DB2® ensures referential integrity between your tables when you define referential constraints.
Referential integrity is the state in which all values of all foreign keys are valid. Referential integrity is based on entity integrity. Entity integrity requires that each entity have a unique key. For example, if every row in a table represents relationships for a unique entity, the table should have one column or a set of columns that provides a unique identifier for the rows of the table. This column (or set of columns) is called the parent key of the table. To ensure that the parent key does not contain duplicate values, a unique index must be defined on the column or columns that constitute the parent key. Defining the parent key is called entity integrity.
A referential constraint is the rule that the nonnull values of a foreign key are valid only if they also appear as values of a parent key. The table that contains the parent key is called the parent table of the referential constraint, and the table that contains the foreign key is a dependent of that table.
The relationship between some rows of the DEPT and EMP tables, shown in the following figure, illustrates referential integrity concepts and terminology. For example, referential integrity ensures that every foreign key value in the DEPT column of the EMP table matches a primary key value in the DEPTNO column of the DEPT table.
Two parent and dependent relationships exist between the DEPT and EMP tables.
- The foreign key on the DEPT column establishes a parent and dependent relationship. The DEPT column in the EMP table depends on the DEPTNO in the DEPT table. Through this foreign key relationship, the DEPT table is the parent of the EMP table. You can assign an employee to no department (by specifying a null value), but you cannot assign an employee to a department that does not exist.
- The foreign key on the MGRNO column also establishes a parent and dependent relationship. Because MGRNO depends on EMPNO, EMP is the parent table of the relationship, and DEPT is the dependent table.
You can define a primary key on one or more columns. A primary key that includes two or more columns is called a composite key. A foreign key can also include one or more columns. When a foreign key contains multiple columns, the corresponding primary key must be a composite key. The number of foreign key columns must be the same as the number of columns in the parent key, and the data types of the corresponding columns must be compatible. (The sample project activity table, DSN8A10.PROJACT, is an example of a table with a primary key on multiple columns, PROJNO, ACTNO, and ACSTDATE.)
A table can be a dependent of itself; this is called a self-referencing table. For example, the DEPT table is self-referencing because the value of the administrative department (ADMRDEPT) must be a department ID (DEPTNO). To enforce the self-referencing constraint, DB2 requires that a foreign key be defined.
Similar terminology applies to the rows of a parent-and-child relationship. A row in a dependent table, called a dependent row, refers to a row in a parent table, called a parent row. But a row of a parent table is not always a parent row—perhaps nothing refers to it. Likewise, a row of a dependent table is not always a dependent row—the foreign key can allow null values, which refer to no other rows.
Referential constraints are optional. You define referential constraints by using CREATE TABLE and ALTER TABLE statements.
DB2 enforces referential constraints when the following actions occur:
- An INSERT statement is applied to a dependent table.
- An UPDATE statement is applied to a foreign key of a dependent table or to the parent key of a parent table.
- A MERGE statement that includes an insert operation is applied to a dependent table.
- A MERGE statement that includes an update operation is applied to a foreign key of a dependent table or to the parent key of a parent table.
- A DELETE statement is applied to a parent table. All affected referential constraints and all delete rules of all affected relationships must be satisfied in order for the delete operation to succeed.
- The LOAD utility with the ENFORCE CONSTRAINTS option is run on a dependent table.
- The CHECK DATA utility is run.
Another type of referential constraint is an informational referential constraint. This type of constraint is not enforced by DB2 during normal operations. An application process should verify the data in the referential integrity relationship. An informational referential constraint allows queries to take advantage of materialized query tables.
The order in which referential constraints are enforced is undefined. To ensure that the order does not affect the result of the operation, there are restrictions on the definition of delete rules and on the use of certain statements. The restrictions are specified in the descriptions of the SQL statements CREATE TABLE, ALTER TABLE, INSERT, UPDATE, MERGE, and DELETE.
- parent key
- A primary key or a unique key of a referential constraint.
- parent table
- A table that is a parent in at least one referential constraint. A table can be defined as a parent in an arbitrary number of referential constraints.
