| 223 |
ashish |
1 |
'''
|
|
|
2 |
Created on 06-Jun-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.DataAccessor import *
|
|
|
19 |
from datastore.DataCodeAccessor import *
|
|
|
20 |
from html2text import *
|
|
|
21 |
from babel.messages.pofile import unescape
|
|
|
22 |
import urllib
|
|
|
23 |
|
|
|
24 |
class babuchak3(BaseSpider):
|
|
|
25 |
|
|
|
26 |
def __init__(self):
|
|
|
27 |
initialize_table()
|
|
|
28 |
#BABUCHAK_DOMAINNAME2 = "babuchak2"
|
|
|
29 |
BABUCHAK_DOMAINNAME2 = get_code_word("BABUCHAK_DOMAINNAME2")
|
|
|
30 |
self.domain_name = BABUCHAK_DOMAINNAME2
|
|
|
31 |
da = DataHelper()
|
|
|
32 |
for item in da.get_allbabuchakphoneurls():
|
|
|
33 |
self.start_urls.append(item.url)
|
|
|
34 |
|
|
|
35 |
def start_requests(self):
|
|
|
36 |
listreq = []
|
|
|
37 |
#for each request a referer has to be set
|
|
|
38 |
#BABUCHAK_REFERER = "www.google.com/search"
|
|
|
39 |
BABUCHAK_REFERER = get_code_word("BABUCHAK_REFERER")
|
|
|
40 |
for url1 in self.start_urls:
|
|
|
41 |
request = Request(url = str(url1), callback=self.parse)
|
|
|
42 |
request.headers.setdefault("Referer", BABUCHAK_REFERER)
|
|
|
43 |
listreq.append(request)
|
|
|
44 |
return listreq
|
|
|
45 |
|
|
|
46 |
def parse(self, response):
|
|
|
47 |
#url1 needed to get complete urls
|
|
|
48 |
da = DataHelper()
|
|
|
49 |
hxs = HtmlXPathSelector(response)
|
|
|
50 |
#BABUCHAK_XPATH5 = '//td[@class="text-header"]/text()'
|
|
|
51 |
BABUCHAK_XPATH5 = get_code_word("BABUCHAK_XPATH5")
|
|
|
52 |
#BABUCHAK_XPATH6 = '//td[@class="xl63"]//strong/span/text()'
|
|
|
53 |
BABUCHAK_XPATH6 = get_code_word("BABUCHAK_XPATH6")
|
|
|
54 |
#BABUCHAK_XPATH7 = '//td[@class="mod-item-body-title"]/b/text()'
|
|
|
55 |
BABUCHAK_XPATH7 = get_code_word("BABUCHAK_XPATH7")
|
|
|
56 |
#BABUCHAK_REMOVELIST = ["Rs.","Rs",",","-","/"]
|
|
|
57 |
#list separated by ';'
|
|
|
58 |
BABUCHAK_REMOVELIST = get_code_word("BABUCHAK_REMOVELIST")
|
|
|
59 |
BABUCHAK_REMOVELIST = BABUCHAK_REMOVELIST.split(';')
|
|
|
60 |
name = hxs.select(BABUCHAK_XPATH5)[0].extract()
|
|
|
61 |
try:
|
|
|
62 |
shown_price = hxs.select(BABUCHAK_XPATH6)[0].extract()
|
|
|
63 |
final_price = hxs.select(BABUCHAK_XPATH6)[2].extract()
|
|
|
64 |
except:
|
|
|
65 |
final_price = shown_price = hxs.select(BABUCHAK_XPATH7)[0].extract()
|
|
|
66 |
|
|
|
67 |
|
|
|
68 |
name = name.strip()
|
|
|
69 |
shown_price = shown_price.strip()
|
|
|
70 |
final_price = final_price.strip()
|
|
|
71 |
|
|
|
72 |
if shown_price != '':
|
|
|
73 |
for r in BABUCHAK_REMOVELIST:
|
|
|
74 |
while shown_price.find(r) != -1:
|
|
|
75 |
shown_price = shown_price.replace(r, "")
|
|
|
76 |
shown_price = shown_price.strip()
|
|
|
77 |
|
|
|
78 |
if final_price != '':
|
|
|
79 |
for r in BABUCHAK_REMOVELIST:
|
|
|
80 |
while final_price.find(r) != -1:
|
|
|
81 |
final_price = final_price.replace(r, "")
|
|
|
82 |
final_price = final_price.strip()
|
|
|
83 |
ps1 = shown_price.find('.')
|
|
|
84 |
if ps1 != -1:
|
|
|
85 |
shown_price = shown_price[0:ps1]
|
|
|
86 |
final_price = shown_price
|
|
|
87 |
shown_price = int(shown_price)
|
|
|
88 |
final_price = int(final_price)
|
|
|
89 |
if shown_price>final_price:
|
|
|
90 |
try:
|
|
|
91 |
shown_price = hxs.select(BABUCHAK_XPATH6)[1].extract()
|
|
|
92 |
final_price = hxs.select(BABUCHAK_XPATH6)[2].extract()
|
|
|
93 |
except:
|
|
|
94 |
final_price = shown_price = hxs.select(BABUCHAK_XPATH7)[0].extract()
|
|
|
95 |
|
|
|
96 |
name = name.strip()
|
|
|
97 |
shown_price = shown_price.strip()
|
|
|
98 |
final_price = final_price.strip()
|
|
|
99 |
|
|
|
100 |
if shown_price != '':
|
|
|
101 |
for r in BABUCHAK_REMOVELIST:
|
|
|
102 |
while shown_price.find(r) != -1:
|
|
|
103 |
shown_price = shown_price.replace(r, "")
|
|
|
104 |
shown_price = shown_price.strip()
|
|
|
105 |
|
|
|
106 |
if final_price != '':
|
|
|
107 |
for r in BABUCHAK_REMOVELIST:
|
|
|
108 |
while final_price.find(r) != -1:
|
|
|
109 |
final_price = final_price.replace(r, "")
|
|
|
110 |
final_price = final_price.strip()
|
|
|
111 |
ps1 = shown_price.find('.')
|
|
|
112 |
if ps1 != -1:
|
|
|
113 |
shown_price = shown_price[0:ps1]
|
|
|
114 |
final_price = shown_price
|
|
|
115 |
|
|
|
116 |
shown_price = int(shown_price)
|
|
|
117 |
final_price = int(final_price)
|
|
|
118 |
|
|
|
119 |
print name
|
|
|
120 |
print shown_price
|
|
|
121 |
print final_price
|
|
|
122 |
da.add_babuchakphone(name,shown_price,final_price)
|
|
|
123 |
SPIDER = babuchak3()
|