Related
The Problem
I need to update a table so that any duplicate rows are updated to have unique values.
The Catch
I need to dynamically ensure that the value I am updating the duplicate row to is also unique.
My Solution So Far (with test case)
CREATE TABLE #temp (name nvarchar(100), ID uniqueidentifier)
INSERT INTO #temp (Name, ID)
VALUES ('Duplicate', '32208C09-C0C3-408C-AB60-273811722194')
INSERT INTO #temp (Name, ID)
VALUES ('Duplicate', '32208C09-C0C3-408C-AB60-273811722194')
INSERT INTO #temp (Name, ID)
VALUES ('Duplicate (2)', '32208C09-C0C3-408C-AB60-273811722194')
;WITH cte AS (
SELECT Name
, ROW_NUMBER() OVER (PARTITION BY Name, ID ORDER BY Name) RowNum
FROM #temp
)
UPDATE cte
SET Name = CONCAT(Name, ' (', RowNum, ')')
WHERE RowNum > 1
SELECT * FROM #temp
DROP TABLE #temp
As you can tell, this will update the table so there is only one row with the name 'Duplicate' but two rows with the name 'Duplicate (2)'. How can I check and account for duplicates in the value I am updating to?
You could use another CTe which gets you the highest Number and then use that to generate the "next" number.
for 2 or more digits you need to adapt it
CREATE TABLE #temp (name nvarchar(100), AssetMakeID uniqueidentifier)
INSERT INTO #temp (Name, AssetMakeID)
VALUES ('Duplicate', '32208C09-C0C3-408C-AB60-273811722194')
INSERT INTO #temp (Name, AssetMakeID)
VALUES ('Duplicate', '32208C09-C0C3-408C-AB60-273811722194')
INSERT INTO #temp (Name, AssetMakeID)
VALUES ('Duplicate (2)', '32208C09-C0C3-408C-AB60-273811722194')
;WITH CTE1 AS (SELECT
MAX( COALESCE(REPLACE(REPLACE(SUBSTRING(name, PATINDEX('%([0-9])%', name), PATINDEX('%)%', name + 't') - PATINDEX('%(%',
name) + 1),'(','') ,')','') ,0)) hinum
,SUBSTRING(name,1, PATINDEX('% ([0-9])%', name) ) name
FROM #temp
WHERE SUBSTRING(name,1, PATINDEX('% ([0-9])%', name) ) IS NOT NULL
GROUP BY SUBSTRING(name,1, PATINDEX('% ([0-9])%', name) ) ),
cte AS (
SELECT #temp.Name
, CASE WHEN ROW_NUMBER() OVER (PARTITION BY #temp.Name, AssetMakeID ORDER BY #temp.Name) > 1 THEN
ROW_NUMBER() OVER (PARTITION BY #temp.Name, AssetMakeID ORDER BY #temp.Name) + hinum -1
ELSe ROW_NUMBER() OVER (PARTITION BY #temp.Name, AssetMakeID ORDER BY #temp.Name) END RowNum
FROM #temp LEFT JOIN CTE1 ON #temp.name = CTE1.name
)
UPDATE cte
SET Name = CONCAT(Name, ' (', RowNum, ')')
WHERE RowNum > 1
SELECT * FROM #temp
name
AssetMakeID
Duplicate
32208c09-c0c3-408c-ab60-273811722194
Duplicate (3)
32208c09-c0c3-408c-ab60-273811722194
Duplicate (2)
32208c09-c0c3-408c-ab60-273811722194
fiddle
Well the easy way is to use a unique string in the update so there is no way your update can cause a duplicate. The current timestamp (with milliseconds) works well. Like this:
UPDATE cte
SET Name = CONCAT(Name, ' (', RowNum, ') at ',convert(varchar(22),getdate(),126))
WHERE RowNum > 1
This will cope with one level of duplication e.g. 'Duplicate (2)' but not two e.g. 'Duplicate (2) (2)'.
Essentially just apply the same logic again in a second cte. In fact you should be able to do this using a recursive CTE to get it to work for all levels.
That said you could use a more unique method of de-duplicating names e.g. just add a guid and it will be unique.
