6 Applying an Object Model to Relational Data

Why Use Object Views

Just as a view is a virtual table, an object view is a virtual object table. Each row in the view is an object: you can call its methods, access its attributes using the dot notation, and create a REF that points to it.

Object views are useful in prototyping or transitioning to object-oriented applications because the data in the view can be taken from relational tables and accessed as if the table were defined as an object table. You can run object-oriented applications without converting existing tables to a different physical structure.

Object views can be used like relational views to present only the data that you want users to see. For example, you might create an object view that presents selected data from an employee table but omits sensitive data about salaries.

Using object views can lead to better performance. Relational data that make up a row of an object view traverse the network as a unit, potentially saving many round trips.

You can fetch relational data into the client-side object cache and map it into C structures or C++ or Java classes, so 3GL applications can manipulate it just like native classes. You can also use object-oriented features like complex object retrieval with relational data.

• By synthesizing objects from relational data, you can query the data in new ways. You can view data from multiple tables by using object de-referencing instead of writing complex joins with multiple tables.

• Because the objects in the view are processed within the server, not on the client, this can result in significantly fewer SQL statements and much less network traffic.

• The object data from object views can be pinned and used in the client side object cache. When you retrieve these synthesized objects in the object cache by means of specialized object-retrieval mechanisms, you reduce network traffic.

• You gain great flexibility when you create an object model within a view in that you can continue to develop the model. If you need to alter an object type, you can simply replace the invalidated views with a new definition.

• Using objects in views does not place any restrictions on the characteristics of the underlying storage mechanisms. By the same token, you are not limited by the restrictions of current technology. For example, you can synthesize objects from relational tables which are parallelized and partitioned.

• You can create different complex data models from the same underlying data.

  See Also:

  • Oracle Database SQL Language Reference for a complete description of SQL syntax and usage.

  • Oracle Database PL/SQL User's Guide and Reference for a complete discussion of PL/SQL capabilities

  • Oracle Database Java Developer's Guide for a complete discussion of Java.

  • Oracle Call Interface Programmer's Guide for a complete discussion of those facilities.

Defining Object Views

The procedure for defining an object view is:

1. Define an object type, where each attribute of the type corresponds to an existing column in a relational table.

2. Write a query that specifies how to extract the data from relational tables. Specify the columns in the same order as the attributes in the object type.

3. Specify a unique value, based on attributes of the underlying data, to serve as an object identifier, which enables you to create pointers (REFs) to the objects in the view. You can often use an existing primary key.

If you want to be able to update an object view, you may have to take another step, if the attributes of the object type do not correspond exactly to columns in existing tables:

Write an INSTEAD OF trigger procedure for Oracle to execute whenever an application program tries to update data in the object view. See "Updating Object Views".

After these steps, you can use an object view just like an object table.

For example, the following SQL statements define an object view, where each row in the view is an object of type employee_t:

CREATE TABLE emp_table (
    empnum   NUMBER (5),
    ename    VARCHAR2 (20),
    salary   NUMBER (9,2),
    job      VARCHAR2 (20));

CREATE TYPE employee_t AS OBJECT (
    empno    NUMBER (5),
    ename    VARCHAR2 (20),
    salary   NUMBER (9,2),
    job      VARCHAR2 (20));
/

CREATE VIEW emp_view1 OF employee_t
    WITH OBJECT IDENTIFIER (empno) AS
        SELECT e.empnum, e.ename, e.salary, e.job
            FROM emp_table e
            WHERE job = 'Developer';

To access the data from the empnum column of the relational table, you would access the empno attribute of the object type.

Using Object Views in Applications

Data in the rows of an object view may come from more than one table, but the object still traverses the network in one operation. The instance appears in the client side object cache as a C or C++ structure or as a PL/SQL object variable. You can manipulate it like any other native structure.

You can refer to object views in SQL statements in the same way you refer to an object table. For example, object views can appear in a SELECT list, in an UPDATE-SET clause, or in a WHERE clause.

You can also define object views on object views.

