aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorCédric Bonhomme <kimble.mandel@gmail.com>2013-10-13 23:32:15 +0200
committerCédric Bonhomme <kimble.mandel@gmail.com>2013-10-13 23:32:15 +0200
commit8dbb9486cf4972929299de48c91a1e4667313186 (patch)
tree0e138d229ff41857ddfa72c5b9f024cb4f4ea4cc
parentTrying to sort articles (SortedListField(ReferenceField('Article')). (diff)
downloadnewspipe-8dbb9486cf4972929299de48c91a1e4667313186.tar.gz
newspipe-8dbb9486cf4972929299de48c91a1e4667313186.tar.bz2
newspipe-8dbb9486cf4972929299de48c91a1e4667313186.zip
Added mongodb.py (to load articles from previous version of pyAggr3g470r).
-rw-r--r--pyaggr3g470r/mongodb.py283
1 files changed, 283 insertions, 0 deletions
diff --git a/pyaggr3g470r/mongodb.py b/pyaggr3g470r/mongodb.py
new file mode 100644
index 00000000..04cd44fa
--- /dev/null
+++ b/pyaggr3g470r/mongodb.py
@@ -0,0 +1,283 @@
+#! /usr/bin/env python
+# -*- coding: utf-8 -*-
+
+# pyAggr3g470r - A Web based news aggregator.
+# Copyright (C) 2010-2013 Cédric Bonhomme - http://cedricbonhomme.org/
+#
+# For more information : http://bitbucket.org/cedricbonhomme/pyaggr3g470r/
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU General Public License as published by
+# the Free Software Foundation, either version 3 of the License, or
+# (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program. If not, see <http://www.gnu.org/licenses/>
+
+__author__ = "Cedric Bonhomme"
+__version__ = "$Revision: 0.8 $"
+__date__ = "$Date: 2012/03/03 $"
+__revision__ = "$Date: 2013/06/25 $"
+__copyright__ = "Copyright (c) Cedric Bonhomme"
+__license__ = "GPLv3"
+
+import pymongo
+
+class Articles(object):
+ """
+ This class is responsible of the management of the MongoDB
+ database.
+ """
+ def __init__(self, url='localhost', port=27017, db_name="pyaggr3g470r", user="", password=""):
+ """
+ Instantiates the connection.
+ """
+ self.db_name = db_name
+ self.connection = pymongo.connection.Connection(url, port)
+ self.db = pymongo.database.Database(self.connection, self.db_name)
+ if password != "":
+ self.db.authenticate(user, password)
+ collections = self.db.collection_names()
+ for collection_name in collections:
+ if collection_name != "system.indexes":
+ self.db[collection_name].ensure_index([("article_date", pymongo.DESCENDING)], \
+ name="date_index", unique=False, \
+ background=True)
+
+ def add_collection(self, new_collection):
+ """
+ Creates a new collection for a new feed.
+ """
+ collection = self.db[new_collection["feed_id"]]
+ collection.insert(new_collection)
+
+ def add_articles(self, articles, feed_id):
+ """
+ Add article(s) in a collection.
+ """
+ collection = self.db[str(feed_id)]
+ for article in articles:
+ cursor = collection.find({"article_id":article["article_id"]})
+ if cursor.count() == 0:
+ collection.insert(article)
+
+ def delete_feed(self, feed_id):
+ """
+ Delete a collection (feed with all articles).
+ """
+ self.db.drop_collection(feed_id)
+
+ def delete_article(self, feed_id, article_id):
+ """
+ Delete an article.
+ """
+ collection = self.db[str(feed_id)]
+ collection.remove(spec_or_id={"article_id":article_id}, safe=True)
+
+ def get_feed(self, feed_id):
+ """
+ Return information about a feed (collection).
+ Return None if the collection does not exist.
+ """
+ try:
+ return next(self.db[str(feed_id)].find())
+ except:
+ return None
+
+ def get_all_feeds(self, condition=None):
+ """
+ Return all feeds object. The returned list
+ is sorted by alphabetically (by feed name).
+ """
+ feeds = []
+ collections = self.db.collection_names()
+ for collection_name in collections:
+ if collection_name != "system.indexes":
+ if condition is None:
+ cursor = self.db[collection_name].find({"type":0})
+ else:
+ cursor = self.db[collection_name].find({"type":0, condition[0]:condition[1]})
+ if cursor.count() != 0:
+ feeds.append(next(cursor))
+ feeds.sort(key = lambda elem: elem['feed_title'].lower())
+ return feeds
+
+ def get_articles(self, feed_id=None, article_id=None, condition=None, limit=1000000000):
+ """
+ Return one or several articles.
+ The parameter "condition" is an optional requirement, for example:
+ get_articles(feed_id, condition=("article_readed", False)) will
+ return all unread articles of the feed 'feed_id'.
+ """
+ if feed_id == None and article_id == None:
+ # Return all articles.
+ articles = []
+ collections = self.db.collection_names()
+ for collection_name in collections:
+ collection = self.db[collection_name]
+ if condition is None:
+ articles.extend(collection.find({"type":1}, limit=limit))
+ else:
+ articles.extend(collection.find({"type":1, condition[0]:condition[1]}, limit=limit))
+ return articles
+
+ elif feed_id != None and article_id == None:
+ # Return all the articles of a collection.
