T-SQL插入表中而不必指定每一列

Question

In our db there is a table that has a little over 80 columns. It has a primary key and Identity insert is turned on. I'm looking for a way to insert into this table every column EXCEPT the primary key column from an identical table in a different DB.

Is this possible?

Answer 1

You can do this quite easily actually:

-- Select everything into temp table
Select * Into 
    #tmpBigTable
    From [YourBigTable]

-- Drop the Primary Key Column from the temp table  
Alter Table #tmpBigTable Drop Column [PrimaryKeyColumn]

-- Insert that into your other big table
Insert Into [YourOtherBigTable]
    Select * From #tmpBigTable

-- Drop the temp table you created
Drop Table #tmpBigTable

Provided you have Identity Insert On in "YourOtherBigTable" and columns are absolutely identical you will be okay.

Answer 2

CREATE TABLE Tests
(
    TestID int IDENTITY PRIMARY KEY,
    A int,
    B int,
    C int
)

INSERT INTO dbo.Tests
VALUES (1,2,3)

SELECT * FROM Tests

This works in SQL2012

Answer 3

You could query Information_Schema to get a list of all the columns and programatically generate the column names for your query. If you're doing this all in t-sql it would be cumbersome, but it could be done. If you're using some other client language, like C# to do the operation, it would be a little less cumbersome.

Answer 4

No, that's not possible. You could be tempted to use

INSERT INTO MyLargeTable SELECT * FROM OtherTable

But that would not work, because your identity column would be included in the *.

You could use

SET IDENTITY_INSERT MyLargeTable ON
INSERT INTO MyLargeTable SELECT * FROM OtherTable
SET IDENTITY_INSERT MyLargeTable OFF

first you enable inserting identity values, than you copy the records, then you enable the identity column again.

But this won't work neither. SQL server won't accept the * in this case. You have to explicitly include the Id in the script, like :

SET IDENTITY_INSERT MyLargeTable ON
INSERT INTO MyLargeTable (Id, co1, col2, ...., col80) SELECT Id, co1, col2, ...., col80 FROM OtherTable
SET IDENTITY_INSERT MyLargeTable OFF

So we're back from where we started.

The easiest way is to right click the table in Management Studio, let it generate the INSERT and SELECT scripts, and edit them a little to let them work together.

Answer 5

Why not just create a VIEW of the original data, removing the unwanted fields? Then 'Select * into' your hearts desire.

• Localized control within a single view
• No need to modify SPROC
• Add/change/delete fields easy
• No need to query meta-data
• No temporary tables

Answer 6

Really, honestly it takes ten seconds or less to pull all of the columns over from the object browser and then delete the identity column from the list. It is a bad idea to use select * for anything but quick ad hoc query.

Answer 7

In answer to a related question ( SELECT * EXCEPT ), I point out the truly relational language Tutorial D allows projection to be expressed in terms of the attributes to be removed instead of the ones to be kept eg

my_relvar { ALL BUT description }

However its INSERT syntax requires tuple value constructors to include attribute name / value pairs eg

INSERT P
   RELATION 
   {
      TUPLE { PNO PNO ( 'P1' ) , PNAME CHARACTER ( 'Nut' ) }, 
      TUPLE { PNO PNO ( 'P2' ) , PNAME CHARACTER ( 'Bolt' ) }
   };

Of course, using this syntax there is no column ordering (because it is truly relational!) eg this is semantically equivalent:

INSERT P
   RELATION 
   {
      TUPLE { PNO PNO ( 'P1' ) , PNAME CHARACTER ( 'Nut' ) }, 
      TUPLE { PNAME CHARACTER ( 'Bolt' ) , PNO PNO ( 'P2' ) }
   };

The alternative would be to rely fully on attribute ordering, which SQL does partially eg this is a close SQL equivalent to the the above:

INSERT INTO P ( PNO , PNAME ) 
   VALUES        
      ( PNO ( 'P1' ) , CAST ( 'Nut'  AS VARCHAR ( 20 ) ) ) , 
      ( PNO ( 'P2' ) , CAST ( 'Bolt' AS VARCHAR ( 20 ) ) );

Once the commalist of columns has been specified the VALUES row constructors have the maintain this order, which is not ideal. But at least the order is specified: your proposal would rely on some default order which may be possibly non-deterministic.