SQL - Convert columns to rows (use a pivot or other means)? - sql

Question

I am struggling with reformatting an answer set I have from a query. This is data from a monitoring tool that I am calculating a monthly average along with a linear regression projection. This is all performed in other functions and working fine. These functions all feed the below two tables with the final data that I use. The only thing I am currently grappling with is re-formatting the output to show more columns than rows.
To start, let me give you the sample data that I'm working with Using SQL Server 2008R2
create table finalactual(Target varchar(100), MonthNum int, [Actual] real)
create table finalprojection(Target varchar(100), MonthNum int, [monthname] varchar(20), [Forecast] real)
Insert into finalprojection values ('C:\', 2, 'February', 65609.29)
Insert into finalprojection values ('C:\', 3, 'March', 65850.27)
Insert into finalprojection values ('C:\', 4, 'April', 66091.26)
Insert into finalprojection values ('C:\', 5, 'May', 66332.25)
Insert into finalprojection values ('C:\', 6, 'June', 66573.23)
Insert into finalprojection values ('C:\', 7, 'July', 66814.22)
Insert into finalprojection values ('C:\', 8, 'August', 67055.2)
Insert into finalprojection values ('C:\', 9, 'September', 67296.19)
Insert into finalprojection values ('C:\', 10, 'October', 67537.17)
Insert into finalprojection values ('C:\', 11, 'November', 67778.16)
Insert into finalprojection values ('C:\', 12, 'December', 68019.14)
Insert into finalprojection values ('C:\', 13, 'January', 68260.13)
Insert into finalprojection values ('E:\', 2, 'February', 385251.0)
Insert into finalprojection values ('E:\', 3, 'March', 401171.1)
Insert into finalprojection values ('E:\', 4, 'April', 417091.2)
Insert into finalprojection values ('E:\', 5, 'May', 433011.3)
Insert into finalprojection values ('E:\', 6, 'June', 448937.3)
Insert into finalprojection values ('E:\', 7, 'July', 464851.4)
Insert into finalprojection values ('E:\', 8, 'August', 480771.4)
Insert into finalprojection values ('E:\', 9, 'September', 496691.5)
Insert into finalprojection values ('E:\', 10, 'October', 512611.6)
Insert into finalprojection values ('E:\', 11, 'November', 528531.6)
Insert into finalprojection values ('E:\', 12, 'December', 544451.8)
Insert into finalprojection values ('E:\', 13, 'January', 560371.8)
Insert into finalactual values ('C:\', 2, 62927.88)
Insert into finalactual values ('C:\', 3, 64534.62)
Insert into finalactual values ('C:\', 4, 67215.3)
Insert into finalactual values ('C:\', 5, 70991.05)
Insert into finalactual values ('C:\', 6, 69857.11)
Insert into finalactual values ('C:\', 7, 64440.7)
Insert into finalactual values ('C:\', 8, 64359.08)
Insert into finalactual values ('E:\', 2, 382691.4)
Insert into finalactual values ('E:\', 3, 400543.9)
Insert into finalactual values ('E:\', 4, 418160.4)
Insert into finalactual values ('E:\', 5, 435643.2)
Insert into finalactual values ('E:\', 6, 451213.5)
Insert into finalactual values ('E:\', 7, 466608.2)
Insert into finalactual values ('E:\', 8, 476218.2)
If you run the following query
select a.target, a.monthnum as 'Month Value', case when a.monthnum <= 12 then
a.monthnum else a.monthnum-12 end as 'Month Number', a.monthname as [Month],
b.Actual as 'Actual', a.forecast from finalprojection a
left join
finalactual b on a.Target = b.Target AND a.monthnum = b.MonthNum
order by Target ASC, [Month Value] ASC
The resulting answer set looks like the following:
target Month Value Month Number Month Actual forecast
C:\ 2 2 February 62927.88 65609.29
C:\ 3 3 March 64534.62 65850.27
C:\ 4 4 April 67215.3 66091.26
C:\ 5 5 May 70991.05 66332.25
C:\ 6 6 June 69857.11 66573.23
C:\ 7 7 July 64440.7 66814.22
C:\ 8 8 August 64359.08 67055.2
C:\ 9 9 September NULL 67296.19
C:\ 10 10 October NULL 67537.17
C:\ 11 11 November NULL 67778.16
C:\ 12 12 December NULL 68019.14
C:\ 13 1 January NULL 68260.13
E:\ 2 2 February 382691.4 385251
E:\ 3 3 March 400543.9 401171.1
E:\ 4 4 April 418160.4 417091.2
E:\ 5 5 May 435643.2 433011.3
E:\ 6 6 June 451213.5 448931.3
E:\ 7 7 July 466608.2 464851.4
E:\ 8 8 August 476218.2 480771.4
E:\ 9 9 September NULL 496691.5
E:\ 10 10 October NULL 512611.6
E:\ 11 11 November NULL 528531.6
E:\ 12 12 December NULL 544451.8
