash table，表示已抓取的用户，对于已抓取过的用户判断是否存在于already_get_user以去除重复抓取，同时将个性化url放入user_queue的队列中，需要抓取新用户时pop队列获取新的用户）
4.获取用户的关注列表和粉丝列表，继续插入到redis
5.从redis的user_queue队列中获取新用户继续重复步骤3
模拟登陆知乎
首先是登陆，登陆功能作为一个包封装了在login里面，方便整合调用
header部分，这里connection最好设为close，不然可能会碰到max retireve exceed的错误
原因在于普通的连接是keep-alive的但是却又没有关闭
# http请求的header headers = { user-agent: mozilla/5.0 (windows nt 10.0; win64; x64) applewebkit/537.36 (khtml, like gecko) chrome/53.0.2785.143 safari/537.36, accept: text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8, host: www.zhihu.com, referer: https://www.zhihu.com/, origin: https://www.zhihu.com/, upgrade-insecure-requests: 1, content-type: application/x-www-form-urlencoded; charset=utf-8, pragma: no-cache, accept-encoding: gzip, deflate, br, 'connection': 'close' } # 验证是否登陆 def check_login(self): check_url = 'https://www.zhihu.com/settings/profile' try: login_check = self.__session.get(check_url, headers=self.headers, timeout=35) except exception as err: print(traceback.print_exc()) print(err) print(验证登陆失败，请检查网络) sys.exit() print(验证登陆的http status code为： + str(login_check.status_code)) if int(login_check.status_code) == 200: return true else: return false
进入首页查看http状态码来验证是否登陆，200为已经登陆，一般304就是被重定向所以就是没有登陆
# 获取验证码 def get_captcha(self): t = str(time.time() * 1000) captcha_url = 'http://www.zhihu.com/captcha.gif?r=' + t + &type=login r = self.__session.get(captcha_url, headers=self.headers, timeout=35) with open('captcha.jpg', 'wb') as f: f.write(r.content) f.close() # 用pillow 的 image 显示验证码 # 如果没有安装 pillow 到源代码所在的目录去找到验证码然后手动输入 '''try: im = image.open('captcha.jpg') im.show() im.close() except:''' print(u'请到 %s 目录找到captcha.jpg 手动输入' % os.path.abspath('captcha.jpg')) captcha = input(请输入验证码\n>) return captcha
获取验证码的方法。当登录次数太多有可能会要求输入验证码，这里实现这个功能
# 获取xsrf def get_xsrf(self): index_url = 'http://www.zhihu.com' # 获取登录时需要用到的_xsrf try: index_page = self.__session.get(index_url, headers=self.headers, timeout=35) except: print('获取知乎页面失败，请检查网络连接') sys.exit() html = index_page.text # 这里的_xsrf 返回的是一个list bs = beautifulsoup(html, 'html.parser') xsrf_input = bs.find(attrs={'name': '_xsrf'}) pattern = r'value=\(.*?)\' print(xsrf_input) self.__xsrf = re.findall(pattern, str(xsrf_input)) return self.__xsrf[0]
获取xsrf，为什么要获取xsrf呢，因为xsrf是一种防止跨站攻击的手段，具体介绍可以看这里csrf
在获取到xsrf之后把xsrf存入cookie当中，并且在调用api的时候带上xsrf作为头部，不然的话知乎会返回403
# 进行模拟登陆 def do_login(self): try: # 模拟登陆 if self.check_login(): print('您已经登录') return else: if self.config.get(zhihu_account, username) and self.config.get(zhihu_account, password): self.username = self.config.get(zhihu_account, username) self.password = self.config.get(zhihu_account, password) else: self.username = input('请输入你的用户名\n> ') self.password = input(请输入你的密码\n> ) except exception as err: print(traceback.print_exc()) print(err) sys.exit() if re.match(r^1\d{10}$, self.username): print(手机登陆\n) post_url = 'http://www.zhihu.com/login/phone_num' postdata = { '_xsrf': self.get_xsrf(), 'password': self.password, 'remember_me': 'true', 'phone_num': self.username, } else: print(邮箱登陆\n) post_url = 'http://www.zhihu.com/login/email' postdata = { '_xsrf': self.get_xsrf(), 'password': self.password, 'remember_me': 'true', 'email': self.username, } try: login_page = self.__session.post(post_url, postdata, headers=self.headers, timeout=35) login_text = json.loads(login_page.text.encode('latin-1').decode('unicode-escape')) print(postdata) print(login_text) # 需要输入验证码 r = 0为登陆成功代码 if login_text['r'] == 1: sys.exit() except: postdata['captcha'] = self.get_captcha() login_page = self.__session.post(post_url, postdata, headers=self.headers, timeout=35) print(json.loads(login_page.text.encode('latin-1').decode('unicode-escape'))) # 保存登陆cookie self.__session.cookies.save()
这个就是核心的登陆功能啦，非常关键的就是用到了requests库，非常方便的保存到session
我们这里全局都是用单例模式，统一使用同一个requests.session对象进行访问功能，保持登录状态的一致性
最后主要调用登陆的代码为
