Revision as of 12:17, 13 December 2007

Searching WhitePages

The following Python script uses HTTPLib to search against whitepages.com.au (via a proxy).

#!/usr/bin/env python
#
#
#-------------------------------------------------------------------------------

import re
import sys
import base64
import pprint
import urllib
import httplib

from copy import copy

#-------------------------------------------------------------------------------

PROXY       = 'PROXY:8080'
SITE        = 'www.whitepages.com.au'

connection  = None

#===== Headers =================================================================

#    'Accept' : 'text/plain, text/html',

get_headers = {
   'Accept-Encoding'    : 'gzip, deflate',
   'Accept'             : '*/*',
   'Accept-Language'    : 'en-au',
   'Host'               : SITE,
   'Connection'         : 'Keep-Alive',
   'User-Agent'         : 'Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1; .NET CLR 1.1.4322; .NET CLR 2.0.50727'
}

post_headers = {
   'Content-type'       : 'application/x-www-form-urlencoded',
   'Accept'             : 'text/plain'
}

idx      = 0

#===== Logging =================================================================

def log_req_header(idx, hdr):
   of = open('log/%04d.req' % idx, 'w')
   of.write("%s\n" % pprint.pformat(hdr))
   of.close()

#-------------------------------------------------------------------------------

def log_resp_header(idx, resp):
   of = open('log/%04d.hdr' % idx, 'w')

   of.write("resp.__dict__ ->\n%s\n\n" % pprint.pformat(resp.__dict__))
   of.write("Status %s  Reason [%s]\n" % (resp.status, resp.reason))
   of.write("Msg ->\n%s\n\n" % resp.msg)
   of.write("Msg.__dict__ ->\n%s\n\n" % pprint.pformat(resp.msg.__dict__))

   of.close()

#-------------------------------------------------------------------------------

def log_resp_body(idx, resp_body):
   of = open('log/%04d.bdy' % idx, 'w')
   of.write(resp_body);
   of.close()

#===== Encapsulate the request code ============================================

def request(method, url, params, headers):
   global idx

   print '>>>> %s %s <<<<' % (method, url)

   connection.request(method, url, params, headers)

   resp = connection.getresponse()

   log_req_header(idx, headers)
   log_resp_header(idx, resp)

   resp_body = resp.read()

   log_resp_body(idx, resp_body)

   print resp_body

   idx += 1

   return resp

#===============================================================================

def do():
   global connection
   connection  = httplib.HTTPConnection(PROXY)

   BASE_URL    = 'http://%s' % SITE

   #------------------------------------------------------------------------

   DO       = 'GET'
   URL      = BASE_URL + '/'

   headers = copy(get_headers)

   request(DO, URL, None, headers)


   #------------------------------------------------------------------------

   DO  = 'GET'
   URL = BASE_URL + '/wp/index.jsp'

   headers = copy(get_headers)

   resp = request(DO, URL, None, headers)

   m = re.search('JSESSIONID=(.*);', resp.msg.__dict__['dict']['set-cookie'])

   if m:
      print m.group(1)
      JSESSIONID = m.group(1)

   print JSESSIONID

   #---------------------------------------------------------------------

   DO  = 'POST'
   URL = BASE_URL + '/wp/busSearch.do;jsessionid=%s' % JSESSIONID

   headers = copy(post_headers)

   form_data = {
      'subscriberName' : 'Hard',
      'state'          : 'VIC',
      'suburb'         : ,
      'street'         : ,
      'Search'         : 'Search'
   }

   params = urllib.urlencode(form_data)

   headers['Content-Length'] = len(params)
   headers['Cookie']         = 'JSESSIONID=%s' % JSESSIONID

   request(DO, URL, params, headers)

   #---------------------------------------------------------------------

   URL = BASE_URL + '/wp/busSearch.do'

   form_data = {
      'subscriberName' : 'Hard',
      'state'          : 'VIC',
      'page'           : '2'
   }

   params = urllib.urlencode(form_data)

   headers['Content-Length'] = len(params)
   headers['Cookie']         = 'JSESSIONID=%s' % JSESSIONID

   request(DO, URL, params, headers)
 
#===============================================================================

do()

#-------------------------------------------------------------------------------

This scripts writes the search results into files (page 1 => 'log/0002.bdy' and page 2 => 'log/0003.bdy'). Amend the above code handle more pages of search results being produced.

