TSQL Pivoting multiple columns - sql

Question

Code:
DECLARE #Employee TABLE
(
[Employee_Id] INT IDENTITY(1, 1)
, [Code] NVARCHAR(10)
) ;
INSERT INTO #Employee
VALUES ( N'E1' ), ( N'E2' ), ( N'E3' ) ;
DECLARE #Contact TABLE
(
[Employee_Id] INT
, [PhoneType] CHAR(1)
, [PhoneNumber] VARCHAR(20)
, [IsMainNumber] BIT
) ;
INSERT INTO #Contact
VALUES (1, 'M', '1234567890', 1), (1, 'H', '1234567891', 0),
(1, 'M', '1234567892', 0), (1, 'B', '1234567893', 0),
(2, 'M', '2234567890', 0), (2, 'H', '2234567891', 1),
(2, 'B', '2234567892', 0), (2, 'M', '2234567893', 0),
(3, 'M', '3234567890', 0), (3, 'H', '3234567891', 0),
(3, 'M', '3234567892', 0), (3, 'B', '3234567893', 1);
SELECT
[E].[Employee_Id],
[E].[Code],
[COA].[MainPhoneNumber],
[COA].[NonMainNumber]
FROM
#Employee AS [E]
OUTER APPLY
(SELECT
MAX (IIF([C].[IsMainNumber] = 1, [C].[PhoneNumber], NULL)) [MainPhoneNumber],
MAX (IIF([C].[IsMainNumber] = 0, [C].[PhoneNumber], NULL)) [NonMainNumber]
FROM
#Contact AS [C]
WHERE
[E].[Employee_Id] = [C].[Employee_Id]
GROUP BY
[C].[Employee_Id]) AS [COA] ;
Current output
Employee_Id Code MainPhoneNumber NonMainNumber
1 E1 1234567890 1234567893
2 E2 2234567891 2234567893
3 E3 3234567893 3234567892
Goal
I need to return the MAX main phone number and its phone type and MAX non-main phone number and its phone type. I'm able to get the MAX main/non-main phone numbers, but need to somehow get their phone types. I don't want to make two additional joins based on Employee_Id and PhoneNumber and get the type, because original table is huge and that would slow things down a lot. Trying to figure out an alternative that performs well.
Desired Output
Employee_Id Code MainPhoneType MainPhoneNumber NonMainPhoneType NonMainNumber
1 E1 M 1234567890 B 1234567893
2 E2 H 2234567891 M 2234567893
3 E3 B 3234567893 M 3234567892

Answer 1

Seems you need two apply :
select e.Employee_Id, e.Code,
c.PhoneType as MainPhoneType, c.PhoneNumber as MainPhoneNumber,
c1.PhoneType as NonMainPhoneType, c1.PhoneNumber as NonMainNumber
from #Employee e outer apply
(select top (1) c.PhoneType, c.PhoneNumber
from #Contact c
where c.Employee_Id = e.Employee_Id and
c.IsMainNumber = 1
order by c.phonetype
) c outer apply
(select top (1) c1.PhoneType, c1.PhoneNumber
from #Contact c1
where c1.Employee_Id = e.Employee_Id and
c1.IsMainNumber = 0
order by c1.phonetype
) c1;
If you don't want to do JOIN two times then you can use temp table just dump the contacts with relevant index
#temp (Employee_Id, IsMainNumber) include (PhoneType, PhoneNumber)
insert into #temp (Employee_Id, PhoneType, PhoneNumber, IsMainNumber)
select Employee_Id, PhoneType, PhoneNumber, IsMainNumber
from (select *, row_number() over (partition by Employee_Id, IsMainNumber order by PhoneType) as seq
from #Contact
) c
where seq = 1
Now, you don't need to use #Contact again :
select e.*, m.*
from #Employee e cross apply
(select max(case when t.IsMainNumber = 1 then t.PhoneType end) as MainPhoneType,
max(case when t.IsMainNumber = 1 then t.PhoneNumber end) as MainPhoneNumber,
max(case when t.IsMainNumber = 0 then t.PhoneType end) as NonMainPhoneType,
max(case when t.IsMainNumber = 0 then t.PhoneNumber end) as NonMainNumber
from #temp t
where t.Employee_Id = e.Employee_Id
) m;

Answer 2

Not really sure how you determine which nonMainNumber is the one you want. Seems that most of your sample data has several rows that could be returned. I will leave that exercise to you. Here is how you could use some conditional aggregation for this.
select x.Employee_Id
, x.Code
, MainPhoneType = max(case when x.RowNum = 1 then x.PhoneType end)
, MainPhoneNumber = max(case when x.RowNum = 1 then x.PhoneNumber end)
, NonMainPhoneType = max(case when x.RowNum = 2 then x.PhoneType end)
, NonMainPhoneNumber = max(case when x.RowNum = 2 then x.PhoneNumber end)
from
(
select e.Employee_Id
, e.Code
, c.PhoneType
, c.PhoneNumber
, RowNum = ROW_NUMBER() over(partition by e.Employee_Id order by c.IsMainNumber desc, c.PhoneType) --Not sure how you determine the non MainNumber when there are several to pick from
from #Employee e
join #Contact c on c.Employee_Id = e.Employee_Id
) x
group by x.Employee_Id
, x.Code

Answer 3

You can do this with conditional aggregation:
select e.Employee_Id, e.Code
max(case when seqnum = 1 and c.PhoneType = 'M' then c.PhoneType end) as MainPhoneType
max(case when seqnum = 1 and c.PhoneType = 'M' then x.PhoneNumber end) as MainPhoneNumber,
max(case when seqnum = 1 and c.PhoneType <> 'M' then c.PhoneType end) as NonMainPhoneType
max(case when seqnum = 1 and c.PhoneType <> 'M' then c.PhoneNumber end) as NonMainPhoneNumber
from #Employee e join
(select c.*,
row_number() over (partition by c.Employee_Id
(case when PhoneType = 'M' then 'M' end)
order by c.PhoneNumber desc
) as seqnum
from #Contact c
) c
on c.Employee_Id = e.Employee_Id
group by e.Employee_Id, e.Code;
The key idea in this logic is the partition by clause. It divides the two types of phones into two groups -- with 'M' for "main" and NULL for all else.