基于PyHusky的分布式爬虫原理及实现

比眉伴天荒 2021-09-23 02:58 281阅读 0赞

转载自：happengft的博客

http://blog.csdn.net/happengft/article/details/69248401

爬虫是我们获取互联网数据的一个非常有效的方法，而分布式爬虫则是利用许多台机器协调工作来加快抓取数据效率的不二途径。分布式爬虫是由访问某些原始网址开始，在获取这些网址的内容后，根据某些规则从中提取出下一轮的爬虫网址。依此类推，直到将全部链接访问完毕或达到我们设置的某些限定方可结束。这里我们需要选定一个运算效率较高的机器作为master, 和一群正常的workers来协助。这个master要负责统筹url的爬取记录、去重和分配等任务，而workers则是通过同时请求不同网页来解决单台机器请求网页效率较低导致的时间瓶颈。这里，我们需要关注下面这几个方面：

1.  *如何高效的通过url请求服务器返回需要的pagesource, 例如借助代理躲避反爬机制或者使用selenium处理动态类网页等等。*
2.  *使用xpath可以快速从网页源码中得到包含于其中的链接，而通过正则表达式re可以过滤出满足爬取条件的url。*
3.  *为了避免重复爬取，需要将所有爬取过的链接放到历史记录中，所以下一轮的链接都是需要经过去重并且不包含于历史记录中的方可。*
4.  *有效信息抓取及保存，例如可以使用BeautifulSoup提取数据特征并将其存储到HDFS。*
5.  *整个过程中master需要掌控全局并紧密联系各个workers，使各部分协调完成自己的任务。*

--------------------

## 实现 ##

Husky作为高效的大数据分布式计算系统，其python实现的PyHusky也非常适合用来实现这里的分布式爬虫的设计。若是大家对爬虫基础了解不足或对PyHusky的使用不够熟悉，可以先查看前面的几篇博文：

1.  [*Python分布式爬虫前菜(1):关于静态动态网页内容获取的N种方法*][Python_1_N]
2.  [*Python分布式爬虫前菜(2):关于提取网页源码中特定信息的技巧*][Python_2]
3.  [*PyHusky接口功能及实例介绍*][PyHusky]

首先，创建一个CrawlerConfig的类来实现爬虫初始化和相关规则的设定：

#crawler_config.py
    import re
    class CrawlerConfig:
        def __init__(self):
            # Time out after 10 seconds
            self.timeout = 10
            # Seed set of urls
            self.urls_init = []
            self.proxies = {
       'http': '127.0.0.1:8118',}
    
            # urls that match these regex will be processed
            self.rules_include = []
            # after processing rules_include, urls that match
            # the following regex will be dropped
            self.rules_exclude = []
    
            self.history = None
            self.hdfspath_output = '/datasets/crawl/openrice/'
            self.parse_handlers = dict()
            #whether to use selenium
            self.is_selenium = False
    
        def set_rules_include(self, rules):
            self.rules_include = [re.compile(rule) for rule in rules]
        def set_rules_exclude(self, rules):
            self.rules_exclude = [re.compile(rule) for rule in rules]
        def add_parse_handler(self, pattern, handler):
            self.parse_handlers[re.compile(pattern)] = handler
    
     
      
       1
       2
       3
       4
       5
       6
       7
       8
       9
       10
       11
       12
       13
       14
       15
       16
       17
       18
       19
       20
       21
       22
       23
       24
       25
       26
       27
       28
       29

接着创建一个通用函数工具来具体实现网页请求、链接解析等具体的实现，命名为crawler\_utils.py：

#crawler_utils.py导入必要的头文件
    import requests
    import grequests
    import urllib2
    import sys
    import time
    import random
    import urlparse
    from lxml import html
    from selenium.webdriver.common.proxy import *
    from selenium import webdriver
     
      
       1
       2
       3
       4
       5
       6
       7
       8
       9
       10
       11

