TODOS

REWRITE?
Matching BA/RB was a PITA and RB has more/better data, so let's write a new scraper to continuously crawl RB for all their data!
- beers
- users
- reviews
- breweries
- locations



BUGS:
When a user has entered reviews before entering their username, previously saved beers should not be clobbered by the new list, they should be merged
User ratings should be restricted to 1-10, some users enter 0 which is ignored
Still returning doubles sometimes

'Wells & Young's Ltd' and 'Speakeasy Ales & Lagers' don't seem to pass beers to typeahead for some reason; suspect ampersand is screwing things up...
When beer is 'Death & Taxes Ale' on recommend: Cannot call method slice of undefined bi.js:236, then recommender error...
possible solution:
http://www.digitalbart.com/blog/jquery-and-urlencode/
or http://stackoverflow.com/questions/6544564/url-encode-a-string-in-jquery-for-an-ajax-request
in js: encodeURI(urlstring)
http://stackoverflow.com/questions/3563126/url-encoding-decoding-with-python
in python: urllib2.unquote(urlstring).decode("utf8")


Updates:
recommender validation - compare algorithm with netflix challenge
integrate sparse_mult to module for pull request to scipy
setup typeahead to tokenize against all words instead of whole names, also check for case sensitivity

Add option to dislike an entire brewery


Extras:
    update geocode to mapquest api for speed increase
    clean bad matches/dupes between BA and RB
    try speeding up recommender by filtering users to 1000 closest pearson correlations to user

    switch user data stores to SQL
    switch map markers and my ratings button to click to toggle popup

    for top rated beers, add feature for 'other people who liked this beer said...'
    include relational maps to visualize recommended beers
    include sentiment/frequency analysis of major keywords in text reviews
    add search beer by keyword (e.g. sour, hoppy, bitter, floral, grapefruit etc)