| 183 |
ashish |
1 |
'''
|
|
|
2 |
Created on 20-May-2010
|
|
|
3 |
|
|
|
4 |
@author: gaurav
|
|
|
5 |
'''
|
|
|
6 |
|
|
|
7 |
from scrapy.spider import BaseSpider
|
|
|
8 |
from scrapy.selector import HtmlXPathSelector
|
|
|
9 |
from scrapy.http import Request
|
|
|
10 |
|
|
|
11 |
from demo.items import DemoItem
|
|
|
12 |
from scrapy.contrib.spidermiddleware import referer
|
|
|
13 |
from scrapy.http.headers import Headers
|
|
|
14 |
from scrapy.http.request.form import FormRequest
|
|
|
15 |
from scrapy.log import msg
|
|
|
16 |
from scrapy.http.response import Response
|
|
|
17 |
|
|
|
18 |
from datastore import DataAccessor
|
|
|
19 |
from datastore.DataAccessor import DataHelper
|
| 231 |
ashish |
20 |
from html2text.unescaping import *
|
| 183 |
ashish |
21 |
|
|
|
22 |
class mobilestore_spider(BaseSpider):
|
|
|
23 |
|
|
|
24 |
def __init__(self):
|
| 231 |
ashish |
25 |
MOBILESTORE_DOMAINNAME = "mobilestore"
|
|
|
26 |
self.domain_name = MOBILESTORE_DOMAINNAME
|
|
|
27 |
MOBILESTORE_URL = "http://www.themobilestore.in/sitemap.xml"
|
|
|
28 |
self.start_urls.append(MOBILESTORE_URL)
|
| 183 |
ashish |
29 |
|
|
|
30 |
def start_requests(self):
|
| 231 |
ashish |
31 |
#adding entry for the supplier i.e its name and site
|
|
|
32 |
MOBILESTORE_HOMEPAGE = "www.themobilestore.in"
|
| 183 |
ashish |
33 |
da = DataHelper()
|
| 231 |
ashish |
34 |
da.add_supplier(self.domain_name, MOBILESTORE_HOMEPAGE)
|
| 183 |
ashish |
35 |
listreq = []
|
| 231 |
ashish |
36 |
|
|
|
37 |
#for each request a referer has to be set
|
|
|
38 |
MOBILESTORE_REFERER = "www.google.com/search"
|
| 183 |
ashish |
39 |
for url1 in self.start_urls:
|
|
|
40 |
request = Request(url = str(url1), callback=self.parse)
|
| 231 |
ashish |
41 |
request.headers.setdefault("Referer", MOBILESTORE_REFERER)
|
| 183 |
ashish |
42 |
listreq.append(request)
|
|
|
43 |
return listreq
|
| 231 |
ashish |
44 |
|
| 183 |
ashish |
45 |
def parse(self, response):
|
|
|
46 |
da = DataHelper()
|
|
|
47 |
hxs = HtmlXPathSelector(response)
|
| 231 |
ashish |
48 |
MOBILESTORE_XPATH1 = '//url/loc/text()'
|
|
|
49 |
phone_urls = hxs.select(MOBILESTORE_XPATH1)
|
|
|
50 |
#elements in chk_list are specific to this site for determining valid sites
|
|
|
51 |
MOBILESTORE_CHKLIST1 = ["Mobile_Phones"]
|
| 183 |
ashish |
52 |
for i in phone_urls:
|
|
|
53 |
site = i.extract()
|
| 231 |
ashish |
54 |
site = unescape(site)
|
| 183 |
ashish |
55 |
pos1 = pos2 = pos3 = 0
|
|
|
56 |
temp =""
|
|
|
57 |
pos1 = site.rfind('/')
|
|
|
58 |
if pos1 != -1:
|
|
|
59 |
pos2 = site.rfind('/',0,pos1-1)
|
|
|
60 |
if pos2 != -1:
|
|
|
61 |
pos3 = site.rfind('/',0,pos2-1)
|
|
|
62 |
if pos3 > 0:
|
|
|
63 |
temp = site[pos3+1:pos2]
|
| 231 |
ashish |
64 |
# adding valid urls to the DB from the site-map
|
|
|
65 |
for m in MOBILESTORE_CHKLIST1:
|
|
|
66 |
if temp == m:
|
|
|
67 |
da.add_mobstoreurl(site)
|
| 183 |
ashish |
68 |
SPIDER = mobilestore_spider()
|