E:\ 13 1 January NULL 560371.8
This is fine, as I have 6 or 7 months of real data, combined with a linear regression calculation extending an additional 6 months (which is why the "null" values for Sept to Jan) under actual.
What I would like to have for the output would be something along this line:
Target A/F Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan
C:\ A 62927 64534 67215 70991 69857 64440 64359 NULL NULL NULL NULL NULL
C:\ F 65609 65850 66091 66332 66573 66814 67055 67296 67537 67778 68019 68260
I have abbreviated the decimals along with the A/F (meaning Actual vs Forecast) for space reasons. I have also left off the E:\ drive...again, in the interest of brevity.
One issue that will likely be a problem is that the starting month (in the example above, it is February) will be dynamic. Meaning if the query runs in September, the first month will be March...October will have April as the first month...and so on.
I have created a SQL Fiddle: http://www.sqlfiddle.com/#!3/901c4/1 to save some time for you all.
I thank you all for your help in advance.
Updated info related to the sliding date need:
Hi guys...all excellent solutions.
Let me explain what I meant by the starting month.
Running the report today (in August) will need to have the first month of the result set being February and ending in January of 2015. This is because the report reaches back 6 months in history to gather the monthly results. From there, it extends (the linear regression) past the actual data for an additional 6 months (forecast). This is one quirk that I really need to accommodate (i know...pain in the butt).
So if this query were to run next month (in September), the starting month would need to be March and end in February 2015. If I run the report in October, the starting month would need to be April, which would increment through May of 2015. I hope that clarifies how that sliding scale of the months needs to be presented.
Just adding a clarification on how the month scale is being used:
the column "month value" is how the output is presented in the necessary order. This column will increment based from the starting month and continue for 12 increments. For example:
If the query is run in August (like the above sample data) the starting month for the data will be February (Run date is "month value" of 8 data begins 6 months prior or "month value" of 2. This sets up the order you see above where "month value" begins in the answer set at 2 and continues to 13. Part of the routine that generates these tables evaluates the "month value" column and generates the "month number" column (if it is < 12 then use the original value...if > 12 then subtract 12). This is why the rows that have a "month value" of 13 show a "month number" of 1 which, using a datename() conversion returns "January".
If this same query were to be run next month (September or month 10), everything would shift by 1 in the answer set. The starting "month value" would be 3 or March and the end "month value" would be 14 or February (the system will subtract 12 from the value over 12 to give us the 2). In this case, the answer set returned would look like this:
target Month Value Month Number Month Actual forecast
C:\ 3 3 March 62927.88 65609.29
C:\ 4 4 April 64534.62 65850.27
C:\ 5 5 May 67215.3 66091.26
C:\ 6 6 June 70991.05 66332.25
C:\ 7 7 July 69857.11 66573.23
C:\ 8 8 August 64440.7 66814.22
C:\ 9 9 Sept 64359.08 67055.2
C:\ 10 10 October NULL 67296.19
C:\ 11 11 November NULL 67537.17
C:\ 12 12 December NULL 67778.16
C:\ 13 1 January NULL 68019.14
C:\ 14 2 February NULL 68260.13
Note: removed the E:\ drive for brevity above
The pivoted answer should then look like below:
Target A/F Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb
C:\ A 62927 64534 67215 70991 69857 64440 64359 NULL NULL NULL NULL NULL
C:\ F 65609 65850 66091 66332 66573 66814 67055 67296 67537 67778 68019 68260
The people seeing the output don't need to see years mentioned as they understand the report is showing a 12 month span of real and forecast data...so as a "January" shows up, they know that they are spanning into a new year.
I hope this helps give you the idea of how the answer set is structured.
Again...thank you for the current ideas...very good stuff.

