import requests
from lxml import etree
import time
import csv
#定义函数抓取每页前30条商品信息
def crow_first(n):
#构造每一页的url变化
url='https://search.jd.com/Search?keyword=%E6%89%8B%E6%9C%BA&enc=utf-8&qrst=1&rt=1&stop=1&vt=2&cid2=653&cid3=655&page='+str(2*n-1)
head = {'authority': 'search.jd.com',
'method': 'GET',
'path': '/s_new.php?keyword=%E6%89%8B%E6%9C%BA&enc=utf-8&qrst=1&rt=1&stop=1&vt=2&wq=%E6%89%8B%E6%9C%BA&cid2=653&cid3=655&page=4&s=84&scrolling=y&log_id=1529828108.22071&tpl=3_M&show_items=7651927,7367120,7056868,7419252,6001239,5934182,4554969,3893501,7421462,6577495,26480543553,7345757,4483120,6176077,6932795,7336429,5963066,5283387,25722468892,7425622,4768461',
'scheme': 'https',
'referer': 'https://search.jd.com/Search?keyword=%E6%89%8B%E6%9C%BA&enc=utf-8&qrst=1&rt=1&stop=1&vt=2&wq=%E6%89%8B%E6%9C%BA&cid2=653&cid3=655&page=3&s=58&click=0',
'user-agent': 'Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/66.0.3359.139 Safari/537.36',
'x-requested-with': 'XMLHttpRequest',
'Cookie':'qrsc=3; pinId=RAGa4xMoVrs; xtest=1210.cf6b6759; ipLocation=%u5E7F%u4E1C; _jrda=5; TrackID=1aUdbc9HHS2MdEzabuYEyED1iDJaLWwBAfGBfyIHJZCLWKfWaB_KHKIMX9Vj9_2wUakxuSLAO9AFtB2U0SsAD-mXIh5rIfuDiSHSNhZcsJvg; shshshfpa=17943c91-d534-104f-a035-6e1719740bb6-1525571955; shshshfpb=2f200f7c5265e4af999b95b20d90e6618559f7251020a80ea1aee61500; cn=0; 3AB9D23F7A4B3C9B=QFOFIDQSIC7TZDQ7U4RPNYNFQN7S26SFCQQGTC3YU5UZQJZUBNPEXMX7O3R7SIRBTTJ72AXC4S3IJ46ESBLTNHD37U; ipLoc-djd=19-1607-3638-3638.608841570; __jdu=930036140; user-key=31a7628c-a9b2-44b0-8147-f10a9e597d6f; areaId=19; __jdv=122270672|direct|-|none|-|1529893590075; PCSYCityID=25; mt_xid=V2_52007VwsQU1xaVVoaSClUA2YLEAdbWk5YSk9MQAA0BBZOVQ0ADwNLGlUAZwQXVQpaAlkvShhcDHsCFU5eXENaGkIZWg5nAyJQbVhiWR9BGlUNZwoWYl1dVF0%3D; __jdc=122270672; shshshfp=72ec41b59960ea9a26956307465948f6; rkv=V0700; __jda=122270672.930036140.-.1529979524.1529984840.85; __jdb=122270672.1.930036140|85.1529984840; shshshsID=f797fbad20f4e576e9c30d1c381ecbb1_1_1529984840145'
}
r = requests.get(url, headers=head)
#指定编码方式,不然会出现乱码
r.encoding='utf-8'
html1 = etree.HTML(r.text)
#定位到每一个商品标签li
datas=html1.xpath('//li[contains(@class,"gl-item")]')
#将抓取的结果保存到本地CSV文件中
with open('JD_Phone.csv','w',newline='',encoding='utf-8')as f:
write=csv.writer(f)
for data in datas:
p_price = data.xpath('//div[@id="J_goodsList"]//strong//i')
p_comment = data.xpath('div/div[5]/strong/a/text()')
p_name = data.xpath('div/div[@class="p-name p-name-type-2"]/a/em')
#这个if判断用来处理那些价格可以动态切换的商品,比如上文提到的小米MIX2,他们的价格位置在属性中放了一个最低价
if len(p_price) == 0:
p_price = data.xpath('div/div[@class="p-price"]/strong/@data-price')
#xpath('string(.)')用来解析混夹在几个标签中的文本
write.writerow([p_name[0].xpath('string(.)'),p_price[0],p_comment[0]])
