zoukankan      html  css  js  c++  java
  • 爬取某些网站的弹幕和评论数据 Python

    本文仅用于学习与交流使用,不具有任何商业价值,如有问题,请与我联系,我会即时处理。---Python逐梦者。

    首先是某果TV。

    弹幕。以电影《悬崖之上》为例。弹幕数据所在的文件是动态加载的,打开开发者工具,让它加载很多数据,然后搜索某一条数据就看到在哪个包里了,然后就是参数变化不同分析。某果TV的视频播放一分钟它就会更新一个json数据包,里面包含需要的弹幕数据。动手干。

     1 import csv
     2 import pprint
     3 import random
     4 import time
     5 import requests
     6 import pandas as pd
     7 
     8 f = open('悬崖之上.csv', mode='a', encoding='utf-8-sig', newline='')
     9 csvWriter = csv.DictWriter(f, fieldnames=[
    10     '用户id',
    11     '弹幕内容',
    12     '获赞数',
    13 ])
    14 csvWriter.writeheader()
    15 # 请求头
    16 headers = {
    17     "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/95.0.4638.69 Safari/537.36",
    18 }
    19 # 开始请求并多页爬取
    20 for page in range(0, 120 + 1):
    21     print(f'=====正在爬取第{page}页数据=====')
    22     time.sleep(random.randint(2, 5))
    23     url = f'https://bullet-ali.xxx.com/bullet/2021/11/15/005204/12281642/{page}.json'
    24     response = requests.get(url=url, headers=headers)
    25     # pprint.pprint(response.json())
    26     # 直接提取需要的,这里提取一下id和内容
    27     for i in response.json()['data']['items']:
    28         # 首先是id
    29         try:
    30             id = i['id']
    31         except:
    32             id = '未知!'
    33         # 其次是内容
    34         content = i['content']
    35         # 点赞数
    36         try:
    37             like = i['v2_up_count']
    38         except:
    39             like = '未获得!'
    40 
    41         # # 组织数据
    42         # text = pd.DataFrame({'用户id':[id], '弹幕内容':[content], '获赞数':[like]})
    43         # # print(text)
    44         # df = pd.concat([text])
    45         dit = {
    46             '用户id':id,
    47             '评论内容':content,
    48             '获赞数':like,
    49         }
    50         print(dit) # 打印是否符合预期
    51         csvWriter.writerow(dit) # 逐行写入内容
    52     break # 调试的时候只怕一页

    程序运行结果:

    评论。照样《悬崖之上》为例,爬一波评论数据。本来评论请求的url地址长这样:https://comment.xxx.com/v4/comment/getCommentList?page=1&subjectType=hunantv2014&subjectId=12281642&callback=jQuery182024636113438271012_1636961381836&_support=10000000&_=1636961383307,有人说callbackhi干扰数据解析,而时间戳不会破坏数据完整性。所以把url改成了:https://comment.xxx.com/v4/comment/getCommentList?page=1&subjectType=hunantv2014&subjectId=12281642&_support=10000000。

