--- a/documents/genericScrapers.py +++ b/documents/genericScrapers.py @@ -3,7 +3,49 @@ import scrape from bs4 import BeautifulSoup import parsedatetime as pdt +from time import mktime +from datetime import datetime +import feedparser import abc + +class GenericRSSDisclogScraper(object): + __metaclass__ = abc.ABCMeta + @abc.abstractmethod + def getAgencyID(self): + """ disclosr agency id """ + return + + @abc.abstractmethod + def getURL(self): + """ disclog URL""" + return + + def getDescription(self, entry, doc): + """ get description from rss entry""" + doc['description'] = entry.summary + return + + def doScrape(self): + foidocsdb = scrape.couch['disclosr-foidocuments'] + (url,mime_type,content) = scrape.fetchURL(scrape.docsdb, self.getURL(), "foidocuments", self.getAgencyID()) + feed = feedparser.parse(content) + for entry in feed.entries: + #print entry + print entry.id + hash = scrape.mkhash(entry.id) + #print hash + doc = foidocsdb.get(hash) + #print doc + if doc == None: + print "saving" + edate = datetime.fromtimestamp(mktime( entry.published_parsed)).strftime("%Y-%m-%d") + doc = {'_id': hash, 'agencyID': self.getAgencyID(), 'url': entry.link, 'docID': entry.id, + "date": edate,"title": entry.title} + self.getDescription(entry, doc) + foidocsdb.save(doc) + else: + print "already saved" + class GenericOAICDisclogScraper(object): __metaclass__ = abc.ABCMeta @abc.abstractmethod @@ -42,7 +84,7 @@ doc = foidocsdb.get(hash) descriptiontxt = "" for string in description.stripped_strings: - descriptiontxt = descriptiontxt + string + descriptiontxt = descriptiontxt + " \n" + string if doc == None: print "saving"