Python实战计划学习第二周

在MongoDB中筛选房源

import pymongo
from bs4 import BeautifulSoup
import requests
import time

def get_seed(url='http://bj.xiaozhu.com/', page=1):
    return url if page <= 1 else '{}search-duanzufang-p{}-0/'.format(url, page)

def parse_fangzi(url, data=None):
    wb_data = requests.get(url)
    soup = BeautifulSoup(wb_data.text, 'lxml')
    title = soup.select('h4 > em') # 标题
    address = soup.select('p > span.pr5') # 地址
    price = soup.select('#pricePart > div.day_l > span') # 价格
    head = soup.select('#floatRightBox > div > div > a > img') # 房主头像
    name = soup.select('#floatRightBox > div > div > h6 > a') # 房主昵称
    gender = soup.select('#floatRightBox > div > div.member_pic > div') # 房主性别
    image = soup.select('#curBigImage') # 首张房子图片
    data = {
        'title': title[0].get_text() if title != None else None
        , 'address': address[0].get_text().strip() if address != None else None
        , 'price': int(price[0].get_text()) if price != None else None
        , 'head': head[0].get('src') if head != None else None
        , 'name': name[0].get_text() if name != None else None
        , 'gender': gender[0].get('class') if gender != None else None
        , 'image': image[0].get('src') if image != None else None
    }
    if data['gender'] != None:
        data['gender'] = '男' if 'member_ico' in data['gender'] else '女'
    return data

def save(sheet):
    seed = 'http://bj.xiaozhu.com/'
    urls = []
    for page in range(1,4) :
        wb_data = requests.get(get_seed(url=seed,page=page))
        soup = BeautifulSoup(wb_data.text, 'lxml')
        for div in soup.select('#page_list > ul > li > div.result_btm_con.lodgeunitname'):
            urls.append(div.get('detailurl'))

    for url in urls:
        sheet.insert_one(parse_fangzi(url))
        time.sleep(2)

client = pymongo.MongoClient('localhost',27017)
walden = client['walden']
xiaozhu = walden['xiaozhu']

# save(xiaozhu)
for item in xiaozhu.find({'price':{'$gte':500}}):
    print(item)

屏幕快照 2016-05-30 上午12.37.33.png

• 学习如何操作mongodb数据库
• 学习Python '三元操作符'

爬取手机号

import pymongo
from bs4 import BeautifulSoup
import requests
import time

count = 0
client = pymongo.MongoClient('localhost', 27017)
walden = client['walden']
shoujihao = walden['shoujihao']

for page in range(1, 2):
    url = 'http://bj.58.com/shoujihao/pn{}/'.format(page)
    wb_data = requests.get(url)
    soup = BeautifulSoup(wb_data.text, 'lxml')
    infocont = soup.select('#infocont > span')
    infocont = int(infocont[0].get_text()) if len(infocont) != 0 else 0
    # 如果找不到计数器 则直接退出
    if infocont == 0:
        print('共插入 {} 条, 终止于 {}'.format(count, url))
        break
    phones = soup.select('ul > div.boxlist > ul > li > a.t')
    for phone in phones:
        data = {
            'href': phone.get('href')
            , 'title': phone.find('strong', 'number').get_text()
        }
        shoujihao.insert_one(data)
    count += len(phones)
    print('{} -> {} : {}/{}'.format(url,len(phones),count,infocont))
    time.sleep(2)

屏幕快照 2016-05-30 上午12.43.40.png

设计断点续传程序

channel_extract.py

from bs4 import BeautifulSoup
import requests
import re


# 抽取58北京跳蚤市场下的所有类目 除却手机号
def get_channel_urls(url):
    wb_data = requests.get(url)
    soup = BeautifulSoup(wb_data.text, 'lxml')
    base = re.search(r'^(https?://[^/]+).*$', url).group(1)
    urls = []
    links = soup.select('ul.ym-submnu > li > b > a')
    for link in links:
        href = link.get('href')
        if href.startswith('/'):
            href = '{}{}'.format(base, href)
        if href not in urls and 'shoujihao' not in href:
            urls.append(href)
    return urls

