CSE Colloquium

1
CSE Colloquium

• Frank Tsui
• Software Engineering

2
4 Topics

• Software Cohesion and Coupling
• Software Style and Metrics
• Software Configuration Management
• Software Process and Communications

3
1. Software Cohesion and Coupling

• Cohesion and Coupling are attributes of software
• Conjectured and shown to be associated with
• software quality and
• maintainability

• Cohesion addresses the relatedness within a
  module or Class.
• Coupling addresses the relatedness among
  modules or Classes

4
High Cohesion and Low Coupling
Research
predictor
numerically measurable
philosophical conceptual

• If we consider a software system, the entities
• in the system (e.g. for a Class, methods and
• attributes form the entities of a Class) are
  evaluated
• syntactically.
• The number of attributes shared or the number of
• methods evoked with in a class formulate the
• basis for the cohesion metric of a Class.
• The number of attributes passed or the number
• of evocation of methods from one class to
  another
• class formulate the basis for Class coupling
  metric

5
Example of a simple Class Cohesion Metric
Instance variables in a class
5
1
2
3
4
M1
M2
M3
methods in a class
Let ivj instance variable j
N S ivj sivk
shared instance variable k M S
sivk then Class
Cohesion M/N
Many variants of this simple Class Cohesion
metrics exist today.
6
2. Software Style

• Research
• Is Style a software attribute ?
• Style is still a philosophical and conceptual
  discussion

• There is no clear concept or accepted definition
• for Style of software.
• It exists in other disciplines
• Music
• Architecture
• Clothes fashion
• Etc.

7
Example from Kernighan and Plaugeron identity
matrix
for (int i 1 i n i ) for (
int j 1 j n j ) M ( i,j ) 0
M( i,i ) 1
for (int i 1 i n i ) for ( int j
1 j n j ) M ( i,j ) ( i/j ) (
j/I )
8
Consider two more sample code for answering the
sum of the first n integers
Using Recursion Sum ( n ) if ( n 1 )
sum 1 else sum n sum ( n 1 )
return sum
Using for Loop Sum ( n ) sum 0 for (
int z 1 z lt n z ) sum sum z
What do you think about these two in terms of
- cohesion and coupling ? - control
structure? - algorithm ? - data ? -
and STYLE ?
9
Example from Pfleeger and Atlee on Tax Problem
Develop a software program that will compute tax
according to the following set of rules
0 tax if income is zero .
10 tax for the first 10,000 or less
12 tax for next 10,000 above 10,000 OR
(10,001 to 20,000)
15 tax for next 10,000 above 20,000 OR
(20,001 to 30,000)
18 tax for next 10,000 above 30,000 OR
(30,001 to 40,000)
20 tax for above 40,000
10
tax 0 if (taxable_income 0) go to EXIT if
(taxable_income gt 10000) tax tax 1000
else tax tax .10 taxable_income
goto EXIT if (taxable_income gt
20000) tax tax 1200 else tax
tax .12 (taxable_income 10000) goto
EXIT if (taxable_income gt 30000) tax
tax 1500 else tax tax .15
(taxable_income 20000) goto EXIT
if (taxable_income 40000) tax
tax .18 (taxable_income 30000) goto
EXIT else tax tax
1800 .20 (taxable_income 40000) EXIT
10 tax for the first 10,000 or less
12 tax for 10,001 to 20,000
15 tax for 20,001 to 30,000
18 tax for 30,001 to 39,999
20 tax for above 40,000
11
for (int i2 level 1 i lt 5 i )
If (taxable_income gt bracket i )
level level 1 else tax
base level percent level
(taxable_income bracket level )
i 6

• Tax table showing the bracket, base, and the
  percent columns
• bracket base percent
• 0 0 .10
• 1000 .12
• 2200 .15
• 3700 .18
• 40000 5500 .20

Code structure is different because the algorithm
is dependent on a table of Information, which may
be implemented with 3 arrays
e.g. float percent .10, .12, .15, .18,
.20 and assume array indexing start with 1
12
Computational form vs role

