How to track iterations of a value in sql - sql

Question

I am trying to track usage of a blade in a manufacturing process using SSMS 2017. The blade is loaded and used on product until it is seen to dull and then taken out for sharpening while another blade replaces it. We have 30 blades that are cycled for use and sharpening.
Using a table that provides product lot number (sequential) and blade name I would like to separate each batch use of the blade name into groups.
My sql skills are pretty basic so I've been trying row_number, rank, and some attempts at utilizing the lead/lag functions. So far this has only enabled me to break down each product into order based on blade name and identify the product on which a blade change is made. I feel like that could be useful but I'm having trouble figuring out exactly how to do it.
I would like to be able to assign each group of product manufactured with an iteration of a blade a identifying number. For example:
LotNo BladeID Iteration
418211 BH40 1
418212 BH40 1
418213 BH40 1
418214 ES11 2
418215 ES11 2
418216 BH40 3
I'm currently able to produce these incorrect results:
Using:
SELECT b.LotNo,
b.BladeID,
ROW_NUMBER() OVER (PARTITION BY b.BladeID ORDER BY b.BladeID)
FROM blades AS b
ORDER BY b.LotNo ASC;
I get:
LotNo BladeID Iteration
418211 BH40 1
418212 BH40 2
418213 BH40 3
418214 ES11 1
418215 ES11 2
418216 BH40 4

Answer 1

Here's a possible solution to your problem. It first creates a series of groups to identify when the number must change. Then it gets an order to assign the correct value to each group. And finally, it assigns the value for the iteration. I'm including the sample data in a consumable way so anyone can use it for testing purposes.
CREATE TABLE #Sample(
LotNo int,
BladeID varchar(10),
Iteration int
);
INSERT INTO #Sample
VALUES
(418211, 'BH40', 1),
(418212, 'BH40', 1),
(418213, 'BH40', 1),
(418214, 'ES11', 2),
(418215, 'ES11', 2),
(418216, 'BH40', 3);
GO
WITH cteGroups AS(
SELECT *,
ROW_NUMBER() OVER(ORDER BY LotNo) - ROW_NUMBER() OVER(PARTITION BY BladeID ORDER BY LotNo) AS island
FROM #Sample
),
cteOrdering AS(
SELECT *, MIN( LotNo) OVER( PARTITION BY island, BladeID) AS OrderCol
FROM cteGroups
)
SELECT LotNo,
BladeID,
Iteration,
DENSE_RANK() OVER( ORDER BY OrderCol) AS IterationCalc
FROM cteOrdering;

Answer 2

You can do this with lag() and a cumulative sum:
select s.*,
sum(case when prev_BladeID = BladeId then 0 else 1 end) over (order by LotNo) as Iteration
from (select s.*,
lag(s.BladeID) over (order by s.LotNo) as prev_BladeID
from #sample s
) s;
In addition to being simpler code than the difference of row numbers, I think this is also simpler conceptually. This is simply counting the number of times that the BladeID changes from one lot to the next.