def __init__(self, service_user=None):
if service_user:
self.service_user = service_user
self.scraper_settings_key = SolutionNewsScraperSettings.create_key(self.service_user)
self.sln_settings = get_solution_settings(self.service_user)
if BROADCAST_TYPE_NEWS not in self.sln_settings.broadcast_types:
raise NewsScraperException(
'Cannot check for news in %s because no broadcast type with name \'%s\' is found',
self.sln_settings.name, BROADCAST_TYPE_NEWS)
self.broadcast_type = transl(BROADCAST_TYPE_NEWS, self.sln_settings.main_language)
def _check_for_news(service_user):
sln_settings = get_solution_settings(service_user)
if BROADCAST_TYPE_NEWS not in sln_settings.broadcast_types:
logging.error(
"check_for_news_in_be_sint_lievens_houtem failed no broadcast type found with name '%s'",
BROADCAST_TYPE_NEWS)
return
broadcast_type = transl(BROADCAST_TYPE_NEWS, sln_settings.main_language)
url = u"https://www.sint-lievens-houtem.be/rss.xml"
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.error(
"Could not check for news in be_sint_lievens_houtem.\n%s" %
response.content)
return
sln_news_scraper_settings_key = SolutionNewsScraperSettings.create_key(
service_user)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
return []
return sln_news_scraper_settings.urls
urls = db.run_in_transaction(trans)
doc = minidom.parseString(response.content)
for item in doc.getElementsByTagName('item'):
try:
title = item.getElementsByTagName("title")[0].firstChild.nodeValue
url = u'%s' % item.getElementsByTagName(
"link")[0].firstChild.nodeValue
description_html = item.getElementsByTagName(
"description")[0].firstChild.nodeValue
message, _, _ = parse_html_content(description_html)
except:
logging.debug("title: %s", title)
logging.debug(item.childNodes)
raise
if url in urls:
continue
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
sln_news_scraper_settings = SolutionNewsScraperSettings(
key=sln_news_scraper_settings_key)
sln_news_scraper_settings.urls = []
if url not in sln_news_scraper_settings.urls:
sln_news_scraper_settings.urls.append(url)
sln_news_scraper_settings.put()
deferred.defer(create_news_item,
sln_settings,
broadcast_type,
message,
title,
url,
_transactional=True)
db.run_in_transaction(trans)
def _check_for_news(service_user):
sln_settings = get_solution_settings(service_user)
if BROADCAST_TYPE_NEWS not in sln_settings.broadcast_types:
logging.error(
"check_for_news_in_be_laarne failed no broadcast type found with name '%s'",
BROADCAST_TYPE_NEWS)
return
broadcast_type = transl(BROADCAST_TYPE_NEWS, sln_settings.main_language)
url = u"http://www.laarne.be/website/437-www"
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.error("Could not check for news in be_laarne.\n%s" %
response.content)
return
tree = html.fromstring(response.content.decode("utf8"))
sln_news_scraper_settings_key = SolutionNewsScraperSettings.create_key(
service_user)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
return []
return sln_news_scraper_settings.urls
urls = db.run_in_transaction(trans)
ul = tree.xpath('//div[@id="nieuws"]//ul')
if not ul:
logging.error("Could not find ul for news in be_laarne")
return
news_items = []
for item in ul[0].getchildren():
if item.tag == "li":
news_items.append([])
if item.tag == "li" or item.tag == "a":
news_items[-1].append(item)
for news_item in news_items:
if len(news_item) != 2:
continue
url = news_item[1].xpath("@href")[0]
if not (url.startswith("http://") or url.startswith("https://")):
url = u"http://www.laarne.be%s" % url
url = unicode(url) if not isinstance(url, unicode) else url
if url in urls:
continue
title = u'%s' % news_item[0].text.replace('\n', ' ')
message = _get_news_details(url)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
sln_news_scraper_settings = SolutionNewsScraperSettings(
key=sln_news_scraper_settings_key)
sln_news_scraper_settings.urls = []
if url not in sln_news_scraper_settings.urls:
sln_news_scraper_settings.urls.append(url)
sln_news_scraper_settings.put()
deferred.defer(create_news_item,
sln_settings,
broadcast_type,
message,
title,
url,
_transactional=True)
db.run_in_transaction(trans)
def _check_for_news(service_user):
sln_settings = get_solution_settings(service_user)
if BROADCAST_TYPE_NEWS not in sln_settings.broadcast_types:
logging.error("check_for_news_in_be_lokeren failed no broadcast type found with name '%s'", BROADCAST_TYPE_NEWS)
return
broadcast_type = transl(BROADCAST_TYPE_NEWS, sln_settings.main_language)
url = u"https://lokeren.be/nieuws"