# 创建login对象 lo = login.login.login(self.session) # 模拟登陆 if lo.check_login(): print('您已经登录') else: if self.config.get(zhihu_account, username) and self.config.get(zhihu_account, username): username = self.config.get(zhihu_account, username) password = self.config.get(zhihu_account, password) else: username = input('请输入你的用户名\n> ') password = input(请输入你的密码\n> ) lo.do_login(username, password)
知乎模拟登陆到此就完成啦
知乎用户抓取
def __init__(self, threadid=1, name=''): # 多线程 print(线程 + str(threadid) + 初始化) threading.thread.__init__(self) self.threadid = threadid self.name = name try: print(线程 + str(threadid) + 初始化成功) except exception as err: print(err) print(线程 + str(threadid) + 开启失败) self.threadlock = threading.lock() # 获取配置 self.config = configparser.configparser() self.config.read(config.ini) # 初始化session requests.adapters.default_retries = 5 self.session = requests.session() self.session.cookies = cookielib.lwpcookiejar(filename='cookie') self.session.keep_alive = false try: self.session.cookies.load(ignore_discard=true) except: print('cookie 未能加载') finally: pass # 创建login对象 lo = login(self.session) lo.do_login() # 初始化redis连接 try: redis_host = self.config.get(redis, host) redis_port = self.config.get(redis, port) self.redis_con = redis.redis(host=redis_host, port=redis_port, db=0) # 刷新redis库 # self.redis_con.flushdb() except: print(请安装redis或检查redis连接配置) sys.exit() # 初始化数据库连接 try: db_host = self.config.get(db, host) db_port = int(self.config.get(db, port)) db_user = self.config.get(db, user) db_pass = self.config.get(db, password) db_db = self.config.get(db, db) db_charset = self.config.get(db, charset) self.db = pymysql.connect(host=db_host, port=db_port, user=db_user, passwd=db_pass, db=db_db, charset=db_charset) self.db_cursor = self.db.cursor() except: print(请检查数据库配置) sys.exit() # 初始化系统设置 self.max_queue_len = int(self.config.get(sys, max_queue_len))
这个是get_user.py的构造函数，主要功能就是初始化mysql连接、redis连接、验证登陆、生成全局的session对象、导入系统配置、开启多线程。
# 获取首页html def get_index_page(self): index_url = 'https://www.zhihu.com/' try: index_html = self.session.get(index_url, headers=self.headers, timeout=35) except exception as err: # 出现异常重试 print(获取页面失败，正在重试......) print(err) traceback.print_exc() return none finally: pass return index_html.text # 获取单个用户详情页面 def get_user_page(self, name_url): user_page_url = 'https://www.zhihu.com' + str(name_url) + '/about' try: index_html = self.session.get(user_page_url, headers=self.headers, timeout=35) except exception as err: # 出现异常重试 print(失败name_url： + str(name_url) + 获取页面失败，放弃该用户) print(err) traceback.print_exc() return none finally: pass return index_html.text # 获取粉丝页面 def get_follower_page(self, name_url): user_page_url = 'https://www.zhihu.com' + str(name_url) + '/followers' try: index_html = self.session.get(user_page_url, headers=self.headers, timeout=35) except exception as err: # 出现异常重试 print(失败name_url： + str(name_url) + 获取页面失败，放弃该用户) print(err) traceback.print_exc() return none finally: pass return index_html.text def get_following_page(self, name_url): user_page_url = 'https://www.zhihu.com' + str(name_url) + '/followers' try: index_html = self.session.get(user_page_url, headers=self.headers, timeout=35) except exception as err: # 出现异常重试 print(失败name_url： + str(name_url) + 获取页面失败，放弃该用户) print(err) traceback.print_exc() return none finally: pass return index_html.text # 获取首页上的用户列表，存入redis def get_index_page_user(self): index_html = self.get_index_page() if not index_html: return bs = beautifulsoup(index_html, html.parser) self.get_xsrf(index_html) user_a = bs.find_all(a, class_=author-link) # 获取用户的a标签 for a in user_a: if a: self.add_wait_user(a.get('href')) else: continue
这一部分的代码就是用于抓取各个页面的html代码