Answer 1

While the other two answers will get you the result that you want, I should do this slightly different. You can unpivot the data in your Actual and Forecast columns first, then pivot the months.
You didn't specify what version of SQL Server you are using but starting in SQL Server 2005+ you can use CROSS APPLY to unpivot.
The basic syntax would be similar to:
select a.target,
Left(a.monthname, 3) as [Month],
AorF,
value
from finalprojection a
left join finalactual b
on a.Target = b.Target
AND a.monthnum = b.MonthNum
cross apply
(
select 'A', b.Actual union all
select 'F', a.forecast
) c(AorF, value);
See Demo. This is going to convert your multiple columns Actual and Forecast into multiple rows. Once the data is in this format, you can easily pivot the months making the full script:
select target,
AorF,
Jan, Feb, Mar, Apr, May, Jun, Jul,
Aug, Sep, Oct, Nov, Dec
from
(
select a.target,
Left(a.monthname, 3) as [Month],
AorF,
value
from finalprojection a
left join finalactual b
on a.Target = b.Target
AND a.monthnum = b.MonthNum
cross apply
(
select 'A', b.Actual union all
select 'F', a.forecast
) c(AorF, value)
) d
pivot
(
max(value)
for month in (Jan, Feb, Mar, Apr, May, Jun, Jul,
Aug, Sep, Oct, Nov, Dec)
) piv
order by target;
See SQL Fiddle with Demo. This gives a final result of:
| TARGET | AORF | JAN | FEB | MAR | APR | MAY | JUN | JUL | AUG | SEP | OCT | NOV | DEC |
|--------|------|---------------|----------------|----------------|---------------|--------------|---------------|----------------|--------------|------------|--------------|-------------|--------------|
| C:\ | A | (null) | 62927.87890625 | 64534.62109375 | 67215.296875 | 70991.046875 | 69857.109375 | 64440.69921875 | 64359.078125 | (null) | (null) | (null) | (null) |
| C:\ | F | 68260.1328125 | 65609.2890625 | 65850.2734375 | 66091.2578125 | 66332.25 | 66573.2265625 | 66814.21875 | 67055.203125 | 67296.1875 | 67537.171875 | 67778.15625 | 68019.140625 |
| E:\ | A | (null) | 382691.40625 | 400543.90625 | 418160.40625 | 435643.1875 | 451213.5 | 466608.1875 | 476218.1875 | (null) | (null) | (null) | (null) |
| E:\ | F | 560371.8125 | 385251 | 401171.09375 | 417091.1875 | 433011.3125 | 448937.3125 | 464851.40625 | 480771.40625 | 496691.5 | 512611.59375 | 528531.625 | 544451.8125 |
If you needs the columns to be ordered differently based on a the start month, then you'll have to use dynamic SQL similar to:
DECLARE #cols AS NVARCHAR(MAX),
#query AS NVARCHAR(MAX),
#start int
set #start = 8
select #cols = STUFF((SELECT ',' + QUOTENAME(Left(monthname, 3))
from finalprojection
where MonthNum > (#start - 6)
and MonthNum <= (#start + 6)
group by monthNum,MonthName
order by MonthNum
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
set #query = 'SELECT target, AorF,' + #cols + '
from
(
select a.target,
Left(a.monthname, 3) as [Month],
AorF,
value
from finalprojection a
left join finalactual b
on a.Target = b.Target
AND a.monthnum = b.MonthNum
cross apply
(
select ''A'', b.Actual union all
select ''F'', a.forecast
) c (AorF, value)
) x
pivot
(
max(value)
for month in (' + #cols + ')
) p
order by target'
exec sp_executesql #query;
See Demo

