How to reverse string in column a word by word? - sql

I have a table with a varchar(50) column 'Relation' having more than 1000 rows of data like:
P1_P10_P45_P20
P1_P14_P5_P22
P1_P3
P3_P4_P5_P2_P100_P2_P1
I want the output to have reverse order:
P20_P45_P10_P1
P22_P5_P14_P1
P3_P1
P1_P2_P100_P2_P5_P4_P3
Could you please help me achieve this in single query?

Aditi you can use a Tally table to find all _ and then join them back using STUFF + FOR XML PATH combination like below.
I recommend that you read about Tally tables at earliest possible time here
Also the demo link is here
--create table yourtable(Relation nvarchar(50));
--insert into yourtable values
-- ('P1_P14_P5_P22'),
-- ('P1_P3'),
-- ('P3_P4_P5_P2_P100_P2_P1'), ('P1_P3'),
-- ('P3_P4_P5_P2_P100_P2_P1');
;WITH Tally AS (
SELECT 1 as Num
UNION ALL
SELECT Num + 1 FROM Tally WHERE Num < 51
)
,
InputSet AS
(
select *, RN=row_number() over (order by (select 1)) from yourtable
)
,TempSet AS
(
SELECT
Relation,
Num,
RN,
partBetweenUnderscore = SUBSTRING(Relation, Num, ISNULL(LEAD(Num) OVER (Partition by RN ORDER BY Num ASC),LEN('_'+Relation)+1)-Num-1)
FROM
(
SELECT *
FROM InputSet CROSS JOIN Tally
WHERE CHARINDEX('_','_'+Relation,Num)=Num
)T
)
SELECT
Relation,
NewRelation = STUFF(
(SELECT '_' + T1.partBetweenUnderscore FROM TempSet T1 WHERE T1.RN=T2.RN ORDER BY T1.Num DESC FOR XML PATH ('')
),1,1,'')
FROM TempSet T2
GROUP BY RN, Relation

You need to split the stored strings using a splitter, that returns the substrings and the position of each substring. After that you can easily build the desired output.
If you use SQL Server 2017+, you may try a JSON-based approach. You need to transform each string into a valid JSON array (for example P1_P10_P45_P20 into ["'P1","P10","P45","P20"]), parse this array as a table with OPENJSON() and join the rows with STRING_AGG() to generate the expected output:
Table:
CREATE TABLE Data (Relation varchar(50))
INSERT INTO Data (Relation)
VALUES
('P1_P10_P45_P20'),
('P1_P14_P5_P22'),
('P1_P3'),
('P3_P4_P5_P2_P100_P2_P1')
Statement:
SELECT c.Relation
FROM Data d
OUTER APPLY (
SELECT STRING_AGG([value], '_') WITHIN GROUP (ORDER BY CONVERT(int, [key]) DESC)
FROM OPENJSON(CONCAT('["', REPLACE(d.Relation, '_', '","'), '"]'))
) c (Relation)
Result:
Relation
----------------------
P20_P45_P10_P1
P22_P5_P14_P1
P3_P1
P1_P2_P100_P2_P5_P4_P3

First all of the prievious comments are correct, especially that this is a data model problem. Here is a very kludgy solution. I offer it because you only have 1000 records. This is not efficient, nor will it scale up. The following works on MS SQL Server 2017.
Drop table if exists Relation
create table Relation (Relation varchar(50))
INSERT INTO Relation (Relation)
VALUES
('P1_P10_P45_P20'),
('P1_P14_P5_P22'),
('P1_P3'),
('P3_P4_P5_P2_P100_P2_P1');
DROP TABLE IF EXISTS Rev
create table Rev (Relation varchar(50), Rev varchar(50))
DROP TABLE IF EXISTS Z
create table Z (token varchar(50))
declare #Reverse varchar(50)
set #Reverse = ''
declare #token varchar(50)
declare #Relation varchar(50)
declare cur cursor for select * from Relation
open cur
fetch next from cur into #Relation
while ##FETCH_STATUS = 0
begin
with T(Relation, starts, pos) as (
select #Relation, 1, charindex('_', #Relation)
union all
select #Relation, pos + 1, charindex('_', #Relation, pos + 1)
from T
where pos > 0)
insert into Z select substring(#Relation, starts, case when pos > 0 then pos - starts else len(#Relation) end) token
from T
declare cur2 cursor for select token from Z
open cur2
fetch next from cur2 into #token
while ##FETCH_STATUS = 0
begin
set #Reverse = #token + '_' + #Reverse
fetch next from cur2 into #token
end
close cur2
deallocate cur2
set #Reverse = (select left(#Reverse,len(#Reverse)-1))
insert into Rev select #Relation, #Reverse
set #Reverse = ''
delete Z
fetch next from cur into #Relation
end;
close cur
deallocate cur
select * from Rev
SELECT ##version

