def add_task_record_time(task_name, submit_date):
status = 0
#start_ts = datetime2ts(submit_date)
start_ts = date2ts(submit_date)
r_task.hset('monitor_task_time_record', task_name, start_ts)
status = 1
return status
def add_task_record_time(task_name, submit_date):
status = 0
#start_ts = datetime2ts(submit_date)
start_ts = date2ts(submit_date)
r_task.hset('monitor_task_time_record', task_name, start_ts)
status = 1
return status
def get_network(task_exist):
task_name = task_exist['task_name']
submit_date = task_exist['submit_date']
submit_ts = date2ts(submit_date)
time_segment = 24*3600
now_ts = time.time()
now_date = ts2datetime(now_ts)
now_date_ts = datetime2ts(now_date)
#test
now_date_ts = datetime2ts('2013-09-07')
iter_date_ts = now_date_ts
iter_count = 1
date_list = []
top_list_dict = {}
while True:
if iter_count >= 8 or iter_date_ts < submit_ts:
break
iter_date = ts2datetime(iter_date_ts)
date_list.append(iter_date)
key = 'inner_' + str(iter_date)
try:
task_date_result = es.get(index=monitor_index_name, doc_type=task_name, id=key)['_source']
except:
task_date_result = {}
#print 'task_name, key, task_date_result:', task_name, key, task_date_result
iter_field = ['top1', 'top2', 'top3', 'top4', 'top5']
for field in iter_field:
user_count_item = json.loads(task_date_result[field])
uid = user_count_item[0]
uname = uid2uname(uid)
count = user_count_item[1]
try:
top_list_dict[field].append([uid, uname, count])
except:
top_list_dict[field] = [[uid, uname, count]]
iter_date_ts -= time_segment
# get inner-retweet group from es---field: inner_graph
'''
try:
inner_graph = json.loads(task_date_result['inner_graph'])
except:
inner_graph = {}
'''
abnormal_index = compute_inner_polarization(top_list_dict)
return [date_list, top_list_dict, abnormal_index]
def compute_mid_result(task_name, task_submit_date):
result = {'count_0':{}, 'count_1':{}, 'sentiment_0_126':{}, 'sentiment_0_127':{}, 'sentiment_0_128':{},\
'sentiment_0_129':{}, 'sentiment_0_130':{}, 'sensitive_score':{}, 'geo_0':{}, 'geo_1':{},\
'hashtag_0':{}, 'hashtag_1':{}, 'sentiment_1_126':{}, 'sentiment_1_127':{}, \
'sentiment_1_128':{}, 'sentiment_1_129':{}, 'sentiment_1_130':{}}
#geo & hashtag: day
#other: 15min
search_time_segment = 3600 * 4
#start_ts = datetime2ts(task_submit_date)
start_ts = date2ts(task_submit_date)
now_ts = time.time()
now_date = ts2datetime(now_ts)
#test
now_ts = datetime2ts('2013-09-08')
date_ts = datetime2ts(now_date)
segment = int((now_ts - date_ts) / 900) + 1
end_ts = date_ts + segment * 900
#every search time-range: 4 hour----bulk action to search
begin_ts = start_ts
while True:
if begin_ts >= end_ts:
break
compute_ts = ts2date(begin_ts)
#print 'compute ts:', compute_ts
query_body = {'range':{'timestamp':{'from': begin_ts, 'to':begin_ts+search_time_segment}}}
try:
mid_result_list = es.search(index=monitor_index_name, doc_type=task_name, body={'query':query_body, 'size':100000, 'sort':[{'timestamp':{'order': 'asc'}}]})['hits']['hits']
except Exception, e:
raise e
if mid_result_list:
for mid_result_item in mid_result_list:
result_item = mid_result_item['_source']
timestamp = result_item['timestamp']
#attr_count
#print 'compute_count'
count_dict = json.loads(result_item['count'])
for sensitive in count_dict:
count_key = 'count_' + sensitive
result[count_key][str(timestamp)] = count_dict[sensitive]
#attr_sentiment
#print 'compute_sentiment'
sensitive_sentiment_dict = json.loads(result_item['sentiment'])
for sensitive in sensitive_sentiment_dict:
sentiment_dict = sensitive_sentiment_dict[sensitive]
for sentiment in sentiment_dict:
sentiment_key = 'sentiment_'+sensitive+'_'+sentiment
result[sentiment_key][str(timestamp)] = sentiment_dict[sentiment]
#attr_sensitive_score
#print 'compute_sensitive_word'
if 'sensitive_word' in result_item:
sensitive_word_dict = json.loads(result_item['sensitive_word'])
else:
sensitive_word_dict = {}
ts_word_score = 0
for word in sensitive_word_dict:
#print 'word:', json.dumps(word.encode('utf-8')), word.encode('utf-8'), type(word.encode('utf-8'))
search_word = word.encode('utf-8')
#print 'search_word:', search_word, type(search_word)
try:
word_identify = json.loads(word_r.hget('sensitive_words', search_word))
except:
word_identify = [2]
ts_word_score += sensitive_word_dict[word] * word_identify[0]
