[python]抓取沪深股市交易龙虎榜数据
python 3.5.0下运行
没做自动建立files文件夹,需要手动在py文件目录下建立files文件夹后运行
#coding=utf-8 import gzipimport http.cookiejar import urllib.request import urllib.parse import json import os import time import datetime def getOpener(head): # deal with the Cookies cj = http.cookiejar.CookieJar() pro = urllib.request.HTTPCookieProcessor(cj) opener = urllib.request.build_opener(pro) header = [] for key, value in head.items(): elem = (key, value) header.append(elem) opener.addheaders = header return opener def ungzip(data): try: # 尝试解压 print('正在解压.....') data = gzip.decompress(data) print('解压完毕!') except: print('未经压缩, 无需解压') return datadef writeFile(fname,data): filename = r'files/'+fname+'.txt' if os.path.exists(filename): message = '文件 + '+filename +' 已存在,跳过' else: message = '文件 + '+filename +' 不存在,新建' f=open(filename,'w') f.write(data) f.close() print ('文件:'+fname+' 处理完毕。') ''' 读取抓取数据开始日期 如果不存在该日期,从10日前开始读取 如果存在从文件内日期开始读取 读到今天 ''' header = { 'Connection': 'Keep-Alive', 'Accept': '*/*', 'Accept-Language': 'zh-CN,zh;q=0.8', 'User-Agent': 'Mozilla/5.0 (Windows NT 6.2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/40.0.2214.111 Safari/537.36', 'Accept-Encoding': 'gzip, deflate', 'Host': '', 'Referer' : '' } shUrl = 'http://query.sse.com.cn/infodisplay/showTradePublicFile.do?dateTx='#2015-09-28 szUrl = ['http://www.szse.cn/szseWeb/common/szse/files/text/jy/jy',#150923.txt 'http://www.szse.cn/szseWeb/common/szse/files/text/smeTxt/gk/sme_jy',#150708.txt 'http://www.szse.cn/szseWeb/common/szse/files/text/nmTxt/gk/nm_jy']#150902.txt startFileName = r'startDay.txt' endDay = datetime.datetime.now() if os.path.exists(startFileName): print('日期配置文件存在,开始读取') f=open(startFileName,'rt') s = f.readline() f.close() if s!='': print('将从日期:'+s+' 开始读取') timeArray = time.strptime(s, "%Y%m%d") timeStamp = int(time.mktime(timeArray)) fromDay = datetime.datetime.utcfromtimestamp(timeStamp) else: print('日期配置文件为空,将从10日前日期开始读取') fromDay = endDay - datetime.timedelta(days = 10) else: print('日期配置文件不存在,将从10日前日期开始读取') fromDay = endDay - datetime.timedelta(days = 10) endDay = endDay + datetime.timedelta(days = 1) while fromDay.strftime("%Y%m%d")!=endDay.strftime("%Y%m%d"): print(fromDay.strftime("%Y%m%d")) ''' 循环上面日期 抓取上证,深证,中小创交易龙虎榜数据 如果内容不为空 文件不存在 写入文件 ''' #抓取上证龙虎榜数据 url = shUrl + fromDay.strftime("%Y-%m-%d") print('读取上证龙虎榜 '+url) header['Host'] = 'query.sse.com.cn' header['Referer'] = 'http://www.sse.com.cn/disclosure/diclosure/public/' try: opener = getOpener(header) op = opener.open(url) data = op.read() data = data.decode() jsonData = json.loads(data) outData = '' if (jsonData['fileContents']!=''): for info in jsonData['fileContents']: outData= outData+ info+' ' writeFile(fromDay.strftime("%Y-%m-%d")+'_上证',outData) except: print(fromDay.strftime("%Y-%m-%d")+'跳过') #抓取深证,中小创交易龙虎榜数据 i=1 for url in szUrl: if(i==1): name = '深证' elif(i==2): name = '中小板' else: name = '创业板' url = url + fromDay.strftime("%y%m%d")+'.txt' print('读取'+name+'龙虎榜 '+url) header['Host'] = 'www.szse.cn' header['Referer'] = 'http://www.szse.cn' try: opener = getOpener(header) op = opener.open(url) data = op.read() data = ungzip(data) data = data.decode('gbk') writeFile(fromDay.strftime("%Y-%m-%d")+'_'+name,data) except: print(fromDay.strftime("%Y-%m-%d")+'跳过') i=i+1 fromDay = fromDay + datetime.timedelta(days = 1) #最后更新日期为当前日期 print('设置最新日期') fromDay = fromDay - datetime.timedelta(days = 1) f=open(startFileName,'w') f.write(fromDay.strftime("%Y%m%d")) f.close() print('读取完成')