1. Lecture #2 October 5 th, 2000 Conceptual Modeling Administration: –HW1 available –Details on projects –Exam date –XML comment

2. Building an Application with a Database System Requirements modeling (conceptual, pictures) –Decide what entities should be part of the application and how they should be linked. Schema design and implementation –Decide on a set of tables, attributes. –Define the tables in the database system. –Populate database (insert tuples). Write application programs using the DBMS –way easier now that the data management is taken care of.

3. Database Design Why do we need it? – Agree on structure of the database before deciding on a particular implementation. Consider issues such as: –What entities to model –How entities are related –What constraints exist in the domain –How to achieve good designs

4. Database Design Formalisms Object Definition Language (ODL): – Closer in spirit to object-oriented models Entity/Relationship model (E/R): –More relational in nature. Both can be translated (semi-automatically) to relational schemas (with varying amount of pain). ODL to OO-schema: direct transformation (C++ or Smalltalk based system).

5. Outline ODL (rather briefly) E/R diagrams Some high-level design principles Modeling constraints Introduction to the relational model From E/R & ODL to relations

6. Object Definition Language Is part of ODMG, which also gave us OQL. Resembles C++ (and Smalltalk). Basic design paradigm in ODL: –Model objects and their properties. For abstraction purposes: –Group objects into classes. What qualifies as a good class? –Objects should have common properties.

7. ODL Class Declarations Interface { attributes: ; relationships ; methods } Method example: float gpa(in: Student) raises (noGrades) Arbitrary function can compute the value of gpa, based on a student object given as input.

8. ODL Example Product Person Company category name price name stockprice name addressssn

9. ODL Declarations Interface Product { attribute string name; attribute float price; attribute enum Categories {electronics, communications, sports …} category } Interface Company { attribute string name; attribute float stockprice; } Interface Person { attribute integer ssn; attribute string name; attribute Struct Address {string street, string city} address; }

10. ODL Example Extended Product Person Company category name price name stockprice name addressssn buys worksFor madeBy

11. ODL Declarations, Extended Interface Product { attribute string name; attribute float price; attribute enum Categories {electronics, communications, sports …} category; relationship madeBy; } Interface Person { attribute integer ssn; attribute string name; attribute Struct Address {string street, string city} address; relationship set buys; relationship set worksFor;}

12. ODL Example, Extended Again Product Person Company category name price name stockprice name addressssn buys worksFor madeBy employs makes

13. ODL Declarations, Extended Again Interface Company { attribute string name; attribute float stockprice; relationship set makes inverse Product::madeBy; relationship set employs inverse Person::worksFor; }

14. Types in ODL Basic types: Atomic types (e.g., string, integer, …) Interface types (e.g., Person, Product, Company) Constructors: Set: (1, 5, 6) Bag: (1, 1, 5, 6, 6 ) List: (1, 5, 6, 1, 6 ) Array: Integer[17] Struct: {string street, string city, integer zipcode}

15. Allowable Types in ODL For attributes: start with atomic or struct, and apply a collection type. OK: string, set of integer, bag of Address. Not OK: Product, set of set of integer. For relationships: start with interface type and apply a collection type. OK: Product, set of Product, list of Person. Not OK: struct {pname Product, cname Company} set of bag of Product integer

16. Entity / Relationship Diagrams Objects entities Classes entity sets Attributes are like in ODL. Relationships: like in ODL except - not associated with classes (I.e., first class citizens) - not necessarily binary Product address buys

17. address namessn Person buys makes employs Company Product namecategory stockprice name price

18. Multi-way Relationships Purchase How do we model a purchase relationship between buyers, products and stores? Product Person Store

19. Roles in Relationships Purchase What if we need an entity set twice in one relationship? Product Person Store salesperson buyer

20. Roles in Relationships Purchase Product Person Store salesperson buyer Note the multiplicity of the relationships: we cannot express all possibilities

21. Attributes on Relationships Purchase Product Person Store date

22. Design Principles Purchase Product Person What’s wrong? President PersonCountry Moral: be faithful!

23. What’s Wrong? Purchase Product Store date person weather Moral: don’t talk too much.

24. What’s Wrong? Purchase Product Person Store date Dates Moral: don’t complicate life more than it already is.

25. Do we really need 3-way relationships? Purchase Person Store Product StoreOf ProductOf BuyerOf Moral: Find a nice way to say things.

