如何在SQL中编写select语句，以从提供的集合中返回以下结果

Question

Firstly, apologies for the vagueness of my title, I will edit it once I have a better understanding of what I need to do. Hopefully the question is a bit more revealing.

Consider a set that looks like this;

Fruit | Quantity | Month
-------------------------
Apple |    5     | Jan
Pear  |    2     | May
Orange|    1     | Jan
Apple |    10    | May
Pear  |    2     | May
Orange|    1     | Jul
Apple |    5     | Jan
Pear  |    2     | Jul
Orange|    1     | Jul

I would like a select statement that returns this;

Fruit | Jan % | May % | Jul % | Total %
---------------------------------------
Apple | 90.9  | 71.43 | 0     | 68.97
Pear  | 0     | 28.57 | 50    | 20.69
Orange| 9.1   | 0     | 50    | 10.34

The values in the results are the percentages of each fruit for each month and the total for all months. Eg the first result (90.9%) is stating that 90.9% of all fruit for January were apples ((Total apples for Jan / Total fruit for Jan) * 100)

I have successfully written a select statement that returns the first and last columns, the trouble I'm having is with the middle three columns and how to get the percentages of each fruit for each month.

Obviously this is a fake set that I contrived to try to simplify my problem, in the real data "month" is actually a DateTime and can be any date. I do not care what order the fruit appear in the results, nor am I worried about (small) rounding errors in the percentages.

Answer 1

The window functions in concert with aa conditional aggregation

Declare @YourTable table (Fruit varchar(25),Quantity  int, Month varchar(25))
Insert Into @YourTable values
('Apple',5,'Jan'),
('Pear',2,'May'),
('Orange',1,'Jan'),
('Apple',10,'May'),
('Pear',2,'May'),
('Orange',1,'Jul'),
('Apple',5,'Jan'),
('Pear',2,'Jul'),
('Orange',1,'Jul')

Select Fruit
      ,Jan   = max(case when Month='Jan' then value else 0 end)
      ,May   = max(case when Month='May' then value else 0 end)
      ,Jul   = max(case when Month='Jul' then value else 0 end)
      ,Total = max(Total)
 From  (
            Select Distinct Fruit
                  ,Month
                  ,Value = sum(Quantity*100.0) over (Partition By Fruit,Month) / sum(Quantity) over (Partition By Month)
                  ,Total = sum(Quantity*100.0) over (Partition By Fruit) / sum(Quantity) over ()
             From  @YourTable
        ) A
 Group By Fruit

Returns

Fruit   Jan         May         Jul         Total
Apple   90.909090   71.428571   0.000000    68.965517
Orange  9.090909    0.000000    50.000000   10.344827
Pear    0.000000    28.571428   50.000000   20.689655

Answer 2

This can be done easier with versions from 2012, but you are using 2008. Hence you can make use of the below query which seems to be quiet complicated

WITH totalData AS 
    (SELECT Month, SUM(Quantity) totalpermonth 
    FROM fruitsData 
    GROUP BY Month), -- Calculating the total fruits per month

totalFriutData as 
    (SELECT Sum(Quantity) totalFruit 
    FROM fruitsData) -- Calculating the total fruits


SELECT Fruit, Jan, May, Jul, CAST(totalperfruit*100 AS decimal)/tf.totalFruit as Total
FROM(
    SELECT Fruit, SUM(CASE WHEN Month = 'Jan' THEN percentage_value ELSE 0 END) Jan
    , SUM(CASE WHEN Month = 'May' THEN percentage_value ELSE 0 END) May
    , SUM(CASE WHEN Month = 'Jul' THEN percentage_value ELSE 0 END) Jul
    , SUM(totalperfruit) totalperfruit
    FROM(
        SELECT Fruit, CAST(SUM(Quantity)*100 AS decimal)/td.totalpermonth percentage_value, fd.Month, td.totalpermonth, SUM(Quantity) totalperfruit
        FROM fruitsData fd
        LEFT JOIN totalData td on fd.Month = td.Month
        GROUP BY Fruit, fd.Month, td.totalpermonth)t1
    GROUP BY Fruit
) t1
CROSS JOIN totalFriutData tf;

Below is the output I got,

Hope this would help you out.

