|
123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172 |
-
- import urllib2
- import urllib
- import urlparse
-
- import datetime, time
- import cgi
-
- from BeautifulSoup import BeautifulSoup
-
- from PlanningUtils import PlanningApplication, \
- PlanningAuthorityResults, \
- getPostcodeFromText
-
- date_format = "%d/%m/%Y"
-
- class KensingtonParser:
-
- def __init__(self, *args):
-
- self.authority_name = "The Royal Borough of Kensington and Chelsea"
- self.authority_short_name = "Kensington and Chelsea"
- self.base_url = "http://www.rbkc.gov.uk/Planning/scripts/weeklyresults.asp"
-
- self._results = PlanningAuthorityResults(self.authority_name, self.authority_short_name)
-
-
- def getResultsByDayMonthYear(self, day, month, year):
- search_day = datetime.date(year, month, day)
-
- # We want the sunday of the week being searched for.
- # (sunday is at the end of the week).
- friday = search_day - datetime.timedelta(search_day.weekday()) + datetime.timedelta(4)
-
- # Not using urllib.urlencode as it insists on turning the "+" into "%2B"
- post_data = "WeekEndDate=%d%%2F%d%%2F%d&order=Received+Date&submit=search" %(friday.day, friday.month, friday.year)
-
-
- # Now get the search page
- response = urllib2.urlopen(self.base_url, post_data)
- soup = BeautifulSoup(response.read())
-
- trs = soup.find("table", summary="Planning Application search results table").findAll("tr")[1:]
-
- for tr in trs:
- application = PlanningApplication()
-
- tds = tr.findAll("td")
-
- # Not sure why these are entities. We'll convert them back.
- application.council_reference = tds[0].a.contents[1].strip().replace("/", "/")
- application.info_url = urlparse.urljoin(self.base_url, tds[0].a['href'])
- application.comment_url = application.info_url
-
- application.date_received = datetime.datetime(*(time.strptime(tds[1].string.strip(), date_format)[0:6]))
-
- application.address = tds[2].string.strip()
- application.postcode = getPostcodeFromText(application.address)
-
- application.description = tds[3].string.strip()
-
- self._results.addApplication(application)
-
- return self._results
-
- def getResults(self, day, month, year):
- return self.getResultsByDayMonthYear(int(day), int(month), int(year)).displayXML()
-
- if __name__ == '__main__':
- parser = KensingtonParser()
- print parser.getResults(11,6,2008)
-
|