TOWARDS PATTERN MANAGEMENT SYSTEM

Erki Eessaar

Institute of Informatics, Tallinn Technical University, Raja 15,12618 Tallinn, Estonia

Keywords: knowledge management, reuse, patterns, Pattern Management System, reuse repository.

Abstract: Patterns allow division of knowledge in

to manageable pieces. It is necessary to develop Pattern

Management System (PMS) in order to effectively use patterns. In this article components of PMS are

proposed. PMS uses metadata about patterns and database of patterns to guide management of patterns.

Textual descriptions of patterns and models what specify pattern are stored in the database of patterns.

1 INTRODUCTION

Reuse is a practice that uses existing knowledge to

solve problems. Patterns permit division of

knowledge into manageable and reusable pieces.

Knowledge about the problem, its context and

solution can be written down as a pattern.

Researchers have proposed repositories of

reusa

ble software components (Gaedke & Rehse,

2000), (Sugumuran & Storey, 2003). Authors like

(Halaris & Geroupoulus, 1996) and (Backlund &

Jonasson, 2002) have stressed that patterns can be

used during all the phases of Information System

Development (ISD) Process and therefore Pattern

repository is needed.

Author’s goal is to develop Pattern Management

Syste

m (PMS), which would allow creation / storage

/ modification / classification / search of different

types of patterns. It could be used for example

during ISD. PMS could also assist development of

methods for ISD (Method Engineering). According

to (Baskerville & Stage, 2001), information systems

development practitioners invent, combine and adapt

fragments of methods to a development situation

rather than wholly adopt published methods. It is a

process that is similar to the usage of patterns.

PMS must support current and evolving

andards, for example Pattern and Component

Markup Language (PCML), which describes among

other things metamodel of patterns.

The rest of the paper is organised as follows.

Sectio

n 2 introduces architecture of PMS. It mostly

concentrates on the structure of database. Section 3

summarizes and points to the future work with

current topic.

2 ARCHITECTURE OF PMS

(Marco, 2000) describes meta-data management

systems (MDMS) what help to collect and use

information about the structure of information

systems, processes that occur in the information

systems and parties that use these systems. (Marco,

2000) predicts that bidirectional metadata

architecture will evolve in the future where “meta

data is changed in the meta data repository and fed

back into the meta data's original source." (Marco,

2000, p. 311) Author of current article suggests that

PMS should be one additional component of

MDMS. For example logical and physical data

models can be created using the assistance of PMS

and then changes can be automatically made in the

databases based on them.

This article proposes architecture of Pattern

anagement System (PMS) what uses relational

database to store patterns data. PMS should have

web-based interface so that developers around the

world could use this.

2.1 Description of patterns

Description of patterns consists of two equally

important portions:

– Structured textual descriptions. For example

description of the problem, context where the

problem occurs and solution to the problem.

– Models that complement textual descriptions,

describing the problem, solutions to the problem

and also examples of pattern usage.

Models present more structured view to the

omain than descriptions in natural language and

therefore they provide better base for searching

655

Eessaar E. (2004).

TOWARDS PATTERN MANAGEMENT SYSTEM.

In Proceedings of the Sixth International Conference on Enterprise Information Systems, pages 655-658

DOI: 10.5220/0002617106550658

 SciTePress

patterns. Models can be created for example using

the CASE tools what have capabilities to save

models as XMI files. Information from these files

could be loaded to the database of PMS. For

example (Hay, 1996) describes universal data

models, which can be starting point for the

development of logical data models and are actually

patterns. It is possible to search model-fragments

using database queries if information about entities,

relationships and attributes is stored in a database. If

model element is found, then it is possible to search

other model-fragments and their associated

documentation, what contain the same or

semantically similar element. For example if user

decides that database must contain information

about persons and organizations then we can find

other patterns what reference to the entity “party” or

its subtypes. It helps to find new requirements to the

database.

2.2 Main components of PMS

Architecture of PMS consists of following basic

layers:

– Meta layer, which allows management of meta-

data about patterns. It will be used to guide

queries and changes in the database, because this

layer manages information about the structure

elements of patterns, pattern types and

corresponding database objects.

– Pattern layer, which enables management of data

of actual patterns and objects what are associated

with them.

Pattern management will be performed using the

meta-data about patterns. For example user must

select pattern type for the registration of pattern.

Program will identify its structure elements (name,

problem, context etc.) based on selected type and

will generate user interface for the registration of

pattern. Program saves data about patterns to the

relational database based on the mapping between

pattern structure elements and corresponding

relations and attributes in the database. Next the

main functional subsystems of PMS will be

presented.

2.2.1 Main functional subsystems of PMS

– Subsystem for managing meta-data of patterns. It

permits definition of the structure elements of

pattern description styles and construction of the

styles using predefined elements.

