Pivot vs Case T-sql efficiency

问题:

I am trying to figure out which way is more efficient. I tried to use Sql Fiddle but it didn't seem to recognize PIVOT or UNION What tools or statistics can I look at to determine which is more efficient.

CREATE TABLE T
(
  PersonNum INT
 ,WeekOf DATETIME
 ,ActivityType1 INT
 ,ActivityType2 INT
 ,Hours INT
)

INSERT INTO T VALUES(1,'11/2/2014',5,0,40)
INSERT INTO T VALUES(1,'11/2/2014',1,0,5)
INSERT INTO T VALUES(1,'11/2/2014',1,0,8)
INSERT INTO T VALUES(1,'11/2/2014',2,1,6)
INSERT INTO T VALUES(1,'11/2/2014',2,2,2)
INSERT INTO T VALUES(1,'11/2/2014',2,3,9)
INSERT INTO T VALUES(1,'11/2/2014',2,4,7)

INSERT INTO T VALUES(1,'11/9/2014',5,0,40)
INSERT INTO T VALUES(1,'11/9/2014',1,0,2)
INSERT INTO T VALUES(1,'11/9/2014',1,0,6)
INSERT INTO T VALUES(1,'11/9/2014',2,1,7)
INSERT INTO T VALUES(1,'11/9/2014',2,2,2)
INSERT INTO T VALUES(1,'11/9/2014',2,3,3)
INSERT INTO T VALUES(1,'11/9/2014',2,4,5)

Method 1

SELECT 
  PersonNum
, WeekOf
, SUM(CASE WHEN ActivityType1 = 5 THEN Hours ELSE 0 END) AS Beginning
, SUM(CASE WHEN ActivityType1 = 1 THEN Hours ELSE 0 END) AS Plus
, SUM(CASE WHEN ActivityType1 = 2 AND ActivityType2 <> 3 THEN Hours ELSE 0 END) AS Minus
, SUM(CASE WHEN ActivityType1 = 2 AND ActivityType2 = 3 THEN Hours ELSE 0 END) AS MinusSpecial
FROM T
GROUP BY 
  PersonNum
, WeekOf

Method 2

SELECT
  PersonNum
, WeekOf
, [Beginning]
, [Plus]
, [Minus]
, [Minus Special]
FROM
(
  SELECT
    PersonNum
  , WeekOf
  , 'Beginning' AS ColumnType
  , Hours
  FROM T
  WHERE
    ActivityType1 = 5 
  UNION
  SELECT
    PersonNum
  , WeekOf
  , 'Plus' AS ColumnType
  , Hours
  FROM T
  WHERE
    ActivityType1 = 1 
  UNION
  SELECT
    PersonNum
  , WeekOf
  , 'Minus' AS ColumnType
  , Hours
  FROM T
  WHERE
      ActivityType1 = 2 
  AND ActivityType2 <> 3
   UNION
  SELECT
    PersonNum
  , WeekOf
  , 'Minus Special' AS ColumnType
  , Hours
  FROM T
  WHERE
      ActivityType1 = 2 
  AND ActivityType2 = 3
    ) Data  
PIVOT 
(SUM(Hours)
FOR ColumnType IN ([Beginning]                
                  ,[Plus]
                  ,[Minus]
                  ,[Minus Special])
) pvt 

回答1:

You can do this

set statistics time on

--first query

set statistics time off

set statistics time on

--second query

set statistics time off

You can look in the Messages window to see the execution times

回答2:

May I suggest a rewrite of your PIVOT query so you don't have to hit the table 5 times.

select P.PersonNum,
       P.WeekOf,
       P.Beginning,
       P.Plus,
       P.Minus,
       P.MinusSpecial
from (
     select T.PersonNum,
            T.WeekOf,
            T.Hours,
            case ActivityType1 
             when 5 then 'Beginning'  
             when 1 then 'Plus'
             when 2 then case ActivityType2
                           when 3 then 'MinusSpecial'
                           else 'Minus'
                         end
            end as ColumnType
     from T
     ) as T
pivot (sum(T.Hours) for T.ColumnType in (Beginning, Plus, MinusSpecial, Minus)) as P

With that you will probably get the same query plan as your group by query and about the same performance.

回答3:

The method 1 will be faster for sure . This is because there is only 1 select statement. From the other side there are 5 select statement in Method 2. The general rule is that the more statements query contains the longer it will take to execute.