Advanced Database Technologies

Instructor: Eric Umuhoza

Email: [email protected]

June 18, 2023

1 / 47
 Active Databases

Active Databases
Preliminary definition
• Active Database: is a DBMS that supports an integrated
subsystem for the definition and management of
production rules – active rules.
• Event Condition Action (ECA) Paradigm
– each rule reacts to some events, evaluates a condition and,
based on the truth value of the condition, might carry out an
action.
• Rule Engine: an autonomous subsystem which keeps
track of the events that have occurred and schedules the
rules for execution.
• An active database system can execute either transactions,
which are initiated by the users, or rules, which are under
the control of the system.
2 / 47
 Active Databases

Active Databases

• A database which offers active rules

• Normally the word trigger is used
• We describe here more in depth:
– Definition in SQL:1999
– Execution model
– Termination analysis
– Design issues
– Several examples

3 / 47
 Active Databases

Passive Database

• Periodically polling
– Too frequent: expensive
– Infrequent: miss the right time to reach
• The polling must be done for all items in stock

4 / 47
 Active Databases

Active Database

• Triggers define actions when situations occur

• Actions are usually database updates

5 / 47
 Active Databases

Base Structure of a Trigger

• Base paradigm: Event–Condition–Action

When an event occurs

If the condition is true
The action is executed

• Other examples of rules in DBMS world:

– Integrity constraints
– Business rules
• Problem: difficult to realize complex systems

6 / 47
 Active Databases

Event-Condition-Action

• Event
– Normally a modification of the state of the database: insert, delete,
update
– When the event occurs, the trigger is activated
• Condition
– A predicate identifying the situations in which the application of a
trigger is necessary
– When the condition is evaluated, the trigger is considered
• Action
– A generic update statement or a stored procedure
– When the action is elaborated, the trigger is executed
• DBMSs already provide all the required components. One only
needs to integrate them

7 / 47
 Active Databases

Triggers in SQL:1999, Syntax

• SQL:1999 (aka SQL-3) proposes a syntax similar to what Oracle

Server and IBM DB2 offered before 1999
• Systems will tend to comply to it
• Each trigger is characterized by:
– name
– name of the monitored table
– execution mode (before or after)
– monitored event (insert, delete or update)
– granularity (statement-level or row-level)
– names and aliases for transition values and transition tables
– action
– creation timestamp

8 / 47
 Active Databases

Triggers in SQL:1999, Syntax

9 / 47
 Active Databases

The Referencing clause

• The format depends on the granularity

– For row-level mode, there are two transition variables –old and new–
that represent the value prior to and, resp., following the modification
of the tuple under consideration
– For statement-level mode, there are two transition tables –old table
and new table– that contain the old and, resp., new value of all
modified tuples
• Variables old and old table can’t be used with triggers whose
event is insert
• Variables new and new table can’t be used with triggers whose
event is delete
• Variables and transition tables are extremely important for the
realization of efficient triggers

10 / 47
 Active Databases

A Simple Example

• Table T2 is a replication of table

T1: whenever an update occurs
on T1, this is replicated on T2.
• Replication is realized by means
of triggers.

11 / 47
 Active Databases

Insertion of a New Tuple into T1

12 / 47
 Active Databases

Deletion of a Tuple from T1

13 / 47
 Active Databases

Update of a Tuple in T1

14 / 47
 Active Databases

Trigger with a condition

• Variant: Table T2 is a replication of table T1, but only the tuples

whose value is > 10 are replicated.
• Insertion operation:

15 / 47
 Active Databases

Execution of a Single Trigger

• Execution mode:
1 BEFORE
The trigger is considered and possibly executed before the action that
activated it is applied
Normally this mode is used when one wants to check a modification
before it takes place

1 AFTER
The trigger is considered and possibly executed after the action that
activated it is applied
It’s the most common mode, suitable for most applications
It’s the simplest one to be used correctly

16 / 47
 Active Databases

Example of BEFORE Trigger

• Extension: no tuples with negative values must be inserted (for

the update operation a similar trigger is defined)

17 / 47
 Active Databases

Granularity of Events

• Statement-level mode (default mode)

– The trigger is considered and possibly executed only once for each
statement that activated it, independently of the number of modified
tuples
– It’s closer to the traditional approach of SQL statements, which
normally are set-oriented
• Row-level mode ( for each row option)
– The trigger is considered and possibly executed once for each tuple
modified by the statement
– It allows writing triggers more simply
– It may be less efficient

