def search(request):
"""
Find songs by searching on their fields.
Query Mongo to get all matching documents. Return the ID, title
and artist values of each match.
"""
query_dict = {}
for k, v in request.GET.iteritems():
if v:
query_dict[k] = re.compile('.*%s.*' % v, re.IGNORECASE)
res = col.find(query_dict, {
'song_id': 1, 'title': 1, 'artist': 1
})
return render(request, 'results.html', {
'results': [i for i in res]
})
def results(request, code):
k = int(request.GET.get('k', 1))
query = queries.find_one({
'md5': code
})
if not query:
raise Http404
segments = query['segments']
all_song_ids = []
q_timbre_values = []
q_loudness_values = []
q_pitch_values = []
for segment in segments:
loudness = segment['loudness_max']
timbre = scipy.array(segment['timbre']).var()
pitch = scipy.array(segment['pitches']).mean()
q_timbre_values.append(timbre)
q_loudness_values.append(loudness)
q_pitch_values.append(pitch)
bbox = (
timbre, loudness, pitch,
timbre, loudness, pitch
)
nearest = idx.nearest(bbox, 1, objects=True)
node_ids = [x.object for x in nearest]
all_song_ids.extend(node_ids)
# Drop duplicates.
ids = set(all_song_ids)
# Fetch the nearest documents from Mongo.
res = col.find({
'song_id': {'$in': list(ids)}
}, {
'song_id': 1, 'title': 1, 'artist': 1, 'features': 1
})
# Compute distances.
results = []
for doc in res:
features = doc['features']
timbre_values = map(lambda x: x['timbre'], features)
loudness_values = map(lambda x: x['loudness'], features)
pitch_values = map(lambda x: x['pitch'], features)
# Calculate distances
timbre_dis = distance(timbre_values, q_timbre_values)
loudness_dis = distance(loudness_values, q_loudness_values)
pitch_dis = distance(pitch_values, q_pitch_values)
doc['timbre_dis'] = timbre_dis
doc['loudness_dis'] = loudness_dis
doc['pitch_dis'] = pitch_dis
doc['all'] = (timbre_dis * pitch_dis) + 100*loudness_dis
results.append(doc)
sorted_res = sorted(results, key=operator.itemgetter('all'))
return render(request, 'results_content.html', {
'results': sorted_res[:k],
'query': query
})
