Combinatorial

Testing

03/08/10 1
1
 Combinatorial Testing: Motivation
• The behavior of a program may be affected by many factors:
– Input parameters, Combinatorial: Relating to,
or involving combinations
– Environment configurations (global variables),
– State variables. ..
• Equivalence partitioning of an input variable:
– Identify the possible types of input values requiring different
processing.
• If the factors are many:
– It is impractical to test all possible combinations of values of all factors.

2
 Combinatorial Testing: Motivation
• Many times, the specific action to be performed depends
on the value of a set of Boolean variable:
– Controller applications
– User interfaces

3
 Combinatorial Testing
• Several combinatorial testing strategies exist:
– Decision table-based testing

– Cause-effect graphing

– Pair-wise testing (reduced number of test cases)

4
• Applicable to requirements involving conditional actions.
• This is represented as a decision table: Decision table-
–Conditions = inputs based Testing
–Actions = outputs (DTT)
–Rules =test cases Condition1
Rule1
Yes
Rule2
Yes
Rule3
No
Rule4
No

• Assume independence of inputs Condition2

Condition3
Yes
No
X
Yes
No
No
X
X

• Example Condition4
Action1
No
Yes
Yes
Yes
No
No
Yes
No

–If c1 AND c2 OR c3 then A1 Action2

Action3
No
No
No
No
Yes
No
No
Yes

5
 Combinations

Rule1 Rule2 Rule3 Rule4

Condition1 Yes Yes No No

Condition2 Yes X No X
Conditions
Condition3 No Yes No X

Condition4 No Yes No Yes

Action1 Yes Yes No No

Actions Action2 No No Yes No

Action3 No No No Yes

6
• A decision table consists of a number of columns r1 r2 … rn
(rules) that comprise all test situations C1 0 1 0
• Example: the triangle problem
c2 - 1 0
– C1: a, b,c form a triangle
– C2: a=b C3 - 1 1

– C3: a= c C4 - 1 0
– C4: b= c a1 1 0 0
– A1: Not a triangle
– A2:scalene
Sample a2 0 0 1

– A3: Isosceles Decision a3 0 0 0

– A4:equilateral table a4 0 1 0
– A5: Right angled a5 0 0

7
 Test cases from Decision Tables
C1: a, b,c form
Test Case
ID a b c Expected
output a triangle
C2: a=b
TC1 4 1 2 Not a
Triangle
C3: a= c
C4: b= c
TC2 2888 2888 2888 Equilateral
TC3 ? | ) Impossible
TC4
…

TC11

8
 Conditions
C1: a < b+c? F T T T T T T T T T T
C2: b < a+c? - F T T T T T T T T T
C3: c < a+b? - - F T T T T T T T T
C4: a=b? - - - T T T T F F F F
More Complete C5: a=c? - - - T T F F T T F F
Decision Table C6: b-c? - - - T F T F T F T F
for the Triangle Actions

Problem A1: Not a Triangle X X X

A2: Scalene X
A3: Isosceles X X X
A4: Equilateral X
A5: Impossible X X X

9
 Case ID a b c Expected
Output
DT1 4 1 2 Not a
Triangle
DT2 1 4 2 Not a
Test Cases for Triangle
the Triangle DT3 1 2 4 Not a
Triangle
Problem
DT4 5 5 5 Equilateral

DT5 ? ? ? Impossible

DT6 ? ? ? Impossible

DT7 2 2 3 Isosceles

DT8 ? ? ? Impossible

DT9 2 3 2 Isosceles

DT10 3 2 2 Isosceles

DT11 3 4 5 Scalene 10
 Printer does not print Y Y Y Y N N N N
Decision Table
Conditions A red light is flashing
– Example 2 Y Y N N Y Y N N

Printer is unrecognized Y N Y N Y N Y N

Check the power cable X

Check the printer-computer cable

X X
Printer Actions
Troubleshooting Ensure printer software is installed
X X X X

Check/replace ink X X X X

Check for paper jam X X

11
 Quiz: Develop BB Test Cases
• Policy for charging customers for certain in-flight services:

If the flight is more than half-full and ticket cost is more than
Rs. 3000, free meals are served unless it is a domestic flight.
Otherwise, no meals are served. Meals are charged on all
domestic flights.

