问题
I'm trying to check if there's an existing record on the day before the selected schedule date. How do we do this? I tried using LAG(), but I'm having problem when it comes to sorting out the data because of the NULL.
DATEDIFF() won't solve my problem as well because I need to check if there's an existing data on that date, and not -1 the date.
What I want to query if example. Day before 2020-01-02 has recordin and recordout, it should reflect NULL if necessary. Is there a way to do this?
This will be the result.
scheduledate schedulein scheduleout recordin recordout prevrecordin prevrecordout
2020-01-02 08:00:00.0000000 17:00:00.0000000 07:41:12.0000000 17:16:54.0000000 NULL NULL
prevrecordin and prevrecordout will be both NULL since 2020-01-01 is NULL.
This is first table.
badgenumber checktype recordin checkdate
10 I 2019-12-20 07:35:58.000 2019-12-20
10 I 2019-12-21 05:18:14.000 2019-12-21
10 I 2019-12-23 07:35:33.000 2019-12-23
10 I 2019-12-26 07:48:20.000 2019-12-26
10 I 2019-12-27 07:41:03.000 2019-12-27
10 I 2019-12-28 07:35:42.000 2019-12-28
10 I 2020-01-02 07:41:12.000 2020-01-02
10 I 2020-01-03 07:50:12.000 2020-01-03
10 I 2020-01-04 07:41:12.000 2020-01-04
This query is being processed by this.
.....
OUTER APPLY (
SELECT TOP(1) t1.recordin, t1.badgenumber
FROM (SELECT MAX(userinfo.badgenumber) AS badgenumber, MAX(RTRIM(checkinout.checktype)) AS 'checktype',
MIN(checkinout.checktime) as 'recordin', MIN(CONVERT(date,checkinout.checktime)) as checkdate,
MAX(RTRIM(employeemasterfile.employeeidno)) AS 'employeeidno' FROM ((checkinout INNER JOIN userinfo
ON checkinout.userid = userinfo.userid) INNER JOIN employeemasterfile ON userinfo.badgenumber = employeemasterfile.fingerscanno)
INNER JOIN departmentmasterfile ON LEFT(employeemasterfile.employeeidno, 4) = LEFT(departmentmasterfile.departmentcode, 4)
WHERE CONVERT(date,checkinout.checktime) BETWEEN DATEADD(DAY, -1,'2019-12-21') AND DATEADD(DAY, 1,'2020-01-05') and badgenumber = '10'
AND CHECKINOUT.CHECKTYPE = 'I' COLLATE SQL_Latin1_General_CP1_CS_AS GROUP BY userinfo.badgenumber, LEFT(checkinout.checktime,14)) AS t1
WHERE
t1.recordin BETWEEN DATEADD(HOUR,-(t0.noofhoursduty),t0.mergetimeinorig) AND DATEADD(HOUR, (t0.noofhoursduty),t0.mergetimeinorig)
AND t1.badgenumber = t0.fingerscanno
AND t0.schedulename !='REST'
ORDER BY abs(datediff(minute, t0.mergetimeinorig, t1.recordin )) DESC
) t1
This is the second table.
badgenumber checktype recordout checkdate
10 O 2019-12-20 20:41:46.000 2019-12-20
10 O 2019-12-21 14:12:34.000 2019-12-21
10 O 2019-12-23 17:03:44.000 2019-12-23
10 O 2019-12-26 17:05:16.000 2019-12-26
10 O 2019-12-27 17:02:32.000 2019-12-27
10 O 2019-12-28 17:07:38.000 2019-12-28
10 O 2020-01-02 17:16:54.000 2020-01-02
10 O 2020-01-03 17:05:11.000 2020-01-03
10 O 2020-01-04 17:04:42.000 2020-01-04
This is being processed by this query.
OUTER APPLY (
SELECT TOP(1) t2.recordout, t2.badgenumber
FROM (SELECT MAX(userinfo.badgenumber) AS badgenumber, MAX(RTRIM(checkinout.checktype)) AS 'checktype',
MAX(checkinout.checktime) as 'recordout', MAX(CONVERT(date,checkinout.checktime)) as checkdate,
MAX(RTRIM(employeemasterfile.employeeidno)) AS 'employeeidno' FROM ((checkinout INNER JOIN userinfo
ON checkinout.userid = userinfo.userid) INNER JOIN employeemasterfile ON userinfo.badgenumber = employeemasterfile.fingerscanno)
INNER JOIN departmentmasterfile ON LEFT(employeemasterfile.employeeidno, 4) = LEFT(departmentmasterfile.departmentcode, 4)
WHERE CONVERT(date,checkinout.checktime) BETWEEN DATEADD(DAY, -1,'2019-12-21') AND DATEADD(DAY, 1,'2020-01-05') and badgenumber = '10'
AND CHECKINOUT.CHECKTYPE = 'O' COLLATE SQL_Latin1_General_CP1_CS_AS GROUP BY userinfo.badgenumber, LEFT(checkinout.checktime,14)) AS t2
WHERE
t2.recordout BETWEEN DATEADD(HOUR,-(t0.noofhoursduty),t0.mergetimeoutorig) AND DATEADD(HOUR, (t0.noofhoursduty),t0.mergetimeoutorig)
AND t2.badgenumber = t0.fingerscanno
AND t0.schedulename !='REST'
ORDER BY abs(datediff(minute, t0.mergetimeoutorig, t2.recordout )) DESC
) t2
This is the query result.