     代码分页的标准的参数,就是2533/15 = 168.86,也就是最大页数为169。开干。

     1 """
     2     爬取芒果TV的评论数据
     3 """
     4 import csv
     5 import pprint
     6 import random
     7 import time
     8 
     9 import requests
    10 
    11 f = open('悬崖之上评论数据.csv', mode='a', encoding='utf-8-sig', newline='')
    12 csvWriter = csv.DictWriter(f, fieldnames=[
    13     '评论者',
    14     '评论创建时间',
    15     '评论内容',
    16     '被点赞数',
    17 ])
    18 csvWriter.writeheader() # 写入头
    19 headers = {
    20     'user-agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/95.0.4638.69 Safari/537.36',
    21 }
    22 for page in range(1, 169):
    23     print(f'=====正在爬取第{page}页内容=====')
    24     time.sleep(random.randint(2, 5)) # 随机休眠
    25     url = f'https://comment.xxxx.com/v4/comment/getCommentList?page={page}&subjectType=hunantv2014&subjectId=12281642&_support=10000000'
    26     response = requests.get(url=url, headers=headers)
    27     # print(response.json()['data']['list']) # 这是个列表
    28     # pprint.pprint(response.json()['data']['list'])
    29     # 提取评论人姓名,评论日期,评论内容和被点赞数
    30     for item in response.json()['data']['list']:
    31         name = item['user']['nickName'] # 评论人姓名
    32         contentCreated = item['date'] # 评论时间,也可以获取时间戳转换成本地时间
    33         content = item['content'] # 评论内容
    34         praiseNum = item['praiseNum'] # 被点赞数
    35         dit = {
    36             '评论者':name,
    37             '评论创建时间':contentCreated,
    38             '评论内容':content,
    39             '被点赞数':praiseNum,
    40         }
    41         print(dit)
    42         # 写入到csv
    43         csvWriter.writerow(dit) # 逐行写入
    44 print('爬取完成!')

    程序运行截图:

     

    其次是某讯视频

    弹幕。进入开发者工具,当视频播放30秒它就会更新一个json数据包,里面包含需要的弹幕数据。

    对比请求的url发现,第一次请求的是15个,后面请求的都是30。

    https://mfm.video.xx.com/danmu?otype=json&callback=jQuery19109701649659612566_1637029736329&target_id=7220956568&vid=t0040z3o3la&session_key=0,38,1637029735&timestamp=15&_=1637029736342
    https://mfm.video.xx.com/danmu?otype=json&callback=jQuery19109701649659612566_1637029736329&target_id=7220956568&vid=t0040z3o3la&session_key=0,38,1637029735&timestamp=45&_=1637029736342

    照样干掉不必要的callback参数。开干:

     1 import csv
     2 import random
     3 import time
     4 import requests
     5 
     6 f = open('某讯视频弹幕数据.csv', mode='a', encoding='utf-8-sig', newline='')
     7 csvWriter = csv.DictWriter(f, fieldnames=[
     8     '弹幕发送ID',
     9     '弹幕内容',
    10     '获赞数',
    11 ])
    12 # 写入头
    13 csvWriter.writeheader()
    14 # 请求头
    15 headers = {
    16     'user-agent':'https://mfm.video.xxx.com/danmu?otype=json&callback=jQuery19109701649659612566_1637029736329&target_id=7220956568&vid=t0040z3o3la&session_key=0,38,1637029735&timestamp=165&_=1637029736342',
    17     'referer':'https://v.qq.com/',
    18 }
    19 # 多页爬取
    20 for timestamp in range(15, 7245, 30): #初始为15,7245为视频总秒长,后面以30递增
    21     time.sleep(random.randint(2, 5)) # 休眠
    22     url = f'https://mfm.video.xxx.com/danmu?otype=json&target_id=7220956568&vid=t0040z3o3la&session_key=0,38,1637029735&timestamp={timestamp}&_=1637029736342'
    23     # 请求数据
    24     response = requests.get(url=url, headers=headers)
    25     # print(response.json()) # 如果不干掉url里的callback参数的话得到的数据需要处理才能用json加载
    26     # 提取数据
    27     for item in response.json()['comments']:
    28         id = item['commentid'] # 弹幕发送者id
    29         danmu = item['content'] # 获取到弹幕内容
    30         like = item['upcount'] # 获赞数
    31         dit = {
    32             '弹幕发送者ID':id,
    33             '弹幕内容':danmu,
    34             '获赞数':like,
    35         }
    36         print(dit)
    37         csvWriter.writerow(dit)
    38 
    39 print('爬取完成!')

