• Data Modelling
• Some Design Ideas
• Exercise: GMail Data Model
• Quality of Designs

❖ Data Modelling

Aims of data modelling:

• describe what information is contained in the database
  (e.g., entities: students, courses, accounts, branches, patients, ...)
• describe relationships between data items
  (e.g., John is enrolled in COMP3311, Tom's account is held at Coogee)
• describe constraints on data
  (e.g., 7-digit IDs, students can enrol in no more than 3 courses per term)

Data modelling is a design process

• converts requirements into a data model

❖ Data Modelling (cont)

Kinds of data models:

• logical: abstract, for conceptual design, e.g., ER, ODL, UML
• physical: record-based, for implementation, e.g., relational, SQL

Strategy: design using abstract model; map to physical model

❖ Some Design Ideas

Consider the following while working through exercises:

• start simple ... evolve design as problem better understood
• identify objects (and their properties), then relationships
• most designs involve kinds (classes) of people
• keywords in requirements suggest data/relationships
  (rule-of-thumb: nouns → data, verbs → relationships)
• don't confuse operations with relationships
  (operation: he buys a book; relationship: the book is owned by him)
• consider all possible data, not just what is available

❖ Exercise: GMail Data Model

Consider the GMail system (or any other modern mail client)

Develop an informal data model for it by identifying:

• the data items involved (objects and their attributes)
• relationships between these data items
• constraints on the data and relationships

❖ Exercise: GMail Data Model (cont)

Objects in GMail data model:

users
        gmail-address, name, password, ...

messages
        timestamp, sender*, title, content, ...

tags
        owner, name, colour parent*

settings
        name, value, user*

Relationships in GMail data model:

recipients
        user - message

sent
        user - message

tag-hierarchy
        child-tag - parent-tag

settings
        user - setting

Constraints in GMail data model:

gmail-address values are unique

users must have a password (strong?)

every message has a sender

every message has a non-empty title and content

values for each setting are valid for that setting

❖ Quality of Designs

There is no single "best" design for a given application.

Most important aspects of a design (data model):

• correctness   (satisfies requirements accurately)
• completeness   (all reqs covered, all assumptions explicit)
• consistency   (no contradictory statements)

Potential inadequacies in a design:

• omits information that needs to be included
• contains redundant information (⇒ inconsistency)
• leads to an inefficient implementation
• violates syntactic or semantic rules of data model