I'm trying to create a table variable with primary key. What is the difference between these two statements? Why the first one does not work?
--This does not work
declare @tabvar table (
rowid int identity(1, 1) not null,
var1 int null,
constraint PK_tabvar_rowid primary key clustered (rowid)
)
--This works
declare @tabvar1 table (
rowid int identity(1, 1) not null primary key clustered,
var1 int null
)
Looks like this behavior is well documented in BOL. Look at the syntax definitions for DECLARE TABLE Variable vs CREATE TABLE.
When you create a table like:
CREATE TABLE tabvar (
rowid int identity(1, 1) not null,
var1 int null
, constraint PK_tabvar_rowid primary key clustered (rowid))
you create a separate SQL object called PK_tabvar_rowid.
This method is preferred for permanent tables as above, because you specifically name the constraint and it exists independently from the table object.
You CAN use the form:
CREATE TABLE tabvar (
rowid int identity(1, 1) not null primary key,
var1 int null)
but this creates a randomly named constraint, which makes future management more difficult.
For table variables (which are transient) - you CANNOT have an independent constraint - so you MUST use the inline primary key definition.
There are differences in the syntax between the CREATE TABLE (Transact-SQL) and DECLARE @local_variable (Transact-SQL) used in creating table variables. You can not use the syntax of the first statement to create a variable table.
The syntax you're using is for CREATE TABLE
, not table variables.
CREATE TABLE tabvar (
rowid int identity(1, 1) not null,
var1 int null
, constraint PK_tabvar_rowid primary key clustered (rowid)
)
The above works fine, as expected. However if you look at the syntax for declaring table variables you'll see what you can and can't do:
declare @tabvar table (
rowid int identity(1, 1) not null,
var1 int null,
primary key (rowid)
)