Aktywny suma stanu jak SCD w SQL Server 2012

Question

mam te dane:

CREATE TABLE #student 
(
    student_id INT, 
    status  VARCHAR(50), 
    created_dt DATE 
) 

CREATE TABLE #student_status_history 
(
    student_id  INT, 
    from_status  VARCHAR(50), 
    to_status   VARCHAR(50), 
    status_changed_dt DATE 
) 

INSERT INTO #student (student_id, status, created_dt) 
VALUES (1, 'Active', '2016-10-02'), 
     (2, 'Active', '2016-10-02'), 
     (3, 'Active', '2016-10-02') 

SELECT * 
FROM #student 

5 października student2 status jest zaktualizowany do inactive

UPDATE #student 
SET status = 'Inactive' 
WHERE student_id = 2 

INSERT INTO #student_status_history (student_id, from_status, to_status, status_changed_dt) 
VALUES (2, 'Active', 'Inactive', '2016-10-05') 

SELECT * 
FROM #student 

SELECT * 
FROM #student_status_history 

8 października student2 stan jest aktualizowany do active:

UPDATE #student 
SET status = 'Active' 
WHERE student_id = 2 

INSERT INTO #student_status_history (student_id, from_status, to_status, status_changed_dt) 
VALUES (2, 'InActive', 'Active', '2016-10-08') 

9 października tworzę d innego studenta:

INSERT INTO #student (student_id, status, created_dt) 
VALUES (4, 'Active', '2016-10-09') 

10 października mam te dane w tabelach.

select * from #student 
    select * from #student_status_history 

Stosując powyższe tabele

powinien wygenerować raport w dniu 10 października począwszy od Oct 1 do Oct 10 aktywnych studentów, z tego dnia

Wyjście powinno być jak poniżej

Date     ActiveCount 
    -----------   ----------- 
    2016-10-01     0 
    2016-10-02     3 
    2016-10-04     3 
    2016-10-05     2 
    2016-10-06     2 
    2016-10-07     2 
    2016-10-08     3 
    2016-10-09     4 
    2016-10-10     4 

Answer 1

Oto jeden sposób, aby to zrobić:

Potrzebujesz calendar tabela z listą dat. Użyłem Recursive CTE do wygenerowania dat między zakresem dat.

;WITH calendar 
    AS (SELECT dates = CONVERT(DATETIME, '2016-10-01') 
     UNION ALL 
     SELECT dates = Dateadd(DAY, 1, dates) 
     FROM calendar 
     WHERE dates < '2016-10-10') 
SELECT c.dates, 
     Count(s.created_dt) AS ActiveCount 
FROM calendar c 
     LEFT JOIN #student s 
       ON s.created_dt <= c.dates 
WHERE NOT EXISTS (SELECT 1 
        FROM #student_status_history sh 
        WHERE sh.student_id = s.student_id 
        HAVING c.dates BETWEEN Min(CASE 
               WHEN from_status = 'active ' 
                AND to_status = 'Inactive' THEN status_changed_dt 
               END) AND Max(CASE 
                  WHEN to_status = 'active ' 
                    AND from_status = 'Inactive' THEN Dateadd(dd, -1, status_changed_dt) 
                  END)) 
GROUP BY c.dates 
OPTION (MAXRECURSION 0) 

Wynik:

╔═════════════════════════╦═════════════╗ 
║   dates   ║ ActiveCount ║ 
╠═════════════════════════╬═════════════╣ 
║ 2016-10-01 00:00:00.000 ║   0 ║ 
║ 2016-10-02 00:00:00.000 ║   3 ║ 
║ 2016-10-03 00:00:00.000 ║   3 ║ 
║ 2016-10-04 00:00:00.000 ║   3 ║ 
║ 2016-10-05 00:00:00.000 ║   2 ║ 
║ 2016-10-06 00:00:00.000 ║   2 ║ 
║ 2016-10-07 00:00:00.000 ║   2 ║ 
║ 2016-10-08 00:00:00.000 ║   3 ║ 
║ 2016-10-09 00:00:00.000 ║   4 ║ 
║ 2016-10-10 00:00:00.000 ║   4 ║ 
╚═════════════════════════╩═════════════╝ 

Answer 2

zrobiłbym coś takiego:

;with cte_dates as (
    select convert(date, '20161001', 112) as [date] 
    union all 
    select dateadd(day, 1, [date]) as [date] 
    from cte_dates as d 
    where 
     d.[date] < convert(date, '20161010', 112) 
) 
select 
    d.[date] as [Date], 
    sum(case when isnull(sh.to_status, s.[status]) = 'Active' then 1 else 0 end) as ActiveCount 
from cte_dates as d 
    left join #student as s on 
     s.created_dt <= d.[date] 
    outer apply (
     select top 1 sh.to_status 
     from #student_status_history as sh 
     where 
      sh.student_id = s.student_id and 
      sh.status_changed_dt <= d.[date] 
     order by 
      sh.status_changed_dt desc 
    ) as sh 
group by 
    d.[date]