pandas (一)

1. 修改列名的三种常见方法

1. columns属性

# ①暴力
df.columns = ['a', 'b', 'c', 'd', 'e']

# ②修改
df.columns = df.columns.str.strip('$')

# ③修改
df.columns = df.columns.map(lambda x:x[1:])

2. rename方法、columns参数

# ④暴力（好处：也可只修改特定的列）
df.rename(columns=('$a': 'a', '$b': 'b', '$c': 'c', '$d': 'd', '$e': 'e'}, inplace=True) 

# ⑤修改
df.rename(columns=lambda x:x.replace('$',''), inplace=True)

2. 筛选特定值（删除），范围内的值

df[(True-df['appPlatform'].isin([2]))]
# 过滤掉某个范围的值
df[df['creativeID']<=10000]

3. 日期格式数据转换

数据格式： 1990/9/26 这种 ， 结合前面time那篇有如下处理转为时间戳

df_read['birth_date'] = df_read['birth_date'].astype(str)
df_read = df_read[~df_read['birth_date'].isin(['0003-11-21'])]   # 选取不是 0003-11-21 这个的异常数据
# df1=df1[df1['A'].isin([xxx])]  选取是的值
df_read['timestamp'] = df_read['birth_date'].map(lambda x:arrow.get(x, "YYYY/M/D").timestamp*1000)df_read['timestamp'] = df_read['birth_date'].map(lambda x:arrow.get(x, "YYYY/M/D").timestamp*1000)
df2=df_read.drop(["birth_date"],axis=1)

参考： https://www.cnblogs.com/hhh5460/tag/pandas/

4. 写入csv

df2.to_csv('nfis_data.csv',index=False)  # 不写索引
dt.to_csv(‘C:/Users/think/Desktop/Result.csv‘,sep=‘?‘)#使用?分隔需要保存的数据，如果不写，默认是,

dt.to_csv(‘C:/Users/think/Desktop/Result1.csv‘,na_rep=‘NA‘) #确实值保存为NA，如果不写，默认是空

dt.to_csv(‘C:/Users/think/Desktop/Result1.csv‘,float_format=‘%.2f‘) #保留两位小数

dt.to_csv(‘C:/Users/think/Desktop/Result.csv‘,columns=[‘name‘]) #保存索引列和name列

dt.to_csv(‘C:/Users/think/Desktop/Result.csv‘,header=0) #不保存列名

dt.to_csv(‘C:/Users/think/Desktop/Result1.csv‘,index=0) #不保存行索引
df = data.drop_duplicates()   # 去重

5. python csv 模块

# 读
csvFile = open("foo.csv", "r")
reader = csv.reader(csvFile)
next(reader)
for i, tt in enumerate(reader):
    print(i, tt)
# 写
columns = ["uid", "application_id", "pic_url", "ic_no", "name", "gender", "birthday"]
with open('foo_02.csv', 'w', newline='') as outcsv:
    writer = csv.writer(outcsv, delimiter=',', quotechar='"', quoting=csv.QUOTE_ALL,
                        lineterminator='
')
    writer.writerow(columns)
...
new_row = (uid, d.id, pic_url, ic_no, name, gender, birthday)
writer.writerow(new_row)        

6. 拆分表格(大json拆分为一个个小字段)

# 整个流程
for i in range(df_read.shape[0]):
# for i in range(50):
    try:
        one_line = df_read[i:i+1]
        uid = one_line['uid'][i]   #
        one_line_json = list(one_line['data_json'])[0]
        py_obj = json.loads(one_line_json)
#         print(py_obj)
        adv_list = py_obj['result'][0]['adv']
        params = py_obj['result'][0]['param']
        bday = params['bday']  # 
        name = params['name']  # 

        if len(adv_list)>0:   # if adv_list
            for u in adv_list:
                report = u['reported']
                advdate = u['advdate']
                amount = u['secuamt']
                advtype = u['advtype']
                loantype = u['loantype']
                bank = u['bank']
                data_list = [uid, bday, name, report, advdate, amount, advtype, loantype, bank]
                df_save = pd.DataFrame(data_list)
                m = df_save.T
                m.columns = ['uid', 'bday', 'name', 'report', 'advdate', 'amount', 'advtype', 'loantype', 'bank']
                m.to_csv('test3.csv',mode='a',header=False)
    except:
        continue