You can access object view data on the client side using the same OCI calls you use for objects from object tables. For example, you can use OCIObjectPin() for pinning a REF and OCIObjectFlush() for flushing an object to the server. When you update or flush to the server an object in an object view, Oracle updates the object view.

Nesting Objects in Object Views

An object type can have other object types nested in it as attributes.

If the object type on which an object view is based has an attribute that itself is an object type, then you must provide column objects for this attribute as part of the process of creating the object view. If column objects of the attribute type already exist in a relational table, you can simply select them; otherwise, you must synthesize the object instances from underlying relational data just as you synthesize the principal object instances of the view. You synthesize, or create, these objects by calling the respective object type's constructor method to create the object instances, and you populate their attributes with data from relational columns that you specify in the constructor.

For example, consider the department table dept in Example 6-2. You might want to create an object view where the addresses are objects inside the department objects. That would allow you to define reusable methods for address objects, and use them for all kinds of addresses.

First, create the types for the address and department objects, then create the view containing the department number, name and address. The address objects are constructed from columns of the relational table.

CREATE TABLE dept (
    deptno       NUMBER PRIMARY KEY,
    deptname     VARCHAR2(20),
    deptstreet   VARCHAR2(20),
    deptcity     VARCHAR2(10),
    deptstate    CHAR(2),
    deptzip      VARCHAR2(10));

CREATE TYPE address_t AS OBJECT (
   street   VARCHAR2(20),
    city    VARCHAR2(10),
    state   CHAR(2),
    zip     VARCHAR2(10));
/
CREATE TYPE dept_t AS OBJECT (
   deptno     NUMBER,
   deptname   VARCHAR2(20),
   address    address_t );
/

CREATE VIEW dept_view OF dept_t WITH OBJECT IDENTIFIER (deptno) AS
    SELECT d.deptno, d.deptname,
      address_t(d.deptstreet,d.deptcity,d.deptstate,d.deptzip) AS 
      deptaddr
      FROM dept d;

Identifying Null Objects in Object Views

Because the constructor for an object never returns a null, none of the address objects in the preceding view can ever be null, even if the city, street, and so on columns in the relational table are all null. The relational table has no column that specifies whether the department address is null. If we define a convention so that a null deptstreet column indicates that the whole address is null, then we can capture the logic using the DECODE function, or some other function, to return either a null or the constructed object:

CREATE OR REPLACE VIEW dept_view AS
  SELECT d.deptno, d.deptname,
        DECODE(d.deptstreet, NULL, NULL, 
        address_t(d.deptstreet, d.deptcity, d.deptstate, d.deptzip)) AS deptaddr
  FROM dept d;

Using such a technique makes it impossible to directly update the department address through the view, because it does not correspond directly to a column in the relational table. Instead, we would define an INSTEAD OF trigger over the view to handle updates to this column.

Using Nested Tables and Varrays in Object Views

Collections, both nested tables and VARRAYs, can be columns in views. You can select these collections from underlying collection columns or you can synthesize them using subqueries. The CAST-MULTISET operator provides a way of synthesizing such collections.

CREATE TABLE emp (
   empno    NUMBER PRIMARY KEY,
   empname  VARCHAR2(20),
   salary   NUMBER,
   job      VARCHAR2 (20), 
   deptno   NUMBER REFERENCES dept(deptno));

CREATE TYPE employee_list_t AS TABLE OF employee_t;
/
CREATE TYPE dept_t AS OBJECT (
    deptno     NUMBER,
    deptname   VARCHAR2(20),
    address    address_t,
    emp_list   employee_list_t);
/
CREATE VIEW dept_view OF dept_t WITH OBJECT IDENTIFIER (deptno) AS
    SELECT d.deptno, d.deptname,
     address_t(d.deptstreet,d.deptcity,d.deptstate,d.deptzip) AS deptaddr,
             CAST( MULTISET (
                           SELECT e.empno, e.empname, e.salary, e.job
                           FROM emp e 
                           WHERE e.deptno = d.deptno) 
                        AS employee_list_t)
                   AS emp_list
   FROM dept d;

CREATE TABLE depts
  ( deptno     NUMBER,
    deptname   VARCHAR2(20));

