dd-re/recommendation_engine.py
2017-11-01 16:38:58 +00:00

124 lines
4.1 KiB
Python

'''
Reccomendation Engine Example
'''
import json
import os
import random
import ox
class Pandora:
def __init__(self, url, username, password):
self.api = ox.API(url)
self.api.signin(username=username, password=password)
def find_annotations(self, query, keys):
# print('FIND ANNOTATIONS', query, keys)
return self.api.findAnnotations({
'keys': keys,
'query': query,
'range': [0, 1000000]
})['data']['items']
def find_entities(self, query, keys):
# print('FIND ENTITIES', query, keys)
return self.api.findEntities({
'keys': keys,
'query': query,
'range': [0, 1000000]
})['data']['items']
def get(self, id, keys):
# print('GET', id, keys)
return self.api.get({
'id': id,
'keys': keys
})['data']
class Engine:
def __init__(self, path):
self.path = path
self.pandora = Pandora(
url='http://pandora.dmp/api/',
username='dd.re',
password='dd.re'
)
filename = os.path.join(self.path, 'playlists.json')
if os.path.exists(filename):
with open(filename) as f:
self.playlists = json.loads(f.read())
else:
self.playlists = []
def _shift_clips(self, clips):
index = random.randrange(len(clips))
return clips[index:] + clips[:index - 1]
def get_videos(self, user):
products = []
for event in user['events']:
if 'product' in event['data']:
products.append(event['data']['product'])
def update(self):
# Get all storylines with tags
storylines = [{
'name': entity['name'],
'tags': entity['tags']
} for entity in self.pandora.find_entities({
'conditions': [
{'key': 'type', 'operator': '==', 'value': 'storylines'},
],
'operator': '&'
}, ['id', 'name', 'tags']) if entity.get('tags', [])]
# Get list of storyline names
names = list(set([storyline['name'] for storyline in storylines]))
# Get all clips annotated with storyline references
clips = [clip for clip in self.pandora.find_annotations({
'conditions': [
{'key': 'layer', 'operator': '==', 'value': 'storylines'}
],
'operator': '&'
}, ['id', 'in', 'out', 'value']) if clip['value'] in names]
# Get list of ids for videos with clips
ids = list(set([clip['id'].split('/')[0] for clip in clips]))
# Get (and cache) order (and code + name) for each video
filename = os.path.join(self.path, 'videos.json')
if os.path.exists(filename):
with open(filename) as f:
videos_ = json.loads(f.read())
ids_ = [video['id'] for video in videos_]
else:
videos_, ids_ = [], []
videos = sorted(videos_ + [
self.pandora.get(id, ['code', 'id', 'order', 'title'])
for id in ids if not id in ids_
], key=lambda video: video['order'])
with open(filename, 'w') as f:
f.write(json.dumps(videos, indent=4, sort_keys=True))
order = {video['id']: video['order'] for video in videos}
# Sort clips
clips = sorted(
clips,
key=lambda clip: order[clip['id'].split('/')[0]] * 1000000 + clip['in']
)
# Get playlists
self.playlists = [playlist for playlist in [{
'name': storyline['name'],
'tags': storyline['tags'],
'clips': [
'{}_{:.3f}-{:.3f}'.format(
clip['id'].split('/')[0], clip['in'], clip['out']
) for clip in clips if clip['value'] == storyline['name']
]
} for storyline in storylines] if playlist['clips']]
with open(os.path.join(self.path, 'playlists.json'), 'w') as f:
f.write(json.dumps(self.playlists, indent=4, sort_keys=True))
if __name__ == '__main__':
engine = Engine('json')
engine.update()