|Oracle® Database Application Developer's Guide - Object-Relational Features
10g Release 1 (10.1)
Part Number B10799-01
The chapter contains these topics:
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.
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.
Define an object type, where each attribute of the type corresponds to an existing column in a relational table.
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.
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:
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
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.
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
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
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.
See Also:See Oracle Call Interface Programmer's Guide for more information about OCI calls.
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
CREATE TABLE dept ( deptno NUMBER PRIMARY KEY, deptname VARCHAR2(20), deptstreet VARCHAR2(20), deptcity VARCHAR2(10), deptstate CHAR(2), deptzip VARCHAR2(10));
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.
Create the type for the address object:
CREATE TYPE address_t AS OBJECT ( street VARCHAR2(20), city VARCHAR2(10), state CHAR(2), zip VARCHAR2(10)); /
Create the type for the department object:
CREATE TYPE dept_t AS OBJECT ( deptno NUMBER, deptname VARCHAR2(20), address address_t ); /
Create the view containing the department number, name and address. The
address objects are constructed from columns of the relational table.
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;
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 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
OF trigger over the view to handle updates to this column.
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.
Taking the previous example as our starting point, we represent each employee in an
emp relational table that has the following structure:
CREATE TABLE emp ( empno NUMBER PRIMARY KEY, empname VARCHAR2(20), salary NUMBER, job VARCHAR2 (20), deptno NUMBER REFERENCES dept(deptno));
Using this relational table, we can construct a
dept_view with the department number, name, address and a collection of employees belonging to the department.
Define a nested table type for the employee type
CREATE TYPE employee_list_t AS TABLE OF employee_t; /
Define a department type having a department number, name, address, and a nested table of employees:
CREATE TYPE dept_t AS OBJECT ( deptno NUMBER, deptname VARCHAR2(20), address address_t, emp_list employee_list_t); /
Define the object view
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;
SELECT subquery inside the
CAST-MULTISET block selects the list of employees that belong to the current department. The
MULTISET keyword indicates that this is a list as opposed to a singleton value. The
CAST operator casts the result set into the appropriate type, in this case to the
employee_list_t nested table type.
A query on this view could give us the list of departments, with each department row containing the department number, name, the address object and a collection of employees belonging to the department.
Multilevel collections and single-level collections are created and used in object views in the same way. The only difference is that, for a multilevel collection, you must create an additional level of collections.
The following example builds an object view containing a multilevel collection. The view is based on flat relational tables (that contain no collections). As a preliminary to building the object view, the example creates the object and collection types it uses. An object type (for example,
emp_t) is defined to correspond to each relational table, with attributes whose types correspond to the types of the respective table columns. In addition, the employee type has a nested table (attribute) of projects, and the department type has a nested table (attribute) of employees. The latter nested table is a multilevel collection. The
CAST-MULTISET operator is used in the
VIEW statement to build the collections.
These are the underlying relational tables:
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));
These are the object and collection types the view will use:
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 ); /
The following statement creates the object view:
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;
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
IDENTIFIER clause, or specify
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
IDENTIFIER clause has an object identifier derived from the primary key. If the
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.
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
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
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;
deptref column in the view holds the department reference. We write the following simple query to determine all employees whose department is located in the city of San Francisco:
SELECT e.eno, 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.
Views with objects can be used to model inverse 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 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.
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
OF trigger for each kind of DML statement you want to execute on the view. In the
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
OF trigger, Oracle transparently runs the associated trigger.
See Also:"Using INSTEAD OF Triggers to Control Mutating and Validation" for an example of an
Something you want to be careful of: In an object view hierarchy,
DELETE statements operate polymorphically just as
SELECT statements do: the set of rows picked out by an
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
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 = ...
A nested table can be modified by inserting new elements and updating or deleting existing elements. Nested table columns that are virtual or synthesized, as in a view, are not usually updatable. To overcome this, Oracle allows
OF triggers to be created on these columns.
OF trigger defined on a nested table column (of a view) is fired when the column is modified. Note that if the entire collection is replaced (by an update of the parent row), the
OF trigger on the nested table column is not fired.
