Object Oriented

1
Object Oriented Object Relational Databases

• Ranga Raju Vatsavai
• Teaching Mentor (Prof. Shekhar)
• CSci 5708 Architecture and Implementation of
  Database Management Systems, Fall 2003
• Week 15 (11/24, 11/26/03) Lecture Notes

2
Last Class (11/24/03)

• RDBMS limitations
• No support for complex data types and predicates
• OO
• Rich set of features encapsulation,
  inheritance, ..
• Helps manage complexity
• Conceptual Modeling Simple Extensions
• EER, UML, PEER

3
Learning Objectives

• Basic concepts of OO and OR models
• Extensions at conceptual modeling
• Mapping Conceptual Model into Logical Model
• Exposure to additional features in SQL1999
  standard.
• Ability to model and implement a broader class of
  applications (spatial, multimedia, engineering,
  biological, scientific, )

4
Mapping Conceptual Model onto Logical Model
5
Mapping Conceptual Model onto Logical Model

• OO Object Definition Language
• OR SQL3 DDL
• ODL is an extension of Interface Description
  Language.
• ODL class definition includes
• Attributes
• Relationships
• Methods

6
Mapping - ODL

• Example
• interface Student
• attribute string status
• attribute Department major
• relationship setltDepartmentgt majorOf inverse
  Departmentstudents
• .

7
ORDBMS Fundamentals

• Try to unify aspects of both relational and
  object databases
• Relation is still central
• No standard of what constitutes an ORDBMS
• Won Kims (UniSQL) white paper
• Michael Stonebraker ORDBMS, The Next Great Wave
• Query language SQL3

8
ORDBMS Fundamentals (SQL3)

• OO Features in SQL3
• Objects
• Type constructors
• Collection types
• User-defined functions and procedures
• Support for large objects
• Inheritance

9
ORDBMS Fundamentals (SQL3)

• Objects in SQL3 comes in two flavors
• ADTs and Row objects
• ADTs user defined arbitrary data types is a key
  feature of ORDBMS
• ADTs are a combination of atomic data types and
  associated methods
• DBMS doesnt need to know about how ADTs are
  stored or their methods work? It just has to know
  about ADTs signature
• Hiding ADT internals is called encapsulation

10
ORDBMS Fundamentals (SQL3)

• Here is a general form of ADT specification
• CREATE TYPE lttype-namegt (
• list of component attributes with individual
  types
• Optional declaration of and lt functions for the
  type
• declaration of other functions (methods) for the
  type
• )
• Example
• CREATE TYPE DepartmentADT (
• Code int,
• Name char(10),
• EQUALS deptEQ, //deptEQ we will define it
  later
• LESS THAN NONE //DEFAULT
• Defintion of other functions goes here
• )

11
ORDBMS Fundamentals (SQL3)

• Defining ADTs methods
• FUNCTION ltnamegt ( ltargumentsgt ) RETURNS lttypegt
• Functions are of two types internal and
  external
• External functions are written in host language
  and only signature appears in ADT definition.
• Internal functions are written in extended SQL
• assignment
• local variables can be declared inside function
  (a DepartmentADT)
• dot op is used to access components of structure
• BEGIN and END are used to collect several stmts.

12
ORDBMS Fundamentals (SQL3)

• Here is an example constructor method
• FUNCTION DepartmentADT(id INT, dname CHAR(10))
• RETURNS DepartmentADT
• d DepartmentADT
• BEGIN
• d DepartmentADT()
• d.code id
• d.name dname
• RETURN d
• END
• Discussion question define method deptEQ

13
ORDBMS Fundamentals (SQL3)

• Function deptEQ(d1 DepartmentADT, d2
  DepartmentADT)
• RETURNS BOOLEAN
• RETURN (d1.code d2.code
• AND d1.name d1.name)
• We could have used DEFAULT (system defined)

14
ORDBMS Fundamentals (SQL3)

• External functions
• ADTs can have methods that are written in host
  language (e.g. C, C, )
• Only signature appears in ADT definition
• DECLARE EXTERNAL ltfunctionNamegt ltsignaturegt
• LANGUAGE ltlanguage namegt
• Example
• DECLARE EXTERNAL Square
• POLYGON
• RETURNS BOOLEAN
• LANGUAGE C

15
ORDBMS Fundamentals (SQL3)

• Row type objects
• Essentially tuples and they roughly resembles
  struct/class
• CREATE ROW TYPE lttypenamegt (ltlistOfAttributes-and-
  their-typesgt)
• CREATE ROW TYPE DepartmentType(
• code INT,
• name CHAR(10)
• )