CREATE TABLE emps
  ( ename VARCHAR2(20),
    salary     NUMBER,
    deptname   VARCHAR2(20));

CREATE TABLE projects
  ( projname   VARCHAR2(20),
    mgr        VARCHAR2(20));

CREATE TYPE project_t AS OBJECT
  ( projname   VARCHAR2(20),
    mgr        VARCHAR2(20));
/
CREATE TYPE nt_project_t AS TABLE OF project_t;
/
CREATE TYPE emp_t AS OBJECT
(  ename      VARCHAR2(20),
   salary     NUMBER,
   deptname   VARCHAR2(20),
   projects   nt_project_t );
/
CREATE TYPE nt_emp_t AS TABLE OF emp_t;
/
CREATE TYPE depts_t AS OBJECT
  ( deptno     NUMBER,
    deptname   VARCHAR2(20),
    emps       nt_emp_t );
/
CREATE VIEW v_depts OF depts_t WITH OBJECT IDENTIFIER (deptno) AS
  SELECT d.deptno, d.deptname, 
    CAST(MULTISET(SELECT e.ename, e.salary, e.deptname,
        CAST(MULTISET(SELECT p.projname, p.mgr
          FROM projects p
          WHERE p.mgr = e.ename)
        AS nt_project_t)
      FROM emps e
      WHERE e.deptname = d.deptname)
    AS nt_emp_t)
  FROM depts d;

Specifying Object Identifiers for Object Views

You can construct pointers (REFs) to the row objects in an object view. Because the view data is not stored persistently, you must specify a set of distinct values to be used as object identifiers. The notion of object identifiers allows the objects in object views to be referenced and pinned in the object cache.

If the view is based on an object table or an object view, then there is already an object identifier associated with each row and you can reuse them. Either omit the WITH OBJECT IDENTIFIER clause, or specify WITH OBJECT IDENTIFIER DEFAULT.

However, if the row object is synthesized from relational data, you must choose some other set of values.

Oracle lets you specify object identifiers based on the primary key. The set of unique keys that identify the row object is turned into an identifier for the object. These values must be unique within the rows selected out of the view, because duplicates would lead to problems during navigation through object references.

The object view created with the WITH OBJECT IDENTIFIER clause has an object identifier derived from the primary key. If the WITH OBJECT IDENTIFIER DEFAULT clause is specified, the object identifier is either system generated or primary key based, depending on the underlying table or view definition.

For example, note the definition of the object type dept_t and the object view dept_view described in "Single-Level Collections in Object Views".

Because the underlying relational table has deptno as the primary key, each department row has a unique department number. In the view, the deptno column becomes the deptno attribute of the object type. Once we know that deptno is unique within the view objects, we can specify it as the object identifier.

See "Storage Considerations for Object Identifiers (OIDs)".

Creating References to View Objects

In the example we have been developing, each object selected out of the dept_view view has a unique object identifier derived from the department number value. In the relational case, the foreign key deptno in the emp employee table matches the deptno primary key value in the dept department table. We used the primary key value for creating the object identifier in the dept_view. This allows us to use the foreign key value in the emp_view in creating a reference to the primary key value in dept_view.

We accomplish this by using MAKE_REF operator to synthesize a primary key object reference. This takes the view or table name to which the reference points and a list of foreign key values to create the object identifier portion of the reference that will match with a particular object in the referenced view.

In order to create an emp_view view which has the employee's number, name, salary and a reference to the department in which she works, we need first to create the employee type emp_t and then the view based on that type as shown in Example 6-6.

