47
Banaras Hindu University
Banaras Hindu University
A Course on
Software Reuse by Design Patterns and Frameworks
by
Dr. Manjari Gupta Department of Computer Science
Banaras Hindu University
Lecture 7
Issues and Challenges in Design Pattern Detection
∗ not only beneficial to the forward engineering process
∗ patterns can also indicate the design rationale behind the system’s implementation
∗ can aid in code comprehension
Design Pattern Detection
∗ no support for documenting the presence and usage of patterns in code
∗ reflect the designers’ intents ∗ ongoing research into correlations between the use of
Design Patterns and defect frequency ∗ Design document is often missing in many legacy
systems ∗ helps on program comprehension and design
visualization ∗ Code maintenance would be made easier
Importance of Design Pattern Detection
∗ should automatic ∗ should detect nonstandard implementations of design
patterns along with their standard implementations ∗ should fast ∗ should not specific to few particular design patterns. ∗ should be customizable ∗ should detect as less number of false positive and false
negative patterns as possible.
Important features of a method for detecting Design Patterns
∗ executable code- static and dynamic analyses, ∗ documentation of source codes, ∗ design model (UML designs) and ∗ other specifications.
Product used for DPD
∗ must accurately reflect the traits of each pattern ∗ Must differentiate one pattern from other patterns ∗ strictness of definition ∗ more redundancy in a pattern’s definition, the
more false positives ∗ formal design patterns definition is required
Design Pattern Definition
∗ based on the decomposition approach. ∗ getting idea about EDPs or meta patterns ∗ Design patterns are expressed exclusively through EDPs ∗ Hayashi et al [Hayashi08] used Pree’s meta patterns to
represent these common properties ∗ checked if the pre-defined conditions of the design pattern,
including the combination conditions of the meta patterns, hold or not.
∗ If it fails in this stage, they moved to other patterns that have the same meta patterns. ∗ reduce any redundant repetition of checking meta patterns
Design pattern detection based on their common properties
∗ pattern detection tools includes three parts ∗ a parser, a detector, and a database.
∗ challenges in pattern detection ∗ the size of the exploration space for large software
systems.
Approaches to detect design patterns
∗ similarity score algorithm to detect design patterns. ∗ structural relations between classes are encoded in
multiple graphs and matrices. ∗ similarly the design patterns are also encoded in
matrices. ∗ similarity score between the matrices of system
source and those of patterns.
Tsantalis et al. approach
∗ SPQR ∗ System for Pattern Query and Recognition ∗ developed at the University of Carolina ∗ uses Elemental Design Patterns, a theorem prover
and ρ-calculus to detect DPs.
Tools for DPD
∗ Fujaba [Niere02] ∗ forward and reverse engineering tool ∗ developed at the University of Paderborn ∗ uses graph matching on abstract syntax graphs
representing source code, combined with fuzzy logic. ∗ treat design and implementation variants of patterns
separately.
Tools for DPD Cont…
∗ Ptidej [Gueheneuc05] ∗ visual tool for design recovery and DP recognition. ∗ Detection is performed through constraints
satisfaction with explanation
Tools for DPD Cont…
∗ system and patterns, they can be represented in the form of matrices.
