This chapter provides an example of starting with a schema for a data cartridge. Object types are crucial to building data cartridges in that they enable domain-level abstractions to be captured in the database.
This chapter contains these topics:
See Also:The following manuals for additional information about creating and using object types:
In the Oracle Object-Relational Database Management System (ORDBMS), you use object types to model real-world entities. An object type has attributes, which reflect the entity's structure, and methods, which implement the operations on the entity. Attributes are defined using built-in types or other object types. Methods are functions or procedures written in PL/SQL or an external language, like C, and stored in the database.
A typical use for an object type is to impose structure on some part of the data in the database. For example, an object type named
DataStream could be used by a cartridge to store large amounts of data in a character LOB (a data type for large objects). This object type has attributes such as an identifier, a name, a date, and so on. The statement in Example 3-1 defines the
create or replace type DataStream as object ( id integer, name varchar2(20), createdOn date, data clob, MEMBER FUNCTION DataStreamMin return pls_integer, MEMBER FUNCTION DataStreamMax return pls_integer, MAP MEMBER FUNCTION DataStreamToInt return integer, PRAGMA restrict_references(DataStreamMin, WNDS, WNPS), PRAGMA restrict_references(DataStreamMax, WNDS, WNPS));
A method is a procedure or function that is part of the object type definition and that can operate on the object type data attributes. Such methods are called member methods, and they take the keyword
MEMBER when you specify them as a component of the object type. The
DataStream type definition declares three methods. The first two,
DataStreamMax, calculate the minimum and maximum values, respectively, in the data stream stored inside the character
The third method,
DataStreamToInt, a map method, governs comparisons between instances of data stream type.
See Also:"Object Comparison" for information about map methods
The pragma (compiler directive)
RESTRICT_REFERENCES is necessary for security, and is discussed in the following sections.
After declaring the type, define the type body. The body contains the code for type methods. Example 3-2 shows the type body definition for the
DataStream type. It defines the member function methods,
DataStreamMax, and the map method
CREATE OR REPLACE TYPE BODY DataStream IS MEMBER FUNCTION DataStreamMin return pls_integer is a pls_integer := DS_Package.ds_findmin(data); begin return a; end; MEMBER FUNCTION DataStreamMax return pls_integer is b pls_integer := DS_Package.ds_findmax(data); begin return b; end; MAP MEMBER FUNCTION DataStreamToInt return integer is c integer := id; begin return c; end; end;
DataStreamMax are call routines in a PL/SQL package named
DS_Package. Since these methods are likely to be compute-intensive (they process numbers stored in the
CLOB to determine minimum and maximum values), they are defined as external procedures and implemented in C. The external dispatch is routed through a PL/SQL package named
DS_Package. Such packages are discussed in Oracle Database PL/SQL Packages and Types Reference.
The third method,
DataStreamToInt, is implemented in PL/SQL. Because we have a identifier,
id, attribute in
DataStream, this method can return the value of the identifier attribute. Most map methods, however, are more complex than
Oracle Database SecureFiles and Large Objects Developer's Guide for general information about LOBs
TYPE statement has an optional keyword
OID, which associates a user-specified object identifier (
OID) with the type definition. It should be used by anyone who creates an object type that will be used in more than one database.
Each type has an
OID. If you create an object type and do not specify an
OID, Oracle generates an
OID and assigns it to the type. Oracle uses the
OID internally for operations pertaining to that type. Using the same
OID for a type is important if you plan to share instances of the type across databases for such operations as export/import and distributed queries.
OIDis assigned to the type itself. Each row in a table with a column of the specified type will have a row-specific
Suppose that you want to create a
SpecialPerson type, and then instantiate this type in two different databases in tables named
SpecialPersonTable2. The RDBMS needs to know that the
SpecialPerson type is the same type in both instances, and therefore the type must be defined using the same
OID in both databases. If you do not specify an
TYPE, a unique identifier is created automatically by the RDBMS. The syntax for specifying an
OID for an object type is in Example 3-3.
CREATE OR REPLACE TYPE type_name OID 'oid' AS OBJECT (attribute datatype [,...]);
In Example 3-4, the
SELECT statement generates an
OID, and the
TYPE statement uses the
OID in creating an object type named
mytype. Be sure to use the
SELECT statement to generate a different
OID for each object type to be created, because this is the only way to guarantee that each
OID is valid and globally unique.
Oracle implicitly defines a constructor method for each object type that you define. The name of the constructor method is the same as the name of the object type. The parameters of the constructor method are exactly the data attributes of the object type, and they occur in the same order as the attribute definition for the object type. Only one constructor method can be defined for each object type.
In Example 3-5, the system creates a type named
rational_type and implicitly creates a constructor method for this object type.
CREATE TYPE rational_type ( numerator integer, denominator integer);
When you instantiate an object of
rational_type, you invoke the constructor method, as demonstrated in Example 3-6:
SQL performs comparison operations on objects. Some comparisons are explicit, using the comparison operators (=, <, >, <>, <=, >=, !=) and the
IN predicates. Other comparisons are implicit, as in the
Comparison of objects uses special member functions of the object type: map methods and order methods. To perform object comparison, you must implement either a map method or an order method in the
BODY statements. In Example 3-7, the type body for the
DataStream type implements the map member function:
MAP MEMBER FUNCTION DataStreamToInt return integer is c integer := id; begin return c; end;
This definition of the map member function relies on the presence of the
id attribute of the
DataStream type to map instances to integers. Whenever a comparison operation is required between objects of type
DataStream, the map function
DataStreamToInt() is called implicitly by the system.
The object type
rational_type does not have a simple
id attribute like
DataStream. Instead, its map member function is complicated, as demonstrated in Example 3-8. Because a map function can return any of the built-in types,
rational_type can return a value or type
MAP MEMBER FUNCTION RationalToReal RETURN REAL IS BEGIN RETURN numerator/denominator; END; ...
If you have not defined a map or order function for an object type, it can only support equality comparisons. Oracle SQL performs the comparison by doing a field-by-field comparison of the attributes of that type.