CREATE TYPE emp_t AS OBJECT (
  empno    NUMBER,
  ename    VARCHAR2(20),
  salary   NUMBER,
  deptref  REF dept_t);
/
CREATE OR REPLACE VIEW emp_view OF emp_t WITH OBJECT IDENTIFIER(empno)
   AS SELECT e.empno, e.empname, e.salary, 
                      MAKE_REF(dept_view, e.deptno) 
         FROM emp e;

The deptref column in the view holds the department reference. The following simple query retrieves all employees whose department is located in the city of San Francisco:

SELECT e.empno, e.salary, e.deptref.deptno
  FROM emp_view e
 WHERE e.deptref.address.city = 'San Francisco';

Note that we could also have used the REF modifier to get the reference to the dept_view objects:

CREATE OR REPLACE VIEW emp_view OF emp_t WITH OBJECT IDENTIFIER(empno)
   AS SELECT e.empno, e.empname, e.salary, REF(d)
         FROM emp e, dept_view d
          WHERE e.deptno = d.deptno;

In this case we join the dept_view and the emp table on the deptno key. The advantage of using MAKE_REF operator instead of the REF modifier is that in using the former, we can create circular references. For example, we can create employee view to have a reference to the department in which she works, and the department view can have a list of references to the employees who work in that department.

Note that if the object view has a primary key based object identifier, the reference to such a view is primary key based. On the other hand, a reference to a view with system generated object identifier will be a system generated object reference. This difference is only relevant when you create object instances in the OCI object cache and need to get the reference to the newly created objects. This is explained in a later section.

As with synthesized objects, we can also select persistently stored references as view columns and use them seamlessly in queries. However, the object references to view objects cannot be stored persistently.

Modelling Inverse Relationships with Object Views

Views with objects can be used to model inverse relationships.

One-to-One Relationships

One-to-one relationships can be modeled with inverse object references. For example, let us say that each employee has a particular computer on her desk, and that the computer belongs to that employee only. A relational model would capture this using foreign keys either from the computer table to the employee table, or in the reverse direction. Using views, we can model the objects so that we have an object reference from the employee to the computer object and also have a reference from the computer object to the employee.

One-to-Many and Many-to-One Relationships

One-to-many relationships (or many-to-many relationships) can be modeled either by using object references or by embedding the objects. One-to-many relationship can be modeled by having a collection of objects or object references. The many-to-one side of the relationship can be modeled using object references.

Consider the department-employee case. In the underlying relational model, we have the foreign key in the employee table. Using collections in views, we can model the relationship between departments and employees. The department view can have a collection of employees, and the employee view can have a reference to the department (or inline the department values). This gives us both the forward relation (from employee to department) and the inverse relation (department to list of employees). The department view can also have a collection of references to employee objects instead of embedding the employee objects.

Updating Object Views

You can update, insert, and delete data in an object view using the same SQL DML you use for object tables. Oracle updates the base tables of the object view if there is no ambiguity.

A view is not directly updatable if its view query contains joins, set operators, aggregate functions, or GROUP BY or DISTINCT clauses. Also, individual columns of a view are not directly updatable if they are based on pseudocolumns or expression in the view query.

If a view is not directly updatable, you can still update it indirectly using INSTEAD OF triggers. To do so, you define an INSTEAD OF trigger for each kind of DML statement you want to execute on the view. In the INSTEAD OF trigger, you code the operations that must take place on the underlying tables of the view to accomplish the desired change in the view. Then, when you issue a DML statement for which you have defined an INSTEAD OF trigger, Oracle transparently runs the associated trigger. See "Using INSTEAD OF Triggers to Control Mutating and Validation" for an example of an INSTEAD OF trigger.

Something you want to be careful of: In an object view hierarchy, UPDATE and DELETE statements operate polymorphically just as SELECT statements do: the set of rows picked out by an UPDATE or DELETE statement on a view implicitly includes qualifying rows in any subviews of the specified view as well. See "Object View Hierarchies" for a discussion of object view hierarchy and examples defining Student_v and Employee_v.

For example, the following statement, which deletes all persons from Person_v, also deletes all students from Student_v and all employees from the Employee_v view.

DELETE FROM Person_v;

To exclude subviews and restrict the affected rows just to those in the view actually specified, use the ONLY keyword. For example, the following statement updates only persons and not employees or students.

UPDATE ONLY(Person_v) SET address = ...

