class Words:
def __init__(self, environment, hashtag, tags_list):
self.env = environment
self.hashtag = hashtag
self.boring_words = {}
self.banned_tags = {}
self.data = {}
self.t_new = 0
self.t_foreign = 0
self.t_skip = 0
self.t_retweet = 0
# self.ns_tweet_count = []
# self.ns_total_score = []
self.ns_score_log = []
# self.ns_index = -1
self.retweets = []
self.date = today()
self.db = DB(environment, self.date)
# self.c = self.db.connect(self.date)
self.db_summary = DBSummary(environment)
self.load_metadata()
self.CONSUMER_KEY = self.env.consumer_key
self.CONSUMER_SECRET = self.env.consumer_secret
self.current_token = -1
self.hash_tags_re = re.compile(r'(?i)(?<!\w)#[\w\u064b-\u0657]+',
re.UNICODE)
self.twitters = list()
for token in self.db_summary.get_all_tokens():
api = twitter.Api(consumer_key=self.CONSUMER_KEY,
consumer_secret=self.CONSUMER_SECRET,
access_token_key=token.key,
access_token_secret=token.secret,
sleep_on_rate_limit=True)
self.twitters.append(api)
self.today_skey = 0
self.score_names = False
if hashtag == 'trends':
if os.path.isfile('metadata/name_score.csv'):
self.score_names = True
logger.info(
"metadata/name_score.csv will be used for name scoring.")
else:
logger.info(
"Warning: metadata/name_score.csv does not exist so name scoring is disabled."
)
self.batch_id = self.db.get_next_batch_id()
self.baseline_tweet_id = self.db.get_baseline_tweet_id()
self.today_skey = self.db.get_date_skey(self.date)
self.loop_pos = -1
self.all_trends = None
if hashtag == 'trends':
self.all_trends = self.db.get_trends()
self.loop_pos = 0
if tags_list is None:
tags_list = []
for (tag, result) in self.all_trends.items():
if result in ('AUTO_ADD', 'MAN_ADD'):
tags_list.append({'tag': tag})
orig_tags_list = tags_list
tags_list = []
for tagdata in orig_tags_list:
tags_list.append(
self.db.get_tag_ranges(tagdata['tag'],
self.baseline_tweet_id))
print('Tags_list:', tags_list)
self.pull_trends(tags_list)
self.write_data()
elif hashtag == 'home_timeline':
status_count = self.pull_data(hashtag)
logger.info('{} statuses pulled.'.format(status_count))
self.write_data()
elif hashtag == 'lists':
lists = self.twitters[
self.db_summary.polling_token_index].GetLists(
screen_name=self.env.polling_account)
logger.info('{} lists for account {}.'.format(
len(lists), self.env.polling_account))
for l in lists:
status_count = self.pull_data(l.slug)
logger.info('{} statuses pulled for list {}.'.format(
status_count, l.slug))
self.write_data()
self.db.disconnect()
@property
def api(self):
self.current_token += 1
if self.current_token >= len(self.twitters):
self.current_token = 0
return self.twitters[self.current_token]
def pull_trend(self, trend, trend_count, trend_position):
self.ns_score_log = []
status_count = 0
request_count = 0
index = 0
while index < len(trend.ranges):
# for id_range in trend.ranges:
id_range = trend.ranges[index]
if not id_range.processed:
max_id = id_range.max_id
since_id = id_range.min_id
statuses = None
logger.info('Range: {:>9} {:35} {:20} {:20}'.format(
'{:4d}/{:4d}'.format(trend_position,
trend_count), trend.name, since_id,
'None' if max_id is None else max_id))
while statuses is None or len(statuses) >= 50:
if request_count >= 100:
new_range = Range(min_id=since_id, max_id=max_id)
id_range.min_id = max_id
trend.ranges.insert(index + 1, new_range)
return request_count, status_count, False
statuses = self.api.GetSearch(term=trend.name,
result_type='recent',
count=100,
include_entities=False,
max_id=max_id,
since_id=since_id)
self.get_words(statuses, trend=trend, source=trend.name)
status_count += len(statuses)
score = trend.get_average_score(10)
if len(statuses) > 0:
max_id = statuses[-1].id - 1
if id_range.max_id is None and len(statuses) > 0:
id_range.max_id = statuses[0].id
id_range.processed = True
request_count += 1
logger.info(
'{:40} {:20} {:20} {:3} {:5} {:5.2f} {}'.format(
trend.name, since_id,
'None' if max_id is None else max_id,
request_count, trend.get_status_count(10), score,
trend.state))
if score < 0.0 and status_count > 150:
trend.state = 'AUTO_DEL'
self.save_score_log(self.ns_score_log, trend.name,
'Negative')
id_range.min_id = max_id
return request_count, status_count, True
if score < 0.5 and status_count > 1000:
trend.state = 'AUTO_DEL'
self.save_score_log(self.ns_score_log, trend.name,
'Hot_Ambiguous')
id_range.min_id = max_id
return request_count, status_count, True
if score > 2.0:
if trend.state not in ('AUTO_ADD', 'MAN_ADD'):
trend.state = 'AUTO_ADD'
# after 500 tweets if we still haven't got an indication, give up
if trend.get_status_count(
10) > 500 and trend.state == 'AUTO_DEL':
self.save_score_log(self.ns_score_log, trend.name,
'Ambiguous')
id_range.min_id = max_id
return request_count, status_count, True
# Not needed for raspberry pi
if request_count % 100 == 0:
logger.info(
f'Sleeping 20 seconds at {request_count} requests.'
