Oracle SQL 滞后 function 带隔板

Question

I have a table which contains year, month, period, region_code, item_code, current_price columns in a mixed order as shown below:

YEAR	MONTH	PERIOD	REGION_CODE	ITEM_CODE	CURRENT_PRICE
2020	12	41	2500	010111	15
2020	12	41	1500	010111	15
2021	01	41	2500	030333	32.4
2021	01	21	2500	010111	17
2020	11	21	2500	030333	15
2020	12	41	1500	020222	25
2020	11	41	2500	010111	15
2021	01	21	2500	030333	21.5
2020	12	21	2500	020222	25.2
2020	12	41	1500	030333	35
2020	12	21	1500	010111	15.1
2020	11	41	2500	020222	15
2021	01	41	2500	020222	19.8
2020	12	41	2500	030333	35
2020	12	21	1500	020222	25.2
2020	12	21	1500	030333	35.3
2020	12	41	2500	020222	25
2020	11	41	1500	030333	12.9
2020	11	41	1500	010111	15
2020	12	21	2500	030333	35.3
2021	01	21	2500	020222	22.6
2020	11	41	1500	020222	15
2020	11	21	1500	030333	15
2020	11	21	1500	020222	28
2020	11	21	1500	010111	65
2021	01	41	2500	010111	31
2020	11	41	2500	030333	12.9
2021	01	41	1500	030333	32.4
2021	01	21	1500	010111	17
2021	01	41	1500	020222	19.8
2020	12	21	2500	010111	15.1
2021	01	21	1500	030333	21.5
2021	01	21	1500	020222	22.6
2021	01	41	1500	010111	31
2020	11	21	2500	020222	28
2020	11	21	2500	010111	65

I want to get previous_price for each item using lag fuction but could not manage to do that. After that i want to calculate current and previous month's geometric mean for each year, month, region_code and item. ANd there will be a TOTAL row for each month for each item. The result table i wanna get look like that: (I didn't calculate CUR_GEO_MEAN and PREV_GEO_MEAN values for the demo table)

YEAR	MONTH	REGION_CODE	ITEM_CODE	CUR_GEO_MEAN	PREV_GEO_MEAN
2020	11	1500	010111	...	...
2020	11	1500	020222	...	...
2020	11	1500	030333	...	...
2020	11	2500	010111	...	...
2020	11	2500	020222	...	...
2020	11	2500	030333	...	...
2020	11	TOTAL	010111	...	...
2020	11	TOTAL	020222	...	...
2020	11	TOTAL	030333	...	...
2020	12	1500	010111	...	...
2020	12	1500	020222	...	...
2020	12	1500	030333	...	...
2020	12	2500	010111	...	...
2020	12	2500	020222	...	...
2020	12	2500	030333	...	...
2020	12	TOTAL	010111	...	...
2020	12	TOTAL	020222	...	...
2020	12	TOTAL	030333	...	...
2021	01	1500	010111	...	...
2021	01	1500	020222	...	...
2021	01	1500	030333	...	...
2021	01	2500	010111	...	...
2021	01	2500	020222	...	...
2021	01	2500	030333	...	...
2021	01	TOTAL	010111	...	...
2021	01	TOTAL	020222	...	...
2021	01	TOTAL	030333	...	...

I have tried to calculate by writing loads of queries but all failed. I am kinda newbie to this field so everything you show/suggest will help me much. Thanks in advance.

Answer 1

This query should be pretty self-explained and should at least point you to the solution if there are some details that I didn't catch.

with geo as (
 select YEAR, MONTH, REGION_CODE,ITEM_CODE, exp(avg(ln(current_price))) CUR_GEO_MEAN   from tab
 group by YEAR, MONTH,REGION_CODE,ITEM_CODE 
   union all
 select YEAR, MONTH, 'TOTAL' REGION_CODE,ITEM_CODE, exp(avg(ln(current_price))) CUR_GEO_MEAN   from tab
 group by YEAR, MONTH, ITEM_CODE)
select 
 YEAR, MONTH, REGION_CODE, ITEM_CODE, CUR_GEO_MEAN,
 lag(CUR_GEO_MEAN) over (partition by REGION_CODE, ITEM_CODE order by YEAR, MONTH) as PREV_GEO_MEAN
from geo
order by YEAR, MONTH, REGION_CODE, ITEM_CODE; 

You may want to adapt the LAG function with a different partitioning and/or order or if you want to account for the missign months.