def bunch_user_tweets_dataframe(dbname, comname, timename, filename, num_batch=2, n=-1):
'''
:param dbname: db name
:param comname: user collection name
:param timename: timeline collection name
:param n: split tweets every n
:return: pandas dataframe
'''
db = dbt.db_connect_no_auth(dbname)
com = db[comname]
times = db[timename]
liwc_results = []
indices = []
user_dis = []
user_create_time = []
first_tweet_time = []
last_tweet_time = []
counts = []
fields = []
split_k = False
if n == -1:
split_k = True
for user in com.find({'timeline_count': {'$gt': 100}}, ['id', 'id_str', 'created_at', 'timeline_count'], no_cursor_timeout=True):
uid = user['id']
tweet_count = user['timeline_count']
if split_k == True:
n = (tweet_count-1)/num_batch
print '---------------------------------------------'
print '%d tweets batched in every %d' %(tweet_count, n)
count = 0
index = 0
tweets = []
for tweet in times.find({'user.id': uid}).sort([("id", 1)]):
if count < n:
tweets.append(tweet)
count += 1
else:
result = liwcp.process_tweet(tweets, Trim_rt=False)
if result:
liwc_results.append([result[k] for k in result.keys()])
if len(fields) == 0:
fields = [k for k in result.keys()]
user_dis.append(user['id_str'])
indices.append(index)
counts.append(n)
user_create_time.append(datetime.strptime(user['created_at'], '%a %b %d %H:%M:%S +0000 %Y'))
first_tweet_time.append(datetime.strptime(tweets[0]['created_at'], '%a %b %d %H:%M:%S +0000 %Y'))
last_tweet_time.append(datetime.strptime(tweets[-1]['created_at'], '%a %b %d %H:%M:%S +0000 %Y'))
print 'User %s, in time %d with %d tweets ---- verify %d tweets' %(user['id_str'], index, count, len(tweets))
index += 1
tweets = [tweet]
count = 1
liwc_results = np.array(liwc_results)
size = len(user_create_time)
counts = np.reshape(counts, (size, 1))
user_create_time = np.reshape(user_create_time, (size, 1))
first_tweet_time = np.reshape(first_tweet_time, (size, 1))
last_tweet_time = np.reshape(last_tweet_time, (size, 1))
user_dis = np.reshape(user_dis, (size, 1))
indices = np.reshape(indices, (size, 1))
user_dis = np.append(user_dis, indices, axis=1)
user_dis = np.append(user_dis, user_create_time, axis=1)
user_dis = np.append(user_dis, first_tweet_time, axis=1)
user_dis = np.append(user_dis, last_tweet_time, axis=1)
user_dis = np.append(user_dis, counts, axis=1)
liwc_results = np.append(user_dis, liwc_results, axis=1)
print 'user matrix', liwc_results.shape
df = pd.DataFrame(data=liwc_results,
columns=['user_id', 'time_index', 'user_created_time', 'first_tweet_time', 'last_tweet_time', 'count'] + fields)
df.to_csv(filename)
df.to_pickle(filename+'.pick')
def bunch_user_tweets_dataframe(dbname,
comname,
timename,
filename,
num_batch=2,
n=-1):
'''
:param dbname: db name
:param comname: user collection name
:param timename: timeline collection name
:param n: split tweets every n
:return: pandas dataframe
'''
db = dbt.db_connect_no_auth(dbname)
com = db[comname]
times = db[timename]
liwc_results = []
indices = []
user_dis = []
user_create_time = []
first_tweet_time = []
last_tweet_time = []
counts = []
fields = []
split_k = False
if n == -1:
split_k = True
for user in com.find({'timeline_count': {
'$gt': 0
}}, ['id', 'id_str', 'created_at', 'timeline_count'],
no_cursor_timeout=True):
uid = user['id']
tweet_count = user['timeline_count']
if split_k == True:
n = (tweet_count - 1) / num_batch
print '---------------------------------------------'
print '%d tweets batched in every %d' % (tweet_count, n)
count = 0
index = 0
tweets = []
for tweet in times.find({'user.id': uid}).sort([("id", 1)]):
if count < n:
tweets.append(tweet)
count += 1
else:
result = liwcp.process_tweet(tweets, Trim_rt=True)
if result:
liwc_results.append([result[k] for k in result.keys()])
if len(fields) == 0:
fields = [k for k in result.keys()]
user_dis.append(user['id_str'])
indices.append(index)
counts.append(n)
user_create_time.append(
datetime.strptime(user['created_at'],
'%a %b %d %H:%M:%S +0000 %Y'))
first_tweet_time.append(
datetime.strptime(tweets[0]['created_at'],
'%a %b %d %H:%M:%S +0000 %Y'))
last_tweet_time.append(
datetime.strptime(tweets[-1]['created_at'],
'%a %b %d %H:%M:%S +0000 %Y'))
print 'User %s, in time %d with %d tweets ---- verify %d tweets' % (
user['id_str'], index, count, len(tweets))
index += 1
tweets = [tweet]
count = 1
liwc_results = np.array(liwc_results)
size = len(user_create_time)
counts = np.reshape(counts, (size, 1))
user_create_time = np.reshape(user_create_time, (size, 1))
first_tweet_time = np.reshape(first_tweet_time, (size, 1))