Answer 3

Conditional sums might be your answer. You still have the problem of figuring out which months to process, but that's not too bad.

SELECT
    SUM(CASE WHEN MONTH(SalesDate) = 1 THEN 1 ELSE 0 END)*100.0/(SELECT COUNT(*) FROM SALES WHERE SalesDate BETWEEN @Start and @End) as JAN
    SUM(CASE WHEN MONTH(SalesDate) = 2 THEN 1 ELSE 0 END)*100.0/(SELECT COUNT(*) FROM SALES WHERE SalesDate BETWEEN @Start and @End) as FEB
FROM
    FruitSales

Answer 4

You can sum quantities per month, add total and then pivot them in line

;with
m as (
    select fruit, Quantity, DATEPART(mm,dt) mm from d
),
tt as (
    select sum(Quantity)*1.0 TotalQt from m
),
tf as (
    select fruit, sum(Quantity)*1.0 TotalQtFr from m group by fruit
),

tm as (
    select mm, sum(Quantity)*1.0 TotalQtMM
    from m 
    group by mm
),
d2 as (
    select m.Fruit, m.mm, cast(100*Quantity/tm.TotalQtMM as decimal(5,2)) QuantityPerc
    from m
    join tm on m.mm=tm.mm
    union all
    select tf.Fruit, 13, cast(100*TotalQtFr/TotalQt as decimal(5,2)) QuantityPerc
    from tf 
    cross join tt
),
p as (
    select *
    from d2
    pivot (sum(QuantityPerc) for mm in ([1],[2],[3],[4],[5],[6],[7],[8],[9],[10],[11],[12],[13])) p
)
-- select * -- to see all months
select [Fruit], Isnull([1], 0) [Jan %], Isnull([5], 0) [May %], Isnull([7], 0) [Jul %], Isnull([13], 0) [Total %]
from p
order by 1

The output:

Fruit   Jan %   May %   Jul %   Total %
Apple   90.90   71.43   0.00    68.97
Orange  9.09    0.00    50.00   10.34
Pear    0.00    28.58   50.00   20.69

Answer 5

Although there is already an accepted answer, I wanted to experiment with rollup/cube in order to find an elegant way of pre-aggregating the required sums per month, fruit and in total:

with groupedFruits as
      (select f.fruit, f.month, cast(CASE WHEN sum(quantity)=0 THEN NULL else sum(quantity) END as decimal) as quantity
      from fruits f
      group by f.month, f.fruit with cube)
select f2.fruit,
  (select g.quantity from groupedFruits g where f2.fruit = g.fruit and g.month='Jan') /
  (select g.quantity from groupedFruits g where g.fruit is null and g.month='Jan') as jan,
  (select g.quantity from groupedFruits g where f2.fruit = g.fruit and g.month='May') /
  (select g.quantity from groupedFruits g where g.fruit is null and g.month='May') as may, 
  (select g.quantity from groupedFruits g where f2.fruit = g.fruit and g.month='Jul') /
  (select g.quantity from groupedFruits g where g.fruit is null and g.month='Jul') as jul,
  (select g.quantity from groupedFruits g where f2.fruit = g.fruit and g.month='Aug') /
  (select g.quantity from groupedFruits g where g.fruit is null and g.month='Aug') as aug,
  (select g.quantity from groupedFruits g where f2.fruit = g.fruit and g.month='Sep') /
  (select g.quantity from groupedFruits g where g.fruit is null and g.month='Sep') as sep,
  sum(f2.quantity) / (select g.quantity from groupedFruits g where g.fruit is null and g.month is null) as total
from fruits f2
group by f2.fruit

The query can also handle data sets where quantity entries are 0 either for a fruit in a specific month or for a complete month, eg when entering tuples like ('Orange', 0, 'Aug') and ('Kiwi', 0, 'Sep') , which then yields:

fruit   |jan    |may    |jul    |aug    |sep    |total
--------|-------|-------|-------|-------|-------|-----
Apple   |0,91   |0,71   |NULL   |NULL   |NULL   |0,69
Kiwi    |NULL   |NULL   |NULL   |NULL   |NULL   |0,00
Orange  |0,09   |NULL   |0,50   |NULL   |NULL   |0,10
Pear    |NULL   |0,29   |0,50   |NULL   |NULL   |0,21