- dependent table
- A table that is a dependent in at least one referential constraint. A table can be defined as a dependent in an arbitrary number of referential constraints. A dependent table can also be a parent table.
- descendent table
- A table that is a dependent of another table or a table that is a dependent of a descendent table.
- referential cycle
- A set of referential constraints in which each associated table is a descendent of itself.
- parent row
- A row that has at least one dependent row.
- dependent row
- A row that has at least one parent row.
- descendent row
- A row that is dependent on another row or a row that is a dependent of a descendent row.
- self-referencing row
- A row that is a parent of itself.
- self-referencing table
- A table that is both parent and dependent in the same referential constraint. The constraint is called a self-referencing constraint.
A temporal referential constraint can be defined for a table that contains a BUSINESS_TIME period. The PERIOD BUSINESS_TIME clause is used in both the FOREIGN KEY clause and the REFERENCES clause to indicate that there must not be a row in the child table for which the period of time represented by the BUSINESS_TIME period value for that row is not contained in the BUSINESS_TIME period of one or more corresponding rows in the parent table. Unlike normal referential constraints, it is not necessary that there be exactly one corresponding row in the parent table where the BUSINESS_TIME period contains the BUSINESS_TIME period of the child row. As long as the BUSINESS_TIME period of a row in the child table is contained in the union of the BUSINESS_TIME periods of two or more contiguous matching rows in the parent table, the temporal referential constraint is satisfied.
- A unique index on the parent table with the BUSINESS_TIME WITHOUT OVERLAPS clause.
- A non-unique index on the child table with the BUSINESS_TIME WITH OVERLAPS clause. Alternatively, the index on the child table, can be defined without the BUSINESS_TIME WITH OVERLAPS clause if the end of the index key includes the end column followed by the begin column of the BUSINESS_TIME period (both in ascending order).
- insert rule
- A nonnull insert value of the foreign key must match some value of the parent key of the parent table. The value of a composite foreign key is null if any component of the value is null.
- update rule
- A nonnull update value of the foreign key must match some value of the parent key of the parent table. The value of a composite foreign key is treated as null if any component of the value is null.
- delete rule
- Controls what happens when a row of the parent table is deleted. The choices of action, made when the referential constraint is defined, are RESTRICT, NO ACTION, CASCADE, or SET NULL. SET NULL can be specified only if some column of the foreign key allows null values.
More precisely, the delete rule applies when a row of the parent table is the object of a delete or propagated delete operation and that row has dependents in the dependent table of the referential constraint. A propagated delete refers to the situation where dependent rows are deleted when parent rows are deleted. Let P denote the parent table, let D denote the dependent table, and let p denote a parent row that is the object of a delete or propagated delete operation. If the delete rule is:
- RESTRICT or NO ACTION, an error occurs and no rows are deleted.
- CASCADE, the delete operation is propagated to the dependent rows of p in D.
- SET NULL, each nullable column of the foreign key of each dependent row of p in D is set to null.
Each referential constraint in which a table is a parent has its own delete rule, and all applicable delete rules are used to determine the result of a delete operation. Thus, a row cannot be deleted if it has dependents in a referential constraint with a delete rule of RESTRICT or NO ACTION or the deletion cascades to any of its descendents that are dependents in a referential constraint with the delete rule of RESTRICT or NO ACTION.
The deletion of a row from parent table P involves other tables and can affect rows of these tables:
- If D is a dependent of P and the delete rule is RESTRICT or NO ACTION, D is involved in the operation but is not affected by the operation and the deletion from the parent table P does not take place.
- If D is a dependent of P and the delete rule is SET NULL, D is involved in the operation and rows of D might be updated during the operation.
- If D is a dependent of P and the delete rule is CASCADE, D is involved in the operation and rows of D might be deleted during the operation. If rows of D are deleted, the delete operation on P is said to be propagated to D. If D is also a parent table, the actions described in this list apply, in turn, to the dependents of D.
Any table that can be involved in a delete operation on P is said to be delete-connected to P. Thus, a table is delete-connected to table P if it is a dependent of P or a dependent of a table to which delete operations from P cascade.