SQL-两个字符串行之间的聚合

Question

I have a following table in which I have employee's transactions having salary as Code 'S' in it:

Id | Code | Amount  | Date
1  | B    | 40      | 2017-01-01
1  | S    | 45000   | 2017-01-02
1  | D    | 30000   | 2017-01-15
1  | B    | 15000   | 2017-01-20
1  | S    | 45000   | 2017-02-02
1  | B    | -20000  | 2017-02-04
1  | B    | -10000  | 2017-02-05

My objective is to see how many days does it take for the employee to drain all his salary

Here is the output I want:

Id | Code | Amount | Month | # days when all salary drains
1  | S    | 45000  | 01    | 20
1  | S    | 45000  | 01    | 0

I have followed this and tried following SQL query:

with cte as
     (
       select *,
        --CASE WHEN Transaction_Code in ('521', '522') then Transaction_Amt else null end as Last_V_ID
          -- find the latest 'V' ID per ArtNo
          max(case when Transaction_Code in ('521', '522') then Transaction_Amt end) 
          over (partition by INTERNAL_ACCT_NUM order by value_date) as Last_V_ID
       from [10.16.42.25].[Cross_Sell_PL].[dbo].[PL_NONPL_TRANS]
    WHERE INTERNAL_ACCT_NUM = '0103PBS8T6001'
    --order by value_date
     )
    select *,
     case when Transaction_Code in ('521', '522') then Transaction_Amt else
      ( lag(Last_V_ID,1,0) OVER (PARTITION BY INTERNAL_ACCT_NUM ORDER BY VALUE_DATE) ) 
      + Transaction_Amt end as running_balance,

       sum(case when Transaction_Amt < Last_V_ID then null else Transaction_Amt end)
       over (partition by INTERNAL_ACCT_NUM order by VALUE_DATE rows unbounded preceding)
    from cte
    order by Internal_Acct_num, value_date

The problem is I am not able to restrict the calculation between the code and also unable to calculate days.

Answer 1

Revised my answer based on your feedback.

The query now uses a common table expression to build a derived table for the salary amount and the dates per month on which they are payed, followed by another derived table that compiles a list of all of the transactions summed per calendar month.

The summed per month transactions are compared with the salaried amount and only where there is an amount that crosses the salary threshold do the summed days show (+1 is included as this includes the day the salary is paid as day 1).

The query as such shows 3 rows as the sample data provided includes what would be Decembers wage month, however there is no salaried amount to compare to from December hence the extra 0. I hope this helps.

declare @employee_transaction table (
    Id int,
    Code nvarchar(1),
    Amount int,
    Date Date
);

insert into @employee_transaction (id, Code, Amount, Date)
values
(1  , 'B'    , 40      , '2017-01-01'),
(1  , 'S'    , 45000   , '2017-01-02'),
(1  , 'D'    , 30000   , '2017-01-15'),
(1  , 'B'    , 15000   , '2017-01-20'),
(1  , 'S'    , 45000   , '2017-02-02'),
(1  , 'B'    , -20000  , '2017-02-04'),
(1  , 'B'    , -10000  , '2017-02-05');


with Salaries as (Select INTERNAL_ACCT_NUM, Transaction_Code, Value_Date, 
Transaction_Amt as Salary_Per_Month from #test1 e where 
Transaction_Code='S'), 
Deductions as (select sum(case when Transaction_Amt<0 then Transaction_Amt 
else 0 end) as Amount, Max(Value_Date) as Drain_date from #TEST1 e2 
where Transaction_Code <>'S' and Value_Date>(select Value_Date from #TEST1 
e3 where Transaction_Code='S' and 
DATEPART(month,e2.Value_Date)=DATEPART(month,e3.Value_Date) 
and e3.Internal_Acct_Num=e2.Internal_Acct_Num) 
group by DATEPART(month,e2.Value_Date)) 

select distinct s.INTERNAL_ACCT_NUM, s.Transaction_Code, s.Salary_Per_Month, 
DATEPART(Month,e.Value_Date) as Month, 
coalesce(DATEDIFF(Day,S.Value_Date,(select d.Drain_Date from Deductions d 
where d.Amount+S.Salary_Per_Month<=0 and 
DatePart(month,d.Drain_Date)=DatePart(month,e.Value_Date) and 
d.drain_date>=s.value_date))+1,0) 
from #TEST1 e 
inner join Salaries s on e.Transaction_Code=s.Transaction_Code 
and e.INTERNAL_ACCT_NUM=s.INTERNAL_ACCT_NUM and s.Value_Date>=e.Value_Date 
and DATEPART(month, s.Value_Date)=DATEPART(month, e.Value_Date) 
order by Month;