26. Modeling Subclasses The world is not flat! Some objects in a class may have properties not shared by other members: Products Software products Educational products So --- we define subclasses (in ODL and in E/R).

27. Subclasses in ODL Interface SoftwareProduct: Product{ attribute Set platform; attribute Set requiredMemory; } Interface EducationalProduct: Product{ attribute Struct Interval {integer begin, integer end} ageGroup; attribute string topic } The two classes also inherit all the properties of Product.

28. Product namecategory price isa Educational ProductSoftware Product Age Groupplatforms Subclasses in E/R Diagrams

29. Multiple Inheritance Product Educational Product Educ-software Product Software Product ageGroup topic Platforms required memory Educational-method

30. How do we resolve conflicts? Product Educational Product Educ-software Product Software Product ageGroup topic Platforms required memory Educational-method Rating (ATA) Rating (ASA) Rating?

31. Product namecategory price isa Educational ProductSoftware Product Age Groupplatforms In ODL: Every object belongs to a single class In E/R: An entity may be spread out in multiple sets.

32. Modeling Constraints Extracting constraints is what modeling is all about. But how do we express them? Examples: Keys: social security number uniquely identifies a person. Single-value constraints: a person can have only one father. Referential integrity constraints: if you work for a company, it must exist in the database. Domain constraints: peoples’ ages are between 0 and 150. Why are these constraints useful in the implementation?

33. Keys A set of attributes that uniquely identify an object or entity: Person: social security number name name + address name + address + age Perfect keys are often hard to find, so organizations usually invent something. An object may have multiple keys: employee number, social-security number

34. Keys in ODL Interface Person (key ssn) { properties… } Defining multiple keys: (key ssn employeID (name address age))

35. Keys in E/R Diagrams address namessn Person Product namecategory price No formal way to specify multiple keys in E/R diagrams

36. Single Value Constraints An entity (or object) may have at most one value for a given attribute or relationship. Person: name, social-security number Company: stock price How do we do this in ODL? In E/R, every attribute has at most one value. Arrows tell us about multiplicity of relations. If we have a single-valued constraint, we can either: 1. Require that the value exist (see referential integrity shortly) 2. Allow null values.

37. Referential Integrity Constraints A relationship has one value and the value must exist. Example: Product madeBy Company: company must exist. How do we enforce referential integrity constraints? (otherwise, we get dangling pointers) - forbid to delete a reference object, or - delete the objects that reference an object we’re deleting. CompanyProduct makes In E/R diagrams:

38. Weak Entity Sets Entity sets are weak when their key attributes come from other classes to which they are related. This happens if: - part-of hierarchies - splitting n-ary relations to binary. UniversityTeam affiliation numbersportname

39. The Relational Data Model Database Model (ODL, E/R) Relational Schema Physical storage ODL definitions Diagrams (E/R) Tables: row names: attributes rows: tuples Complex file organization and index structures.

40. Terminology Name Price Category Manufacturer gizmo $19.99 gadgets GizmoWorks Power gizmo $29.99 gadgets GizmoWorks SingleTouch $149.99 photography Canon MultiTouch $203.99 household Hitachi tuples Attribute names What can’t you say in the relational model?

41. More Terminology Every attribute has an atomic type. Relation Schema: relation name + attribute names + attribute types Relation instance: a set of tuples. Only one copy of any tuple! Database Schema: a set of relation schemas. Database instance: a relation instance for every relation in the schema.

42. More on Tuples Formally, a mapping from attribute names to (correctly typed) values: name gizmo price $19.99 category gadgets manufacturer GizmoWorks Sometimes we refer to a tuple by itself: (note order of attributes) (gizmo, $19.99, gadgets, GizmoWorks) or Product (gizmo, $19.99, gadgets, GizmoWorks).

