ALTER TABLE (SQL)

Arguments

Privileges and Locking

The ALTER TABLE command is a privileged operation. The user must have %ALTER_TABLE administrative privilege to execute ALTER TABLE. Failing to do so results in an SQLCODE –99 error with the %msg User 'name' does not have %ALTER_TABLE privileges.

The user must have %ALTER privilege on the specified table. If the user is the Owner (creator) of the table, the user is automatically granted %ALTER privilege for that table. Otherwise, the user must be granted %ALTER privilege for the table. Failing to do so results in an SQLCODE –99 error with the %msg User 'name' does not have required %ALTER privilege needed to change the table definition for 'Schema.TableName'.

To determine if the current user has %ALTER privilege, invoke the %CHECKPRIV command. To determine if a specified user has %ALTER privilege, invoke the $SYSTEM.SQL.Security.CheckPrivilege()Opens in a new tab method.

To assign the required administrative privilege, use the GRANT command with %ALTER_TABLE privilege; this requires the appropriate granting privileges. To assign the %ALTER object privilege, you can use:

• The GRANT command with the %ALTER privilege. This requires the appropriate granting privileges.

• The ALTER check box for the table on the SQL Tables tab in the Management Portal on the page for editing a role or user. This requires the appropriate granting privileges.

In embedded SQL, you can use the $SYSTEM.Security.Login()Opens in a new tab method to log in as a user with appropriate privileges:

   DO $SYSTEM.Security.Login("_SYSTEM","SYS")
   &sql(      )

You must have the %Service_Login:Use privilege to invoke the $SYSTEM.Security.Login method. For further information, see %SYSTEM.SecurityOpens in a new tab.

• ALTER TABLE cannot be used on a table projected from a persistent class, unless the table class definition includes [DdlAllowed]. Otherwise, the operation fails with an SQLCODE -300 error with the %msg DDL not enabled for class 'Schema.tablename'.

• ALTER TABLE cannot be used on a table projected from a deployed persistent class. This operation fails with an SQLCODE -400 error with the %msg Unable to execute DDL that modifies a deployed class: 'classname'.

ALTER TABLE acquires a table-level lock on table. This prevents other processes from modifying the table’s data. This lock is automatically released at the conclusion of the ALTER TABLE operation. When ALTER TABLE locks the corresponding class definition, it uses the SQL Lock Timeout setting for the current process.

To alter a table, the table cannot be locked by another process in either EXCLUSIVE MODE or SHARE MODE. Attempting to alter a locked table results in an SQLCODE -110 error, with a %msg such as the following: Unable to acquire exclusive table lock for table 'Sample.MyTest'.

RENAME Table

You can rename an existing table using the following syntax:

ALTER TABLE schema.TableName RENAME NewTableName

This operation renames the existing table in its existing schema. You can only change the table name, not the table schema. Specifying a schema name in the NewTableName results in an SQLCODE -1 error. Specifying the same table name for both old and new tables generates an SQLCODE -201 error.

Renaming a table changes the SQL table name. It does not change the corresponding persistent class name.

Renaming a table does not change references to the old table name in triggers.

If a view references the existing table name, attempting to rename the table will fail. This is because attempting to rename the table is an atomic operation that causes a recompile of the view, which generates an SQLCODE -30 error “Table 'schema.oldname' not found”.

ADD COLUMN Restrictions

ADD COLUMN can add a single column, or can add a comma-separated list of columns.

If you attempt to add a field to a table through an ALTER TABLE tablename ADD COLUMN statement:

• If a column of that name already exists, the statement fails with an SQLCODE -306 error.

• If the statement specifies a NOT NULL constraint on the column and there is no default value for the column, then the statement fails if data already exists in the table. This is because, after the completion of the DDL statement, the NOT NULL constraint is not satisfied for all the pre-existing rows. This generates the error code SQLCODE -304 (Attempt to add a NOT NULL field with no default value to a table which contains data).

• If the statement specifies a NOT NULL constraint on the column and there is a default value for the column, the statement updates any existing rows in the table and assigns the default value for the column to the field. This includes default values such as CURRENT_TIMESTAMP.

• If the statement DOES NOT specify a NOT NULL constraint on the column and there is a default value for the column, then there are no updates of the column in any existing rows. The column value is NULL for those rows.

To change this default NOT NULL constraint behaviors, refer to the COMPILEMODE=NOCHECK option of the SET OPTION command.