Answer 2

I don't understand exactly what you mean by the columns are in different order based on the month this is run. If you can explain that I can help figure out a solution to that. This query looks lengthy but it is actually quite simple once you break it down. This is using what is known as a cross tab. To break the Actual and Forecast I did them separately and then used a UNION to put them back together.
with MyCte as
(
select a.target
, a.monthnum
, b.Actual
, a.forecast
, ROW_NUMBER() over(partition by a.target order by a.monthnum) as RowNum
from finalprojection a
left join finalactual b on a.Target = b.Target AND a.monthnum = b.MonthNum
)
select Target
, 'A' as [A/F]
, max(case when monthnum = 2 then Actual end) as Feb
, max(case when monthnum = 3 then Actual end) as Mar
, max(case when monthnum = 4 then Actual end) as Apr
, max(case when monthnum = 5 then Actual end) as May
, max(case when monthnum = 6 then Actual end) as Jun
, max(case when monthnum = 7 then Actual end) as Jul
, max(case when monthnum = 8 then Actual end) as Aug
, max(case when monthnum = 9 then Actual end) as Sep
, max(case when monthnum = 10 then Actual end) as Oct
, max(case when monthnum = 11 then Actual end) as Nov
, max(case when monthnum = 12 then Actual end) as Dec
, max(case when monthnum = 1 then Actual end) as Jan
from MyCte
group by Target
union all
select Target
, 'F' as [A/F]
, max(case when monthnum = 2 then Forecast end) as Feb
, max(case when monthnum = 3 then Forecast end) as Mar
, max(case when monthnum = 4 then Forecast end) as Apr
, max(case when monthnum = 5 then Forecast end) as May
, max(case when monthnum = 6 then Forecast end) as Jun
, max(case when monthnum = 7 then Forecast end) as Jul
, max(case when monthnum = 8 then Forecast end) as Aug
, max(case when monthnum = 9 then Forecast end) as Sep
, max(case when monthnum = 10 then Forecast end) as Oct
, max(case when monthnum = 11 then Forecast end) as Nov
, max(case when monthnum = 12 then Forecast end) as Dec
, max(case when monthnum = 1 then Forecast end) as Jan
from MyCte
group by Target
order by Target, [A/F]

Answer 3

Here is a pivot table based on the query you gave the results from. It does Actuals and Forecast separately. To have the columns in a different order you would need to select them in a different order.
SELECT Target, 'Actual' AS [A/F]
, [February], [March], [April], [May], [June], [July], [August]
, [September], [October], [November], [December], [January]
FROM (
SELECT [Target], [Month], [Actual] FROM (
select a.target, a.monthnum as 'Month Value', case when a.monthnum <= 12 then
a.monthnum else a.monthnum-12 end as 'Month Number', a.monthname as [Month],
b.Actual as 'Actual', a.forecast
from #finalprojection a left join #finalactual b on a.Target = b.Target AND a.monthnum = b.MonthNum
) AS YourTable
) AS SourceTable
PIVOT (
SUM([Actual]) FOR [Month] IN ([February], [March], [April], [May], [June], [July], [August]
, [September], [October], [November], [December], [January])
) AS PivotData
UNION ALL
SELECT Target, 'Forecast' AS [A/F]
, [February], [March], [April], [May], [June], [July], [August]
, [September], [October], [November], [December], [January]
FROM (
SELECT [Target], [Month], [forecast] FROM (
select a.target, a.monthnum as 'Month Value', case when a.monthnum <= 12 then
a.monthnum else a.monthnum-12 end as 'Month Number', a.monthname as [Month],
b.Actual as 'Actual', a.forecast
from #finalprojection a left join #finalactual b on a.Target = b.Target AND a.monthnum = b.MonthNum
) AS YourTable
) AS SourceTable
PIVOT (
SUM([forecast]) FOR [Month] IN ([February], [March], [April], [May], [June], [July], [August]
, [September], [October], [November], [December], [January])
) AS PivotData
When I ran this with the data you provided I got the following:
Target A/F February March April May June July August September October November December January
C:\ Actual 62927.87890625 64534.62109375 67215.296875 70991.046875 69857.109375 64440.69921875 64359.078125 NULL NULL NULL NULL NULL
E:\ Actual 382691.40625 400543.90625 418160.40625 435643.1875 451213.5 466608.1875 476218.1875 NULL NULL NULL NULL NULL
C:\ Forecast 65609.2890625 65850.2734375 66091.2578125 66332.25 66573.2265625 66814.21875 67055.203125 67296.1875 67537.171875 67778.15625 68019.140625 68260.1328125
E:\ Forecast 385251 401171.09375 417091.1875 433011.3125 448937.3125 464851.40625 480771.40625 496691.5 512611.59375 528531.625 544451.8125 560371.8125
Note that the pivot uses a sum of all the data - so could give you dodgy results if you have more than one entry for each target & month. This is in there because the pivot needs something to group on. This might be where your problem of the order of the months comes in, but I need you to be more specific there to be sure.