非顺序SQL分区

Question

Blockquote

All,

I am trying to partition a set of data but I don't want my partitions to be "sequential".

Traditionally by partitioning by ID and then by sale stage I get this. Sequential partitioning.

ROW_NUMBER() OVER(PARTITION BY  ID,  sale stage ORDER BY  Converge, modified date desc)

ID               sale stage      modified date      partition
01_AAI_73133    Closed Lost        5/1/2015             1
01_AAI_73133    Closed Lost        4/26/2015            2
01_AAI_73133    Closed Lost        4/20/2015            3
01_AAI_73133    Locked and Loaded  4/5/2015             1
01_AAI_73133    Locked and Loaded  3/29/2015            2
01_AAI_73133    Pitching           3/7/2015             1
01_AAI_73133    Pitching           2/14/2015            2
01_AAI_73133    Pitching           2/1/2015             3
01_AAI_73134    Pitching           1/20/2015            4

What I actually want is this

ID               sale stage      modified date      partition
01_AAI_73133    Closed Lost        5/1/2015             1
01_AAI_73133    Closed Lost        4/26/2015            1
01_AAI_73133    Closed Lost        4/20/2015            1
01_AAI_73133    Locked and Loaded  4/5/2015             2
01_AAI_73133    Locked and Loaded  3/29/2015            2
01_AAI_73133    Pitching           3/7/2015             3
01_AAI_73133    Pitching           2/14/2015            3
01_AAI_73133    Pitching           2/1/2015             3
01_AAI_73134    Pitching           1/20/2015            3

Does anyone have a idea how to achieve this?

Thanks

Answer 1

You are probably looking for dense_rank() :

DENSE_RANK() OVER (ORDER BY ID, salestage)

If you are trying to order by the maximum date for each id and sale stage, then you might need a subquery:

select dense_rank() over (order by maxmd, id, salestate)
from (select . . .,
             MAX(ModifiedDate) OVER (PARTITION BY ID, salestage) as maxmd
      from . . .
     ) t;