row_number - over partition sql server ejemplo

Cálculo basado en la fila de SQL anterior

Question

Estoy intentando calcular el valor de la factura de meses anteriores y colocarlo en la misma fila que el valor actual.

Estoy usando SQL Server.Microsoft SQL Server 2008 R2 (SP2)

He intentado con CTE y todo tipo de combinaciones extrañas, pero mi resultado nunca es correcto. ¿Puede alguien señalarme en la dirección correcta?

Tabla de ejemplo:

invoice_month   value
--------------------------
    201510      337265.386
    201511      335466.456
    201512      338646.500
    201601      333440.380
    201602      330731.208
    201603      339299.752
    201604      340878.168
    201605      359585.382
    201606      339616.430
    201607      316457.486
    201608      308009.976
    201609      232196.268
    201610      894839.180
    201611      232196.268
    201612      232196.268
    201701      232196.268
    201702      232196.268
    201703      232196.268

El resultado esperado sería:

invoice_month   value,         prevValue
--------------------------------------------
    201510      337265.386,    NULL
    201511      335466.456,    337265.386
    201512      338646.500,    335466.456
    201601      333440.380,    333440.380
    201602      330731.208,    333440.380
    201603      339299.752,    330731.208
    201604      340878.168,    339299.752
    201605      359585.382,    340878.168
    201606      339616.430,    359585.382
    201607      316457.486,    339616.430
    201608      308009.976,    316457.486
    201609      232196.268,    308009.976
    201610      894839.180,    232196.268
    201611      232196.268,    894839.180
    201612      232196.268,    232196.268
    201701      232196.268,    232196.268
    201702      232196.268,    232196.268
    201703      232196.268,    232196.268

Answer 1

En SQL Server 2012+, usaría lag() :

select t.invoice_month, t.value,
       lag(t.value) over (order by t.invoice_month) as prev_value
from t;

Esta es una función estándar ANSI. En versiones anteriores de SQL Server, hay métodos alternativos.

Answer 2

para versiones anteriores a 2012:

;with cte
 as
(select invoice_month,value,
row_number() over (order by invoice_month) as rn
from  #temp  )
select c1.*,c2.value
from cte c1
left join
cte c2
on c1.rn=c2.rn+1

Answer 3

SELECT *
FROM   (SELECT TOP 1 invoice_month,
                     value,
                     NULL AS [Previous value]
        FROM   Invoice
        ORDER  BY invoice_month ASC) T
UNION
SELECT I.invoice_month,
       I.value,
       I1.value AS [Previous value]
FROM   Invoice I
       INNER JOIN Invoice I1
               ON I.invoice_month - 1 = I1.invoice_month