如何删除 SQL Server 2005 中列的 IDENTITY 属性

Question

I want to be able to insert data from a table with an identity column into a temporary table in SQL Server 2005.

The TSQL looks something like:

-- Create empty temp table
SELECT *
INTO #Tmp_MyTable
FROM MyTable
WHERE 1=0
...
WHILE ...
BEGIN
    ...
    INSERT INTO #Tmp_MyTable
    SELECT TOP (@n) *
    FROM MyTable
    ...

END

The above code created #Tmp_Table with an identity column, and the insert subsequently fails with an error "An explicit value for the identity column in table '#Tmp_MyTable' can only be specified when a column list is used and IDENTITY_INSERT is ON."

Is there a way in TSQL to drop the identity property of the column in the temporary table without listing all the columns explicitly ? I specifically want to use "SELECT *" so that the code will continue to work if new columns are added to MyTable.

I believe dropping and recreating the column will change its position, making it impossible to use SELECT *.

Update:

I've tried using IDENTITY_INSERT as suggested in one response. It's not working - see the repro below. What am I doing wrong?

-- Create test table
CREATE TABLE [dbo].[TestTable](
    [ID] [numeric](18, 0) IDENTITY(1,1) NOT NULL,
    [Name] [varchar](50) NULL,
 CONSTRAINT [PK_TestTable] PRIMARY KEY CLUSTERED 
(
    [ID] ASC
)
) 
GO
-- Insert some data
INSERT INTO TestTable
(Name)
SELECT 'One'
UNION ALL
SELECT 'Two'
UNION ALL
SELECT 'Three'
GO
-- Create empty temp table
SELECT *
INTO #Tmp
FROM TestTable
WHERE 1=0

SET IDENTITY_INSERT #Tmp ON -- I also tried OFF / ON
INSERT INTO #Tmp
SELECT TOP 1 * FROM TestTable

SET IDENTITY_INSERT #Tmp OFF 
GO
-- Drop test table
DROP TABLE [dbo].[TestTable]
GO

Note that the error message "An explicit value for the identity column in table '#TmpMyTable' can only be specified when a column list is used and IDENTITY_INSERT is ON." - I specifically don't want to use a column list as explained above.

Update 2 Tried the suggestion from Mike but this gave the same error:

-- Create empty temp table
SELECT *
INTO #Tmp
FROM (SELECT
      m1.*
      FROM TestTable                 m1
          LEFT OUTER JOIN TestTable  m2 ON m1.ID=m2.ID
      WHERE 1=0
 ) dt

INSERT INTO #Tmp
SELECT TOP 1 * FROM TestTable

As for why I want to do this: MyTable is a staging table which can contain a large number of rows to be merged into another table. I want to process the rows from the staging table, insert/update my main table, and delete them from the staging table in a loop that processes N rows per transaction. I realize there are other ways to achieve this.

Update 3

I couldn't get Mike's solution to work, however it suggested the following solution which does work: prefix with a non-identity column and drop the identity column:

SELECT CAST(1 AS NUMERIC(18,0)) AS ID2, *
INTO #Tmp
FROM TestTable
WHERE 1=0
ALTER TABLE #Tmp DROP COLUMN ID

INSERT INTO #Tmp
SELECT TOP 1 * FROM TestTable

Mike's suggestion to store only the keys in the temporary table is also a good one, though in this specific case there are reasons I prefer to have all columns in the temporary table.

Answer 1

You could try

SET IDENTITY_INSERT #Tmp_MyTable ON 
-- ... do stuff
SET IDENTITY_INSERT #Tmp_MyTable OFF

This will allow you to select into #Tmp_MyTable even though it has an identity column.

But this will not work:

-- Create empty temp table
SELECT *
INTO #Tmp_MyTable
FROM MyTable
WHERE 1=0
...
WHILE ...
BEGIN
    ...
    SET IDENTITY_INSERT #Tmp_MyTable ON 

    INSERT INTO #Tmp_MyTable
    SELECT TOP (@n) *
    FROM MyTable

    SET IDENTITY_INSERT #Tmp_MyTable OFF 
    ...    
END

(results in the error "An explicit value for the identity column in table '#Tmp' can only be specified when a column list is used and IDENTITY_INSERT is ON.")

It seems there is no way without actually dropping the column - but that would change the order of columns as OP mentioned. Ugly hack: Create a new table based on #Tmp_MyTable ...

I suggest you write a stored procedure that creates a temporary table based on a table name ( MyTable ) with the same columns (in order), but with the identity property missing.

