How to get a Datediff calculation based on two different case statements?

Question

I have a very long select query but have pasted the relevant pieces here.

I need to figure out how to get the datediff between the OrderSignedDate and the OrderIssuedDate that are each in a separate case statement within my select query. Any help would be greatly appreciated:

select tkh.task_id
,cas.case_number as CaseNo
,cst.description as CaseStatus
,cas.title as Title
,ztt.description as OrderType
,tkh.task_status

--Order Signed Date
,(select (case when 
zrt.description = 'Administrative Law Judge'
and tka.completed_date is not null
then convert(varchar(10), tka.completed_date, 101) END) 
from ecl_task_actor tka
join ecl_resource res on tka.entity_pk_id = res.resource_id
join ecl_z_resource_type zrt on res.resource_type_id = zrt.resource_type_id
where tka.task_id = tkh.task_id
and tka.task_actor_status <> 'D'
and zrt.description = 'Administrative law judge') as OrderSignedDate

--Date Order was Issued
,(select (case when 
zrt.description = 'Order Processor Team'
and tka.completed_date is not null
then convert(varchar(10), tka.completed_date,101)
END) 
from ecl_task_actor tka
join ecl_resource res on tka.entity_pk_id = res.resource_id
join ecl_z_resource_type zrt on res.resource_type_id = zrt.resource_type_id
where tka.task_id = tkh.task_id
and tka.task_actor_status <> 'D'
and zrt.description = 'Order processor Team') as OrderIssuedDate


from ecl_task_header tkh
join ecl_case cas on tkh.case_id = cas.case_id
JOIN ecl_z_case_status_type as cst on cst.case_status_type_id = cas.case_status_type_id
join ecl_z_task_type ztt on tkh.task_type_id = ztt.task_type_id

where ztt.description like '%Issue%'
order by cas.case_number asc

Answer 1

It's actually not difficult at all. Your selection looks like this:

select ...,
(...) as OrderSignedDate,
(...) as OrderIssueDate,
...

The ... denotes that some stuff is there which remains unchanged. Now, let's use this as a table :

select ..., OrderSignedDate, OrderIssueDate, ...,
from (
select ...,
(...) as OrderSignedDate,
(...) as OrderIssueDate,
...
) t

and at the outer select which is wrapped around your initial select you can use OrderSignedDate and OrderIssueDate as columns and you can datediff them.

Answer 2

I'm assuming you're looking to add this inline to your existing query.

Since the sub-selects that define the two existing dates are correlated to the outer query through their WHERE clauses, there's not a great way to avoid repeating the code. On the other hand, that means it's as simple as just repeating the code. Use the existing sub-selects as the parameters for your DATEDIFF .

This ought to work:

...<The current SELECT list, plus>
  ,DATEDIFF
  (
    DAY,
    (--Order Signed Date
    SELECT (CASE
              WHEN zrt.description = 'Administrative Law Judge'
                   AND tka.completed_date IS NOT NULL
                THEN CONVERT(varchar(10), tka.completed_date, 101)
            END
           )
    FROM
      ecl_task_actor AS tka
    JOIN
      ecl_resource AS res
        ON
        tka.entity_pk_id = res.resource_id
    JOIN
      ecl_z_resource_type AS zrt
        ON
        res.resource_type_id = zrt.resource_type_id
    WHERE
      tka.task_id = tkh.task_id
      AND tka.task_actor_status <> 'D'
      AND zrt.description = 'Administrative law judge'
    ),
   (--Date Order was Issued
    SELECT (CASE
              WHEN zrt.description = 'Order Processor Team'
                   AND tka.completed_date IS NOT NULL
                THEN CONVERT(varchar(10), tka.completed_date, 101)
            END
           )
    FROM
      ecl_task_actor AS tka
    JOIN
      ecl_resource AS res
        ON
        tka.entity_pk_id = res.resource_id
    JOIN
      ecl_z_resource_type AS zrt
        ON
        res.resource_type_id = zrt.resource_type_id
    WHERE
      tka.task_id = tkh.task_id
      AND tka.task_actor_status <> 'D'
      AND zrt.description = 'Order processor Team'
   )
  ) AS DaysBetween
... <The rest of the query>

Edit: This should work because those two sub-selects both resolve into dates for each row of data, so we're just reusing those two dates for the DATEDIFF function. It'll add a little overhead, but my guess is that with a query that's already this big, split-second performance isn't necessarily the primary concern.