ICFCA08.ppt

Refactorings of Design Defects using
Relational Concept Analysis

Naouel Moha, Amine Mohamed Rouane Hacene,
Petko Valtchev, and Yann-Gaël Guéhéneuc

LORIA, France
DIRO, University of Montréal, Canada
LATECE, Université du Québec à Montréal, Montréal, Canada

ICFCA’08
Montréal (Qc), Canada, February 25-28, 2008

Context

Moha © Refactorings of Design Defects using RCA 2

Context


Context: Overall Process
Textual descriptions
of design defects

1 S 3
P
S
E
U
C
G
I
G
F
E
I
S
C
T
A
I
T
O
I
N
O
N

Rules Refactorings

2 D
C 4
O
E
R
T
R
E
E
C
C
T
T
I
I
O
O
N
N

Classes having Better
design defects Design

Textual descriptions
of design defects

1 S 3
P
S
E
U
C
G
I
G
F
E
I
S
C
T
A
I
T
O
I
N
O
N

Rules Refactorings

2 D
C 4
O
E
R
T
R
E
E
C
C
T
T
I
I
O
O
N
N

Classes having Better
design defects Design

1-2. Specification and Detection of DDs
More and more tools and techniques
Actively researched area
Method DECOR (Defect dEtection for CORrection) [Moha 06]

3. Suggestion
3 S ti
Manual identification of the modifications
Time-, resource-consuming, error-prone activity
, g, p y

4. Correction
Refactorings
g
Technique used to change “the internal structure of software to make it
easier to understand and cheaper to modify without changing its
observable behavior”


What Kind of Defects ?
Design Patterns are “good” solutions to recurring design problems

Design Defects (DDs)
are “bad” solutions to recurring problems, Antipatterns [Brown 98]

hinder d
hi d development and maintenance by making programs h d t
l t d i t b ki harder to
comprehend and/or evolve

DDs of interest: infected by low cohesion and high coupling
Cohesion: how closely the methods are related to the variables in
the class

Coupling: the degree of its reliance on services provided by other
classes


What Kind of Defects ?
An example of DDs [Brown 98]

Blob (God Class)
“ Procedural-style design leads to one object with a
lion’s share of the responsibilities while most other
objects only hold data or execute simple processes ”

Large controller class

Many fields and methods with a low cohesion

High coupled with the data stored in
associated data classes


An Example
Blob

Before


An Example
How t correct th d f t ?
H to t the defect

Before After
Large and complex class Large class becomes less complex
Small data classes Data classes gain more behaviour
Not object-oriented More object-oriented style

An Example
Redistribute l
R di t ib t class members among existing classes ( ith possibly
b i ti l (with ibl
new classes) to increase cohesion and–or decrease coupling

Before After


Contribution
Suggestion
RCA [Rouane 07, Ph.D. Thesis]
Extension of FCA to relational d
E i f l i l data
Models the inter-object links and infers description logics role like
relations between concepts
RCA provides a suitable framework for clustering individuals along
the properties they share and their links with other individuals
Identify methods that share common fields and methods that call
common methods cohesive sets low coupled


Contribution
Suggestion
RCA [Rouane 07, Ph.D. Thesis]
Extension of FCA to relational d
E i f l i l data
Models the inter-object links and infers description logics role like
relations between concepts
RCA provides a suitable framework for clustering individuals along
the properties they share and their links with other individuals
Identify methods that share common fields and methods that call
common methods cohesive sets low coupled

«D
Develop an automated approach f suggesting d f t
l t t d h for ti defect-
correcting refactorings using RCA »


An Example
H to t the defect

Before After
Large and complex class Large class becomes less complex
Small data classes Data classes gain more behaviour
Not object-oriented More object-oriented style

An Example
H to t the defect

Blob Cohesive Sets


An Example
H to t the defect

I = {W-borrow_date_Book, W-return_date_book}
E = {}
I = {‘check_Availability_Book()’}
E = {check_Availability_Book()}

I = {call:c4}
E = {}

I = {‘borrow_Book()’} I = {‘reserve_Book()’, W-reserved_book}
E = {borrow_Book()} E = {reserve_Book()}

Lattice Cohesive Set


An Example
H to t the defect

Cohesive Sets After Refactoring *

* Move Method Mode Field, Create Class
Method, Field


Suggestion


Suggestion
1. RCF Modeling


Suggestion
2. RCA Al
Algorithms
ith


Suggestion
2. RCA Al
Algorithms
ith

9
I = {W-library opened}
{W library_opened}
E={}

10
I = {‘open_Library()’}
E = {open Library()}
{open_Library()}
7
I = {‘close_Library()’}
E = {close_Library()}


Suggestion
2. RCA Al
Algorithms
ith


Suggestion
2. RCA Al
Algorithms
ith

4
I = {‘check_Availability_Book()’}
E = {check_Availability_Book()}

19
I = {call:c4}
E={}

3 12
I = {‘borrow_Book()’} I = {‘reserve_Book()’,W-reserved_book}
E = {borrow_Book() } E = {reserve_Book()}


Suggestion
3. Interpretation

Cohesive Sets After Refactoring

* Move Method Mode Field, Create Class
Method, Field


Experimental Study
Tooling
PADL to model source code and generate contexts
Galicia to construct and visualize the concept lattices

Experiments
E i t
Goal: evaluate the relevance of the cohesive sets suggested
4 different open-source programs


System Experimental Study
Nb of fields/ Nb of Nb of real
Size (fields +
Blob Class LOC methods cohesive cohesive Precision
methods)
moved sets sets
DHTTransportUDPImpl 2,049 (42+66) 108 (27+32) 59 10 7 70%
V2.3.0.6
Azureus

DHTControlImpl 1,868 (47+80) 127 (35+62) 97 19 11 58%
TRTrackerBTAnnouncerImpl 1,393 (36+47) 83 (24+33) 57 16 5 31%
V0.7.1 V1.4 V1.2.1

LogBrokerMonitor 1,142 (29+105)134 (23+85) 108 31 17 55%
ene Log4j

Category 387 (9+53) 62 (8+44) 52 18 9 50%
IndexReader 236 (7+52) 59 (5+30) 35 4 2 50%
Nutch Luce

QueryParser 829 (36+48) 84 (24+37) 61 13 10 77%
FSNamesystem 710 (24+35) 59 (17+25) 42 18 9 50%
JobTracker 555 (
(22+31) 53
) (
(17+18) 35
) 11 8 73%
Average Precision: 57%


Conclusion
Contribution
an approach that uses RCA to suggest refactorings to correct
certain DDs in particular Blob
DDs,

Validation
4 different programs
Relevant refactorings to improve programs

Future Work
- Generalise to other DDs
- Assess more programs via our approach
- Discuss the suggested refactorings with their developers and
apply them


Questions
Contact: mohanaou@iro.umontreal.ca
http://www-etud.iro.umontreal.ca/~mohanaou

Thanks for your attention !


ICFCA08.ppt

More Related Content

Viewers also liked (7)

Similar to ICFCA08.ppt (20)

More from Ptidej Team (20)

Recently uploaded (20)

ICFCA08.ppt