You could use following code:

select t.name as tablename, typ.name as typename, c.*
from sys.columns c inner join
     sys.tables t on c.object_id = t.[object_id] inner join
     sys.types typ on c.system_type_id = typ.system_type_id
order by t.name, c.column_id

to get a glimpse on how reflection works in TSQL. I believe you will have to loop over the columns for the table in question and execute dynamic (hand-crafted, stored in strings and then evaluated) alter statements to the generated table.

Would you mind posting such a stored procedure for the rest of the world? This question seems to come up quite a lot in other forums as well...

Answer 2

IF you are just processing rows as you describe, wouldn't it be better to just select the top N primary key values into a temp table like:

CREATE TABLE #KeysToProcess
(
     TempID    int  not null primary key identity(1,1)
    ,YourKey1  int  not null
    ,YourKey2  int  not null
)

INSERT INTO #KeysToProcess (YourKey1,YourKey2)
SELECT TOP n YourKey1,YourKey2  FROM MyTable

The keys should not change very often (I hope) but other columns can with no harm to doing it this way.

get the @@ROWCOUNT of the insert and you can do a easy loop on TempID where it will be from 1 to @@ROWCOUNT

and/or

just join #KeysToProcess to your MyKeys table and be on your way, with no need to duplicate all the data.

This runs fine on my SQL Server 2005, where MyTable.MyKey is an identity column.

-- Create empty temp table
SELECT *
INTO #TmpMikeMike
FROM (SELECT
      m1.*
      FROM MyTable                 m1
          LEFT OUTER JOIN MyTable  m2 ON m1.MyKey=m2.MyKey
      WHERE 1=0
 ) dt

INSERT INTO #TmpMike
SELECT TOP 1 * FROM MyTable

SELECT * from #TmpMike

EDIT
THIS WORKS, with no errors...

-- Create empty temp table
SELECT *
INTO #Tmp_MyTable
FROM (SELECT
          m1.*
          FROM MyTable                 m1
              LEFT OUTER JOIN MyTable  m2 ON m1.KeyValue=m2.KeyValue
          WHERE 1=0
     ) dt
...
WHILE ...
BEGIN
    ...
    INSERT INTO #Tmp_MyTable
    SELECT TOP (@n) *
    FROM MyTable
    ...

END

however, what is your real problem? Why do you need to loop while inserting "*" into this temp table? You may be able to shift strategy and come up with a much better algorithm overall.

Answer 3

EDIT Toggling IDENTITY_INSERT as suggested by Daren is certainly the more elegant approach, in my case I needed to eliminate the identity column so that I could reinsert selected data into the source table

The way that I addressed this was to create the temp table just as you do, explicitly drop the identity column, and then dynamically build the sql so that I have a column list that excludes the identity column (as in your case so the proc would still work if there were changes to the schema) and then execute the sql here's a sample

declare @ret int
Select * into #sometemp from sometable
Where
id = @SomeVariable

Alter Table #sometemp Drop column SomeIdentity 

Select @SelectList = ''
Select @SelectList = @SelectList 
+ Coalesce( '[' + Column_name + ']' + ', ' ,'')
from information_schema.columns
where table_name = 'sometable'
and Column_Name <> 'SomeIdentity'

Set @SelectList = 'Insert into sometable (' 
+ Left(@SelectList, Len(@SelectList) -1) + ')'
Set @SelectList = @SelectList 
+ ' Select * from #sometemp '
exec @ret  =  sp_executesql  @selectlist

Answer 4

I have wrote this procedure as compilation of many answers to automatically and fast drop column identity:

CREATE PROCEDURE dbo.sp_drop_table_identity @tableName VARCHAR(256) AS
BEGIN
    DECLARE @sql VARCHAR (4096);
    DECLARE @sqlTableConstraints VARCHAR (4096);
    DECLARE @tmpTableName VARCHAR(256) = @tableName + '_noident_temp';