)
sleep(20)
index += 1
return request_count, status_count, True
def pull_trends(self, trends):
total_status_count = 0
last_write = 0
total_request_count = 0
trend_count = len(trends)
for i, trend in enumerate(trends):
completed = False
while not completed:
request_count, status_count, completed = self.pull_trend(
trend, trend_count, i + 1)
total_request_count += request_count
total_status_count += status_count
self.db.tag_history.append(trend)
if total_request_count >= last_write + 20:
self.write_data()
self.batch_id += 1
last_write = total_request_count
return total_status_count
def load_metadata(self):
f = open('metadata/boring.txt', 'r')
for line in f:
self.boring_words[line.rstrip()] = 1
f.close()
f = open('metadata/banned_tags.txt', 'r')
for line in f:
self.banned_tags[line.rstrip()] = 1
f.close()
def cleanup_exit(self):
logger.info("Deleting batch %i" % self.batch_id)
self.db.delete_batch(self.batch_id)
self.db.disconnect()
exit(1)
def twitter_search(self, q, sinceid, maxid):
self.current_token += 1
if self.current_token >= len(self.twitters):
self.current_token = 0
if maxid is None:
result = self.twitters[self.current_token].GetSearch(
term=q,
result_type='recent',
count='100',
include_entities='false',
since_id=sinceid)
else:
result = self.twitters[self.current_token].GetSearch(
term=q,
result_type='recent',
count='100',
include_entities='false',
since_id=sinceid,
max_id=maxid)
return result
# def oauthReq(self, url, key, secret, http_method="GET", post_body='',
# http_headers=None):
# consumer = oauth.Consumer(key=self.CONSUMER_KEY, secret=self.CONSUMER_SECRET)
# if self.hashtag == 'home_timeline':
# token = self.db.getDefaultToken()
# else:
# token = self.db.getNextToken()
# client = oauth.Client(consumer, token)
# resp, content = client.request(
# url,
# method=http_method,
# body=post_body,
# headers=http_headers # , force_auth_header=True
# )
# # print "*** %s ***" % content
# # exit()
# return content
def write_data(self):
logger.info("Writing data.")
for tweetdate, stats in self.data.items():
logger.info("Saving data for %s." % tweetdate)
for tag, words in self.data[tweetdate]['tag_words'].items():
for i, v in words.items():
self.db.write_hashtag_word(tweetdate, tag, i, v)
for tweeter, words in self.data[tweetdate]['tweeter_words'].items(
):
for i, v in words.items():
self.db.write_tweeter_word(tweetdate, tweeter, i, v)
for tweeter, words in self.data[tweetdate][
'tweeter_mentions'].items():
for i, v in words.items():
self.db.write_tweeter_mention(tweetdate, tweeter, i, v)
for tag, count in self.data[tweetdate]['tags'].items():
self.db.write_tag(tweetdate, tag, count)
for tag, tweeters in self.data[tweetdate]['tag_tweeters'].items():
for i, v in tweeters.items():
self.db.write_tag_tweeter(tweetdate, tag, i, v)
for tag, tags in self.data[tweetdate]['tag_tags'].items():
for i, v in tags.items():
self.db.write_tag_tag(tweetdate, tag, i, v)
self.db.write_tweets()
self.db.write_tag_history()
self.db.commit()
logger.info("Data saved.")