+ collection = self.db[str(feed_id)]
+ if condition is None:
+ cursor = collection.find({"type":1}, limit=limit)
+ else:
+ cursor = collection.find({"type":1, condition[0]:condition[1]}, limit=limit)
+ return cursor.sort([("article_date", pymongo.DESCENDING)])
+
+ elif feed_id != None and article_id != None:
+ # Return a precise article.
+ collection = self.db[str(feed_id)]
+ try:
+ return next(collection.find({"article_id":article_id}))
+ except:
+ return []
+
+ def get_favorites(self, feed_id=None):
+ """
+ Return favorites articles.
+ """
+ if feed_id is not None:
+ # only for a feed
+ collection = self.db[feed_id]
+ cursor = collection.find({'type':1, 'article_like':True})
+ return cursor.sort([("article_date", pymongo.DESCENDING)])
+ else:
+ favorites = []
+ for feed_id in self.db.collection_names():
+ favorites += self.get_favorites(feed_id)
+ return favorites
+
+ def nb_articles(self, feed_id=None):
+ """
+ Return the number of articles of a feed
+ or of all the database.
+ """
+ if feed_id is not None:
+ collection = self.db[feed_id]
+ cursor = collection.find({'type':1})
+ return cursor.count()
+ else:
+ nb_articles = 0
+ for feed_id in self.db.collection_names():
+ nb_articles += self.nb_articles(feed_id)
+ return nb_articles
+
+ def nb_unread_articles(self, feed_id=None):
+ """
+ Return the number of unread articles of a feed
+ or of all the database.
+ """
+ if feed_id is not None:
+ return self.get_articles(feed_id=feed_id, condition=("article_readed", False)).count()
+ else:
+ return len(self.get_articles(condition=("article_readed", False)))
+
+ def like_article(self, like, feed_id, article_id):
+ """
+ Like or unlike an article.
+ """
+ collection = self.db[str(feed_id)]
+ collection.update({"article_id": article_id}, {"$set": {"article_like": like}})
+
+ def nb_favorites(self, feed_id=None):
+ """
+ Return the number of favorites articles of a feed
+ or of all the database.
+ """
+ if feed_id is not None:
+ return self.get_favorites(feed_id).count()
+ else:
+ return len(self.get_favorites())
+
+ def nb_mail_notifications(self):
+ """
+ Return the number of subscribed feeds.
+ """
+ nb_mail_notifications = 0
+ for feed_id in self.db.collection_names():
+ collection = self.db[feed_id]
+ cursor = collection.find({'type':0, 'mail':True})
+ nb_mail_notifications += cursor.count()
+ return nb_mail_notifications
+
+ def mark_as_read(self, readed, feed_id=None, article_id=None):
+ """
+ Mark one or several articles as read.
+ """
+ if feed_id != None and article_id != None:
+ collection = self.db[str(feed_id)]
+ collection.update({"article_id": article_id, "article_readed":not readed}, {"$set": {"article_readed": readed}})
+ elif feed_id != None and article_id == None:
+ collection = self.db[str(feed_id)]
+ collection.update({"type": 1, "article_readed":not readed}, {"$set": {"article_readed": readed}}, multi=True)
+ else:
+ for feed_id in self.db.collection_names():
+ self.mark_as_read(readed, feed_id, None)
+
+ def update_feed(self, feed_id, changes):
+ """
+ Update a feed.
+ """
+ collection = self.db[str(feed_id)]
+ collection.update({"type": 0, "feed_id":feed_id}, {"$set": changes}, multi=True)
+ if "feed_id" in changes.keys():
+ self.db[str(feed_id)].rename(str(changes["feed_id"]))
+
+ # Functions on database
+ def drop_database(self):
+ """
+ Drop all the database
+ """
+ self.connection.drop_database(self.db_name)
+
+
+if __name__ == "__main__":
+ # Point of entry in execution mode.
+ articles = Articles()
+ # Create a collection for a stream
+ collection_dic = {"collection_id": 42,\
+ "feed_image": "Image", \
+ "feed_title": "Title", \
+ "feed_link": "Link", \
+ "site_title": "Site link", \
+ "mail": True, \
+ }
+ #articles.add_collection(collection_dic)
+
+ # Add an article in the newly created collection
+ article_dic1 = {"article_id": 51, \
+ "article_date": "Today", \
+ "article_link": "Link of the article", \
+ "article_title": "The title", \
+ "article_content": "The content of the article", \
+ "article_readed": True, \
+ "article_like": True \
+ }
+ article_dic2 = {"article_id": 52, \
+ "article_date": "Yesterday", \
+ "article_link": "Link", \
+ "article_title": "Hello", \
+ "article_content": "The content of the article", \
+ "article_readed": True, \
+ "article_like": True \
+ }
+
+ #articles.add_articles([article_dic1, article_dic2], 42)
+
+ print("All articles:")
+ #print articles.get_all_articles()
+
+
+ # Drop the database
+ #articles.drop_database()
bgstack15