Update sequence based on pattern SQL Server - sql-server

Question

I was not able to quote correct tile for the question.
Below is my table.
Expected Output : With ID and SequenceNo column. Was unable to upload image.
ID Act SequenceNo1
1 1 1
2 2 NULL
3 3 NULL
4 4 NULL
5 1 5
6 2 NULL
7 3 NULL
8 4 NULL
9 5 NULL
10 6 NULL
11 5 11
12 6 NULL
13 1 13
14 2 NULL
15 3 NULL
16 4 NULL
17 5 NULL
18 6 NULL
19 1 19
20 2 NULL
21 3 NULL
22 4 NULL
23 5 NULL
24 6 NULL
At the start last column SequenceNo is NULL.`
My requirement is to update value of ID column to SequenceNo column whenever new series of Act column is started.
Act column has value of 1 to 6. There can be case where any number from 1 to 6 is missing from Act.
Example1 : ID 1 to 4 - Act is correct but, in next row (ID=5) Act is restarted. Hence need to update SequenceNo column.
Example2 : ID 5 to 10 are correct. But next row (ID=11;Act=5) has new sequence hence need to update SequenceNo column.
CREATE TABLE #tmp
(
ID int
, ScheduleID varchar(50)
,DCNumber VARCHAR(50)
, BuildingID varchar(10)
, StoreNumber int
, [DayOfWeek] int
, [Tm] varchar(10)
,[Act] int
, SequenceNo int
)
INSERT INTO #tmp SELECT 1,'WAS',9003,900301,254,1,'00:00',1,NULL
INSERT INTO #tmp SELECT 2,'WAS',9003,900301,254,1,'00:00',2,NULL
INSERT INTO #tmp SELECT 3,'WAS',9003,900301,254,1,'00:00',3,NULL
INSERT INTO #tmp SELECT 4,'WAS',9003,900301,254,1,'00:00',4,NULL
INSERT INTO #tmp SELECT 5,'WAS',9003,900301,254,2,'00:00',1,NULL
INSERT INTO #tmp SELECT 6,'WAS',9003,900301,254,2,'00:00',2,NULL
INSERT INTO #tmp SELECT 7,'WAS',9003,900301,254,2,'00:00',3,NULL
INSERT INTO #tmp SELECT 8,'WAS',9003,900301,254,2,'00:00',4,NULL
INSERT INTO #tmp SELECT 9,'WAS',9003,900301,254,2,'00:00',5,NULL
INSERT INTO #tmp SELECT 10,'WAS',9003,900301,254,2,'00:00',6,NULL
INSERT INTO #tmp SELECT 11,'WAS',9003,900301,254,3,'00:00',5,NULL
INSERT INTO #tmp SELECT 12,'WAS',9003,900301,254,3,'00:00',6,NULL
INSERT INTO #tmp SELECT 13,'WAS',9003,900301,254,4,'00:00',1,NULL
INSERT INTO #tmp SELECT 14,'WAS',9003,900301,254,4,'00:00',2,NULL
INSERT INTO #tmp SELECT 15,'WAS',9003,900301,254,4,'00:00',3,NULL
INSERT INTO #tmp SELECT 16,'WAS',9003,900301,254,4,'00:00',4,NULL
INSERT INTO #tmp SELECT 17,'WAS',9003,900301,254,5,'00:00',5,NULL
INSERT INTO #tmp SELECT 18,'WAS',9003,900301,254,5,'00:00',6,NULL
INSERT INTO #tmp SELECT 19,'WAS',9003,900301,254,6,'00:00',1,NULL
INSERT INTO #tmp SELECT 20,'WAS',9003,900301,254,6,'00:00',2,NULL
INSERT INTO #tmp SELECT 21,'WAS',9003,900301,254,6,'00:00',3,NULL
INSERT INTO #tmp SELECT 22,'WAS',9003,900301,254,6,'00:00',4,NULL
INSERT INTO #tmp SELECT 23,'WAS',9003,900301,254,7,'00:00',5,NULL
INSERT INTO #tmp SELECT 24,'WAS',9003,900301,254,7,'00:00',6,NULL
I have build one logic but that is time consuming.
DECLARE #Act INT, #iStart INT, #iMax INT, #iFirst INT
SET #iFirst = 7
SET #iMax = (SELECT MAX(ID) FROM #tmp)
SET #iStart = 1
WHILE(#iStart <= #iMax)
BEGIN
SET #Act = (SELECT Act FROM #tmp WHERE ID = #iStart)
IF(#iFirst > #Act )
BEGIN
UPDATE #tmp SET SequenceNo = #iStart WHERE ID = #iStart
END
SET #iFirst = #Act
SET #iStart = #iStart + 1
END
I am looking out for any alternative optimized solution.

Answer 1

Is this what you wanted?
update tupd
set
SequenceNo = tupd.ID
from (
select
*
, LAG(t.Act, 1, null) over(order by t.ID) as LagAct
from #tmp t
) tt
inner join #tmp tupd on tt.ID = tupd.ID
where tt.Act < tt.LagAct
select
*
from #tmp tt
order by tt.ID

Answer 2

You can find breaks in the Act sequence if you compare the current row with the previous. In current SQL Server versions, ie 2012+ you can do that with the LAG() method. In previous versions, you have to perform a self join on the current ID value and its previous value. If the difference between the current and previous Act values isn't 1, there is a break in the sequence. This only works if there are no gaps in ID.
This query will return 1 in the SeqBreak column for each row where Act restarts
select
t1.ID,
t1.SequenceNo ,
case when t1.act = t2.act+1 then '0' else '1' end as SeqBreak
from #tmp t1 left join #tmp t2 on t1.id=t2.id+1
You can use this in a CTE to select only the rows where Act breaks. You can update the CTE directly, at least in SQL Server 2014.
with x as (
select
t1.ID,
t1.SequenceNo ,
case when t1.act = t2.act+1 then '0' else '1' end as SeqBreak
from #tmp t1 left join #tmp t2 on t1.id=t2.id+1 )
update x
set SequenceNo=x.ID
where seqbreak=1 and id>1
If that doesn't work with SQL Server 2008, you'll have to join between the cte and the table:
with x as (
select
t1.ID,
t1.SequenceNo ,
case when t1.act = t2.act+1 then '0' else '1' end as SeqBreak
from #tmp t1 left join #tmp t2 on t1.id=t2.id+1 )
update #tmp
set SequenceNo=x.ID
from #tmp inner join x on x.ID=#tmp.ID
where seqbreak=1 and #tmp.id>1
The equivalent query using LAG(), one of the windowing functions in SQL Server 2012 is simpler and twice as fast, since it avoids the self join:
with x as (
select
ID,
SequenceNo ,
case when act = 1 +LAG(act,1) OVER (ORDER BY ID) then '0' else '1' end as SeqBreak
from #tmp)
update x
set SequenceNo=x.ID
where seqbreak=1 and id>1

Answer 3

;WITH CTE
AS
(
SELECT ID,Act,RANK() OVER (PARTITION BY [DayOfWeek] ORDER BY Act) RN
FROM #tmp
)
UPDATE t
SET t.SequenceNo = t.id
FROM #tmp t
inner join CTE C
ON t.ID = C.ID
WHERE C.RN = 1

Answer 4

If I'm getting your question correctly below query should fetch the desired result.
update t3 set sequenceno = t3.id
from
#temp t1
cross apply (select id+1 as idsum
from #temp
tmp where t1.id = tmp.id)t2
inner join #temp t3
on t3.id = t2.idsum
where t1.act>t3.act