| 5291 |
varun.gupt |
1 |
'''
|
|
|
2 |
Created on 24-May-2012
|
|
|
3 |
|
|
|
4 |
@author: Varun Gupta
|
|
|
5 |
'''
|
|
|
6 |
from BeautifulSoup import BeautifulSoup
|
|
|
7 |
from BaseScraper import BaseScraper
|
|
|
8 |
from Utils import removePriceFormatting
|
|
|
9 |
import json
|
|
|
10 |
|
|
|
11 |
class SnapdealScraper(BaseScraper):
|
|
|
12 |
|
|
|
13 |
pageCount = {}
|
|
|
14 |
productCountPerScraping = 20
|
|
|
15 |
|
|
|
16 |
def __init__(self):
|
|
|
17 |
BaseScraper.__init__(self)
|
|
|
18 |
self.url = None
|
|
|
19 |
self.id = None
|
|
|
20 |
self.currentPage = None
|
|
|
21 |
self.category = None
|
|
|
22 |
|
|
|
23 |
def setUrl(self, url):
|
|
|
24 |
self.url = url
|
|
|
25 |
urlChunks = url.split('/')
|
|
|
26 |
|
|
|
27 |
self.category = urlChunks[7]
|
|
|
28 |
self.currentPage = 1 + (int(urlChunks[8]) / SnapdealScraper.productCountPerScraping)
|
|
|
29 |
|
|
|
30 |
def scrape(self):
|
|
|
31 |
str = BaseScraper.read(self, self.url)
|
|
|
32 |
self.json = json.loads(str)
|
|
|
33 |
self.setPageCount()
|
|
|
34 |
|
|
|
35 |
def getPhones(self):
|
|
|
36 |
phones = []
|
|
|
37 |
|
|
|
38 |
for product in self.json['productResponseDTO']['productDtos']:
|
|
|
39 |
phones.append({
|
|
|
40 |
'name': str(product['name']),
|
|
|
41 |
'price': product['voucherPrice'],
|
|
|
42 |
'source': 'snapdeal',
|
|
|
43 |
'product_url': str(product['pageUrl']),
|
|
|
44 |
'in_stock': int(not bool(product['soldOut']))
|
|
|
45 |
})
|
|
|
46 |
return phones
|
|
|
47 |
|
|
|
48 |
def getNextUrl(self):
|
|
|
49 |
if self.currentPage < SnapdealScraper.pageCount[self.category]:
|
|
|
50 |
return 'http://www.snapdeal.com/json/product/get/search/%s/%s/%s?q=&sort=plrty&keyword=' % (self.category, self.currentPage * SnapdealScraper.productCountPerScraping, SnapdealScraper.productCountPerScraping)
|
|
|
51 |
else:
|
|
|
52 |
return None
|
|
|
53 |
|
|
|
54 |
def setPageCount(self):
|
|
|
55 |
if SnapdealScraper.pageCount is None or self.category not in SnapdealScraper.pageCount:
|
|
|
56 |
resultCount = int(self.json['productResponseDTO']['numberFound'])
|
|
|
57 |
SnapdealScraper.pageCount[self.category] = 1 + int(resultCount / SnapdealScraper.productCountPerScraping)
|
|
|
58 |
|
|
|
59 |
def getDataFromProductPage(self, url):
|
|
|
60 |
html = BaseScraper.read(self, url)
|
|
|
61 |
soup = BeautifulSoup(html)
|
|
|
62 |
name = str(soup.find('div', {'class': 'prodtitle-head'}).find('h1').string.strip())
|
|
|
63 |
price = removePriceFormatting(str(soup.find('span', {'id': 'selling-price-id'}).string.strip()))
|
|
|
64 |
|
|
|
65 |
data = {
|
|
|
66 |
"product_url": str(url),
|
|
|
67 |
"source": "snapdeal",
|
|
|
68 |
"price": price,
|
|
|
69 |
"in_stock": 1,
|
|
|
70 |
"name": name
|
|
|
71 |
}
|
|
|
72 |
return data
|
|
|
73 |
|
|
|
74 |
if __name__ == '__main__':
|
|
|
75 |
s = SnapdealScraper()
|
| 5401 |
varun.gupt |
76 |
s.setUrl('http://www.snapdeal.com/json/product/get/search/175/480/20?q=&sort=plrty&keyword=')
|
|
|
77 |
s.scrape()
|
|
|
78 |
print s.getPhones()
|
|
|
79 |
print s.getNextUrl()
|
|
|
80 |
#print s.getDataFromProductPage('http://www.snapdeal.com/product/mobiles-mobile-phones/sony-ericsson-xperia-mini-pro-sk17i-black-10052?pos=31;493')
|