Data Definition

Data Definition and Integrity Constraints Reading: C&B, Chap 6

In this lecture you will learn • the different SQL data types & related scalar functions • how to define new data types with DDL statements (this is all DDL stuff) • some of the integrity constraints used in DBMSs • SQL's Integrity Enhancement Features (IEF) • how integrity constraints can affect row operations • the notion of schemas Dept. of Computing Science, University of Aberdeen

2

SQL's Integrity Enhancement Features (IEF) • So far, we have thought of databases as static repositories. In fact, real databases are often very ‘active’ with 100's of users simultaneously querying and updating the DB. • So database integrity is important • IEFs allow the DB designer to specify & enforce: – – – – –

domain constraints required data entity integrity referential integrity enterprise constraints (business rules) Dept. of Computing Science, University of Aberdeen

3

Creating Tables - Data Definition • CREATE TABLE is used to define relational tables • it defines the data type for each column • defines rules for how data may be inserted and deleted

•

CREATE TABLE Staff (StaffNo VARCHAR(5), Lname VARCHAR(20), Salary FLOAT, HireDate DATE);

VARCHAR, FLOAT, and DATE are examples of domains • Domains specify type & range of allowed data values Dept. of Computing Science, University of Aberdeen

4

Built-in Data Types (Domains) in ANSI SQL • ANSI SQL supports many data types (vendors often also have own dialects):

– CHARACTER (CHAR), CHARACTER VARYING (VARCHAR) – NUMERIC, DECIMAL (DEC), INTEGER (INT), SMALLINT – FLOAT, REAL, DOUBLE PRECISION – DATE, TIME, TIMESTAMP – BOOLEAN, BIT – BINARY LARGE OBJECT (BLOB), etc.

• Some types have an associated size . e.g. CHAR(5) Dept. of Computing Science, University of Aberdeen

5

User-Defined Domains in ANSI SQL Domains are sort of like data types you create yourself… CREATE DOMAIN SexType AS CHAR(1) DEFAULT 'M' CHECK (VALUE IN ('M', 'F')); CREATE TABLE Staff (StaffNo VARCHAR(5), Lname VARCHAR(20), Salary FLOAT, HireDate DATE, Sex SexType); INSERT INTO Staff VALUES ('S0057', 'Smith', 12075.50, '12-JAN-1990', 'F'); . . OK INSERT INTO Staff VALUES ('S0023', 'Jones', 14250.50, '14-FEB-1997', 'X'); . . Fails

•

SexType acts as a constraint on allowed range of values Dept. of Computing Science, University of Aberdeen

6

Required Data & More Domain Constraints

• Required Data: some data can’t contain nulls  NOT NULL column specifier • Example: CREATE TABLE Staff ( StaffNo VARCHAR(5) NOT NULL, Lname VARCHAR(20) NOT NULL, Salary FLOAT CHECK (Salary BETWEEN 50 and 20000), HireDate DATE, Sex SexType);

• StaffNo & Lname are required - may not be NULL • Domain Constraints: specifies what’s ok in col/table • The CHECK clause gives a domain constraint for Salary • Updates & insertions will fail if constraints not satisfied Dept. of Computing Science, University of Aberdeen

7

Dynamic Domain Constraints • Domains may be defined ‘dynamically’ using values that already exist in the database: CREATE DOMAIN StaffNoDomain AS VARCHAR(5) CHECK (VALUE IN (SELECT StaffNo FROM Staff)); CREATE TABLE PropertyForRent (PropertyNo VARCHAR(5) NOT NULL, StaffNo StaffNoDomain);

• This could be used to ensure every StaffNo in PropertyForRent is valid • Domains can be deleted: DROP DOMAIN DomainName [RESTRICT | CASCADE]

Dept. of Computing Science, University of Aberdeen

8

Scalar Functions • Scalar functions may be used to convert/manipulate data values (remember aggregates: MIN, MAX, etc.) • Example: SELECT SUBSTRING(Lname FROM 1 TO 3), CONVERT(INTEGER Salary), EXTRACT(YEAR FROM HireDate) FROM Staff;

Result SMI

12075 1990

• ANSI SQL supports many scalar functions... • See CB, Table 6.2, p163 Dept. of Computing Science, University of Aberdeen

