SQL: How to create new columns depending on other column's value in the same table

Question

I have 2 tables, Main and Units.

Main table:

Todate   Unit   CategoryID   Quantity
1/7/2012  1        S            300
1/7/2012  1        U            350
2/7/2012  2        S            220
3/7/2012  2        S             50
3/7/2012  2        U            330
4/7/2012  1        S            200
4/7/2012  1        U            180

S = Sales, U = Upgrades

Units table:

UnitNum   UnitName
1         Measures
2         Performance

I need to get this result:

Todate   UnitNum   UnitName    Sales    Upgrades
1/7/2012    1      Measures     300       350
2/7/2012    2      Performance  220
3/7/2012    2      Performance   50       330
4/7/2012    1      Measures     200       180

Meaning i need to create 2 columns - sales and upgrades, depending on the value in CategoryID, and i need them to be in the same row. What i have so far is this

select Todate, Main.Unit, UnitName,
case when CategoryID = 'S' then Quantity end as Sales,
case when CategoryID = 'U' then Quantity end as Upgrades
from Main join Units on Main.UnitNum = Units.UnitNum
group by  Todate, Main.Unit, UnitName

It gives me 2 new columns but they are in two separate rows..

I would really appreciate any help resolving this! Thank you

Answer 1

You just need an aggregate query around the case statements.

select m.todate
     , m.unit
     , u.unitname
     , sum(case when m.categoryid = 'S' then quantity end ) as sales
     , sum(case when m.categoryid = 'U' then quantity end ) as upgrages
  from main m
  join units u
    on m.unit = u.unitnum
 group by m.todate, m.unit, u.unitname

Answer 2

You need something like this:

SELECT
    Todate, m.Unit, UnitName,
    Sales = (SELECT SUM(Quantity)
             FROM dbo.Main m2 
             WHERE m.Todate = m2.Todate AND CategoryID = 'S'),
    Updates = (SELECT SUM(Quantity)
               FROM dbo.Main m2 
               WHERE m.Todate = m2.Todate AND CategoryID = 'U')
FROM
    dbo.Main m 
INNER join 
    dbo.Units u on m.Unit = u.UnitNum
GROUP BY  
    Todate, m.Unit, UnitName
ORDER BY  
    Todate, m.Unit, UnitName

This seems to be returning the output you're looking for:

Answer 3

You need to do a self-join of your main table with itself:

 SELECT m1.todate, m1.unit AS unitnum, u.unitname,
        SUM(m1.quantity) AS sales, SUM(m2.quantity) AS upgrades
   FROM (SELECT todate, unit, quantity FROM Main WHERE category = 'S') AS m1
   FULL OUTER JOIN
        (SELECT todate, unit, quantity FROM Main WHERE category = 'U') AS m2
     ON m1.todate = m2.todate AND m1.unit = m2.unit
   JOIN units AS u
     ON m1.unit = u.unitnum
  GROUP BY m1.todate, m1.unit, u.unitname
  ORDER BY m1.todate, m1.unit, u.unitname;

There are many other equivalent ways of writing the same query; this one preserves the symmetry of the problem fairly well, though. ( Updated to use a FULL OUTER JOIN between the m1 and m2 sub-queries, to deal with cases where there are sales with no upgrades or upgrades with no sales. )

Answer 4

You could use PIVOT for this too:

WITH data AS (
  SELECT
    m.Todate,
    m.UnitNum,
    u.UnitName,
    Category = CASE m.CategoryID
                 WHEN 'S' THEN 'Sales'
                 WHEN 'U' THEN 'Upgrades'
               END,
    Quantity
  FROM Main m
  INNER JOIN Units u
  ON m.UnitNum = u.UnitNum
)
SELECT
  Todate,
  UnitNum,
  UnitName,
  Sales,
  Upgrades
FROM data
PIVOT (
  SUM(Quantity) FOR Category IN (Sales, Upgrades)
) p
;