If you specify an ordinary data field named “ID” and the RowID field is already named “ID” (the default), the ADD COLUMN operation succeeds. ALTER TABLE adds the ID data column, and renames the RowId column as “ID1” to avoid duplicate names.

ALTER COLUMN Restrictions

ALTER COLUMN can modify the definition of a single column:

• Rename the column using the syntax ALTER TABLE tablename ALTER COLUMN oldname RENAME newname. Renaming a column changes the SQL field name. It does not change the corresponding persistent class property name. ALTER COLUMN oldname RENAME newname replaces oldfield name references in trigger code and ComputeCode.

• Change the column characteristics: data type, default value, NULL/NOT NULL, and collation type.

If the table contains data, you cannot change the data type of a column that contains data if this change would result in stream data being typed as non-stream data or non-stream data being typed as stream data. Attempting to do so results in an SQLCODE -374 error. If there is no existing data, this type of datatype change is permitted.

You can use ALTER COLUMN to add, change, or drop a field default value.

If the table contains data, you cannot specify NOT NULL for a column if that column contains null values; this results in an SQLCODE -305 error.

If you change the collation type for a column that contains data, you must rebuild all indexes for that column.

MODIFY column Restrictions

MODIFY can modify the definitions of a single column or a comma-separated list of columns.

• Rename the column using the syntax ALTER TABLE tablename MODIFY oldname RENAME newname. Renaming a column changes the SQL field name. It does not change the corresponding persistent class property name. MODIFY oldname RENAME newname replaces oldfield name references in trigger code and ComputeCode.

• Change the column characteristics: data type, default value, and other characteristics.

If the table contains data, you cannot change the data type of a column that contains data to an incompatible data type:

• A data type with a lower (less inclusive) data type precedence if this conflicts with existing data values. Attempting to do so results in an SQLCODE -104 error, with the %msg specifying which field and which data value caused the error.

• A data type with a smaller MAXLEN or a MAXVAL/MINVAL if this conflicts with existing data values. Attempting to do so results in an SQLCODE -104 error, with the %msg specifying which field and which data value caused the error.

• A data type change from a stream data type to a non-stream data type or a non-stream data type to a stream data type. Attempting to do so results in an SQLCODE -374 error. If there is no existing data, this type of datatype change is permitted.

You can use MODIFY to add or change a field default value. You cannot use MODIFY to drop a field default value.

If the table contains data, you cannot specify NOT NULL for a column if that column contains null values; this results in an SQLCODE -305 error. The syntax forms ALTER TABLE mytable MODIFY field1 NOT NULL and ALTER TABLE mytable MODIFY field1 CONSTRAINT nevernull NOT NULL perform the same operation. The optional CONSTRAINT identifier clause is a no-op provided for compatibility. InterSystems IRIS does not retain or use this field constraint name. Attempting to drop this field constraint by specifying this field constraint name results in an SQLCODE -315 error.

DROP COLUMN Restrictions

DROP COLUMN can delete multiple column definitions, specified as a comma-separated list. Each listed column name must be followed by its RESTRICT or CASCADE (if unspecified, the default is RESTRICT) and %DELDATA or %NODELDATE (if unspecified, the default is %NODELDATA) options.

By default, deleting a column definition does not delete any data that has been stored in that column from the data map. To delete both the column definition and the data, specify the %DELDATA option.

Deleting a column definition does not delete the corresponding column-level privileges. For example, the privilege granted to a user to insert, update, or delete data on that column. This has the following consequences:

• If a column is deleted, and then another column with the same name is added, users and roles will have the same privileges on the new column that they had on the old column.

• Once a column is deleted, it is not possible to revoke object privileges for that column.

For these reasons, it is generally recommended that you use the REVOKE command to revoke column-level privileges from a column before deleting the column definition.

RESTRICT keyword (or no keyword): You cannot drop a column if that column appears in an index, or is defined in a foreign key constraint or other unique constraint. Attempting a DROP COLUMN for that column fails with an SQLCODE -322 error. RESTRICT is the default. See DROP INDEX.

CASCADE keyword: If the column appears in an index, the index will be deleted. There may be multiple indexes. If the column appears in a foreign key, the foreign key will be deleted. There may be multiple foreign keys.

You cannot drop a column if that column is used in COMPUTECODE or in a COMPUTEONCHANGE clause. Attempting to do so results in an SQLCODE -400 error.

