looking for cursor replacement - sql

Question

I have 2 tables #TblFinal and #TblData.. I need to compare #TblData data with"#TblFinal based on unique key F_U_KEY vs D_U_KEY.
if #TblData data for a unique key is not exist in #TblFinal, then #TblData data record will insert into #TblFinal table. F_COUNTER will be 1 (default) and F_IS_CLEAR is 0 default
if #TblData data for a unique key is exist in #TblFinal, then first we get key F_ID from #TblFinal
case A: if D_SEV_ID of table "#TblData" is not equal to -11
then need to update "F_COUNTER" for table "#TblFinal" for "F_ID"
case B: if D_SEV_ID of table "#TblData" is equal to -11
then new entry for D_SEV_ID = -11 and F_IS_CLEAR =1 for "F_ID"
Here is the cursor I wrote, need some optimized solution as my actual #TblFinal having huge data and data compare table "#TblData" always have 100 records to compare.
CREATE TABLE #TblFinal
(F_ID INT IDENTITY(1,1), F_VAL NVARCHAR(20), F_SEV_ID INT, F_U_KEY NVARCHAR(200), F_COUNTER INT DEFAULT(1), F_IS_CLEAR BIT DEFAULT(0))
CREATE TABLE #TblData
(D_ID INT, D_VAL NVARCHAR(20), D_SEV_ID INT, D_U_KEY NVARCHAR(200))
INSERT INTO #TblData VALUES(1, 'test 1', 2, '1:100002135::::15124:9334'), (2, 'test 1', 2, '1:100002135::::15124:9334'),
(3, 'test', -11, '1:100002135::::15124:9334'), (4, 'test 1', 2, '1:100002135::::15124:9334'), (5, 'test 1', 2, '1:1024:9334')
DECLARE #D_ID INT
DECLARE #D_SEV_ID INT
DECLARE #D_U_KEY NVARCHAR(200)
DECLARE A_CUR CURSOR FOR
SELECT D_ID, D_SEV_ID, D_U_KEY FROM #TblData ORDER BY D_ID ASC
OPEN A_CUR
FETCH NEXT FROM A_CUR INTO #D_ID, #D_SEV_ID, #D_U_KEY
WHILE ##FETCH_STATUS = 0
BEGIN
IF EXISTS(SELECT 1 FROM #TblFinal(NOLOCK) WHERE F_U_KEY = #D_U_KEY AND F_IS_CLEAR = 0 AND F_SEV_ID <> -11)
BEGIN
DECLARE #FId INT
SELECT #FId = F_ID FROM #TblFinal(NOLOCK) WHERE F_U_KEY = #D_U_KEY AND F_IS_CLEAR = 0 AND F_SEV_ID <> -11
--IF #D_SEV_ID != -11
IF (#D_SEV_ID <> -11)
BEGIN
UPDATE #TblFinal
SET F_COUNTER = F_COUNTER + 1 WHERE F_ID = #FId
END
--IF #D_SEV_ID = -11
ELSE IF(#D_SEV_ID = -11)
BEGIN
INSERT INTO #TblFinal (F_VAL, F_SEV_ID, F_U_KEY)
SELECT D_VAL, D_SEV_ID, D_U_KEY FROM #TblData(NOLOCK) WHERE D_ID = #D_ID
UPDATE #TblFinal
SET F_IS_CLEAR = 1 WHERE F_ID = #FId
END
ELSE
BEGIN
PRINT 'DO NOTHING'
END
END
ELSE
BEGIN
INSERT INTO #TblFinal (F_VAL, F_SEV_ID, F_U_KEY)
SELECT D_VAL, D_SEV_ID, D_U_KEY FROM #TblData(NOLOCK) WHERE D_ID = #D_ID
END
FETCH NEXT FROM A_CUR INTO #D_ID, #D_SEV_ID, #D_U_KEY
END
CLOSE A_CUR
DEALLOCATE A_CUR
--SELECT * FROM #TblData
SELECT * FROM #TblFinal
DROP TABLE #TblData
DROP TABLE #TblFinal