WITH cte1 AS (
SELECT Name, Id
, ROW_NUMBER() OVER (PARTITION BY Name, ID ORDER BY Name) RowNum
-- You should already have one, but if not generate it
, ROW_NUMBER() OVER (ORDER BY Name) UniqueId
FROM #temp
), cte2 as (
SELECT NewName Name, RowNum, UniqueId
, ROW_NUMBER() OVER (PARTITION BY NewName, ID ORDER BY NewName) RowNum2
FROM cte1
CROSS APPLY (
VALUES (CASE WHEN RowNum = 1 THEN Name ELSE CONCAT(Name, ' (', RowNum, ')') END)
) n (NewName)
)
UPDATE c1 SET
Name = CASE WHEN RowNum2 = 1 THEN c2.Name ELSE CONCAT(c2.Name, ' (', RowNum2, ')') END
FROM cte1 c1
INNER JOIN cte2 c2 on c2.UniqueId = c1.UniqueId
WHERE c1.RowNum > 1 or RowNum2 > 1;
I am going to choose another answer as the correct answer since I personally prefer it, but I thought I'd post what I ended up doing myself.
DROP TABLE IF EXISTS #temp
CREATE TABLE #temp (name nvarchar(100), ID uniqueidentifier)
INSERT INTO #temp (Name, ID)
VALUES ('Duplicate', '32208C09-C0C3-408C-AB60-273811722194')
INSERT INTO #temp (Name, ID)
VALUES ('Duplicate', '32208C09-C0C3-408C-AB60-273811722194')
INSERT INTO #temp (Name, ID)
VALUES ('Duplicate (2)', '32208C09-C0C3-408C-AB60-273811722194')
DECLARE #doWhileTrueFlag bit = 1
WHILE (#doWhileTrueFlag = 1)
BEGIN
;WITH cte AS (
SELECT
Name,
ROW_NUMBER() OVER (PARTITION BY Name, ID ORDER BY Name) RowNum
FROM #temp
)
UPDATE cte
SET Name = CONCAT(Name, ' (', RowNum, ')')
WHERE RowNum > 1
SET #doWhileTrueFlag = CASE
WHEN ##ROWCOUNT > 0 THEN 1
ELSE 0
END
END
SELECT * FROM #temp
DROP TABLE #temp
This performs the update I was already doing in a loop until no more updates are done. A rather inelegant solution, but the names created are prettier for the clients.
I have a scenario wherein I need to find the ID which only has start and END in it. Below is the table for reference.
Declare #T Table ( ID int, Name varchar(100))
Insert into #T values (1,'Start')
Insert into #T values (1,'END')
Insert into #T values (1,'Stuart')
Insert into #T values (1,'robin')
Insert into #T values (2,'Start')
Insert into #T values (2,'END')
Insert into #T values (3,'Start')
Insert into #T values (4,'END')
I want the Output as:
ID Name
2 Start
2 END
I want those ID which only has start and end in it.
What I tried so far:
SELECT * FROM #T t
WHERE EXISTS (SELECT * FROM #T WHERE id = t.id AND name = 'start')
AND EXISTS (SELECT * FROM #T WHERE id = t.id AND name = 'END')
But my query is giving ID 1 as well.
Can someone please help me rectify the problem.
I presume your issue is that record 1 has a 'Stuart' in it too?
As such, you can do a similar check in the WHERE e.g.,
SELECT * FROM #T t
WHERE EXISTS (SELECT * FROM #T WHERE id = t.id AND name = 'start')
AND EXISTS (SELECT * FROM #T WHERE id = t.id AND name = 'END')
AND NOT EXISTS (SELECT * FROM #T WHERE id = t.id AND name NOT IN ('start','END'))
Note that you may want to consider
What happens if you have two 'start' rows or two 'end' rows (e.g., start-start-end)? Can you even have two 'start' rows (e.g., start-start)?
What happens if you have a blank/NULL (e.g., start-NULL-end)?
EDIT: removed 'What happens if they're out of order (e.g., end-start)?' as a question as there is no sorting in the data at all (e.g., not even an implicit sort).
You can go for CTE to get group wise count and total count as 2.