# 加入带抓取用户队列，先用redis判断是否已被抓取过 def add_wait_user(self, name_url): # 判断是否已抓取 self.threadlock.acquire() if not self.redis_con.hexists('already_get_user', name_url): self.counter += 1 print(name_url + 加入队列) self.redis_con.hset('already_get_user', name_url, 1) self.redis_con.lpush('user_queue', name_url) print(添加用户 + name_url + 到队列) self.threadlock.release() # 获取页面出错移出redis def del_already_user(self, name_url): self.threadlock.acquire() if not self.redis_con.hexists('already_get_user', name_url): self.counter -= 1 self.redis_con.hdel('already_get_user', name_url) self.threadlock.release()
用户加入redis的操作，在数据库插入出错时我们调用del_already_user删除插入出错的用户
# 分析粉丝页面获取用户的所有粉丝用户 # @param follower_page get_follower_page()中获取到的页面，这里获取用户hash_id请求粉丝接口获取粉丝信息 def get_all_follower(self, name_url): follower_page = self.get_follower_page(name_url) # 判断是否获取到页面 if not follower_page: return bs = beautifulsoup(follower_page, 'html.parser') # 获取关注者数量 follower_num = int(bs.find('span', text='关注者').find_parent().find('strong').get_text()) # 获取用户的hash_id hash_id = \ json.loads(bs.select(#zh-profile-follows-list)[0].select(.zh-general-list)[0].get('data-init'))[ 'params'][ 'hash_id'] # 获取关注者列表 self.get_xsrf(follower_page) # 获取xsrf post_url = 'https://www.zhihu.com/node/profilefollowerslistv2' # 开始获取所有的关注者 math.ceil(follower_num/20)*20 for i in range(0, math.ceil(follower_num / 20) * 20, 20): post_data = { 'method': 'next', 'params': json.dumps({offset: i, order_by: created, hash_id: hash_id}) } try: j = self.session.post(post_url, params=post_data, headers=self.headers, timeout=35).text.encode( 'latin-1').decode( 'unicode-escape') pattern = re.compile(rclass=\zm-item-link-avatar\[^\]*\([^\]*), re.dotall) j = pattern.findall(j) for user in j: user = user.replace('\\', '') self.add_wait_user(user) # 保存到redis except exception as err: print(获取正在关注失败) print(err) traceback.print_exc() pass # 获取正在关注列表 def get_all_following(self, name_url): following_page = self.get_following_page(name_url) # 判断是否获取到页面 if not following_page: return bs = beautifulsoup(following_page, 'html.parser') # 获取关注者数量 following_num = int(bs.find('span', text='关注了').find_parent().find('strong').get_text()) # 获取用户的hash_id hash_id = \ json.loads(bs.select(#zh-profile-follows-list)[0].select(.zh-general-list)[0].get('data-init'))[ 'params'][ 'hash_id'] # 获取关注者列表 self.get_xsrf(following_page) # 获取xsrf post_url = 'https://www.zhihu.com/node/profilefolloweeslistv2' # 开始获取所有的关注者 math.ceil(follower_num/20)*20 for i in range(0, math.ceil(following_num / 20) * 20, 20): post_data = { 'method': 'next', 'params': json.dumps({offset: i, order_by: created, hash_id: hash_id}) } try: j = self.session.post(post_url, params=post_data, headers=self.headers, timeout=35).text.encode( 'latin-1').decode( 'unicode-escape') pattern = re.compile(rclass=\zm-item-link-avatar\[^\]*\([^\]*), re.dotall) j = pattern.findall(j) for user in j: user = user.replace('\\', '') self.add_wait_user(user) # 保存到redis except exception as err: print(获取正在关注失败) print(err) traceback.print_exc() pass
调用知乎的api，获取所有的关注用户列表和粉丝用户列表，递归获取用户
这里需要注意的是头部要记得带上xsrf不然会抛出403
# 分析about页面，获取用户详细资料 def get_user_info(self, name_url): about_page = self.get_user_page(name_url) # 判断是否获取到页面 if not about_page: print(获取用户详情页面失败，跳过，name_url： + name_url) return self.get_xsrf(about_page) bs = beautifulsoup(about_page, 'html.parser') # 获取页面的具体数据 try: nickname = bs.find(a, class_=name).get_text() if bs.find(a, class_=name) else '' user_type = name_url[1:name_url.index('/', 1)] self_domain = name_url[name_url.index('/', 1) + 1:] gender = 2 if bs.find(i, class_=icon icon-profile-female) else (1 if bs.find(i, class_=icon icon-profile-male) else 3) follower_num = int(bs.find('span', text='关注者').find_parent().find('strong').get_text()) following_num = int(bs.find('span', text='关注了').find_parent().find('strong').get_text()) agree_num = int(re.findall(r'<strong>(.*)</strong>.*赞同', about_page)[0]) appreciate_num = int(re.findall(r'<strong>(.*)</strong>.*感谢', about_page)[0]) star_num = int(re.findall(r'<strong>(.*)</strong>.*收藏', about_page)[0]) share_num = int(re.findall(r'<strong>(.*)</strong>.