    程序运行截图:

     评论获取。 某讯视频评论数据在网页底部,是动态加载的,需要抓包进行分析。

     请求的url是:

    https://video.coral.xxx.com/varticle/6655100451/comment/v2?callback=_varticle6655100451commentv2&orinum=10&oriorder=o&pageflag=1&cursor=0&scorecursor=0&orirepnum=2&reporder=o&reppageflag=1&source=132&_=1637030980407
    https://video.coral.xxx.com/varticle/6655100451/comment/v2?callback=_varticle6655100451commentv2&orinum=10&oriorder=o&pageflag=1&cursor=6829967729286321250&scorecursor=0&orirepnum=2&reporder=o&reppageflag=1&source=132&_=1637030980410

    变化在cursor参数,开始的时候cursor是0,翻页后的cursor是上一个url的json数据里的data,last字段。也就是可以通过获取response.json()['data']['last']来获取curosr。开干:

     1 import csv
     2 import random
     3 import re
     4 import time
     5 import requests
     6 from urllib.parse import unquote
     7 
     8 f = open('某讯视频评论.csv', mode='a', encoding='utf-8-sig', newline='')
     9 csvWriter = csv.DictWriter(f, fieldnames=[
    10     'Id',
    11     '评论人',
    12     '评论时间',
    13     '获赞数',
    14     '评论内容',
    15 ])
    16 # 写入头
    17 csvWriter.writeheader()
    18 # 请求头
    19 headers = {
    20     'user-agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/95.0.4638.69 Safari/537.36',
    21 }
    22 
    23 # 每页的评论数是10,总共有3203,所以最大值是321
    24 page = 1 # 初始计数,一定从1开始,从0开始的话,while循环里获取不到cursor
    25 while page < 321:
    26     print(f'=====开始爬取{page}页的内容=====')
    27     time.sleep(random.randint(2, 5)) # 随机休眠
    28     # 第一页的情况,第一页很重要,因为我们要获取cursor以供else分支使用
    29     if page == 1:
    30         # 照样干掉url里的callback参数
    31         url = 'https://video.coral.xxx.com/varticle/6655100451/comment/v2?orinum=10&oriorder=o&pageflag=1&cursor=0&scorecursor=0&orirepnum=2&reporder=o&reppageflag=1&source=132'
    32     else:
    33         url = f'https://video.coral.xxxx.com/varticle/6655100451/comment/v2?orinum=10&oriorder=o&pageflag=1&cursor={cursor}&scorecursor=0&orirepnum=2&reporder=o&reppageflag=1&source=132'
    34     res = requests.get(url=url, headers=headers).json() # 请求数据 第一次请求的实际上是page==1的情况
    35     cursor = res['data']['last'] # 第一次请求url后就能得到一个cursor,然后通过循环传递给循环分支else
    36     print(f'=====要爬取的url为{url}=====')
    37     time.sleep(2)
    38     # 每次请求完url后获取数据
    39     for item in res['data']['oriCommList']:
    40         id = item['id'] # 评论者id
    41         # 使用urllib的unquote函数对url进行解码
    42         nickname = unquote(item['custom'])
    43         commentTime = item['time'] # 这个是时间是时间戳,存储的时候再做转换
    44         # 尝试获取获赞数
    45         try:
    46             like = item.get('up')
    47         except:
    48             like = 'No'
    49         content = item['content'] # 评论内容
    50         dit = {
    51             'Id':id,
    52             '评论人':re.findall('nick=(.*?)&head', nickname)[0], # 通过正则把解码后的字符串里的名字提取出来
    53             '评论时间':time.strftime("%Y-%m-%d %H:%M", time.localtime(int(commentTime))), # 将获取到的时间戳转换成本地时间
    54             '获赞数':like,
    55             '评论内容':content,
    56         }
    57         print(dit)
    58         csvWriter.writerow(dit) # 逐行写入csv文档
    59     page+=1 # 翻页递增 进入下一次循环
    60     time.sleep(random.uniform(2, 3))
    61 print('评论爬完成!')