Parsing the Search Results

The search results HTML looks as follows:

Hard On Tools
- 5 Scoresby Rd Bayswater 3153
- (03) 9720 5199
- <form class='mapForm' action='whereIs.do' method='post'><input type='hidden' name='streetNumber' value='5' /><input type='hidden' name='streetName' value='Scoresby' /><input type='hidden' name='streetType' value='Rd' /><input type='hidden' name='locality' value='Bayswater' /><input type='hidden' name='state' value='VIC' /><input type='hidden' name='placeName' value='Hard On Tools' /><input type='hidden' name='address' value='5 Scoresby Rd Bayswater 3153' /><input type='hidden' name='phoneNumber' value='(03) 9720 5199' /><input type='hidden' name='link' value='1197346075688' /><input type='hidden' name='hashCode' value='152335231501250' /><input type='hidden' name='brandId' value='5' /><input type='hidden' name='logData' value= /><input type='hidden' name='subscriberName' value='Hard On Tools' /><input name='mapSubmit' src='images/result-map.gif' alt='Map' type='image' /></form>
- OR...
- (03) 9738 2882
- Fax
- (03) 9720 0966

Hard Parts Australia Pty Ltd
14 Yiannis Crt Springvale 3171
0418 756 340
<form class='mapForm' action='whereIs.do' method='post'><input type='hidden' name='streetNumber' value='14' /><input type='hidden' name='streetName' value='Yiannis' /><input type='hidden' name='streetType' value='Crt' /><input type='hidden' name='locality' value='Springvale' /><input type='hidden' name='state' value='VIC' /><input type='hidden' name='placeName' value='Hard Parts Australia Pty Ltd' /><input type='hidden' name='address' value='14 Yiannis Crt Springvale 3171' /><input type='hidden' name='phoneNumber' value='0418 756 340' /><input type='hidden' name='link' value='1197346075688' /><input type='hidden' name='hashCode' value='152335231501250' /><input type='hidden' name='brandId' value='5' /><input type='hidden' name='logData' value= /><input type='hidden' name='subscriberName' value='Hard Parts Australia Pty Ltd' /><input name='mapSubmit' src='images/result-map.gif' alt='Map' type='image' /></form>

Hard Parts Victoria

This script makes use of the BeautifulSoup Package (see [xxx]);

#!/usr/bin/env python

import pprint

from BeautifulSoup import BeautifulSoup

doc = open('log/0002.html', 'r')

soup = BeautifulSoup(doc)

# print len(soup('table', { "class" : "table_style"}))

# tables = soup.findAll('table', { "class" : "table_style"})
objs = soup.findAll('div', { "class" : "encap_result"})

pp = pprint.PrettyPrinter(3)

for obj in objs:
   t = obj.find(text=True)

   if t:
      print "==========================================="
      # print t

   #print '%s\n\n' % obj.__dict__
   # print '%s\n\n' % obj

      f    = obj.findAll('span',  { 'class' : 'black'})

      for s in f:
         print 'span="black" -> "%s"' % s.find(text=True)

      f    = obj.findAll('input',  { "name" : 'placeName'})

      for s in f:
         # pp.pprint(s.__dict__)
         # print 'attrMap -> "%s"' % s.attrMap
         print 'placeName -> "%s"' % s.attrMap['value']


      for s in obj.findAll('input',  { "name" : 'placeName'}):
         print 'placeName -> "%s"' % s.attrMap['value']

      for s in obj.findAll('input',  { "name" : 'address'}):
         print 'address -> "%s"' % s.attrMap['value']

      for s in obj.findAll('input',  { "name" : 'locality'}):
         print 'locality -> "%s"' % s.attrMap['value']

      for s in obj.findAll('input',  { "name" : 'streetNumber'}):
         print 'streetNumber -> "%s"' % s.attrMap['value']

      for s in obj.findAll('input',  { "name" : 'streetName'}):
         print 'streetName -> "%s"' % s.attrMap['value']

      for s in obj.findAll('input',  { "name" : 'streetType'}):
         print 'streetType -> "%s"' % s.attrMap['value']


      lis       = obj.findAll('li',  { "class" : None})

      for li in lis:
         print 'li -> "%s"' % li.find(text=True)

      addresses      = obj.findAll('li',  { "class" : "entryData address"})

      for address in addresses:
         print 'addr -> "%s"' % address.find(text=True)

      phone_numbers  = obj.findAll('li',  { "class" : "entryData phone"})

      for phone in phone_numbers:
         print 'phone -> "%s"' % phone.find(text=True)

Difference between revisions of "Parsing WhitePages Search Results HTML"

Revision as of 12:17, 13 December 2007

Contents

Searching WhitePages

Parsing the Search Results

Hard On Tools

OR...

Fax

Hard Parts Australia Pty Ltd

Hard Parts Victoria

Navigation menu

Revision as of 12:17, 13 December 2007 (view source) PeterHarding (talk \| contribs) ← Older edit		Revision as of 12:17, 13 December 2007 (view source) PeterHarding (talk \| contribs) Newer edit →
Line 93:		Line 93:

	return resp		return resp

	#===============================================================================		#===============================================================================