写在前面
最近系统地学习了正则表达式，发现正则表达式最大的应用之一——网络爬虫之前一直没有涉猎，遂借此契机顺带写一个爬虫的demo备用。选择对象是CSDN排行榜前100用户，各自按访问量从高到低排序的前20篇文章，使用一些简单的数据分析手段看看技术热点，方便今后拓宽技术栈。
项目总述
主要爬取的数据是文章标题和访问量，先总体可视化总体文章的技术关键词；然后按访问量分组，可视化每个访问段的技术热点。
数据爬取
获得服务器API
首先我们要知道通过什么接口可以获得网站数据：首先进入博客总榜，按F12进入控制台，选中Network选项卡监视网络请求，然后刷新网页。从下图可以看到在API"https://blog.csdn.net/phoenix/web/blog/all-rank?page=1&pageSize=20"中我们可以拿到我们想要的用户信息——主要是用户名
现在到用户博客首页，同样地，按F12进入控制台，选中Network选项卡监视网络请求，然后点击按访问量排序，则可以发现另一个关键APIhttps://blog.csdn.net/community/home-api/v1/get-business-list?page=1&size=20&businessType=blog&orderby=ViewCount&noMore=false&username={}，如下图所示。
我们与服务器的交互就依靠这两个API进行。
程序总体设计
思考一下，我们总共有如下的公共变量：
# 请求头 headers = { 'User-Agent': 'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/63.0.3239.132 Safari/537.36' } # 排行榜url rankUrl = "https://blog.csdn.net/phoenix/web/blog/all-rank?page={}&pageSize=20" # 按访问量排行的文章列表 mostViewArtical = "https://blog.csdn.net/community/home-api/v1/get-business-list?page=1&size=20&businessType=blog&orderby=ViewCount&noMore=false&username={}" userNames =[] # 用户名列表 titleList = [] # 文章标题列表 viewCntList = [] # 访问量列表
为便于管理，引入一个类进行爬虫，专门负责与服务器进行数据交互
class GetInfo: def __init__(self) -> None: # 请求头 self.headers = { 'User-Agent': 'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/63.0.3239.132 Safari/537.36' } # 排行榜url self.rankUrl = "https://blog.csdn.net/phoenix/web/blog/all-rank?page={}&pageSize=20" # 按访问量排行的文章列表 self.mostViewArtical = "https://blog.csdn.net/community/home-api/v1/get-business-list?page=1&size=20&businessType=blog&orderby=ViewCount&noMore=false&username={}" self.userNames = [] self.titleList, self.viewCntList = [], []
交互完成后，再使用别的库进行数据分析，将两个过程分离开
用户名爬取
定义一个私有的初始化函数
def __initRankUsrName(self): usrNameList = [] for i in range(5): response = requests.get(url=self.rankUrl.format(i), headers=self.headers) response.encoding = 'utf-8' response.raise_for_status() soup = BeautifulSoup(response.text, 'html.parser') information = json.loads(str(soup)) for item in information['data']['allRankListItem']: usrNameList.append(item['userName']) return usrNameList
这里获取用户名主要是为了动态生成第二个API
文章爬取
再定义一个私有函数，输入参数是用户名列表：
def __initArticalInfo(self, usrList): titleList = [] viewCntList = [] for name in usrList: url = self.mostViewArtical.format(name) # print(url) response = requests.get(url=url, headers=self.headers) response.encoding = 'utf-8' response.raise_for_status() titleList.extend(re.findall(r"\"title\":\"(.*?)\"", response.text)) viewCntList.extend(re.findall(r"\"viewCount\":(.*?),", response.text)) return titleList, viewCntList
这里我使用正则表达式直接处理字符串，并返回文章标题列表、访问量列表。可以随便访问一个API做实验，这里以我的用户名为例，可以看到要获取文章标题就是以\"title\":\"(.*?)\"去匹配，其中\用于转义；要获取访问量就是以\"viewCount\":(.*?),去匹配，访问数字没有加引号。
事实上，用正则匹配不需要将返回的字符串加载为Json字典，可能有更快的处理效率(但不如json灵活)
这个爬虫类就设计好了，完整代码如下：
