Dynamically generated sql pivot with calculated columns general methodology - sql

Question

I have the following scenario:
Generation of a pivot with dynamic columns which I can manage via generating dynamic sql either on the client or in a stored procedure
I then need to calculate additional values based on the values already calculated by the pivot.
Here is the sql to generate all the steps in the process :
IF OBJECT_ID('dbo.departmentStats', 'U') IS NOT NULL
DROP TABLE dbo.departmentStats;
CREATE TABLE departmentStats
([Facility] varchar(50), [Department] varchar(50),[Stat] varchar(5),
[January] int, [February] int, [March] int);
INSERT INTO departmentStats
([Facility],[Department],[Stat], [January], [February], [March])
VALUES
('f1','d1','Stat1', 90, 40, 60),
('f1','d1','Stat2', 30, 20, 10);
select 'Original Data' as src,* from departmentStats
IF OBJECT_ID('tempdb.dbo.#pivotstats', 'U') IS NOT NULL
DROP TABLE #pivotstats;
CREATE TABLE #pivotstats
(facility varchar(50),department varchar(50),stat VARCHAR(50),period
varchar(50),value int)
insert into #pivotstats
select facility,department,Stat, period, value
from departmentStats
unpivot
(
value
for Period in (January, February, March)
) u
select 'Original after unpivot' as src,* from #pivotstats
IF OBJECT_ID('dbo.departmentStats2', 'U') IS NOT NULL
DROP TABLE dbo.departmentStats2;
CREATE TABLE departmentStats2
([Facility] varchar(50), [Department] varchar(50),[Period] varchar(20),
[stat1] int, [stat2] int, [Sum] numeric(10,2),[Division] numeric(10,2));
insert into departmentStats2
select
facility
,department
,period
,Stat1
,stat2
,Stat1+Stat2 as [sum]
,case when Stat2<>0 then cast(Stat1/Stat2 AS numeric) else 0.0 end as
division
from #pivotstats
Pivot(
sum(value)
for Stat in ([Stat1],[Stat2])
)
as x
select 'Added formula columns' as src,* from departmentStats2
select 'Union and repivot' as src,* from (
select
facility
,department
,Period
,'Stat1' as Stat
,stat1 as value
from departmentStats2
union all
select
facility
,department
,Period
,'Stat2' as Stat
,stat2 as value
from departmentStats2
union all
select
facility
,department
,Period
,'sum' as Stat
,[SUM] as value
from departmentStats2
union all
select
facility
,department
,Period
,'Division' as Stat
,[Division] as value
from departmentStats2
) as bunion
pivot(
sum(value)
for Period in (January, February, March)
)as P
I have only included three periods for the sample.
I have no doubt that I can generate SQL in this format at will.
I don't foresee any real issues considering my data source will be static (updated daily) and heavily indexed. I don't yet have a full data set ready to test but I will update the q when I do
The reason for the pivot olympics is that the final output is actually as spreadsheet which uses index/match to lookup the values for each stat by facility/department/name to produce graphs and charts
Opinions ?
Obvious flaws?
Is this better done in C# by generating a class at runtime?
Remember the idea is to provide the ability to add stats based on known stats with a minimum of effort.
Here is a better way to do the last step using cross apply instead of union all
select 'Union and repivot' as src,* from (
select
departmentStats2.Facility,
departmentStats2.Department,Departmentstats2.period,x.stat,x.value
from
departmentStats2
cross Apply(
values(Facility,Department,Period,'1_Sales',Sales),
(Facility,Department,Period,'2_Cost',Cost),
(Facility,Department,Period,'3_GrossProfit',GrossProfit),
(Facility,Department,Period,'4_GPP',GPP)
) as x(Facility,Department,Period,Stat,[Value])
)as P
pivot(
sum(value)
for Period in (January, February, March)
)as P
order by Facility,department,Stat
from
http://sqlsunday.com/2014/03/02/unpivot-using-cross-apply/