Declare #T Table ( ID int, Name varchar(100))
Insert into #T values (1,'Start')
Insert into #T values (1,'END')
Insert into #T values (1,'Stuart')
Insert into #T values (1,'robin')
Insert into #T values (2,'Start')
Insert into #T values (2,'END')
Insert into #T values (3,'Start')
Insert into #T values (4,'END')
;WITH CTE_Total_StartEnd AS
(
select id, count(*) AS Total_Cnt
, COUNT( case when Name IN ('Start') THEN 1 END) as start_cnt
, COUNT( case when Name IN ('End') THEN 1 END) as end_cnt
from #t
group by id
having COUNT( case when Name IN ('Start') THEN 1 END) =1 and
COUNT( case when Name IN ('End') THEN 1 END) = 1 and
count(*) = 2
)
SELECT t.* from #t t
inner join CTE_Total_StartEnd as c
ON c.id = t.id
+----+-------+
| ID | Name |
+----+-------+
| 2 | Start |
| 2 | END |
+----+-------+
You can do this by using group by function also like below
WITH cte AS
(
SELECT 1 AS id , 'Start' AS name
UNION ALL
SELECT 1 AS id ,'END' AS name
UNION ALL
SELECT 1 AS id ,'Stuart' AS name
UNION ALL
SELECT 1 AS id ,'robin' AS name
UNION ALL
SELECT 2 AS id ,'Start' AS name
UNION ALL
SELECT 2 AS id ,'END' AS name
UNION ALL
SELECT 3 AS id ,'Start' AS name
UNION ALL
SELECT 4 AS id ,'END' AS name
)
SELECT T.ID,SUM(T.VAL)AS SUM
FROM
(
SELECT id,name , CASE WHEN name='Start' THEN 1
WHEN name='END' THEN 2
ELSE 3
END AS VAL
FROM cte
)T
GROUP BY T.ID
HAVING SUM(T.VAL) =3
could you please try this? Pls note i added collate command in the end of sql.
SQL Server check case-sensitivity?
SELECT * FROM #T t
WHERE EXISTS (SELECT * FROM #T WHERE id = t.id AND name = 'start' COLLATE SQL_Latin1_General_CP1_CS_AS)
AND EXISTS (SELECT * FROM #T WHERE id = t.id AND name = 'END' COLLATE SQL_Latin1_General_CP1_CS_AS)
I have a scenario where I have to select multiple rows from table, I have multiple rows of one record but with different status,
at times I have two identical rows with identical data for status < for that case I canted to select Non zero for the first occurrence and set 0 for the remaining occurrences.
Below is the Image to show and I have marked strike-out and marked 0 for the remaining occurrence.
And body could suggest better SQL Query:
Here is the Query: I am getting zero value for status 1 for ID =1 but I need to show first as regular and then 0 if that status repeats again.
CREATE TABLE #Temp
(ID INT,
ItemName varchar(10),
Price Money,
[Status] INT,
[Date] Datetime)
INSERT INTO #Temp VALUES(1,'ABC',10,1,'2014-08-27')
INSERT INTO #Temp VALUES(1,'ABC',10,2,'2014-08-27')
INSERT INTO #Temp VALUES(1,'ABC',10,1,'2014-08-28')
INSERT INTO #Temp VALUES(2,'DEF',25,1,'2014-08-26')
INSERT INTO #Temp VALUES(2,'DEF',25,3,'2014-08-27')
INSERT INTO #Temp VALUES(2,'DEF',25,1,'2014-08-28')
INSERT INTO #Temp VALUES(3,'GHI',30,1,'2014-08-27')
SELECT CASE WHEN Temp.Status = 1 THEN
0
ELSE
Temp.Price END AS Price,
* FROM (SELECT * FROM #Temp) Temp
DROP TABLE #Temp
Here is the result:
You might modify your inner select using Row_Number() and set price to Zero for RowNumber > 1.
SELECT CASE WHEN Temp.RowNumber > 1 THEN
0
ELSE
Temp.Price END AS Price,
* FROM (
SELECT *,Row_Number() over (PARTITION by ID,Status ORDER BY ID,Date) AS 'RowNumber'
FROM #Temp
) Temp
Order by ID,Date
You can try this:
;WITH DataSource AS
(
SELECT RANK() OVER (PARTITION BY [ID], [ItemName], [Price], [Status] ORDER BY Date) AS [RankID]
,*
FROM #Temp
)
SELECT [ID]
,[ItemName]
,IIF([RankID] = 1, [Price], 0)
,[Status]
,[Date]
FROM DataSource
ORDER BY [ID]
,[Date]
Here is the output:
please try this below code . it is working for me.