class GetInfo: def __init__(self) -> None: # 请求头 self.headers = { 'User-Agent': 'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/63.0.3239.132 Safari/537.36' } # 排行榜url self.rankUrl = "https://blog.csdn.net/phoenix/web/blog/all-rank?page={}&pageSize=20" # 按访问量排行的文章列表 self.mostViewArtical = "https://blog.csdn.net/community/home-api/v1/get-business-list?page=1&size=20&businessType=blog&orderby=ViewCount&noMore=false&username={}" self.userNames = self.__initRankUsrName() self.titleList, self.viewCntList = self.__initArticalInfo( self.userNames) def __initArticalInfo(self, usrList): titleList = [] viewCntList = [] for name in usrList: url = self.mostViewArtical.format(name) # print(url) response = requests.get(url=url, headers=self.headers) response.encoding = 'utf-8' response.raise_for_status() titleList.extend(re.findall(r"\"title\":\"(.*?)\"", response.text)) viewCntList.extend( re.findall(r"\"viewCount\":(.*?),", response.text)) return titleList, viewCntList def __initRankUsrName(self): usrNameList = [] for i in range(5): response = requests.get(url=self.rankUrl.format(i), headers=self.headers) response.encoding = 'utf-8' response.raise_for_status() soup = BeautifulSoup(response.text, 'html.parser') information = json.loads(str(soup)) for item in information['data']['allRankListItem']: usrNameList.append(item['userName']) return usrNameList info = GetInfo()
使用也很方便，只需要实例化调用其中的列表属性即可。
数据分析
数据存储
将文本数据存成csv格式，先设计表头：
if not os.path.exists("articalInfo.csv"): #创建存储csv文件存储数据 with open('articalInfo.csv', "w", encoding="utf-8-sig", newline='') as f: csv_head = csv.writer(f) csv_head.writerow(['title', 'viewCnt'])
注意编码格式为utf-8-sig，否则会乱码
接下来存数据：
length = len(info.titleList) for i in range(length): if info.titleList[i]: with open('articalInfo.csv', 'a+', encoding='utf-8-sig') as f: f.write(info.titleList[i] + ',' + info.viewCntList[i] + '\n')
总体数据可视化
新建一个模块专门用于可视化数据，与爬虫分离开，因为爬虫是慢IO过程，会影响调试效率，后面可以试试用协程来处理爬虫。
首先，把爬虫的信息读取到txt文件去
df = pd.read_csv('articalInfoNor.csv', encoding='utf-8-sig',usecols=['title', 'viewCnt']) titleList = '，'.join(df['title'].values) with open('text.txt','a+', encoding='utf-8-sig') as f: f.writelines(titleList)
如何返回分词结果:
def getKeyWordText(): # 读取文件信息 file = open(path.join(path.dirname(__file__), 'text.txt'), encoding='utf-8-sig').read() return ' '.join(jieba.cut(file))
借助词云库可视化一下:
bg_pic = imread('2.jpg') #生成词云 wordcloud = WordCloud(font_path=r'C:\Windows\Fonts\simsun.ttc',mask=bg_pic,background_color='white',scale=1.5).generate(text) image_colors = ImageColorGenerator(bg_pic) #显示词云图片 plt.imshow(wordcloud) plt.axis('off') plt.show() #保存图片 wordcloud.to_file('test.jpg')
这个大大的“的”是什么鬼？显然高频关键词里有太多语气助词、连接词，我们最好设置一个停用词列表把这些明显不需要的词屏蔽掉。我这里采用修饰器的方法让代码更简洁，关于修饰器的内容可以参考Python修饰器
def splitText(mode): stopWords = ["的","与","和","建议","收藏","使用","了","实现","我","中","你","在","之","年","月","日"] def warpper(func): def warp(): textSplit = func() if mode: temp = [word for word in textSplit if word not in stopWords] return ' '.join(temp) else: return ' '.join(textSplit) return warp return warpper
当mode=True时启用屏蔽，否则关闭屏蔽，那么之前的函数应该修改为:
# 返回关键词文本 @splitText(False) def getKeyWordText(): # 读取文件信息 file = open(path.join(path.dirname(__file__), 'text.txt'), encoding='utf-8-sig').read() return jieba.cut(file)
再来一次：
现在就正常多了。可以看到Python和Java是绝对的领先，之后是各位总结的方法论等等，算法的词频反而不高？
数据分组
我把数据进一步分层为
1、访问量>10W
2、访问量5W~10W
3、访问量1W~5W
4、访问量5K~1W
5、访问量5K以下
先来看看数据分布情况：
我猜如果分段分得再细一点可能趋于正态分布~
分组可视化看看：
>10W的词云
5~10W的词云
1~5W的词云