counts.py

from bs4 import BeautifulSoup
import requests
import time
import pymongo

client = pymongo.MongoClient('localhost', 27017)
walden = client['walden']
_58_channels = walden['_58_channels']
_58_urls = walden['_58_urls']
_58_infos = walden['_58_infos']
while True:
    print('{}/{}/{}'.format(_58_channels.count(), _58_urls.count(), _58_infos.count()))
    time.sleep(5)

main.py

from multiprocessing import Pool
from channel_extract import get_channel_urls
from page_parsing import get_item_info
from page_parsing import get_links_from
import time

import pymongo

client = pymongo.MongoClient('localhost', 27017)
walden = client['walden']
_58_channels = walden['_58_channels']
_58_urls = walden['_58_urls']


def get_all_links_from(channel):
    if _58_channels.count({'channel': channel}) > 0:
        return
    count = 0
    for page in range(1, 101):
        links = get_links_from(channel, page, 0)
        time.sleep(0.1)
        if links <= 0:
            break
        count += links
    print('{} -> {}'.format(channel, count))
    _58_channels.insert_one({'channel': channel, 'count': count})


def get_info(url):
    get_item_info(url)
    time.sleep(0.1)


if __name__ == '__main__':
    channels = get_channel_urls('http://bj.58.com/sale.shtml')
    print(len(channels))
    # 首先执行解析类目程序
    pool = Pool()
    pool.map(get_all_links_from, channels)
    pool.close()
    pool.join()
    _58_urls.find_and_modify(query={'flag': {'$eq':0}}, update={'$set': {'flag': 0}})
    urls = list(map(lambda url: url['url'], _58_urls.find({'flag': {'$eq': 0}})))
    pool = Pool()
    pool.map(get_info, urls)
    pool.close()
    pool.join()

page_parsing.py

from bs4 import BeautifulSoup
import requests
import time
import pymongo
import sys

client = pymongo.MongoClient('localhost', 27017)
walden = client['walden']
_58_urls = walden['_58_urls']
_58_infos = walden['_58_infos']


def get_links_from(channel, page, who_sells=0):
    # http://bj.58.com/iphonesj/0/pn2/
    list_url = '{}{}/pn{}'.format(channel, who_sells, page)
    wb_data = requests.get(list_url)
    soup = BeautifulSoup(wb_data.text, 'lxml')
    # #infocont > span > b
    infocont = soup.select('#infocont > span > b')
    infocont = int(infocont[0].get_text()) if len(infocont) > 0 else 0
    print('{} -> {}'.format(list_url, infocont))
    if infocont <= 0:
        return 0
    links = soup.select('table > tr > td > a.t')
    for link in links:
        if not link.has_attr('onclick') and not link.has_attr('data-addtype'):
            item_link = link.get('href').split('?')[0]
            data = {
                'url': item_link
                , 'flag': 0
            }
            if _58_urls.find({'url': {'$eq': data['url']}}).count() == 0:
                _58_urls.insert_one(data)
    return len(links)


def get_item_info(url):
    wb_data = requests.get(url)
    soup = BeautifulSoup(wb_data.text, 'lxml')
    res = list(filter(lambda a: a.src != None and '404' in a.src, soup.find_all('script')))
    if len(res) > 0:
        return
    title = soup.select('div.col_sub.mainTitle > h1')
    price = soup.select('span.price.c_f50')
    date = soup.select('li.time')
    area = soup.select('li > div.su_con > span.c_25d')
    # 去除匹配到的 借钱买
    area = list(filter(lambda a: '借钱买' not in a.get_text(), area))
    try:
        data = {
            'url': url
            , 'title': title[0].get_text().strip() if len(title) > 0 else None
            , 'price': price[0].get_text().strip().strip('元').strip() if len(price) > 0 else None
            , 'date': date[0].get_text().strip() if len(date) > 0 else None
            , 'area': ''.join(area[0].stripped_strings) if len(area) > 0 else None
        }
    except:
        print('{} -> 异常'.format(url))
        print(sys.exc_info())
    else:
        print('{} -> {}'.format(url, data))
        record = _58_infos.find_one({'url': {'$eq': url}})
        if record == None:
            _58_infos.insert_one(data)
        else:
            _58_infos.update({'_id': record['_id']}, data)
        _58_urls.find_and_modify(query={'url': url}, update={'$inc': {'flag': 1}})