CREATE TRIGGER emp_instr INSTEAD OF INSERT on emp_view 
FOR EACH ROW
DECLARE
  dept_var dept_t;
  emp_count integer;
BEGIN
  -- Enforce the constraint
  -- First get the department number from the reference
  UTL_REF.SELECT_OBJECT(:NEW.deptref, dept_var);

  SELECT COUNT(*) INTO emp_count
    FROM emp
   WHERE deptno = dept_var.deptno;
  IF emp_count < 9 THEN
     -- Do the insert
     INSERT INTO emp (empno, empname, salary, deptno) 
        VALUES (:NEW.empno, :NEW.ename, :NEW.salary, dept_var.deptno);
  END IF;
END;
/

Applying the Object Model to Remote Tables

Although you cannot directly access remote tables as object tables, object views let you access remote tables as if they were object tables.

Consider a company with two branches; one in Washington D.C. and another in Chicago. Each site has an employee table. The headquarters in Washington has a department table with the list of all the departments. To get a total view of the entire organization, we can create views over the individual remote tables and then a overall view of the organization.

In Example 6-8, we begin by creating an object view for each employee table. Then we can create the global view.

CREATE VIEW emp_washington_view (eno, ename, salary, job)
   AS SELECT e.empno, e.empname, e.salary, e.job
          FROM emp@washington e;

CREATE VIEW emp_chicago_view (eno, ename, salary, job)
   AS SELECT e.empno, e.empname, e.salary, e.job
          FROM emp@chicago e;

CREATE VIEW orgnzn_view OF dept_t WITH OBJECT IDENTIFIER (deptno) 
    AS SELECT d.deptno, d.deptname,
          address_t(d.deptstreet,d.deptcity,d.deptstate,d.deptzip) AS deptaddr,
          CAST( MULTISET (
                      SELECT e.eno, e.ename, e.salary, e.job
                      FROM emp_washington_view e) 
                   AS employee_list_t) AS emp_list
       FROM dept d
       WHERE d.deptcity = 'Washington'
   UNION ALL
       SELECT d.deptno, d.deptname,
           address_t(d.deptstreet,d.deptcity,d.deptstate,d.deptzip) AS deptaddr,
              CAST( MULTISET (
                       SELECT e.eno, e.ename, e.salary, e.job
                       FROM emp_chicago_view e) 
                    AS employee_list_t) AS emp_list
       FROM dept d
       WHERE d.deptcity = 'Chicago';

This view has the list of all employees for each department. We use UNION ALL because we cannot have two employees working in more than one department.

Defining Complex Relationships in Object Views

You can define circular references in object views using the MAKE_REF operator: view_A can refer to view_B which in turn can refer to view_A. This allows an object view to synthesize a complex structure such as a graph from relational data.

For example, in the case of the department and employee, the department object currently includes a list of employees. To conserve space, we may want to put references to the employee objects inside the department object, instead of materializing all the employees within the department object. We can construct (pin) the references to employee objects, and later follow the references using the dot notation to extract employee information.

Because the employee object already has a reference to the department in which the employee works, an object view over this model contains circular references between the department view and the employee view.

You can create circular references between object views in two different ways:

• Method 1: Re-create First View After Creating Second View

  1. Create view A without any reference to view B.

  2. Create view B, which includes a reference to view A.

  3. Replace view A with a new definition that includes the reference to view B.

• Method 2: Create First View Using FORCE Keyword

  1. Create view A with the reference to view B using the FORCE keyword.

  2. Create view B with reference to view A. When view A is used, it is validated and re-compiled.

Method 2 has fewer steps, but the FORCE keyword may hide errors in the view creation. You need to query the USER_ERRORS catalog view to see if there were any errors during the view creation. Use this method only if you are sure that there are no errors in the view creation statement.

Also, if errors prevent the views from being recompiled upon use, you must recompile them manually using the ALTER VIEW COMPILE command.

We will see the implementation for both the methods.

