sql server delete sql

Thanks
Don't I have to use VERSION along with ID in partition clause?

You are right, I have missed it. Here is the correct query:

;with dup as (
SELECT *,
  ROW_NUMBER() OVER (PARTITION BY [TYPE], [ID], [VERSION] ORDER BY [count]) rn
  FROM Table1
  WHERE [TYPE] = 'SHOE')
DELETE FROM dup where rn > 1;

Select allOpen in new window

I think I got it. You are assigning a rank and deleting ranks >1
Is therr any other way.
Thanks.

Yes, there are other ways too, but this one is the easiest (and fastest in my opinion). Do you wish to explore them?

yeah, I was curious to know if we can do it using one single delete statement  instead of a Select ..Delete

Thanks.

@Thandava Vallepalli: Your statement will not work for the cases with duplicate count values. I.e. if there are two or more records with count=4 for ID=126, they all will be left and not deleted

There is no way in my opinion to delete is using a single delete. There must be a select. However, you will be able to insert records to a temp table using this statement:

select [ID], [version], [type], min(Cnt) Cnt
INTO #Temp
from TabVer
group by [ID], [version], [type]

Select allOpen in new window

then truncate the original table and re-insert the records from the temp table.

TRUNCATE TabVer
INSERT INTO TabVer ([ID], [version], [type], Cnt)
SELECT [ID], [version], [type], Cnt
FROM #Temp

Select allOpen in new window

@Chaau,

Generally most of the tables in RDMBS contain primary key.  I think, [Cnt] is a primary key.
If [Cnt] is not primary key, here is another single delete statement which takes care of duplicate [Cnt] as well.

=========================================
delete t1
from TabVer t1
            inner join (
                                    select row_number() over (partition by [ID], [version], [type] order by [Cnt]) as row_id
                                                ,[Cnt]
                                                ,[ID]
                                                ,[Version]
                                                ,[Type]
                                    from TabVer t2
                                    where t2.[Type] = 'SHOE'
                              ) t3
                        on t1.cnt = t3.cnt
                              and t1.[id] = t3.[id]
                              and t1.[version] = t3.[version]
                              and t1.[type] = t3.[type]
where t1.[Type] = 'SHOE'
            and row_id > 1;
======================================

--itsvtk

@Chaau,

Quick advice....Never ever use truncate table in production database.  
In case of mistakes in queires, we will loose entire data and can't rollback it.


--itsvtk

Hi,

You can try this ..

; WITH CTE AS
(
SELECT *,ROW_NUMBER() OVER (PARTITION BY id,version ORDER BY counts,id) CTN FROM  [dbo].[T3323]
) 
--SELECT *  FROM CTE WHERE CTN > 1
DELETE FROM CTE WHERE CTN > 1

Select allOpen in new window

And we can't write direct delete without select since
Windowed functions can only appear in the SELECT or ORDER BY clauses

@Vikas,

You can write single delete statement (without using window funciton) to delete duplicates from a table, if you have a primary key in that table (I think, [Cnt] is primary key).  Please check my first answer above for single delete statement.

If you don't have primary key in a table, you can still delete duplicate records using window function in a single delete statement. Please check my second answer above.


--itsvtk

@Thandava,

Thanks for you valuable comment but in your second example is the Window function is used with Select Statement
(
select row_number() over (partition by [ID], [version], [type] order by [Cnt]) as row_id
                                                ,[Cnt]
)

and that is what my comment says that without Select you can't use Windows Function..

Thanks again......

@Vikas,

The select statement with window function is required only if a table doesn't have a primary key.  
Most of the cases, in RDBDS, every table contains a primary key / composite primary key.

I assumed that [Cnt] is the primary key column in customer table. If my assumption is correct, then no need of window function.

Here is one statement without any window function to delete duplicates.

==============================================================
DELETE t1
FROM TabVer t1
WHERE Cnt > ( SELECT min(Cnt)
                        FROM TabVer t2
                        WHERE t1.[ID] = t2.[ID]
                                    AND t1.[version] = t2.[version]
                                    AND t1.[type] = t2.[type]
                  )
            AND t1.[Type] = 'SHOE';
=======================================================



If [Cnt] is not a primary key, then have to use window function. But still it is a single statement which deletes duplicates, bit complicated though.
=======================================================
DELETE t1
FROM TabVer t1
            INNER JOIN (
                                    SELECT ROW_NUMBER() OVER (PARTITION BY [ID], [version], [type] ORDER BY [Cnt]) AS row_id
                                                ,[Cnt]
                                                ,[ID]
                                                ,[Version]
                                                ,[Type]
                                    FROM TabVer t2
                                    WHERE t2.[Type] = 'SHOE'
                              ) t3
                        ON t1.cnt = t3.cnt
                              AND t1.[id] = t3.[id]
                              AND t1.[version] = t3.[version]
                              AND t1.[type] = t3.[type]
WHERE t1.[Type] = 'SHOE'
            and row_id > 1;
=======================================================


--itsvtk

If count can have duplicates, you must use a Windowing function to do this with any efficiency at all.


Btw:
>> Never ever use truncate table in production database.  
In case of mistakes in queries, we will lose entire data and can't rollback it. <<

FALSE: Truncate CAN be rolled back.

Thanks for the clarification @ScottPletcher

--itsvtk

thanks all for the answers.