    BEGIN TRANSACTION

    -- 1) Create temporary table with edentical structure except identity
    -- Idea borrowed from https://stackoverflow.com/questions/21547/in-sql-server-how-do-i-generate-a-create-table-statement-for-a-given-table
    -- modified to ommit Identity and honor all constraints, not primary key only!
    SELECT
        @sql = 'CREATE TABLE [' + so.name + '_noident_temp] (' + o.list + ')'
        + ' ' + j.list
    FROM sysobjects so
    CROSS APPLY (
        SELECT
            ' [' + column_name + '] '
            + data_type
            + CASE data_type
                WHEN 'sql_variant' THEN ''
                WHEN 'text' THEN ''
                WHEN 'ntext' THEN ''
                WHEN 'xml' THEN ''
                WHEN 'decimal' THEN '(' + CAST(numeric_precision as VARCHAR) + ', ' + CAST(numeric_scale as VARCHAR) + ')'
                ELSE COALESCE('(' + CASE WHEN character_maximum_length = -1 THEN 'MAX' ELSE CAST(character_maximum_length as VARCHAR) END + ')', '')
            END
            + ' '
            /* + case when exists ( -- Identity skip
            select id from syscolumns
            where object_name(id)=so.name
            and name=column_name
            and columnproperty(id,name,'IsIdentity') = 1
            ) then
            'IDENTITY(' +
            cast(ident_seed(so.name) as varchar) + ',' +
            cast(ident_incr(so.name) as varchar) + ')'
            else ''
            end + ' ' */
            + CASE WHEN IS_NULLABLE = 'No' THEN 'NOT ' ELSE '' END
            + 'NULL'
            + CASE WHEN information_schema.columns.column_default IS NOT NULL THEN ' DEFAULT ' + information_schema.columns.column_default ELSE '' END
            + ','
        FROM
            INFORMATION_SCHEMA.COLUMNS
        WHERE table_name = so.name
        ORDER BY ordinal_position
        FOR XML PATH('')
    ) o (list)
    CROSS APPLY(
        SELECT
            CHAR(10) + 'ALTER TABLE ' + @tableName + '_noident_temp ADD ' + LEFT(alt, LEN(alt)-1)
        FROM(
            SELECT
                CHAR(10)
                + ' CONSTRAINT ' + tc.constraint_name  + '_ni ' + tc.constraint_type + ' (' + LEFT(c.list, LEN(c.list)-1) + ')'
                + COALESCE(CHAR(10) + r.list, ', ')
            FROM
                information_schema.table_constraints tc
                CROSS APPLY(
                    SELECT
                        '[' + kcu.column_name + '], '
                    FROM
                        information_schema.key_column_usage kcu
                    WHERE
                        kcu.constraint_name = tc.constraint_name
                    ORDER BY
                        kcu.ordinal_position
                    FOR XML PATH('')
                ) c (list)
                OUTER APPLY(
                    -- https://stackoverflow.com/questions/3907879/sql-server-howto-get-foreign-key-reference-from-information-schema
                    SELECT
                        '  REFERENCES [' + kcu1.constraint_schema + '].' + '[' + kcu2.table_name + ']' + '([' + kcu2.column_name + ']) '
                        + CHAR(10)
                        + '    ON DELETE ' + rc.delete_rule
                        + CHAR(10)
                        + '    ON UPDATE ' + rc.update_rule + ', '
                    FROM information_schema.referential_constraints as rc
                        JOIN information_schema.key_column_usage as kcu1 ON (kcu1.constraint_catalog = rc.constraint_catalog AND kcu1.constraint_schema = rc.constraint_schema AND kcu1.constraint_name = rc.constraint_name)
                        JOIN information_schema.key_column_usage as kcu2 ON (kcu2.constraint_catalog = rc.unique_constraint_catalog AND kcu2.constraint_schema = rc.unique_constraint_schema AND kcu2.constraint_name = rc.unique_constraint_name AND kcu2.ordinal_position = KCU1.ordinal_position)
                    WHERE
                        kcu1.constraint_catalog = tc.constraint_catalog AND kcu1.constraint_schema = tc.constraint_schema AND kcu1.constraint_name = tc.constraint_name
                ) r (list)
            WHERE tc.table_name = @tableName
            FOR XML PATH('')
        ) a (alt)
    ) j (list)
    WHERE
        xtype = 'U'
    AND name NOT IN ('dtproperties')
    AND so.name = @tableName

    SELECT @sql as '1) @sql';
    EXECUTE(@sql);

    -- 2) Obtain current back references on our table from others to reenable it later
    -- https://stackoverflow.com/questions/3907879/sql-server-howto-get-foreign-key-reference-from-information-schema
    SELECT
        @sqlTableConstraints = (
            SELECT
                'ALTER TABLE [' + kcu1.constraint_schema + '].' + '[' + kcu1.table_name + ']'
                + ' ADD CONSTRAINT ' + kcu1.constraint_name + '_ni FOREIGN KEY ([' + kcu1.column_name + '])'
                + CHAR(10)
                + '  REFERENCES ['  + kcu2.table_schema + '].[' + kcu2.table_name + ']([' + kcu2.column_name + '])'
                + CHAR(10)
                + '    ON DELETE ' + rc.delete_rule
                + CHAR(10)
                + '    ON UPDATE ' + rc.update_rule + ' '
            FROM information_schema.referential_constraints as rc
                JOIN information_schema.key_column_usage as kcu1 ON (kcu1.constraint_catalog = rc.constraint_catalog AND kcu1.constraint_schema = rc.constraint_schema AND kcu1.constraint_name = rc.constraint_name)
                JOIN information_schema.key_column_usage as kcu2 ON (kcu2.constraint_catalog = rc.unique_constraint_catalog AND kcu2.constraint_schema = rc.unique_constraint_schema AND kcu2.constraint_name = rc.unique_constraint_name AND kcu2.ordinal_position = KCU1.ordinal_position)
            WHERE
                kcu2.table_name = 'department'
            FOR XML PATH('')
        );
    SELECT @sqlTableConstraints as '8) @sqlTableConstraints';
    -- Execute at end

