def get_sensitive_user_detail(uid_list, date, sensitive):
results = []
index_name = str(date).replace('-','') # index_name:20130901
user_bci_results = es_cluster.mget(index=index_name, doc_type='bci', body={'ids':uid_list}, _source=True)['docs']
user_profile_results = es_user_profile.mget(index="weibo_user", doc_type="user", body={"ids":uid_list}, _source=True)['docs']
for i in range(0, len(uid_list)):
personal_info = ['']*6
uid = uid_list[i]
personal_info[0] = uid_list[i]
if user_profile_results[i]['found']:
profile_dict = user_profile_results[i]['_source']
personal_info[1] = profile_dict['nick_name']
personal_info[2] = profile_dict['user_location']
personal_info[3] = profile_dict['fansnum']
personal_info[4] = profile_dict['statusnum']
if user_bci_results[i]['found']:
personal_info[5] = user_bci_results[i]['_source'].get('user_index', 0)
else:
personal_info[5] = 0
if sensitive:
sensitive_words = r_cluster.hget('sensitive_' + index_name, str(uid))
if sensitive_words:
sensitive_dict = json.loads(sensitive_words)
personal_info.append(sensitive_dict.keys())
else:
personal_info.append([])
results.append(personal_info)
return results
def get_sensitive_user_detail(uid_list, date, sensitive):
es_cluster = es_user_profile
ts = datetime2ts(date)
results = []
index_name = pre_influence_index + str(date).replace(
'-', '') # index_name:20130901
user_bci_results = es_bci.mget(index=index_name,
doc_type='bci',
body={'ids': uid_list},
_source=False,
fields=['user_index'])['docs']
user_profile_results = es_user_profile.mget(index="weibo_user",
doc_type="user",
body={"ids": uid_list},
_source=True)['docs']
top_influnce_value = get_top_value("user_index", es_bci, index_name, "bci")
for i in range(0, len(uid_list)):
personal_info = [''] * 6
uid = uid_list[i]
personal_info[0] = uid_list[i]
personal_info[1] = uid_list[i]
if user_profile_results[i]['found']:
profile_dict = user_profile_results[i]['_source']
uname = profile_dict['nick_name']
if uname:
personal_info[1] = uname
personal_info[2] = profile_dict['user_location']
personal_info[3] = profile_dict['fansnum']
personal_info[4] = profile_dict['statusnum']
if user_bci_results[i]['found']:
try:
tmp_bci = user_bci_results[i]['fields']['user_index'][0]
influence = math.log(
tmp_bci / float(top_influnce_value) * 9 + 1, 10) * 100
personal_info[5] = influence
except:
personal_info[5] = 0
else:
personal_info[5] = 0
if sensitive:
sensitive_words = redis_cluster.hget('sensitive_' + str(ts),
str(uid))
if sensitive_words:
sensitive_dict = json.loads(sensitive_words)
personal_info.append(sensitive_dict.keys())
else:
personal_info.append([])
else:
personal_info.append([])
results.append(personal_info)
return results
def get_sensitive_user_detail(uid_list, date, sensitive):
es_cluster = es_user_profile
ts = datetime2ts(date)
results = []
index_name = pre_influence_index + str(date).replace('-','') # index_name:20130901
user_bci_results = es_bci.mget(index=index_name, doc_type='bci', body={'ids':uid_list}, _source=False, fields=['user_index'])['docs']
user_profile_results = es_user_profile.mget(index="weibo_user", doc_type="user", body={"ids":uid_list}, _source=True)['docs']
top_influnce_value = get_top_value("user_index", es_bci, index_name, "bci")
for i in range(0, len(uid_list)):
personal_info = ['']*6
uid = uid_list[i]
personal_info[0] = uid_list[i]
personal_info[1] = uid_list[i]
if user_profile_results[i]['found']:
profile_dict = user_profile_results[i]['_source']
