SQL Server: Rewrite recursive CTE to substitute for option maxrecursion in a View

Question

I ran into the issue that we cannot use underneath query inside a view, but only in a table. Unfortunately we deal with the situation that we don't have tables as an option for this project.

I am curious whether someone knows in what direction I should look for to substitute for underneath logic:

What I try to do is create a record for each day within a date range, by doing the following query:

WITH CTE_PerDay AS (
    SELECT 
         TableDateRange.objectId
        ,TableDateRange.amount
        ,TableDateRange.beginDate
        ,COALESCE(TableDateRange.endDate, '2099-12-31') AS endDate
    FROM TableDateRange
    UNION ALL
    SELECT
         CTE_PerDay.objectId
        ,CTE_PerDay.amount
        ,DATEADD(DAY, 1, CTE_PerDay.beginDate) AS beginDate
        ,CTE_PerDay.endDate
    FROM CTE_PerDay 
    WHERE GETDATE() > DATEADD(DAY, 1, CTE_PerDay.beginDate)

)
SELECT * FROM CTE_PerDay
OPTION (maxrecursion 0)

Sample Dataset TableDataRange

ObjectId	Amount	beginDate	endDate
1	500	2020-01-03
2	35	2015-05-31	2019-10-01
3	200	2017-03-15	2020-06-02

CREATE TABLE TableDateRange
(
     ObjectId   varchar(300),
     Amount     int,
     beginDate  date,
     endDate    date
);

INSERT INTO TableDateRange ( ObjectId , Amount , beginDate , endDate )
VALUES
    ('1', 500, '2020-01-03', NULL),
    ('2', 35, '2015-05-31', '2019-10-01'),
    ('3', 200, '2017-03-15', '2020-06-02');

So the query runs fine, however in a view I can't use the OPTION functionality, and without it I get the error 'The statement terminated. The maximum recursion 100 has been exhausted before statement completion.' Any suggestions?

Answer 1

You could use a tally: that's a set-based solution, that performs better than recursion when the number of iterations increases - and it is supported in views.

Here is an approach:

select t.objectid, t.amount, dateadd(day, x.n, t.begindate) as dt
from (
    select row_number() over (order by (select null)) - 1
    from (values(0),(0),(0),(0),(0),(0),(0),(0),(0),(0)) a(n)
    cross join (values(0),(0),(0),(0),(0),(0),(0),(0),(0),(0)) b(n)
    cross join (values(0),(0),(0),(0),(0),(0),(0),(0),(0),(0)) c(n)
) x(n)
inner join tabledatarange t
    on dateadd(day, x.n, t.begindate) <= case 
        when enddate <= convert(date, getdate()) then enddate
        else convert(date, getdate())
    end

The tally generates all numbers between 0 and 999 (you can easily expand it by adding cross joins). We use it to "multiply" the rows of the original table and generate the date range.

I attempted to rewrite the part the handles the end date. I understand that you don't want future dates, so that's what the condition in the on clause does.

For this sample data:

ObjectId | Amount | beginDate  | endDate   
-------: | -----: | :--------- | :---------
       1 |    500 | 2020-12-28 | null      
       2 |     35 | 2019-09-26 | 2019-10-01
       3 |    200 | 2020-05-28 | 2020-06-02

The query returns:

objectid | amount | dt        
-------: | -----: | :---------
       1 |    500 | 2020-12-28
       1 |    500 | 2020-12-29
       1 |    500 | 2020-12-30
       1 |    500 | 2020-12-31
       2 |     35 | 2019-09-26
       2 |     35 | 2019-09-27
       2 |     35 | 2019-09-28
       2 |     35 | 2019-09-29
       2 |     35 | 2019-09-30
       2 |     35 | 2019-10-01
       3 |    200 | 2020-05-28
       3 |    200 | 2020-05-29
       3 |    200 | 2020-05-30
       3 |    200 | 2020-05-31
       3 |    200 | 2020-06-01
       3 |    200 | 2020-06-02

Demo on DB Fiddle