# get_links_from('http://bj.58.com/iphonesj/', 1)
# get_item_info('http://bj.58.com/zixingche/26131404258880x.shtml')

屏幕快照 2016-05-30 上午1.09.55.png

• 断点续传主要思路：在mongodb中渠道解析时存储解析到的url并给定判断位flag＝0,在解析完详情页的时候将对应url的flag自增，每次运行取待爬取url的时候仅取flag＝0的
• 另外发现启动多进程的时候 mongodb会报警(根据内容应该是同时连接了多个mongoclient):

UserWarning: MongoClient opened before fork. Create MongoClient with connect=False, or create client after forking. See PyMongo's documentation for details: http://api.mongodb.org/python/current/faq.html#using-pymongo-with-multiprocessing>
  "MongoClient opened before fork. Create MongoClient "

爬取10万商品数据

channel_extract.py ：渠道抓取

from bs4 import BeautifulSoup
import requests
import re


# 收取赶集网北京下面的所有二级类目
def get_channel_urls(url):
    wb_data = requests.get(url)
    soup = BeautifulSoup(wb_data.text, 'lxml')
    base = re.search(r'^(https?://[^/]+).*$', url).group(1)
    urls = []
    links = soup.select('dl.fenlei dt a')
    for link in links:
        href = link.get('href')
        if href.startswith('/'):
            href = '{}{}'.format(base, href)
            # 去除转转
        if href not in urls :
            urls.append(href)
    return urls

# #wrapper > div.content > div:nth-child(1) > div:nth-child(1) > dl > dt > a:nth-child(1)
# channels = get_channel_urls('http://bj.ganji.com/wu/')
# print('{}\n{}'.format(len(channels),'\n'.join(channels)))

counts.py ：监控小程序

import time
import pymongo

client = pymongo.MongoClient('localhost', 27017)
walden = client['walden']
_ganji_channels = walden['_ganji_channels']
_ganji_urls = walden['_ganji_urls']
_ganji_infos = walden['_ganji_infos']

while True:
    print('{}/{}/{}'.format(_ganji_channels.count(), _ganji_urls.count(), _ganji_infos.count()))
    time.sleep(5)

main.py ： 主程序

from multiprocessing import Pool
from channel_extract import get_channel_urls
from page_parsing import get_item_info
from page_parsing import get_links_from

import pymongo

client = pymongo.MongoClient('localhost', 27017)
walden = client['walden']
_ganji_channels = walden['_ganji_channels']
_ganji_urls = walden['_ganji_urls']


def get_all_links_from(channel):
    if _ganji_channels.count({'channel': channel}) > 0:
        return
    count = 0
    for page in range(1, 201):
        links = get_links_from(channel, page)
        if links <= 0:
            continue
        count += links
    print('{} -> {}'.format(channel, count))
    _ganji_channels.insert_one({'channel': channel, 'count': count})


def get_info(url):
    get_item_info(url)


# _ganji_urls.drop()
# _ganji_urls.find_and_modify(query={'flag': {'$eq': 0}}, update={'$set': {'flag': 0}})
if __name__ == '__main__':
    _ganji_channels.drop()
    channels = get_channel_urls('http://bj.ganji.com/wu/')
    print(len(channels))
    # 首先执行解析类目程序
    pool = Pool()
    pool.map(get_all_links_from, channels)
    pool.close()
    pool.join()
    urls = list(map(lambda url: url['url'], _ganji_urls.find({'flag': {'$eq': 0}})))
    pool = Pool()
    pool.map(get_info, urls)
    pool.close()
    pool.join()