16
ORDBMS Fundamentals (SQL3)

• Creating Relations of Row Type
• OF TYPE ltrow-type-namegt
• Example
• CREATE TABLE Department OF TYPE DepartmentType
• References A component attribute of tuple may
  be a reference to a tuple of another (or posibly
  same) relation.
• CREATE ROW TYPE Department (
• code INT,
• name CHAR(10),
• chair REF (Faculty_Row_Type)
• )

17
ORDBMS Fundamentals (SQL3)

• CREATE ROW TYPE Department_Row_Type (
• code INT,
• name CHAR(10),
• chair REF (Faculty_Row_Type)
• )
• CREATE TABLE Department OF TYPE
  Department_Row_Type
• Q? Print chair name of EECS department
• SELECT d.chair-gtname
• FROM Department d
• WHERE d..name EECS

18
ORDBMS Fundamentals (SQL3)

• Collection types
• setof
• Example
• CREATE TABLE Rectangles (rname CHAR(10), pnts
  setof(Points)).
• SELECT r.rname
• FROM Rectangles r
• WHERE count(r.pnts) 5

19
ORDBMS Fundamentals (SQL3)

• Support for large objects
• blob, clob
• Consider class recording
• Class_video_recordings(cid integer, cmdate
  date, loc char(10), video BLOB)

20
ORDBMS Fundamentals (SQL3)

• Inheritance
• Used in two ways
• reusing and refining types, and
• for creating hierarchies of collections of
  similar but not identical objects
• UNDER
• CREATE TYPE Student UNDER Person (addr address)
• Creates an explicit relationship between subtype
  Student and supertype Person.
• An object of subtype is considered to be an
  object of supertype.

21
ORDBMS Fundamentals (SQL3)

• OO Features in SQL3
• Objects
• Type constructors
• Collection types
• User-defined functions and procedures
• Support for large objects
• Inheritance

22
Outline for todays class 11/26/03

• Objectives
• Mapping Conceptual Model into Logical Model
• ORDBMS Fundamentals
• How ORDBMS incorporates OO ideas
• SQL3
• Physical Design and efficiency Issues
• Demo
• Conclusions

23
Physical design and efficiency issues

• Need
• efficiently storage of ADTs and structured
  objects
• efficient indexed access
• Main problem is size, so disk layout is important
• BLOBS require special storage, typically
  different location on disk (separate from the
  tuple)
• Indexing
• Domain specific (large collection of indices)
• Extendible indices

24
Physical design and efficiency issues

• Query processing
• User defined aggregate functions (SQL defined may
  not be useful)
• ORDBMS allows registering new aggregate functions
  
• Security external methods of ADTs can
  compromise the database or crash (if its buggy).
• One solution interpreted rather than compiled
  (Java and procedural portions of SQL1999).
• Run in different address space than the DBMS
• Method Caching
• Cache of input and output of methods
• Pointer Swizzling
• Technique to reduce cost of cached object

25
Comparison of OO and ORDBMS

• Fundamental difference is in philosophy
• OODMSs try to add DBMS functionality to Prog.
  Lang.
• ORDBMS try to add richer data types and
  predicates to an RDBMS
• OODBMS aimed at applications where
  object-centric viewpoint is appropriate
• ORDBMS aimed at applications where large data
  collections are the focus
• Query (OQL) is not efficiently supported in
  OODBMS
• Query processing is the centerpiece of an ORDBMS

26
Outline for todays class 11/26/03

• Objectives
• Mapping Conceptual Model into Logical Model
• ORDBMS Fundamentals
• How ORDBMS incorporates OO ideas
• SQL3
• Physical design and efficiency issues
• Demo
• Summary and Conclusions

27
Demo

• Postgresql/PostGIS

28
Summary and Conclusions

• Basic concepts of OO and OR models
• Extensions at conceptual modeling
• Mapping Conceptual Model into Logical Model
• Exposure to additional features in SQL1999
  standard.
• Ability to model and implement a broader class of
  applications (spatial, multimedia, engineering,
  biological, scientific, )
• ORDBMS/SQL3 offer much promise

29
Additional Readings

• http//www.cs.umn.edu/vatsavai/oo-ordbms.ppt
• Main Database Management Systems by Raghu
  Ramakrishnan (Chapter 24 Object Database
  Systems).
• Additional References (relevant chapters)
• Fundamentals of Database Systems Elmasri
  Navathe
• A First Course In Database Systems Ullman
  Widom.
• http//www-users.cs.umn.edu/shekhar/5705/
• Spatial Database A Tour (Chapter 2 and 3)

30
Acknowledgements

• PFF Faculty
• Prof. Shekhar
• Class
• Happy Thanks Giving and good luck on final exams.