写在前面
这次的爬虫是关于房价信息的抓取,目的在于练习10万以上的数据处理及整站式抓取。
数据量的提升最直观的感觉便是对函数逻辑要求的提高,针对Python的特性,谨慎的选择数据结构。以往小数据量的抓取,即使函数逻辑部分重复,I/O请求频率密集,循环套嵌过深,也不过是1~2s的差别,而随着数据规模的提高,这1~2s的差别就有可能扩展成为1~2h。
因此对于要抓取数据量较多的网站,可以从两方面着手降低抓取信息的时间成本。
1)优化函数逻辑,选择适当的数据结构,符合Pythonic的编程习惯。例如,字符串的合并,使用join()要比“+”节省内存空间。
2)依据I/O密集与CPU密集,选择多线程、多进程并行的执行方式,提高执行效率。
一、获取索引
包装请求request,设置超时timeout
1 # 获取列表页面 2 def get_page(url): 3 headers = { 4 'User-Agent': r'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) ' 5 r'Chrome/45.0.2454.85 Safari/537.36 115Browser/6.0.3', 6 'Referer': r'http://bj.fangjia.com/ershoufang/', 7 'Host': r'bj.fangjia.com', 8 'Connection': 'keep-alive' 9 } 10 timeout = 60 11 socket.setdefaulttimeout(timeout) # 设置超时 12 req = request.Request(url, headers=headers) 13 response = request.urlopen(req).read() 14 page = response.decode('utf-8') 15 return page
一级位置:区域信息
二级位置:板块信息(根据区域位置得到板块信息,以key_value对的形式存储在dict中)
以dict方式存储,可以快速的查询到所要查找的目标。-> {'朝阳':{'工体','安贞','健翔桥'......}}
三级位置:地铁信息(搜索地铁周边房源信息)
将所属位置地铁信息,添加至dict中。 -> {'朝阳':{'工体':{'5号线','10号线' , '13号线'},'安贞','健翔桥'......}}
对应的url:http://bj.fangjia.com/ershoufang/--r-%E6%9C%9D%E9%98%B3%7Cw-5%E5%8F%B7%E7%BA%BF%7Cb-%E6%83%A0%E6%96%B0%E8%A5%BF%E8%A1%97
解码后的url:http://bj.fangjia.com/ershoufang/--r-朝阳|w-5号线|b-惠新西街
根据url的参数模式,可以有两种方式获取目的url:
1)根据索引路径获得目的url
1 # 获取房源信息列表(嵌套字典遍历) 2 def get_info_list(search_dict, layer, tmp_list, search_list): 3 layer += 1 # 设置字典层级 4 for i in range(len(search_dict)): 5 tmp_key = list(search_dict.keys())[i] # 提取当前字典层级key 6 tmp_list.append(tmp_key) # 将当前key值作为索引添加至tmp_list 7 tmp_value = search_dict[tmp_key] 8 if isinstance(tmp_value, str): # 当键值为url时 9 tmp_list.append(tmp_value) # 将url添加至tmp_list 10 search_list.append(copy.deepcopy(tmp_list)) # 将tmp_list索引url添加至search_list 11 tmp_list = tmp_list[:layer] # 根据层级保留索引 12 elif tmp_value == '': # 键值为空时跳过 13 layer -= 2 # 跳出键值层级 14 tmp_list = tmp_list[:layer] # 根据层级保留索引 15 else: 16 get_info_list(tmp_value, layer, tmp_list, search_list) # 当键值为列表时,迭代遍历 17 tmp_list = tmp_list[:layer] 18 return search_list
2)根据dict信息包装url
{'朝阳':{'工体':{'5号线'}}}
参数:
—— r-朝阳
—— b-工体
—— w-5号线
组装参数:http://bj.fangjia.com/ershoufang/--r-朝阳|w-5号线|b-工体
1 # 根据参数创建组合url 2 def get_compose_url(compose_tmp_url, tag_args, key_args): 3 compose_tmp_url_list = [compose_tmp_url, '|' if tag_args != 'r-' else '', tag_args, parse.quote(key_args), ] 4 compose_url = ''.join(compose_tmp_url_list) 5 return compose_url
二、获取索引页最大页数
1 # 获取当前索引页面页数的url列表 2 def get_info_pn_list(search_list): 3 fin_search_list = [] 4 for i in range(len(search_list)): 5 print('>>>正在抓取%s' % search_list[i][:3]) 6 search_url = search_list[i][3] 7 try: 8 page = get_page(search_url) 9 except: 10 print('获取页面超时') 11 continue 12 soup = BS(page, 'lxml') 13 # 获取最大页数 14 pn_num = soup.select('span[class="mr5"]')[0].get_text() 15 rule = re.compile(r'd+') 16 max_pn = int(rule.findall(pn_num)[1]) 17 # 组装url 18 for pn in range(1, max_pn+1): 19 print('************************正在抓取%s页************************' % pn) 20 pn_rule = re.compile('[|]') 21 fin_url = pn_rule.sub(r'|e-%s|' % pn, search_url, 1) 22 tmp_url_list = copy.deepcopy(search_list[i][:3]) 23 tmp_url_list.append(fin_url) 24 fin_search_list.append(tmp_url_list) 25 return fin_search_list
三、抓取房源信息Tag
这是我们要抓取的Tag:
['区域', '板块', '地铁', '标题', '位置', '平米', '户型', '楼层', '总价', '单位平米价格']