ADD CONSTRAINT Restrictions

You can add a constraint to a comma-separated list of fields. For example, you can add the UNIQUE (FName,SurName) constraint, which establishes a UNIQUE constraint on the combined values of the two fields FName and Surname. Similarly, you can add a primary key constraint or a foreign key constraint on a comma-separated list of fields.

A constraint can be named or unnamed. If unnamed, InterSystems SQL generates a constraint name using the table name. For example, MYTABLE_Unique1 or MYTABLE_PKEY1.

The following example creates two unnamed constraints, adding both the unique constraint and the primary key constraint to comma-separated lists of fields:

 ALTER TABLE SQLUser.MyStudents 
    ADD UNIQUE (FName,SurName),PRIMARY KEY (Fname,Surname)  

• A field must exist to be used in a constraint. Specifying a non-existent field generates an SQLCODE -31 error.

• The RowId field cannot be used in a constraint. Specifying the RowId (ID) field generates SQLCODE -31 error.

• A stream field cannot be used in a constraint. Specifying a stream field generates an SQLCODE -400 error: “invalid index attribute”

• A constraint can only be applied once to a field. Specifying the same constraint twice to a field generates an SQLCODE -400 error: “index name conflict”.

By using the optional CONSTRAINT identifier keyword clause, you can create a named constraint. A named constraint must be a valid identifier; constraint names are not case-sensitive. This provides a name for the constraint for future use. This is shown in the following example:

 ALTER TABLE SQLUser.MyStudents 
    ADD CONSTRAINT UnqFullName UNIQUE (FName,SurName)  

You can specify multiple constraints as a comma-separated list; the constraint name is applied to the first constraint, the other constraints receive default names.

A constraint name must be unique for the table. Specifying the same constraint name twice to a field generates an SQLCODE -400 error: “index name conflict”.

ADD PRIMARY KEY Restrictions

A primary key value is required and unique. Therefore, adding a primary key constraint to an existing field or combination of fields makes each of these fields a required field. If you add a primary key constraint to a list of existing fields, the combined values of these fields must be unique. You cannot add a primary key constraint to an existing field if that field permits NULL values. You cannot add a primary key constraint to a field (or list of fields) if that field (or list of fields) contain non-unique values.

If you add a primary key constraint to an existing field, the field may also be automatically defined as an IDKey index. This depends on whether data is present and upon a configuration setting established in one of the following ways:

• The SQL SET OPTION PKEY_IS_IDKEY statement.

• The system-wide $SYSTEM.SQL.Util.SetOption()Opens in a new tab method configuration option DDLPKeyNotIDKey. To determine the current setting, call $SYSTEM.SQL.CurrentSettings()Opens in a new tab which displays Are primary keys created through DDL not ID keys; the default is 1.

• Go to the Management Portal, select System Administration, Configuration, SQL and Object Settings, SQL. View the current setting of Define primary key as ID key for tables created via DDL.

  • If the check box is not selected (the default), the Primary Key does not becomes the IDKey index in the class definition. Access to records using a primary key that is not the IDKEY is significantly less efficient; however, this type of primary key value can be modified.

  • If the check box is selected, when a Primary Key constraint is specified through DDL, and the field does not contain data, the primary key index is also defined as the IDKey index. If the field does contain data, the IDKey index is not defined. If the primary key is defined as the IDKey index, data access is more efficient, but a primary key value, once set, can never be modified.

If CREATE TABLE defined a bitmap extent index and later you use ALTER TABLE to add a primary key that is also the IDKey, the system automatically drops the bitmap extent index.

ADD PRIMARY KEY When Already Exists

You can only define one primary key. By default, InterSystems IRIS rejects an attempt to define a primary key when one already exists, or to define the same primary key twice, and issues an SQLCODE -307 error. The SQLCODE -307 error is issued even if the second definition of the primary key is identical to the first definition. To determine the current configuration, call $SYSTEM.SQL.CurrentSettings()Opens in a new tab, which displays a Allow create primary key through DDL when key exists setting. The default is 0 (No), which is the recommended configuration setting. If this option is set to 1 (Yes), an ALTER TABLE ADD PRIMARY KEY causes InterSystems IRIS to remove the primary key index from the class definition, and then recreates this index using the specified primary key field(s).

From the Management Portal, System Administration, Configuration, SQL and Object Settings, SQL you can set this option (and other similar create, alter, and drop options) system-wide by selecting the Ignore redundant DDL statements check box.