self.data = {}
@staticmethod
def save_score_log(score_log, trend, reject_reason):
filename = "log/reject_%s_%s_%s.log" % (trend, reject_reason,
file_timestamp())
secommon.save_list(score_log, filename)
def calculate_name_score(self, status, trend):
tweeter = status.user.screen_name
tweeter_name = status.user.name
score_candidate = (tweeter_name if status.retweeted_status is None else
status.retweeted_status.user.name)
score_candidate_sn = (tweeter if status.retweeted_status is None else
status.retweeted_status.user.screen_name)
trend.name_scores[-1].status_count += 1
# self.ns_tweet_count[self.ns_index] += 1
name_score = self.db.get_name_score(score_candidate,
score_candidate_sn,
status.user.location,
status.user.time_zone)
# self.ns_total_score[self.ns_index] += name_score
trend.name_scores[-1].total_score += name_score
score3 = '{:.2f}'.format(trend.get_average_score(3))
score6 = '{:.2f}'.format(trend.get_average_score(6))
self.ns_score_log.append([
score_candidate, score_candidate_sn,
trend.name_scores[-1].status_count, name_score,
trend.name_scores[-1].total_score, score3, score6, tweeter_name,
tweeter, status.id
])
def process_status_words(self, status_id, status_date, status_text,
tweeter):
if status_date not in self.data:
self.data[status_date] = {}
self.data[status_date]['tweeter_mentions'] = {}
self.data[status_date]['tag_words'] = {}
self.data[status_date]['tweeter_words'] = {}
self.data[status_date]['tags'] = {}
self.data[status_date]['tag_tweeters'] = {}
self.data[status_date]['tag_tags'] = {}
# get all relevant hashtags
relevant_hashtags = re.findall(r'(?<![A-Za-z0-9_])#([A-Za-z0-9_]+)',
status_text.lower())
tweet_hashtags = set(self.hash_tags_re.findall(status_text))
tweet_tags = [ht[1:] for ht in tweet_hashtags]
for tag in tweet_tags:
if tag in self.data[status_date]['tags']:
self.data[status_date]['tags'][tag] += 1
else:
self.data[status_date]['tags'][tag] = 1
if tag not in self.data[status_date]['tag_tweeters']:
self.data[status_date]['tag_tweeters'][tag] = {}
if tweeter in self.data[status_date]['tag_tweeters'][tag]:
self.data[status_date]['tag_tweeters'][tag][tweeter] += 1
else:
self.data[status_date]['tag_tweeters'][tag][tweeter] = 1
if tag not in self.data[status_date]['tag_tags']:
self.data[status_date]['tag_tags'][tag] = {}
for tag2 in tweet_tags:
if tag2 != tag:
if tag2 in self.data[status_date]['tag_tags'][tag]:
self.data[status_date]['tag_tags'][tag][tag2] += 1
else:
self.data[status_date]['tag_tags'][tag][tag2] = 1
# remove links
text = re.sub(r"(?<![A-Za-z0-9_])https?://[^ ,;'()\[\]<>{}]+",
'',
status_text,
flags=re.IGNORECASE)
alist = re.split('[, .;\'\"(){\}\[\]<>:?/=+\\\`~!#^&*\\r\\n\-]+', text)
tweetwords = list()
for item in alist:
nitem = item.strip(' ,.-+()[]:\'\"').lower()
if u"\u2026" in nitem: # ignore words truncated with ellipsis (...)
