def analyze_detail(self, html, company_id, url):
xpath_html = etree.HTML(html)
company = Company(company_id=company_id, company_source=self.source)
#转换为一段json字符串，几乎包含所有的信息
# company_text = xpath_html.xpath("//script[@id='companyInfoData']/text()")
# if company_text[0]:
# company_text = company_text[0]
try:
company_text = xpath_html.xpath(
"//script[@id='companyInfoData']/text()")[0]
except:
return
company_info = json.loads(company_text)
# 公司基本信息，包括人数，类型等
baseinfo = company_info['baseInfo']
# #地址列表，，里面包含很多地址信息
# address = company_info['addressList'][0]
# #里面包含公司基本信息，包括名字、简介等，
# coreInfo = company_info['coreInfo']
try:
address = company_info['addressList'][0]
company.address = address['detailAddress'] #详细地址
company.city = address['city'] #城市
company.company_name = company_info['coreInfo'][
'companyName'] #公司名称
except Exception as e:
log.error('{}缺少必要字段,error={}'.format(url, e))
return
#长简介
if company_info['introduction'].get('companyProfile'):
company.company_info = company_info['introduction'][
'companyProfile']
#短简介
if company_info['coreInfo'].get('companyIntroduce'):
# if company_info['coreInfo']['companyIntroduce']:
company.company_short_info = company_info['coreInfo'][
'companyIntroduce']
if baseinfo.get('industryField'):
company.business = company_info['baseInfo']['industryField']
if baseinfo.get('financeStage'):
company.development_stage = company_info['baseInfo'][
'financeStage']
if baseinfo.get('companySize'):
company.company_size = company_info['baseInfo']['companySize']
#所在区域
if address.get('district'):
company.region = address['district']
company.url = url
# result = company.serialization_info()
# # print(result)
company.insert_db()
def analyze_detail(self, html, company_id, url):
xpath_html = etree.HTML(html)
company = Company(company_id=company_id, company_source=self.source)
company.address = xpath_html.xpath('string(//*[@id="location_container"]/div[2]/div[2])').strip()
company.company_info = xpath_html.xpath('string(//*[@id="company_intro"])').strip()
company.company_short_info = xpath_html.xpath('/html/body/div[2]/div/div/div[1]/div/text()')[0].strip()
company.city = xpath_html.xpath('//*[@id="basic_container"]/div[2]/ul/li[4]/span/text()')[0].strip()
company.business = xpath_html.xpath('//*[@id="basic_container"]/div[2]/ul/li[1]/span/text()')[0].strip()
company.development_stage = xpath_html.xpath('//*[@id="basic_container"]/div[2]/ul/li[2]/span/text()')[
0].strip()
company.company_name = xpath_html.xpath('/html/body/div[2]/div/div/div[1]/h1/a/text()')[0].strip()
company.company_size = xpath_html.xpath('//*[@id="basic_container"]/div[2]/ul/li[3]/span/text()')[0].strip()
company.url = url
company.insert_db()