uname = profile_dict['nick_name']
if uname:
personal_info[1] = uname
personal_info[2] = profile_dict['user_location']
personal_info[3] = profile_dict['fansnum']
personal_info[4] = profile_dict['statusnum']
if user_bci_results[i]['found']:
try:
tmp_bci = user_bci_results[i]['fields']['user_index'][0]
influence = math.log(tmp_bci/float(top_influnce_value)*9+1, 10)*100
personal_info[5] = influence
except:
personal_info[5] = 0
else:
personal_info[5] = 0
if sensitive:
sensitive_words = redis_cluster.hget('sensitive_' + str(ts), str(uid))
if sensitive_words:
sensitive_dict = json.loads(sensitive_words)
personal_info.append(sensitive_dict.keys())
else:
personal_info.append([])
else:
personal_info.append([])
results.append(personal_info)
return results
def get_sensitive_user_detail(uid_list, date, sensitive):
results = []
index_name = str(date).replace('-', '') # index_name:20130901
user_bci_results = es_cluster.mget(index=index_name,
doc_type='bci',
body={'ids': uid_list},
_source=True)['docs']
user_profile_results = es_user_profile.mget(index="weibo_user",
doc_type="user",
body={"ids": uid_list},
_source=True)['docs']
for i in range(0, len(uid_list)):
personal_info = [''] * 6
uid = uid_list[i]
personal_info[0] = uid_list[i]
if user_profile_results[i]['found']:
profile_dict = user_profile_results[i]['_source']
personal_info[1] = profile_dict['nick_name']
personal_info[2] = profile_dict['user_location']
personal_info[3] = profile_dict['fansnum']
personal_info[4] = profile_dict['statusnum']
if user_bci_results[i]['found']:
personal_info[5] = user_bci_results[i]['_source'].get(
'user_index', 0)
else:
personal_info[5] = 0
if sensitive:
sensitive_words = r_cluster.hget('sensitive_' + index_name,
str(uid))
if sensitive_words:
sensitive_dict = json.loads(sensitive_words)
personal_info.append(sensitive_dict.keys())
else:
personal_info.append([])
results.append(personal_info)
return results
def full_text_search(keywords, uid, start_time, end_time, size):
results = []
uid_list = []
user_profile_list = []
query_body = {
"query": {
"filtered":{
"filter":{
"bool": {
"must": []
}
}
}
},
"size":size,
"sort":{"timestamp":{"order": 'desc'}}
}
if RUN_TYPE:
query_body["sort"] = {"user_fansnum":{"order": 'desc'}}
if uid:
query_body["query"]["filtered"]["filter"]["bool"]["must"].append({"term":{"uid":uid}})
if keywords:
keywords_list = keywords.split(',')
for word in keywords_list:
query_body["query"]["filtered"]["filter"]["bool"]["must"].append({'wildcard':{'text':{'wildcard':'*'+word+'*'}}})
index_list = []
exist_bool = es_flow_text.indices.exists(index="flow_text_"+end_time)
if start_time:
start_ts = datetime2ts(start_time)
end_ts = datetime2ts(end_time)
ts = end_ts
while 1:
index_name = "flow_text_"+ts2datetime(ts)
exist_bool = es_flow_text.indices.exists(index=index_name)
if exist_bool:
index_list.append(index_name)
if ts == start_ts:
break
else:
ts -= 3600*24
print index_list
# 没有可行的es
if not index_list:
return []
search_results = es_flow_text.search(index=index_list, doc_type="text", body=query_body)["hits"]["hits"]
for item in search_results:
uid_list.append(item['_source']['uid'])
history_max = get_history_max()
personal_field = ["nick_name", "fansnum", "statusnum","user_location"]
user_info = get_user_profile(uid_list, personal_field)
bci_results = ES_CLUSTER_FLOW1.mget(index="bci_history", doc_type="bci", body={"ids":uid_list}, _source=False, fields=["bci_day_last"])["docs"]
