代码改变世界

python爬虫多进程,多线程,协程以及组合应用的效率对比(multiprocessing库)--以爬取单本小说全文为例

2020-03-23 09:56  冻雨冷雾  阅读(807)  评论(0)  编辑  收藏  举报

本篇将测试爬取单本小说下： 利用多进程，多线程，协程，以及多进程加多线程，多进程加协程组合应用的效率。

以爬取--笔趣阁--大道争锋为例，测试相关组合的性能。

 

  • 多线程

代码如下：

# -*- coding: utf-8 -*-
"""
Created on Wed Mar  4 10:39:55 2020

@author: wenzhe.tian

多进程+多线程
多进程+协程
"""

book_name_list=['大道争锋']




####### 开始工作
import time
from concurrent.futures import ThreadPoolExecutor
import requests
from lxml import etree
import os
import urllib.parse as parse


save_path='D:\\bqg_novel\\'
headers = {'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.95 Safari/537.36'}
target_url='https://m.52bqg.com'
try:
    os.mkdir(save_path)
except:
    pass


### 定义两个函数获取书的url和章节的url
def get_chapter_content(i): # 根据章节url返回章节内容
    chapter_now=requests.get(target_url+i,headers)
    chapter_now.encoding='gbk'
    chapter_now=chapter_now.text; #源码
    #    chapter_now_ori=chapter_now #检测用
    chapter_now=etree.HTML(chapter_now)
    chapter_content='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
    next_page_num=1
    while '下一页' in chapter_now.xpath('//div[@class="nr_page"]//td[@class="next"]/descendant::text()'):
        chapter_content=chapter_content.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        next_page_num=next_page_num+1;
        chapter_now=requests.get(target_url+i.replace('.html','_'+str(next_page_num)+'.html'),headers)
        chapter_now.encoding='gbk'
        chapter_now=chapter_now.text; #源码
        #        chapter_now_ori=chapter_now #检测用
        chapter_now=etree.HTML(chapter_now)
        chapter_content_next='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
        chapter_content_next=chapter_content_next.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        chapter_content=chapter_content+chapter_content_next;
    return chapter_content


def get_chapter_link(i): ########## 确定章节的数目,爬取所有章节link ######
    global url_all,headers
    if i==0:
        req_next=requests.get(url_all,headers)
    else:
        req_next=requests.get(url_all+'/'+str(i+1),headers)
    req_next.encoding='gbk'
    html_next=etree.HTML(req_next.text)
    chapter_name_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]//a/descendant::text()|//ul[@class="last9"]//li//a/descendant::text()')
    chapter_url_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]/a/@href|//ul[@class="last9"]//li/a/@href')
    chapter_name=chapter_name_next[1:]
    chapter_url=chapter_url_next[1:]
    return chapter_name,chapter_url



################################# 对于所有章节的url内容爬取 #####################################
novel=[]

for k in book_name_list:
    start=time.time()
    url='https://m.52bqg.com/modules/article/waps.php?searchtype=articlename&searchkey='+parse.quote(k,encoding="gbk")+'&t_btnsearch='
    req=requests.get(url,headers)
    req.encoding='gbk'
    if 'book_' in req.url and 'search' not in req.url: #搜索结果不是清单则直接开始爬
        url_all=req.url
        url_all=url_all.replace('book','chapters')
    else: #是清单则判断是否有完全匹配项，若无则只爬榜1
        search_result=req.text
        html_search=etree.HTML(search_result)
        search_book=html_search.xpath('//div[@class="article"]/a/text()')
        search_book_url=html_search.xpath('//div[@class="article"]/a[1]/@href')
        if k in search_book:
            url_all=target_url+search_book_url[search_book.index(k)]
            url_all=url_all.replace('book','chapters')
        else:
            url_all=target_url+search_book_url[0]
            url_all=url_all.replace('book','chapters')

