it just pulls out from the stream of tweets those fragments in which 'bobuk' is mentioned (bobuk is here just for the sake of example as the most active user of the Twitter API in Runet).
All this without having to think about all the details for the right API request.
TWITTER_SAMPLE is a tweet stream that contains approximately 1% of the total. If you need to query the shared stream, specify TWITTER as the data source.
Filtered tweets can be saved to a database on a local disk (its parameters are set in settings.py). For performance reasons, writing to the database is performed only when 1000 records are reached, so if the filtered records are smaller, they will not be saved.