    程序运行部分截图:

     第三是B站

    以《EDG夺冠时刻》,B站的上的纪录片为例进行爬取。视频地址:https://www.bilibili.com/bangumi/play/ss39849/?from=search&seid=6112754044363142537&spm_id_from=333.337.0.0。

    弹幕:

    找到视频,点开右边的弹幕列表,加载弹幕的时候,得到的数据如下:

     

     假如登录了的话,可以点击“查看历史弹幕”,如果没登录,这个按钮是灰色的。

     历史弹幕包括2021年12月1日到8日的弹幕。打点击“查看历史弹幕”的时候,会出现每天弹幕的数据,得到类似url:https://api.bilibili.com/x/v2/dm/web/history/seg.so?type=1&oid=445826862&date=2021-12-02。发现其中的关联,开始构造url并开干。

     1 import requests
     2 import pandas as pd
     3 import re
     4 
     5 def data_response(url):
     6     headers = {
     7         "cookie":"_uuid=BE35640F-EB4E-F87D-53F2-7A8FD5D50E3330964infoc; buvid3=D0213B95-F001-4A46-BE4F-E921AE18EB67167647infoc; CURRENT_BLACKGAP=1; CURRENT_QUALITY=0; rpdid=|(u))ku~m)kJ0J'uYJuRRRYmk; video_page_version=v_old_home_17; blackside_state=1; LIVE_BUVID=AUTO1516364619569495; b_lsid=E27592910_17D990B450B; bsource=search_baidu; buvid_fp=D0213B95-F001-4A46-BE4F-E921AE18EB67167647infoc; innersign=1; sid=ipqajpj8; CURRENT_FNVAL=80; PVID=2; fingerprint=23eb07890bf96775d60093211947fae4; buvid_fp_plain=2919B0C8-360F-47D1-8DD1-51FA81536F4E34777infoc; DedeUserID=603136708; DedeUserID__ckMd5=2e5e771f4e696459; SESSDATA=93ba949a,1654503622,fb700*c1; bili_jct=9d8bc6e01fc089192a6aeed373a0333c",
     8         "referer":"https://www.bilibili.com/bangumi/play/ss39849/?from=search&seid=6112754044363142537&spm_id_from=333.337.0.0",
     9         "user-agent":"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/96.0.4664.45 Safari/537.36",
    10     }
    11     response = requests.get(url=url, headers=headers)
    12 
    13     return response # 返回
    14 
    15 def main(oid, month):
    16     df = pd.DataFrame()
    17     url = f'https://api.bilibili.com/x/v2/dm/history/index?month={month}&type=1&oid={oid}'
    18     # url = f'https://api.bilibili.com/x/v2/dm/history/index?type=1&oid={oid}&month={month}'
    19     list_data = data_response(url).json()['data']  # 拿到所有日期
    20     print(list_data)
    21     for data in list_data:
    22         urls = f'https://api.bilibili.com/x/v2/dm/web/history/seg.so?type=1&oid={oid}&date={data}' # 网址
    23         text = re.findall(".*?([\u4E00-\u9FA5]+).*?", data_response(urls).text)
    24         for e in text:
    25             print(e)
    26             data = pd.DataFrame({'弹幕': [e]})
    27             df = pd.concat([df, data])
    28     df.to_csv('弹幕.csv', encoding='utf-8-sig', index=False, mode='a+')
    29 
    30 
    31 if __name__ == "__main__":
    32     oid = '445826862' # 视频弹幕链接的id值
    33     month = '2021-12' # 开始日期,这里至于哦2021-12-01到2021-12-08
    34     main(oid, month) # 运行程序
    35     

    这样就把所有弹幕数据都爬下来了。还有一种爬b站弹幕的方法。有人说B站的弹幕藏在:https://comment.bilibili.com/445826862.xml,其中的数字就是视频的oid。用浏览器打开试了下,确实都在,也来用python爬下。

