更改编号列

Question

我有 NAME 和 PAY，但在这个例子中我需要 CHANGEGROUP：

NAME   PAY  DATE    CHANGEGROUP
Sally   12  10/01/2011  1
Sally   12  10/01/2011  1
Sally   12  11/02/2011  1
Sally   12  11/02/2011  1
Sally   12  12/01/2012  1
Sally   13  04/23/2013  2
Sally   12  04/24/2013  3
Sally   10  05/01/2013  4
Sally   10  10/01/2014  4

我尝试了RANK() 和DENSE_RANK()，但它们是根据价值分组的——因为工资下降了，所以我的分组混乱了。我看到了this，但它与这个旧版本的 SQL 2005 不兼容

Answer 1

这是一个空白和孤岛问题。

一种方法。 SQL Fiddle

WITH T1
     AS (SELECT *,
                ROW_NUMBER()
                  OVER (
                    PARTITION BY NAME
                    ORDER BY DATE) - ROW_NUMBER()
                                       OVER (
                                         PARTITION BY NAME, [PAY]
                                         ORDER BY DATE) AS Grp
         FROM   Table1),
     T2
     AS (SELECT *,
                MIN(DATE)
                  OVER (
                    PARTITION BY NAME, Grp) AS MinDate
         FROM   T1)
SELECT [NAME],
       [PAY],
       [DATE],
       DENSE_RANK()
         OVER (
           PARTITION BY NAME
           ORDER BY MinDate) AS CHANGEGROUP
FROM   T2
ORDER  BY NAME,
          MinDate