Ruby多进程并行抓取网页

 

想研究一下各大网站首页有多少个链接，于是用Ruby写了一个脚本，用Beanstalk消息队列，把任务放入Beanstalk中，同时开启多个子进程，如20个，并行运行，同时抓取网页，先抓取完成的进程再次读取Beanstalk任务，继续执行，直到没有任务后，进程退出，主进程等待所有子进程退出后，打印抓取的消息。 
Ruby的多线程实际是只能跑在单cpu上，并且同一时刻cpu只处理一个线程，所以采用多进程抓取，消息队列采用最简单的Beanstalk，需要安装Beanstalkd服务。 
示例代码主要抓取电商网站测试。 
代码可以直接运行，需要ruby 1.9版本，1.9一下要稍微修改才能运行。

 

[代码] [Ruby]代码

001 #!/usr/bin/env ruby
002 #encoding: UTF-8
003
004 # 抓取每一个站点的首页链接数量
005 # require 'rubygems'            # 1.8.7
006 require 'ap'                # gem install awesome_print
007 require 'json'
008 require 'net/http'
009 require 'nokogiri'          # gem install nokogiri
010 require 'forkmanager'       # gem install parallel-forkmanager
011 require 'beanstalk-client'  # gem install beanstalk-client
012
013 class MultipleCrawler
014
015     class Crawler
016         def initialize(user_agent, redirect_limit=1)
017             @user_agent = user_agent
018             @redirect_limit = redirect_limit
019             @timeout = 20
020         end
021         attr_accessor :user_agent, :redirect_limit, :timeout
022
023         def fetch(website)
024             print "Pid:#{Process.pid}, fetch: #{website}\n"
025             redirect, url = @redirect_limit, website
026             start_time = Time.now
027             redirecting = false
028             begin
029                 begin
030                     uri = URI.parse(url)
031                     req = Net::HTTP::Get.new(uri.path)
032                     req.add_field('User-Agent', @user_agent)
033                     res = Net::HTTP.start(uri.host, uri.port) do |http|
034                         http.read_timeout = @timeout
035                         http.request(req)
036                     end
037                     if res.header['location'] # 遇到重定向，则url设定为location，再次抓取
038                         url = res.header['location']
039                         redirecting = true
040                     end
041                     redirect -= 1
042                 end while redirecting and redirect>=0
043                 opened_time = (Time.now - start_time).round(4) # 统计打开网站耗时
044                 encoding = res.body.scan(/<meta.+?charset=["'\s]*([\w-]+)/i)[0]
045                 encoding = encoding ? encoding[0].upcase : 'GB18030'
046                 html = 'UTF-8'==encoding ? res.body : res.body.force_encoding('GB2312'==encoding || 'GBK'==encoding ? 'GB18030' : encoding).encode('UTF-8')
047                 doc = Nokogiri::HTML(html)
048                 processed_time = (Time.now - start_time - opened_time).round(4)# 统计分析链接耗时, 1.8.7, ('%.4f' % float).to_f 替换 round(4)
049                 [opened_time, processed_time, doc.css('a[@href]').size, res.header['server']]
050             rescue =>e
051                 e.message
052             end
053         end
054     end
055
056     def initialize(websites, beanstalk_jobs, pm_max=1, user_agent='', redirect_limit=1)
057         @websites = websites                # 网址数组
058         @beanstalk_jobs = beanstalk_jobs    # beanstalk服务器地址和管道参数
059         @pm_max = pm_max                    # 最大并行运行进程数
060         @user_agent = user_agent            # user_agent 伪装成浏览器访问
061         @redirect_limit = redirect_limit    # 允许最大重定向次数
062
063         @ipc_reader, @ipc_writer = IO.pipe # 缓存结果的 ipc 管道
064     end
065
066     attr_accessor :user_agent, :redirect_limit
067
068     def init_beanstalk_jobs # 准备beanstalk任务
069         beanstalk = Beanstalk::Pool.new(*@beanstalk_jobs)
070         #清空beanstalk的残留消息队列
071         begin
072             while job = beanstalk.reserve(0.1)
073                 job.delete
074             end
075         rescue Beanstalk::TimedOut
076             print "Beanstalk queues cleared!\n"
077         end
078         @websites.size.times{|i| beanstalk.put(i)} # 将所有的任务压栈
079         beanstalk.close
080         rescue => e
081             puts e
082             exit
083     end
084
085     def process_jobs # 处理任务
086         start_time = Time.now
087         pm = Parallel::ForkManager.new(@pm_max)
088         @pm_max.times do |i|
089             pm.start(i) and next # 启动后，立刻 next 不会等待进程执行完，这样才可以并行运算
090             beanstalk = Beanstalk::Pool.new(*@beanstalk_jobs)
091             @ipc_reader.close    # 关闭读取管道，子进程只返回数据
092             loop{
093                 begin
094                     job = beanstalk.reserve(0.1) # 检测超时为0.1秒，因为任务以前提前压栈
095                     index = job.body
096                     job.delete
097                     website = @websites[index.to_i]
098                     result = Crawler.new(@user_agent).fetch(website)
099                     @ipc_writer.puts( ({website=>result}).to_json )
100                 rescue Beanstalk::DeadlineSoonError, Beanstalk::TimedOut, SystemExit, Interrupt
101                     break
102                 end
103             }
104             @ipc_writer.close
105             pm.finish(0)
106         end
107         @ipc_writer.close
108         begin
109             pm.wait_all_children        # 等待所有子进程处理完毕
110         rescue SystemExit, Interrupt    # 遇到中断，打印消息
111             print "Interrupt wait all children!\n"
112         ensure
113             results = read_results
114             ap results, :indent => -4 , :index=>false # 打印处理结果
115             print "Process end, total: #{@websites.size}, crawled: #{results.size}, time: #{'%.4f' % (Time.now - start_time)}s.\n"
116         end
117     end
118
119     def read_results # 通过管道读取子进程抓取返回的数据
120         results = {}
121         while result = @ipc_reader.gets
122             results.merge! JSON.parse(result)
123         end
124         @ipc_reader.close
125         results
126     end
127
128     def run # 运行入口
129         init_beanstalk_jobs
130         process_jobs
131     end
132 end
133
134 websites = %w(
135 http://www.51buy.com/ http://www.360buy.com/ http://www.tmall.com/ http://www.taobao.com/
136 http://china.alibaba.com/ http://www.paipai.com/ http://shop.qq.com/ http://www.lightinthebox.com/
137 http://www.amazon.cn/ http://www.newegg.com.cn/ http://www.vancl.com/ http://www.yihaodian.com/
138 http://www.dangdang.com/ http://www.m18.com/ http://www.suning.com/ http://www.hstyle.com/
139 )
140 beanstalk_jobs = [['localhost:11300'],'crawler-jobs']
141 user_agent = 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10.7; rv:13.0) Gecko/20100101 Firefox/13.0'
142 pm_max = 10
143
144 MultipleCrawler.new(websites, beanstalk_jobs, pm_max, user_agent).run