result['sensitive_score'][str(timestamp)] = ts_word_score
#attr_geo
#print 'compute geo'
timestamp_date = ts2datetime(timestamp)
sensitive_geo_dict = json.loads(result_item['geo'])
for sensitive in sensitive_geo_dict:
if timestamp_date not in result['geo_'+sensitive]:
result['geo_'+sensitive][timestamp_date] = {}
geo_dict = sensitive_geo_dict[sensitive]
for geo in geo_dict:
try:
result['geo_'+sensitive][timestamp_date][geo] += geo_dict[geo]
except:
result['geo_'+sensitive][timestamp_date][geo] = geo_dict[geo]
#attr_hashtag
#print 'compute hashtag'
if 'hashtag' in result_item:
sensitive_hashtag_dict = json.loads(result_item['hashtag'])
else:
sensitive_hashtag_dict = {}
result['hashtag_0'][timestamp_date] = {}
result['hashtag_1'][timestamp_date] = {}
for sensitive in sensitive_hashtag_dict:
for sensitive in sensitive_hashtag_dict:
if timestamp_date not in result['hashtag_'+sensitive]:
result['hashtag_'+sensitive][timestamp_date] = {}
hashtag_dict = sensitive_hashtag_dict[sensitive]
for hashtag in hashtag_dict:
try:
result['hashtag_'+sensitive][timestamp_date][hashtag] += hashtag_dict[hashtag]
except:
result['hashtag_'+sensitive][timestamp_date][hashtag] = hashtag_dict[hashtag]
begin_ts += search_time_segment
def get_user_comment_retweet(task_exist):
result = {} # result = {'uid1_comment':{ts:value}, 'uid1_retweet':{ts_value}, 'uid2_comment'}
submit_date = task_exist['submit_date']
start_ts = date2ts(submit_date)
task_status = task_exist['status']
if task_status == 1:
now_ts = time.time()
now_date = ts2datetime(now_ts)
now_date_ts = datetime2ts(now_date)
segment = int((now_ts - now_date_ts) / 900) + 1
end_ts = now_date_ts + segment * 900
#test
end_ts = datetime2ts('2013-09-02')
else:
end_ts = date2ts(task_exist['end_date'])
task_user = task_exist['uid_list']
select_top_dict = {} # {uid:[ave_retweet_count, ave_peak_retweet_count]}
#select union top5 ave_retweet_count and top5 ave_peak_retweet_count
for user in task_user:
result[user+'_comment'] = {}
result[user+'_retweet'] = {}
comment_retweet_dict = monitor_r.hgetall(user)
for item in comment_retweet_dict:
item_type_ts = item.split('_')
item_type = item_type_ts[0]
item_ts = item_type_ts[1]
result[user+'_'+item_type][item_ts] = int(comment_retweet_dict[item])
# use to detect peaks
comment_dict = result[user+'_comment']
complement_comment_dict = complement_ts(comment_dict, start_ts, end_ts)
sort_comment_dict = sorted(complement_comment_dict.items(), key=lambda x:int(x[0]))
detect_peaks_comment_input = [item[1] for item in sort_comment_dict]
#print 'detect_peaks_comment_input:', detect_peaks_comment_input
result[user+'_comment_peak'] = detect_peaks(detect_peaks_comment_input)
retweet_dict = result[user+'_retweet']
complement_retweet_dict = complement_ts(retweet_dict, start_ts, end_ts)
sort_retweet_dict = sorted(complement_retweet_dict.items(), key=lambda x:int(x[0]))
detect_peaks_retweet_input = [item[1] for item in sort_retweet_dict]
result[user+'_retweet_peak'] = detect_peaks(detect_peaks_retweet_input)
ave_retweet_count = sum(detect_peaks_retweet_input) / len(detect_peaks_retweet_input)
peak_count_list = [detect_peaks_retweet_input[peak_location] for peak_location in result[user+'_retweet_peak']]
ave_peak_count = sum(peak_count_list) / len(peak_count_list)
select_top_dict[user] = [ave_retweet_count, ave_peak_count]
#select union top5
sort_select_top_count_dict = sorted(select_top_dict.items(), key=lambda x:x[1][0], reverse=True)
top5_count_user_list = sort_select_top_count_dict[:5]
top5_count_user = [item[0] for item in top5_count_user_list]
sort_select_top_peak_dict = sorted(select_top_dict.items(), key=lambda x:x[1][1], reverse=True)
top5_peak_user_list = sort_select_top_peak_dict[:5]
top5_peak_user = [item[0] for item in top5_peak_user_list]
union_user = list(set(top5_count_user) & set(top5_peak_user))
new_result = {}
for user in union_user:
new_result[user+'_retweet'] = result[user+'_retweet']
new_result[user+'_retweet_peak'] = result[user+'_retweet_peak']
new_result[user+'_comment'] = result[user+'_comment']
new_result[user+'_comment_peak'] = result[user+'_comment_peak']
new_result['profile'] = get_top_user_profile(union_user)
#compute abnormal index
new_result['abnormal_index'] = compute_comment_retweet_abnormal(new_result, union_user)
return new_result
