SQL-不使用GROUP BY的情况下使用CASE

Question

How to get the sum of the value without GROUP BY using CASE expression in SQL (or please suggest any other best way to achieve this)

Query :

SELECT 
    CASE
        WHEN [lab_date] <= '2018-03-24'
             AND [lab_date] >= '2018-01-01'
        THEN(ISNULL(SUM(ISNULL([cost], 0) * 5 / 100), 0))
        WHEN [lab_date] >= '2018-03-25'
        THEN(ISNULL((SUM(ISNULL([cost], 0))) - (SUM(ISNULL([cost], 0) * 5 / 100)), 0))
    END AS [tax]
FROM [LABOR];

+-------------+--------+
|  lab_date   |  cost  |
+-------------+--------+
| 2018-03-12  |  75.00 |
| 2018-03-01  | 150.00 |
| 2018-03-11  | 450.00 |
| 2018-03-13  |  37.50 |
| 2018-03-15  | 150.00 |
+-------------+--------+

Getting this with GROUP by lab_date

+-------+
|  tax  |
+-------+
| 15.00 |
| 22.50 |
|  3.75 |
|  1.88 |
+-------+

How we can get the result like below

+-------+
|  tax  |
+-------+
| 43.13 |
+-------+

Answer 1

I think you want conditional aggregation . . . where the case is the argument to SUM() :

SELECT SUM(CASE WHEN lab_date <= '2018-03-24' AND lab_date >= '2018-01-01' THEN cost * 5.0/100
                WHEN lab_date >= '2018-03-25' THEN cost - cost * 5.0/100
                ELSE 0
           END)  as tax         
FROM LABOR ;

Note that SQL Server does integer division. So I don't recommend ever writing 5/100, because that will evaluate to 0 . Even if the multiplication is performed first, not including the decimal point is a bad habit.

Answer 2

Why circumvent the very feature that is exactly designed for your purpose?

Try something like this:

with
taxes_before(lab_date, tax) as (
  select lab_date, cost*0.05
  from labor
  where lab_date between '2018-01-01' and '2018-03-24'
),
taxes_after(lab_date, tax) as (
  select lab_date, cost*0.95
  from labor
  where lab_date >= '2018-01-25'
),
taxes_all(lab_date, tax) as (
  select lab_date, tax from taxes_before
  union select lab_date, tax from taxes_after
)
select sum(tax)
from taxes_all

Disclaimer: I don't have a MS SQL Server here, so shame on me if there is a syntax error.

Explanation: You compute the two sets for the time before day X and the time after day X separately. Then you union those two sets and just group as usual.

Why is this better than the case expression? The WHERE clause benefits from a (possible) index on your date column, but the case statement is evaluated for each row and hence your query scans over every row in your table.