12
 Fill all POSSIBLE COMBINATIONS
combinations more than half-
N N N N Y Y Y Y
in the table. full

CONDITONS
more than
Rs.3000 per N N Y Y N N Y Y
seat

domestic flight N Y N Y N Y N Y
ACTIONS

13
 POSSIBLE COMBINATIONS
Analyze
column by more than half-full N N N N Y Y Y Y
column to
determine

CONDITONS
which actions more than Rs. 3000
N N Y Y N N Y Y
per seat
are appropriate
for each
combination domestic flight N Y N Y N Y N Y

Y Y Y Y
ACTIONS

serve meals

free Y
14
 POSSIBLE COMBINATIONS
Reduce the more than half-full N N N N Y Y Y Y
table by

CONDITONS
eliminating more than Rs. 3000
N N Y Y N N Y Y
redundant per seat
columns.
domestic flight N Y N Y N Y N Y
ACTIONS

serve meals X X X X

free X

15
 Final Combinations
solution more than half-full N Y Y Y

CONDITONS
more than 3000 per seat - N Y Y

ACTIONS domestic flight - - N Y

serve meals X X X

free X

16
 –Rules need to be complete:
Assumptions •That is, every combination of decision table
regarding
values including default combinations are
rules
present.
–Rules need to be consistent:
•That is, there is no two different actions for the
same combinations of conditions

17
 Guidelines and Observations
• Decision table testing is appropriate for programs:
– There is a lot of decision making
– Output is a logical relationship among input variables
– Results depend on calculations involving subsets of inputs
– There are cause and effect relationships between input and
output

• Decision tables do not scale up very well

18
 Quiz: Design test Cases
• Customers on a e-commerce site get following
discount:
– A member gets 10% discount for purchases lower than
Rs. 2000, else 15% discount
– Purchase using SBI card fetches 5% discount
– If the purchase amount after all discounts exceeds Rs.
2000/- then shipping is free.
19
 Cause-effect Graphs
• Overview:
–Explores combinations of possible inputs
–Specific combination of inputs (causes) results in outputs
(effects)
–Represented as nodes of a cause effect graph
–The graph also includes constraints and a number of
intermediate nodes linking causes and effects
20
• If depositing less than Rs. 1 Lakh, rate of interest:
– 6% for deposit upto 1 year Cause-Effect
– 7% for deposit over 1 year but less than 3 yrs Graph
– 8% for deposit 3 years and above Example

• If depositing more than Rs. 1 Lakh, rate of interest:

– 7% for deposit upto 1 year
– 8% for deposit over 1 year but less than 3 yrs
– 9% for deposit 3 years and above

21
 Cause-Effect Graph Example
Causes Effects
C1: Deposit<1yr e1: Rate 6%

C2: 1yr<Deposit<3yrs e2: Rate 7%

C3: Deposit>3yrs e3: Rate 8%

C4:Deposit <1 Lakh e4: Rate 9%

C5: Deposit >=1Lakh

22
c4 e10 e1

c5 e2
e20
Cause-Effect
c1 e30 e3 Graphing
c2 e40
e4
c3 e50

e60
23
 Develop a Decision Table
C1 C2 C3 C4 C5 e1 e2 e3 e4
1 0 0 1 0 1 0 0 0
1 0 0 0 1 0 1 0 0
0 1 0 1 0 0 1 0 0
0 1 0 0 1 1 0 1 0
• Convert each row to a test case

24
 Pair-wise Testing

03/08/10 25
25
Combinatorial Testing of User
Interface

0 = effect off
1 = effect on

210 = 1,024 tests for all combinations

* 10 3 = 1024 * 1000 …. Just too many to tests

26
26
 Combinatorial Testing Problem
X1 X2 X3 . . . Xn

System S

•Combinatorial testing problems generally follow a simple input-process-

output model;
•The “state” of the system is not the focus of combinatorial testing.