page_parsing.py ： 解析模块，主要用来解析商品列表和商品详情

from bs4 import BeautifulSoup
import requests
import pymongo
import sys
import random

client = pymongo.MongoClient('localhost', 27017)
walden = client['walden']
_ganji_urls = walden['_ganji_urls']
_ganji_infos = walden['_ganji_infos']

headers = {
    'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_11_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/47.0.2526.111 Safari/537.36',
    'Connection': 'keep-alive'
}
# 代理网站 http://cn-proxy.com/
proxy_list = [
    '101.96.11.47:80'
    , '101.96.11.43:80'
    , '101.96.11.42:80'
    , '101.96.11.44:80'
    , '112.5.220.199:80'
    , '111.13.109.56:8080'
]
# 随机获取代理ip
proxy_ip = random.choice(proxy_list)
proxies = {'http': proxy_ip}


def get_links_from(channel, page, who_sells=1):
    # http://bj.ganji.com/jiaju/a1o1/
    list_url = '{}a{}o{}/'.format(channel, who_sells, page)
    wb_data = requests.get(list_url, headers=headers, proxies=proxies)
    if wb_data.status_code != 200:  # 页面不存在则返回 -1
        return -1
    soup = BeautifulSoup(wb_data.text, 'lxml')
    if len(soup.select('ul.pageLink')) == 0:  # 页面下方无页面跳转模块
        print('{} -> {} 结束'.format(list_url, soup.select('ul.pageLink')))
        return 0
    links = soup.select('dl.list-bigpic > dt > a')
    for link in links:
        data = {
            'url': link.get('href')
            , 'flag': 0
        }
        # 跳过转转
        if 'zhuanzhuan' not in data['url'] and _ganji_urls.find({'url': {'$eq': data['url']}}).count() == 0:
            _ganji_urls.insert_one(data)
    return len(links)


def get_item_info(url):
    wb_data = requests.get(url, headers=headers, proxies=proxies)
    # wb_data = requests.get(url, headers=headers)
    soup = BeautifulSoup(wb_data.text, 'lxml')
    if wb_data.status_code != 200:  # 如果没有该页面则跳过
        return
    title = soup.select('h1.title-name')
    date = soup.select('i.pr-5')
    types = soup.select('ul.det-infor > li:nth-of-type(1) > span > a')
    price = soup.select('ul.det-infor > li > i.f22')
    areas = soup.select('ul.det-infor > li:nth-of-type(3) > a')
    newer = soup.select('ul.second-det-infor > li:nth-of-type(1)')
    try:
        data = {
            'url': url
            , 'title': title[0].get_text().strip()
            , 'date': date[0].get_text().strip().strip('发布').strip() if len(date) > 0 else None
            , 'type': [type.text.strip() for type in types]
            , 'price': price[0].get_text().strip() if len(price) > 0 else None
            , 'area': [area.text.strip() for area in areas if area.text.strip() != "-"]
            , 'newer': '{}{}'.format(newer[0].find('label').get_text(),
                                     newer[0].contents[1].strip()) if len(newer) > 0 else None
        }
    except:
        print(
            '异常 : {} -> title={},date={},types={},price={},areas={},newer={}'.format(url, title, date, types, price,
                                                                                     areas, newer))
        print(sys.exc_info())
    else:
        print('{} -> {}'.format(url, data))
        record = _ganji_infos.find_one({'url': {'$eq': url}})
        if record == None:
            _ganji_infos.insert_one(data)
        else:
            _ganji_infos.update({'_id': record['_id']}, data)
        _ganji_urls.find_and_modify(query={'url': url}, update={'$inc': {'flag': 1}})

# print(get_links_from('http://bj.ganji.com/ershoubijibendiannao/', 1))
# get_item_info('http://bj.ganji.com/bangong/2136884202x.htm')

屏幕快照 2016-05-30 上午1.17.13.png

• 一共抓取了20个类目5w多条商品记录
• 发现有些时候自己本地可以访问的页面代理却访问不了