T-SQL Group by multiple batches of 2 - sql-server

Question

So the wonderful people here on stackoverflow helped me with a "find consecutive failures" type query. (Status =4 is a failure). I thought I had cracked the second part of my problem because my test case seems to work fine but whenever I run it on our test environment I get dodgy results, so I must be doing something wrong. The goal is to find X number of consecutive failures. So the below is set to find 2 consecutive failures. I'm using SQL Server 2008 R2
DECLARE #t TABLE (
[InstructionId] INT,
[InstructionDetailId] INT,
[Sequence] INT,
[Status] INT
)
INSERT INTO #t SELECT 222,111,1, 2
INSERT INTO #t SELECT 222,112,2,2
INSERT INTO #t SELECT 222,113,3,4
INSERT INTO #t SELECT 222,114,4,4
INSERT INTO #t SELECT 222,115,5,2
INSERT INTO #t SELECT 222,116,6,4
INSERT INTO #t SELECT 222,117,7,2
INSERT INTO #t SELECT 222,118,8,4
INSERT INTO #t SELECT 222,119,9,4
INSERT INTO #t SELECT 222,120,10,2
INSERT INTO #t SELECT 222,121,11,2
INSERT INTO #t SELECT 222,124,12,4
INSERT INTO #t SELECT 222,126,13,4
INSERT INTO #t SELECT 222,128,14,4
INSERT INTO #t SELECT 223,126,13,4
INSERT INTO #t SELECT 223,128,14,4
INSERT INTO #t SELECT 223,129,15,2
INSERT INTO #t SELECT 223,130,16,4
INSERT INTO #t SELECT 224,111,17,4
INSERT INTO #t SELECT 224,112,18,4
INSERT INTO #t SELECT 223,160,33,4
INSERT INTO #t SELECT 223,161,34,4
INSERT INTO #t SELECT 223,162,35,4
INSERT INTO #t SELECT 223,163,40,4
;with HardcoreCTE AS
(
select t.*,
t.[Sequence] - ROW_NUMBER() OVER(PARTITION BY t.instructionId ORDER BY
t.InstructionDetailId) AS ItemCount
from #t t outer apply
( select top (1) t1.*
from #t t1
where t1.InstructionId = t.InstructionId and
t1.Sequence < t.Sequence
order by t1.Sequence desc
) t1 outer apply
( select top (1) t2.*
from #t t2
where t2.InstructionId = t.InstructionId and
t2.Sequence > t.Sequence
order by t2.Sequence
) t2
where t.status = 4 and (t.status = t1.status or t.status = t2.status)
)
,
HardCoreCTE2
AS
(
select *, Count(1) OVER(PARTITION BY ItemCount) AS ItemCount2 from
HardcoreCTE
)
select * from HardCoreCTE2
where ItemCount2 =2
So the above works brilliants to find results where there are specifically only 2 consecutive failures with these
results:
Now from the above results the only ones it finds are the records where there are 2 consecutive failures but whenever I convert the above to the actual test environment tables it doesn't seem to work.
Test Env Results: As you can see for the "InstructionId" of 2518380 it brought back one record and the for "InstructionId" 2614351. It's meant to bring back sets of 2 records.
Test Env Query: (Pretty much identical)
;with InitialDataCTE
AS
(
SELECT Instruction.InstructionID,InstructionDetail.InstructionDetailID,
InstructionDetail.InstructionDetailStatusID AS [Status],
InstructionDetail.Sequence
FROM Instruction INNER JOIN
InstructionDetail ON Instruction.InstructionID =
InstructionDetail.InstructionID
where InstructionDetailStatusID =4
and InstructionDetail.PaymentDateOriginal between '2015-01-05'
AND '2018-09-08'
),
HardCoreCTE
AS
(
select t.*,
t.Sequence - ROW_NUMBER() OVER(PARTITION BY t.instructionId ORDER BY
t.InstructionDetailId) AS ItemCount
from InitialDataCTE t outer apply
( select top (1) t1.*
from InitialDataCTE t1
where t1.InstructionId = t.InstructionID and
t1.Sequence < t.Sequence
order by t1.Sequence desc
) t1 outer apply
( select top (1) t2.*
from InitialDataCTE t2
where t2.InstructionId = t.InstructionId and
t2.Sequence > t.Sequence
order by t2.Sequence
) t2
where t.Status = 4 and (t.Status = t1.Status or t.Status = t2.Status)
)
,
HardCoreCTE2
AS
(
select *, Count(1) OVER(PARTITION BY ItemCount) AS ItemCount2 from
HardCoreCTE
)
select * from HardCoreCTE2
where ItemCount2 =2
order by InstructionID, Sequence
Really appreciate if someone can tell me where I am going wrong, I've been messing around with variations of the Count(*) but nothing successful yet. Thanx alot

Answer 1

I came to the next query:
with
a as (
select *,
row_number() over(partition by InstructionId order by Sequence)-
row_number() over(partition by InstructionId, [Status] order by Sequence) g
from #t
),
b as (
select *,
count(*) over(partition by InstructionId, [Status], g) c
from a
where [Status] = 4
)
select *
from b
where c > 2
order by 1, 3;
For your test data, I got the following result:
InstructionId InstructionDetailId Sequence Status g c
222 224 312 4 6 3
222 226 413 4 6 3
222 228 514 4 6 3
223 161 84 4 2 3
223 162 95 4 2 3
223 163 140 4 2 3
You can test this query here.