Answer 1

It does not look like a set-based approach will work for you since the order of your data looks important. That said, there are some things you can do which should speed up your query by about 50%.
Loop from min(D_ID) to max(D_ID), incrementing by 1.
Only look for the item in #TblFinal once instead of once in an exists and then again to get the #FId. Also, make sure #TblFinal has an index on F_U_KEY + F_IS_CLEAR + F_SEV_ID and that F_ID is either part of the clustered index or within an INCLUDE clause. Ideally it is the first field in the clustered index for the updates to be fast.
By storing D_VAL in a variable, you can insert directly into #TblFinal from variables rather than having to SELECT again from #TblData.
Here's the code:
DECLARE #D_ID INT
DECLARE #FId INT
DECLARE #D_SEV_ID INT
DECLARE #D_U_KEY NVARCHAR(200)
DECLARE #D_VAL NVARCHAR(20)
DECLARE #Min_D_ID INT, #Max_D_ID INT
SELECT #Min_D_ID = min(D_ID), #Max_D_ID = max(D_ID) FROM #TblData(NOLOCK)
SELECT #D_ID = #Min_D_ID
WHILE #D_ID <= #Max_D_ID
BEGIN
SELECT #D_SEV_ID = D_SEV_ID, #D_U_KEY = D_U_KEY, #D_VAL = D_VAL FROM #TblData(NOLOCK) WHERE D_ID = #D_ID
IF #D_U_KEY IS NOT NULL
BEGIN
SELECT #FId = F_ID FROM #TblFinal(NOLOCK) WHERE F_U_KEY = #D_U_KEY AND F_IS_CLEAR = 0 AND F_SEV_ID <> -11
IF #FId IS NOT NULL
BEGIN
IF (#D_SEV_ID <> -11)
BEGIN
UPDATE #TblFinal
SET F_COUNTER = F_COUNTER + 1 WHERE F_ID = #FId
END
ELSE IF(#D_SEV_ID = -11)
BEGIN
INSERT INTO #TblFinal (F_VAL, F_SEV_ID, F_U_KEY)
SELECT #D_VAL, #D_SEV_ID, #D_U_KEY
UPDATE #TblFinal
SET F_IS_CLEAR = 1 WHERE F_ID = #FId
END
ELSE
BEGIN
PRINT 'DO NOTHING'
END
END
ELSE
BEGIN
INSERT INTO #TblFinal (F_VAL, F_SEV_ID, F_U_KEY)
SELECT #D_VAL, #D_SEV_ID, #D_U_KEY
END
END
SELECT #D_ID = #D_ID + 1, #FId = NULL
END

Answer 2

You should be able to do this set based. Here is one way to tackle this problem. The only part I don't really understand is F_IS_CLEAR. The business rule for that is a bit fuzzy. If you need help with part I need to understand what the rationale is for that.
CREATE TABLE #Output
(F_ID INT IDENTITY(1,1), F_VAL NVARCHAR(20), F_SEV_ID INT, F_U_KEY NVARCHAR(200), F_COUNTER INT DEFAULT(1), F_IS_CLEAR BIT DEFAULT(0))
insert #Output
select x.D_VAL
, x.D_SEV_ID
, x.D_U_KEY
, x.F_COUNTER
, x.F_IS_CLEAR
from
(
select D_VAL = MIN(D_VAL)
, D_SEV_ID = MIN(D_SEV_ID)
, D_U_KEY
, F_COUNTER = COUNT(*)
, F_IS_CLEAR = 1
, SortOrder = 1
from #TblData d
where D_ID < (select d2.D_ID from #TblData d2 where d2.D_U_KEY = d.D_U_KEY and d2.D_SEV_ID = -11)
group by d.D_U_KEY
UNION ALL
select D_VAL
, D_SEV_ID
, D_U_KEY
, F_COUNTER = 1
, F_IS_CLEAR = 0
, SortOrder = 2
from #TblData d
where d.D_SEV_ID = -11
UNION ALL
select D_VAL = MIN(D_VAL)
, D_SEV_ID = MIN(D_SEV_ID)
, D_U_KEY
, F_COUNTER = COUNT(*)
, F_IS_CLEAR = 0
, SortOrder = 3
from #TblData d
where D_ID > (select d2.D_ID from #TblData d2 where d2.D_U_KEY = d.D_U_KEY and d2.D_SEV_ID = -11)
group by d.D_U_KEY
UNION ALL
select D_VAL = MIN(D_VAL)
, D_SEV_ID = MIN(D_SEV_ID)
, D_U_KEY
, F_COUNTER = COUNT(*)
, F_IS_CLEAR = 0
, SortOrder = 4
from #TblData d
where NOT EXISTS (select d2.D_ID from #TblData d2 where d2.D_U_KEY = d.D_U_KEY and d2.D_SEV_ID = -11)
group by d.D_U_KEY
) x
order by D_U_KEY, SortOrder
select * from #Output