Python的Pandas库与Oracle Sql操作的一一对应关系

1、where

sql：

select * from temp1
   where position_dt >= 20170704
     and position_dt between 20170701 and 20170704
     and sk_f_prod in ('pd001', 'pd002')
     and secu_type_code not in ('期货', '远期')
     and account_project_name like '证券投资%'
     and (mkt_suspension_info is null or secu_mkt_code = '001')
     and case when prod_name is null then '无名' else prod_name end = '无名'
     )

pandas:

df = df[(df['position_dt'] >= 20170701)
    &(df['position_dt'] <= 20170704)
    &(df.sk_f_prod.isin(['pd001', 'pd002']))
    &(~df.secu_type_code.isin(['期货', '远期']))
    &(df.account_project_name.str.contains('证券投资'))
    &(df.mkt_suspension_info.isnull() | (df.secu_mkt_code == '001'))
    &(df['prod_name'].map((lambda x: x if x else '无名')) == '无名')]

2、group by:求最大值、求和、求数量

sql：

select position_dt,
       sk_f_prod,
       max(locc_position_cost),
       sum(orgc_posi_mkt_val),
       count(case
               when substr(subj_code, 1, 4) = '1102' then  1
               when secu_var = '股票品种' and secu_inner_code like '%600038%' then 0
               else 2
             end) quantity
  from sql_where
 group by position_dt, sk_f_prod

pandas:

df['amount'] = df.apply(lambda x: 1 if x.subj_code[:4] == '1102' else (0 if x.secu_var == '股票品种' and x.secu_inner_code.str.contains('600038') else 2), axis=1)
df = df.groupby(['sk_f_prod', 'position_dt']).agg({'locc_position_cost': 'max', 'orgc_posi_mkt_val': 'sum', 'amount': 'count'})

3、order by, top

sql：

select * from 
(select row_number() over(order by sk_primary_key desc) rn,a.*  from sql_where a) b
where rn <= 3

pandas:

df = df.sort_values('sk_primary_key', ascending=False).head(3)

4、distinct

sql：

select distinct sk_f_prod from sql_where

pandas:

print(df['sk_f_prod'].unique())

5、replace

sql：

select case when subj_class = '证券投资' then '123' else subj_class end subj_class from sql_where a

pandas:

print(df['subj_class'].replace('证券投资', '123'))

6、累加和累乘

sql：

select sk_f_prod, position_dt, locc_position_cost,
 sum(position_quantity) over(order by sk_primary_key) posi_sum
 ,exp(sum(ln(position_quantity/10000)) over(order by sk_primary_key)) posi_prod
 from sql_where

pandas:

 df = df.sort_values('sk_primary_key', ascending=True)['position_quantity'].cumsum()
 df = (df.sort_values('sk_primary_key', ascending=True)['position_quantity']/10000).cumprod()

7、union和union all

sql：

select * from sql_where
where sk_primary_key = '99148311'
union all
select * from sql_where
where sk_primary_key in('99148311', '99153271'

select * from sql_where
where sk_primary_key = '99148311'
union
select * from sql_where
where sk_primary_key in('99148311', '99153271')

pandas:

df1 = df[df['sk_primary_key'] == '99148311']
df2 = df[df['sk_primary_key'].isin(['99148311', '99153271'])]
df = pd.concat([df1, df2])
df = pd.concat([df1, df2]).drop_duplicates()

8、inner join、left join、right join、full join、rename

sql：

--inner join
select a.*, b.* from
(select sk_primary_key from sql_where
where sk_primary_key in ('99148311', '99153271', '99157749')) a
inner join 
(select sk_primary_key from sql_where
where sk_primary_key in ('99148311', '99158720', '99183591')) b
on a.sk_primary_key = b.sk_primary_key

--left join
select a.*, b.* from
(select sk_primary_key from sql_where
where sk_primary_key in ('99148311', '99153271', '99157749')) a
left join 
(select sk_primary_key from sql_where
where sk_primary_key in ('99148311', '99158720', '99183591')) b
on a.sk_primary_key = b.sk_primary_key

--right join
select a.*, b.* from
(select sk_primary_key from sql_where
where sk_primary_key in ('99148311', '99153271', '99157749')) a
right join 
(select sk_primary_key from sql_where
where sk_primary_key in ('99148311', '99158720', '99183591')) b
on a.sk_primary_key = b.sk_primary_key

--full join
select a.*, b.* from
(select sk_primary_key sk_primary_key1, sk_f_prod sk_f_prod_left  from sql_where
where sk_primary_key in ('99148311', '99153271', '99157749')) a
full join 
(select sk_primary_key sk_primary_key2, sk_f_prod sk_f_prod_right from sql_where
where sk_primary_key in ('99148311', '99158720', '99183591')) b
on a.sk_primary_key1 = b.sk_primary_key2

pandas:

   

df1 = df[df['sk_primary_key'].isin(['99148311', '99153271', '99157749'])][['sk_primary_key', 'sk_f_prod']]
    df1 = df1.rename(columns={'sk_primary_key': 'sk_primary_key1'})
    df2 = df[df['sk_primary_key'].isin(['99148311', '99158720', '99183591'])][['sk_primary_key', 'sk_f_prod']]
    df2 = df2.rename(columns={'sk_primary_key': 'sk_primary_key2'})
    # inner join
    # df = pd.merge(left=df1, right=df2, left_on='sk_primary_key1', right_on='sk_primary_key2', how='inner', suffixes={'_left', '_right'})
    # left join
    # df = pd.merge(left=df1, right=df2, left_on='sk_primary_key1', right_on='sk_primary_key2', how='left', suffixes={'_left', '_right'})
    # right join
    # df = pd.merge(left=df1, right=df2, left_on='sk_primary_key1', right_on='sk_primary_key2', how='right', suffixes={'_left', '_right'})
    # outer join
    df = pd.merge(left=df1, right=df2, left_on='sk_primary_key1', right_on='sk_primary_key2', how='outer', suffixes={'_left', '_right'})

9、lead、lag

sql：

select id,
       date_dt,
       nav,
       lead(nav, 1) over(order by trade_dt) lead_nav,
       lag(nav, 1) over(order by trade_dt) nav_nav
  from temp1

pandas:

df = df.sort_values(by='TRADE_DT').reset_index(drop=True)
lag:df['RIGHT_UNIT_NAV'].shift(1), lead:df['RIGHT_UNIT_NAV'].shift(1)