44271: Database Design & Implementation Logical Data Modelling (Avoiding Database Anomalies) Ian Perry Room: C49 Tel Ext.:

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling What is a Logical Data Model? A ‘robust’ representation of the initial decisions made when building our Conceptual Data Model, which was composed of: Entities Attributes Relationships When I say ‘robust’ I mean that this model MUST ‘perform’ well with respect to a specific style/type of software.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Database Theories & Software Hardware independent, the match to ‘type’ of software is only concern, e.g.: Hierarchical DBMS Relational DBMS Object-based DBMS Each Database Theory addresses: Data Structure Data Integrity Data Manipulation

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Database Theory = Relational Model First proposed by Dr. E. F. Codd in June Codd E F, (1970), A Relational Model of Data for Large Shared Data Banks, Communications of the ACM, Vol. 13, No. 6, Pgs 377 – 387. Codd's model is now accepted as the definitive model for relational database management systems (RDBMS). Structured English QUEry Language ("SEQUEL") was developed by IBM Corporation, Inc., to use Codd's model. SEQUEL later became SQL. In 1979, Relational Software, Inc. (now Oracle Corporation) introduced the first commercial implementation of SQL. SQL is the most widely used RDBMS manipulation language.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Relations look like Entities, but … Entity Staff(SCode, Name, Address, DoB, …) May discover requirement for ‘extra’ Attributes, and also need to ‘complete’ our list of Attributes for each Relation. Relation Staff(SCode, Name, Address, DoB, DoE) Entity Contract(CCode, Site, Begin, End, …) Can’t draw relationship lines, so need to ‘add’ extra attributes to Relations at the ‘M’ end of any ‘1:M’ relationships; e.g. 1 Staff “take part in” M Contract. Relation Contract(CCode, Site, Begin, End, SCode)

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Use Tables to ‘flesh-out’ your Logical Model Staff(SCode, Name, Address, DoB, DoE)  Contract (CCode, Site, Begin, End, SCode)  NB. Tables ARE NOT Relations!

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Primary & Foreign Keys Most important Attributes in a Relation are know as ‘Keys’: of which there are two types. Primary Key: One, or more, Attribute(s) that identify a unique occurrence of the ‘Entity’ that this ‘Relation’ represents. Foreign Key: Attributes used (i.e. instead of the lines of an ER Diagram) to represent the presence of relationships. Often referred to as: The Primary/Foreign Key Mechanism.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Attributes, Domains & Relationships Attribute Values should be atomic (i.e. simple/single values only); e.g.: ‘address’ should be separated into ‘street’ & ‘town’ & ‘postcode’. Set of eligible Attribute Values is known as an Attribute’s Domain; e.g.: if we only have 100 members of staff, then the Domain of the ‘SCode’ Attribute could be “whole numbers between 1 & 100”. The Relational Model is weak at explicitly modelling relationships: Attributes in different Relations MUST HAVE same Attribute Domain for relationship to be possible.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Codd’s Rules Each Tuple (i.e. row) MUST BE unique, i.e.: need a way to discriminate between Tuples. Therefore: each Relation MUST HAVE a Primary Key. There may be many Candidates for the job of Primary Key, so select on basis of: uniqueness AND/OR minimality. Keys with more than one Attribute: are know as composite keys.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Rules for Integrity No Attribute that is part of the Primary Key can assume a ‘null’ value, else: how could we discriminate between Tuples? Foreign Key Attributes must take values that are either ‘null’, or from same Domain as the Primary Key Attribute to which they are logically linked, else: we will lose the possibility of making relationships.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Avoiding Database Anomalies Most Database books have a section describing a mathematically-based technique called Normalisation: I will show you a much easier way of achieving the same result. What we want to achieve is a ‘robust’ Logical Data Model; i.e. by: Transforming a Conceptual Data Model into a set of Relations. Checking these Relations for any Anomalies. Documenting them as a Database Schema.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling What is an Anomaly? Anything we try to do with a database that may lead to unexpected and/or unpredictable results. Three types of Anomaly; i.e.: insert delete update Need to check your database design carefully: the only good database is an anomaly free database.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Insert Anomaly When we want to enter a value into a data cell but the attempt is prevented, as the primary key value is not known.  e.g. We have built a new Room (e.g. B123), but it has not yet been timetabled for any courses (so we don’t have a CoNo value).

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Delete Anomaly When a value we want to delete also means we will delete values we wish to keep.  e.g. CoNo 351 has ended, but Room C320 will be used elsewhere.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Update Anomaly When we want to change a single data item value, but must update multiple entries  e.g. Room H940 has been improved, it is now of RSize = 500.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Conceptual Model & Translation Process Conceptual Model: Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB,...) Student(Enrol-No, Name, Address, OLevelPoints,...) Course(CourseCode, Name, Duration,...) Staff Course Student 1MMM  Translation Process:  Entities become Relations  Attributes become Attributes(?)  Key Attribute(s) become Primary Key(s)  Relationships are represented by additional Foreign Key Attributes;  for those Relations that are at the ‘M’ end of each 1:M Relationship.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling The ‘Staff’ & ‘Student’ Relations Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB,...) becomes: Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB) Student(Enrol-No, Name, Address, OLevelPoints,...) becomes: Student(Enrol-No, Name, Address, OLevelPoints, Tutor) NB. Foreign Key Tutor references Staff.Staff-ID