9

Entity Integrity - Primary Keys

• Reminder: the primary key of each row in a table must be unique and non-null. • Example: The primary key of the Viewing table is composed of two attributes (composite key): CREATE TABLE Viewing ( ClientNo VARCHAR(5) NOT NULL, PropertyNo VARCHAR(5) NOT NULL, PRIMARY KEY (ClientNo, PropertyNo));

• SQL will reject operations that would violate primary key uniqueness • Can use UNIQUE(Colname) to enforce uniqueness of alternate keys; UNIQUE keys must also be declared as NOT NULL Dept. of Computing Science, University of Aberdeen

10

Referential Integrity - Foreign Keys • Reminders: • A foreign key links a child table to its parent table. • If a foreign key is non-null, it must match an existing row in the parent table. • So... SQL has more keywords for this: CREATE TABLE PropertyForRent (... StaffNo VARCHAR(5) NOT NULL, FOREIGN KEY (StaffNo) REFERENCES Staff);

• SQL will reject operations that would violate referential integrity Dept. of Computing Science, University of Aberdeen

11

Referential Integrity and Referential Actions • Question: if a key attribute in the parent table is modified, what should happen in the child table ? - SQL provides 4 alternative referential actions: • ON UPDATE & ON DELETE specify referential actions which determines what happens when an attempt to UPDATE or DELETE a candidate key in a parent table which has matching rows in child table: FOREIGN KEY (Key) REFERENCES Table [ON DELETE | UPDATE Action] – – – –

CASCADE - apply changes to child rows (could trickle down) SET NULL - set child keys to NULL SET DEFAULT - set child keys to DEFAULT value NO ACTION - reject the operation (default)

• Suppose a client is removed from the DreamHome DBMS. What's the most appropriate action to specify for ClientNo (primary key) in the Viewing table? Dept. of Computing Science, University of Aberdeen

12

Enterprise Constraints (Business Rules) • Sometimes, real-world business rules involve constraints that refer to more than one table. Its useful to define enterprise constraints just once. • Example: A member of staff may manage no more than 100 properties: CREATE ASSERTION StaffNotOverLoaded CHECK (NOT EXISTS (SELECT StaffNo FROM PropertyForRent GROUP BY StaffNo HAVING COUNT (*) > 100)); CREATE TABLE PropertyForRent ( ... CONSTRAINT StaffNotOverLoaded);

Dept. of Computing Science, University of Aberdeen

13

Triggers • Often, real-world business rules cannot be implemented using constraints. • Example: The branch manager is notified by e-mail if a client views more than 10 properties. • Different DBMSs often provide a trigger mechanism • Triggers may contain procedural code (if/then/else, function calls)—discussed in C&B pp.967-971 • Triggers can implement complex database operations • However, triggers can add to database complexity (hidden rules) • Triggers are not ANSI standard Dept. of Computing Science, University of Aberdeen

14

Putting It All Together - Schemas • A schema is a collection of named, related DBMS objects: • Tables, Domains, Constraints, Views (virtual relations, more on this later), Triggers, and more ... • A multi-user DBMS may contain multiple schemas: • Each schema is owned by a given user • A Database Administrator (DBA) manages schemas (CREATE, DROP) • Schemas are maintained in special system tables • However, different DBMSs have different ways of managing schemas... Dept. of Computing Science, University of Aberdeen

15

Simplified Data Model of a DBMS

Dept. of Computing Science, University of Aberdeen

16

Database Schemas Evolve Over Time • Ideally, a database is created once and then used for many years ... BUT • The data model may be improved (integrity, performance) ... • New features may be added in new releases ... • Enterprise rules may change ... • Therefore, SQL provides many options for changing tables: • See ALTER TABLE, CB Ch. 6, p173 Dept. of Computing Science, University of Aberdeen

17

Summary So Far... DBs are ‘active’ or ‘alive’ - contents always changing The structure of a DB can also evolve over time... DB contents should always be consistent - integrity ANSI SQL provides several Integrity Enhancement Features (IEFs) • IEF => domain constraints, entity/referential integrity, business rules... • IEFs imply additional design choices for new DBs • One DBMS can manage multiple DBs - notion of schemas & privileges • • • •

Dept. of Computing Science, University of Aberdeen

18