def _require_session_owned(session_id):
"""
Require that the session is owned by the logged in user
"""
session_r = _session.find_one({'_id': bson.ObjectId(session_id)})
if not session_r:
raise Exception('Session not found')
search_r = _search.find_one({'_id': bson.ObjectId(session_r['search_id'])})
if not search_r:
raise Exception('Search not found')
if search_r['username'] != session.get('username', ''):
raise Exception('You do not have permission to access this snapshot')
return (search_r, session_r)
def _require_session_owned(session_id):
"""
Require that the session is owned by the logged in user
"""
session_r = _session.find_one({'_id': bson.ObjectId(session_id)})
if not session_r:
raise Exception('Session not found')
search_r = _search.find_one({'_id': bson.ObjectId(session_r['search_id'])})
if not search_r:
raise Exception('Search not found')
if search_r['username'] != session.get('username', ''):
raise Exception('You do not have permission to access this snapshot')
return (search_r, session_r)
def filter(session_id):
"""
Get histogram and tweets
@filter: comma-delimited list of elements to filter by
if element starts with '#', then it is a hashtag
else, it is a stem
"""
try:
_require_session_access(session_id)
session_r = _session.find_one(
{'_id': bson.ObjectId(session_id)})
if not session_r:
raise Exception('Session not found')
search_r = _search.find_one(
{'_id': bson.ObjectId(session_r['search_id'])})
if not search_r:
raise Exception('Search not found')
# Find tweets
params = {'session_id': session_id}
filter = request.args.getlist('filter[]')
filter_stems = []
filter_hashtags = []
filter_urls = []
for element in filter:
if element.startswith('#'):
filter_hashtags.append(element)
elif element.startswith('http'):
filter_urls.append(element)
else:
filter_stems.append(element)
if filter_urls:
params['urls'] = {'$all': filter_urls}
if filter_stems:
params['stems'] = {'$all': filter_stems}
if filter_hashtags:
params['hashtags'] = {'$all': filter_hashtags}
cursor = _tweets.find(params, {
'embed': 1,
'id_str': 1,
'created_at': 1,
'user.name': 1,
'user.screen_name': 1,
'retweeted_status.id_str': 1,
'stems': 1,
'hashtags': 1,
'urls': 1
}, sort=[('dt', pymongo.DESCENDING)])
# Process tweets
stem_counter = Counter()
hashtag_counter = Counter()
url_counter = Counter()
tweets = []
retweets = 0
id_set = set()
for tweet in cursor:
stem_counter.update(tweet['stems'])
hashtag_counter.update(tweet['hashtags'])
url_counter.update(tweet['urls'])
if tweet['id_str'] in id_set:
retweets += 1
continue
id_set.add(tweet['id_str'])
if 'retweeted_status' in tweet:
retweeted_id = tweet['retweeted_status']['id_str']
if retweeted_id in id_set:
retweets += 1
continue
id_set.add(retweeted_id)
tweets.append({
'text': tweet['embed'],
'user_name': tweet['user']['name'],
'user_screen_name': tweet['user']['screen_name'],
'id_str': tweet['id_str'],
'created_at': tweet['created_at']
})
stem_counts = [x for x in stem_counter.most_common() \
if x[0] not in filter_stems]
hashtag_counts = [x for x in hashtag_counter.most_common() \
if x[0] not in filter_hashtags]
url_counts = [x for x in url_counter.most_common() \
if x[0] not in filter_urls]
return _jsonify(
search=search_r,
session=session_r,
stem_counts=stem_counts,
hashtag_counts=hashtag_counts,
url_counts=url_counts,
tweets=tweets,
retweets=retweets
)
except Exception, e:
traceback.print_exc()
return _jsonify(error=str(e))
