Programing Slicing and Its applications

Indian Institute of Technology, Kharagpur

Date : 27 March-2014 Ankur Jain, Dept of Computer Sc & Engg., IIT Kgp

Presented By- Ankur Jain (13CS60D02)

Programmer’s Nightmare…


If I change this statement in program:

What other statements would be affected?

(Impact analysis)

What other statement needs to be tested?

(Regression test)

The values live at this statement:

Defined where?

Modified Where?

(Manual Checking)

How can I abstract code?

Studies show….


Software testing contributes more then

50% of cost and time of SDLC

Maintainers spend nearly 50% of their time trying to understand the program

Source: http://www.ece.cmu.edu/~koopman/des_s99/sw_testing/

Try...


Program Slicing!

Outline


Introduction

Types of Slicing

Program Models

Flow based

Dependency based

Slicing Algorithms

Challenges

Directions of research

Proposed by…


Proposed by Mark Weiser in 1981

Chief scientist at Xerox PARC

As his PhD thesis

Father of ubiquitous computing

Mark D. Weiser (July 23, 1952 - April 27, 1999)

Basic Concepts…


“For statement S and variable V, the slice of program P includes only those statements of P needed to capture the behavior of V at S.”

Slicing Criterion (SC): <S, V>

S = Point of interest in the program (statement) V = Subset of variables used in the program

A program slice is a subset of a program

Example…


1 begin

2 read(x, y)

3 total := 0.0

4 sum := 0.0

5 if x <= 1

6 then sum := y

7 else begin

8 read(z)

9 total := x*y*z

10 end

11 write(total, sum)

12 end

SC = <11, sum>

2 read(x, y)

4 sum := 0.0

5 if x <= 1

6 then sum := y

Slicing Criterion

Why is Program Slicing Useful?


• Program slices are more manageable for

testing and debugging

• During testing, debugging, or understanding a program, most of the code in the program is irrelevant to what you are interested in.

• Program slicing provides a convenient way of filtering out “irrelevant” code.

Slice Variants…


Types of slices: Static Dynamic

Direction of slice: Forward Backward

Executability of slice: Executable Non-executable

Amorphous, etc.

Froward Slices Vs Backward Slices


Statements which might affect the value of variables in V

Statements which might be

affected by the values of

variables in V

int i, sum, prd;

1. read(i);

2. prd = 1;

3. sum = 0;

4. while (i<10)

5. sum = sum + i;

6. prd = prd * i;

7. i = i + 1;

8. write(sum);

9. write(prd); For SC = <9, prd>

Slice: {1, 2, 4, 6, 7}

For SC = <2, prd>

Slice: {6, 9}

Backward Slices:

Froward Slices

Static Slice | Dynamic Slice


Using dynamic information (an execution trace)

Debugging

Set of all statements that actually affect the value of a variable at a program point

Using static information (a source program)

Regression Testing

Set of all statements that may affect the value of a variable at a program point

Example…


int i, sum, prd;

1. read(i);

2. prd = 1;

3. sum = 0;

4. while (i<10)

5. sum = sum + i;

6. prd = prd * i;

7. i = i + 1;

8. write(sum);

9. write(prd);

Static Slice

{1, 2, 4, 6, 7}

Dynamic Slice

For Input ‘i’ = 15

{2}

For Slicing Criteria [SC] = <9, prd>

Flow based model: Control flow Data flow Dependency based model: Data Dependency Control Dependency

Slicing Criterion

Dependency Based Model


Program Code

Control Flow Graph (CFG)

Data Dependency Graph (DDG)

Program Dependency Graph (PDG)

Forward Dominance Tree

Control Dependency Graph (CDG)

Data Dependency Graph (DDG)


int a, b, sum;

1. read(a);

2. read(b);

3. sum = 0;

4. while(a<8)

5. sum = sum + b;

6. a = a + 1;

7. write(sum);

8. sum = b;

9. write(sum);

4 5

6

7

8

9

2 1 3

Data Dependency Graph

Control Flow Graph (CFG)


int a, b, sum;

1. read(a);

2. read(b);

3. sum = 0;

4. while(a<8)

5. sum = sum + b;

6. a = a + 1;

7. write(sum);

8. sum = b;

9. write(sum);

Start 1

Stop

4

5

3

6

2

8

7

9

True

False

True

True

True

True

True

True

True

True

True

Dominance


Let X and Y be two nodes in a Control flow graph

X dominates Y

If and only if

Every path from Start to Y passes through X

Y forward dominates X

The forward Dominance Tree (FDT) 1. Vertices represent statement 2. Root node of tree is exit node of the CFG 3. Arcs represent immediate forward dominance

Forward Dominance Tree (FDT): Using CFG


9

4 6

7

5

8

2

3

1

Start 1

Stop

4

5

3

6

2

8

7

9

True

False

True

True

True

True

True

True

True

True

True

Control Dependency Graph: Using CFG & FDT


4

5 6

7

8

9

2

1

3

S

1. Y is control dependent on X

2. Path in the CFG from X to Y

3. Doesn’t contain the immediate forward dominator of X

Y

X

X

Ifdom(4)=7

X

Y

Control Dependency Graph (CDG)


int a, b, sum;

