Complicated irregular pivot in SQL Server 2012 - sql

Question

I have two tables:
a (column1, column2, column3)
b (column6, column7, column8)
a.column1 is foreign key as in b.column6.
One row from table a sometimes matches 3 rows in table b, sometimes 5, sometimes 1.... no definite count of returned rows.
I have a business requirement to flip all corresponding columns in table b into one row.. like this:
a.column1, a.column2, a.column3, b.column7, b.column8, b.column7, b.column8
a.column1, a.column2, a.column3, b.column7, b.column8
a.column1, a.column2, a.column3, b.column7, b.column8, b.column7, b.column8, b.column7, b.column8
a.column1, a.column2, a.column3, b.column7, b.column8, b.column7, b.column8, b.column7, b.column8b.column7, b.column8, b.column7, b.column8
You see , the number of columns in each row from Table a is always 3... but from table b, you might have a variable number of columns.... And column7 and column8 have to repeated appear in that order.
How can I do this? Thanks.

Answer 1

It sounds like you are going to need to unpivot and then pivot the data. If you have an unknown number of values you will have to use dynamic SQL but I would first suggest writing a hard-code or static version of the query first, then convert it to dynamic SQL.
The process to unpivot the data is going to take your multiple columns in tableB and convert it into multiple rows. Since you are using SQL Server 2012 you can use CROSS APPLY to unpivot the data:
select column1, column2, column3,
col = col + '_' + cast(seq as varchar(10)),
value
from
(
select a.column1, a.column2, a.column3,
b.column6, b.column7, b.column8,
row_number() over(partition by a.column1
order by a.column1) seq
from tablea a
inner join tableb b
on a.column1 = b.column6
) d
cross apply
(
select 'column6', column6 union all
select 'column7', column7 union all
select 'column8', column8
) c (col, value);
See SQL Fiddle with Demo. This will give you a result similar to:
| COLUMN1 | COLUMN2 | COLUMN3 | COL | VALUE |
| 1 | 2 | 3 | column6_1 | 1 |
| 1 | 2 | 3 | column7_1 | 18 |
| 1 | 2 | 3 | column8_1 | 56 |
| 1 | 2 | 3 | column6_2 | 1 |
| 1 | 2 | 3 | column7_2 | 25 |
| 1 | 2 | 3 | column8_2 | 89 |
As you can see you now have multiple rows that you can easily apply the pivot function to. The PIVOT code will be:
select column1, column2, column3,
column6_1, column7_1, column8_1,
column6_2, column7_2, column8_2,
column6_3, column7_3, column8_3
from
(
select column1, column2, column3,
col = col + '_' + cast(seq as varchar(10)),
value
from
(
select a.column1, a.column2, a.column3,
b.column6, b.column7, b.column8,
row_number() over(partition by a.column1
order by a.column1) seq
from tablea a
inner join tableb b
on a.column1 = b.column6
) d
cross apply
(
select 'column6', column6 union all
select 'column7', column7 union all
select 'column8', column8
) c (col, value)
) src
pivot
(
max(value)
for col in (column6_1, column7_1, column8_1,
column6_2, column7_2, column8_2,
column6_3, column7_3, column8_3)
) piv;
See SQL Fiddle with Demo. Since you stated that you might have an unknown or dynamic number of entries in tableB you will need to use dynamic SQL. This will generate a sql string that will be executed to get you the final result:
DECLARE #cols AS NVARCHAR(MAX),
#query AS NVARCHAR(MAX)
select #cols = STUFF((SELECT ',' + QUOTENAME(col +'_'+cast(seq as varchar(10)))
from
(
select row_number() over(partition by column6
order by column6) seq
from tableB
) t
cross apply
(
select 'column6', 1 union all
select 'column7', 2 union all
select 'column8', 3
) c (col, so)
group by col, so, seq
order by seq, so
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
set #query = 'SELECT column1, column2, column3,' + #cols + '
from
(
select column1, column2, column3,
col = col + ''_'' + cast(seq as varchar(10)),
value
from
(
select a.column1, a.column2, a.column3,
b.column6, b.column7, b.column8,
row_number() over(partition by a.column1
order by a.column1) seq
from tablea a
inner join tableb b
on a.column1 = b.column6
) d
cross apply
(
select ''column6'', column6 union all
select ''column7'', column7 union all
select ''column8'', column8
) c (col, value)
) x
pivot
(
max(value)
for col in (' + #cols + ')
) p '
execute sp_executesql #query;
See SQL Fiddle with Demo. Both versions give a result:
| COLUMN1 | COLUMN2 | COLUMN3 | COLUMN6_1 | COLUMN7_1 | COLUMN8_1 | COLUMN6_2 | COLUMN7_2 | COLUMN8_2 | COLUMN6_3 | COLUMN7_3 | COLUMN8_3 |
| 1 | 2 | 3 | 1 | 18 | 56 | 1 | 25 | 89 | (null) | (null) | (null) |
| 2 | 4 | 6 | 2 | 78 | 245 | (null) | (null) | (null) | (null) | (null) | (null) |
| 3 | 8 | 9 | 3 | 10 | 15 | 3 | 45 | 457 | 3 | 89 | 50 |