def analyze():
"""
Get tweets from twitter and analyze them
@language = language code, e.g. 'en'
@query = query string
OR
@list_id = list id
"""
try:
language = request.args.get('language') or 'en'
query = request.args.get('query')
list_id = request.args.get('list_id')
if query:
query_lower = query.lower()
elif list_id:
list_map = _get_list_map()
else:
raise Exception('No query or list specified')
# Get api object
api = tweepy.API(get_oauth())
# Get/create search record
param = {
'username': session['username'],
'language': language
}
if query:
param['query_lower'] = query_lower
else:
param['list_id'] = list_id
search_r = _search.find_one(param)
if not search_r:
search_r = param
if query:
search_r['query'] = query
else:
search_r['list_name'] = list_map[list_id]
search_r['_id'] = _search.save(search_r, manipulate=True)
search_id = str(search_r['_id'])
# Create new search session
session_r = {
'search_id': search_id,
'dt': datetime.datetime.now().isoformat(),
'stem_counts': [], # [[stem, post count]]
'stem_map': {}, # {stem: [term, count]}
}
session_r['_id'] = _session.save(session_r, manipulate=True)
session_id = str(session_r['_id'])
# Process tweets
stopwords = extract.get_stopwords(language).copy()
stoptags = set()
stemmer = extract.get_stemmer(language)
stem_map = defaultdict(Counter)
tweet_list = []
if query:
stoptags.update([x.lower().lstrip('#') for x in query_lower.split()])
stopwords.update(stoptags)
cursor = tweepy.Cursor(api.search, q=query, lang=language, \
count=100, result_type='recent', include_entities=True)
else:
cursor = tweepy.Cursor(api.list_timeline, list_id=list_id, \
count=100, include_entities=True)
for tweet in cursor.items(limit=settings.TWITTER_SEARCH_LIMIT):
tweet_dict = twutil.tweepy_model_to_dict(tweet)
tweet_dict['session_id'] = session_id
tweet_dict['embed'] = twutil.format_text(tweet_dict)
tweet_dict['tokens'] = extract.tokenize(tweet_dict['text'])
# Filter hashtags from query
# tweet_dict['hashtags'] = list(set(['#'+x['text'].lower() \
# for x in tweet_dict['entities']['hashtags']]))
tweet_dict['hashtags'] = list(set([
'#'+x['text'].lower() \
for x in tweet_dict['entities']['hashtags'] \
if x['text'].lower() not in stoptags
]))
tweet_dict['urls'] = list(set([x['expanded_url'] \
for x in tweet_dict['entities']['urls']]))
tweet_list.append(tweet_dict)
# ------------------------------------------------------------
# Process trigrams
trigram_counter = Counter()
for tweet in tweet_list:
grams = []
for tokens in tweet['tokens']:
for g in nltk.ngrams(tokens, 3):
if extract.stoplist_iter(g, stopwords):
continue
if g[0].startswith('@') or g[1].startswith('@') or g[2].startswith('@'):
continue
grams.append(g)
stems = extract.stems_from_grams(grams, stemmer)
for s, g in zip(stems, grams):
stem_map[s].update([g])
tweet['stems_3'] = list(set(stems))
trigram_counter.update(tweet['stems_3'])
# Ignore trigrams that only appear once
for g, n in trigram_counter.items():
if n < 2:
del trigram_counter[g]
del stem_map[g]
# ------------------------------------------------------------
# Process bigrams
bigram_counter = Counter()
for tweet in tweet_list:
grams = []
stems = []
for tokens in tweet['tokens']:
gram_list = nltk.ngrams(tokens, 2)
stem_list = extract.stems_from_grams(gram_list, stemmer)
last_i = len(gram_list) - 1
for i, g in enumerate(gram_list):
if extract.stoplist_iter(g, stopwords):
continue
if g[0].startswith('@') or g[1].startswith('@'):
