Reverse Engineering Goal Models from Legacy Code

1
Reverse Engineering Goal Models from Legacy Code

• Yijun Yu1
• Yiqiao Wang1
• John Mylopoulos1
• Sotirios Liaskos1
• Alexei Lapouchnian1
• Julio Cesar Sampaio do Prado Leite2
• http//www.cs.toronto.edu/yijun/

2
1. Motivation

• In legacy software, the purpose (goals,
  requirements) of a software system is almost
  always lost in legacy code
• In the new era, open and dynamic systems require
  high-variability software
• So, we need to recover the purpose of the legacy
  software in order to adapt it for the new context

3
Put into reengineering perspective
Horseshoe Model
abstract
Reverse engineering
Forward engineering
Requirements
IntentionsTradeoffs
GOAL MODEL
GOAL MODEL
customizablearchitecture
SOA
Architectures
BEHAVIOUR MODEL
Web services
Functions
components
LEGACY CODE
Code

time
4
Hard goal model
5
Soft goal model
6
Put into reengineering perspective
Horseshoe Model
Reverse engineering
Forward engineering
Requirements
IntentionsTradeoffs
GOAL MODEL
GOAL MODEL
customizablearchitecture
SOA
Architectures
BEHAVIOUR MODEL
Web services
Functions
components
LEGACY CODE
Code

7
2. A tool supported reverse engineering
2.1
FPT compiler
FPT compiler
2.2
2.3
2.4
2.4
8
2.1 Software Refactoring

• Software refactoring Restructuring existing
  code by altering its internal structure without
  changing its external behaviour - Martin
  Fowler, also Comments signal the semantic gap
  between the code and the programmers purpose

entry
// the following does S S1(I1, O1) S2(I2,
O2) // other
// refactored S(I, O)
exit
I Variables defined before the entry of the
block O Variables defined in the block that will
be used after the exit
9
2.2 Statecharts refactoring

• Statecharts are used to bridge the semantics
  abstraction gap between source code and goal
  models
• refactored source code
• equivalent statechart
• higher-level statechart
• unstructured high-level program
• structured high-level program

10
2.3 Goal model from structured programs
11
2.4 Identifying NFR and softgoals

• Create a set of functional tests
• If removing a method does not break the
  functional test, then the goal associated with
  the method is a NFR
• If the identified NFR improves some quality
  attribute, then
• There is a soft goal for the quality attribute
• There is a contribution from the NFR to the
  softgoal

12
3. Case studies

• Case study selection criteria
• Email software systems as they are the target for
  personal RE
• Software popularity with a large user base, must
  support wide-range of requirements
• Open-source the validity of the findings can be
  verified
• Applicable with our tool support

13
3.1 The Columba case study

• 1 ColumbaLogger.createDefaultHandler()
• 2 registerCommandLineArguments()
• 3 // handle commandline parameters
• 4 if (handleCoreCommandLineParameters(args))
• 5 System.exit(0)
• ...

• Why Columba?
• it is an email system
• popular
• 140 KLoc
• Open-source
• Java
• Ver. 1.0RC2
• Whats New!
• Ver. 1.0RC3 has put one of our resulting NFR
  into comments

1 ColumbaLogger.createDefaultHandler() 2
registerCommandLineArguments() 3
handle_commandline_parameters(args) ...

• 1 // ColumbaLogger.createDefaultHandler()
• 2 registerCommandLineArguments()
• 3 handle_commandline_parameters(args)
• ...

• boolean maintainability_logging false
• ...
• 1 if (maintainability_logging)
• 2 ColumbaLogger.createDefaultHandler()
• 3 registerCommandLineArguments()
• ...

14
(No Transcript)
15
A running Columba
16
The core functional Columba
17
JUnit testing for Preserved Functionality
18
3.2 The Squirrel Mail case study

• Why Squirrel Mail?
• it is also an email system
• popular
• PHP HTML
• 70 KLoc
• Open-source
• unstructured
• Check this
• Its reliable, secure, easy to use
• but not that fast

19
Restructure AST Goal Graph into Stakeholder Goals
20
4. Conclusion

• A reverse engineering process is proposed to
  recover a goal model from legacy code.
• This is a critical step in reengineering legacy
  code to improve its quality through variability.
• Reverse Engineering research has focused almost
  exclusively on recovering design information
  (rather than purposes)
• Future work
• Need to further evaluate the effectiveness of
  heuristics used
• Need more experiments

21
Welcome to the 1st RETR workshop

• Reverse engineering to Requirements
• Collocated with WCRE05, November Pittsburg, USA
• http//www.cs.toronto.edu/km/retr