     1 import requests
     2 import pandas as pd
     3 import re
     4 
     5 def data_get(oid):
     6     headers = {
     7         "cookie": "_uuid=BE35640F-EB4E-F87D-53F2-7A8FD5D50E3330964infoc; buvid3=D0213B95-F001-4A46-BE4F-E921AE18EB67167647infoc; CURRENT_BLACKGAP=1; CURRENT_QUALITY=0; rpdid=|(u))ku~m)kJ0J'uYJuRRRYmk; video_page_version=v_old_home_17; blackside_state=1; LIVE_BUVID=AUTO1516364619569495; b_lsid=E27592910_17D990B450B; bsource=search_baidu; buvid_fp=D0213B95-F001-4A46-BE4F-E921AE18EB67167647infoc; innersign=1; sid=ipqajpj8; CURRENT_FNVAL=80; PVID=2; fingerprint=23eb07890bf96775d60093211947fae4; buvid_fp_plain=2919B0C8-360F-47D1-8DD1-51FA81536F4E34777infoc; DedeUserID=603136708; DedeUserID__ckMd5=2e5e771f4e696459; SESSDATA=93ba949a,1654503622,fb700*c1; bili_jct=9d8bc6e01fc089192a6aeed373a0333c",
     8         "referer": "https://www.bilibili.com/bangumi/play/ss39849/?from=search&seid=6112754044363142537&spm_id_from=333.337.0.0",
     9         "user-agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/96.0.4664.45 Safari/537.36",
    10     }
    11     url = f'https://comment.bilibili.com/{oid}.xml'
    12     # 开始请求网页
    13     response = requests.get(url=url, headers=headers)
    14     response.encoding = response.apparent_encoding # 自动识别编码
    15     response.encoding = 'utf-8' # 将编码转换成utf-8
    16     print(response.text) #
    17     textData = re.findall('<d p=".*?">(.*?)</d>', response.text) # 表达式的括号里才是需要的弹幕数据,是个列表
    18     df = pd.DataFrame()
    19     # 开始保存数据
    20     for item in textData:
    21         print(item)
    22         data = pd.DataFrame({'弹幕': [item]})
    23         df = pd.concat([df, data])
    24     df.to_csv('弹幕1.csv', encoding='utf-8-sig', index=False, mode='a+')
    25 
    26 if __name__ == "__main__":
    27     oid = '445826862'
    28     data_get(oid)

    看了几行数据,完全不一样,可能存储格式不一样。不纠结了,接下来爬评论。

    评论:

    打开开发者工具,按F12查找评论请求的链接。得到如下url。

    https://api.bilibili.com/x/v2/reply/main?callback=jQuery172024432989634133118_1638953989760&jsonp=jsonp&next=0&type=1&oid=506840377&mode=3&plat=1&_=1638954002015
    https://api.bilibili.com/x/v2/reply/main?callback=jQuery172024432989634133118_1638953989760&jsonp=jsonp&next=2&type=1&oid=506840377&mode=3&plat=1&_=1638954002015
    https://api.bilibili.com/x/v2/reply/main?callback=jQuery172024432989634133118_1638953989760&jsonp=jsonp&next=3&type=1&oid=506840377&mode=3&plat=1&_=1638954002015

    链接还是比较奇葩的,参数只有一个有变化就是next=数字。第一页是0,第二页是 2,第三页是3。不是常规的012这种格式。但是实际在网页中打开012这样的格式,也是会返回数据的。这里以它的格式为准,照样把不必要的参数干掉。就是callback参数和最后的时间戳,callback参数会影响json数据解析,时间戳不会。