    -- 3) Drop outer references for switch (structure must be identical: http://msdn.microsoft.com/en-gb/library/ms191160.aspx) and rename table
    SELECT
        @sql = (
            SELECT
                ' ALTER TABLE [' + kcu1.constraint_schema + '].' + '[' + kcu1.table_name + '] DROP CONSTRAINT ' + kcu1.constraint_name
            FROM information_schema.referential_constraints as rc
                JOIN information_schema.key_column_usage as kcu1 ON (kcu1.constraint_catalog = rc.constraint_catalog AND kcu1.constraint_schema = rc.constraint_schema AND kcu1.constraint_name = rc.constraint_name)
                JOIN information_schema.key_column_usage as kcu2 ON (kcu2.constraint_catalog = rc.unique_constraint_catalog AND kcu2.constraint_schema = rc.unique_constraint_schema AND kcu2.constraint_name = rc.unique_constraint_name AND kcu2.ordinal_position = KCU1.ordinal_position)
            WHERE
                kcu2.table_name = @tableName
            FOR XML PATH('')
        );
    SELECT @sql as '3) @sql'
    EXECUTE (@sql);

    -- 4) Switch partition
    -- http://www.calsql.com/2012/05/removing-identity-property-taking-more.html
    SET @sql = 'ALTER TABLE ' + @tableName + ' switch partition 1 to ' + @tmpTableName;
    SELECT @sql as '4) @sql';
    EXECUTE(@sql);

    -- 5) Rename real old table to bak
    SET @sql = 'EXEC sp_rename ' + @tableName + ', ' + @tableName + '_bak';
    SELECT @sql as '5) @sql';
    EXECUTE(@sql);

    -- 6) Rename temp table to real
    SET @sql = 'EXEC sp_rename ' + @tmpTableName + ', ' + @tableName;
    SELECT @sql as '6) @sql';
    EXECUTE(@sql);

    -- 7) Drop bak table
    SET @sql = 'DROP TABLE ' + @tableName + '_bak';
    SELECT @sql as '7) @sql';
    EXECUTE(@sql);

    -- 8) Create again doped early constraints
    SELECT @sqlTableConstraints as '8) @sqlTableConstraints';
    EXECUTE(@sqlTableConstraints);


    -- It still may fail if there references from objects with WITH CHECKOPTION
    -- it may be recreated - https://stackoverflow.com/questions/1540988/sql-2005-force-table-rename-that-has-dependencies
    COMMIT
END

Use is pretty simple:

EXEC sp_drop_table_identity @tableName = 'some_very_big_table'

Benefits and limitations:

1. It uses switch partition (applicable to not partitioned tables too) statement for fast move without full data copy . It also apply some conditions for applicability.
2. It make on the fly table copy without identity. Such solution I also post separately and it also may need tuning on not so trivial structures like compound fields (it cover my needs).
3. If table included in objects with schema bound by CHECKOUPTION (sp, views) it prevent do switching (see last comment in code). It may be additionally scripted to temporary drop such binding. I had not do that yet.

All feedback welcome.

Answer 5

Most efficient way to drop identity columns (especially for large databases) on SQL Server is to modify DDL metadata directly, on SQL Server older than 2005 this can be done with:

sp_configure 'allow update', 1
go
reconfigure with override
go

update syscolumns set colstat = 0 --turn off bit 1 which indicates identity column
where id = object_id('table_name') and name = 'column_name'
go

exec sp_configure 'allow update', 0
go
reconfigure with override
go

SQL Server 2005+ doesn't support reconfigure with override , but you can execute Ad Hoc Queries when SQL Server instance is started in single-user mode (start db instance with -m flag, ie "C:\\Program Files\\Microsoft SQL Server\\MSSQL10.MSSQLSERVER\\MSSQL\\Binn\\sqlservr.exe -m", make sure to run as Administrator) with Dedicated Admin Console (from SQL Management Studio connect with ADMIN: prefix, ie ADMIN:MyDatabase). Column metdata is stored in sys.sysschobjs internal table (not shown without DAC):

use myDatabase

update sys.syscolpars set status = 1, idtval = null -- status=1 - primary key, idtval=null - remove identity data
where id = object_id('table_name') AND name = 'column_name' 

More on this approach on this blog