Exemplo n.º 1
0
class WikiSpider(CrawlSpider):
    ''' Can spider MediaWiki and Twiki wikis.
    '''
    
    domain_name = config.get('wiki', 'domain_name')
    extra_domain_names = config.get('wiki', 'extra_domain_names').split(',')
    start_urls = config.get('wiki', 'start_urls').split(',')
    
    rules = [
        Rule(SgmlLinkExtractor(deny=["\?", "Special:.*"]), 
             callback='parse_wiki', 
             follow=True),
    ]
    
    def parse_wiki(self, response):
        hxs = HtmlXPathSelector(response)
        
        item = IndexableItem()
        item['type'] = 'wiki'
        item['title'] = hxs.select('//title/text()').extract()[0]
        item['url'] = response.url
        item['mod_datetime'] = datetime.now().isoformat()

        to_exclude = ["[not(self::%s)]" % x for x in ["script", "style"]]
        content_xpath = "//body//*%s/text()" % ("".join(to_exclude))
        item['contents'] = "\n".join([
            s.strip() for s 
            in hxs.select(content_xpath).extract() 
            if s.strip()
        ])
        
        return item
Exemplo n.º 2
0
    def __init__(self, domain_name=None):
        BaseSpider.__init__(self, domain_name)

        consumer_key = config.get('yammer', 'consumer_key')
        consumer_secret = config.get('yammer', 'consumer_secret')
        app_token = config.get('yammer', 'app_token')

        self.consumer = OAuthConsumer(consumer_key, consumer_secret)
        self.signature = OAuthSignatureMethod_PLAINTEXT()
        self.token = OAuthToken.from_string(app_token)
Exemplo n.º 3
0
    def __init__(self, domain_name=None):
        BaseSpider.__init__(self, domain_name)

        consumer_key    = config.get('yammer', 'consumer_key')
        consumer_secret = config.get('yammer', 'consumer_secret')
        app_token       = config.get('yammer', 'app_token')

        self.consumer  = OAuthConsumer(consumer_key, consumer_secret)
        self.signature = OAuthSignatureMethod_PLAINTEXT()
        self.token     = OAuthToken.from_string(app_token)
Exemplo n.º 4
0
class YammerSpider(BaseSpider):
    domain_name = 'yammer.com'
    start_urls = [config.get('yammer', 'json_messages_url')]

    def __init__(self, domain_name=None):
        BaseSpider.__init__(self, domain_name)

        consumer_key = config.get('yammer', 'consumer_key')
        consumer_secret = config.get('yammer', 'consumer_secret')
        app_token = config.get('yammer', 'app_token')

        self.consumer = OAuthConsumer(consumer_key, consumer_secret)
        self.signature = OAuthSignatureMethod_PLAINTEXT()
        self.token = OAuthToken.from_string(app_token)

    def make_requests_from_url(self, url):
        oauth_request = OAuthRequest.from_consumer_and_token(self.consumer,
                                                             token=self.token,
                                                             http_method='GET',
                                                             http_url=url)
        oauth_request.sign_request(self.signature, self.consumer, self.token)

        return Request(oauth_request.to_url(),
                       callback=self.parse,
                       dont_filter=True)

    def parse(self, response):
        data = json.loads(response.body)

        for message in data['messages']:
            item = YammerItem()
            item['type'] = 'yammer'
            item['url'] = message['web_url']

            body = message['body']['plain']
            item['contents'] = body

            max_title_length = 40
            if len(body) > 40:
                item['title'] = body[0:max_title_length - 3] + '...'
            else:
                item['title'] = body

            item['mod_datetime'] = message['created_at']
            item['author'] = message['sender_id']
            item['parent_url'] = message['replied_to_id']
            item['thread_url'] = message['thread_id']
            item['likes'] = message['liked_by']

            yield item
Exemplo n.º 5
0
 def __init__(self):
     server = Server(config.get('couchdb', 'host'))
     self.conn = server[config.get('couchdb', 'db')]