# returns all recipes with the search string in the name or description

results = []

# pickling doesn't seem to be any faster than just loading from redis

#try:

# recipes = pickle.load(open( "recipes.p", "rb" ))

#except:

# recipes = [redis.get(k) for k in redis.keys('*recipe*')]

# pickle.dump( recipes, open( "recipes.p", "wb" ) )

recipes = [redis.get(k) for k in redis.keys('*recipe*')]

for r in recipes:

r = loads(r)

score = compute_match(search, r)

if score > 0:

results.append((r, score))

sorted_r = sorted(results, key=operator.itemgetter(1))

sorted_r.reverse()

i = min(len(sorted_r), 10)

#print sorted_r[:i]

#raw_input()

try:

s = wordnet.similarity(w1,w2)

return s

except:

#print search

pall = re.compile('('+search.lower()+')') #pall is the regex for the whole search string

ps = [re.compile('('+s.lower()+' )') for s in search.split(' ')] #ps is the list of regexs for each word.

rname = recipe['name'].lower()

#print [p.findall('chili') for p in ps]

#recipe_2grams = [' '.join(r) for r in zip(*[rname.split(' ')[i:] for i in range(2)])]

#recipe_2grams += ['-'.join(r) for r in zip(*[rname.split(' ')[i:] for i in range(2)])]

#search_2grams = [' '.join(s) for s in zip(*[search.lower().split(' ')[i:] for i in range(2)])]

#search_2grams += ['-'.join(s) for s in zip(*[search.lower().split(' ')[i:] for i in range(2)])]

#

#filters = [

# 'vegan', # no meat, cheese, egg, or animal-derived ingredients

# 'vegetarian', # no meat

# 'low fat', 'low-fat', # not a lot of saturated fat

# 'slow cooker', 'slow-cooker', # uses a slow-cooker or crockpot

# 'gluten', 'gluten-free', 'gluten free', # no gluten

# 'paleo', # high protein, low carb

# 'raw', # no cooking

# 'easy', 'quick', # low cook time, few ingredients

# 'low-sodium' # no high salt ingredients

# ]

#_filters = set(filters) & \

# (set([s.lower() for s in search.split(' ')]) | set(search_2grams)) & \

# (set(rname.split(' ')) | set(recipe_2grams))

#if not _filters:

# return 0 # if the filter isn't in the name, don't return the recipe(?)

#dashes = [' '.join(f.split('-')) for f in _filters if '-' in f]

#if dashes:

# _filters.update([' '.join(dashes)])

#for f in _filters:

# rname = rname.replace(f, '')

# scoring is this way:

# One or more of the search terms must appear in the recipe title.

# If both do, the score is doubled.

# The frequency of the search terms in the description and instructions determines the score.

# The body score is normalized by the length of the recipe body.

try:

rbody = (recipe['description'] + ' ' + ' '.join(recipe['recipeInstructions'])).lower()

except:

rbody = recipe['description']

name_score = sum([len(p.findall(rname)) for p in ps])

#print [p.findall(' chili') for p in ps]

#print name_score

#raw_input()

body_score = 0

if len(pall.findall(rname)) > 0 and len(search.split(' ')) > 1:

name_score = name_score * 2 #double name score is the whole search string is in the name

if name_score > 0: # only return recipe if search in name(?)

body_score = sum([sum([try_similarity(wsearch, wbody) for wsearch in search.split(' ')]) for wbody in rbody])

body_score = body_score / float(len(rbody))

return name_score + 100*body_score

else:

new_recipes = list(recipes)

# adjust order of recipes based on similarity of ingredient lists with recipes above

for i in xrange(len(recipes)):

ind = len(recipes) - i - 1

#