• Consider someone asking you to add the first two
  integers
• You might say y 1 2
• Suppose we say add the first 5 integers you
  may still do the following
• y 1 2 3 4
  5
• Suppose we keep this up and ask you to add the
  first 50 integers!
• y 1 2
  ---------------------- 49 50
• How about first 500 integers? ---- you may pause
  and try the following
• set iterator 1
• compute set sum 0
• sum lt sum iterator
• If iterator 500, compute iterator lt (
  iterator 1) and repeat the above statement
• otherwise, stop

The role did not change ---- at some point we
changed the form ---
13
Inspirations of Previous form change ?

• Rephrasing of 18th century English poet, William
  Blakes stanza
• Can you see the world in a grain of sand?
• Can you imagine heaven in a wild flower?
• Can you grasp infinity in the palm of your hand?
• Can you experience eternity in an hour?

Are these different forms of the same thing?
Expressing infinite computation requires a form
change. There may be many forms ---- is style
defined by form ?
14
Application to general Iterator pattern

• The general iterator function (role) is to
  provide
• First ( )
• Next ( )
• Done ( )

• Easy and similar to the previous example of
  adding integers
• if the data type is integers.
• - We also know the algorithms for data expressed
  with tree structure, by
• specifying the traversal preference.
• What about a non-structured set? A, T, C, G,
  AAT, AATTCG do we
• have to have a mapping to (some ordinal scale)
  ? If so, does that
• mapping dictate the form of the iterator
  algorithm and thus the style?

15
Towards Polymorphism

• ADD (max, data dtype)
• This ADD function is parametric polymorphic in
  two ways
• Max parameter gives us the capability of ADD to
  perform the sum of any arbitrary number of items,
  not just the first 25 items or first 600 items,
  etc.
• Data parameter with potentially different data
  typing allows us to add different types of items.

Expanding the role in two dimensionalities,
required us to consider expanding the form
differently ---- via parameters.
16
Related to Polymorphism?

• Studying the various forms of design/coding leads
  us to some specific areas
• Varieties of polymorphism
• Are the varieties of polymorphism a matter of
• form or
• role
• Would the number and type of forms and roles
  be a good starting point for the metric of
  style? ------ for programs since all programs
  range from monomorphism to some degree of
  polymorphism.

The Component Based Design advocates will also
find these concepts closely related to
interfaces and services.
17
3. Software Configuration Management

• Software Configuration Management is the
  discipline of controlling the evolution of
  software.

• Has 2 main components
• Identifying Defining the artifacts
• that need to be controlled
• 2. Mechanism to control the artifacts

18
Configuration Management Research
Understanding and Controlling Relationships Amon
g Artifacts
Focused on the Control Mechanism -
versioning - conflicts - security
-access speed -etc.
Moving towards
19
Intra Inter Relationships
Requirements
Design
Code
Test Cases
general
general
3
general
general
general
1
general
general
general
general
French
2
4
French
French
French
French
Japanese
Japanese
Japanese
Japanese
Brazilian
Brazilian
Brazilian
Brazilian
20
Inter-Artifacts Relationship Matrix Example
21
4. Software Process and Communications

• It is conjectured that communications affects the
  success of software projects , especially with
  Global Software Development

Current Research
How to Relate ?

• Software Project
• good morale
• good product
• meets schedule
• meets cost

• Communications
• types
• structure
• amount

Future Research
Predict ?
22
Project Teams by Success and Amount of
Communications
100
95
Success
90
85
80
75
1 k
2 k
3 k
4 k
5 k
6 k
7 k
8 k
Communications in person-minutes
23
Distribution of Communications
39.7
40
30
of Total Comm.
26.5
23.1
20
10.7
10
Req.
Des/code
Test
Proj. Mgmt
24
Distribution of Communications by Groups
45
40
35
B
30
9 teams overall
of Total Comm.
25
20
15
A
10
C
5
Req.
Des/code
Test
Proj. Mgmt