SQLServer 2008 PIVOT on multiple fields issue - sql

Question

I am trying to using PIVOT on 3 fields/Columns i.e Site, Value, Date.I am using SQLServer 2008.Please check below my actual table and desired output table and the pivot query i used
Actual Table:
**tblReference**
**ID** **Refer_code** **Site** **Value** **Date**
1 9290 CA 12.5 2014-01-01 20:20:41
5 9290 TX 12.6 2014-01-05 18:20:30
2 6651 CA 13.5 2014-01-01 21:20:21
3 7442 TX 14.5 2014-01-05 19:15:14
4 8093 CA 15.5 2014-01-01 19:20:20
6 8093 TX 16.5 2014-01-05 20:20:20
**Desired output table:**
**Refer_code** **Site_1** **Site_2** **Val_1** **Val_2** **StartDate** **EndDate**
9290 CA TX 12.5 12.6 2014-01-01 2014-01-05
20:20:41 18:20:30
6651 CA NULL 13.5 NULL 2014-01-01 NULL
21:20:21
7442 NULL TX NULL 14.5 NULL 2014-01-05
19:15:14
8093 CA TX 15.5 16.5 2014-01-01 2014-01-05
19:20:20 20:20:20
**Query:**
SELECT Refer_code, [Site_1], [Site_2], [Date_1] AS StartDate, [Date_2] AS EndDate, [Val_1], [Val_2]
FROM
(
SELECT Refer_code, Site, 'Site_'+ cast(row_number() over(partition by Refer_code order by Date) as nvarchar(50)) SiteVal,Date,'Date_'+cast(row_number() over(partition by Refer_code order by Date) as nvarchar(50)) DateVal,Value,'Val_'+cast(row_number() over(partition by Refer_code order by Date) as nvarchar(50)) Val
FROM tblReference
) x
pivot
(
min(Site)
for SiteVal in ([Site_1], [Site_2])
) p
pivot
(
min(Date)
for DateVal in ([Date_1], [Date_2])
) s
pivot
(
min(Value)
for Val in ([Val_1], [Val_2])
) t
The above query is not returning the result as expected.Please help me on the solution.

Answer 1

There are several ways to get the desired result. The easiest way might be to use an aggregate function with a CASE expression but you can also use the PIVOT function.
Aggregate with CASE:
I would take the following steps, first create a unique sequence for each refer_code using row_number():
select refer_code, site,
row_number() over(partition by refer_code
order by date) seq,
value,
date
from tblReference;
See SQL Fiddle with Demo.
Once you have the unique sequence you can easily convert your existing data into multiple columns by applying an aggregate function with a CASE expression:
;with cte as
(
select refer_code, site,
row_number() over(partition by refer_code
order by date) seq,
value,
date
from tblReference
)
select refer_code,
max(case when seq = 1 then site end) site1,
max(case when seq = 2 then site end) site2,
max(case when seq = 1 then value end) value1,
max(case when seq = 2 then value end) value2,
max(case when seq = 1 then date end) startdate,
max(case when seq = 2 then date end) enddate
from cte
group by refer_code;
See SQL Fiddle with Demo
PIVOT:
If you want to use the PIVOT function and you need to pivot multiple columns I would suggest looking at unpivoting those columns first, then applying the pivot function. You will need to create the unique sequence which is used for your new column names, then you need to unpivot the site, date and value columns. Since you are using SQL Server 2008+ you can use CROSS APPLY with VALUEs to unpivot the multiple columns into multiple rows:
;with cte as
(
select refer_code, site,
row_number() over(partition by refer_code
order by date) seq,
value,
date
from tblReference
)
select *
from
(
select refer_code,
col = col + cast(seq as varchar(10)),
val
from cte
cross apply
(
values
('Site', site),
('Value', cast(value as varchar(10))),
('Date', convert(varchar(10), date, 120))
) c (col, val)
) d;
See SQL Fiddle with Demo. This gets your data into the format of:
| REFER_CODE | COL | VAL |
|------------|--------|------------|
| 6651 | Site1 | CA |
| 6651 | Value1 | 13.50 |
| 6651 | Date1 | 2014-01-01 |
| 7442 | Site1 | TX |
| 7442 | Value1 | 14.50 |
| 7442 | Date1 | 2014-01-05 |
Now you can easily apply the pivot function to get the final result, so the entire code will be:
;with cte as
(
select refer_code, site,
row_number() over(partition by refer_code
order by date) seq,
value,
date
from tblReference
)
select refer_code,
site1, site2,
startdate = date1, enddate = date2,
value1, value2
from
(
select refer_code,
col = col + cast(seq as varchar(10)),
val
from cte
cross apply
(
values
('Site', site),
('Value', cast(value as varchar(10))),
('Date', convert(varchar(10), date, 120))
) c (col, val)
) d
pivot
(
max(val)
for col in (site1, site2, date1, date2,
value1, value2)
) piv;
See SQL Fiddle with Demo. Both will give a result of:
| REFER_CODE | SITE1 | SITE2 | STARTDATE | ENDDATE | VALUE1 | VALUE2 |
|------------|-------|--------|------------|------------|--------|--------|
| 6651 | CA | (null) | 2014-01-01 | (null) | 13.50 | (null) |
| 7442 | TX | (null) | 2014-01-05 | (null) | 14.50 | (null) |
| 8093 | CA | TX | 2014-01-01 | 2014-01-05 | 15.50 | 16.50 |
| 9290 | CA | TX | 2014-01-01 | 2014-01-05 | 12.50 | 12.60 |
Edit:
Based on your comment that you need the columns to be CA as site1, etc. then the easiest way to get the result would be using an aggregate function with CASe:
select refer_code,
max(case when site = 'CA' then site end) site1,
max(case when site = 'TX' then site end) site2,
max(case when site = 'CA' then value end) value1,
max(case when site = 'TX' then value end) value2,
max(case when site = 'CA' then date end) startdate,
max(case when site = 'TX' then date end) endate
from tblReference
group by refer_code;
See SQL Fiddle with Demo. This gives a result:
| REFER_CODE | SITE1 | SITE2 | VALUE1 | VALUE2 | STARTDATE | ENDATE |
|------------|--------|--------|--------|--------|--------------------------------|--------------------------------|
| 6651 | CA | (null) | 13.5 | (null) | January, 01 2014 21:20:21+0000 | (null) |
| 7442 | (null) | TX | (null) | 14.5 | (null) | January, 05 2014 19:15:14+0000 |
| 8093 | CA | TX | 15.5 | 16.5 | January, 01 2014 19:20:20+0000 | January, 05 2014 20:20:20+0000 |
| 9290 | CA | TX | 12.5 | 12.6 | January, 01 2014 20:20:41+0000 | January, 05 2014 18:20:30+0000 |