    # 根据书名判断章节页数
    req_all=requests.get(url_all,headers)
    req_all.encoding='gbk'
    html_all=etree.HTML(req_all.text)
    chapter_page_all=html_all.xpath('//table[@class="page-book"]//td/a/@href')
    chapter_page_all=chapter_page_all[1].split('/')
    chapter_page_all=int(chapter_page_all[-1])
    # 开始多线程抓取
    with ThreadPoolExecutor(250) as executor:
        # 根据章节页数,得到章节url
        chapter=list(executor.map(get_chapter_link,range(chapter_page_all)))
        chapter=list(zip(*chapter))
        chapter_url=list(chapter[1])
        chapter_name=list(chapter[0])
        chapter_url = sum(chapter_url, [])
        chapter_name = sum(chapter_name, [])
        chapter_all=list(executor.map(get_chapter_content,chapter_url))
    end=time.time()
    print("耗时: "+str(int(end-start))+'秒') #计时统计
    for i in range(len(chapter_all)):
        chapter_all[i]=chapter_name[i]+'\n'+chapter_all[i]
    target='/n'.join(chapter_all)
    f = open(save_path+'\\'+k+'.txt','a+',encoding='utf-8')
    f.read()
    f.write(target)
    f.close()
    print(k+'已完成')
View Code

耗时: 70s

 

  • 协程

代码如下：

# -*- coding: utf-8 -*-
"""
Created on Wed Mar  4 10:39:55 2020

@author: wenzhe.tian



"""

book_name_list=['大道争锋']




####### 开始工作

import gevent
from gevent import monkey,pool
pool=pool.Pool(200)
monkey.patch_all(thread=False)
import requests
import time
from lxml import etree
import os
import urllib.parse as parse


save_path='D:\\bqg_novel\\'
headers = {'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.95 Safari/537.36'}
target_url='https://m.52bqg.com'
try:
    os.mkdir(save_path)
except:
    pass


### 定义两个函数获取书的url和章节的url
def get_chapter_content(i): # 根据章节url返回章节内容
    chapter_now=requests.get(target_url+i,headers)
    chapter_now.encoding='gbk'
    chapter_now=chapter_now.text; #源码
    #    chapter_now_ori=chapter_now #检测用
    chapter_now=etree.HTML(chapter_now)
    chapter_content='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
    next_page_num=1
    while '下一页' in chapter_now.xpath('//div[@class="nr_page"]//td[@class="next"]/descendant::text()'):
        chapter_content=chapter_content.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        next_page_num=next_page_num+1;
        chapter_now=requests.get(target_url+i.replace('.html','_'+str(next_page_num)+'.html'),headers)
        chapter_now.encoding='gbk'
        chapter_now=chapter_now.text; #源码
        #        chapter_now_ori=chapter_now #检测用
        chapter_now=etree.HTML(chapter_now)
        chapter_content_next='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
        chapter_content_next=chapter_content_next.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        chapter_content=chapter_content+chapter_content_next;
    return chapter_content


def get_chapter_link(i): ########## 确定章节的数目,爬取所有章节link ######
    global url_all,headers
    if i==0:
        req_next=requests.get(url_all,headers)
    else:
        req_next=requests.get(url_all+'/'+str(i+1),headers)
    req_next.encoding='gbk'
    html_next=etree.HTML(req_next.text)
    chapter_name_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]//a/descendant::text()|//ul[@class="last9"]//li//a/descendant::text()')
    chapter_url_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]/a/@href|//ul[@class="last9"]//li/a/@href')
    chapter_name=chapter_name_next[1:]
    chapter_url=chapter_url_next[1:]
    return chapter_name,chapter_url



################################# 对于所有章节的url内容爬取 #####################################
novel=[]

for k in book_name_list:
    start=time.time()
    url='https://m.52bqg.com/modules/article/waps.php?searchtype=articlename&searchkey='+parse.quote(k,encoding="gbk")+'&t_btnsearch='
    req=requests.get(url,headers)
    req.encoding='gbk'
    if 'book_' in req.url and 'search' not in req.url: #搜索结果不是清单则直接开始爬
        url_all=req.url
        url_all=url_all.replace('book','chapters')
    else: #是清单则判断是否有完全匹配项，若无则只爬榜1
        search_result=req.text
        html_search=etree.HTML(search_result)
        search_book=html_search.xpath('//div[@class="article"]/a/text()')
        search_book_url=html_search.xpath('//div[@class="article"]/a[1]/@href')
        if k in search_book:
            url_all=target_url+search_book_url[search_book.index(k)]
            url_all=url_all.replace('book','chapters')
        else:
            url_all=target_url+search_book_url[0]
            url_all=url_all.replace('book','chapters')