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.error("Could not check for news in be_lokeren.\n%s" % response.content)
return
tree = html.fromstring(response.content.decode("utf8"))
sln_news_scraper_settings_key = SolutionNewsScraperSettings.create_key(service_user)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
return []
return sln_news_scraper_settings.urls
urls = db.run_in_transaction(trans)
master_div = tree.xpath('//div[@id="content_container"]//div[@class="content"]//div[contains(@class, "newspost")]')[0]
master_div_children = master_div.getchildren()
news_items = []
for item in master_div_children:
if item.tag == "h2":
news_items.append([])
if item.tag == "h2" or item.tag == "p":
news_items[-1].append(item)
for news_item in news_items:
if len(news_item) > 2:
url = news_item[-1].getchildren()[0].xpath("@href")[0]
if not (url.startswith("http://") or url.startswith("https://")):
url = u"https://lokeren.be%s" % url
url = unicode(url) if not isinstance(url, unicode) else url
if url in urls:
continue
title = u'%s' % news_item[0].getchildren()[0].text
message = u''
for i in xrange(1, len(news_item) - 1):
child = news_item[i]
while True:
message = child.text and child.text.strip()
if message:
break
news_short_children = child.getchildren()
if not news_short_children:
break
child = news_short_children[0]
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
sln_news_scraper_settings = SolutionNewsScraperSettings(key=sln_news_scraper_settings_key)
sln_news_scraper_settings.urls = []
if url not in sln_news_scraper_settings.urls:
sln_news_scraper_settings.urls.append(url)
sln_news_scraper_settings.put()
deferred.defer(create_news_item, sln_settings, broadcast_type, message, title, url,
_transactional=True)
db.run_in_transaction(trans)
def _check_for_news(service_user):
sln_settings = get_solution_settings(service_user)
if BROADCAST_TYPE_NEWS not in sln_settings.broadcast_types:
logging.error("check_for_news_in_be_dendermonde failed no broadcast type found with name '%s'", BROADCAST_TYPE_NEWS)
return
broadcast_type = transl(BROADCAST_TYPE_NEWS, sln_settings.main_language)
url = u"http://www.dendermonde.be/rssout.aspx?cat=N"
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.error("Could not check for news in be_dendermonde.\n%s" % response.content)
return
sln_news_scraper_settings_key = SolutionNewsScraperSettings.create_key(service_user)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
return []
return sln_news_scraper_settings.urls
urls = db.run_in_transaction(trans)
doc = minidom.parseString(response.content)
for item in doc.getElementsByTagName('item'):
try:
title = item.getElementsByTagName("title")[0].firstChild.nodeValue
url = item.getElementsByTagName("link")[0].firstChild.nodeValue
url = unicode(url) if not isinstance(url, unicode) else url
if url in urls:
continue
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.warn('Received status code %d from %s with content:\n', response.status_code, url,
response.content)
continue
tree = html.fromstring(response.content.decode("utf8"))
div = tree.xpath('//div[@class="short box"]')
if not div:
logging.error('News scraper for dendermonde needs to be updated')
continue
message = u'%s' % div[0].text
except Exception:
logging.debug("title: %s", title)
logging.debug(item.childNodes)
raise
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
sln_news_scraper_settings = SolutionNewsScraperSettings(key=sln_news_scraper_settings_key)
sln_news_scraper_settings.urls = []
if url not in sln_news_scraper_settings.urls:
sln_news_scraper_settings.urls.append(url)
sln_news_scraper_settings.put()
deferred.defer(create_news_item, sln_settings, broadcast_type, message, title, url,
_transactional=True)
db.run_in_transaction(trans)
def _check_for_news(service_user):
sln_settings = get_solution_settings(service_user)
if BROADCAST_TYPE_NEWS not in sln_settings.broadcast_types:
logging.error(
"check_for_news_in_be_lede failed no broadcast type found with name '%s'",
BROADCAST_TYPE_NEWS)
return
broadcast_type = transl(BROADCAST_TYPE_NEWS, sln_settings.main_language)
url = u"http://www.lede.be/nieuwsoverzicht.Aspx"
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.error("Could not check for news in be_lede.\n%s" %
response.content)
return
tree = html.fromstring(response.content.decode("utf8"))
sln_news_scraper_settings_key = SolutionNewsScraperSettings.create_key(
service_user)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
return []
return sln_news_scraper_settings.urls
urls = db.run_in_transaction(trans)
news_items = tree.xpath(