CREATE TABLE #Temp
(ID INT,
ItemName varchar(10),
Price Money,
[Status] INT,
[Date] Datetime)
INSERT INTO #Temp VALUES(1,'ABC',10,1,'2014-08-27')
INSERT INTO #Temp VALUES(1,'ABC',10,2,'2014-08-27')
INSERT INTO #Temp VALUES(1,'ABC',10,1,'2014-08-28')
INSERT INTO #Temp VALUES(2,'DEF',25,1,'2014-08-26')
INSERT INTO #Temp VALUES(2,'DEF',25,3,'2014-08-27')
INSERT INTO #Temp VALUES(2,'DEF',25,1,'2014-08-28')
INSERT INTO #Temp VALUES(3,'GHI',30,1,'2014-08-27')
select *,case when a.rn=1 and status!=2 then price else 0 end as price from
(select *,ROW_NUMBER() over(partition by status,date order by date asc) rn from #Temp) a
order by ItemName asc
You can do this with UNION:
SELECT * FROM #Temp t
WHERE NOT EXISTS
(SELECT * FROM #Temp
WHERE t.id = id and t.status = status and t.date < date)
UNION ALL
SELECT ID, ItemName, 0 as Price, status, date
WHERE EXISTS
(SELECT * FROM #Temp
WHERE t.id = id and t.status = status and t.date < date)
Or subquery:
SELECT CASE
WHEN (SELECT COUNT(*)
FROM #Temp
WHERE t.id = id and t.status = status
and t.date > date) > 1 THEN 0 ELSE price END as NewPrice, t.*
FROM #Temp t
Or possibly RANK() function:
SELECT CASE
WHEN RANK() OVER (PARTITION BY id, status ORDER BY date) > 1
THEN 0 ELSE Price END,
t.*
FROM #Temp t
I am looking for a query which fetches me the data that is different compared to the previous row,
A sample code (with table creation and data)
create table #temp
(id int, eid int, name char(10),estid int, ecid int, epid int, etc char(5) )
insert into #temp values (1,1,'a',1,1,1,'a')
insert into #temp values (2,1,'a',1,1,1,'a')
insert into #temp values (3,1,'a',2,1,1,'a')
insert into #temp values (4,1,'a',1,1,1,'a')
insert into #temp values (5,1,'a',1,1,1,'a')
insert into #temp values (6,1,'a',1,2,1,'a')
insert into #temp values (7,1,'a',1,1,1,'a')
insert into #temp values (8,1,'a',2,1,1,'a')
insert into #temp values (9,1,'a',1,1,1,'a')
insert into #temp values (10,1,'a',1,1,1,'a')
insert into #temp values (11,2,'a',1,1,1,'a')
insert into #temp values (12,2,'a',1,1,1,'a')
insert into #temp values (13,2,'a',2,1,1,'a')
insert into #temp values (14,2,'a',1,1,1,'a')
insert into #temp values (15,2,'a',1,1,1,'a')
insert into #temp values (16,2,'a',1,2,1,'a')
insert into #temp values (17,2,'a',1,1,1,'a')
insert into #temp values (18,2,'a',2,1,1,'a')
insert into #temp values (19,2,'a',1,1,1,'a')
insert into #temp values (20,2,'a',1,1,1,'a')
I tried with some ways of getting the data as the way that i expected
SELECT * INTo #Temp_Final
FROM #temp
WHERE #temp.%%physloc%%
NOT IN (SELECT Min(b.%%physloc%%)
FROM #temp b
GROUP BY eid,name,estid,ecid,epid,etc)
ORDER BY id
SELECT * FROM #temp WHERE id not in (SELECT id FROM #Temp_Final) ORDER BY id
But i wasn't getting the result as i expected...
This is how the result needs to be
select * from #temp where id in (1,3,4,6,7,8,9,11,13,14,16,17,18,19)
You can do this with a simple self-join and appropriate comparison:
select t.*
from #temp t left outer join
#temp tprev
on t.id = tprev.id + 1
where tprev.id is null or
t.name <> tprev.name or
t.estid <> tprev.estid or
t.ecid <> tprev.ecid or
t.epid <> tprev.epid or
t.etc <> tprev.etc;
This assumes that the ids are sequential with no gaps. If the ids are not, you can get the previous id using a correlated subquery or the lag() function.
Your title says "delete" but the question seems to just want the list of such rows. You can phrase this as a delete query if you need to.