continue
# Filter by trigrams
if i > 0 and \
(stem_list[i-1][0], stem_list[i][0], stem_list[i][1]) in trigram_counter:
continue
if i < last_i and \
(stem_list[i][0], stem_list[i][1], stem_list[i+1][1]) in trigram_counter:
continue
grams.append(g)
stems.append(stem_list[i])
for s, g in zip(stems, grams):
stem_map[s].update([g])
tweet['stems_2'] = list(set(stems))
bigram_counter.update(tweet['stems_2'])
# Ignore bigrams that only appear once
for g, n in bigram_counter.items():
if n < 2:
del bigram_counter[g]
del stem_map[g]
# ------------------------------------------------------------
# Process unigrams
for tweet in tweet_list:
grams = []
stems = []
for tokens in tweet['tokens']:
gram_list = nltk.ngrams(tokens, 1)
stem_list = extract.stems_from_grams(gram_list, stemmer)
last_i = len(gram_list) - 1
for i, g in enumerate(gram_list):
if extract.stoplist_iter(g, stopwords):
continue
# Filter bigram terms
if i > 0 and \
(stem_list[i-1][0], stem_list[i][0]) in bigram_counter:
continue
if i < last_i and \
(stem_list[i][0], stem_list[i+1][0]) in bigram_counter:
continue
# Filter trigram terms
if i > 1 and \
(stem_list[i-2][0], stem_list[i-1][0], stem_list[i][0]) in trigram_counter:
continue
if i > 0 and i < last_i and \
(stem_list[i-1][0], stem_list[i][0], stem_list[i+1][0]) in trigram_counter:
continue
if i < (last_i - 1) and \
(stem_list[i][0], stem_list[i+1][0], stem_list[i+2][0]) in trigram_counter:
continue
grams.append(g)
stems.append(stem_list[i])
for s, g in zip(stems, grams):
stem_map[s].update([g])
# Process stems
tweet['stems'] = [' '.join(x) for x in set(stems)]
tweet['stems'].extend([' '.join(x) for x in tweet['stems_2'] if x in bigram_counter])
tweet['stems'].extend([' '.join(x) for x in tweet['stems_3'] if x in trigram_counter])
del tweet['stems_2']
del tweet['stems_3']
# Update session
for stem, c in stem_map.iteritems():
session_r['stem_map'][' '.join(stem)] = \
[' '.join(k) for k, v in c.most_common()]
# Save tweets
if tweet_list:
_tweets.insert(tweet_list)
session_r['tweet_count'] = len(tweet_list)
_session.save(session_r)
return _jsonify(session=session_r)
except tweepy.TweepError, e:
traceback.print_exc()
return _jsonify(error=e.message[0]['message'])
def filter(session_id):
"""
Get histogram and tweets
@filter: comma-delimited list of elements to filter by
if element starts with '#', then it is a hashtag
else, it is a stem
"""
try:
_require_session_access(session_id)
session_r = _session.find_one(
{'_id': bson.ObjectId(session_id)})
if not session_r:
raise Exception('Session not found')
search_r = _search.find_one(
{'_id': bson.ObjectId(session_r['search_id'])})
if not search_r:
raise Exception('Search not found')
# Find tweets
params = {'session_id': session_id}
filter = request.args.getlist('filter[]')
filter_stems = []
filter_hashtags = []
filter_urls = []
for element in filter:
if element.startswith('#'):
filter_hashtags.append(element)
elif element.startswith('http'):
filter_urls.append(element)
else:
filter_stems.append(element)
if filter_urls:
params['urls'] = {'$all': filter_urls}
if filter_stems:
params['stems'] = {'$all': filter_stems}
if filter_hashtags:
params['hashtags'] = {'$all': filter_hashtags}
cursor = _tweets.find(params, {
'embed': 1,
'id_str': 1,
'created_at': 1,
'user.name': 1,
'user.screen_name': 1,
'retweeted_status.id_str': 1,
'stems': 1,
'hashtags': 1,
'urls': 1
}, sort=[('dt', pymongo.DESCENDING)])