在crawler\_utils.py中添加请求网页源码的函数实现：

# 静态网页请求
    def download_content(url, to_secs, proxies):
        retries = 0
        timeout = 10
        while True:
            try:
                (proto, addr) = random.choice(proxies)
                agent = random.choice(['Mozilla/5.0 (Windows NT 6.1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/50.0.2661.102 Safari/537.36', 'Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/50.0.2661.75 Safari/537.36'])
                headers = {
       'User-Agent':agent}
                response=requests.get(url, timeout=timeout, proxies={proto:addr}, headers=headers)
                sys.stderr.write('\033[0;32mdownloaded: \033[0m'+url+'\033[0;31m'+'---retries: '+str(retries)+' times\033[0m\n')
                return response.content
            except requests.exceptions.RequestException as e:
                retries += 1
                timeout = min(to_secs, timeout+10)
                if retries == 10:
                    break
                sys.stderr.write('request failed: '+url+" "+str(e)+"\n")
                time.sleep(random.choice([5, 8]))
                continue
    # 动态网页请求解决方案，调用selenium和phantomjs
    def download_content_selenium(url, to_secs):
        retries = 0
        timeout = 20
        while True:
            try:
                phantomjs_path = '/data/opt/brew/lib/node_modules/phantomjs/lib/phantom/bin/phantomjs'
                service_args = [
                '--proxy=http://xxxx:8000/',
                '--proxy-type=http',
                ]
                driver = webdriver.PhantomJS(executable_path=phantomjs_path,service_args=service_args)
                driver.set_page_load_timeout(timeout)
                driver.set_script_timeout(timeout)
                driver.get(url)
                html_source = driver.page_source.encode('utf8')
                sys.stderr.write('\033[0;32mThis is selenium-----downloaded: \033[0m'+url+'\033[0;31m'+'---retries: '+str(retries)+' times\033[0m\n')
                driver.quit()
                return html_source
            except:
                try:
                    driver.quit()
                    sys.stderr.write('\033[0;33mDriver quit success '+str(retries)+'\033[0m')
                except:
                    sys.stderr.write('\033[0;31mQuit fail \033[0m')
                    pass
                retries += 1
                if retries == 4:
                    break
                timeout = min(to_secs, timeout+10)
                sys.stderr.write('request failed: '+url+" "+"\n")
                time.sleep(random.choice([5, 8]))
                continue
    
    
     
      
       1
       2
       3
       4
       5
       6
       7
       8
       9
       10
       11
       12
       13
       14
       15
       16
       17
       18
       19
       20
       21
       22
       23
       24
       25
       26
       27
       28
       29
       30
       31
       32
       33
       34
       35
       36
       37
       38
       39
       40
       41
       42
       43
       44
       45
       46
       47
       48
       49
       50
       51
       52
       53
       54
       55

在crawler\_utils.py中添加提取网页源码内所包含的链接的函数实现：

def urlprocess(base_url, url):
        return urlparse.urljoin(base_url, url.split('#')[0])
    def urlpredicate(url, rules_include, rules_exclude):
        if(not url.startswith('http')):
            return None
        included = False
        matched_url = ""
        for pattern in rules_include:
            if pattern.match(url) != None:
                included = True
                matched_url = pattern.match(url).group()
                break
        if not included:
            return None
        for pattern in rules_exclude:
            if pattern.match(url) != None:
                return None
        return matched_url
    #按规则返回特定格式的url
    def parse_browser_removetag_userules(_url, response, rules_include, rules_exclude):
        try:
            parsed_body=html.fromstring(response)
            joined_link = []
            for url in parsed_body.xpath('//a/@href'):
                url = urlpredicate(unicode(urlprocess(_url, url)), rules_include, rules_exclude)
                if url == None:
                    continue
                joined_link.append(unicode(urlprocess(_url, url)))
            return joined_link
        except:
            print 'parse failed: '+response
            return []
    
    
     
      
       1
       2
       3
       4
       5
       6
       7
       8
       9
       10
       11
       12
       13
       14
       15
       16
       17
       18
       19
       20
       21
       22
       23
       24
       25
       26
       27
       28
       29
       30
       31
       32
       33
       34

以上完成了相关设置类和网页源码请求和链接解析等函数工具，下面我将详细介绍如何使用PyHusky调用以上工具来完成分布式爬虫的具体实现，命名为crawler\_dist.py：

#crawler_dist.py
    import timeit
    import re
    import json
    #PyHusky
    import bindings.frontend as fe
    import bindings.frontend.env
    #crawler_utils.py存放在hcrawl文件夹中
    from hcrawl import crawler_utils as ct
    
    #为map()选择所需的静态或动态网页请求函数，默认is_selenium=False
    def mapper_download(url, timeout, proxies,is_selenium):
        if is_selenium:
            response=ct.download_content_selenium(url, timeout, proxies)
        else:
            response=ct.download_content(url, timeout, proxies)
        return (url, response)
    #为flat_map()返回对应网页包含的符合规则之链接列表 
    def mapper_parse(htmltuple, rules_include, rules_exclude):
        if(htmltuple[1] is not None):
            res=ct.parse_browser_removetag_userules(htmltuple[0], htmltuple[1], rules_include, rules_exclude)
            return res
        else:
            return []
    #为map()提取对应源码中有用的数据特征，如何提取在config中定义 
    def parse_html(config, url_html_pair):
        parse_handlers = config.parse_handlers
        def mapper((url, html)):
            for pattern in parse_handlers:
                if pattern.match(url) != None:
                    return parse_handlers[pattern](pattern.match(url).group(), html)
            return json.dumps({
       "type":"handlers_fail", "url":url})
        return url_html_pair.map(mapper)
    
    def log_msg(msg):
        print "[pyhusky-log]:"+msg
    
    
     