DPD based on only structural and static analysis [Pande10]
Calculation of matrices for system design
Client
AbstractFactory
+CreateProduct()
AbstractProduct
ConcreteFactory ConcreteProduct
Calculation of matrices for system design
Client
AbstractFactory
+CreateProduct()
AbstractProduct
ConcreteFactory ConcreteProduct Direct Association matrix of Fig. 1 (root = 3)
Calculation of matrices for system design
Client
AbstractFactory
+CreateProduct()
AbstractProduct
ConcreteFactory ConcreteProduct Direct Association matrix of Fig. 1 (root = 3)
Dependency matrix Fig. 1 (root = 2)
Calculation of matrices for system design
Client
AbstractFactory
+CreateProduct()
AbstractProduct
ConcreteFactory ConcreteProduct Direct Association matrix Dependency matrix Generalization matrix of Fig. 1 (root = 3) of Fig. 1 (root = 2) of Fig. 1 (root = 5)
Calculation of matrices for system design
Client
AbstractFactory
+CreateProduct()
AbstractProduct
ConcreteFactory ConcreteProduct Direct Association matrix Dependency matrix Generalization matrix of Fig. 1 (root = 3) of Fig. 1 (root = 2) of Fig. 1 (root = 5)
Overall Matrix
Calculation of matrices for Façade design pattern
Facade
Subsystem Classes
Façade Design Pattern
Calculation of matrices for Façade design pattern
Facade
Subsystem Classes
Façade Design Pattern
Direct Association matrix of fig 5 (root = 2)
Calculation of matrices for Façade design pattern
Facade
Subsystem Classes
Façade Design Pattern
Direct Association matrix of fig 5 (root = 2)
Overall Matrix
Calculation of matrices for Façtory Method design pattern
Factory Method Design Pattern
Product
ConcreteProduct
Creator
+FactoryMethod()
ConcreteCreator
Calculation of matrices for Façtory Method design pattern
Factory Method Design Pattern
Product
ConcreteProduct
Creator
+FactoryMethod()
ConcreteCreator
Dependency matrix of fig 2 (root = 3)
Calculation of matrices for Façtory Method design pattern
Factory Method Design Pattern
Product
ConcreteProduct
Creator
+FactoryMethod()
ConcreteCreator
Dependency matrix of fig 2 Generalization matrix of fig 2 (root = 3) (root = 5)
Calculation of matrices for Façtory Method design pattern
Factory Method Design Pattern
Product
ConcreteProduct
Creator
+FactoryMethod()
ConcreteCreator
Dependency matrix of fig 2 Generalization matrix of fig 2 (root = 3) (root = 5)
Overall Matrix
Calculation of matrices for singleton design pattern
Singleton
+Instance(): Singleton -instance
Singleton Design Pattern
Calculation of matrices for singleton design pattern
Singleton
+Instance(): Singleton -instance
Singleton Design Pattern
[ 1 ] Direct Association matrix of fig 5 (root = 2)
Calculation of matrices for singleton design pattern
Singleton
+Instance(): Singleton -instance
Singleton Design Pattern
[ 1 ] Direct Association matrix of fig 5 (root = 2)
[21]
Calculation of matrices for singleton design pattern
Singleton
+Instance(): Singleton -instance
Singleton Design Pattern
[ 1 ] Direct Association matrix of fig 5 (root = 2)
[21]
[2] Overall Matrix
∗ determine whether a graph is isomorphic to a subgraph of another graph.
∗ is NP-complete. ∗ G1 (V1, E1) and G2 (V2, E2) be two graphs ∗ M1 and M2 be the adjacency matrices ∗ A permutation matrix is a square (0, 1)-matrix ∗ G1 (M1, Lv, Le) and G2 (M2, Lv, Le) are said to be
isomorphic if there exist a permutation matrix P such that
M2 = P M1 PT
Sub-graph isomorphism detection
∗ Given an n x n matrix M = (mij), let Sk,m(M) denote the k x m matrix that is obtained from M by deleting rows k + 1 , . . . , n and columns m + 1 , . . . , n, where k, m < n.
∗ Let G1 and G2 be graphs with adjacency matrices M1 and M2 of dimensions m x m and n x n respectively, where m < =n. There is a subgraph isomorphism from G1 to G2 if there exists an n x n permutation matrix P such that
M1 = Sm, m (P M2 PT) M2 -system design matrix and M1 -design pattern matrix
Sub-graph isomorphism detection
Design Pattern Detection as Façade Design Pattern
Design Pattern Detection as Façade Design Pattern
P =
Design Pattern Detection as Façade Design Pattern
P = PT =
Design Pattern Detection as Façade Design Pattern
P = PT =
PSPT =
Design Pattern Detection as Façade Design Pattern
P = PT =
PSPT =
After eliminating entries from 3rd row and 3rd column we have reduced matrix as (because Façade Design Pattern is of 2 x 2 order)
Design Pattern Detection as Factory Method Design Pattern
Design Pattern Detection as Factory Method Design Pattern
Design Pattern Detection as Factory Method Design Pattern
Design Pattern Detection as Factory Method Design Pattern
Design Pattern Detection as Factory Method Design Pattern
Design Pattern Detection as Factory Method Design Pattern
After eliminating 5th row and 5th column (since factory method has order 4 x 4),
∗ For singleton there will be no permutation matrix for which we can find out (after row and column elimination) a matrix which is equivalent to design pattern matrix.
Particular Design Pattern may or may not exist
