SQL - first in first out - sql

Question

I want to implement FIFO in my stock table
table looks like:
---+------------+-----------+--------+--------------+-----------
id | shift_type | item_type | amount | name | date
---+------------+-----------+--------+--------------+-----------
1 | in | apple | 50 | apple type 1 | 2017-12-01
2 | out | apple | 30 | apple type 1 | 2017-12-02
3 | in | apple | 40 | apple type 2 | 2017-12-04
4 | in | apple | 60 | apple type 3 | 2017-12-05
5 | out | apple | 20 | apple type 1 | 2017-12-07
6 | out | apple | 10 | apple type 1 | 2017-12-07
7 | in | apple | 20 | apple type 3 | 2017-12-09
and it keeps info about stock shifts.
If I want to take just the oldest incomes - I can take records with lowest id or oldest date.
But... there are also "out" shifts.
So if I want to take for example 50 apples, query should return me 2 records:
---+------------+-----------+--------+--------------+------------+-----
id | shift_type | item_type | amount | name | date | take
---+------------+-----------+--------+--------------+------------+-----
1 | in | apple | 50 | apple type 1 | 2017-12-01 | 20
3 | in | apple | 40 | apple type 2 | 2017-12-04 | 30
because - first out(id 2) takes 30 apples, so there is 20 left from income id 1, and rest of it should be taken from income id 3
How can I implement it with SQL?

Answer 1

I did something here. But need to test with different inputs.
Given a no, say 50 in the outer select query it will give you list of stacks that you need to take apples from.
WITH view_t
AS (
SELECT 1 id
,'in' shift_type
,'apple' item_type
,50 amount
,'apple type 1' NAME
,TO_DATE('2017-12-01', 'yyyy-mm-dd') date1
FROM dual
UNION ALL
SELECT 2
,'out'
,'apple'
,30
,'apple type 1'
,TO_DATE('2017-12-02', 'yyyy-mm-dd')
FROM dual
UNION ALL
SELECT 3
,'in'
,'apple'
,40
,'apple type 2'
,TO_DATE('2017-12-04', 'yyyy-mm-dd')
FROM dual
UNION ALL
SELECT 4
,'in'
,'apple'
,60
,'apple type 3'
,TO_DATE('2017-12-05', 'yyyy-mm-dd')
FROM dual
UNION ALL
SELECT 5
,'out'
,'apple'
,30
,'apple type 2'
,TO_DATE('2017-12-07', 'yyyy-mm-dd')
FROM dual
)
SELECT id
,date1
,shift_type
,NAME
,itemrem
,nvl((
50 - LAG(itemrem) OVER (
ORDER BY id
)
), itemrem) take
FROM (
SELECT id
,date1
,shift_type
,NAME
,SUM(amtretain) OVER (
ORDER BY id
) itemrem
FROM (
SELECT id
,date1
,shift_type
,amount
,signamt
,NAME
,CASE
WHEN LEAD(shift_type) OVER (
ORDER BY id
) = 'out'
THEN LEAD(signamt) OVER (
ORDER BY id
) + signamt
ELSE signamt
END amtretain
--date1,shift_type,sum(signamt) over(order by id)
FROM (
SELECT id
,shift_type
,NAME
,item_type
,date1
,DECODE(shift_type, 'in', amount, 'out', amount * - 1) signamt
,amount
FROM view_t
)
)
WHERE shift_type = 'in'
);
The idea behind the approach is to,
convert the signs of the amount based on the shift_type
Use lead to subtract the signamt based on its shift type
Use a running total to identify the remaining apples at each date
Finally using LAG with the input no of apples to calculate how much to take