Use oracle LAG function data in calculations - sql

Question

I want get STOCK and weighted average unit cost WAUC.
TABLE T1:
ROW
ITEM
IN
OUT
PRICE
STOCK
WAUC
1
A
1000
-
20
1000
20
2
A
2000
-
25
-
-
3
A
1500
-
15
-
-
4
A
500
-
20
-
-
I have the first 5 columns and the first two records of last two columns, and want to get the rest of last two columns STOCK and WAUC.
WAUC = ((PREVIOUS_PRICE * PREVIOUS_STOCK) + (CURRENT_IN * CURRENT_PRICE)) / CURRENT_STOCK
So, I write the query:
SELECT ROW,
SUM(IN - OUT) OVER(ORDER BY ROW) STOCK,
((LAG(STOCK * WAUC) OVER (ORDER BY ROW)) + (IN * PRICE)) / STOCK AS WAUC
FROM T1
What I want is :
ROW
ITEM
IN
OUT
PRICE
STOCK
WAUC
1
A
1000
-
20
1000
20
2
A
2000
-
25
3000
23.33
3
A
1500
-
15
4500
20.55
4
A
500
-
20
5000
20.49
In other words, I want to use LAG results in calculation data.

Answer 1

Your formula should be:
WAUC = (
PREVIOUS_WAUC * PREVIOUS_STOCK
+ (CURRENT_IN - CURRENT_OUT) * CURRENT_PRICE
)
/ CURRENT_STOCK
You can use a MODEL clause (with some extra measurements to make the calculation simpler):
SELECT "ROW", item, "IN", "OUT", price, stock, wauc
FROM t1
MODEL
DIMENSION BY ("ROW")
MEASURES (item, "IN", "OUT", price, 0 AS change, 0 AS stock, 0 AS total, 0 AS wauc)
RULES (
change["ROW"] = COALESCE("IN"[cv()], 0) - COALESCE("OUT"[cv()], 0),
stock["ROW"] = change[cv()] + COALESCE(stock[cv()-1], 0),
total["ROW"] = change[cv()] * price[cv()] + COALESCE(total[cv()-1], 0),
wauc["ROW"] = total[cv()] / stock[cv()]
);
Or, from Oracle 12, using MATCH_RECOGNIZE:
SELECT "ROW",
item,
"IN",
"OUT",
price,
total_stock AS stock,
total_cost / total_stock AS wauc
FROM t1
MATCH_RECOGNIZE(
ORDER BY "ROW"
MEASURES
SUM(COALESCE("IN", 0) - COALESCE("OUT", 0)) AS total_stock,
SUM((COALESCE("IN", 0) - COALESCE("OUT", 0))*price) AS total_cost
ALL ROWS PER MATCH
PATTERN (all_rows+)
DEFINE
all_rows AS 1 = 1
)
Or analytic functions:
SELECT "ROW",
item,
"IN",
"OUT",
price,
SUM(COALESCE("IN",0) - COALESCE("OUT", 0)) OVER (ORDER BY "ROW")
AS stock,
SUM((COALESCE("IN",0) - COALESCE("OUT", 0))*price) OVER (ORDER BY "ROW")
/ SUM(COALESCE("IN",0) - COALESCE("OUT", 0)) OVER (ORDER BY "ROW")
AS wauc
FROM t1
Which, for the sample data:
CREATE TABLE t1 ("ROW", ITEM, "IN", "OUT", PRICE, STOCK, WAUC) AS
SELECT 1, 'A', 1000, CAST(NULL AS NUMBER), 20, CAST(NULL AS NUMBER), CAST(NULL AS NUMBER) FROM DUAL UNION ALL
SELECT 2, 'A', 2000, NULL, 25, NULL, NULL FROM DUAL UNION ALL
SELECT 3, 'A', 1500, NULL, 15, NULL, NULL FROM DUAL UNION ALL
SELECT 4, 'A', 500, NULL, 20, NULL, NULL FROM DUAL;
All output:
ROW
ITEM
IN
OUT
PRICE
STOCK
WAUC
1
A
1000
20
1000
20
2
A
2000
25
3000
23.33333333333333333333333333333333333333
3
A
1500
15
4500
20.55555555555555555555555555555555555556
4
A
500
20
5000
20.5
Note: ROW, IN and OUT are keywords and you should not use them as identifiers as you would have to use quoted identifiers everywhere they occur.
db<>fiddle here

Answer 2

Your problem has nothing to do with "lag".
Using MT0's sample data:
select "ROW", item, "IN", "OUT", price,
sum(nvl("IN", 0) - nvl("OUT", 0))
over (partition by item order by "ROW") as stock,
round(sum((nvl("IN", 0) - nvl("OUT", 0)) * price)
over (partition by item order by "ROW")
/ sum(nvl("IN", 0) - nvl("OUT", 0))
over (partition by item order by "ROW"), 2 ) as wauc
from t1
;
ROW ITEM IN OUT PRICE STOCK WAUC
------ ---- ------ ------ ------ ------ ------
1 A 1000 20 1000 20
2 A 2000 25 3000 23.33
3 A 1500 15 4500 20.56
4 A 500 20 5000 20.5