'//div[@class="nwsOverview"]')[0].getchildren()[3].getchildren()
for li in news_items:
for item in li.getchildren():
if item.tag == "a":
url = item.xpath("@href")[0]
if not (url.startswith("http://")
or url.startswith("https://")):
url = u"http://www.lede.be%s" % url
url = unicode(url) if not isinstance(url, unicode) else url
if url in urls:
continue
message = _get_news_details(url)
title = u'%s' % item.text
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
sln_news_scraper_settings = SolutionNewsScraperSettings(
key=sln_news_scraper_settings_key)
sln_news_scraper_settings.urls = []
if url not in sln_news_scraper_settings.urls:
sln_news_scraper_settings.urls.append(url)
sln_news_scraper_settings.put()
deferred.defer(create_news_item,
sln_settings,
broadcast_type,
message,
title,
url,
_transactional=True)
db.run_in_transaction(trans)
def _check_for_news(service_user):
sln_settings = get_solution_settings(service_user)
if BROADCAST_TYPE_NEWS not in sln_settings.broadcast_types:
logging.error(
"check_for_news_in_be_zele failed no broadcast type found with name '%s'",
BROADCAST_TYPE_NEWS)
return
broadcast_type = transl(BROADCAST_TYPE_NEWS, sln_settings.main_language)
url = u"https://www.zele.be/nieuws"
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.error("Could not check for news in be_zele.\n%s" %
response.content)
return
tree = html.fromstring(response.content.decode("utf8"))
sln_news_scraper_settings_key = SolutionNewsScraperSettings.create_key(
service_user)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
return []
return sln_news_scraper_settings.urls
urls = db.run_in_transaction(trans)
news_items = tree.xpath(
'//div[@class="news-partial "]//div[@class="partial-container"]//ul//li'
)
for news_item in news_items:
a = news_item.getchildren()[0]
url = u'%s' % a.xpath("@href")[0]
if not (url.startswith("http://") or url.startswith("https://")):
url = u"https://www.zele.be%s" % url
if url in urls:
continue
title = None
message = u''
for item in a.getchildren():
if item.tag == "h3":
title = u'%s' % item.text
if item.tag == "div" and item.xpath("@class")[0] == "short":
message = item.text
if title:
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
sln_news_scraper_settings = SolutionNewsScraperSettings(
key=sln_news_scraper_settings_key)
sln_news_scraper_settings.urls = []
if url not in sln_news_scraper_settings.urls:
sln_news_scraper_settings.urls.append(url)
sln_news_scraper_settings.put()
deferred.defer(create_news_item,
sln_settings,
broadcast_type,
message,
title,
url,
_transactional=True)
db.run_in_transaction(trans)
def _check_for_news(service_user):
sln_settings = get_solution_settings(service_user)
if BROADCAST_TYPE_NEWS not in sln_settings.broadcast_types:
logging.error(
"check_for_news_in_be_wetteren failed no broadcast type found with name '%s'",
BROADCAST_TYPE_NEWS)
return
broadcast_type = transl(BROADCAST_TYPE_NEWS, sln_settings.main_language)
url = u"http://www.wetteren.be/nieuwsoverzicht/505/default.aspx?_vs=0_N"
response = urlfetch.fetch(url, deadline=60)
if response.status_code != 200:
logging.error("Could not check for news in be_wetteren.\n%s" %
response.content)
return
tree = html.fromstring(response.content.decode("utf8"))
sln_news_scraper_settings_key = SolutionNewsScraperSettings.create_key(
service_user)
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
return []
return sln_news_scraper_settings.urls
urls = db.run_in_transaction(trans)
for title_node in tree.xpath(
'//div[@id="overzicht"]//ul[@class="nieuws"]//li[@class="normal"]'
):
_, a_title, _, span_detail = title_node.getchildren()
title = u'%s' % a_title.text
url = u'%s' % a_title.xpath("@href")[0]
if not (url.startswith("http://") or url.startswith("https://")):
url = u"http://www.wetteren.be%s" % url
if url in urls:
continue
child = span_detail
message = u''
while True:
message = child.text and child.text.strip()
if message:
break
news_short_children = child.getchildren()
if not news_short_children:
break
child = news_short_children[0]
if message:
message = u'%s' % message
def trans():
sln_news_scraper_settings = SolutionNewsScraperSettings.get(
sln_news_scraper_settings_key)
if not sln_news_scraper_settings:
sln_news_scraper_settings = SolutionNewsScraperSettings(
key=sln_news_scraper_settings_key)
sln_news_scraper_settings.urls = []
if url not in sln_news_scraper_settings.urls:
sln_news_scraper_settings.urls.append(url)
sln_news_scraper_settings.put()
deferred.defer(create_news_item,
sln_settings,
broadcast_type,
message,
title,
url,
_transactional=True)
db.run_in_transaction(trans)