continue
if nitem == '':
continue
if nitem in self.boring_words:
continue
if nitem[:1] == '@' and len(nitem) > 2:
# Tweeter mentions
if tweeter not in self.data[status_date]['tweeter_mentions']:
self.data[status_date]['tweeter_mentions'][tweeter] = {}
if nitem[1:] in self.data[status_date]['tweeter_mentions'][
tweeter]:
self.data[status_date]['tweeter_mentions'][tweeter][
nitem[1:]] += 1
else:
self.data[status_date]['tweeter_mentions'][tweeter][
nitem[1:]] = 1
continue
tweetwords.append(nitem)
for tag in relevant_hashtags:
if tag not in self.data[status_date]['tag_words']:
self.data[status_date]['tag_words'][tag] = {}
if nitem in self.data[status_date]['tag_words'][tag]:
self.data[status_date]['tag_words'][tag][nitem] += 1
else:
self.data[status_date]['tag_words'][tag][nitem] = 1
# Tweeter words
if tweeter.lower() in self.db.rated_tweeters:
if tweeter not in self.data[status_date]['tweeter_words']:
self.data[status_date]['tweeter_words'][tweeter] = {}
if nitem in self.data[status_date]['tweeter_words'][tweeter]:
self.data[status_date]['tweeter_words'][tweeter][
nitem] += 1
else:
self.data[status_date]['tweeter_words'][tweeter][nitem] = 1
tweet_words_text = u'~' + u'~'.join([
self.db.get_word_skey(x, self.date)[1]
for x in sorted(set(tweetwords))
]) + u'~'
self.db.update_tweet_words(status_id, tweet_words_text)
def get_words(self, statuses, trend=None, source=None):
# max_id = 0
# min_id = MAX_STATUS_ID
for status in statuses:
# max_id = max(status.id, max_id)
# min_id = min(status.id, min_id)
tweeter = status.user.screen_name
tweeter_name = status.user.name
tweeter_created_at = self.env.get_local_date(
status.user.created_at)
tweeter_skey = self.db.get_tweeter_skey(
screen_name=tweeter,
name=tweeter_name,
followers_count=status.user.followers_count,
friends_count=status.user.friends_count,
lang=status.user.lang,
time_zone=status.user.time_zone,
verified=status.user.verified,
statuses_count=status.user.statuses_count,
profile_image_url=status.user.profile_image_url,
created_at=tweeter_created_at,
location=status.user.location)
tweet_text = status.text
retweet_id = status.retweeted_status.id if status.retweeted_status is not None else 0
if retweet_id != 0:
tweet_text = "RT " + status.retweeted_status.user.screen_name + ": " + \
status.retweeted_status.text
if self.score_names:
self.calculate_name_score(status, trend)
if status.user.followers_count > 0 and tweeter.lower(
) in self.db.rated_tweeters:
self.db.write_daily_followers(tweeter_skey, self.today_skey,
status.user.followers_count)
retweet_created_at = ''
retweet_screen_name = ''
retweet_count = status.retweet_count
if status.retweeted_status is not None:
self.retweets.append(status.retweeted_status)
retweet_created_at = self.env.get_local_timestamp(
status.retweeted_status.created_at)
retweet_screen_name = status.retweeted_status.user.screen_name
# check if duplicate and insert if not duplicate
status_date = self.env.get_local_date(status.created_at)
status_created_at = self.env.get_local_timestamp(status.created_at)
date_skey = self.db.get_date_skey(status_date)
if self.db.tweet_is_duplicate(
id_=status.id,
created_at=status_created_at,
screen_name=status.user.screen_name,
text=tweet_text,
tweeter_skey=tweeter_skey,
retweet_count=retweet_count,
in_reply_to_status_id=status.in_reply_to_status_id,
date_skey=date_skey,
retweet_id=retweet_id,
retweet_created_at=retweet_created_at,
retweet_screen_name=retweet_screen_name,
batch_id=self.batch_id,
source=source):
self.t_skip += 1
continue
self.process_status_words(status_id=status.id,
status_date=status_date,
status_text=tweet_text,
tweeter=tweeter)
# @timeout(7)
def pull_data(self, list_name):
since_id = self.db.get_baseline_tweet_id()
max_id = None
all_statuses = []
statuses = None
while statuses is None or len(statuses) > 0:
if list_name == 'home_timeline':
statuses = self.twitters[
self.db_summary.default_token_index].GetHomeTimeline(
count=200,
since_id=since_id,
max_id=max_id,
include_entities=False)
else:
statuses = self.twitters[
self.db_summary.polling_token_index].GetListTimeline(
owner_screen_name=self.env.polling_account,
slug=list_name,
count=200,
since_id=since_id,
max_id=max_id,
include_entities=False)
if len(statuses) > 0:
self.get_words(statuses, source=list_name)
all_statuses.extend(statuses)
max_id = statuses[-1].id - 1
logger.info('{} {}'.format(statuses[-1].id, len(statuses)))
if len(all_statuses) > 0:
max_id = max([status.id for status in all_statuses])
min_id = min([status.id for status in all_statuses])
self.db.write_list_max_id(list_name, max_id, min_id)
return len(all_statuses)
def main():