5k~1W的词云
感觉从这里开始更百花齐放一些，似乎也更关注具体问题的解决
5k以下的词云
不得不感叹python在每个阶段都是牌面
完整代码
import requests from bs4 import BeautifulSoup import os, json, re, csv class GetInfo: def __init__(self) -> None: # 请求头 self.headers = { 'User-Agent': 'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/63.0.3239.132 Safari/537.36' } # 排行榜url self.rankUrl = "https://blog.csdn.net/phoenix/web/blog/all-rank?page={}&pageSize=20" # 按访问量排行的文章列表 self.mostViewArtical = "https://blog.csdn.net/community/home-api/v1/get-business-list?page=1&size=20&businessType=blog&orderby=ViewCount&noMore=false&username={}" self.userNames = self.__initRankUsrName() self.titleList, self.viewCntList = self.__initArticalInfo( self.userNames) def __initArticalInfo(self, usrList): titleList = [] viewCntList = [] for name in usrList: url = self.mostViewArtical.format(name) # print(url) response = requests.get(url=url, headers=self.headers) response.encoding = 'utf-8' response.raise_for_status() titleList.extend(re.findall(r"\"title\":\"(.*?)\"", response.text)) viewCntList.extend( re.findall(r"\"viewCount\":(.*?),", response.text)) return titleList, viewCntList def __initRankUsrName(self): usrNameList = [] for i in range(5): response = requests.get(url=self.rankUrl.format(i), headers=self.headers) response.encoding = 'utf-8' response.raise_for_status() soup = BeautifulSoup(response.text, 'html.parser') information = json.loads(str(soup)) for item in information['data']['allRankListItem']: usrNameList.append(item['userName']) return usrNameList info = GetInfo() if not os.path.exists("articalInfo.csv"): #创建存储csv文件存储数据 with open('articalInfo.csv', "w", encoding="utf-8-sig", newline='') as f: csv_head = csv.writer(f) csv_head.writerow(['title', 'viewCnt']) length = len(info.titleList) for i in range(length): if info.titleList[i]: with open('articalInfo.csv', 'a+', encoding='utf-8-sig') as f: f.write(info.titleList[i] + ',' + info.viewCntList[i] + '\n')
from wordcloud import WordCloud,ImageColorGenerator import matplotlib.pyplot as plt from imageio import imread import jieba import pandas as pd from os import path df = pd.read_csv('articalInfoCom.csv', encoding='utf-8-sig',usecols=['title', 'viewCnt']) titleList = '，'.join(df['title'].values) with open('text.txt','a+', encoding='utf-8-sig') as f: f.writelines(titleList) def splitText(mode): stopWords = ["的","与","和","建议","收藏","使用","了","实现","我","中","你","在","之","年","月","日"] def warpper(func): def warp(): textSplit = func() if mode: temp = [word for word in textSplit if word not in stopWords] return ' '.join(temp) else: return ' '.join(textSplit) return warp return warpper # 返回关键词文本 @splitText(True) def getKeyWordText(): # 读取文件信息 file = open(path.join(path.dirname(__file__), 'text.txt'), encoding='utf-8-sig').read() return jieba.cut(file) text = getKeyWordText() #读取txt文件、背景图片 bg_pic = imread('2.jpg') #生成词云 wordcloud = WordCloud(font_path=r'C:\Windows\Fonts\simsun.ttc',mask=bg_pic,background_color='white',scale=1.5).generate(text) image_colors = ImageColorGenerator(bg_pic) #显示词云图片 plt.imshow(wordcloud) plt.axis('off') plt.show() #保存图片 wordcloud.to_file('test.jpg')