     1 import csv
     2 import random
     3 import threading
     4 import time
     5 import requests
     6 import pandas as pd
     7 from threading import Thread
     8 
     9 # 请求头
    10 headers = {
    11     'user-agent':'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/96.0.4664.45 Safari/537.36',
    12 }
    13 
    14 def pdToCsv():
    15     df = pd.DataFrame() # pandas数据格式,用于保存
    16     try:
    17         a = 1 # 约定
    18         while True:
    19             if a == 1: # 如果a=1,就让url里的next=0
    20                 url = 'https://api.bilibili.com/x/v2/reply/main?jsonp=jsonp&next=0&type=1&oid=506840377&mode=3&plat=1'
    21             else:
    22                 url = f'https://api.bilibili.com/x/v2/reply/main?jsonp=jsonp&next={a}&type=1&oid=506840377&mode=3&plat=1'
    23             # 开始请求url
    24             response = requests.get(url=url, headers=headers)
    25             time.sleep(random.uniform(2, 5))
    26             for i in response.json()['data']['replies']:
    27                 uname = i['member']['uname']  # 用户名称
    28                 sex = i['member']['sex']  # 用户性别
    29                 mid = i['mid']  # 用户id
    30                 current_level = i['member']['level_info']['current_level']  # vip等级
    31                 message = i['content']['message'].replace('\n', '')  # 用户评论
    32                 like = i['like']  # 评论点赞次数
    33                 ctime = i['ctime']  # 评论时间
    34                 data = pd.DataFrame({'用户名': [uname], '性别': [sex], 'id': [mid],
    35                                      'vip等级': [current_level], '评论': [message], '获赞数': [like],
    36                                      '评论时间': [ctime]})
    37                 df = pd.concat([df, data])
    38             a += 1 # 自增1以便下一次循环
    39     except Exception as e:
    40         print(e)
    41     df.to_csv('我们是冠军pd.csv', encoding='utf-8-sig') # 保存数据
    42     print(df.shape)
    43 
    44 
    45 """也可以向下面这样写"""
    46 def stringToCsv():
    47     f = open('我们是冠军评论csv.csv', mode='a', encoding='utf-8-sig', newline='') # 打开文件
    48     csvWriter = csv.DictWriter(f, fieldnames=[
    49         '用户名',
    50         '性别',
    51         'id',
    52         'vip等级',
    53         '评论内容',
    54         '获赞数',
    55         '评论时间',
    56     ])
    57     csvWriter.writeheader() #  写入头
    58     n = 1
    59     while n < 5426 / 10 + 1: # 因为总共有5426条,所以要爬550页
    60         time.sleep(random.uniform(2,5))
    61         if n == 1: # 循环的第一个请求url
    62             url = 'https://api.bilibili.com/x/v2/reply/main?jsonp=jsonp&next=0&type=1&oid=506840377&mode=3&plat=1'
    63         else:
    64             url = f'https://api.bilibili.com/x/v2/reply/main?jsonp=jsonp&next={n}&type=1&oid=506840377&mode=3&plat=1'
    65 
    66         # 开始请求数据
    67         response = requests.get(url=url, headers=headers)
    68         for i in response.json()['data']['replies']:
    69             uname = i['member']['uname']  # 用户名称
    70             sex = i['member']['sex']  # 用户性别
    71             mid = i['mid']  # 用户id
    72             current_level = i['member']['level_info']['current_level']  # vip等级
    73             message = i['content']['message'].replace('\n', '')  # 用户评论
    74             like = i['like']  # 评论点赞次数
    75             ctime = i['ctime']  # 评论时间
    76             print(uname, sex, mid, current_level, message, like, ctime, sep='|')
    77             dit = {
    78                 '用户名': uname,
    79                 '性别': sex,
    80                 'id': mid,
    81                 'vip等级': current_level,
    82                 '评论内容': message,
    83                 '获赞数': like,
    84                 '评论时间': ctime,
    85             }
    86             # 逐行写入
    87             csvWriter.writerow(dit)
    88     n += 1 # 循环条件,不然会死循环
    89     f.close() # 关闭文件
    90 
    91 if __name__ == "__main__":
    92     thread1 = threading.Thread(target=pdToCsv)
    93     thread2 = threading.Thread(target=stringToCsv)
    94     thread1.start()
    95     thread2.start()
    96     thread1.join()
    97     thread2.join()
  • 相关阅读:
    凹透镜
    三角形动点和将军饮马
    数学
    壮壮学习准则
    均值不等式,求极值
    2020年自贡中考数学真题,用的是花钱买的"几何画板",wechat:QZCS12
    90年高考题
    裂项:2005年初中数学竞赛题p32,4
    02-需求来源
    01-产品需求的内涵
  • 原文地址:https://www.cnblogs.com/mafu/p/15557207.html
Copyright © 2011-2022 走看看