# parser = argparse.ArgumentParser(description='Draft stats for the given day and push to cloud for approval.')
# parser.add_argument('date', metavar='yyyy-mm-dd',
# help='the date to process')
#
# args = parser.parse_args()
environment = defaults.get_environment()
db = DB(environment, today())
db_summary = DBSummary(environment)
jdata = Publisher.get_pending(environment)
# c = db_connect(env.summary_database)
trenders_published = list()
trenders_all = list()
already_processed = list()
if 'tweets' in jdata:
for tweet in jdata['tweets']:
tweet_status = db_summary.get_tweet_status(tweet['t_id'])
if tweet_status is None:
db_summary.save_tweet(tweet)
for item in tweet['items']:
db_summary.save_tweet_item(tweet, item)
if tweet['type'] == 'trenders' and item['selected'] == 'Y':
trenders_all.append(item['tweet_text'][1:])
if tweet['status'] == 'pend-post':
trenders_published.append(item['tweet_text'][1:])
elif tweet_status in ['posted', 'rejected']:
tweet['status'] = tweet_status
already_processed.append(tweet)
if len(trenders_published) > 0:
with open(
f'{environment.bot_data_directory}/trenders_all_{yesterday_file()}.txt',
'a') as f:
for sn in trenders_all:
f.write("%s\n" % sn)
with open(
f'{environment.bot_data_directory}/trenders_published_{yesterday_file()}.txt',
'a') as f:
for sn in trenders_published:
f.write("%s\n" % sn)
db_summary.disconnect()
trend_date = now()
# now = now()
# yesterday = (datetime.datetime.now() - timedelta(days=2)).strftime('%Y-%m-%d')
if 'trends' in jdata:
if len(jdata['trends']) > 0:
# c = db_connect()
# t = (yesterday,)
# c.execute('SELECT max(max_id) max_id FROM tag_history where date <= ?', t)
# row = c.fetchone()
# max_id = 0
# if row != None:
# max_id = row[0]
relevant_words = db.get_relevant_words()
generic_words = db.get_generic_words()
trs = list()
for trend in jdata['trends']:
tag = '#' + trend['hashtag'].lower()
tr = {
'hashtag': trend['hashtag'],
'status': 'posted',
'trend_at': trend_date
}
trs.append(tr)
tag_discovery_result = db.get_tag_discovery_result(tag)
status = nvl(tag_discovery_result, 'NONE')
if trend['status'] == 'pend-post' and status in ('NONE',
'AUTO_DEL',
'MAN_DEL'):
logger.info('Adding: ' + tag)
db.save_tag_discovery(tag, 'MAN_ADD')
elif trend['status'] == 'pend-del' and status in ('AUTO_ADD',
'MAN_ADD'):
logger.info('Deleting: ' + tag)
db.save_tag_discovery(tag, 'MAN_DEL')
# Trend relevance
if 'relevance' in trend:
relevance = relevant_words[
trend['hashtag'].lower()] if trend['hashtag'].lower(
) in relevant_words else 'neutral'
if trend['relevance'] != relevance:
new_relevance = None if trend[
'relevance'] == 'neutral' else trend['relevance']
db.set_word_relevance(trend['hashtag'], new_relevance)
# Trend generic
if 'generic' in trend:
generic = generic_words[trend['hashtag'].lower(
)] if trend['hashtag'].lower() in generic_words else ''
if trend['generic'] != generic:
new_relevance = None if trend[
'generic'] == 'neutral' else trend['generic']
db.set_word_generic(trend['hashtag'], new_relevance)
data = {'trends': trs}
Publisher.publish(environment, data, 'trends')
db.commit()
open(f'{environment.temp_file_directory}/compute_daily',
'a').close()
if 'categories' in jdata:
if len(jdata['categories']) > 0:
for cat in jdata['categories']:
db.set_tweeter_category(cat['screen_name'], cat['category'])
logger.info("Category for", cat['screen_name'], "changed to",
cat['category'])
db.commit()
if 'words' in jdata:
if len(jdata['words']) > 0:
for word in jdata['words']:
category = word['category']
if category == '':
category = None
db.set_word_generic(word['word'], category)
logger.info("Generic for", word['word'], "changed to",
category)
db.disconnect()
if len(already_processed) > 0:
data = {'tweets': already_processed}
Publisher.publish(environment, data, 'posted')