18 / 47
 Active Databases

For Each Row vs. For Each Statement /1

• A set of tuples is deleted from T1:

19 / 47
 Active Databases

For Each Row vs. For Each Statement /2

• A set of tuples is deleted from T1:

20 / 47
 Active Databases

Execution of Multiple Triggers: Conflicts between Triggers

• If several triggers are associated to the same event, SQL:1999

prescribes the following policy
1 BEFORE statement-level triggers are executed
2 BEFORE row-level triggers are executed
3 The modification is applied and the integrity constraints defined on the
DB are checked
4 AFTER row-level triggers are executed
5 AFTER statement-level triggers are executed

• If there are several triggers belonging to the same category, the

order of execution chosen by the system depends on the
implementation

21 / 47
 Active Databases

Execution Model
• SQL:1999 states that triggers are handled within a Trigger
Execution Context (TEC)
• The execution of a trigger action may produce events that
activate other triggers, which will have to be evaluated within a
new, internal TEC
• In every moment there may be several, nested TECs for a
transaction, but only one may be active
• For row-level triggers, the TEC considers which tuples have
already been considered and which ones are to be considered
• The structure is therefore stack-like:
– TEC0 → TEC1 → ... → TECn
• When a trigger has considered all the events, the TEC closes and
the system moves to the next trigger
• The model is complex, but precise and relatively easy to
implement
22 / 47
 Active Databases

Example: Salary Management

Employee
RegNum Name Salary DeptN ProjN
50 Smith 59.000 1 20
51 Black 56.000 1 10
52 Jones 50.000 1 20

Project
Department ProjNum Objective
DeptNum MGRRegNum 10 NO
1 50 20 NO

23 / 47
 Active Databases

Example Trigger T1: Bonus

Event: update of Objective in Project
Condition: Objective = ’YES’
Action: Increase by 10% the salary of the employees
involved in
the project

24 / 47
 Active Databases

Example Trigger T2: CheckIncrement

Event: update of Salary in Employee
Condition: New salary greater than manager’s salary
Action: Decrease salary and make it the same as the
manager’s

25 / 47
 Active Databases

Example Trigger T3: CheckDecrement

Event: update of Salary in Employee

Condition: Decrement greater than 3%
Action: Decrement salary only by 3%

26 / 47
 Active Databases

Activation of T1
update Project
set Objective = ’yes’ where ProgNum = 10

27 / 47
 Active Databases

Activation of T2

Event: update of Salary in Employee

Condition: true (employee Black’s salary is greater than
manager Smith’s salary)
Action: Black’s salary is modified and set to Smith’s salary

Employee
RegNum Name Salary DeptN ProjN
50 Smith 59.000 1 20
51 Black 59.000 1 10
52 Jones 50.000 1 20

• T2 is activated again – the condition in false

• T3 is activated

28 / 47
 Active Databases

Activation of T3

Event: update of Salary in Employee

Condition: true (Black’s salary salary was decreased
by more than 3%)
Action: Black’s salary is decreased by only 3%

Employee
RegNum Name Salary DeptN ProjN
50 Smith 59.000 1 20
51 Black 59.752 1 10
52 Jones 50.000 1 20

• T3 is activated again – the condition is false

• T2 is activated again – the condition is true

29 / 47
 Active Databases

Activation of T2

Employee
RegNum Name Salary DeptN ProjN
50 Smith 59.000 1 20
51 Black 59.000 1 10
52 Jones 50.000 1 20

Activation of T3
• The trigger condition is false
– Salary was decreased by less than 3%
• Trigger activation has reached termination

30 / 47
 Active Databases

Design -Trigger Properties

• It’s important to ensure that interferences among triggers and

chain activations do not produce anomalies in the behavior of the
system
• 3 classical properties
1 Termination: for any initial state and sequence of modifications, a
final state is produced (no infinite activation cycles)
2 Confluence: triggers terminate and produce a unique final state,
independent of the order in which triggers are executed
3 Determinism of observable behavior: triggers are confluent and
produce the same sequence of messages
• Termination is by far the most important property

31 / 47
 Active Databases

Termination Analysis

• There are several conceptual tools, most of which graph-based

• The simplest is the triggering graph
– A node for each trigger
– An arc from a node ti to a node tj if the execution of ti ’s action can
activate trigger tj (can be done with a simple syntactic analysis)