1 # 获取tag信息 2 def get_info(fin_search_list, process_i): 3 print('进程%s开始' % process_i) 4 fin_info_list = [] 5 for i in range(len(fin_search_list)): 6 url = fin_search_list[i][3] 7 try: 8 page = get_page(url) 9 except: 10 print('获取tag超时') 11 continue 12 soup = BS(page, 'lxml') 13 title_list = soup.select('a[class="h_name"]') 14 address_list = soup.select('span[class="address]') 15 attr_list = soup.select('span[class="attribute"]') 16 price_list = soup.find_all(attrs={"class": "xq_aprice xq_esf_width"}) # select对于某些属性值(属性值中间包含空格)无法识别,可以用find_all(attrs={})代替 17 for num in range(20): 18 tag_tmp_list = [] 19 try: 20 title = title_list[num].attrs["title"] 21 print(r'************************正在获取%s************************' % title) 22 address = re.sub(' ', '', address_list[num].get_text()) 23 area = re.search('d+[u4E00-u9FA5]{2}', attr_list[num].get_text()).group(0) 24 layout = re.search('d[^0-9]d.', attr_list[num].get_text()).group(0) 25 floor = re.search('d/d', attr_list[num].get_text()).group(0) 26 price = re.search('d+[u4E00-u9FA5]', price_list[num].get_text()).group(0) 27 unit_price = re.search('d+[u4E00-u9FA5]/.', price_list[num].get_text()).group(0) 28 tag_tmp_list = copy.deepcopy(fin_search_list[i][:3]) 29 for tag in [title, address, area, layout, floor, price, unit_price]: 30 tag_tmp_list.append(tag) 31 fin_info_list.append(tag_tmp_list) 32 except: 33 print('【抓取失败】') 34 continue 35 print('进程%s结束' % process_i) 36 return fin_info_list
四、分配任务,并行抓取
对任务列表进行分片,设置进程池,并行抓取。
1 # 分配任务 2 def assignment_search_list(fin_search_list, project_num): # project_num每个进程包含的任务数,数值越小,进程数越多 3 assignment_list = [] 4 fin_search_list_len = len(fin_search_list) 5 for i in range(0, fin_search_list_len, project_num): 6 start = i 7 end = i+project_num 8 assignment_list.append(fin_search_list[start: end]) # 获取列表碎片 9 return assignment_list
1 p = Pool(4) # 设置进程池 2 assignment_list = assignment_search_list(fin_info_pn_list, 3) # 分配任务,用于多进程 3 result = [] # 多进程结果列表 4 for i in range(len(assignment_list)): 5 result.append(p.apply_async(get_info, args=(assignment_list[i], i))) 6 p.close() 7 p.join() 8 for result_i in range(len(result)): 9 fin_info_result_list = result[result_i].get() 10 fin_save_list.extend(fin_info_result_list) # 将各个进程获得的列表合并
通过设置进程池并行抓取,时间缩短为单进程抓取时间的3/1,总计时间3h。
电脑为4核,经过测试,任务数为3时,在当前电脑运行效率最高。
五、将抓取结果存储到excel中,等待可视化数据化处理
1 # 存储抓取结果 2 def save_excel(fin_info_list, file_name): 3 tag_name = ['区域', '板块', '地铁', '标题', '位置', '平米', '户型', '楼层', '总价', '单位平米价格'] 4 book = xlsxwriter.Workbook(r'C:UsersAdministratorDesktop\%s.xls' % file_name) # 默认存储在桌面上 5 tmp = book.add_worksheet() 6 row_num = len(fin_info_list) 7 for i in range(1, row_num): 8 if i == 1: 9 tag_pos = 'A%s' % i 10 tmp.write_row(tag_pos, tag_name) 11 else: 12 con_pos = 'A%s' % i 13 content = fin_info_list[i-1] # -1是因为被表格的表头所占 14 tmp.write_row(con_pos, content) 15 book.close()
附上源码
1 #! -*-coding:utf-8-*- 2 # Function: 房价调查 3 # Author:蘭兹 4 5 from urllib import parse, request 6 from bs4 import BeautifulSoup as BS 7 from multiprocessing import Pool 8 import re 9 import lxml 10 import datetime 11 import cProfile 12 import socket 13 import copy 14 import xlsxwriter 15 16 17 starttime = datetime.datetime.now() 18 19 base_url = r'http://bj.fangjia.com/ershoufang/' 20 21 22 test_search_dict = {'昌平': {'霍营': {'13号线': 'http://bj.fangjia.com/ershoufang/--r-%E6%98%8C%E5%B9%B3|w-13%E5%8F%B7%E7%BA%BF|b-%E9%9C%8D%E8%90%A5'}}} 23 24 search_list = [] # 房源信息url列表 25 tmp_list = [] # 房源信息url缓存列表 26 layer = -1 27 28 29 # 获取列表页面 30 def get_page(url): 31 headers = { 32 'User-Agent': r'Mozilla/5.0 (Windows NT 6.1; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) ' 33 r'Chrome/45.0.2454.85 Safari/537.36 115Browser/6.0.3', 34 'Referer': r'http://bj.fangjia.com/ershoufang/', 35 'Host': r'bj.fangjia.com', 36 'Connection': 'keep-alive' 37 } 38 timeout = 60 39 socket.setdefaulttimeout(timeout) # 设置超时 40 req = request.Request(url, headers=headers) 41 response = request.urlopen(req).read() 42 page = response.decode('utf-8') 43 return page 44 45 46 # 获取查询关键词dict 47 def get_search(page, key): 48 soup = BS(page, 'lxml') 49 search_list = soup.find_all(href=re.compile(key), target='') 50 search_dict = {} 51 for i in range(len(search_list)): 52 soup = BS(str(search_list[i]), 'lxml') 53 key = soup.select('a')[0].get_text() 54 value = soup.a.attrs['href'] 55 search_dict[key] = value 56 return search_dict 57 58 59 # 获取房源信息列表(嵌套字典遍历) 60 def get_info_list(search_dict, layer, tmp_list, search_list): 61 layer += 1 # 设置字典层级 62 for i in range(len(search_dict)): 63 tmp_key = list(search_dict.keys())[i] # 提取当前字典层级key 64 tmp_list.append(tmp_key) # 将当前key值作为索引添加至tmp_list 65 tmp_value = search_dict[tmp_key] 66 if isinstance(tmp_value, str): # 当键值为url时 67 tmp_list.append(tmp_value) # 将url添加至tmp_list 68 search_list.append(copy.deepcopy(tmp_list)) # 将tmp_list索引url添加至search_list 69 tmp_list = tmp_list[:layer] # 根据层级保留索引 70 elif tmp_value == '': # 键值为空时跳过 71 layer -= 2 # 跳出键值层级 72 tmp_list = tmp_list[:layer] # 根据层级保留索引 73 else: 74 get_info_list(tmp_value, layer, tmp_list, search_list) # 当键值为列表时,迭代遍历 75 tmp_list = tmp_list[:layer] 76 return search_list 77 78 79 # 获取房源信息详情 80 def get_info_pn_list(search_list): 81 fin_search_list = [] 82 for i in range(len(search_list)): 83 print('>>>正在抓取%s' % search_list[i][:3]) 84 search_url = search_list[i][3] 85 try: 86 page = get_page(search_url) 87 except: 88 print('获取页面超时') 89 continue 90 soup = BS(page, 'lxml') 91 # 获取最大页数 92 pn_num = soup.select('span[class="mr5"]')[0].get_text() 93 rule = re.compile(r'd+') 94 max_pn = int(rule.findall(pn_num)[1]) 95 # 组装url 96 for pn in range(1, max_pn+1): 97 print('************************正在抓取%s页************************' % pn) 98 pn_rule = re.compile('[|]') 99 fin_url = pn_rule.sub(r'|e-%s|' % pn, search_url, 1) 100 tmp_url_list = copy.deepcopy(search_list[i][:3]) 101 tmp_url_list.append(fin_url) 102 fin_search_list.append(tmp_url_list) 103 return fin_search_list 104 105 106 # 获取tag信息 107 def get_info(fin_search_list, process_i): 108 print('进程%s开始' % process_i) 109 fin_info_list = [] 110 for i in range(len(fin_search_list)): 111 url = fin_search_list[i][3] 112 try: 113 page = get_page(url) 114 except: 115 print('获取tag超时') 116 continue 117 soup = BS(page, 'lxml') 118 title_list = soup.select('a[class="h_name"]') 119 address_list = soup.select('span[class="address]') 120 attr_list = soup.select('span[class="attribute"]') 121 price_list = soup.find_all(attrs={"class": "xq_aprice xq_esf_width"}) # select对于某些属性值(属性值中间包含空格)无法识别,可以用find_all(attrs={})代替 122 for num in range(20): 123 tag_tmp_list = [] 124 try: 125 title = title_list[num].attrs["title"] 126 print(r'************************正在获取%s************************' % title) 127 address = re.sub(' ', '', address_list[num].get_text()) 128 area = re.search('d+[u4E00-u9FA5]{2}', attr_list[num].get_text()).group(0) 129 layout = re.search('d[^0-9]d.', attr_list[num].get_text()).group(0) 130 floor = re.search('d/d', attr_list[num].get_text()).group(0) 131 price = re.search('d+[u4E00-u9FA5]', price_list[num].get_text()).group(0) 132 unit_price = re.search('d+[u4E00-u9FA5]/.', price_list[num].get_text()).group(0) 133 tag_tmp_list = copy.deepcopy(fin_search_list[i][:3]) 134 for tag in [title, address, area, layout, floor, price, unit_price]: 135 tag_tmp_list.append(tag) 136 fin_info_list.append(tag_tmp_list) 137 except: 138 print('【抓取失败】') 139 continue 140 print('进程%s结束' % process_i) 141 return fin_info_list 142 143 144 # 分配任务 145 def assignment_search_list(fin_search_list, project_num): # project_num每个进程包含的任务数,数值越小,进程数越多 146 assignment_list = [] 147 fin_search_list_len = len(fin_search_list) 148 for i in range(0, fin_search_list_len, project_num): 149 start = i 150 end = i+project_num 151 assignment_list.append(fin_search_list[start: end]) # 获取列表碎片 152 return assignment_list 153 154 155 # 存储抓取结果 156 def save_excel(fin_info_list, file_name): 157 tag_name = ['区域', '板块', '地铁', '标题', '位置', '平米', '户型', '楼层', '总价', '单位平米价格'] 158 book = xlsxwriter.Workbook(r'C:UsersAdministratorDesktop\%s.xls' % file_name) # 默认存储在桌面上 159 tmp = book.add_worksheet() 160 row_num = len(fin_info_list) 161 for i in range(1, row_num): 162 if i == 1: 163 tag_pos = 'A%s' % i 164 tmp.write_row(tag_pos, tag_name) 165 else: 166 con_pos = 'A%s' % i 167 content = fin_info_list[i-1] # -1是因为被表格的表头所占 168 tmp.write_row(con_pos, content) 169 book.close() 170 171 172 if __name__ == '__main__': 173 file_name = input(r'抓取完成,输入文件名保存:') 174 fin_save_list = [] # 抓取信息存储列表 175 # 一级筛选 176 page = get_page(base_url) 177 search_dict = get_search(page, 'r-') 178 # 二级筛选 179 for k in search_dict: 180 print(r'************************一级抓取:正在抓取【%s】************************' % k) 181 url = search_dict[k] 182 second_page = get_page(url) 183 second_search_dict = get_search(second_page, 'b-') 184 search_dict[k] = second_search_dict 185 # 三级筛选 186 for k in search_dict: 187 second_dict = search_dict[k] 188 for s_k in second_dict: 189 print(r'************************二级抓取:正在抓取【%s】************************' % s_k) 190 url = second_dict[s_k] 191 third_page = get_page(url) 192 third_search_dict = get_search(third_page, 'w-') 193 print('%s>%s' % (k, s_k)) 194 second_dict[s_k] = third_search_dict 195 fin_info_list = get_info_list(search_dict, layer, tmp_list, search_list) 196 fin_info_pn_list = get_info_pn_list(fin_info_list) 197 p = Pool(4) # 设置进程池 198 assignment_list = assignment_search_list(fin_info_pn_list, 2) # 分配任务,用于多进程 199 result = [] # 多进程结果列表 200 for i in range(len(assignment_list)): 201 result.append(p.apply_async(get_info, args=(assignment_list[i], i))) 202 p.close() 203 p.join() 204 for result_i in range(len(result)): 205 fin_info_result_list = result[result_i].get() 206 fin_save_list.extend(fin_info_result_list) # 将各个进程获得的列表合并 207 save_excel(fin_save_list, file_name) 208 endtime = datetime.datetime.now() 209 time = (endtime - starttime).seconds 210 print('总共用时:%s s' % time)
总结:
当抓取数据规模越大,对程序逻辑要求就愈严谨,对python语法要求就越熟练。如何写出更加pythonic的语法,也需要不断学习掌握的。
推荐阅读《编写高质量代码 改善Python程序的91个建议》
大家可以尝试抓取一下,分析一下房价的走势也是蛮有意思的ლ(^o^ლ)
欢迎交流,转载请注明出处~ (^ _ ^)/~~
更多python爬虫实例,请访问:http://www.landsblog.com/blog/category/pachong