Pivoting without aggregation - sql-server

Question

I'm trying to pivot without aggregation, and running into a bit of a wall.
Here's the sample T-SQL I'm trying to get to work:
declare #optionalFields table (ParentId int, Name nvarchar(50), Value nvarchar(50));
insert into #optionalFields values (1, 'Field1', 'Foo');
insert into #optionalFields values (1, 'Field2', 'Bar');
insert into #optionalFields values (1, 'Field3', '42');
insert into #optionalFields values (2, 'Field1', 'Bar');
insert into #optionalFields values (2, 'Field2', 'Foo');
insert into #optionalFields values (2, 'Field3', '24');
declare #data table (Id int, Name nvarchar(50));
insert into #data values (1, 'Test record 1');
insert into #data values (2, 'Test record 2');
declare #joined table (Id int, Name nvarchar(50), OptionalFieldName nvarchar(50), OptionalFieldValue nvarchar(50));
insert into #joined
select
data.Id
,data.Name
,opt.Name
,opt.Value
from #data data
inner join #optionalFields opt on data.Id = opt.ParentId
declare #cols as nvarchar(max) =
stuff((select distinct ',' + quotename(OptionalFieldName) from #joined for xml path(''), type).value('.', 'nvarchar(max)'), 1, 1, '');
select * into #tmp from #joined
-- just to see that it's returning the expected values (it does)
select
Id
,Name
,OptionalFieldName
,OptionalFieldValue
,row_number() over (partition by Id order by Id) RN
from #tmp -- this is the FROM clause in the below dynamic-sql query
declare #query as nvarchar(max) = '
select Id, Name, ' + #cols + '
from (select Id, Name, OptionalFieldName, OptionalFieldValue, row_number() over (partition by Id order by Id) RN from #tmp) src
pivot (max(OptionalFieldName) for RN in (' + #cols + ')) pvt';
execute(#query);
drop table #tmp;
SSMS is giving me 2 errors:
Msg 8114, Level 16, State 1, Line 4
Error converting data type nvarchar to bigint.
Msg 473, Level 16, State 1, Line 4
The incorrect value "Field1" is supplied in the PIVOT operator.
The "debug" select statement is returning this:
The article (link above) seemed very promising, however I can't seem to be able to get it to work. What am I doing wrong? Or is this article outright wrong and what I'm trying to do is impossble?
I've seen a number of similar SO questions, but either they involved all-numeric fields that could "just work" with aggregation, or they involved known columns that could be implemented as simple joins - I don't know what OptionalFieldName values I'm going to be selecting, and the OptionalFieldValue values are strings that simply can't be aggregated, at least AFAIK.

Answer 1

I'm a bit confused on why you are trying to trick this using row_number(). Even though you have string values, you can still aggregate it - you just need to use max or min to get the result.
I'd always recommend trying to write your query with hard-coded values first, especially when using PIVOT before even attempting to use dynamic SQL. I'm unsure why you can't just write the query this way:
select Id, Name, Field1, Field2, Field3
from
(
select
Id
,Name
,OptionalFieldName
,OptionalFieldValue
from #tmp
) d
pivot
(
max(OptionalFieldValue)
for OptionalFieldName in (Field1, Field2, Field3)
) piv;
See a Demo.
Then if you really need dynamic SQL, you would just write it:
declare #optionalFields table (ParentId int, Name nvarchar(50), Value nvarchar(50));
insert into #optionalFields values (1, 'Field1', 'Foo');
insert into #optionalFields values (1, 'Field2', 'Bar');
insert into #optionalFields values (1, 'Field3', '42');
insert into #optionalFields values (2, 'Field1', 'Bar');
insert into #optionalFields values (2, 'Field2', 'Foo');
insert into #optionalFields values (2, 'Field3', '24');
declare #data table (Id int, Name nvarchar(50));
insert into #data values (1, 'Test record 1');
insert into #data values (2, 'Test record 2');
declare #joined table (Id int, Name nvarchar(50), OptionalFieldName nvarchar(50), OptionalFieldValue nvarchar(50));
insert into #joined
select
data.Id
,data.Name
,opt.Name
,opt.Value
from #data data
inner join #optionalFields opt on data.Id = opt.ParentId
declare #cols as nvarchar(max);
set #cols = stuff((select distinct ',' + quotename(OptionalFieldName)
from #joined
for xml path(''), type).value('.', 'nvarchar(max)'), 1, 1, '');
select * into #tmp from #joined
DECLARE #query AS NVARCHAR(MAX)
set #query = 'SELECT Id, Name,' + #cols + '
from
(
select Id
,Name
,OptionalFieldName
,OptionalFieldValue
from #tmp
) x
pivot
(
max(OptionalFieldValue)
for OptionalFieldName in (' + #cols + ')
) p '
execute(#query);
See Demo. Both versions appear to give the result that you have requested.

Answer 2

EDIT: This answer is here only to show another way to accomplish this pivoting. The answer from #bluefeet is the best solution!
Hope I understand what you need right:
declare #optionalFields table (ParentId int, Name nvarchar(50), Value nvarchar(50));
insert into #optionalFields values (1, 'Field1', 'Foo');
insert into #optionalFields values (1, 'Field2', 'Bar');
insert into #optionalFields values (1, 'Field3', '42');
insert into #optionalFields values (2, 'Field1', 'Bar');
insert into #optionalFields values (2, 'Field2', 'Foo');
insert into #optionalFields values (2, 'Field3', '24');
declare #data table (Id int, Name nvarchar(50));
insert into #data values (1, 'Test record 1');
insert into #data values (2, 'Test record 2');
select
data.Id
,data.Name
,opt.Name as Name1
,opt.Value into #tmp
from #data data
inner join #optionalFields opt on data.Id = opt.ParentId
declare #cols as nvarchar(max) =
stuff((select distinct ',' + quotename(Name1) from #tmp for xml path(''), type).value('.', 'nvarchar(max)'), 1, 1, '');
DECLARE #cols1 as nvarchar(max) =
stuff((select distinct +',MAX(CASE WHEN (pvt1.'+quotename(Name1) +' = ros.RN AND pvt1.id = ros.id) THEN ros.Value ELSE NULL END) as '+quotename(Name1) from #tmp for xml path(''), type).value('.', 'nvarchar(max)'), 1, 1, '');
declare #query as nvarchar(max) = '
;WITH cte AS(
SELECT * FROM #tmp
),
ros AS (
SELECT ROW_NUMBER() OVER (partition by Id order by Id) AS [RN],id,Value
FROM cte),
pvt1 AS (
select *
from (select Id, Name, Name1, row_number() over (partition by Id order by Id) RN
from cte) src
pivot (max(RN) for Name1 in ('+#cols+')) pvt)
SELECT pvt1.ID,
pvt1.Name,
'+#cols1+'
FROM pvt1
CROSS JOIN ros
GROUP BY pvt1.ID,
pvt1.Name'
execute(#query);
drop table #tmp
Result:
ID Name Field1 Field2 Field3
1 Test record 1 Foo Bar 42
2 Test record 2 Bar Foo 24
And if you add more Fields like insert into #optionalFields values (2, 'Field4', '15');, you will get:
ID Name Field1 Field2 Field3 Field4
1 Test record 1 Foo Bar 42 NULL
2 Test record 2 Bar Foo 24 15