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling The ‘Staff’ & ‘Course’ Relations Course(CourseCode, Name, Duration,...) becomes: Course(CourseCode, Name, Duration) NB.Can’t ‘simply’ add extra attributes to act as Foreign Keys; as BOTH Relations have a ‘M’ end:  I warned you about leaving M:M relationships in your Conceptual Data Model. MUST create an ‘artificial’ linking Relation. Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB)

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling ‘ Staff’, ‘Course’ & ‘Team’ Relations NB.In the ‘artificial’ Team Relation: Primary Key is a ‘composite’ of CourseCode & Staff-ID Foreign Key CourseCode references Course.CourseCode Foreign Key Staff-ID references Staff.Staff-ID Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB) Course(CourseCode, Name, Duration) Team(Staff-ID, CourseCode)

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling 4 Relations from 3 Entities? OK, BUT are they anomaly free? Is every Tuple unique? i.e. is there a Primary Key. Are the Attributes Atomic? i.e. do they store only ONE item of data. Does every Attribute within each Relation ‘depend’ upon the Primary Key? Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB) Course(CourseCode, Name, Duration) Team(Staff-ID, CourseCode) Student(Enrol-No, Name, Address, OLevelPoints, Tutor)

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling What if the checks fail? If any Relation fails ‘checks’: especially those checking dependency. we MUST split that Relation into multiple Relations: until they pass the tests. but MUST remember to leave behind a Foreign Key: to ‘point’ forwards to the Primary Key of the ‘new’ split-off Relation.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Are they Anomaly Free? Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB) Course(CourseCode, Name, Duration) Team(Staff-ID, CourseCode) Student(Enrol-No, Name, Address, OLevelPoints, Tutor) NOT Dependent upon Staff-ID; Requires a slightly more complex ‘solution’. NOT very Atomic; Could easily be split into ‘Street’, ‘Town’ & ‘PostCode’.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling ‘Fixing’ the Dependency ‘Problem’ The Attribute ‘RateOfPay’ depends upon ‘ScalePoint’ NOT ‘Staff-ID’. So, we MUST remove ‘RateOfPay’ from the ‘Staff’ Relation, like this: NB. In the ‘Staff’ Relation: Foreign Key ScalePoint references Pay.ScalePoint Staff(Staff-ID, Name, ScalePoint, RateOfPay, DOB) Staff(Staff-ID, Name, ScalePoint, DOB) Pay(ScalePoint, RateOfPay)

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling 5 Relations from 3 Entities Now all we need to do: Is to document our ‘Anomaly Free’ Relations as a Database Schema. Staff(Staff-ID, Name, ScalePoint, DOB) Course(CourseCode, Name, Duration) Team(Staff-ID, CourseCode) Student(Enrol-No, Name, Street, Town, PostCode, OLevelPoints, Tutor) Pay(ScalePoint, RateOfPay)

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Document Relations as a Database Schema A Database Schema: defines all Relations, lists all Attributes (with their Domains), and identifies all Primary & Foreign Keys. We may/should have ‘discovered’ a number of constraints during our analysis of the Business situation, e.g: the College only delivers 10 Courses. there are only 12 Points on the Pay Scale. Staff MUST be at least 21 Years Old. These constraints can/should be expressed as the ‘Domains’ of the Database Schema.

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Logical Schema 1 - Domains Schema College Domains StudentIdentifiers = ; StaffIdentifiers = ; GeneralNames = TextString (15 Characters); Addresses = TextString (20 Characters); PostCodes = TextString (7 or 8 Characters); CourseIdentifiers = ; OLevelPoints = ; ScalePoints = ; StaffBirthDates = Date (dd/mm/yyyy), >21 Years before Today;

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Logical Schema 2 - Relations Relation Student Enrol-No: StudentIdentifiers; Name: GeneralNames; Street: Addresses; Town: Addresses; PostCode: PostCodes; OLevelPoints: OLevelPoints; Tutor: StaffIdentifiers; Primary Key: Enrol-No Foreign Key Tutor references Staff.Staff-ID

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Logical Schema 3 - Relations Relation Staff Staff-ID: StaffIdentifiers; Name: GeneralNames; ScalePoint: ScalePoints; DOB: StaffBirthDates; Primary Key: Staff-ID Foreign Key ScalePoint references Pay.ScalePoint

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling Logical Schema... Relation Course CourseCode: CourseIdentifiers; Name: GeneralNames; … etc. Continue to define each of the Relations in a similar manner. NB. Make sure that you define ALL of the Relations, including: ‘artificial’ ones (e.g. Team) ‘split-off’ ones (e.g. Pay)

Ian PerrySlide : Database Design & Implementation: Logical Data Modelling This Week’s Workshop The purpose of this week’s Workshop is to practice developing ‘robust’ logical data models that conform to the ‘rules’ of Codd’s Relational Model. Exploring the ‘definition’ of Relations. Identifying potential anomalies in a Table of data, and ‘solving’ these ‘problems’. Documenting a Database Schema (i.e. a Logical Model), in the format required by Part 2 of the Assignment.