43. Updates The database maintains a current database state. Updates to the data: 1) add a tuple 2) delete a tuple 3) modify an attribute in a tuple Updates to the data happen very frequently. Updates to the schema: relatively rare. Rather painful. Why?

44. From ODL to Relational Schema Start simple: a class definition has only single valued attributes Interface product{ float price; string name; Enum {telephony, gadgets, books} category} Class becomes a relation, and every attribute becomes a relation attribute: Name Price Category Gizmo $19.99 gadgets Product

45. Adding Non atomic Attributes Name Currency Amount Category Gizmo US$ 19.99 gadgets Power Gizmo US$ 29.99 gadgets Price is a record: {string currency, float amount} Product

46. Set Attributes Name SSN Phone Number Fred 123-321-99 (201) 555-1234 Fred 123-321-99 (206) 572-4312 Joe 909-438-44 (908) 464-0028 Joe 909-438-44 (212) 555-4000 One option: have a tuple for every value in the set: Disadvantages?

47. Modeling Collection Types How can we model bags? Lists? Fixed length arrays? The problem becomes even more significant if a class has several attributes that are set types? Question: how bad is the redundancy for n set type attributes, each with possibly up to m values? Questions:

48. Modeling Relationships Interface Product { attribute string name; attribute float price; relationship madeBy; } Interface Company { attribute string name; attribute float stock-price; attribute string address; } How do we incorporate the relationship madeBy into the schema?

49. Option #1 Name Price made-by-name made-by-stock-price made-by-address Gizmo $19.99 gizmoWorks 0.0001$ Montezuma What’s wrong?

50. Hint Interface Product { attribute string name; attribute float price; relationship madeBy; } Interface Company { attribute string name; attribute float stock-price; attribute string address; relationship set makes; }

51. Better Solution Name Price made-by-name Gizmo $19.99 gizmoWorks Product relation: (assume: name is a key for company) Company relation: Name Stock Price Address Gizmo $0.00001 Montezuma

52. Additional Issues 1. What if there is no key? 2. What if the relationship is multi-valued? 3. How do we represent a relationship and its inverse?

53. From E/R Diagrams to Relational Schema Easier than ODL - relationships are already independent entities - only atomic types exist in the E/R model. Entity sets relations Relationships relations Special care for weak entity sets.

54. address namessn Person buys makes employs Company Product namecategory Stock price name price

55. Entity Sets to Relations Product namecategory price Product: Name Category Price gizmo gadgets $19.99

56. Relationships to Relations makes Company Product namecategory Stock price name Relation Makes (watch out for attribute name conflicts) Product-name Product-Category Company-name Starting-year gizmo gadgets gizmoWorks 1963 Start Year

57. Handling Weak Entity Sets UniversityTeam affiliation numbersportname Relation Team: Sport Number University-name mud wrestling 15 Montezuma State U. - need all the attributes that contribute to the key of Team - don’t need a separate relation for Affiliation.

58. Modeling Subclass Structure Product Educational Product Educ-software Product Software Product ageGroup topic Platforms required memory Educational-method

59. Option #1: the ODL Approach 4 tables: each object can only belong to a single class Product(name, price, category, manufacturer) EducationalProduct( name, price, category, manufacturer, ageGroup, topic) SoftwareProduct( name, price, category, manufacturer, platforms, requiredMemory) EducationalSoftwareProduct( name, price, category, manufacturer, ageGroup, topic, platforms, requiredMemory)

60. Option #2: the E/R Approach Product(name, price, category, manufacturer) EducationalProduct( name, ageGroup, topic) SoftwareProduct( name, platforms, requiredMemory) No need for a relation EducationalSoftwareProduct Unless, it has a specialized attribute: EducationalSoftwareProduct(name, educational-method)

61. Option #3: The Null Value Approach Have one table: Product ( name, price, manufacturer, age-group, topic, platforms, required-memory, educational-method) Some values in the table will be NULL, meaning that the attribute not make sense for the specific product. How many more meanings will NULL have??