For SQL Server 2012 (SQL Fiddle)
WITH CTE
AS (SELECT *,
LAG(eid) OVER (ORDER BY id) AS prev_eid,
LAG(name) OVER (ORDER BY id) AS prev_name,
LAG(estid) OVER (ORDER BY id) AS prev_estid,
LAG(ecid) OVER (ORDER BY id) AS prev_ecid,
LAG(epid) OVER (ORDER BY id) AS prev_epid,
LAG(etc) OVER (ORDER BY id) AS prev_etc
FROM #temp)
DELETE FROM CTE
WHERE EXISTS (SELECT eid,
name,
estid,
ecid,
epid,
etc
INTERSECT
SELECT prev_eid,
prev_name,
prev_estid,
prev_ecid,
prev_epid,
prev_etc)
select
t.id,
t.eid,
t.name,
t.estid,
t.ecid,
t.epid,
t.etc
from #temp t
left join #temp d
on d.id = t.id-1
and d.eid = t.eid
and d.name = t.name
and d.estid = t.estid
and d.ecid = t.ecid
and d.epid = t.epid
and d.etc = t.etc
where d.id is null
I have a dataset like:
type seqID text
A 1 Text1a
A 2 Text2a
A 3 Text3a
B 1 Text1b
B 2 Text2b
How do I get the row back by type with the highest seqID grouped by type? So in the above example I would want the row that has A, 3, Text3a and B, 2, Text2b returned.
SELECT *
FROM tmp t1
WHERE NOT EXISTS
(SELECT 1 FROM tmp t2 WHERE t1.type = t2.type AND t2.seqID > t1.seqID)
It shouldn't exists any other row with the same type and higher seqID.
You kind of need an ID, but since "Text" seems unique for this example
CREATE TABLE #TMP
(type VARCHAR(3), seqID INT, [text] varchar(256))
insert #TMP values ('A' , 1 , 'Text1a')
insert #TMP values ('A' , 2 , 'Text2a')
insert #TMP values ('A' , 3 , 'Text3a')
insert #TMP values ('B' , 1 , 'Text1b')
insert #TMP values ('B' , 2 , 'Text2b')
SELECT * FROM #TMP T
where [text] IN
(SELECT TOP 1 [text] FROM #TMP t2 WHERE t.type = t2.type ORDER BY t2.seqID DESC)
SELECT tbl.*
FROM
( SELECT type, MAX(seqID)
FROM tbl
GROUP BY type) maxes
WHERE
tbl.type= maxes.type AND
tbl.seqID= maxes.seqID
SELECT t.* FROM
(
SELECT type, MAX(seqID) as maxId
FROM Table
GROUP BY type
) m
INNER JOIN Table t ON m.maxId = t.seqId
Using CTE
;WITH maxIds(maxId)
AS
(
SELECT type, MAX(seqID) as maxId
FROM Table
GROUP BY type
)
SELECT t.* FROM
Table t
INNER JOIN maxIds m ON m.maxId = t.seqID
If you are on SQL Server 2005+, you could use a ranking function (more specifically, ROW_NUMBER()):
SELECT
type,
seqID,
text
FROM (
SELECT
*,
rnk = ROW_NUMBER() OVER (PARTITION BY type ORDER BY seqID DESC)
FROM atable
) s
WHERE rnk = 1
create table #tlb1(
[type] VARCHAR(3), seqID INT, [text] varchar(max)
)
declare #type varchar(3), #text varchar(max);
declare #seqID int;
declare seq_cursor cursor for
select [type], max(seqID) from tbl group by [type]
open seq_cursor
fetch next from seq_cursor into #type,#seqID
while(##fetch_status=0)
begin
set #text= (select [text] from tbl where [type]=#type and seqID=#seqid);
insert into #tlb1 values (#type, #seqID,#text);
fetch next from seq_cursor into #type,#seqID
end
select * from #tlb1
close seq_cursor
deallocate seq_cursor
truncate table #tlb1
Try:
SELECT type, max(seqID),text
FROM 'db'
GROUP BY type
As easy as that.
EDITED solution. Consider this a psuedo-code (since I am not familiar with SQL server syntax):
SELECT a.type, a.seqID, a.text FROM table a
JOIN
(SELECT type, max(seqID) seqID FROM table GROUP BY type) b
ON a.seqID = b.seqID AND a.type=b.type