OF triggers provide a way of updating complex views that otherwise could not be updated. They can also be used to enforce constraints, check privileges and validate the DML. Using these triggers, you can control mutation of the objects created though an object view that might be caused by inserting, updating and deleting.
For instance, suppose we wanted to enforce the condition that the number of employees in a department cannot exceed 10. To enforce this, we can write an
OF trigger for the employee view. The trigger is not needed for doing the DML because the view can be updated, but we need it to enforce the constraint.
We implement the trigger by means of the following code:
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;
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.
First, we create an object view for each employee table:
CREATE VIEW emp_washington_view (eno,ename,salary) AS SELECT e.empno, e.empname, e.salary FROM emp@washington_link e; CREATE VIEW emp_chicago_view AS SELECT e.eno, e.name, e.salary FROM emp_tab@chicago_link e;
We can now create the global view:
CREATE VIEW orgnzn_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 e.eno, e.ename, e.salary FROM emp_washington_view e) AS employee_list_t) FROM dept d WHERE d.deptcity = 'Washington' UNION ALL SELECT d.deptno, d.deptname, address_t(d.deptstreet,d.deptcity,d.deptstate,d.deptzip), CAST( MULTISET ( SELECT e.eno, e.name, e.salary FROM emp_chicago_view e) AS employee_list_t) FROM dept d WHERE d.deptcity = 'Chicago';
This view has the list of all employees for each department. We use
ALL because we cannot have two employees working in more than one department.
You can define circular references in object views using the
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:
Create view A without any reference to view B.
Create view B, which includes a reference to view A.
Replace view A with a new definition that includes the reference to view B.
Create view A with the reference to view B using the
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
We will see the implementation for both the methods.
First, we set up some relational tables and associated object types. Although the tables contain some objects, they are not object tables. To access the data objects, we will create object views later.
emp table stores the employee information:
CREATE TABLE emp ( empno NUMBER PRIMARY KEY, empname VARCHAR2(20), salary NUMBER, deptno NUMBER );
The emp_t type contains a reference to the department. We need a dummy department type so that the emp_t type creation succeeds.
CREATE TYPE dept_t; /
The employee type includes a reference to the department:
CREATE TYPE emp_t AS OBJECT ( eno NUMBER, ename VARCHAR2(20), salary NUMBER, deptref REF dept_t ); /
We represent the list of references to employees as a nested table:
CREATE TYPE employee_list_ref_t AS TABLE OF REF emp_t; /
The department table is a typical relational table:
CREATE TABLE dept ( deptno NUMBER PRIMARY KEY, deptname VARCHAR2(20), deptstreet VARCHAR2(20), deptcity VARCHAR2(10), deptstate CHAR(2), deptzip VARCHAR2(10) );
To create object views, we need object types that map to columns from the relational tables:
CREATE TYPE address_t AS OBJECT ( street VARCHAR2(20), city VARCHAR2(10), state CHAR(2), zip VARCHAR2(10)); /
We earlier created an incomplete type; now we fill in its definition:
CREATE OR REPLACE TYPE dept_t AS OBJECT ( dno NUMBER, dname VARCHAR2(20), deptaddr address_t, empreflist employee_list_ref_t); /
Now that we have the underlying relational table definitions, we create the object views on top of them.
We first create the employee view with a null in the
deptref column. Later, we will turn that column into a reference.
CREATE VIEW emp_view OF emp_t WITH OBJECT IDENTIFIER(eno) AS SELECT e.empno, e.empname, e.salary, NULL FROM emp e;
Next, we create the department view, which includes references to the employee objects.
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;
We create a list of references to employee objects, instead of including the entire employee object. We now re-create the employee view with the reference to the department view.
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;
This creates the views.
First, we create an employee view that includes a reference to the department view, which does not exist at this point. This view cannot be queried until the department view is created properly.
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;
Next, we create a department view that includes references to the employee objects. We do not have to use the
FORCE keyword here, because
emp_view already exists.