Related

Compare 2 columns value in a table and show the matching values in a new column . column values are comma seperated

Here I want to compare the comma-separated values from the column roles and userroles, then show the matching values in another column.
I found an example compare comma separated values in sql and used the cursor to iterate rows one by one. it's working for me. but I feel there should be a better way to do it.
Any help is much appreciated.
Create table #temp4
(
user_id int,
permission_id int,
roles varchar(max),
userroles varchar(max),
matchingrolesinthisrow varchar(max))
Insert Into #temp4 values
( 1, -12010, '2341,8760,3546', '1000,1001,1002,1003', null),
( 1, -334, '1002,1001,3467', '2341,1002,3467', null),
( 2, -12349, '9876,9982,6548', '1001,1002,2341', null)
below is the result table I am looking for.
user_id
permission_id
roles
userroles
matchingrolesinthisrow
1
-12010
2341,8760,3546
1000,1001,1002,1003
1
-334
1002,1001,3467
2341,1002,3467
1002,3467
2
-12349
9876,9982,6548
1001,1002,2341
My attempt so far and it's working. please guide me to do this in a better way.
DECLARE #user_id INT
DECLARE #permission_id INT
DECLARE #roles VARCHAR(MAX)
DECLARE #userroles VARCHAR(MAX)
DECLARE #matchingrolesinthisrow VARCHAR(MAX)
declare cur CURSOR LOCAL for
select user_id, permission_id, roles, userroles, matchingrolesinthisrow from #temp4 order by 1
open cur
fetch next from cur into #user_id, #permission_id, #roles, #userroles, #matchingrolesinthisrow
while ##FETCH_STATUS = 0 BEGIN
print (#roles)
print(#userroles)
--execute on each row
UPDATE #temp4
SET matchingrolesinthisrow = T1.[Item]
FROM [developers].[Split](#roles, ',') AS T1
INNER JOIN [developers].[Split](#userroles, ',') AS T2 on T1.[Item] = T2.[Item]
Where roles = #roles and userroles = #userroles and permission_id = #permission_id and user_id = #user_id
fetch next from cur into #user_id, #permission_id, #roles, #userroles, #matchingrolesinthisrow
END
close cur
deallocate cur
--Split function
CREATE FUNCTION [developers].[Split]
(
#s VARCHAR(max),
#split CHAR(1)
)
RETURNS #temptable TABLE ([Item] VARCHAR(MAX))
AS
BEGIN
DECLARE #x XML
SELECT #x = CONVERT(xml,'<root><s>' + REPLACE(#s,#split,'</s><s>') + '</s></root>');
INSERT INTO #temptable
SELECT [Value] = T.c.value('.','varchar(20)')
FROM #X.nodes('/root/s') T(c);
RETURN
END;
Ideally, you should store each inidividual piece of information in separate row, so you should have two separate tables roles and userroles which are foreign-keyed on this one.
Be that as it may, this does not need cursors. You can use STRING_SPLIT and STRING_AGG to get the result you want very easily:
SELECT
t4.user_id,
t4.permission_id,
t4.roles,
t4.userroles,
matchingrolesinthisrow = (
SELECT STRING_AGG(r.value, ',')
FROM STRING_SPLIT(t4.roles, ',') r
JOIN STRING_SPLIT(t4.userroles, ',') ur ON ur.value = r.value
)
FROM #temp4 t4;
SQL Fiddle
If you are on an early version of SQL Server, you will have to use custom Table Valued Functions to do this.
Modified the shared answer to work in the lower version of SQL Server.
SELECT
t4.user_id,
t4.permission_id,
t4.roles,
t4.userroles,
STUFF((SELECT N',' + CONVERT(nvarchar(2000),r.value)
FROM STRING_SPLIT(t4.roles, ',') r
INNER JOIN STRING_SPLIT(t4.userroles, ',') ur ON ur.value = r.value
FOR XML PATH(N''), TYPE).value(N'.[1]', N'nvarchar(max)'), 1, 1, N'') as matchingrolesinthisrow
FROM #temp4 t4;