• If the graph is acyclic, the system is guaranteed to terminate

– There cannot be infinite trigger sequences

• If the graph has some cycles, it maybe non–terminating (but it

might as well be terminating)

32 / 47
 Active Databases

Termination Graph for the Salary Management Triggers

• The graph is cyclic, but the repeated execution of the triggers

reaches termination anyway

33 / 47
 Active Databases

Problems in Designing Trigger Applications

• Triggers have a very high potential

– They enrich the database by adding data management aspects that are
otherwise distributed over all applications that use the data
– but this potential is little used
• Realizing trigger applications is complex
• The available IDEs are not adequate
• DBMS vendors use triggers to realize innovative services, by
introducing mechanisms for automated trigger generation
• Examples:
1 Constraint management
2 Data replication

34 / 47
 Active Databases

Applications of Active Databases

• Classical applications: internal rules within the database

– Triggers generated by the system and not visible by the user
– Main aspects:
Integrity constraint management, computation of derived data,
replicated data management
– Other aspects:
Versioning management, privacy, security, action logging, event
recording, etc.
• External rules (or business rules)
– They express application-specific knowledge

35 / 47
 Active Databases

Applications 1: Referential Integrity Management

• Repair strategies for violations of referential integrity constraints

– The constraint is expressed as a predicate in the condition part
• Example:
CREATE TABLE Employee (
...
FOREIGNKEY(DeptN) REFERENCES Department(DeptNum)
ON DELETE SET NULL,
...);
• Operations that can violate this constraint:
– INSERT in Employee
– UPDATE of Employee.DeptN
– UPDATE of Department.DeptNum
– DELETE in Department

36 / 47
 Active Databases

Actions in the Employee Table

Event: insert in Employee

Condition: the new DeptN value is not among
those in the Dept table
Action: insertion is inhibited, reporting error

• For the update of DeptN in Employee , the trigger only changes

in the event part
37 / 47
 Active Databases

Deletion in the Department Table

Event: delete in Department

Condition: there exist employees belonging to that department
Action: set their DeptN to NULL

38 / 47
 Active Databases

Updates in the Department Table

Event: update of DeptNum in Department

Condition: the new DeptNum value is among those
in the Employee table
Action: DeptN in the Employeeis also modified

39 / 47
 Active Databases

Application 2: Recursion Management

• Triggers for recursion management

• Ex.: representation of a hierarchy of products
– Each product is characterized by a super-product and by a depth level
in the hierarchy
– Can be represented by a recursive view (with recursive construct in
SQL:1999)
– Alternatively: use triggers to build and maintain the hierarchy
Product(Code,Name,Description,SuperProduct,Level)
• Hierarchy represented by SuperProduct and Level
• Products not contained in other products:
SuperProduct = NULL and Level = 0

40 / 47
 Active Databases

Deletion of a Product

• In case of product deletion, all its sub-products must be deleted

as well

41 / 47
 Active Databases

Insertion of a New Product

• In case of an insertion, the appropriate Level value must be

calculated

42 / 47
 Active Databases

Priorities, Activations, and Groups

• Definition of priority
– Allows specifying the execution order of triggers when there are several
triggers activated at the same time
– SQL:1999 states an order based on the execution mode and granularity;
when these coincide, the choice depends on the implementation

• Activation/deactivation of triggers
– Not in the standard, but often available

• Organization of triggers in groups

– Some systems offer trigger grouping mechanisms, so as to
activate/deactivate by groups

43 / 47
 Active Databases

Execution Modes

• The execution mode describes the connection between the

activation (event) and the consideration and execution
phases (condition and action)
– Condition and action are always evaluated together

• 3 classical alternatives
1 Immediate: the trigger is considered and executed with the activating
event
Example: triggers that immediately check satisfaction of integrity
constraints
2 Deferred: handled at the end of the transaction
Example: triggers that check satisfaction of integrity constraints that
require the execution of several operations
3 Detached: the trigger is handled in a separate transaction
Example: efficient management of variations of stock indices values
after several exchanges

44 / 47
 Active Databases

Trigger Syntax in MySQL

45 / 47
 Active Databases

Trigger Syntax in PostgreSQL

46 / 47
Active Databases

The End

47 / 47