sensitive_results = es_sensitive_history.mget(index="sensitive_history", doc_type="sensitive", body={"ids":uid_list}, _source=False, fields=["last_value"])["docs"]
count = 0
for item in search_results:
item = item['_source']
uid_list.append(item['uid'])
iter_item = []
iter_item.append(item['uid'])
iter_item.append(user_info[count][1])
iter_item.append(item['text'])
iter_item.append(ts2date(item['timestamp']))
iter_item.append(item['geo'])
if item.get("sensitive_words_string", ''):
iter_item.append(item['sensitive_words_string'].split('&'))
else:
iter_item.append([])
iter_item.append(item.get('retweeted', 0))
iter_item.append(item.get('comment', 0))
count += 1
results.append(iter_item)
user_set = set()
count = 0
for item in user_info:
if item[0] in user_set:
continue
else:
user_set.add(item[0])
if bci_results[count]["found"]:
bci_value = bci_results[count]["fields"]["bci_day_last"][0]
item.append(normalize_index(bci_value, history_max["max_bci"]))
else:
item.append(0)
if sensitive_results[count]["found"]:
sensitive_value = sensitive_results[count]['fields']['last_value'][0]
item.append(normalize_index(sensitive_value, history_max["max_sensitive"]))
else:
item.append(0)
user_profile_list.append(item)
return results, user_profile_list
def influenced_user_detail(uid, date, origin_retweeted_mid, retweeted_retweeted_mid, message_type, default_number=20):
query_body = {
"query":{
"filtered":{
"filter":{
"bool":{
"must": [
]
}
}
}
},
"size":20000,
}
if RUN_TYPE == 1:
query_body["sort"] = {"user_fansnum":{"order":"desc"}}
#详细影响到的人
date1 = str(date).replace('-', '')
index_name = pre_index + date1
index_flow_text = pre_text_index + date
origin_retweeted_uid = [] # influenced user uid_list
retweeted_retweeted_uid = []
origin_comment_uid = []
retweeted_comment_uid = []
query_origin = copy.deepcopy(query_body)
query_retweeted = copy.deepcopy(query_body)
if origin_retweeted_mid: # 所有转发该条原创微博的用户
query_origin["query"]["filtered"]["filter"]["bool"]["must"].append({"terms": {"root_mid": origin_retweeted_mid}})
query_origin["query"]["filtered"]["filter"]["bool"]["must"].extend([{"term":{"message_type": message_type}}, {"term":{"root_uid": uid}}])
origin_retweeted_result = es.search(index=index_flow_text, doc_type=flow_text_index_type, body=query_origin, fields=["uid"])["hits"]["hits"]
if origin_retweeted_result:
for item in origin_retweeted_result:
origin_retweeted_uid.append(item["fields"]["uid"][0])
if retweeted_retweeted_mid: # 所有评论该条原创微博的用户
query_retweeted["query"]["filtered"]["filter"]["bool"]["must"].append({"terms": {"root_mid": retweeted_retweeted_mid}})
query_retweeted["query"]["filtered"]["filter"]["bool"]["must"].extend([{"term":{"message_type": message_type}},{"term": {"directed_uid": uid}}])
retweeted_retweeted_result = es.search(index=index_flow_text, doc_type=flow_text_index_type, body=query_retweeted, fields=["uid"])["hits"]["hits"]
if retweeted_retweeted_result:
for item in retweeted_retweeted_result:
retweeted_retweeted_uid.append(item["fields"]["uid"][0])
retweeted_uid_list = [] # all retweeted user list
retweeted_results = {} # statistics of all retweeted uid information
retweeted_domain = {}
retweeted_topic = {}
retweeted_geo = {}
bci_results = {}
in_portrait = []
out_portrait = []
average_influence = 0
total_influence = 0
count = 0
all_uid_set = set(origin_retweeted_uid) | set(retweeted_retweeted_uid)
retweeted_uid_list.extend(origin_retweeted_uid)
retweeted_uid_list.extend(retweeted_retweeted_uid)