– Subsystem for pattern management. Its main

functions are creation, modification, search and

evaluation of patterns and pattern languages. It

must also be capable to extract and load model

information from XMI files to the database.

– Subsystem for managing opinions of users about

patterns.

– Supporting information management subsystems

for managing information about objects with

which patterns have associations.

2.2.2 Sub-components of PMS database

Next the sub-components of PMS database will be

presented.

1. Registers of patterns meta-data

Each pattern has description style. Logical structure

elements and domains are the building blocks, which

are used for the construction of the description

styles.

– Register of logical structure elements of pattern

description styles where data about structure

elements and their attributes is registered.

Structure elements are for example "context",

"forces", "problem", "solution", "positive

consequences of pattern usage", "negative

consequences of pattern usage", "examples".

Attributes could be for example "description"

and "summary".

– Register of domains. Author suggests that each

attribute has domain in the context of pattern

description style. Domain describes the rules that

apply to this attribute in the context of the style:

a) Data type of the attribute.

b) Is the value of the attribute required?

c) Is the value of the attribute unique among other

patterns? For example name of the pattern

should be unique.

d) What is the default value of this attribute?

e) Other constraints to the values of the attribute.

– Register of description styles of patterns. Each

style consists of set of logical structure elements.

Appropriate domain from register of domains is

associated with each attribute of logical structure

element. System must also permit creation of

new styles based on existing styles. One

important fact (multiplicity of element) what is

registered about association of structure element

and style, determines amount of values what can

correspond to this structure element in the real

pattern. For example each pattern can have one

base name but many alias names, forces or code

samples.

– Register of metamodels, which describe

elements of the model types. For example class

diagrams and collaboration diagrams illustrate

and explain patterns of object-oriented design.

Metamodels determine the structure of database

where models will be stored.

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

656

– Register of physical structure of patterns

database. It contains information about relations

and their attributes in the database where

components of patterns are kept. Central relation

in the database of patterns is called “Patterns”. It

contains one tuple for each pattern, which is

stored in the database of patterns.

– Register of pattern types. Pattern type is

associated with description style and one or more

domains where patterns can be used.

– Register of mappings between elements of

pattern description styles and physical structure

elements of database of patterns.

Addition or modification of a domain or style

causes PMS to change the structure of database and

record this change in the register of physical

structure of patterns database and in the register of

mappings. If new domain is defined then system

creates domain object in the database. If structure

element is added to the description style, then PMS

creates new relation in the database. It will have

foreign key attribute which possible values are

identifiers of patterns from the relation “Patterns”.

Database might contain more than one relation

where attribute values of same type of structure

element are registered. Attributes in these relations

have different domains. Each relation has associated

triggers what enforce appropriate multiplicity of the

element. Meta-data guides management of data in

these relations.

2. Registers that contain actual patterns and

user opinions about them

– Register of textual descriptions of patterns.

Textual descriptions are organized according to

the logical structure of pattern description style.

– Register of models what are associated with

patterns. PMS has more than one such register -

one for each model type. Each model type has

metamodel, which describes model elements that

can be used to create model of that type.

Structure of each register is developed based on

metamodel.

– Register of pattern languages. Pattern language

is a special case of pattern what describes

solutions to the set of problems. One pattern

could belong to many pattern languages and

language contains more than one pattern. Pattern

language as pattern might have its own structure

elements (name, problem etc.) and associated

models.

– Register of user evaluations of patterns. Patterns

can be searched based on them. Patterns have

different type of descriptive attributes:

a) Attributes, which are common to all the patterns

and have a same set of possible values in all

domains (type A).

b) Attributes, which are common to all the patterns,

but in different domains have different set of

possible values (type B).

c) Attributes, which are used only with patterns

from specific domains and have a set of possible

values (type C).

Value from a pre-agreed structured set of terms

may be associated with such attributes. Sometime it

is difficult to determine exact value of attribute from

the list of values. For example nowadays there is no

governing organization, which standardizes patterns

and decides that one pattern is still a candidate

pattern but another is mature and accepted pattern.

Some of such attributes (they have type A) are:

1. Success of a pattern.

2. Maturity of a pattern.

3. Usability of documentation of a pattern.

4. Level of abstraction of the proposed solution.

Creation of patterns is collaborative work.

Therefore judgements of users could be used to

determine the values of attributes as well.

Users could compare patterns in pairs in regard

to selected attribute, using values from the

fundamental scale of decision method “Analytic

Hierarchies Process (AHP)”. For example result of a

comparison might be that pattern p

has essentially

better usability than p

. If new pattern is created,

then the patterns manager, who has obligations to

encourage the use of patterns, compares it with

patterns which are well known to all or most

developers for their good or bad properties. Pattern

users can compare patterns what they have used.

If patterns are searched based on attributes which

values have been determined by users, then it is

necessary to determine other pattern, compared to

which the value of an attribute is found. For example

query could be: “Find patterns of type t which have

better usability than pattern p

”

Next example presents some attributes of type B

and some of their possible values in the Information

System Development domain.

