问题
I have a table that roughly looks like this
Transaction Date Gain/Loss
15464 7/31/2018 $500
15464 6/30/2018 -$200
59872 7/31/2018 $1000
59872 6/30/2018 $2500
How can i add another column that calculates the difference between the current month's gain/loss and the previous month's for each specific transaction number? i.e.
Transaction Date Gain/Loss Change in Gain/Loss
15464 7/31/2018 $500 $700
15464 6/30/2018 -$200 -$200
59872 7/31/2018 $1000 -$1500
59872 6/30/2018 $2500 $2500
回答1:
I think you can use a correlated subquery:
select t.*,
(select top (1)
from t as t2
where t2.transaction = t.transaction and t2.date < t.date
order by t2.date desc
) as prev_gain_loss
from t;
Calculating the difference is just arithmetic.
回答2:
Using your table as a base - and going on the assumption that each month records a single transaction on the last day of the month:
SELECT T1.Transaction
, T1.[Date]
, T1.[Gain/Loss]
, T2.[Date]
, T1.[Gain/Loss]-T2.[Gain/Loss] AS [Change in Gain/Loss]
FROM MyTable T1 LEFT JOIN MyTable T2 ON T1.Transaction = T2.Transaction AND
T1.[Date] = DATESERIAL(YEAR(T2.[Date]), MONTH(T2.[Date])+2,0)
Consider that you have two tables - one displaying the figures for this month, and one displaying the figures for last month.
To join the two tables together you'd look at the Transaction number as the common attribute between the two tables.
The fact that both tables are the same table is irrelevant to Access - as long as it knows which is which by aliasing the table names as T1 and T2.
FROM MyTable T1 LEFT JOIN MyTable T2 ON T1.Transaction = T2.Transaction
This would give too many records though. The transaction for July would join to itself in the second table, and to the previous month. The previous month transaction would also join to itself and the current month transaction. This results in four records per transaction.
| Transaction | T1.Date | Gain/Loss | T2.Date | Change in Gain/Loss |
|-------------|------------|-----------|------------|---------------------|
| 15464 | 31/07/2018 | £500.00 | 30/06/2018 | £700.00 |
| 15464 | 31/07/2018 | £500.00 | 31/07/2018 | £0.00 |
| 15464 | 30/06/2018 | -£200.00 | 30/06/2018 | £0.00 |
| 15464 | 30/06/2018 | -£200.00 | 31/07/2018 | -£700.00 |
| 59872 | 31/07/2018 | £1,000.00 | 30/06/2018 | -£1,500.00 |
| 59872 | 31/07/2018 | £1,000.00 | 31/07/2018 | £0.00 |
| 59872 | 30/06/2018 | £2,500.00 | 30/06/2018 | £0.00 |
| 59872 | 30/06/2018 | £2,500.00 | 31/07/2018 | £1,500.00 |
To only perform the calculation on the previous month figure we need to join the current month to the previous month.
T1.[Date] = DATESERIAL(YEAR(T2.[Date]), MONTH(T2.[Date])+2,0)
The DATESERIAL calculation in this join adds two months to the date, it then returns day zero of the second month which is equivalent to the last day of the month before that. (Hope that made sense).
| Transaction | Date | Gain/Loss | Change in Gain/Loss |
|-------------|------------|-----------|---------------------|
| 15464 | 31/07/2018 | £500.00 | £700.00 |
| 15464 | 30/06/2018 | -£200.00 | |
| 59872 | 31/07/2018 | £1,000.00 | -£1,500.00 |
| 59872 | 30/06/2018 | £2,500.00 | |
If you must have figures in the starting month then you can use NZ to turn the NULL values to 0.
, T1.[Gain/Loss]-NZ(T2.[Gain/Loss],0) AS [Change in Gain/Loss]
| Transaction | Date | Gain/Loss | Change in Gain/Loss |
|-------------|------------|-----------|---------------------|
| 15464 | 31/07/2018 | £500.00 | £700.00 |
| 15464 | 30/06/2018 | -£200.00 | -£200.00 |
| 59872 | 31/07/2018 | £1,000.00 | -£1,500.00 |
| 59872 | 30/06/2018 | £2,500.00 | £2,500.00 |
来源：https://stackoverflow.com/questions/51986917/get-previous-months-data-alongside-current-month