      
       1
       2
       3
       4
       5
       6
       7
       8
       9
       10
       11
       12
       13
       14
       15
       16
       17
       18
       19
       20
       21
       22
       23
       24
       25
       26
       27
       28
       29
       30
       31
       32
       33
       34
       35
       36
       37
       38

承上，下面的crawler\_run函数便使用到优秀的PyHusky，便是这里实现分布式爬虫的关键：

def crawler_run(config, existing_progress=None):
    #读取urls数据，如果是list则使用parallelize将urls分配到各个机器上
        if type(config.urls_init) is list:
            this_round_urls = fe.env.parallelize(config.urls_init).cache()
            history = fe.env.parallelize(config.urls_init).cache()
    #否则使用load从HDFS地址中读取数据
        if type(config.urls_init) is str:
            this_round_urls = fe.env.load(config.urls_init).cache()
            history = fe.env.load(config.history).cache()
        iter = 0
        start = timeit.default_timer()
        next_len = 1
    #爬虫进行到底，直到下一轮urls个数next_len为零
        while next_len:
            iter += 1
            log_msg("downloading")
    #调用map()，在各个机器上运行mapper_download函数，返回url和对应的pagesource 
            this_round_htmltuple = this_round_urls.map(lambda url: mapper_download(url, config.timeout, config.proxies, config.is_selenium)).cache()
            log_msg("urls: "+str(this_round_htmltuple.count()))
            log_msg("writing to hdfs")
    #parse_html()提取pagesource中的所需的数据特征，写到HDFS
            parse_html(config, this_round_htmltuple).write_to_hdfs(config.hdfspath_output+str(iter))
    #flat_map()可以通过mapper_parse返回的urls收集下一轮的urls
            next_round_urls = this_round_htmltuple.flat_map(lambda htmltuple: mapper_parse(htmltuple, config.rules_include, config.rules_exclude)) \
    #以下的map(),reduce_by_key(),去重
                .map(lambda url: (url, 1)) \
                .reduce_by_key(lambda a, b: a+b) \
                .map(lambda (url, cnt): url) \
                .cache()
            log_msg("next round urls count: " + str(next_round_urls.count()))
    #uncache()，释放内存
            this_round_htmltuple.uncache()
            log_msg("calculating difference with history")
            history = history.map(lambda x:x.decode('utf8')).cache()
    #使用difference(),与历史记录history对比，可以剔除已经爬取过的urls
            url_diff = next_round_urls.difference(history).cache()
    # 释放this_round_urls所占内存，并将url_diff赋值给新的this_round_urls
            this_round_urls.uncache()
            this_round_urls = url_diff.cache()
            log_msg("create new history")
    #将新的urls添加到历史记录中，达到每一轮都不会重复爬取
            next_round_history = history.concat(url_diff).cache()
            history.uncache()
            history = next_round_history
            print 'husky_crawler_info: round '+str(iter)+' speed: '+str(next_len)+' pages within '+str((timeit.default_timer()-start))+' seconds'
    #下一轮有效urls个数
            next_len = url_diff.count()
            url_diff.uncache()
            hlen = history.count()
            start = timeit.default_timer()
            print 'husky_crawler_info: history size: ', hlen-next_len, ', next round size: ', next_len
    #轮数限制，若把num_iters设置为负数，则可逃避限制
            if iter == config.num_iters:
                break
    
     
      
       1
       2
       3
       4
       5
       6
       7
       8
       9
       10
       11
       12
       13
       14
       15
       16
       17
       18
       19
       20
       21
       22
       23
       24
       25
       26
       27
       28
       29
       30
       31
       32
       33
       34
       35
       36
       37
       38
       39
       40
       41
       42
       43
       44
       45
       46
       47
       48
       49
       50
       51
       52
       53
       54
       55

至此，我们大概介绍了分布式爬虫框架的实现思路。为了更好的理解和分析，大家需要熟悉Pyhusky，这个可以参考我之前写的相关博文。接下来的博文，我将分别给出针对静态和动态两种类型的网站的数据获取的具体操作。只需要根据不同的网页特征，设置相关的爬虫规则，便可以愉快的实现同时在多台机器上快速地获取互联网数据的初衷了。

####                         ####

##  ##

![快速回复][blog-icon-reply.png] ![TOP][]

[Python_1_N]: http://blog.csdn.net/HAPPENgft/article/details/59766342
[Python_2]: http://blog.csdn.net/HAPPENgft/article/details/60580588
[PyHusky]: http://blog.csdn.net/HAPPENgft/article/details/56668076
[blog-icon-reply.png]: http://static.blog.csdn.net/images/blog-icon-reply.png
[TOP]: /images/20210923/76cbd4839e3b4bd7ab997142e654766f.png