# Process tweets
stem_counter = Counter()
hashtag_counter = Counter()
url_counter = Counter()
tweets = []
retweets = 0
id_set = set()
for tweet in cursor:
stem_counter.update(tweet['stems'])
hashtag_counter.update(tweet['hashtags'])
url_counter.update(tweet['urls'])
if tweet['id_str'] in id_set:
retweets += 1
continue
id_set.add(tweet['id_str'])
if 'retweeted_status' in tweet:
retweeted_id = tweet['retweeted_status']['id_str']
if retweeted_id in id_set:
retweets += 1
continue
id_set.add(retweeted_id)
tweets.append({
'text': tweet['embed'],
'user_name': tweet['user']['name'],
'user_screen_name': tweet['user']['screen_name'],
'id_str': tweet['id_str'],
'created_at': tweet['created_at']
})
stem_counts = [x for x in stem_counter.most_common() \
if x[0] not in filter_stems]
hashtag_counts = [x for x in hashtag_counter.most_common() \
if x[0] not in filter_hashtags]
url_counts = [x for x in url_counter.most_common() \
if x[0] not in filter_urls]
return _jsonify(
search=search_r,
session=session_r,
stem_counts=stem_counts,
hashtag_counts=hashtag_counts,
url_counts=url_counts,
tweets=tweets,
retweets=retweets
)
except Exception, e:
traceback.print_exc()
return _jsonify(error=str(e))
def analyze():
"""
Get tweets from twitter and analyze them
@language = language code, e.g. 'en'
@query = query string
OR
@list_id = list id
"""
try:
language = request.args.get('language') or 'en'
query = request.args.get('query')
list_id = request.args.get('list_id')
if query:
query_lower = query.lower()
elif list_id:
list_map = _get_list_map()
else:
raise Exception('No query or list specified')
# Get api object
api = tweepy.API(get_oauth())
# Get/create search record
param = {
'username': session['username'],
'language': language
}
if query:
param['query_lower'] = query_lower
else:
param['list_id'] = list_id
search_r = _search.find_one(param)
if not search_r:
search_r = param
if query:
search_r['query'] = query
else:
search_r['list_name'] = list_map[list_id]
search_r['_id'] = _search.save(search_r, manipulate=True)
search_id = str(search_r['_id'])
# Create search session
session_r = {
'search_id': search_id,
'dt': datetime.datetime.now().isoformat(),
'stem_counts': [], # [[stem, post count]]
'stem_map': {}, # {stem: [term, count]}
}
session_r['_id'] = _session.save(session_r, manipulate=True)
session_id = str(session_r['_id'])
# Process tweets
stopwords = extract.get_stopwords(language).copy()
stoptags = set()
stemmer = extract.get_stemmer(language)
stem_map = defaultdict(Counter)
tweet_list = []
if query:
stoptags.update([x.lower().lstrip('#') for x in query_lower.split()])
stopwords.update(stoptags)
cursor = tweepy.Cursor(api.search, q=query, lang=language, \
count=100, result_type='recent', include_entities=True)
else:
cursor = tweepy.Cursor(api.list_timeline, list_id=list_id, \
count=100, include_entities=True)
for tweet in cursor.items(limit=settings.TWITTER_SEARCH_LIMIT):
tweet_dict = twutil.tweepy_model_to_dict(tweet)
tweet_dict['session_id'] = session_id
tweet_dict['embed'] = twutil.format_text(tweet_dict)
tweet_dict['tokens'] = extract.tokenize(tweet_dict['text'])
# Filter hashtags from query
# tweet_dict['hashtags'] = list(set(['#'+x['text'].lower() \
# for x in tweet_dict['entities']['hashtags']]))
tweet_dict['hashtags'] = list(set([
'#'+x['text'].lower() \
for x in tweet_dict['entities']['hashtags'] \
if x['text'].lower() not in stoptags
]))