27
27
• Instead of testing all possible combinations:
– A subset of combinations is generated. t-way Testing
• Key observation:
– It is often the case that a fault is caused by interactions among a
few factors.
• t-way testing can dramatically reduce the number of test
cases:
– but remains effective in terms of fault detection.
28
28
 t-way Interaction Testing
Interest Rate | Amount | Months | Down Pmt | Pmt Frequency

All combinations: etc. . .

every value of
every parameters
.
All pairs: every
value of each
pair of
parameters

t-way interactions:
every value of every t-
way combination of
parameters

29
 Pairwise Testing
Pressure | Temperature | Velocity | Acceleration | Air Density

Α Τ1 1 10 1.1
Β Τ2 2 0 2.1
Τ3 3 20 3.1
Pressure Temperature 4 0
5
A T1 6
A T2
A T3
B T1
B T2
B T3

10/3/2018 30
 Pairwise Reductions
Number of Number of Number of Size of pair
inputs selected test combinations wise test
data values set

7 2 128 8

13 3 1.6 x 106 15

40 3 1.2 x 1019 21

31 31
• A t-way interaction fault: Fault-Model

–Triggered by a certain combination of t input values.

– A simple fault is a 1-way fault

– Pairwise fault is a t-way fault where t = 2.

• In practice, a majority of software faults consist of simple

and pairwise faults.

32
 Single-mode Bugs
• The simplest bugs are single-mode faults:
–Occur when one option causes a problem regardless of
the other settings

–Example: A printout is always gets smeared when you

choose the duplex option in the print dialog box
• Regardless of the printer or the other selected options

33
 Double-mode Faults
• Double-mode faults
–Occurs when two options are combined
–Example: The printout is smeared only when duplex is
selected and the printer selected is model 394

34
34
 Multi-mode Faults
• Multi-mode faults
–Occur when three or more settings produce the bug

–This is the type of problems that make complete coverage

necessary

35
• begin
– int x, y, z;
Example of Pairwise Fault
– input (x, y, z);
– if (x == x1 and y == y2)
• output (f(x, y, z));

– else if (x == x2 and y == y1)

• output (g(x, y));

– Else // Missing (x == x2 and y == y1) f(x, y, z) – g(x, y);

• output (f(x, y, z) + g(x, y))

• end
• Expected: x = x1 and y = y1 => f(x, y, z) – g(x, y);
x = x2, y = y2 => f(x, y, z) + g(x, y)

36
36
 HARDKEYBOARDHIDDEN_NO ORIENTATION_LANDSCAPE
Example: HARDKEYBOARDHIDDEN_UNDEFINED
HARDKEYBOARDHIDDEN_YES
ORIENTATION_PORTRAIT
ORIENTATION_SQUARE

Android smart KEYBOARDHIDDEN_NO

ORIENTATION_UNDEFINED

KEYBOARDHIDDEN_UNDEFINED
phone testing KEYBOARDHIDDEN_YES
SCREENLAYOUT_LONG_MASK
SCREENLAYOUT_LONG_NO
SCREENLAYOUT_LONG_UNDEFINED
• Apps should work on all KEYBOARD_12KEY
combinations of platform options, KEYBOARD_NOKEYS
SCREENLAYOUT_LONG_YES
but there are 3 x 3 x 4 x 3 x 5 x 4 x KEYBOARD_QWERTY
KEYBOARD_UNDEFINED
4 x 5 x 4 = 172,800 configurations SCREENLAYOUT_SIZE_LARGE
NAVIGATIONHIDDEN_NO SCREENLAYOUT_SIZE_MASK
NAVIGATIONHIDDEN_UNDEFINED SCREENLAYOUT_SIZE_NORMAL

NAVIGATIONHIDDEN_YES SCREENLAYOUT_SIZE_SMALL
SCREENLAYOUT_SIZE_UNDEFINED
NAVIGATION_DPAD
NAVIGATION_NONAV TOUCHSCREEN_FINGER
NAVIGATION_TRACKBALL TOUCHSCREEN_NOTOUCH
NAVIGATION_UNDEFINED TOUCHSCREEN_STYLUS
NAVIGATION_WHEEL TOUCHSCREEN_UNDEFINED

37
White-Box Testing

03/08/10 38
 What is White-box Testing?
• White-box test cases designed based on:

–Code structure of program.

