def run_it():
# 读取网页地址列表
sql = '''
SELECT `md5`,`url`,`flag`
FROM `result`.`sys_url_info`
where flag = 0
and url like '%https://zhidao.baidu.com/search%'
'''
res, list_a = applicationDb.read_sql(sql)
# 读取所有配置信息
sql = '''
SELECT
`uuid`,
`delete_flag`,
`discription`,
`seed_name`,
`seed_type`,
`seed_url`,
`request_proxy`
FROM `application`.`sys_seed_url_info`
WHERE delete_flag = 0
LIMIT 0, 100
'''
rec, config_list = resultDb.read_sql(sql)
# 判断每一个地址是否有配置信息
# 引入通用模块
name_model = "Currency"
run_model = __import__(name_model)
flag = True
for a in list_a: # 地址列表
print("验证链接地址： ", a)
for config in config_list: # 配置列表
# 如果有进入采集程序
if a[1].find(config[5]) > -1:
print("开始分析网页")
flag = False
# 通用模块调用
run_model.run_it(uuid=config[0], url=a[1], uri=config[5])
if flag:
print("缺少采集配置信息")
# 如果没有自动添加新的配置信息，并提示用户进行修改 状态修改-1
sql = "update sys_url_info set flag =-1 where md5='" + config[
0] + "'"
rule = Rule()
applicationDb.write_sql(sql)
sql = "INSERT INTO application.sys_seed_url_info(UUID,delete_flag, seed_url)VALUES ('%s','-1','%s') " % (
rule.get_md5_value(rule.get_url_root(a[1])), a[1])
applicationDb.write_sql(sql)
# 并提示用户进行修改 TODO
# 更新url
sql = '''
UPDATE `result`.`sys_url_info`
SET `flag` = -1
WHERE `url` = '%s'
''' % (a)
resultDb.write_sql(sql)
def testBaseProperty(self):
nowTime = datetime.now()
rule = Rule(
id=1,
webName="佛山市人民政府",
webUrl="http://www.foshan.gov.cn/zwgk/zwdt/jryw/",
webModel="normal",
ruleModel="regular",
rulePattern=
r'<li [\s\S]*?([0-9]{4}-[0-9]{2}-[0-9]{2})[\s\S]*?href="([\s\S]*?)"[\s\S]*?title="([\s\S]*?)" >',
titlePosition="2",
hrefPosition="1",
timePosition="0",
isEffect=1,
updateTime=nowTime)
self.assertEqual(rule.id, 1)
self.assertEqual(rule.webName, "佛山市人民政府")
self.assertEqual(rule.webUrl,
"http://www.foshan.gov.cn/zwgk/zwdt/jryw/")
self.assertEqual(rule.webModel, "normal")
self.assertEqual(rule.ruleModel, "regular")
self.assertEqual(
rule.rulePattern,
r'<li [\s\S]*?([0-9]{4}-[0-9]{2}-[0-9]{2})[\s\S]*?href="([\s\S]*?)"[\s\S]*?title="([\s\S]*?)" >'
)
self.assertEqual(rule.titlePosition, "2")
self.assertEqual(rule.hrefPosition, "1")
self.assertEqual(rule.timePosition, "0")
self.assertEqual(rule.isEffect, 1)
self.assertEqual(rule.updateTime, nowTime)
self.assertNotEqual(rule.subscribeUpdateTime, None)
self.assertEqual(rule.subscribeIsPush, None)
self.assertEqual(rule.subscribeLastPushTime, None)
rule = Rule()
self.assertEqual(rule.id, None)
self.assertEqual(rule.webName, None)
self.assertEqual(rule.webUrl, None)
self.assertEqual(rule.webModel, None)
self.assertEqual(rule.ruleModel, None)
self.assertEqual(rule.rulePattern, None)
self.assertEqual(rule.titlePosition, None)
self.assertEqual(rule.hrefPosition, None)
self.assertEqual(rule.timePosition, None)
self.assertEqual(rule.isEffect, None)
self.assertNotEqual(rule.updateTime, None)
return
def testParser(self):
parser = HtmlParser()
rule = Rule(
id=1,
webName="佛山市人民政府",
webUrl="http://www.foshan.gov.cn/zwgk/zwdt/jryw/",
webModel="normal",
ruleModel="regular",
rulePattern=
r'<li [\s\S]*?([0-9]{4}-[0-9]{2}-[0-9]{2})[\s\S]*?href="([\s\S]*?)"[\s\S]*?title="([\s\S]*?)" >',
