如何使用 SQL 中的循环将记录与以前的记录进行比较？

Question

I am looking at a data set of Emergency Room visits. I only want to keep visits per ID that are 30 days apart. So as an example say I have this below.

If I start with ID=1:

• In Row 1 I can see that the lag between row 1 and 2 is 15 days so I will exclude, or for now flag, row 2.
• Then I will continue to use Row 1 to evaluate Row 3. Again this is only 17 days so I will exclude Row 3 and look at Row 4.
• Row 4 is 30 days away so I keep it and then use Row 4 to evaluate Row 5....and so on.

I have been trying to do this with the lag function but I can't figure out how to utilize the lag when I have to continue to use the 'anchor' row to evaluate several rows.

Top is what I have and bottom is what I want. Any ideas?

I am using AZURE data studio.

HAVE

Row#  ID  DATE
 1    1   1/1/2020
 2    1   1/15/2020
 3    1   1/17/2020
 4    1   2/4/2020
 5    1   3/15/2020
 6    2   1/15/2020
 7    2   3/15/2020
 8    2   3/18/2020

WANT

Row#  ID  DATE
 1    1   1/1/2020
 4    1   2/4/2020
 5    1   3/15/2020
 6    2   1/15/2020
 7    2   3/15/2020

Answer 1

This tutorial page should get you started on a cursor based solution.

Answer 2

You don't use a loop. You continue to use LAG, you were on the right way initially.

;WITH dateLagged AS (
    SELECT 
        ID
     ,  Date
     ,  Diff = ISNULL(DATEDIFF(day,LAG(Date,1) OVER(PARTITION BY ID ORDER BY ID, Date), Date),0) 
    FROM dbo.EmergencyRoom),
 DiffCumulated AS (
    SELECT 
       ID
    ,  Date
    ,  CumDiff = SUM(Diff) OVER(PARTITION BY ID  ORDER BY ID, Date) 
    FROM dateLagged
 ),
 AnchorsMarked AS (
    SELECT
       ID
    ,  Date
    ,  Marker =  IIF(CumDiff = 0 
                  OR CumDiff > 30 AND LAG(CumDiff,1) OVER(ORDER BY ID, Date) < 30 
                  OR CumDiff - LAG(CumDiff,1) OVER(ORDER BY ID, Date) > 30, 1,0)
    FROM  DiffCumulated
  )
SELECT 
   ID
,  Date 
FROM AnchorsMarked WHERE Marker = 1

As a rule of thumb: if you want to use looping in SQL then you've taken a wrong turn somewhere. There are very few problems in SQL which do require looping, it is not one of them.