However, even if this option is set to allow the creation of a primary key when one already exists, you cannot recreate a primary key index if it is also the IDKEY index and the table contains data. Attempting to do so generates an SQLCODE -307 error.

ADD FOREIGN KEY Restrictions

For information on foreign keys, refer to Defining Foreign Keys and Foreign Key Referential Action Clause in the CREATE TABLE command, and to Using Foreign Keys.

By default, you cannot have two foreign keys with the same name. Attempting to do so generates an SQLCODE -311 error. To determine the current setting, call $SYSTEM.SQL.CurrentSettings()Opens in a new tab, which displays a Allow DDL ADD foreign key constraint when foreign key exists setting. The default is 0 (No), which is the recommended setting for this option. When 1 (Yes), you can add a foreign key through DDL even if one with the same name already exists.

From the Management Portal, System Administration, Configuration, SQL and Object Settings, SQL you can set this option (and other similar create, alter, and drop options) system-wide by selecting the Ignore redundant DDL statements check box.

Your table definition should not have two foreign keys with different names that reference the same field-commalist field(s) and perform contradictory referential actions. In accordance with the ANSI standard, InterSystems SQL does not issue an error if you define two foreign keys that perform contradictory referential actions on the same field (for example, ON DELETE CASCADE and ON DELETE SET NULL). Instead, InterSystems SQL issues an error when a DELETE or UPDATE operation encounters these contradictory foreign key definitions.

An ADD FOREIGN KEY that specifies a non-existent foreign key field generates an SQLCODE -31 error.

An ADD FOREIGN KEY that references a non-existent parent key table generates an SQLCODE -310 error. An ADD FOREIGN KEY that references a non-existent field in an existing parent key table generates an SQLCODE -316 error. If you do not specify a parent key field, it defaults to the ID field.

Before issuing an ADD FOREIGN KEY, the user must have REFERENCES privilege on the table being referenced or on the columns of the table being referenced. REFERENCES privilege is required if the ALTER TABLE is executed via Dynamic SQL or over a SQL driver connection.

An ADD FOREIGN KEY that references a field (or combination of fields) that can take non-unique values generates an SQLCODE -314 error, with additional details available through %msg.

NO ACTION is the only referential action supported for sharded tables.

An ADD FOREIGN KEY is constrained when data already exists in the table. To change this default constraint behavior, refer to the COMPILEMODE=NOCHECK option of the SET OPTION command.

When you define an ADD FOREIGN KEY constraint for a single field and the foreign key references the idkey of the referenced table, InterSystems IRIS converts the property in the foreign key into a reference property. This conversion is subject to the following restrictions:

• The table must contain no data.

• The property on the foreign key cannot be of a persistent class (that is, it cannot already be a reference property).

• The data types and data type parameters of the foreign key field and the referenced idkey field must be the same.

• The foreign key field cannot be an IDENTITY field.

DROP CONSTRAINT Restrictions

By default, you cannot drop a unique or primary key constraint if it is referenced by a foreign key constraint. Attempting to do so results in an SQLCODE -317 error. To change this default foreign key constraint behavior, refer to the COMPILEMODE=NOCHECK option of the SET OPTION command.

The effects of dropping a primary key constraint depend on the setting of the Are Primary Keys also ID Keys setting (as described above):

• If the PrimaryKey index is not also the IDKey index, dropping the primary key constraint drops the index definition.

• If the PrimaryKey index is also the IDKey index, and there is no data in the table, dropping the primary key constraint drops the entire index definition.

• If the PrimaryKey index is also the IDKey index, and there is data in the table, dropping the primary key constraint just drops the PRIMARYKEY qualifier from the IDKey index definition.

DROP CONSTRAINT When Non-Existent

By default, InterSystems IRIS rejects an attempt to drop a field constraint on a field that does not have that constraint and issues an SQLCODE -315 error. To determine the current setting, call $SYSTEM.SQL.CurrentSettings()Opens in a new tab, which displays a Allow DDL DROP of non-existent constraint setting. The default is 0 (No), which is the recommended setting. If this option is set to 1 (Yes), an ALTER TABLE DROP CONSTRAINT causes InterSystems IRIS to perform no operation and issue no error message.

From the Management Portal, System Administration, Configuration, SQL and Object Settings, SQL you can set this option (and other similar create, alter, and drop options) system-wide by selecting the Ignore redundant DDL statements check box.