–White-box testing is also called structural testing.

 White-Box Testing Strategies
• Coverage-based:
– Design test cases to cover certain program elements.

• Fault-based:
– Design test cases to expose some category of faults
• Several white-box testing strategies have become very popular :
– Statement coverage
– Branch coverage
– Path coverage White-Box Testing
– Condition coverage
– MC/DC coverage
– Mutation testing
– Data flow-based testing
 Why Both BB and WB Testing?
Black-box White-box
• Impossible to write a test case • Does not address the question
for every possible set of inputs of whether a program matches
and outputs the specification
• Some code parts may not be
reachable • Does not tell if all functionalities
have been implemented
• Does not tell if extra
functionality has been • Does not uncover any missing
implemented. program logic
 Coverage-Based Testing Versus Fault-Based Testing
• Idea behind coverage-based testing:
– Design test cases so that certain program elements are executed
(or covered).
– Example: statement coverage, path coverage, etc.
• Idea behind fault-based testing:
– Design test cases that focus on discovering certain types of
faults.
– Example: Mutation testing.
• Statement: each statement executed at least once
• Branch: each branch traversed (and every entry point
taken) at least once
• Condition: each condition True at least once and False at
least once
• Multiple Condition: All combination of Condition covered
• Path:
Types of program element Coverage
• Dependency:
 Stronger and
Coverage Weaker Testing

Stronger
Weaker
 Complementary Testing

Coverage
Strategy
1 Strategy
2
Stronger, Weaker, and Complementary Testing

Complementary

Weaker
 Statement Coverage
• Statement coverage strategy:

–Design test cases so that every statement in the

program is executed at least once.
 Statement Coverage
• The principal idea:
–Unless a statement is executed,

–We have no way of knowing if an error exists in

that statement.
 Statement Coverage Criterion
• However, observing that a statement behaves properly
for one input value:

–No guarantee that it will behave correctly for all input

values!
 Statement Coverage
• Coverage measurement:
# executed statements
# statements

• Rationale: a fault in a statement can only be

revealed by executing the faulty statement
• int f1(int x, int y){ Example
• 1 while (x != y){
• 2 if (x>y) then
•3 x=x-y;
• 4 else y=y-x;
• 5}
• 6 return x; }
int f1(int x,int y){
Example
1 while (x != y){
2 if (x>y) then
3 x=x-y;
4 else y=y-x;
5 }
6 return x; }
 Euclid's GCD Algorithm

• By choosing the test set {(x=3,y=3),(x=4,y=3),

(x=3,y=4)}

–All statements are executed at least once.

 Branch Coverage
• Also called decision coverage.
• Test cases are designed such that:

–Each branch condition

•Assumes true as well as false value.
int f1(int x,int y){
Example
1 while (x != y){
2 if (x>y) then
3 x=x-y;
4 else y=y-x;
5 }
6 return x; }
 Example
• Test cases for branch coverage can be:

• {(x=3,y=3),(x=3,y=2), (x=4,y=3), (x=3,y=4)}

 Branch Testing
• Adequacy criterion: Each branch (edge in the CFG)
must be executed at least once
• Coverage:
# executed branches
# branches
 Quiz 1: Branch and Statement Coverage:
Which is Stronger?
• Branch testing guarantees statement
coverage:

–A stronger testing compared to the statement

coverage-based testing.
 Stronger Testing
• Stronger testing:
–Superset of weaker testing
–A stronger testing covers all the elements covered by a
weaker testing.
–Covers some additional elements not covered by
weaker testing
 Sample
Coverage
Report
 Statements vs Branch Testing
• Traversing all edges of a graph causes all nodes to be visited

– So a test suite that satisfies branch adequacy criterion also satisfies

statement adequacy criterion for the same program.