find circular transactions in database table

I have a table in sql server database in which records of transactions are stored. Table consists of user id of buyer and user id of seller of product. I have to find the circles in the table for example-
I have to get the records of type- A sells to B, B sells to C, C sells to D AND D sells to A.
Please help.
Use following function:
CREATE FUNCTION dbo.CheckIsCircular(#SellerId INT)
RETURNS BIT
AS BEGIN
DECLARE #IsCircular BIT = 0
DECLARE #Sellers TABLE(Id INT)
DECLARE #TempSellers TABLE(Id INT)
DECLARE #Buyers TABLE(Id INT)
INSERT INTO #TempSellers(Id)VALUES(#SellerId)
WHILE EXISTS(SELECT * FROM #TempSellers)BEGIN
IF EXISTS(SELECT *
FROM #Sellers s
INNER JOIN #TempSellers t ON t.Id = s.Id)BEGIN
SET #IsCircular = 1
BREAK;
END
INSERT INTO #Sellers(Id)
SELECT Id FROM #TempSellers
INSERT INTO #Buyers(Id) SELECT BuyerId FROM YourTable
DELETE #TempSellers
INSERT Into #TempSellers(Id)
SELECT YourTable.SellerId
FROM YourTable
INNER JOIN #Buyers ON [#Buyers].Id = YourTable.SellerId
END
RETURN #IsCircular
END
Your problem is a graph traversal challenge; this is not natively supported in TSQL, but you can simulate it.
This is a skeleton how I do it in Teradata, so syntax must be slightly modified for SQL Server:
WITH RECURSIVE cte (..., Path, isCycle) AS
(
SELECT
...
,',' || CAST(seller AS VARCHAR(1000)) || ',' AS path
,0 AS isCycle
FROM tab
UNION ALL
SELECT
...
,cte.Path || cte.buyer || ',',
,case when cte.Path LIKE '%,' || TRIM(tab.buyer) || ',%' then 1 else 0 end
FROM cte, tab
WHERE cte.buyer = tab.seller
AND cte.isCycle <> 1
)
SELECT ...
,Path || Destination
,isCycle
FROM cte
WHERE isCycle = 1
Build a materialized path of the graph while traversing and check if the next buyer is already in this path.
With a recursive cte
declare #trans table (seller int, buyer int)
insert #trans
values (1,2),(2,3),(3,4),(4,1),(1,5),(2,6),(3,5)
begin try
;with cte as
(
select *, convert(varchar(500),'') as route from #trans
union all
select cte.seller, t1.buyer, convert(varchar(500),route + CONVERT(varchar(5),t1.seller)) from cte
inner join #trans t1 on cte.buyer = t1.seller
)
select * from cte
where seller=buyer
option (maxrecursion 50)
end try
begin catch
print 'loops'
end catch