retweeted_uid_list = list(set(retweeted_uid_list) - set([uid])) # filter uids
if retweeted_uid_list:
user_portrait_result = es_user_portrait.mget(index=user_portrait, doc_type=portrait_index_type, body={"ids": retweeted_uid_list}, fields=["domain", "topic_string", "activity_geo_dict","importance", "influence"])["docs"]
bci_index = "bci_" + date.replace('-', '')
bci_results = es_cluster.mget(index=bci_index, doc_type="bci", body={"ids":retweeted_uid_list}, fields=['user_index'])["docs"]
for item in user_portrait_result:
if item["found"]:
temp = []
count += 1
temp.append(item['_id'])
temp.append(item["fields"]["importance"][0])
in_portrait.append(temp)
temp_domain = item["fields"]["domain"][0].split('&')
temp_topic = item["fields"]["topic_string"][0].split('&')
temp_geo = json.loads(item["fields"]["activity_geo_dict"][0])[-1].keys()
#total_influence += item["fields"]["influence"][0]
retweeted_domain = aggregation(temp_domain, retweeted_domain)
retweeted_topic = aggregation(temp_topic, retweeted_topic)
retweeted_geo = aggregation(temp_geo, retweeted_geo)
else:
out_portrait.append(item['_id'])
retweeted_domain = proportion(retweeted_domain)
retweeted_topic = proportion(retweeted_topic)
retweeted_geo = proportion(retweeted_geo)
if bci_results:
total_influence = 0
for item in bci_results:
if item['found']:
total_influence += item['fields']['user_index'][0]
try:
average_influence = total_influence/len(retweeted_uid_list)
except:
average_influence = 0
sorted_retweeted_domain = sorted(retweeted_domain.items(),key=lambda x:x[1], reverse=True)
sorted_retweeted_topic = sorted(retweeted_topic.items(),key=lambda x:x[1], reverse=True)
sorted_retweeted_geo = sorted(retweeted_geo.items(), key=lambda x:x[1], reverse=True)
retweeted_results["domian"] = sorted_retweeted_domain[:5]
retweeted_results["topic"] = sorted_retweeted_topic[:5]
retweeted_results["geo"] = sorted_retweeted_geo[:5]
retweeted_results["influence"] = average_influence
in_portrait = sorted(in_portrait, key=lambda x:x[1], reverse=True)
temp_list = []
for item in in_portrait:
temp_list.append(item[0])
retweeted_results['in_portrait_number'] = len(temp_list)
retweeted_results['out_portrait_number'] = len(out_portrait)
in_portrait_url = get_user_url(temp_list[:default_number])
out_portrait_url = get_user_url(out_portrait[:default_number])
retweeted_results["in_portrait"] = in_portrait_url
retweeted_results["out_portrait"] = out_portrait_url
retweeted_results["total_number"] = len(temp_list) + len(out_portrait)
return retweeted_results
def influenced_people(uid, mid, influence_style, date, default_number=20):
# uid
# which weibo----mid, retweeted weibo ---seek for root_mid
# influence_style: retweeted(0) or comment(1)
date1 = ts2datetime(datetime2ts(date)).replace('-', '')
index_name = pre_index + date1
index_flow_text = pre_text_index + date
text_result = es.get(index=index_flow_text, doc_type=flow_text_index_type, id=mid)["_source"]
temp_mid = text_result.get("root_mid",'') #判断微博是否是原创微博
if temp_mid:
mid_type = 1 # 非原创微博
else:
mid_type = 0 # 原创微博
query_body = {
"query":{
"filtered":{
"filter":{
"bool":{
"must":[
]
}
}
}
},
"size": 30000
}
if RUN_TYPE:
query_body["sort"] = {"user_fansnum":{"order":"desc"}}
if int(mid_type) == 0:
if int(influence_style) == 0: # origin weibo, all retweeted people
query_body["query"]["filtered"]["filter"]["bool"]["must"].extend([{"term": {"root_uid": uid}}, {"term": {"message_type": 3}}, {"term": {"root_mid": mid}}])
else: # commented people