titlePosition="2",
hrefPosition="1",
timePosition="0",
isEffect=1)
htmlDownload = HtmlDownload()
html = htmlDownload.download(rule.webUrl, rule.webModel)
msgs = parser.RegularPraser(html, rule)
self.assertNotEqual(len(msgs), 0)
msgs = parser.parse(html, rule)
self.assertNotEqual(len(msgs), 0)
msgs, matchs = parser.parseForAPI(html, rule)
self.assertNotEqual(len(msgs), 0)
self.assertNotEqual(len(matchs), 0)
msgs = parser.parse('', rule)
self.assertEqual(len(msgs), 0)
pass
class TestHtmlDownload(unittest.TestCase):
nowTime = datetime.now()
rule = Rule(
id=1,
webName="baidu",
webUrl="https://www.baidu.com/",
webModel="normal",
ruleModel="regular",
rulePattern=
r'<li [\s\S]*?([0-9]{4}-[0-9]{2}-[0-9]{2})[\s\S]*?href="([\s\S]*?)"[\s\S]*?title="([\s\S]*?)" >',
titlePosition="2",
hrefPosition="1",
timePosition="0",
isEffect=1,
updateTime=nowTime)
def testBaseProperty(self):
#检测https的下载
self.rule.webUrl = "https://www.baidu.com/"
htmlDownload = HtmlDownload()
r = htmlDownload.download(self.rule.webUrl, self.rule.webModel)
self.assertNotEqual(r, -1)
#检测http的下载
self.rule.webUrl = 'http://www.baidu.com'
r = htmlDownload.download(self.rule.webUrl, self.rule.webModel)
self.assertNotEqual(r, -1)
self.rule.webUrl = 'www.baidu.com'
r = htmlDownload.download(self.rule.webUrl, self.rule.webModel)
self.assertNotEqual(r, -1)
self.rule.webUrl = r'http://paperpost.people.com.cn/all-rmrb-\Y-\m-\d.html'
self.rule.webModel = 'date'
nowTime = datetime(2019, 1, 1, 1, 1)
r = htmlDownload.download(self.rule.webUrl,
self.rule.webModel,
date=nowTime)
self.assertNotEqual(r, -1)
#!/usr/bin/env python
# -*- coding: UTF-8 -*-
from common.HtmlSource import HtmlSource
from common.Rule import Rule
# from common.inc_conn import Conn_mysql
from common.inc_file import File_file, File_floder
from common.inc_csv import Csv_base
import time
htmlSource = HtmlSource()
rule = Rule()
path = 'D:/newpro/6.1'
# 多线程
def read_detial(url, i):
detial_html = htmlSource.get_html(url_p=url, type_p='rg')
#print(detial_html)
# 写html
files = File_file()
names = url.split('/')
file_name = names[len(names) - 1]
files.save_source(path=path,
file=file_name,
all_the_text=detial_html,
encoding_='utf-8')
colum = [
('title', '//h1[@class="articleHead"]//text()', 'l'),
('pushDate',
'//div[@class="info"]//span[@class="infoA"][@id="pubtime_baidu"]//text()',
def testBaseMethod(self):
nowTime = datetime.now()
rule = Rule(
id=1,
webName="佛山市科学技术局",
webUrl="http://www.fskw.gov.cn/tzgg/",
webModel="normal",
ruleModel="regular",
rulePattern=
r'<li><span>[\s\S]*?([0-9]{4}-[0-9]{2}-[0-9]{2})[\s\S]*?href="([\s\S]*?)"[\s\S]*?title="([\s\S]*?)">[\s\S]*?</li>',
titlePosition="2",
hrefPosition="1",
timePosition="0",
isEffect=1)
msg = Message("title", "href", nowTime, 1)
self.assertRaises(AssertionError, rule.addMessage, 1)
rule.addMessage(msg)
msg = rule.getMessages()
self.assertEqual(msg[0].title, "title")
rule.removeMessage(msg[0])
msg = rule.getMessages()
self.assertEqual(len(msg), 0)
msg = Message("title", "href", nowTime, 1)
rule.addMessage(msg)
rule.removeMessageById(1)
msg = rule.getMessages()
self.assertEqual(len(msg), 0)