CREATE TABLE emp
(  empno    NUMBER PRIMARY KEY,
   empname  VARCHAR2(20),
   salary   NUMBER,
   deptno   NUMBER );

CREATE TYPE dept_t;
/

CREATE TYPE emp_t AS OBJECT
( eno NUMBER,
  ename VARCHAR2(20),
  salary  NUMBER,
  deptref REF dept_t );
/

CREATE TYPE employee_list_ref_t AS TABLE OF REF emp_t;
/

CREATE TABLE dept
(   deptno        NUMBER PRIMARY KEY,
    deptname      VARCHAR2(20),
    deptstreet    VARCHAR2(20),
    deptcity      VARCHAR2(10),
    deptstate     CHAR(2),
    deptzip       VARCHAR2(10) );

CREATE TYPE address_t AS OBJECT 
( street        VARCHAR2(20),
   city         VARCHAR2(10),
   state        CHAR(2),
   zip          VARCHAR2(10));
/

CREATE OR REPLACE TYPE dept_t AS OBJECT
( dno           NUMBER,
  dname         VARCHAR2(20),
  deptaddr      address_t,
  empreflist    employee_list_ref_t);
/

CREATE VIEW emp_view OF emp_t WITH OBJECT IDENTIFIER(eno)
   AS SELECT e.empno, e.empname, e.salary, NULL
         FROM emp e;

CREATE VIEW dept_view OF dept_t WITH OBJECT IDENTIFIER(dno) 
   AS SELECT d.deptno, d.deptname, 
                address_t(d.deptstreet,d.deptcity,d.deptstate,d.deptzip),
                CAST( MULTISET (
                           SELECT MAKE_REF(emp_view, e.empno)
                           FROM emp e 
                           WHERE e.deptno = d.deptno) 
                        AS employee_list_ref_t)
   FROM dept d; 

CREATE OR REPLACE VIEW emp_view OF emp_t WITH OBJECT IDENTIFIER(eno)
   AS SELECT e.empno, e.empname, e.salary, 
                       MAKE_REF(dept_view, e.deptno)
         FROM emp e;

CREATE OR REPLACE FORCE VIEW emp_view OF emp_t WITH OBJECT IDENTIFIER(eno)
   AS SELECT e.empno, e.empname, e.salary, 
                       MAKE_REF(dept_view, e.deptno)
         FROM emp e;

CREATE OR REPLACE VIEW dept_view OF dept_t WITH OBJECT IDENTIFIER(dno) 
   AS SELECT d.deptno, d.deptname, 
                address_t(d.deptstreet,d.deptcity,d.deptstate,d.deptzip),
                CAST( MULTISET (
                           SELECT MAKE_REF(emp_view, e.empno)
                           FROM emp e 
                           WHERE e.deptno = d.deptno) 
                        AS employee_list_ref_t)
   FROM   dept d; 

SELECT DEREF(e.COLUMN_VALUE)
  FROM TABLE( SELECT e.empreflist FROM dept_view e WHERE e.dno = 100) e;

SELECT e.COLUMN_VALUE.eno
  FROM TABLE(SELECT e.empreflist FROM dept_view e WHERE e.dno = 100) e
 WHERE e.COLUMN_VALUE.ename like 'John%';

SELECT d.dno, e.COLUMN_VALUE.eno, e.COLUMN_VALUE.ename
  FROM dept_view d, TABLE(d.empreflist) e
 WHERE e.COLUMN_VALUE.ename like 'John%' 
  AND d.dno = 100;

SELECT e.deptref.dno, DEREF(f.COLUMN_VALUE)
  FROM emp_view e, TABLE(e.deptref.empreflist) f
 WHERE e.deptref.dno = 100 
  AND f.COLUMN_VALUE.ename like 'John%';

Object View Hierarchies

An object view hierarchy is a set of object views each of which is based on a different type in a type hierarchy. Subviews in a view hierarchy are created under a superview, analogously to the way subtypes in a type hierarchy are created under a supertype.

Each object view in a view hierarchy is populated with objects of a single type, but queries on a given view implicitly address its subviews as well. Thus an object view hierarchy gives you a simple way to frame queries that can return a polymorphic set of objects of a given level of specialization or greater.

For example, suppose you have the following type hierarchy, with person_typ as the root:

Figure 6-1 Object View Hierarchy

Description of "Figure 6-1 Object View Hierarchy"

If you have created an object view hierarchy based on this type hierarchy, with an object view built on each type, you can query the object view that corresponds to the level of specialization you are interested in. For instance, you can query the view of student_typ to get a result set that contains only students, including part-time students.