for t1.recordin, and t2.recordout, scheduledate comes from t0.scheduledate with the corresponding date on the table respectively.
scheduledate schedulein scheduleout recordin recordout
2019-12-21 06:00:00.0000000 14:00:00.0000000 05:18:14.0000000 14:12:34.0000000
2019-12-23 08:00:00.0000000 17:00:00.0000000 07:35:33.0000000 17:03:44.0000000
2019-12-24 08:00:00.0000000 17:00:00.0000000 NULL NULL
2019-12-25 08:00:00.0000000 17:00:00.0000000 NULL NULL
2019-12-26 08:00:00.0000000 17:00:00.0000000 07:48:20.0000000 17:05:16.0000000
2019-12-27 08:00:00.0000000 17:00:00.0000000 07:41:03.0000000 17:02:32.0000000
2019-12-28 08:00:00.0000000 17:00:00.0000000 07:35:42.0000000 17:07:38.0000000
2019-12-30 08:00:00.0000000 17:00:00.0000000 NULL NULL
2019-12-31 08:00:00.0000000 17:00:00.0000000 NULL NULL
2020-01-01 08:00:00.0000000 17:00:00.0000000 NULL NULL
2020-01-02 08:00:00.0000000 17:00:00.0000000 07:41:12.0000000 17:16:54.0000000
2020-01-03 08:00:00.0000000 17:00:00.0000000 07:50:12.0000000 17:05:11.0000000
2020-01-04 08:00:00.0000000 17:00:00.0000000 07:41:12.0000000 17:04:42.0000000
I tried doing CASE WHEN (LAG()) as well, unfortunately I am unable to do it because of the NULL value as well.
What I need to achieve is a new column that will display somewhat like this.
Similar expected result.
scheduledate schedulein scheduleout recordin recordout prevrecordin prevrecordout
21/12/2019 06:00:00 14:00:00 05:18:14 14:12:34 NULL NULL
23/12/2019 08:00:00 17:00:00 07:35:33 17:03:44 05:18:14 14:12:34
24/12/2019 08:00:00 17:00:00 NULL NULL 07:35:33 17:03:44
25/12/2019 08:00:00 17:00:00 NULL NULL NULL NULL
26/12/2019 08:00:00 17:00:00 07:48:20 17:05:16 NULL NULL
27/12/2019 08:00:00 17:00:00 07:41:03 17:02:32 07:48:20 17:05:16
28/12/2019 08:00:00 17:00:00 07:35:42 17:07:38 07:41:03 17:02:32
30/12/2019 08:00:00 17:00:00 NULL NULL 07:35:42 17:07:38
31/12/2019 08:00:00 17:00:00 NULL NULL NULL NULL
01/01/2020 08:00:00 17:00:00 NULL NULL NULL NULL
02/01/2020 08:00:00 17:00:00 07:41:12 17:16:54 NULL NULL
03/01/2020 08:00:00 17:00:00 07:50:12 17:05:11 07:41:12 17:16:54
04/01/2020 08:00:00 17:00:00 07:41:12 17:04:42 07:50:12 17:05:11
Your help would be greatly appreciated. Thank you.
回答1:
Based on your sample data, I am assuming the following:
- At most record per badge per day.
recordinandrecordouton the same day.checktypeis irrelevant- You know how to generate the data in the original table.
If so, you can use lag():
select t.*,
(case when datediff(day,
lag(recordin) over (partition by badgenumber order by recordin),
recordin
) <> 1
then null
else lag(recordin) over (partition by badgenumber order by recordin)
end),
(case when datediff(day,
lag(recordin) over (partition by badgenumber order by recordin),
recordin
) <> 1
then null
else lag(recordout) over (partition by badgenumber order by recordin)
end),
from t;
If the above are not true, I would suggest that you ask a new question. Try to simplify the problem. Your rather complex query has nothing to do with the question you are asking, so it doesn't help the question.
来源：https://stackoverflow.com/questions/59747074/check-if-theres-an-existing-record-on-the-day-before