query_body["query"]["filtered"]["filter"]["bool"]["must"].extend([{"term": {"directed_uid": uid}}, {"term": {"message_type": 2}}, {"term": {"root_mid": mid}}])
else:
if int(influence_style) == 0: # origin weibo, all retweeted people
query_body["query"]["filtered"]["filter"]["bool"]["must"].extend([{"term": {"directed_uid": uid}}, {"term": {"message_type": 3}}, {"term": {"root_mid": temp_mid}}])
else: # commented people
query_body["query"]["filtered"]["filter"]["bool"]["must"].extend([{"term": {"directed_uid": uid}}, {"term": {"message_type": 2}}, {"term": {"root_mid": temp_mid}}])
search_results = es.search(index=index_flow_text, doc_type=flow_text_index_type, body=query_body, _source=False, fields=["uid"], timeout=30)["hits"]["hits"]
results = [] # uid_list
if search_results:
for item in search_results:
if int(item["fields"]["uid"][0]) == int(uid):
pass
else:
results.append(item["fields"]["uid"][0])
results = list(set(results))
else:
results = []
bci_index = "bci_" + date.replace('-','')
if results:
portrait_results = es_user_portrait.mget(index=user_portrait, doc_type=portrait_index_type, body={"ids": results}, fields=["domain", "topic_string", "activity_geo_dict","importance", "influence"])["docs"]
bci_results = es_cluster.mget(index=bci_index, doc_type='bci', body={"ids":results}, fields=['user_index'])['docs']
else:
portrait_results = {}
bci_results = {}
in_portrait = []
out_portrait = []
in_portrait_info = []
retweeted_domain = {}
retweeted_topic = {}
retweeted_geo = {}
average_influence = 0
total_influence = 0
count = 0
if bci_results:
total_influence = 0
for item in bci_results:
if item['found']:
total_influence += item['fields']['user_index'][0]
try:
average_influence = total_influence/len(results)
except:
average_influence = 0
if portrait_results:
for item in portrait_results:
if item["found"]:
temp = []
count += 1
temp.append(item['_id'])
temp.append(item["fields"]["importance"][0])
in_portrait.append(temp)
temp_domain = item["fields"]["domain"][0].split('&')
temp_topic = item["fields"]["topic_string"][0].split('&')
temp_geo = json.loads(item["fields"]["activity_geo_dict"][0])[-1].keys()
#total_influence += item["fields"]["influence"][0]
retweeted_domain = aggregation(temp_domain, retweeted_domain)
retweeted_topic = aggregation(temp_topic, retweeted_topic)
retweeted_geo = aggregation(temp_geo, retweeted_geo)
else:
out_portrait.append(item['_id'])
retweeted_domain = proportion(retweeted_domain)
retweeted_topic = proportion(retweeted_topic)
retweeted_geo = proportion(retweeted_geo)
#try:
# average_influence = total_influence/count
#except:
# average_influence = 0
sorted_retweeted_domain = sorted(retweeted_domain.items(),key=lambda x:x[1], reverse=True)
sorted_retweeted_topic = sorted(retweeted_topic.items(),key=lambda x:x[1], reverse=True)
sorted_retweeted_geo = sorted(retweeted_geo.items(), key=lambda x:x[1], reverse=True)
retweeted_results = dict()
retweeted_results["domian"] = sorted_retweeted_domain[:5]
retweeted_results["topic"] = sorted_retweeted_topic[:5]
retweeted_results["geo"] = sorted_retweeted_geo[:5]
retweeted_results["influence"] = average_influence
in_portrait = sorted(in_portrait, key=lambda x:x[1], reverse=True)
temp_list = []
for item in in_portrait:
temp_list.append(item[0])
retweeted_results['in_portrait_number'] = len(temp_list)
retweeted_results['out_portrait_number'] = len(out_portrait)
in_portrait_url = get_user_url(temp_list[:default_number])
