sql - 根据 10 天间隔创建组

Question

我有以下数据表

Animal  Immunization_Date
Cat     1/18/2017
Cat     1/27/2017
Cat     5/7/2017
Cat     5/12/2017
Dog     1/1/2017
Dog     1/5/2017
Dog     1/7/2017
Dog     3/25/2017
Dog     4/18/2017

我正在尝试根据动物 10 天的间隔创建排名，这将导致以下结果。 (查找动物的第一个日期，然后在该日期 10 天内的任何日期分配一组 1 。然后为未分配给 1 的动物取下一个日期并将其分配给 2 和然后将 2 分配给该日期 10 天内的任何日期等...)

Animal  Immunization_Date   10_Day_Group_Rank
Cat     1/18/2017           1
Cat     1/27/2017           1
Cat     5/7/2017            2
Cat     5/12/2017           2
Dog     1/1/2017            1
Dog     1/5/2017            1
Dog     1/7/2017            1
Dog     3/25/2017           2
Dog     4/18/2017           3

我一直在尝试以下代码，但我似乎无法让 10 天组工作。

Select
dt.Animal,
dt.Immunization_Date,
sum(dt.10_day_Group) over(partition dt.Animal order by dt.Immunization_Date rows unbounded preceding) as 10_day_Group --creates a running total that is also the group
from
(
Select
Animal,
Immunization_Date,
case when min(Immunization_Date) over (partition by Animal order by Immunization_Date) <=10 then 1 else 0 end as 10_Day_Group --Create intervals of 10 days
from Table_A
) as dt

我不太确定如何将 10 天分组。

case when min(Immunization_Date) over (partition by Animal order by Immunization_Date) <=10 then 1 else 0 end as 10_Day_Group

我可以使用以下命令在 Excel 中执行此操作。我知道 excel 和 SQL 是不同的，但我希望如果有什么可以在 SQL 中完成的事情，我希望能看到如何在 Excel 中完成。

Excel 数据表如下所示(表从单元格 A1 开始)。 (注意 Animal 需要排序，并且 Immunization_Date 需要排序才能使 Excel 公式发挥作用)

Animal  Immunization_Date   Dummy_1 10_Day_Group
Cat     1/18/2017       1/18/2017       1
Cat     1/27/2017       1/18/2017       1
Cat     5/7/2017        5/7/2017        2
Cat     5/12/2017       5/7/2017        2
Dog     1/1/2017        1/1/2017        1
Dog     1/5/2017        1/1/2017        1
Dog     1/7/2017        1/1/2017        1
Dog     3/25/2017       3/25/2017       2
Dog     4/18/2017       4/18/2017       3

Dummy_1 的公式是以下 IFERROR(IF(AND(A2=A1,B2-C1<=10),C1,B2),B2)

10_Day_Group 的公式是以下 IFERROR(IF(AND(C2=C1,A2=A1),D1,IF(AND(A2=A1,C2<>C1),D1+1,1)),1)

Answer 1

@MatBailie 的递归答案非常好，但是当每只动物的行数增加时，性能会变差。

当第一个 CTE 可以在 volatile 表中实现时，它将降低资源使用率(因为 Teradata 的优化器不会实现这个结果，该死):

CREATE VOLATILE TABLE boundaries AS
(
  SELECT
    i.*, -- need to add the alias 
    (
      SELECT MIN(immunization_date)
        FROM immunizations
       WHERE animal             = i.animal
         AND immunization_date >= i.immunization_date + 10
    )
      AS next_boundary_date
  FROM
    immunizations   i
 )
WITH DATA 
UNIQUE PRIMARY INDEX(animal, immunization_date)
ON COMMIT PRESERVE ROWS;

但是当您可以使用临时表时，您也可以使用简单的递归:

CREATE VOLATILE TABLE vt AS
 (
   SELECT
      animal, 
      immunization_date,
      Row_Number() -- add row number to simplify recursive processing
      Over (PARTITION BY animal
            ORDER BY immunization_date) AS rn
   FROM immunizations AS i
 )
WITH DATA 
UNIQUE PRIMARY INDEX(animal, rn)
ON COMMIT PRESERVE ROWS;

WITH RECURSIVE cte AS
 (
   SELECT
      animal, immunization_date, rn,
      immunization_date+10 AS end_date, -- define the end of the range 
      1 AS grp -- SMALLINT = limited to 127 group, CAST to a larger INT for more groups
   FROM vt
   WHERE rn = 1 -- oldest row

   UNION ALL

   SELECT 
      vt.animal, vt.immunization_date, vt.rn,
      -- check if the current row's date is within the 10 day range
      -- otherwise increase the group number and define the new range end
      CASE WHEN vt.immunization_date < end_date THEN cte.end_date ELSE vt.immunization_date +10 END,
      CASE WHEN vt.immunization_date < end_date THEN cte.grp      ELSE cte.grp+1 END
   FROM cte
   JOIN vt
     ON vt.animal = cte.animal
    AND vt.rn = cte.rn+1
 )
SELECT *
FROM cte
ORDER BY 1,2