1. Level of technicality of a problem (sample

values: “Business Environment”; “Enterprise

Model”, "Technical model").

2. View to the system (sample values: “Data”;

“Functions”; “Time”; “People”).

3. Extent of the problem what pattern solves

(sample values: “Subsystem”; “Component”).

4. Activities what the pattern helps to perform

(sample values: “Modeling”; “Programming”).

Users of PMS could associate patterns with

attribute (type B, C) values together with

membership function value (between 0 and 1). It

permits calculation of average membership function

value for each pair of attribute value and pattern.

TOWARDS PATTERN MANAGEMENT SYSTEM

657

3. Registers, containing supporting information

These registers are part of any major information

system. Common patterns what describe their

structure have been worked out and have been

published as patterns (Hay, 1996), (Fowler, 1997).

– Register of documents what contain information

about patterns or what have been used to create

or complement patterns. Patterns of documents

data model (Hay, 1996, p. 205-234) can be used

to develop the structure of this register.

– Register of parties (persons and organizations)

who have some role in the events, which have

happened or have been scheduled to happen with

patterns. Patterns of domain models about parties

(Fowler, 1997, p. 24-27) can be used to develop

the structure of this register.

– Register of planned and happened events, which

have happened with patterns (for example

creation, review or modification of a pattern).

Patterns of proposed and implemented actions

(Fowler, 1997, p. 158-160) can be used to

develop the structure of this register. For each

event next data will be stored: time of an event;

reference to the parties who participated in the

event, together with the information about their

roles in the event; references to the documents

what where used during the event, together with

the information about their roles in the event.

– Register of queries, what users of PMS have

constructed and stored for later usage by them or

other interested parties.

3 CONCLUSION AND FUTURE

In this paper, principles of Pattern Management

System (PMS) were presented. PMS could help find

solutions to the problems and store evolving

knowledge in the form of patterns. Author proposes,

that meta-data about structure of patterns and

database of patterns must guide the operation of

PMS in order to manage different types of patterns.

PMS must manage both narratives and information

extracted from models. It makes possible to use

query capabilities of existing database management

systems to search patterns based on model elements.

Future work will involve the development of

methods for searching patterns. Search based on user

evaluations should take into account the amount of

evaluations and quality of experts who make them.

One expert could evaluate patterns in regard to some

attributes many times. If lot of experts have changed

their opinion recently in one direction, then it could

also be taken into account.

One promising methods is to search patterns

according to semantic similarity. For example during

requirements analysis functional requirements to the

system are documented using use cases. Data model

can be constructed by searching data modeling

pattern which model elements (entities, attributes

and associations) have the biggest semantic

similarity to subset of the use case text.

REFERENCES

Backlund, P., & Jonasson, I., 2002. 'Patterns as a Means

for Managing Knowledge in the Information Systems

Engineering Process', in BalticDB&IS 2002, vol. 2,

eds H. M. Haav & A. Kalja, Tallinn Technical

University, Tallinn, pp. 15-25.

Baskerville, R., & Stage, J., 2001. 'Accommodating

emergent work practices: etnographic choice of

method fragments', in Realigning Research and

Practice in Information Systems Development: The

Social and Organizational Perspective, eds N. L.

Russo, B. Fitzgerald & J. I. DeGross, Kluwer

Academic Publishers, pp. 11-28.

Fowler, M., 1997. Analysis Patterns: Reusable Object

Models, Addison Wesley, Calif.

Gaedke, M., & Rehse, J., 2000. 'Supporting compositional

reuse in component-based Web engineering',

Proceedings of the 2000 ACM symposium on Applied

computing. Retrieved 17 Oct. 2003 from ACM digital

library.

Halaris, J., & Geroupoulus, S.T., 1996. 'Reuse Concepts

and a Reuse Support Repository', IEEE Symposium

and Workshop on Engineering of Computer Based

Systems, pp. 27-34. Retrieved 18 Aug. 2003 from

IEEE Xplore.

Hay, D. C., 1996. Data model patterns: conventions of

thought, Dorset House Pub, New York.

Marco, D., 2000. Building and Managing the Meta Data

Repository: A Full Lifecycle Guide, John Wiley &

Sons, USA.

Sugumuran, V., & Storey, V.C., 2003. 'A semantic-based

approach to component retrieval', ACM SIGMIS

Database, vol. 34, issue 3, pp.8-24. Retrieved 17 Oct.

2003 from ACM digital library.

Pattern and Component Markup Language. Draft 3.

Retrieved June 19 2003 from

http://www.objectventure.net/files/docs/PCMLSpecifi

cation.pdf

ICEIS 2004 - INFORMATION SYSTEMS ANALYSIS AND SPECIFICATION

658