    # 根据书名判断章节页数
    req_all=requests.get(url_all,headers)
    req_all.encoding='gbk'
    html_all=etree.HTML(req_all.text)
    chapter_page_all=html_all.xpath('//table[@class="page-book"]//td/a/@href')
    chapter_page_all=chapter_page_all[1].split('/')
    chapter_page_all=int(chapter_page_all[-1])
    # 开始协程抓取
    g_list=list(map(lambda x:gevent.spawn(get_chapter_link, x),range(chapter_page_all)))
    gevent.joinall(g_list)
    chapter=[]
    for g in g_list:
        chapter.append(g.value)
    chapter=list(zip(*chapter))
    chapter_url=list(chapter[1])
    chapter_name=list(chapter[0])
    chapter_url = sum(chapter_url, [])
    chapter_name = sum(chapter_name, [])
    g_list=list(map(lambda x:gevent.spawn(get_chapter_content, x),chapter_url))
    gevent.joinall(g_list)
    chapter_all=[]
    for g in g_list:
        chapter_all.append(g.value)
    end=time.time()
    print("耗时: "+str(int(end-start))+'秒') #计时统计
    for i in range(len(chapter_all)):
        chapter_all[i]=chapter_name[i]+'\n'+chapter_all[i]
    target='/n'.join(chapter_all)
    f = open(save_path+'\\'+k+'.txt','a+',encoding='utf-8')
    f.read()
    f.write(target)
    f.close()
    print(k+'已完成')
View Code

耗时： 103s

 

  • 多进程调用多线程

代码如下：

# -*- coding: utf-8 -*-
"""
Created on Wed Mar  4 10:39:55 2020

@author: wenzhe.tian

多进程+多线程

"""


####### 开始工作
import time
from concurrent.futures import ThreadPoolExecutor
import requests
from lxml import etree
import os
import urllib.parse as parse
from multiprocessing import Pool
book_name_list=['斗罗大陆3龙王传说']
save_path='D:\\bqg_novel\\'
headers = {'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.95 Safari/537.36'}
target_url='https://m.52bqg.com'
try:
    os.mkdir(save_path)
except:
    pass
novel=[]

##### 定义两个函数获取书的url和章节的url
def get_chapter_content(i): # 根据章节url返回章节内容
    chapter_now=requests.get(target_url+i,headers)
    chapter_now.encoding='gbk'
    chapter_now=chapter_now.text; #源码
    #    chapter_now_ori=chapter_now #检测用
    chapter_now=etree.HTML(chapter_now)
    chapter_content='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
    next_page_num=1
    while '下一页' in chapter_now.xpath('//div[@class="nr_page"]//td[@class="next"]/descendant::text()'):
        chapter_content=chapter_content.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        next_page_num=next_page_num+1;
        chapter_now=requests.get(target_url+i.replace('.html','_'+str(next_page_num)+'.html'),headers)
        chapter_now.encoding='gbk'
        chapter_now=chapter_now.text; #源码
        #        chapter_now_ori=chapter_now #检测用
        chapter_now=etree.HTML(chapter_now)
        chapter_content_next='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
        chapter_content_next=chapter_content_next.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        chapter_content=chapter_content+chapter_content_next;
    return chapter_content


def get_chapter_link(link): ########## 确定章节的数目,爬取所有章节link ######
    i=link[0]
    url_all=link[1]
    if i==0:
        req_next=requests.get(url_all,headers)
    else:
        req_next=requests.get(url_all+str(i+1),headers)
    req_next.encoding='gbk'
    html_next=etree.HTML(req_next.text)
    chapter_name_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]//a/descendant::text()|//ul[@class="last9"]//li//a/descendant::text()')
    chapter_url_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]/a/@href|//ul[@class="last9"]//li/a/@href')
    chapter_name=chapter_name_next[1:]
    chapter_url=chapter_url_next[1:]
    return chapter_name,chapter_url

def run_proc(page):
    with ThreadPoolExecutor(200) as executor:
        # 根据章节页数,得到章节url
        i=list(page[0])
        for k in range(len(i)):
            i[k]=[i[k],page[1]]
        chapter=list(executor.map(get_chapter_link,i))
        chapter=list(zip(*chapter))
        chapter_url=list(chapter[1])
        chapter_name=list(chapter[0])
        chapter_url = sum(chapter_url, [])
        chapter_name = sum(chapter_name, [])
        chapter_all=list(executor.map(get_chapter_content,chapter_url))
        for i in range(len(chapter_all)):
            chapter_all[i]=chapter_name[i]+'\n'+chapter_all[i]
    return chapter_all