• The converse is not true:

– A statement-adequate (or node-adequate) test suite may not be branch-

adequate (edge-adequate).
– Statement coverage
– Branch coverage (aka decision coverage) White-box
– Basic condition coverage Testing
– Condition/Decision coverage
– Multiple condition coverage
– MC/DC coverage
– Path coverage
– Data flow-based testing
– Mutation testing

64
 All Branches can still miss testing specific conditions
• Assume failure occurs when c==DIGIT

if((c == ALPHABET) || (c ==DIGIT))

• Branch adequacy criterion can be satisfied by c==alphabet and

c==splchar

– The faulty sub-expression might not be tested!

– Even though we test both outcomes of the branch

 Basic Condition Coverage
• Also called condition coverage or simple condition
coverage .

• Test case design: ((c == ALPHABET) || (c== DIGIT))

–Each component of a composite conditional expression

• Made to assume both true and false values.
 Basic Condition Testing
• Simple or (basic) Condition Testing:
– Test cases make each atomic condition assume T and F values
– Example: if (a>10 && b<50)
• Following test inputs would achieve basic condition coverage
– a=15, b=30
– a=5, b=60
• Does basic condition coverage subsume decision coverage?
 Example: BCC
• Consider the conditional expression
–((c1.and.c2).or.c3):

• Each of c1, c2, and c3 is exercised with all possible

values,
–That is, given true and false values.
 Basic condition testing
• Adequacy criterion: each basic condition must be
executed at least once
• Coverage:
# truth values taken by all basic conditions
2 * # basic conditions
 Is BCC Stronger than Decision Coverage?
• Consider the conditional statement:
–If(((a>5).and.(b<3)).or.(c==0)) a=10;

• Two test cases can achieve basic condition coverage: (a=10,

b=2, c=2) and (a=1, b=10, c=0)

• BCC does not imply Decision coverage and vice versa

 Condition/Decision Coverage Testing
• Condition/decision coverage:
– Each atomic condition made to assume both T and F values
– Decisions are also made to get T an F values

• Multiple condition coverage (MCC):

– Atomic conditions made to assume all possible combinations of
truth values
 MCC
• Test cases make Conditions to assume all possible
combinations of truth values.
• Consider: if (a || b && c) then …
Test a b c
(1) T T T
(2) T T F Exponential in the
(3) T F T
(4) T F F number of basic
(5) F T T
(6)
(7)
T
F
T
F
F
T
conditions
(8) F F F
 Multiple Condition Coverage (MCC)
• Consider a Boolean expression having n
components:
–For condition coverage we require 2n test cases.
• MCC testing technique:
–Practical only if n (the number of component conditions)
is small.
 MCC for Compound conditions: Exponential complexity
(((a || b) && c) || d) && e

Test a b c d e 25=32
Case
(1) T — T — T
(2) F T T — T •Short-circuit
(3) T — F T T
(4) F T F T T evaluation often
(5) F F — T T
(6) T — T — F reduces number of
(7) F T T — F
(8) T — F T F test cases to a more
(9)
(10)
F
F
T
F
F
—
T
T
F
F manageable
(11)
(12)
T
F
—
T
F
F
F
F
—
—
number, but not
(13) F F — F — always…
 Subsumption
Multiple Condition
• Condition testing:
–Stronger testing than branch Condition/Decision

testing.
Decision
• Branch testing:
–Stronger than statement
Basic Condition Statement
coverage testing.
 Shortcomings of Condition Testing
• Redundancy of test cases: Condition evaluation could be
compiler-dependent:
– Reason: Short circuit evaluation of conditions
• Coverage may be Unachievable: Possible dependencies
among variables:
– Example: ((chr==`A ́)||(chr==`E ́)) can not both be true at the
same time
 Short-circuit Evaluation
• if(a>30 && b<50)…

– If a>30 is FALSE compiler need not evaluate (b<50)

• Similarly, if(a>30 || b<50)…

– If a>30 is TRUE compiler need not evaluate (b<50)