last_tweet_time = np.reshape(last_tweet_time, (size, 1))
user_dis = np.reshape(user_dis, (size, 1))
indices = np.reshape(indices, (size, 1))
user_dis = np.append(user_dis, indices, axis=1)
user_dis = np.append(user_dis, user_create_time, axis=1)
user_dis = np.append(user_dis, first_tweet_time, axis=1)
user_dis = np.append(user_dis, last_tweet_time, axis=1)
user_dis = np.append(user_dis, counts, axis=1)
liwc_results = np.append(user_dis, liwc_results, axis=1)
print 'user matrix', liwc_results.shape
df = pd.DataFrame(data=liwc_results,
columns=[
'user_id', 'time_index', 'user_created_time',
'first_tweet_time', 'last_tweet_time', 'count'
] + fields)
df.to_csv(filename)
df.to_pickle(filename + '.pick')
def bunch_user_tweets_panel(dbname, comname, timename, n=100):
'''
:param dbname: db name
:param comname: user collection name
:param timename: timeline collection name
:param n: split tweets every n
:return: pandas panel
'''
db = dbt.db_connect_no_auth(dbname)
com = db[comname]
times = db[timename]
data = {}
for user in com.find({'timeline_count': {'$gt': 500}}, ['id', 'id_str', 'created_at']):
uid = user['id']
liwc_results = []
indices = []
user_create_time = []
first_tweet_time = []
last_tweet_time = []
fields = []
count = 0
index = 0
tweets = []
for tweet in times.find({'user.id': uid}).sort([("id", 1)]):
if count < n:
tweets.append(tweet)
count += 1
else:
result = liwcp.process_tweet(tweets, Trim_rt=False)
if result:
liwc_results.append([result[k] for k in result.keys()])
if len(fields) == 0:
fields = [k for k in result.keys()]
indices.append(index)
user_create_time.append(datetime.strptime(user['created_at'], '%a %b %d %H:%M:%S +0000 %Y'))
first_tweet_time.append(datetime.strptime(tweets[0]['created_at'], '%a %b %d %H:%M:%S +0000 %Y'))
last_tweet_time.append(datetime.strptime(tweets[-1]['created_at'], '%a %b %d %H:%M:%S +0000 %Y'))
# print index, count
index += 1
count = 0
tweets = []
liwc_results = np.array(liwc_results)
size = len(user_create_time)
user_create_time = np.reshape(user_create_time, (size, 1))
first_tweet_time = np.reshape(first_tweet_time, (size, 1))
last_tweet_time = np.reshape(last_tweet_time, (size, 1))
liwc_results = np.append(liwc_results, user_create_time, axis=1)
liwc_results = np.append(liwc_results, first_tweet_time, axis=1)
liwc_results = np.append(liwc_results, last_tweet_time, axis=1)
print liwc_results.shape
df = pd.DataFrame(data=liwc_results,
columns=fields + ['user_created_time', 'first_tweet_time', 'last_tweet_time'],
index=indices)
data[user['id_str']] = df
pn = pd.Panel(data)
pn.to_pickle('ed-liwc.panel')
def bunch_user_tweets_panel(dbname, comname, timename, n=100):
'''
:param dbname: db name
:param comname: user collection name
:param timename: timeline collection name
:param n: split tweets every n
:return: pandas panel
'''
db = dbt.db_connect_no_auth(dbname)
com = db[comname]
times = db[timename]
data = {}
for user in com.find({'timeline_count': {
'$gt': 500
}}, ['id', 'id_str', 'created_at']):
uid = user['id']
liwc_results = []
indices = []
user_create_time = []
first_tweet_time = []
last_tweet_time = []
fields = []
count = 0
index = 0
tweets = []
for tweet in times.find({'user.id': uid}).sort([("id", 1)]):
if count < n:
tweets.append(tweet)
count += 1
else:
result = liwcp.process_tweet(tweets, Trim_rt=False)
if result:
liwc_results.append([result[k] for k in result.keys()])
if len(fields) == 0:
fields = [k for k in result.keys()]
indices.append(index)
user_create_time.append(
datetime.strptime(user['created_at'],
'%a %b %d %H:%M:%S +0000 %Y'))
first_tweet_time.append(
datetime.strptime(tweets[0]['created_at'],
'%a %b %d %H:%M:%S +0000 %Y'))
last_tweet_time.append(
datetime.strptime(tweets[-1]['created_at'],
'%a %b %d %H:%M:%S +0000 %Y'))
# print index, count
index += 1
count = 0
tweets = []
liwc_results = np.array(liwc_results)
size = len(user_create_time)
user_create_time = np.reshape(user_create_time, (size, 1))
first_tweet_time = np.reshape(first_tweet_time, (size, 1))
last_tweet_time = np.reshape(last_tweet_time, (size, 1))
liwc_results = np.append(liwc_results, user_create_time, axis=1)
liwc_results = np.append(liwc_results, first_tweet_time, axis=1)
liwc_results = np.append(liwc_results, last_tweet_time, axis=1)
print liwc_results.shape
df = pd.DataFrame(
data=liwc_results,
columns=fields +
['user_created_time', 'first_tweet_time', 'last_tweet_time'],
index=indices)
data[user['id_str']] = df
pn = pd.Panel(data)
pn.to_pickle('ed-liwc.panel')
