pandas

注：import pandas as pd

1、读写

df = pd.read_table('******', sep='\t', encoding="UTF-8", names= [**,**,**])

df.to_csv('***', header=True,mode ='a' ,sep=',',index=False)

df.ix[:5, :10]   #查看数据
df[['num']] = df[['num']].astype(float)  #数据类型

2、筛选
#数字
df1 = df[df['order_s'] <5.0]

#字符串
df.loc[df['column_name'].isin(some_values)]

3、转换

df = pd.DataFrame(数组)

df = pd.DataFrame({'a': [1,2,3,4,5], 'b':[5,4,3,2,1]})

df['appid'].values #转为arrary

df['appid'].values.tolist() #转为list

5、nan缺失值
df.column1 = df.column1.fillna('')
df1 = df.replace(np.nan,' ', regex=True)

6、运算

df['visit_days'] = df.apply(lambda row:float(row['visit_days'])/max(float(row['pay_days']),1),axis =1)
df['visit_days'] = data_ft['visit_days'].map(lambda x: float(x) if is_number(x) else 0)

7、合并

#concat 
result = pd.concat([df1, df4], axis=1)  #合并列
axis = 1 合并列，=0    合并行

#merge
data_user = pd.merge(data_user,data_judge,on = ['member_id'],how = 'left')

8、其它

排序
df = df.sort_values(['appid', 'date'], ascending=[True,False]) 

聚合
out = df.groupby(['city_name','city_code','type','state'], as_index=False).\
        aggregate({'member_id': pd.Series.count}).\
         rename(columns={'member_id': '每周期用户数','state':'是否超级会员'})