Scrapy抓取日本乐天的前五大品牌洗衣机的评论

rakuten.co.jp是日本比较老牌的电商网站,跟日本亚马逊差不多,是日本用得比较多的电商购物网站。
不过跟中国的京东和淘宝天猫比起来就是小巫见大巫了。


抓取的逻辑直接看代码即可,不用过多解释。

Scrapy主代码:

import urllib
import re
import json
from random import choice
from scrapy.spiders import Spider
from scrapy.http import FormRequest
from scrapy import Request
import time
import copy
from scrapy.selector import Selector


class Rakuten(Spider):

def __init__(self):
self.d = {}
# self.with_com_item_count = {}
self.url_list = [
# 'https://review.rakuten.co.jp/item/1/270903_10431504/1.1/', # 测试评论
'https://search.rakuten.co.jp/search/mall/-/204491/tg1004749/', # 海尔
'https://search.rakuten.co.jp/search/mall/-/204491/tg1002881/', # 东芝
'https://search.rakuten.co.jp/search/mall/-/204491/tg1004726/', # 日立
'https://search.rakuten.co.jp/search/mall/-/204491/tg1002867/', # 松下
'https://search.rakuten.co.jp/search/mall/-/204491/tg1002860/', # 夏普

]

name = 'r'

def start_requests(self):
self.logger.warning("rakuten start_requests请求入口")
for url in self.url_list:
yield Request(url,dont_filter=True,
meta={'cookiejar': 1},
callback=self.parse_item_pages,
)

def parse_item_pages(self, response):
url = response.url
item = {}
if '/204491/tg1004749/' in url:
item['brand_name'] = '海尔'
elif '/204491/tg1002881/' in url:
item['brand_name'] = '东芝'
elif '/204491/tg1004726/' in url:
item['brand_name'] = '日立'
elif '/204491/tg1002867/' in url:
item['brand_name'] = '松下'
elif '/204491/tg1002860/' in url:
item['brand_name'] = '夏普'

# 主URL搜索的商品条数
total_num = 0
total_num_str = '//h1[@class="section"]/text()[2]'
total_num_result = response.xpath(total_num_str).re('[0-9]+')
if total_num_result:
total_num = int(total_num_result[0])
item['total_items_num'] = total_num
# 分页数 每页45条
page_num = 0
if total_num:
page_num = total_num // 45
if total_num % 45 != 0:
page_num += 1
item['total_items_page_num'] = page_num
meta = {'cookiejar': response.meta['cookiejar'], 'item': item} # 传递item
urls = [url] # 构造url列表
p = 2
while p <= page_num:
urls.append(url + '?p=' + str(p))
p += 1
for url in urls:
yield FormRequest.from_response(response, url=url, meta=meta, callback=self.parse_one_item_page)


def parse_one_item_page(self, response):
self.logger.warning('进入parse 处理一个商品列表分页')
if response.status == 200:
url = response.url
d = copy.deepcopy(response.meta['item'])
self.logger.warning('访问商品列表分页成功:{}'.format(url))
total_num = d['total_items_num']
page_num = d['total_items_page_num']
# 当前分页所有有评论的链接
with_com_links_str = '//a[@class="dui-rating-filter _link"]'
with_com_links_result = response.xpath(with_com_links_str).extract()
for with_com_item in with_com_links_result:
item = copy.deepcopy(response.meta['item'])
item['item_belong2_page_url'] = url # 商品列表分页url
find_link_reg = re.compile('href="(.*?)"')
find_link_result = find_link_reg.findall(with_com_item)
if find_link_result:
item['comment_url'] = find_link_result[0] # 评论详情页
find_score_reg = re.compile('<span class="score">(.*?)</span>')
find_score_result = find_score_reg.findall(with_com_item)
if find_score_result:
item['avg_score'] = float(find_score_result[0]) # 平均评分
find_score_num_reg = re.compile('<span class="legend">\((.*?)件\)</span>')
find_score_num_result = find_score_num_reg.findall(with_com_item)
if find_score_num_result:
item['score_people_count'] = int(find_score_num_result[0]) # 评论人数
comment_page_count = 0 # 评论分页
comment_page_count = item['score_people_count'] // 15
if item['score_people_count'] % 15 != 0:
comment_page_count += 1
item['comment_page_count'] = comment_page_count

if item['comment_url']:
yield Request(item['comment_url'], dont_filter=True,
callback=self.parse_comment_page,
meta={'item': item, 'cookiejar': response.meta['cookiejar']}
)
else:
self.logger.error(response.url)
self.logger.error('状态码为:{}'.format(response.status))


def parse_comment_page(self, response):
self.logger.warning('处理评论页面:{}'.format(response.url))
d = copy.deepcopy(response.meta['item'])
# 商品名
name_str = '//a[@sid_linkname="item_01"]/text()'
name = response.xpath(name_str).extract_first()
d['item_name'] = name.replace(u'\u3000', u' ')