f.close()
#定义函数抓取每页后30条商品信息
def crow_last(n):
#获取当前的Unix时间戳,并且保留小数点后5位
a=time.time()
b='%.5f'%a
url='https://search.jd.com/s_new.php?keyword=%E6%89%8B%E6%9C%BA&enc=utf-8&qrst=1&rt=1&stop=1&vt=2&wq=%E6%89%8B%E6%9C%BA&cid2=653&cid3=655&page='+str(2*n)+'&s='+str(48*n-20)+'&scrolling=y&log_id='+str(b)
head={'authority': 'search.jd.com',
'method': 'GET',
'path': '/s_new.php?keyword=%E6%89%8B%E6%9C%BA&enc=utf-8&qrst=1&rt=1&stop=1&vt=2&wq=%E6%89%8B%E6%9C%BA',
'scheme':'https',
'referer': 'https://search.jd.com/Search?keyword=%E6%89%8B%E6%9C%BA&enc=utf-8&qrst=1&rt=1&stop=1&vt=2&wq=%E6%89%8B%E6%9C%BA&cid2=653&cid3=655&page=3&s=58&click=0',
'user-agent': 'Mozilla/5.0 (Windows NT 6.1; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/66.0.3359.139 Safari/537.36',
'x-requested-with': 'XMLHttpRequest',
'Cookie':'qrsc=3; pinId=RAGa4xMoVrs; xtest=1210.cf6b6759; ipLocation=%u5E7F%u4E1C; _jrda=5; TrackID=1aUdbc9HHS2MdEzabuYEyED1iDJaLWwBAfGBfyIHJZCLWKfWaB_KHKIMX9Vj9_2wUakxuSLAO9AFtB2U0SsAD-mXIh5rIfuDiSHSNhZcsJvg; shshshfpa=17943c91-d534-104f-a035-6e1719740bb6-1525571955; shshshfpb=2f200f7c5265e4af999b95b20d90e6618559f7251020a80ea1aee61500; cn=0; 3AB9D23F7A4B3C9B=QFOFIDQSIC7TZDQ7U4RPNYNFQN7S26SFCQQGTC3YU5UZQJZUBNPEXMX7O3R7SIRBTTJ72AXC4S3IJ46ESBLTNHD37U; ipLoc-djd=19-1607-3638-3638.608841570; __jdu=930036140; user-key=31a7628c-a9b2-44b0-8147-f10a9e597d6f; areaId=19; __jdv=122270672|direct|-|none|-|1529893590075; PCSYCityID=25; mt_xid=V2_52007VwsQU1xaVVoaSClUA2YLEAdbWk5YSk9MQAA0BBZOVQ0ADwNLGlUAZwQXVQpaAlkvShhcDHsCFU5eXENaGkIZWg5nAyJQbVhiWR9BGlUNZwoWYl1dVF0%3D; __jdc=122270672; shshshfp=72ec41b59960ea9a26956307465948f6; rkv=V0700; __jda=122270672.930036140.-.1529979524.1529984840.85; __jdb=122270672.1.930036140|85.1529984840; shshshsID=f797fbad20f4e576e9c30d1c381ecbb1_1_1529984840145'
}
r = requests.get(url, headers=head)
r.encoding = 'utf-8'
html1 = etree.HTML(r.text)
datas = html1.xpath('//li[contains(@class,"gl-item")]')
with open('JD_Phone.csv','a',newline='',encoding='utf-8')as f:
write=csv.writer(f)
for data in datas:
p_price = data.xpath('div/div[@class="p-price"]/strong/i/text()')
p_comment = data.xpath('div/div[5]/strong/a/text()')
p_name = data.xpath('div/div[@class="p-name p-name-type-2"]/a/em')
if len(p_price) == 0:
p_price = data.xpath('div/div[@class="p-price"]/strong/@data-price')
write.writerow([p_name[0].xpath('string(.)'),p_price[0],p_comment[0]])
f.close()
if __name__=='__main__':
for i in range(1,101):
try:
print(' First_Page: ' + str(i))
crow_first(i)
print(' Finish')
except Exception as e:
print(e)
try:
print(' Last_Page: ' + str(i))
crow_last(i)
print(' Finish')
京东商品评论爬虫
最新推荐文章于 2024-09-26 09:51:42 发布