out_portrait_url = get_user_url(out_portrait[:default_number])
return_results = dict()
return_results["influence_users"] = [in_portrait_url, out_portrait_url]
return_results["influence_distribution"] = retweeted_results
return return_results
def full_text_search(keywords, uid, start_time, end_time, size):
results = []
uid_list = []
user_profile_list = []
query_body = {
"query": {
"bool": {
"must": []
}
},
"size":size,
"sort":{"timestamp":{"order": 'desc'}}
}
if RUN_TYPE:
query_body["sort"] = {"user_fansnum":{"order": 'desc'}}
if uid:
query_body["query"]["bool"]["must"].append({"term":{"uid":uid}})
if keywords:
keywords_list = keywords.split(',')
for word in keywords_list:
query_body["query"]["bool"]["must"].append({'wildcard':{'text':{'wildcard':'*'+word+'*'}}})
index_list = []
exist_bool = es_flow_text.indices.exists(index="flow_text_"+end_time)
if start_time:
start_ts = datetime2ts(start_time)
end_ts = datetime2ts(end_time)
ts = end_ts
while 1:
index_name = "flow_text_"+ts2datetime(ts)
exist_bool = es_flow_text.indices.exists(index=index_name)
if exist_bool:
index_list.append(index_name)
if ts == start_ts:
break
else:
ts -= 3600*24
print index_list
# 没有可行的es
if not index_list:
return [[], []]
search_results = es_flow_text.search(index=index_list, doc_type="text", body=query_body)["hits"]["hits"]
for item in search_results:
uid_list.append(item['_source']['uid'])
user_info = []
if uid_list:
history_max = get_history_max()
personal_field = ["nick_name", "fansnum", "statusnum","user_location"]
user_info = get_user_profile(uid_list, personal_field)
bci_results = ES_CLUSTER_FLOW1.mget(index="bci_history", doc_type="bci", body={"ids":uid_list}, _source=False, fields=["bci_day_last"])["docs"]
in_portrait = es_user_portrait.mget(index="sensitive_user_portrait", doc_type="user", body={"ids":uid_list}, _source=False)["docs"]
sensitive_results = es_sensitive_history.mget(index="sensitive_history", doc_type="sensitive", body={"ids":uid_list}, _source=False, fields=["last_value"])["docs"]
print "len search: ", len(search_results)
count = 0
# uid uname text date geo sensitive_words retweeted comment
for item in search_results:
item = item['_source']
uid_list.append(item['uid'])
iter_item = []
iter_item.append(item['uid'])
iter_item.append(user_info[count][1])
iter_item.append(item['text'])
iter_item.append(ts2date(item['timestamp']))
iter_item.append(item['geo'])
if item.get("sensitive_words_string", ''):
iter_item.append(item['sensitive_words_string'].split('&'))
else:
iter_item.append([])
iter_item.append(item.get('retweeted', 0))
iter_item.append(item.get('comment', 0))
count += 1
results.append(iter_item)
user_set = set()
count = 0
# uid "nick_name", "fansnum", "statusnum","user_location", bci, sensitive
for item in user_info:
if item[0] in user_set:
continue
else:
user_set.add(item[0])
if bci_results[count]["found"]:
if bci_results[count].has_key("fields"):
bci_value = bci_results[count]["fields"]["bci_day_last"][0]
else:
bci_value = 0
item.append(normalize_index(bci_value, history_max["max_bci"]))
else:
item.append(0)
if sensitive_results[count]["found"]:
if sensitive_results[count].has_key("fields"):
sensitive_value = sensitive_results[count]['fields']['last_value'][0]
else:
sensitive_value = 0
item.append(normalize_index(sensitive_value, history_max["max_sensitive"]))
else:
item.append(0)
if in_portrait[count]["found"]:
item.append("1")
else:
item.append("0")
user_profile_list.append(item)
return results, user_profile_list