# 商品主页 和 商品id
item_str = '//a[@sid_linkname="item_01"]'
item = response.xpath(item_str).extract_first()
find_href = re.compile('href="(.*?)"')
find_result = find_href.findall(item)
if find_result:
find_result = find_result[0]
d['item_url'] = find_result
d['item_id'] = find_result.split('/')[-2]

if d['item_url'] not in self.d:
self.d[d['item_url']] = d

if 'comment_list' not in self.d[d['item_url']]:
self.d[d['item_url']]['comment_list'] = [] # 元素是dict 存放所有的评论信息

# 处理单页信息 当前页
coms_str = '//div[@class="revRvwUserMain"]'
coms_results = response.xpath(coms_str).extract()

# self.logger.warning('处理当前页的15条评论 开始')
for coms_result in coms_results:
# 处理单个评论
single_com = {}
# 用户评分
user_score = 0
user_score_reg = re.compile('<span class="revUserRvwerNum value">(.*?)</span>')
user_score_result = user_score_reg.findall(coms_result)
if user_score_result:
user_score = int(user_score_result[0])
single_com['user_score'] = user_score

# 用户评论指标
com_subjects_reg = re.compile('<li class="revUserDispList">(.*?):<span class="revDispListTxt">(.*?)</span></li>')
com_subjects_result = com_subjects_reg.findall(coms_result)
if com_subjects_result:
td = {}
for k, v in com_subjects_result:
td[k] = v
single_com['com_subjects'] = td

# 用户主评论
com_summary_reg = re.compile('<dt class="revRvwUserEntryTtl summary">(.*?)</dt>')
com_summary_result = com_summary_reg.findall(coms_result)
if com_summary_result:
single_com['com_summary'] = com_summary_result[0]

# 用户评论详情
# com_detail_reg = re.compile('"revRvwUserEntryCmt description">(.*?)</dd>')
# com_detail_result = com_detail_reg.findall(coms_result)
# if com_detail_result:
# single_com['com_detail'] = com_detail_result[0]

com_list = Selector(text=coms_result).xpath(
'//dd[@class="revRvwUserEntryCmt description"]/text()').extract()
single_com['com_detail'] = ''.join(com_list)

# 评论日期
com_date_reg = re.compile('<span class="revUserEntryDate dtreviewed">(.*?)</span>')
com_date_result = com_date_reg.findall(coms_result)
if com_date_result:
single_com['com_date'] = com_date_result[0]

if single_com not in self.d[d['item_url']]['comment_list']:
self.d[d['item_url']]['comment_list'].append(single_com)


# self.logger.warning('处理当前页的15条评论 结束')
# self.logger.error(self.d)
# self.logger.error(len(self.d[d['item_id']]['comment_list']))

# self.logger.warning('处理下一页')
# # 处理评论分页
item = copy.deepcopy(response.meta['item'])
url = response.url
p = int(url.split('/')[-2].split('.')[0]) # 当前页数
page_num = item['comment_page_count'] # 评论总页数
# self.logger.warning('评论总页数:' + str(page_num))
# self.logger.warning('当前页数:' + str(p))
if p < page_num:
url = url.split('/')
url[-2] = str(p+1) + '.1'
url = '/'.join(url)
yield Request(
url, callback=self.parse_comment_page,
meta={'item': item, 'cookiejar': response.meta['cookiejar']}
)
elif p == page_num:
self.logger.error('处理完一个商品的评论+++++++++++++++++++++++++++++++')
self.d[d['item_url']]['comment_count'] = len(self.d[d['item_url']]['comment_list'])
yield self.d[d['item_url']]

直接写入MongoDB即可,可以方便地转为json和csv等格式。

结果简单说明,TOP5洗衣机品牌的商品评论非常的少,共190个商品条目左右。有评论的也绝大多数在一页以内(15条)。可以说是非常的冷清,相比京东的家电销量和评论。这个也跟之前的预期相符,海尔在日本卖得很好,这个有点意外。

日本电商重实体,实体电商是非常火爆的,服务也是别的国家所不能望其项背的。所以线上电商这么萧条也可以理解。