1. read(a);

2. read(b);

3. sum = 0;

4. while(a<8)

5. sum = sum + b;

6. a = a + 1;

7. write(sum);

8. sum = b;

9. write(sum);

4

5 6

7

8

9

2

1

3

S

Program Dependency Graph (PDG)


int a, b, sum;

1. read(a);

2. read(b);

3. sum = 0;

4. while(a<8)

5. sum = sum + b;

6. a = a + 1;

7. write(sum);

8. sum = b;

9. write(sum);

Union of CDG and DDG

1

4

6

5

8

2 9

3

7

Data dependence Control dependence

Limitation: A PDG can model programs with a single function Not suitable for inter-procedural slicing

System Dependency Graph Model: by- Horwitz


Basic Idea: Connect PDGs with auxiliary dependence edges

• Parse source code - one procedure at a time.

– Construct the CDG for each procedure including main.

• Add actual and formal parameter nodes:

– Connect using parameter-in, parameter-out edges

• Represent function calls

– Using call edges

• Find data dependencies:

– Perform data flow analysis of the CDGs

– Connect data dependence edges

• Add summary edges

Steps in SDG Construction

System Dependency Graph (SDG)


Control Dependence Data Dependence Call, Parameter−in, Parameter−out Summary Edge

bin = b

ain = a

entry add

Entry main

a = 0

b = 1 add(a, b)

a =ain

b = bin

a = a+b

c = aout

aout = a

main(){ int a, b; a = 0; b = 1; c=add(a, b); } void add(int a, int b) { a = a + b; return a;

}

Slicing an SDG: Two phase algorithm


Proposed by Horwitz et al

• Pass1: From the slice point:

• Traverse backward along all edges except parameter-out edges

• Mark the reached vertices

• Pass 2: From vertices marked in Pass 1

• Traverse backwards along all edges:

• Except call and parameter-in edges

Slicing an SDG: Pass-1




}

entry main

a = 0

b = 1 add(a, b)

entry add

a =ain

b = bin

a = a+b

aout = a

c = aout

bin = b

ain = a

Slice Point

Except parameter-out edges

Slicing an SDG: Pass-2




}

entry main

a = 0

b = 1 add(a, b)

entry add

a =ain

b = bin

a = a+b

aout = a

c = aout

bin = b

ain = a

Slice Point

Except call and parameter-in edges

Dynamic Slicing: Example cont…


Consider the following program:

Computing: product of odd (p), sum of even (s)

Execution Trace, N=3 1,2,3,4 //intial 5,6,8,9 //i=1 5,6,7,9 //i=2 5,10 //i=3, end Which part of the program is responsible for computing sum?



We need to compute a backward slice with slicing criterion <(10, s)>

Dependencies: Dynamic Data Dependence: which variable assignment was propagated to the value of `s' ? Dynamic Control Dependence: which are the conditional branches that executed till line 10 and in what order?



For N=3

The Dynamic Slice w.r.t. (10,s)


Slicing Tools


• Unravel : Static slicing tool for ANSI C

• WET : Whole Execution Traces, dynamic slicing

• CodeSurfer : Commercial static slicing tool for C

• Performs data flow and control dependence analysis

• Indus : Static slicer for Java

• Available for Eclipse via Kaveri plugin

• JSlice : Dynamic slicing tool for Java

• SPYDER: Debugging tool

• Performs both static and dynamic slice

Further Reduction in Size of Slice…


Amorphous Slice:

No Syntax Preservation

Retaining the semantic property

Applications:

Re-engineering

Component Re-use

Program Comprehension

Testing & Maintenance

Program Integration

Amorphous Slice: Example


for(i = 0,sum = a[0], biggest = sum; i<19; sum = a[++i])

if (a[i+1] > biggest)

{

biggest = a[i+1];

average = sum/20;

}

Amorphous slice

for(i = 1, biggest = a[0]; i<20; ++i)

{

if (a[i]>biggest)

biggest = a[i];

}

Traditional slice

for(i = 0,sum = a[0], biggest = sum; i<19;

sum = a[++i])

if (a[i+1] > biggest)

{

biggest = a[i+1];

}

Slicing Criterion

Loop unrolling

Challenges: For further reading…


Slicing Object-Oriented Programs

Effect of Exceptions on Control Flow

Slicing UML Models

Slicing of threaded programs

Slicing Concurrent and Distributed Programs

Directions of Research


A Novel Fault Localization Technique

References

[1] M Weiser. 1984. “Program slicing” IEEE Transactions on Software Engineering 10(4):352–57

[2] F. Tip. 1995. A survey of program slicing techniques. Journal of Programming Languages 3(3): 121–89

[3] D. P.Mohapatra, R.Mall, and R. Kumar. 2006. “ An overview of slicing techniques for object-oriented programs” Informatica 30(2):253–77

[4] G. B.Mund, R.Mall, and S. Sarkar. 2003. “Computation of intra-procedural dynamic program slices. Information and Software Technology 45(8):499–512.



Questions



