to frame cost effective query - sql-server-2005

Question

Could you please help me out in building a Query. I Have a table as below
Id Info_Id Type
1 2 2
2 6 2
3 5 3
4 8 3
5 2 3
6 2 2
7 5 2
8 8 2
9 5 2
10 8 2
11 8 2
12 5 3
13 6 3
14 8 3
a query need to be framed so as to group by "Info_Id".
I need output as below for eg:
Info_Id CountOfRec Type2 Type3
2 3 2 1
5 4 2 2
6 2 1 1
8 5 3 2
I Tried as below but I m not able to get the efficient output
select Info_Id, count(Id)as CountOfRec,
(select count(Id)from tbl_TypeInfo where Info_Id = 5 AND Type = 2) as Type2,
(select count(Id)from tbl_ TypeInfo where Info_Id = 5 AND Type = 3) as Type3
from tbl_TypeInfo
where Info_Id = 5
group by Info_Id
output was this ,
Info_Id CountOfRec Type2 Type3
5 4 2 2
(I have to loop for each “Info_id” to get desired OP, there is thousand of records and its time consuming)
I wanted to get the highlighted output from the table. The query I have framed is not efficient and there would be good solution for this can you help me out.

Answer 1

You can use the CASE expression to only count the rows for a specific type:
SELECT Info_Id,
COUNT(*) AS CountOfRec,
COUNT(CASE WHEN Type = 2 THEN 1 ELSE NULL END) AS Type2
COUNT(CASE WHEN Type = 3 THEN 1 ELSE NULL END) AS Type3
FROM tbl_TypeInfo
GROUP BY Info_Id
Add a WHERE Info_Id = 5 to retrieve results for a specific ID only.
Update: as per comments, if you do not store a table of ID's, You need to change your IN(..) list to a virtual "table":
SELECT vt.id,
COUNT(*) AS CountOfRec,
COUNT(CASE WHEN Type = 2 THEN 1 ELSE NULL END) AS Type2,
COUNT(CASE WHEN Type = 3 THEN 1 ELSE NULL END) AS Type3
FROM (
SELECT 1 id
UNION SELECT 2
UNION SELECT 3
UNION SELECT 5
UNION SELECT 8
) AS vt LEFT JOIN tbl_TypeInfo ON vt.id = tbl_TypeInfo.Info_Id
GROUP BY vt.id

Answer 2

You can use SQL Server's PIVOT operator
SELECT Info_ID
, CountOfRec = [2] + [3]
, Type2 = [2]
, Type3 = [3]
FROM (
SELECT *
FROM (
SELECT *
FROM tbl_TypeInfo
) s
PIVOT (COUNT(Id) FOR Type IN ([2], [3])) pvt
) q
Test
;WITH tbl_TypeInfo AS (
SELECT [Id] = 1, [Info_Id] = 2, [Type] = 2
UNION ALL SELECT 2, 6, 2
UNION ALL SELECT 3, 5, 3
UNION ALL SELECT 4, 8, 3
UNION ALL SELECT 5, 2, 3
UNION ALL SELECT 6, 2, 2
UNION ALL SELECT 7, 5, 2
UNION ALL SELECT 8, 8, 2
UNION ALL SELECT 9, 5, 2
UNION ALL SELECT 1, 8, 2
UNION ALL SELECT 1, 8, 2
UNION ALL SELECT 1, 5, 3
UNION ALL SELECT 1, 6, 3
UNION ALL SELECT 1, 8, 3
)
SELECT Info_ID
, CountOfRec = [2] + [3]
, Type2 = [2]
, Type3 = [3]
FROM (
SELECT *
FROM (
SELECT *
FROM tbl_TypeInfo
) s
PIVOT (COUNT(Id) FOR Type IN ([2], [3])) pvt
) q

Answer 3

This is a bit crazy, but if the types are always, always 2 and 3, this could be threated as an equation, where count(type2)+count(type3)=count(*) and 2*count(type2)+3*count(type3)=sum(*) so you could to something like
SELECT 3*c-s as Type2Count, s-2*c as Type3Count
FROM (SELECT COUNT(*) as C, SUM(Type) as S
FROM tbl_TypeInfo
WHERE Info_Id = 5) SourceTable
This will be lightning fast, however, this is extemely breakable!!!!
If ever, the types are changed, or a type is added, this will not work.