Get running no series on basis of one column value - sql

Question

I have a table like
SELECT str AS company, item#, Qty
FROM temp_on_hand
WHERE qty > 2
ORDER BY Item# ASC
output of that query is -
company item# Qty
1 746 3
5 9526 1
1 14096 1
2 14096 2
3 14095 2
I want to generate new item#( with addition of '-0001' to current item#) on basis of Qty column i.e. if Qty column has value 3 for company 1 than query should return three rows like -
company NewItem# Item# Qty
1 746-00001 746 3
1 746-00002 746 3
1 746-00003 746 3
5 9526-00001 9526 1
1 14096-00001 14096 1
2 14096-00002 14096 2
2 14096-00003 14096 2
3 14095-00001 14095 3
3 14095-00002 14095 3
3 14095-00003 14095 3
. . . . . . .
Table structure like that
CREATE TABLE temp_on_hand(str INT, item# INT,Qty INT)
INSERT INTO temp_on_hand VALUES (1, 746, 3)
INSERT INTO temp_on_hand VALUES (5, 9526, 1)
INSERT INTO temp_on_hand VALUES (1, 14096, 1)
INSERT INTO temp_on_hand VALUES (2, 14096, 2)
INSERT INTO temp_on_hand VALUES (3, 14095, 2)
ALTER TABLE temp_on_hand ADD new_item# VARCHAR)
similarly for upcoming values.
Thanks in advance

Answer 1

You can join to a Numbers table.
You can use a real one, but I will use Itzik Ben-Gan's on-the-fly tally table (it's actually better as an inline Table-valued Function).
EDIT: According to your comments, you don't actually need the numbering from Nums, you want a fresh overall numbering. So you can just select from L1
WITH
L0 AS ( SELECT 1 AS c
FROM (VALUES(1),(1),(1),(1),(1),(1),(1),(1),
(1),(1),(1),(1),(1),(1),(1),(1)) AS D(c) ),
L1 AS ( SELECT 1 AS c FROM L0 A, L0 B ), -- add more cross joins for more rows
SELECT
t.str AS company,
t.item# + FORMAT(ROW_NUMBER() OVER (ORDER BY t.item# ASC), '-0000') NewItem#,
t.item#,
t.Qty
FROM temp_on_hand t
CROSS APPLY(
SELECT TOP (t.Qty) c
FROM L1
) n
WHERE t.qty > 2
ORDER BY t.Item#, n.rownum ASC;
db<>fiddle

Answer 2

The key to good performance using a numbers table based approach is to make sure the row expansion is constrained by a row goal, i.e. SELECT TOP(n), without a row goal the full cartesian product is used. Also, the FORMAT function is known to be slow.
You could try something like this
[EDIT]: The sequence assigned to the NewItem# does not reset for each (startdate, enddate) pair.
drop TABLE if exists #temp_on_hand;
go
CREATE TABLE #temp_on_hand(str INT, item# INT,Qty INT)
INSERT INTO #temp_on_hand VALUES
(1, 746, 3),
(5, 9526, 1),
(1, 14096, 1),
(2, 14096, 2),
(3, 14095, 3);
with
l as (select 1 n from (values (1),(1),(1),(1),(1),(1),(1),(1)) as v(n))
select *, concat_ws('-', item#,
right('00000'+cast(row_number() over (order by (select null)) as varchar(5)), 5)) NewItem#
from #temp_on_hand toh
cross apply (select top (toh.Qty) 1 n
from l l1, l l2,l l3, l l4) tally;
str item# Qty n NewItem#
1 746 3 1 746-00001
1 746 3 1 746-00002
1 746 3 1 746-00003
5 9526 1 1 9526-00004
1 14096 1 1 14096-00005
2 14096 2 1 14096-00006
2 14096 2 1 14096-00007
3 14095 3 1 14095-00008
3 14095 3 1 14095-00009
3 14095 3 1 14095-00010

Answer 3

A recursive CTE is a simple method:
WITH cte as (
SELECT str AS company, item#, Qty, 1 as n
FROM temp_on_hand
WHERE qty > 2
UNION ALL
SELECT company, item#, Qty, n + 1
FROM cte
WHERE n < Qty
)
SELECT str, item# + format(n, '0000') as newitem#, item#, qty
FROM cte;
Note that if qty exceeds 100, you will also need option (maxrecursion 0).
EDIT:
If you want the numbering within a company, you can use window functions and a cumulative sum:
WITH cte as (
SELECT str AS company, item#, Qty, 1 as n,
SUM(qty) OVER (PARTITION BY str ORDER BY item#) - qty as start_qty
FROM temp_on_hand
WHERE qty > 2
UNION ALL
SELECT company, item#, Qty, n + 1, start_qty
FROM cte
WHERE n < Qty
)
SELECT str,
item# + format(n + start_qty, '0000') as newitem#, item#, qty
FROM cte;