MS SQL Pivot dat from long with multiple columns - sql

Question

I want to pivot a table from long to wide that has multiple id columns.
I found solutions for one column but not really for multiple columns.
The closest solution that I could adapt for one column was this one
T-SQL PIVOT data from long form to wide by a date
My table looks more or less like this,
create table t (id int, date date, varA_id int, VarB_id int, value int)
insert into t values
(1,'2005-01-20',1, 1,197)
,(2,'2005-01-20',1,2,58)
,(3,'2005-01-20',1,3,90)
,(4,'2005-01-20',2,1,210)
,(5,'2005-01-20',2,2,133)
,(6,'2005-01-20',2,3,67)
,(7,'2005-01-20',3,1,87)
,(8,'2005-01-20',3,2,87)
,(9,'2005-01-20',3,3,87)
Actually without the date, but that's fine. I want to spread in a way that I get columns for each permutation of VarA_id and VarB_id
So my expected result would look like this
My actual table has three _id columns and more permutations, so I really need a generic solution.
Based on the other solution in my link I was hoping something like this would work. I adjust the top part that creates the column names and this would work. I dont know how to realy adjust the bottom part that fetches the values.
declare #cols nvarchar(max);
declare #sql nvarchar(max);
select #cols = stuff((
select distinct
', ' + 'VarA_'+convert(varchar(10),varA_id) + '_VarB_'+convert(varchar(10),varB_id)
from t
order by 1
for xml path (''), type).value('.','nvarchar(max)')
,1,2,'')
select #sql = '
select Id, date, ' + #cols + '
from (
select Id, date, varA_id = ''v''+convert(varchar(10),varA_id), value
from t
) as t
pivot (sum([value]) for [varA_id] in (' + #cols + ') ) p'
select #sql
exec(#sql);

Answer 1

The main problem with your dynamic sql?
It was that the name constructed in the source query didn't match the generated column names.
Here's a fix :
declare #cols varchar(max) = null;
declare #sql nvarchar(max);
select #cols = concat(#cols+', '+char(10), quotename(concat('VarA_', varA_id, '_VarB_', varB_id)))
from test
group by varA_id, varB_id
order by varA_id, varB_id;
-- select #cols as cols;
set #sql = 'select * '+char(10)+
'from ( ' +char(10)+
' select [date], [value], ' +char(10)+
' concat(''VarA_'',varA_id,''_VarB_'',varB_id) as Col ' +char(10)+
' from test ' +char(10)+
') as src ' +char(10)+
'pivot (sum([value]) for Col in ('+char(10)+ #cols +char(10)+')) pvt';
-- select #sql as sql;
exec(#sql);
date
VarA_1_VarB_1
VarA_1_VarB_2
VarA_1_VarB_3
VarA_2_VarB_1
VarA_2_VarB_2
VarA_2_VarB_3
VarA_3_VarB_1
VarA_3_VarB_2
VarA_3_VarB_3
2005-01-20
197
58
90
210
133
67
87
87
87
db<>fiddle here

Answer 2

Ok my own solution so far is to add a help column and basically just do what the other questions does. I need to improve on this, so I dont add a column and I would like better names, but at least this shows what I want.
alter table t add help_col nvarchar(10)
Update t
set help_col=convert(varchar(10),varA_id)+convert(varchar(10),varB_id)
declare #cols nvarchar(max);
declare #sql nvarchar(max);
select #cols = stuff((
select distinct
', ' + 'v'+convert(varchar(10),help_col)
from t
order by 1
for xml path (''), type).value('.','nvarchar(max)')
,1,2,'')
select #sql = '
select date, ' + #cols + '
from (
select date, help_col = ''v''+convert(varchar(10),help_col), value
from t
) as t
pivot (sum([value]) for [help_col] in (' + #cols + ') ) p'
select #sql
exec(#sql);
Which results in
select date, v11, v12, v13, v21, v22, v23, v31, v32, v33 from ( select date, help_col = 'v'+convert(varchar(10),help_col), value from t ) as t pivot (sum([value]) for [help_col] in (v11, v12, v13, v21, v22, v23, v31, v32, v33) ) p
which yields
date v11 v12 v13 v21 v22 v23 v31 v32 v33
2005-01-20 197 58 90 210 133 67 87 87 87