Dynamically Create tables and Insert into it from another table with CSV values

Have a Table with the CSV Values in the columns as below
ID Name text
1 SID,DOB 123,12/01/1990
2 City,State,Zip NewYork,NewYork,01234
3 SID,DOB 456,12/21/1990
What is need to get is 2 tables in this scenario as out put with the corresponding values
ID SID DOB
1 123 12/01/1990
3 456 12/21/1990
ID City State Zip
2 NewYork NewYork 01234
Is there any way of achieving it using a Cursor or any other method in SQL server?
There are several ways that this can be done. One way that I would suggest would be to split the data from the comma separated list into multiple rows.
Since you are using SQL Server, you could implement a recursive CTE to split the data, then apply a PIVOT function to create the columns that you want.
;with cte (id, NameItem, Name, textItem, text) as
(
select id,
cast(left(Name, charindex(',',Name+',')-1) as varchar(50)) NameItem,
stuff(Name, 1, charindex(',',Name+','), '') Name,
cast(left(text, charindex(',',text+',')-1) as varchar(50)) textItem,
stuff(text, 1, charindex(',',text+','), '') text
from yt
union all
select id,
cast(left(Name, charindex(',',Name+',')-1) as varchar(50)) NameItem,
stuff(Name, 1, charindex(',',Name+','), '') Name,
cast(left(text, charindex(',',text+',')-1) as varchar(50)) textItem,
stuff(text, 1, charindex(',',text+','), '') text
from cte
where Name > ''
and text > ''
)
select id, SID, DOB
into table1
from
(
select id, nameitem, textitem
from cte
where nameitem in ('SID', 'DOB')
) d
pivot
(
max(textitem)
for nameitem in (SID, DOB)
) piv;
See SQL Fiddle with Demo. The recursive version will work great but if you have a large dataset, you could have some performance issues so you could also use a user defined function to split the data:
create FUNCTION [dbo].[Split](#String1 varchar(MAX), #String2 varchar(MAX), #Delimiter char(1))
returns #temptable TABLE (colName varchar(MAX), colValue varchar(max))
as
begin
declare #idx1 int
declare #slice1 varchar(8000)
declare #idx2 int
declare #slice2 varchar(8000)
select #idx1 = 1
if len(#String1)<1 or #String1 is null return
while #idx1 != 0
begin
set #idx1 = charindex(#Delimiter,#String1)
set #idx2 = charindex(#Delimiter,#String2)
if #idx1 !=0
begin
set #slice1 = left(#String1,#idx1 - 1)
set #slice2 = left(#String2,#idx2 - 1)
end
else
begin
set #slice1 = #String1
set #slice2 = #String2
end
if(len(#slice1)>0)
insert into #temptable(colName, colValue) values(#slice1, #slice2)
set #String1 = right(#String1,len(#String1) - #idx1)
set #String2 = right(#String2,len(#String2) - #idx2)
if len(#String1) = 0 break
end
return
end;
Then you can use a CROSS APPLY to get the result for each row:
select id, SID, DOB
into table1
from
(
select t.id,
c.colname,
c.colvalue
from yt t
cross apply dbo.split(t.name, t.text, ',') c
where c.colname in ('SID', 'DOB')
) src
pivot
(
max(colvalue)
for colname in (SID, DOB)
) piv;
See SQL Fiddle with Demo
You'd need to approach this as a multi-step ETL project. I'd probably start with exporting the two types of rows into a couple staging tables. So, for example:
select * from yourtable /* rows that start with a number */
where substring(text,1,1) in
('0','1','2','3','4','5','6','7','8','9')
select * from yourtable /* rows that don't start with a number */
where substring(text,1,1)
not in ('0','1','2','3','4','5','6','7','8','9')
/* or simply this to follow your example explicitly */
select * from yourtable where name like 'sid%'
select * from yourtable where name like 'city%'
Once you get the two types separated then you can split them out with one of the already written split functions found readily out on the interweb.
Aaron Bertrand (who is on here often) has written up a great post on the variety of ways to split comma delimted strings using SQL. Each of the methods are compared and contrasted here.
http://www.sqlperformance.com/2012/07/t-sql-queries/split-strings
If your row count is minimal (under 50k let's say) and it's going to be a one time operation than pick the easiest way and don't worry too much about all the performance numbers.
If you have a ton of rows or this is an ETL process that will run all the time then you'll really want to pay attention to that stuff.
A simple solution using cursors to build temporary tables. This has the limitation of making all columns VARCHAR and would be slow for large amounts of data.
--** Set up example data
DECLARE #Source TABLE (ID INT, Name VARCHAR(50), [text] VARCHAR(200));
INSERT INTO #Source
(ID, Name, [text])
VALUES (1, 'SID,DOB', '123,12/01/1990')
, (2, 'City,State,Zip', 'NewYork,NewYork,01234')
, (3, 'SID,DOB', '456,12/21/1990');
--** Declare variables
DECLARE #Name VARCHAR(200) = '';
DECLARE #Text VARCHAR(1000) = '';
DECLARE #SQL VARCHAR(MAX);
--** Set up cursor for the tables
DECLARE cursor_table CURSOR FAST_FORWARD READ_ONLY FOR
SELECT s.Name
FROM #Source AS s
GROUP BY Name;
OPEN cursor_table
FETCH NEXT FROM cursor_table INTO #Name;
WHILE ##FETCH_STATUS = 0
BEGIN
--** Dynamically create a temp table with the specified columns
SET #SQL = 'CREATE TABLE ##Table (' + REPLACE(#Name, ',', ' VARCHAR(50),') + ' VARCHAR(50));';
EXEC(#SQL);
--** Set up cursor to insert the rows
DECLARE row_cursor CURSOR FAST_FORWARD READ_ONLY FOR
SELECT s.Text
FROM #Source AS s
WHERE Name = #Name;
OPEN row_cursor;
FETCH NEXT FROM row_cursor INTO #Text;
WHILE ##FETCH_STATUS = 0
BEGIN
--** Dynamically insert the row
SELECT #SQL = 'INSERT INTO ##Table VALUES (''' + REPLACE(#Text, ',', ''',''') + ''');';
EXEC(#SQL);
FETCH NEXT FROM row_cursor INTO #Text;
END
--** Display the table
SELECT *
FROM ##Table;
--** Housekeeping
CLOSE row_cursor;
DEALLOCATE row_cursor;
DROP TABLE ##Table;
FETCH NEXT FROM cursor_table INTO #Name;
END
CLOSE cursor_table;
DEALLOCATE cursor_table;