You can base the root view of an object view hierarchy on any type in a type hierarchy: you do not need to start the object view hierarchy at the root type. Nor do you need to extend an object view hierarchy to every leaf of a type hierarchy or cover every branch. However, you cannot skip intervening subtypes in the line of descent. Any subview must be based on a direct subtype of the type of its direct superview.

Just as a type can have multiple sibling subtypes, an object view can have multiple sibling subviews. But a subview based on a given type can participate in only one object view hierarchy: two different object view hierarchies cannot each have a subview based on the same subtype.

A subview inherits the object identifier (OID) from its superview. An OID cannot be explicitly specified in any subview.

A root view can explicitly specify an object identifier using the WITH OBJECT ID clause. If the OID is system-generated or the clause is not specified in the root view, then subviews can be created only if the root view is based on a table or view that also uses a system generated OID.

The query underlying a view determines whether the view is updatable. For a view to be updatable, its query must contain no joins, set operators, aggregate functions, GROUP BY, DISTINCT, pseudocolumns, or expressions. The same applies to subviews.

If a view is not updatable, you can define INSTEAD OF triggers to perform appropriate DML actions. Note that INSTEAD OF triggers are not inherited by subviews.

All views in a view hierarchy must be in the same schema.

Creating an Object View Hierarchy

You build an object view hierarchy by creating subviews under a root view. You do this by using the UNDER keyword in the CREATE VIEW statement, as show in Example 6-9.

The same object view hierarchy can be based on different underlying storage models. In other words, a variety of layouts or designs of underlying tables can produce the same object view hierarchy. The design of the underlying storage model has implications for the performance and updatability of the object view hierarchy.

The following examples show three possible storage models. In the first, a flat model, all views in the object view hierarchy are based on the same table. In the second, a horizontal model, each view has a one-to-one correspondence with a different table. And in the third, a vertical model, the views are constructed using joins.

The Flat Model

In the flat model, all the views in the hierarchy are based on the same table. In the following example, the single table AllPersons contains columns for all the attributes of person_typ, student_typ, or employee_typ.

Figure 6-2 Flat Storage Model for Object View Hierarchy

Description of "Figure 6-2 Flat Storage Model for Object View Hierarchy"

CREATE TABLE AllPersons
( typeid NUMBER(1), 
  ssn NUMBER, 
  name VARCHAR2(30), 
  address VARCHAR2(100),
  deptid NUMBER,
  major VARCHAR2(30),
  empid NUMBER, 
  mgr VARCHAR2(30));
  

The typeid column identifies the type of each row. Possible values are:

-- 1 = person_typ
-- 2 = student_typ
-- 3 = employee_typ

CREATE TYPE person_typ AS OBJECT
( ssn NUMBER,
  name VARCHAR2(30),
  address VARCHAR2(100)) NOT FINAL;/

CREATE TYPE student_typ UNDER person_typ 
( deptid NUMBER,
   major VARCHAR2(30)) NOT FINAL;/

CREATE TYPE employee_typ UNDER person_typ
( empid NUMBER, 
  mgr VARCHAR2(30));/

The following statements create the views that make up the object view hierarchy:

Example 6-9 Creating an Object View Hierarchy

CREATE VIEW Person_v OF person_typ
  WITH OBJECT OID(ssn) AS
  SELECT ssn, name, address 
  FROM AllPersons 
  WHERE typeid = 1;

CREATE VIEW Student_v OF student_typ UNDER Person_v
  AS 
  SELECT ssn, name, address, deptid, major
  FROM AllPersons
  WHERE typeid = 2;

CREATE VIEW Employee_v OF employee_typ UNDER Person_v
  AS
  SELECT ssn, name, address, empid, mgr
  FROM AllPersons
  WHERE typeid = 3;

The flat model has the advantage of simplicity and poses no obstacles to supporting indexes and constraints. Its drawbacks are:

• A single table cannot contain more than 1000 columns, so the flat model imposes a 1000-column limit on the total number of columns that the object view hierarchy can contain.

