Struggling with getting results (Views, Variables, StoredProcedures) - sql

Question

As I understand it, you cannot create a view with variables. The dilemma is I have to do some datediff on values. Our third party reporting software can only do basic Selects on Tables or views. I cannot call up any stored procedures or set variables there.
My original source data looks select * from tblquotestatuschangelog
results in this I would like to see these results
Quotenumber UpdatedOn Status UpdatedBy
----------------------------------------------
100001 04102019 Open domain/user
100001 04132019 Closed domain/user
I have done a pivot on this data to get the results in the desired format with this query. (There are more status types than in this example, and more can be added via the CRM we use) (Q1)
DECLARE #cols AS NVARCHAR(MAX),
#query AS NVARCHAR(MAX);
select #cols = STUFF((SELECT distinct ',' + QUOTENAME(T1.STATUS)
FROM tblCglQuoteStatusChangeLog as T1
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
set #query = 'SELECT quotenumber, ' + #cols + ' from
(
select quotenumber, status, updatedon, updatedby
from tblcglquotestatuschangelog
where updatedon > 2019-04-01
) x
pivot
(
max(updatedon)
for status in (' + #cols + ')
) p '
execute #query
To get these results
Quotenumber Open Closed
----------------------------------------------
100001 04102019 04132019
I would like to call (Q1) as a View to perform
Select
QuoteNumber,
case when datediff(day,Open, Closed) = 0 then Cast('SAMEDAY' AS NVARCHAR(20)) else cast(datediff(day,openedon,updatedon) as NVarCHAR(20)) end as TotalAge,
datediff(day,Open,SentToCustomer) as Stage1,
datediff(day,SentToCustomer,) as Stage2,
From V1
Any help or alternate direction to achieve results would be much appreciated.

Answer 1

You can use conditional aggregation to achieve your results - as long as you want to rely on the assumptions about the shape of your data. Those are your answers to my questions - that the tuple is unique and that there are only 2 status values. If you have more status values (and it smells like you do) you can extend the logic
use tempdb;
set nocount on;
go
if object_id('v1') is not null
drop view v1;
if object_id('chglog') is not null
drop table chglog;
go
create table chglog (IdNumber int not null, UpdatedOn date not null, Status varchar(10) not null, UpdatedBy varchar(20) not null
constraint chg_unique unique (IdNumber, Status),
constraint chg_check check (Status in ('Open', 'Closed'))
);
insert chglog(IdNumber, UpdatedOn, Status, UpdatedBy)
values (100001, '20190410', 'Open', 'domain/user'), (100001, '20190413', 'Closed', 'domain/user'),
(9999, '20190401', 'Open', 'zork'),
(99001, '20190402', 'Open', 'bob'),
(99001, '20190402', 'Closed', 'alice')
;
go
-- setup complete, now create a view to do a hard-coded pivot
create view v1 as
select IdNumber,
max(case Status when 'Open' then UpdatedOn else null end) as 'Open',
max(case Status when 'Closed' then UpdatedOn else null end) as 'Closed'
from chglog
group by IdNumber;
go
-- test the view
select * from v1
order by IdNumber;
-- now the query that you wanted to write/use
Select
IdNumber,
case when datediff(day, [Open], Closed) = 0 then N'SAMEDAY' else cast(datediff(day, [Open], Closed) as nvarchar(20)) end as TotalAge
-- datediff(day,Open,SentToCustomer) as Stage1,
-- datediff(day,SentToCustomer,) as Stage2,
from v1
order by IdNumber;
I'll point out that your "table" contained IdNumber but your last query referenced QuoteNumber (as well as other columns you did not mention). I just ignored some and guesssed some. This is just old-style pivoting based on hard-coded values. So long as you know ahead of time the values you need to consider when pivoting and you know that the domain is unchanging (or don't care about other values), this will what you asked.
The question you should consider is whether you should care about other status values that are added via your CRM after you create this view. But you're kinda stuck with the 3rd party reporting software anyways. A reporting tool that can't use anything but a table or view as a source of data seems primitive (to say the least).