t-sql find specific value with a csv string

I need some on help a SQL Query. I have a column with values stored as comma separated values.
I need to write a query which finds the 3rd delimited item within each value in the column.
Is this possible to do this in a Select statement?
ex: ColumnValue: josh,Reg01,False,a0-t0,22/09/2010
So I will need to get the 3rd value (i.e.) False from the above string.
Yes.
Where #s is your string...
select
SUBSTRING (#s,
CHARINDEX(',',#s,CHARINDEX(',',#s)+1)+1,
CHARINDEX(',',#s,CHARINDEX(',',#s,CHARINDEX(',',#s)+1)+1)
-CHARINDEX(',',#s,CHARINDEX(',',#s)+1)-1)
Or more generically...
;with cte as
(
select 1 as Item, 1 as Start, CHARINDEX(',',#s, 1) as Split
union all
select cte.Item+1, cte.Split+1, nullif(CHARINDEX(',',#s, cte.Split+1),0) as Split
from cte
where cte.Split<>0
)
select SUBSTRING(#s, start,isnull(split,len(#s)+1)-start)
from cte
where Item = 3
Now store your data properly :)
Try this (assuming SQL Server 2005+)
DECLARE #t TABLE(ColumnValue VARCHAR(50))
INSERT INTO #t(ColumnValue) SELECT 'josh,Reg01,False,a0-t0,22/09/2010'
INSERT INTO #t(ColumnValue) SELECT 'mango,apple,bannana,grapes'
INSERT INTO #t(ColumnValue) SELECT 'stackoverflow'
SELECT ThirdValue = splitdata
FROM(
SELECT
Rn = ROW_NUMBER() OVER(PARTITION BY ColumnValue ORDER BY (SELECT 1))
,X.ColumnValue
,Y.splitdata
FROM
(
SELECT *,
CAST('<X>'+REPLACE(F.ColumnValue,',','</X><X>')+'</X>' AS XML) AS xmlfilter FROM #t F
)X
CROSS APPLY
(
SELECT fdata.D.value('.','varchar(50)') AS splitdata
FROM X.xmlfilter.nodes('X') as fdata(D)
) Y
)X WHERE X.Rn = 3
//Result
ThirdValue
False
bannana
Also it is not very clear from your question as what version of SQL Server you are using. In case you are using SQL SERVER 2000, you can go ahead with the below approach.
Step 1: Create a number table
CREATE TABLE dbo.Numbers
(
N INT NOT NULL PRIMARY KEY
);
GO
DECLARE #rows AS INT;
SET #rows = 1;
INSERT INTO dbo.Numbers VALUES(1);
WHILE(#rows <= 10000)
BEGIN
INSERT INTO dbo.Numbers SELECT N + #rows FROM dbo.Numbers;
SET #rows = #rows * 2;
END
Step 2: Apply the query below
DECLARE #t TABLE(ColumnValue VARCHAR(50))
INSERT INTO #t(ColumnValue) SELECT 'josh,Reg01,False,a0-t0,22/09/2010'
INSERT INTO #t(ColumnValue) SELECT 'mango,apple,bannana,grapes'
INSERT INTO #t(ColumnValue) SELECT 'stackoverflow'
--Declare a table variable to put the identity column and store the indermediate results
DECLARE #tempT TABLE(Id INT IDENTITY,ColumnValue VARCHAR(50),SplitData VARCHAR(50))
-- Insert the records into the table variable
INSERT INTO #tempT
SELECT
ColumnValue
