A Coding Scheme to Support Systematic Analysis of Software Comprehension

1
A Coding Scheme to Support Systematic Analysis of
Software Comprehension

• IEEE Transactions on Software Engineering
  July/August 1999
• Anneliese von Mayrhauser
• Stephen Lang
• Discussion/Facilitation David Weltman

2
Discussion Outline

• Paper Overview
• Protocol Analysis basis (with ICM) for this
  work
• A Flexible Expandable Coding System AFECS Model
• Overview
• Examples
• Analysis
• Summary/Conclusions

3
Paper Overview

• Model Cognitive Activity
• Software Maintenance but Generalizable
• A New/Better Coding Scheme for Representing
  Software Comprehension
• Standardized Framework
• Easier
• Consistent Results
• More Comprehensive
• Flexible
• Aggregation

4
A Flexible Expandable Coding Scheme - AFECS
Level 1
Level 2
New Coding Scheme Model. Figure 1, page 529.
5
Protocol Analysis
Problem
Segment
Verbage
Parsing
Segment
Segment
Coding Rules
Transcribing
Code generate a hypo
Code added a column
Time Consuming
Code prob. Identification

• Problem
• Verbage
• Parsing
• Coding
• Categorizing
• Analysis
• Conclusions

Code construct a new attribute
Code id a data sink
Inconsistent
Category 1
Category 2
Category 3
Non standard Aggregate Results
A n a l y s i s
6
AFECS Model Attributes

• Segments
• Levels (max 6) General to Specific
• Level 1- Mental Model
• Level 2 - Element
• Other Levels

7
A Flexible Expandable Coding SchemeMental Model
(Level 1) - General

• Program Model SYS
• Not Familiar, build a program model mental
  representation, control flow, model rep.
• Situation Model- SIT
• Select a module for content analysis, data flow
• Top-Down/Domain OPP
• Familiarity
• Function

Bottom up
Integrated Comp. Model
Opportunistic
8
A Flexible Expandable Coding SchemeElement
(Level 2) More Specific

• Goal gol
• What we are trying to accomplish
• Enhance this sort that we are familiar with
• Hypothesis hyp
• Claim, further classification, resolution
• Supporting Action act
• Working with info
• Asking/Answering ?s
• Planning

9
A Flexible Expandable Coding Scheme Verb (Level
3) More Specific

• Goal Actions
• not, sch, chd
• Hypothesis Actions
• gen, end
• Supporting Actions
• inf, ask, ans, pln

10
AFECS Model Code ExamplesEnhancement/Debugging
tasks

• SYS.act.inf
• Unfamiliar with code, highlight part of the code
  and manipulate that information maybe to help
  build a hypothesis
• Airline reservation system we need to enhance
  the pricing module, find out where part of it is,
  and mark it for working with later find another
  part
• OPP.hyp.gen
• We think we are familiar with part of the code
  and what it does, taking a educated guess,
  further classifying that guess
• Airline reservation system- Is this part looking
  for open seats, scanning variables/arrays that
  could represent seat assignments confirm or
  validate
• SYS.act.inf.cod (Fig. 3, page 536)
• Subject spent significant cognitive efforts using
  code as information sources to gain familiarity

11
AFECS Model - Analysis

• Use of Mental Models
• For certain problems, where are most of the
  subjects actions
• How long and frequency of switches between models
  (figs. 4 5, page 537)
• For certain problems how long was spent with
  testing familiarity (program understandability)
• For what type of situations will likely be a high
  probability of switching between domains
• Debugging, much switching between model rep.
  (SYS) and domain/function (OPP) and staying at
  model rep. (SYS)

12
Summary/Conclusions

• Devised a Coding Scheme for representing problem
  solving.
• Flexible
• Generalizable
• Tailorable
• Facilitates automation