• Each row of the table will have NULLs for all the attributes not belonging to its type. Such non-trailing NULLs can adversely affect performance.

The Horizontal Model

On the horizontal model, each view or subview is based on a different table. In the example, the tables are relational, but they could just as well be object tables for which column substitutability is turned off.

Figure 6-3 Horizontal Storage Model for Object View Hierarchy

Description of "Figure 6-3 Horizontal Storage Model for Object View Hierarchy"

CREATE TABLE only_persons
( ssn NUMBER,
  name VARCHAR2(30),
  address VARCHAR2(100));

CREATE TABLE only_students
( ssn NUMBER, 
  name VARCHAR2(30),
  address VARCHAR2(100), 
  deptid NUMBER,  
  major VARCHAR2(30));

CREATE TABLE only_employees
( ssn NUMBER, 
  name VARCHAR2(30),
  address VARCHAR2(100), 
  empid NUMBER, 
  mgr VARCHAR2(30));

These are the views:

CREATE OR REPLACE VIEW Person_v OF person_typ
  WITH OBJECT OID(ssn) AS
  SELECT * 
  FROM only_persons 

CREATE OR REPLACE VIEW Student_v OF student_typ UNDER Person_v
  AS 
  SELECT *
  FROM only_students;

CREATE OR REPlACE VIEW Employee_v OF employee_typ UNDER Person_v
  AS
  SELECT * 
  FROM only_employees;

The horizontal model is very efficient at processing queries of the form:

SELECT VALUE(p) FROM Person_v p
  WHERE VALUE(p) IS OF (ONLY student_typ);

Such queries need access only a single physical table to get all the objects of the specific type. The drawbacks of this model are that queries of the sort SELECT * FROM view require performing a UNION over all the underlying tables and projecting the rows over just the columns in the specified view. (See "Querying a View in a Hierarchy".) Also, indexes on attributes (and unique constraints) must span multiple tables, and support for this does not currently exist.

The Vertical Model

In the vertical model, there is a physical table corresponding to each view in the hierarchy, but each physical table stores only those attributes that are unique to its corresponding subtype.

Figure 6-4 Vertical Storage Model for Object View Hierarchy

Description of "Figure 6-4 Vertical Storage Model for Object View Hierarchy"

CREATE TABLE all_personattrs 
( typeid NUMBER,
  ssn NUMBER,  
  name VARCHAR2(30),
  address VARCHAR2(100));

CREATE TABLE all_studentattrs
( ssn NUMBER, 
  deptid NUMBER, 
  major VARCHAR2(30));

CREATE TABLE all_employeeattrs
( ssn NUMBER,
  empid NUMBER, 
  mgr VARCHAR2(30));

CREATE OR REPLACE VIEW Person_v OF person_typ 
WITH OBJECT OID(ssn) AS
  SELECT ssn, name, address 
  FROM all_personattrs 
  WHERE typeid = 1;

CREATE OR REPLACE VIEW Student_v OF student_typ UNDER Person_v
  AS 
  SELECT x.ssn, x.name, x.address, y.deptid, y.major
  FROM all_personattrs x, all_studentattrs y
  WHERE x.typeid = 2 AND x.ssn = y.ssn;

CREATE OR REPLACE VIEW Employee_v OF employee_typ UNDER Person_v
  AS
  SELECT x.ssn, x.name, x.address, y.empid, y.mgr
  FROM all_personattrs x, all_employeeattrs y
  WHERE x.typeid = 3 AND x.ssn = y.ssn;

The vertical model can efficiently process queries of the kind SELECT * FROM root_view, and it is possible to index individual attributes and impose unique constraints on them. However, to re-create an instance of a type, a join over OIDs must be performed for each level that the type is removed from the root in the hierarchy.

SELECT REF(p), VALUE(p) FROM Person_v p;

SELECT * FROM Person_v;

SELECT VALUE(p) FROM ONLY(Person_v) p;

SELECT * FROM Person_v;

SELECT VALUE(p) FROM Person_v p;

GRANT SELECT ON Person_v TO oe WITH HIERARCHY OPTION;

SELECT * FROM ONLY(Person_v);