################################# 对于所有章节的url内容爬取 #####################################
if __name__ == '__main__':
    for k in book_name_list:
        start=time.time()
        url='https://m.52bqg.com/modules/article/waps.php?searchtype=articlename&searchkey='+parse.quote(k,encoding="gbk")+'&t_btnsearch='
        req=requests.get(url,headers)
        req.encoding='gbk'
        if 'book_' in req.url and 'search' not in req.url: #搜索结果不是清单则直接开始爬
            url_all=req.url
            url_all=url_all.replace('book','chapters')
        else: #是清单则判断是否有完全匹配项，若无则只爬榜1
            search_result=req.text
            html_search=etree.HTML(search_result)
            search_book=html_search.xpath('//div[@class="article"]/a/text()')
            search_book_url=html_search.xpath('//div[@class="article"]/a[1]/@href')
            if k in search_book:
                url_all=target_url+search_book_url[search_book.index(k)]
                url_all=url_all.replace('book','chapters')
            else:
                url_all=target_url+search_book_url[0]
                url_all=url_all.replace('book','chapters')

        # 根据书名判断章节页数
        req_all=requests.get(url_all,headers)
        req_all.encoding='gbk'
        html_all=etree.HTML(req_all.text)
        chapter_page_all=html_all.xpath('//table[@class="page-book"]//td/a/@href')
        chapter_page_all=chapter_page_all[1].split('/')
        chapter_page_all=int(chapter_page_all[-1])

        # 对小说页使用进程处理
        count=0
        page_list=[]
        while count<chapter_page_all:
            next_count=count+10
            if next_count>chapter_page_all:
                next_count=chapter_page_all;
            page_list.append([range(count,next_count),url_all])
            count=count+10

        p = Pool(4)
        result=p.map(run_proc, page_list)
        p.close()
        p.join()
        chapter_all= sum(result,[])
        end=time.time()
        print("耗时: "+str(int(end-start))+'秒') #计时统计

        target='/n'.join(chapter_all)
        f = open(save_path+'\\'+k+'.txt','a+',encoding='utf-8')
        f.read()
        f.write(target)
        f.close()
        print(k+'已完成')
View Code

 

耗时： 40s

 

  • 多进程调用协程

代码如下：

# -*- coding: utf-8 -*-
"""
Created on Wed Mar  4 10:39:55 2020

@author: wenzhe.tian

多进程+多线程
多进程+协程
"""


####### 开始工作
import gevent
from gevent import monkey
monkey.patch_all(thread=False)
import requests
import time
from lxml import etree
import os
import urllib.parse as parse
from multiprocessing import Pool

book_name_list=['大道争锋']
save_path='D:\\bqg_novel\\'
headers = {'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_12_0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/55.0.2883.95 Safari/537.36'}
target_url='https://m.52bqg.com'
try:
    os.mkdir(save_path)
except:
    pass
novel=[]

##### 定义两个函数获取书的url和章节的url
def get_chapter_content(i): # 根据章节url返回章节内容
    chapter_now=requests.get(target_url+i,headers)
    chapter_now.encoding='gbk'
    chapter_now=chapter_now.text; #源码
    #    chapter_now_ori=chapter_now #检测用
    chapter_now=etree.HTML(chapter_now)
    chapter_content='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
    next_page_num=1
    while '下一页' in chapter_now.xpath('//div[@class="nr_page"]//td[@class="next"]/descendant::text()'):
        chapter_content=chapter_content.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        next_page_num=next_page_num+1;
        chapter_now=requests.get(target_url+i.replace('.html','_'+str(next_page_num)+'.html'),headers)
        chapter_now.encoding='gbk'
        chapter_now=chapter_now.text; #源码
        #        chapter_now_ori=chapter_now #检测用
        chapter_now=etree.HTML(chapter_now)
        chapter_content_next='\n'.join(chapter_now.xpath('//div[@id="nr1"]/descendant::text()'))
        chapter_content_next=chapter_content_next.replace('本章未完，点击下一页继续阅读','').replace('-->>','').replace('&n','')
        chapter_content=chapter_content+chapter_content_next;
    return chapter_content


