Unpivot several columns and group by

Question

I have a table that looks like below. The values are number of contracts each customer have each month.

CustomerNo |  CustomerType |  201901  |  201902  |  201903  |
-----------+---------------+----------+----------+----------|
1          |  Private      |    1     |    0     |    0     |
2          |  Private      |    0     |    1     |    1     |
3          |  Enterprise   |    1     |    1     |    0     |
4          |  Enterprise   |    1     |    1     |    1     |

I want to unpivot the YearMonth columns and group by CustomerType, YearMonth so the table looks like this:

CustomerType | YearMonth | Contracts | 
-------------+-----------+-----------|
Private      |  201901   |     1     |
Enterprise   |  201901   |     2     |
Private      |  201902   |     1     |
Enterprise   |  201902   |     2     |
Private      |  201903   |     1     |
Enterprise   |  201903   |     1     |

How

Answer 1

I would just use apply :

select v.*
from t cross apply
     (values (t.customerType, 201901, t.[201901]),
             (t.customerType, 201902, t.[201902]),
             (t.customerType, 201903, t.[201903])
     ) v(CustomerType, YearMonth, Contracts)
where Contracts > 0;

I much prefer apply because this is a good introduction to lateral joins , very handy functionality in SQL databases.

The alternative of unpivot is bespoke syntax and it only does one thing (and in my opinion, not very elegantly either).

Answer 2

For MS SQL Server:

SELECT CustomerType, YearMonth, SUM(Contracts) Contracts 
FROM (
  SELECT CustomerNo, CustomerType, [201901], [201902], [201903]
  FROM Customers
) pvt
UNPIVOT(Contracts FOR YearMonth IN([201901], [201902], [201903])) unpvt
GROUP BY CustomerType, YearMonth;

Result:

CustomerType    YearMonth   Contracts
Enterprise  201901  2
Private     201901  1
Enterprise  201902  2
Private     201902  1
Enterprise  201903  1
Private     201903  1

Link to db<>fiddle

MSDN documentation: UNPIVOT