CODEWITHSUNDEEP
sqlsql-server
Cesar
Cesar

Reputation: 585

Subtracting value from previous row sql by group

I have the following dataset:

DATA_FIM    CENTRO  C   M   ESTOQUE
2018-02-01  HD01    CD  70  539.000
2018-03-01  HD01    CD  70  511.000
2018-04-01  HD01    CD  70  468.000
2018-05-01  HD01    CD  70  447.000
2018-06-01  HD01    CD  70  382.000
2018-07-01  HD01    CD  70  348.000
2018-08-01  HD01    CD  70  285.000
2018-09-01  HD01    CD  70  245.000
2018-10-01  HD01    CD  70  221.000
2018-11-01  HD01    CD  70  207.000
2018-12-01  HD01    CD  70  122.000
2018-12-21  HD01    CD  70  101.000
2018-02-01  HD01    CD  71  164.000
2018-03-01  HD01    CD  71  147.000
2018-04-01  HD01    CD  71  124.000
2018-05-01  HD01    CD  71  107.000
2018-06-01  HD01    CD  71  78.000

1- Considering the 2nd, 3rd, 4th columns as a group, I want to subtract the 5th column values from previous row of same column.

In order to create a new row.

[![Data][2]][2]

It's very similar with this problem: [Issue:13196190][3], however, in my case I have multiple index, and I am not sure how to solve.

UPDATED

I am using the suggested codes

select 
ZBI_FAT_PRODUCT.DATA_FIM,
        ZBI_DIM_EMPRESA.CENTRO, 
        ZBI_DIM_EMPRESA.CANAL, 
        ZBI_DIM_PRODUCT.MATERIAL, 
        ZBI_DIM_PRODUCT.TITULO,
        ZBI_FAT_PRODUCT.ESTOQUE_VENDA, 
        LAG(ZBI_FAT_PRODUCT.ESTOQUE_VENDA, 1, ZBI_FAT_PRODUCT.ESTOQUE_VENDA) OVER (PARTITION BY ZBI_DIM_EMPRESA.CENTRO, ZBI_DIM_PRODUCT.MATERIAL, ZBI_DIM_PRODUCT.TITULO ORDER BY ZBI_FAT_PRODUCT.DATA_FIM) - ZBI_FAT_PRODUCT.ESTOQUE_VENDA AS NEW
FROM ZBI_DIM_EMPRESA INNER JOIN (ZBI_DIM_PRODUCT INNER JOIN ZBI_FAT_PRODUCT ON ZBI_DIM_PRODUCT.BK = ZBI_FAT_PRODUCT.BK_MATERIAL) ON ZBI_DIM_EMPRESA.BK = ZBI_FAT_PRODUCT.BK_EMPRESA;

Upvotes: 0

Views: 1316

Answers (2)

Barry Piccinni
Barry Piccinni

Reputation: 1801

For brevity, and to extract the solution from the comments -

The answer is to use the LAG() function: 

SELECT *, 
LAG(ESTOQUE, 1, ESTOQUE) OVER (PARTITION BY CENTRO, C, M ORDER BY DATA_FIM) - ESTOQUE AS New
FROM EnergyLog

Here is an example of this working: SQLFiddle

Upvotes: 1

Gordon Linoff
Gordon Linoff

Reputation: 1271241

You seem to just want lag():

select el.*,
       (lag(estoque, 0, estoque) over (partition by centro, c, m order by date_mif) -
        estoque
       ) as diff
from energylog el;

EDIT:

For the edited question, only two columns seem to comprise a group

select el.*,
       (lag(estoque, 0, value) over (partition by centro, canal, order by date_mif) -
        estoque
       ) as diff
from energylog el;

Upvotes: 5

Related Questions