def get_chapter_link(link): ########## 确定章节的数目,爬取所有章节link ######
    i=link[0]
    url_all=link[1]
    if i==0:
        req_next=requests.get(url_all,headers)
    else:
        req_next=requests.get(url_all+str(i+1),headers)
    req_next.encoding='gbk'
    html_next=etree.HTML(req_next.text)
    chapter_name_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]//a/descendant::text()|//ul[@class="last9"]//li//a/descendant::text()')
    chapter_url_next=html_next.xpath('//ul[@class="last9"]//li[@class="even"]/a/@href|//ul[@class="last9"]//li/a/@href')
    chapter_name=chapter_name_next[1:]
    chapter_url=chapter_url_next[1:]
    return chapter_name,chapter_url

def run_proc(page):
    # 根据章节页数,得到章节url
    i=list(page[0])
    for k in range(len(i)):
        i[k]=[i[k],page[1]]
    g_list=list(map(lambda x:gevent.spawn(get_chapter_link, x),i))
    gevent.joinall(g_list)
    chapter=[]
    for g in g_list:
        chapter.append(g.value)
    chapter=list(zip(*chapter))
    chapter_url=list(chapter[1])
    chapter_name=list(chapter[0])
    chapter_url = sum(chapter_url, [])
    chapter_name = sum(chapter_name, [])
    g_list=list(map(lambda x:gevent.spawn(get_chapter_content, x),chapter_url))
    gevent.joinall(g_list)
    chapter_all=[]
    for g in g_list:
        chapter_all.append(g.value)

    for i in range(len(chapter_all)):
        chapter_all[i]=chapter_name[i]+'\n'+chapter_all[i]
    return chapter_all

################################# 对于所有章节的url内容爬取 #####################################
if __name__ == '__main__':
    for k in book_name_list:
        start=time.time()
        url='https://m.52bqg.com/modules/article/waps.php?searchtype=articlename&searchkey='+parse.quote(k,encoding="gbk")+'&t_btnsearch='
        req=requests.get(url,headers)
        req.encoding='gbk'
        if 'book_' in req.url and 'search' not in req.url: #搜索结果不是清单则直接开始爬
            url_all=req.url
            url_all=url_all.replace('book','chapters')
        else: #是清单则判断是否有完全匹配项，若无则只爬榜1
            search_result=req.text
            html_search=etree.HTML(search_result)
            search_book=html_search.xpath('//div[@class="article"]/a/text()')
            search_book_url=html_search.xpath('//div[@class="article"]/a[1]/@href')
            if k in search_book:
                url_all=target_url+search_book_url[search_book.index(k)]
                url_all=url_all.replace('book','chapters')
            else:
                url_all=target_url+search_book_url[0]
                url_all=url_all.replace('book','chapters')

        # 根据书名判断章节页数
        req_all=requests.get(url_all,headers)
        req_all.encoding='gbk'
        html_all=etree.HTML(req_all.text)
        chapter_page_all=html_all.xpath('//table[@class="page-book"]//td/a/@href')
        chapter_page_all=chapter_page_all[1].split('/')
        chapter_page_all=int(chapter_page_all[-1])

        # 对小说页使用进程处理
        count=0
        page_list=[]
        while count<chapter_page_all:
            next_count=count+10
            if next_count>chapter_page_all:
                next_count=chapter_page_all;
            page_list.append([range(count,next_count),url_all])
            count=count+10

        p = Pool(4)
        result=p.map(run_proc, page_list)
        p.close()
        p.join()
        chapter_all= sum(result,[])
        end=time.time()
        print("耗时: "+str(int(end-start))+'秒') #计时统计

        target='/n'.join(chapter_all)
        f = open(save_path+'\\'+k+'.txt','a+',encoding='utf-8')
        f.read()
        f.write(target)
        f.close()
        print(k+'已完成')
View Code

耗时： 60s

 

简单的看出多核应用大于单核，多线程当然好于无线程(不要受GIL锁对多线程误解，效率明显是提高的)

本测试中开多线程效率（线程数250）是要大于协程的。

 

下篇将会对全站小说爬取的多进程多线程调用总结以及scrapy应用的对比。

欢迎交流指正。有任何疑问直接丢评论区。