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;
This allows us to query the department view, getting the employee object by de-referencing the employee reference from the nested table
SELECT DEREF(e.COLUMN_VALUE) FROM TABLE( SELECT e.empreflist FROM dept_view e WHERE e.dno = 100) e;
COLUMN_VALUE is a special name that represents the scalar value in a scalar nested table. In this case,
COLUMN_VALUE denotes the reference to the employee objects in the nested table
We can also access only the employee number of all those employees whose name begins with
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%';
To get a tabular output, unnest the list of references by joining the department table with the items in its nested table:
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%';
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 5-1 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,
DISTINCT, pseudocolumns, or expressions. The same applies to subviews.
If a view is not updatable, you can define
OF triggers to perform appropriate DML actions. Note that
OF triggers are not inherited by subviews.
All views in a view hierarchy must be in the same schema.
Note:You can create views of types that are non-instantiable. A non-instantiable type cannot have instances, so ordinarily there would be no point in creating an object view of such a type. However, a non-instantiable type can have subtypes that are instantiable. The ability to create object views of non-instantiable types enables you to base an object view hierarchy on a type hierarchy that contains a non-instantiable type.
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 Student_v OF student_typ UNDER Person_v AS SELECT ssn, name, address, deptid, major FROM AllPersons WHERE typeid = 2;
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.
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
Figure 5-2 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));
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:
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.
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 5-3 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.
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 5-4 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 contraints 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.
You can query any view or subview in an object view hierarchy; rows are returned for the declared type of the view that you query and for any of that type's subtypes. So, for instance, in an object view hierarchy based on the
person_typ type hierarchy, you can query the view of
person_typ to get a result set that contains all persons, including students and employees; or you can query the view of
student_typ to get a result set that contains only students, including part-time students.
SELECT list of a query, you can include either functions such as
VALUE() that return an object instance, or you can specify object attributes of the view's declared type, such as the
ssn attributes of
If you specify functions, to return object instances, the query returns a polymorphic result set: that is, it returns instances of both the view's declared type and any subtypes of that type.
For example, the following query returns instances of persons, employees, and students of all types, as well as
REFs to those instances.
SELECT REF(p), VALUE(p) FROM Person_v p;
If you specify individual attributes of the view's declared type in the
SELECT list or do a
*, again the query returns rows for the view's declared type and any subtypes of that type, but these rows are projected over columns for the attributes of the view's declared type, and only those columns are used. In other words, the subtypes are represented only with respect to the attributes they inherit from and share with the view's declared type.
So, for example, the following query returns rows for all persons and rows for employees and students of all types, but the result uses only the columns for the attributes of
address. It does not show rows for attributes added in the subtypes, such as the
deptid attribute of
SELECT * FROM Person_v;
To exclude subviews from the result, use the
ONLY keyword. The
ONLY keyword confines the selection to the declared type of the view that you are querying:
SELECT VALUE(p) FROM ONLY(Person_v) p;
Generally, a query on a view with subviews requires only the
SELECT privilege on the view being referenced and does not require any explicit privileges on subviews. For example, the following query requires only
SELECT privileges on
Person_v but not on any of its subviews.
SELECT * FROM Person_v;
However, a query that selects for any attributes added in subtypes but not used by the root type requires the
SELECT privilege on all subviews as well. Such subtype attributes may hold sensitive information that should reasonably require additional privileges to access.
The following query, for example, requires
SELECT privileges on
Person_v and also on
Employee_v (and on any other subview of
Person_v) because the query selects object instances and thus gets all the attributes of the subtypes.
SELECT VALUE(p) FROM Person_v p;
To simplify the process of granting
SELECT privileges on an entire view hierarchy, you can use the
HIERARCHY option. Specifying the
HIERARCHY option when granting a user
SELECT privileges on a view implicitly grants
SELECT privileges on all current and future subviews of the view as well. For example:
GRANT SELECT ON Person_v TO scott WITH HIERARCHY OPTION;
A query that excludes rows belonging to subviews also requires
SELECT privileges on all subviews. The reason is that information about which rows belong exclusively to the most specific type of an instance may be sensitive, so the system requires
SELECT privileges on subviews for queries (such as the following one) that exclude all rows from subviews.
SELECT * FROM ONLY(Person_v);