,SUBSTRING(ColumnValue, Numbers.N,CHARINDEX(',', ColumnValue + ',', Numbers.N) - Numbers.N) AS splitdata
FROM #t
JOIN Numbers ON Numbers.N <= DATALENGTH(ColumnValue) + 1
AND SUBSTRING(',' + ColumnValue, Numbers.N, 1) = ','
--Project the filtered records
SELECT ThirdValue = X.splitdata
FROM
--The co-related subquery does the ROW_NUMBER() OVER(PARTITION BY ColumnValue)
(SELECT
Rn = (SELECT COUNT(*)
FROM #tempT t2
WHERE t2.ColumnValue=t1.ColumnValue
AND t2.Id<=t1.Id)
,t1.ColumnValue
,t1.splitdata
FROM #tempT t1)X
WHERE X.Rn =3
-- Result
ThirdValue
False
bannana
Also you can use Master..spt_Values for your number table
DECLARE #t TABLE(ColumnValue VARCHAR(50))
INSERT INTO #t(ColumnValue) SELECT 'josh,Reg01,False,a0-t0,22/09/2010'
INSERT INTO #t(ColumnValue) SELECT 'mango,apple,bannana,grapes'
INSERT INTO #t(ColumnValue) SELECT 'stackoverflow'
--Declare a table variable to put the identity column and store the indermediate results
DECLARE #tempT TABLE(Id INT IDENTITY,ColumnValue VARCHAR(50),SplitData VARCHAR(50))
-- Insert the records into the table variable
INSERT INTO #tempT
SELECT
ColumnValue
,SUBSTRING(ColumnValue, Number ,CHARINDEX(',', ColumnValue + ',', Number ) - Number) AS splitdata
FROM #t
JOIN master..spt_values ON Number <= DATALENGTH(ColumnValue) + 1 AND type='P'
AND SUBSTRING(',' + ColumnValue, Number , 1) = ','
--Project the filtered records
SELECT ThirdValue = X.splitdata
FROM
--The co-related subquery does the ROW_NUMBER() OVER(PARTITION BY ColumnValue)
(SELECT
Rn = (SELECT COUNT(*)
FROM #tempT t2
WHERE t2.ColumnValue=t1.ColumnValue
AND t2.Id<=t1.Id)
,t1.ColumnValue
,t1.splitdata
FROM #tempT t1)X
WHERE X.Rn =3
You can read about this from
1) What is the purpose of system table table master..spt_values and what are the meanings of its values?
2) Why (and how) to split column using master..spt_values?
You really need something like String.Split(',')(2) which unfortunately dos not exist in SQL but this may be helpful to you
You can make some test with this solution and the other ones but, I believe that using XML in such situations almost always gives to you best performance and insure less coding:
DECLARE #InPutCSV NVARCHAR(2000)= 'josh,Reg01,False,a0-t0,22/09/2010'
DECLARE #ValueIndexToGet INT=3
DECLARE #XML XML = CAST ('<d>' + REPLACE(#InPutCSV, ',', '</d><d>') + '</d>' AS XML);
WITH CTE(RecordNumber,Value) AS
(
SELECT ROW_NUMBER() OVER(ORDER BY T.v.value('.', 'NVARCHAR(100)') DESC) AS RecordNumber
,T.v.value('.', 'NVARCHAR(100)') AS Value
FROM #XML.nodes('/d') AS T(v)
)
SELECT Value
FROM CTE WHERE RecordNumber=#ValueIndexToGet
I can confirm that it takes 1 seconds to get value from CSV string with 100 000 values.