tweet_dict['urls'] = list(set([x['expanded_url'] \
for x in tweet_dict['entities']['urls']]))
tweet_list.append(tweet_dict)
# ------------------------------------------------------------
# Process trigrams
trigram_counter = Counter()
for tweet in tweet_list:
grams = []
for tokens in tweet['tokens']:
for g in nltk.ngrams(tokens, 3):
if extract.stoplist_iter(g, stopwords):
continue
if g[0].startswith('@') or g[1].startswith('@') or g[2].startswith('@'):
continue
grams.append(g)
stems = extract.stems_from_grams(grams, stemmer)
for s, g in zip(stems, grams):
stem_map[s].update([g])
tweet['stems_3'] = list(set(stems))
trigram_counter.update(tweet['stems_3'])
# Ignore trigrams that only appear once
for g, n in trigram_counter.items():
if n < 2:
del trigram_counter[g]
del stem_map[g]
# ------------------------------------------------------------
# Process bigrams
bigram_counter = Counter()
for tweet in tweet_list:
grams = []
stems = []
for tokens in tweet['tokens']:
gram_list = nltk.ngrams(tokens, 2)
stem_list = extract.stems_from_grams(gram_list, stemmer)
last_i = len(gram_list) - 1
for i, g in enumerate(gram_list):
if extract.stoplist_iter(g, stopwords):
continue
if g[0].startswith('@') or g[1].startswith('@'):
continue
# Filter by trigrams
if i > 0 and \
(stem_list[i-1][0], stem_list[i][0], stem_list[i][1]) in trigram_counter:
continue
if i < last_i and \
(stem_list[i][0], stem_list[i][1], stem_list[i+1][1]) in trigram_counter:
continue
grams.append(g)
stems.append(stem_list[i])
for s, g in zip(stems, grams):
stem_map[s].update([g])
tweet['stems_2'] = list(set(stems))
bigram_counter.update(tweet['stems_2'])
# Ignore bigrams that only appear once
for g, n in bigram_counter.items():
if n < 2:
del bigram_counter[g]
del stem_map[g]
# ------------------------------------------------------------
# Process unigrams
for tweet in tweet_list:
grams = []
stems = []
for tokens in tweet['tokens']:
gram_list = nltk.ngrams(tokens, 1)
stem_list = extract.stems_from_grams(gram_list, stemmer)
last_i = len(gram_list) - 1
for i, g in enumerate(gram_list):
if extract.stoplist_iter(g, stopwords):
continue
# Filter bigram terms
if i > 0 and \
(stem_list[i-1][0], stem_list[i][0]) in bigram_counter:
continue
if i < last_i and \
(stem_list[i][0], stem_list[i+1][0]) in bigram_counter:
continue
# Filter trigram terms
if i > 1 and \
(stem_list[i-2][0], stem_list[i-1][0], stem_list[i][0]) in trigram_counter:
continue
if i > 0 and i < last_i and \
(stem_list[i-1][0], stem_list[i][0], stem_list[i+1][0]) in trigram_counter:
continue
if i < (last_i - 1) and \
(stem_list[i][0], stem_list[i+1][0], stem_list[i+2][0]) in trigram_counter:
continue
grams.append(g)
stems.append(stem_list[i])
for s, g in zip(stems, grams):
stem_map[s].update([g])
# Process stems
tweet['stems'] = [' '.join(x) for x in set(stems)]
tweet['stems'].extend([' '.join(x) for x in tweet['stems_2'] if x in bigram_counter])
tweet['stems'].extend([' '.join(x) for x in tweet['stems_3'] if x in trigram_counter])
del tweet['stems_2']
del tweet['stems_3']
# Update session
for stem, c in stem_map.iteritems():
session_r['stem_map'][' '.join(stem)] = \
[' '.join(k) for k, v in c.most_common()]
# Save tweets
if tweet_list:
_tweets.insert(tweet_list)
session_r['tweet_count'] = len(tweet_list)
_session.save(session_r)
return _jsonify(session=session_r)
except tweepy.TweepError, e:
traceback.print_exc()
return _jsonify(error=e.message[0]['message'])