*分享', about_page)[0]) browse_num = int(bs.find_all(span, class_=zg-gray-normal)[2].find(strong).get_text()) trade = bs.find(span, class_=business item).get('title') if bs.find(span, class_=business item) else '' company = bs.find(span, class_=employment item).get('title') if bs.find(span, class_=employment item) else '' school = bs.find(span, class_=education item).get('title') if bs.find(span, class_=education item) else '' major = bs.find(span, class_=education-extra item).get('title') if bs.find(span, class_=education-extra item) else '' job = bs.find(span, class_=position item).get_text() if bs.find(span, class_=position item) else '' location = bs.find(span, class_=location item).get('title') if bs.find(span, class_=location item) else '' description = bs.find(p, class_=bio ellipsis).get('title') if bs.find(p, class_=bio ellipsis) else '' ask_num = int(bs.find_all(a, class_='item')[1].find(span).get_text()) if \ bs.find_all(a, class_='item')[ 1] else int(0) answer_num = int(bs.find_all(a, class_='item')[2].find(span).get_text()) if \ bs.find_all(a, class_='item')[ 2] else int(0) article_num = int(bs.find_all(a, class_='item')[3].find(span).get_text()) if \ bs.find_all(a, class_='item')[3] else int(0) collect_num = int(bs.find_all(a, class_='item')[4].find(span).get_text()) if \ bs.find_all(a, class_='item')[4] else int(0) public_edit_num = int(bs.find_all(a, class_='item')[5].find(span).get_text()) if \ bs.find_all(a, class_='item')[5] else int(0) replace_data = \ (pymysql.escape_string(name_url), nickname, self_domain, user_type, gender, follower_num, following_num, agree_num, appreciate_num, star_num, share_num, browse_num, trade, company, school, major, job, location, pymysql.escape_string(description), ask_num, answer_num, article_num, collect_num, public_edit_num) replace_sql = '''replace into user(url,nickname,self_domain,user_type, gender, follower,following,agree_num,appreciate_num,star_num,share_num,browse_num, trade,company,school,major,job,location,description, ask_num,answer_num,article_num,collect_num,public_edit_num) values(%s,%s,%s,%s, %s,%s,%s,%s,%s,%s,%s,%s, %s,%s,%s,%s,%s,%s,%s, %s,%s,%s,%s,%s)''' try: print(获取到数据：) print(replace_data) self.db_cursor.execute(replace_sql, replace_data) self.db.commit() except exception as err: print(插入数据库出错) print(获取到数据：) print(replace_data) print(插入语句： + self.db_cursor._last_executed) self.db.rollback() print(err) traceback.print_exc() except exception as err: print(获取数据出错，跳过用户) self.redis_con.hdel(already_get_user, name_url) self.del_already_user(name_url) print(err) traceback.print_exc() pass
最后，到用户的about页面，分析页面元素，利用正则或者beatifulsoup分析抓取页面的数据
这里我们sql语句用replace into而不用insert into，这样可以很好的防止数据重复问题
# 开始抓取用户，程序总入口 def entrance(self): while 1: if int(self.redis_con.llen(user_queue)) < 1: self.get_index_page_user() else: # 出队列获取用户name_url redis取出的是byte，要decode成utf-8 name_url = str(self.redis_con.rpop(user_queue).decode('utf-8')) print(正在处理name_url： + name_url) self.get_user_info(name_url) if int(self.redis_con.llen(user_queue)) <= int(self.max_queue_len): self.get_all_follower(name_url) self.get_all_following(name_url) self.session.cookies.save() def run(self): print(self.name + is running) self.entrance()
最后，入口
if __name__ == '__main__': login = getuser(999, 登陆线程) threads = [] for i in range(0, 4): m = getuser(i, thread + str(i)) threads.append(m) for i in range(0, 4): threads[i].start() for i in range(0, 4): threads[i].join()
这里就是多线程的开启，需要开启多少个线程就把4换成多少就可以了
docker
嫌麻烦的可以参考一下我用docker简单的搭建一个基础环境：
mysql和redis都是官方镜像
docker run --name mysql -itd mysql:latest docker run --name redis -itd mysql:latest
再利用docker-compose运行python镜像，我的python的docker-compose.yml：
python: container_name: python build: . ports: - 84:80 external_links: - memcache:memcache - mysql:mysql - redis:redis volumes: - /docker_containers/python/www:/var/www/html tty: true stdin_open: true extra_hosts: - python:192.168.102.140 environment: pythonioencoding: utf-8
以上就是利用python实现多线程抓取知乎用户方法的详细内容。