Delete duplicate records in SQL Server?

Consider a column named EmployeeName table Employee. The goal is to delete repeated records, based on the EmployeeName field.
EmployeeName
------------
Anand
Anand
Anil
Dipak
Anil
Dipak
Dipak
Anil
Using one query, I want to delete the records which are repeated.
How can this be done with TSQL in SQL Server?
You can do this with window functions. It will order the dupes by empId, and delete all but the first one.
delete x from (
select *, rn=row_number() over (partition by EmployeeName order by empId)
from Employee
) x
where rn > 1;
Run it as a select to see what would be deleted:
select *
from (
select *, rn=row_number() over (partition by EmployeeName order by empId)
from Employee
) x
where rn > 1;
Assuming that your Employee table also has a unique column (ID in the example below), the following will work:
delete from Employee
where ID not in
(
select min(ID)
from Employee
group by EmployeeName
);
This will leave the version with the lowest ID in the table.
Edit
Re McGyver's comment - as of SQL 2012
MIN can be used with numeric, char, varchar, uniqueidentifier, or datetime columns, but not with bit columns
For 2008 R2 and earlier,
MIN can be used with numeric, char, varchar, or datetime columns, but not with bit columns (and it also doesn't work with GUID's)
For 2008R2 you'll need to cast the GUID to a type supported by MIN, e.g.
delete from GuidEmployees
where CAST(ID AS binary(16)) not in
(
select min(CAST(ID AS binary(16)))
from GuidEmployees
group by EmployeeName
);
SqlFiddle for various types in Sql 2008
SqlFiddle for various types in Sql 2012
You could try something like the following:
delete T1
from MyTable T1, MyTable T2
where T1.dupField = T2.dupField
and T1.uniqueField > T2.uniqueField
(this assumes that you have an integer based unique field)
Personally though I'd say you were better off trying to correct the fact that duplicate entries are being added to the database before it occurs rather than as a post fix-it operation.
DELETE
FROM MyTable
WHERE ID NOT IN (
SELECT MAX(ID)
FROM MyTable
GROUP BY DuplicateColumn1, DuplicateColumn2, DuplicateColumn3)
WITH TempUsers (FirstName, LastName, duplicateRecordCount)
AS
(
SELECT FirstName, LastName,
ROW_NUMBER() OVER (PARTITIONBY FirstName, LastName ORDERBY FirstName) AS duplicateRecordCount
FROM dbo.Users
)
DELETE
FROM TempUsers
WHERE duplicateRecordCount > 1
WITH CTE AS
(
SELECT EmployeeName,
ROW_NUMBER() OVER(PARTITION BY EmployeeName ORDER BY EmployeeName) AS R
FROM employee_table
)
DELETE CTE WHERE R > 1;
The magic of common table expressions.
Try
DELETE
FROM employee
WHERE rowid NOT IN (SELECT MAX(rowid) FROM employee
GROUP BY EmployeeName);
If you're looking for a way to remove duplicates, yet you have a foreign key pointing to the table with duplicates, you could take the following approach using a slow yet effective cursor.
It will relocate the duplicate keys on the foreign key table.
create table #properOlvChangeCodes(
id int not null,
name nvarchar(max) not null
)
DECLARE #name VARCHAR(MAX);
DECLARE #id INT;
DECLARE #newid INT;
DECLARE #oldid INT;
DECLARE OLVTRCCursor CURSOR FOR SELECT id, name FROM Sales_OrderLineVersionChangeReasonCode;
OPEN OLVTRCCursor;
FETCH NEXT FROM OLVTRCCursor INTO #id, #name;
WHILE ##FETCH_STATUS = 0
BEGIN
-- determine if it should be replaced (is already in temptable with name)
if(exists(select * from #properOlvChangeCodes where Name=#name)) begin
-- if it is, finds its id
Select top 1 #newid = id
from Sales_OrderLineVersionChangeReasonCode
where Name = #name
-- replace terminationreasoncodeid in olv for the new terminationreasoncodeid
update Sales_OrderLineVersion set ChangeReasonCodeId = #newid where ChangeReasonCodeId = #id
-- delete the record from the terminationreasoncode
delete from Sales_OrderLineVersionChangeReasonCode where Id = #id
end else begin
-- insert into temp table if new
insert into #properOlvChangeCodes(Id, name)
values(#id, #name)
end
FETCH NEXT FROM OLVTRCCursor INTO #id, #name;
END;
CLOSE OLVTRCCursor;
DEALLOCATE OLVTRCCursor;
drop table #properOlvChangeCodes
delete from person
where ID not in
(
select t.id from
(select min(ID) as id from person
group by email
) as t
);
Please see the below way of deletion too.
Declare #Employee table (EmployeeName varchar(10))
Insert into #Employee values
('Anand'),('Anand'),('Anil'),('Dipak'),
('Anil'),('Dipak'),('Dipak'),('Anil')
Select * from #Employee
Created a sample table named #Employee and loaded it with given data.
Delete aliasName from (
Select *,
ROW_NUMBER() over (Partition by EmployeeName order by EmployeeName) as rowNumber
From #Employee) aliasName
