以下查询帐户周末和节假日。该查询提供了即时包含假期的规定,但为了使查询更清晰,我只是将假期具体化到一个实际的表中。
CREATE TABLE tx
(n varchar(4), d date);
INSERT INTO tx
(n, d)
VALUES
('Bill', '2006-12-29'), -- Friday
-- 2006-12-30 is Saturday
-- 2006-12-31 is Sunday
-- 2007-01-01 is New Year's Holiday
('Bill', '2007-01-02'), -- Tuesday
('Bill', '2007-01-03'), -- Wednesday
('Bill', '2007-01-04'), -- Thursday
('Bill', '2007-01-05'), -- Friday
-- 2007-01-06 is Saturday
-- 2007-01-07 is Sunday
('Bill', '2007-01-08'), -- Monday
('Bill', '2007-01-09'), -- Tuesday
('Bill', '2012-07-09'), -- Monday
('Bill', '2012-07-10'), -- Tuesday
('Bill', '2012-07-11'); -- Wednesday
create table holiday(d date);
insert into holiday(d) values
('2007-01-01');
/* query should return 7 consecutive good
attendance(from December 29 2006 to January 9 2007) */
/* and 3 consecutive attendance from July 7 2012 to July 11 2012. */
询问:
with first_date as
(
-- get the monday of the earliest date
select dateadd( ww, datediff(ww,0,min(d)), 0 ) as first_date
from tx
)
,shifted as
(
select
tx.n, tx.d,
diff = datediff(day, fd.first_date, tx.d)
- (datediff(day, fd.first_date, tx.d)/7 * 2)
from tx
cross join first_date fd
union
select
xxx.n, h.d,
diff = datediff(day, fd.first_date, h.d)
- (datediff(day, fd.first_date, h.d)/7 * 2)
from holiday h
cross join first_date fd
cross join (select distinct n from tx) as xxx
)
,grouped as
(
select *, grp = diff - row_number() over(partition by n order by d)
from shifted
)
select
d, n, dense_rank() over (partition by n order by grp) as nth_streak
,count(*) over (partition by n, grp) as streak
from grouped
where d not in (select d from holiday) -- remove the holidays
输出:
| D | N | NTH_STREAK | STREAK |
-------------------------------------------
| 2006-12-29 | Bill | 1 | 7 |
| 2007-01-02 | Bill | 1 | 7 |
| 2007-01-03 | Bill | 1 | 7 |
| 2007-01-04 | Bill | 1 | 7 |
| 2007-01-05 | Bill | 1 | 7 |
| 2007-01-08 | Bill | 1 | 7 |
| 2007-01-09 | Bill | 1 | 7 |
| 2012-07-09 | Bill | 2 | 3 |
| 2012-07-10 | Bill | 2 | 3 |
| 2012-07-11 | Bill | 2 | 3 |
现场测试:http ://www.sqlfiddle.com/#!3/815c5/1
查询的主要逻辑是将所有日期向后移动两天。这是通过将日期除以 7 并将其乘以 2,然后从原始数字中减去它来完成的。例如,如果给定日期是 15 日,这将计算为 15/7 * 2 == 4;然后从原始数字中减去 4,15 - 4 == 11。15 将成为第 11 天。同样,第 8 天变成第 6 天;8 - (8/7 * 2) == 6。
Weekends are not in attendance(e.g. 6,7,13,14)
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15
将计算应用于所有工作日数字将产生以下值:
1 2 3 4 5
6 7 8 9 10
11
对于假期,您需要在出勤时将它们插入,以便可以轻松确定连续性,然后将它们从最终查询中删除。以上出勤产生连续11次良好出勤。
查询逻辑的详细解释在这里:http ://www.ienablemuch.com/2012/07/monitoring-perfect-attendance.html