Where rowNumber > 1
Select * from #Employee
Result:
I know, this is asked six years ago, posting just incase it is helpful for anyone.
Here's a nice way of deduplicating records in a table that has an identity column based on a desired primary key that you can define at runtime. Before I start I'll populate a sample data set to work with using the following code:
if exists (select 1 from sys.all_objects where type='u' and name='_original')
drop table _original
declare #startyear int = 2017
declare #endyear int = 2018
declare #iterator int = 1
declare #income money = cast((SELECT round(RAND()*(5000-4990)+4990 , 2)) as money)
declare #salesrepid int = cast(floor(rand()*(9100-9000)+9000) as varchar(4))
create table #original (rowid int identity, monthyear varchar(max), salesrepid int, sale money)
while #iterator<=50000 begin
insert #original
select (Select cast(floor(rand()*(#endyear-#startyear)+#startyear) as varchar(4))+'-'+ cast(floor(rand()*(13-1)+1) as varchar(2)) ), #salesrepid , #income
set #salesrepid = cast(floor(rand()*(9100-9000)+9000) as varchar(4))
set #income = cast((SELECT round(RAND()*(5000-4990)+4990 , 2)) as money)
set #iterator=#iterator+1
end
update #original
set monthyear=replace(monthyear, '-', '-0') where len(monthyear)=6
select * into _original from #original
Next I'll create a Type called ColumnNames:
create type ColumnNames AS table
(Columnnames varchar(max))
Finally I will create a stored proc with the following 3 caveats:
1. The proc will take a required parameter #tablename that defines the name of the table you are deleting from in your database.
2. The proc has an optional parameter #columns that you can use to define the fields that make up the desired primary key that you are deleting against. If this field is left blank, it is assumed that all the fields besides the identity column make up the desired primary key.
3. When duplicate records are deleted, the record with the lowest value in it's identity column will be maintained.
Here is my delete_dupes stored proc:
create proc delete_dupes (#tablename varchar(max), #columns columnnames readonly)
as
begin
declare #table table (iterator int, name varchar(max), is_identity int)
declare #tablepartition table (idx int identity, type varchar(max), value varchar(max))
declare #partitionby varchar(max)
declare #iterator int= 1
if exists (select 1 from #columns) begin
declare #columns1 table (iterator int, columnnames varchar(max))
insert #columns1
select 1, columnnames from #columns
set #partitionby = (select distinct
substring((Select ', '+t1.columnnames
From #columns1 t1
Where T1.iterator = T2.iterator
ORDER BY T1.iterator
For XML PATH ('')),2, 1000) partition
From #columns1 T2 )
end
insert #table
select 1, a.name, is_identity from sys.all_columns a join sys.all_objects b on a.object_id=b.object_id
where b.name = #tablename
declare #identity varchar(max)= (select name from #table where is_identity=1)
while #iterator>=0 begin
insert #tablepartition
Select distinct case when #iterator=1 then 'order by' else 'over (partition by' end ,
substring((Select ', '+t1.name
From #table t1
Where T1.iterator = T2.iterator and is_identity=#iterator
ORDER BY T1.iterator
For XML PATH ('')),2, 5000) partition
From #table T2
set #iterator=#iterator-1
end
declare #originalpartition varchar(max)
if #partitionby is null begin
select #originalpartition = replace(b.value+','+a.type+a.value ,'over (partition by','') from #tablepartition a cross join #tablepartition b where a.idx=2 and b.idx=1
select #partitionby = a.type+a.value+' '+b.type+a.value+','+b.value+') rownum' from #tablepartition a cross join #tablepartition b where a.idx=2 and b.idx=1
end
else
begin
select #originalpartition=b.value +','+ #partitionby from #tablepartition a cross join #tablepartition b where a.idx=2 and b.idx=1
set #partitionby = (select 'OVER (partition by'+ #partitionby + ' ORDER BY'+ #partitionby + ','+b.value +') rownum'
from #tablepartition a cross join #tablepartition b where a.idx=2 and b.idx=1)
end
exec('select row_number() ' + #partitionby +', '+#originalpartition+' into ##temp from '+ #tablename+'')
exec(
'delete a from _original a
left join ##temp b on a.'+#identity+'=b.'+#identity+' and rownum=1
where b.rownum is null')
drop table ##temp
end
Once this is complied, you can delete all your duplicate records by running the proc. To delete dupes without defining a desired primary key use this call:
exec delete_dupes '_original'
To delete dupes based on a defined desired primary key use this call:
declare #table1 as columnnames
insert #table1
values ('salesrepid'),('sale')
exec delete_dupes '_original' , #table1