1. Formalizing and Analyzing Feature models in Alloy
Jing Sun
Computer Science
This might seem like an odd title for a talk by someone who claims to do metrics research, but hopefully it will become clear why I chose it.
Roughly what I’m going to do is discuss some of the metrics you might read about, and evaluate how useful they are. To do this, I’ll have to explain a bit of metrics theory, but hopefully not too much.

2. Outline Feature modeling The Alloy language
Formalizing feature diagram
Checking feature models
Conclusion and future work
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

3. Feature & Feature Modeling
Conceptual modeling perspective
an important property of a concept
allows to express the commonalities and differences between concept instance
Domain engineering context
represents reusable, configurable requirements
Feature modeling
is the activity of modeling the common and the variable properties of concepts and their inter-dependencies and organizing them into a coherent model
feature model diagram
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

4. Feature & Software Engineering
Feature-oriented programming (Prehofer, 1997)
Can programs be created by composing features?
Next generation of OO programming?
Feature and advanced separation of concern
Feature is regarded as one dimension of concern (Tarr et al., 1999)
Can we separate feature concerns from other concerns?
Feature and software evolution and maintenance
Locating features from source code (Eisenbarth, 2003)
Understanding the implementation of certain features in programs
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

5. Feature & Software Engineering
Feature and requirements engineering
Representing and structuring the problem domain as features
A well known problem: Feature interaction in telecommunication domain (Zave, 1993)
Feature and software process
Feature-driven development: an agile, highly adaptive, software development process (Palmer and Felsing, 2002)
Feature and software reuse
FORM (Feature-Oriented Reuse Method) (Kang et al., 1998)
Modeling a software product line’s commonalties and variability in terms of features
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

6. Our current interest
Analyzing features and their relationships, organizing them into feature model
Representing feature models
Formalizing feature models
Analyzing and verifying feature models
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

7. Feature modeling Basic relations among features Feature Types
One feature may depend on another feature
One feature may exclude another feature
Feature Types
Mandatory features: must be included in the description of a concept instance
Optional features: may be included in the description of a concept instance
Or features: may have any non-empty subset of the or-features included in the description of a concept instance
Alternative features: …
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

8. Feature diagram example
A car concept feature diagram
Car
Transmission
Engine
Automatic
Manual
Car Body
Pulls trailer
Electric
Gasoline
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

9. Applying formal methods
Alloy – a light-weight formal modeling language
Based on first order predicate logic
A subset of Z
Support automatic analysis
Applying Alloy to feature modeling
Formalizing a feature model
Checking a feature model (e.g., checking valid/invalid combination of features, semantic equivalence, …)
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

10. Feature language in Alloy
sig Feature {} // define a set of features
disj sig Concept extends Feature { // define a concept
holds : set Feature // all possible configuration of a concept }
// Depends - The selection of f2 depended on the selection of f1
fun Depends(c:Concept, f1:Feature, f2: Feature) {
f1 in c.holds => f2 in c.holds
// Alternative feature type
fun Alternative(c:Concept, pf:Feature, s:set Feature) {
all f : s | Excludes(f, s-f) && Depends(f, pf)
pf in c.holds => one f : s | Depends(pf, f)
} …
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

11. Feature model in Alloy open feature/FeatureModel
static disj sig Automatic extends Feature{}
static disj sig Manual extends Feature{}
static disj sig CarBody extends Feature{} …
static disj sig Car extends Concept{}
fact {Mandatory(Car, Car, CarBody+Transmission+Engine)}
fact {Optional(Car, Car, PullsTrailer)}
fact {Alternative(Car, Transmission, Automatic+Manual)}
fact {Or(Car, Engine, Electric+Gasoline)}
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

12. Checking & analyzing feature models
Checking valid/invalid feature configurations
assert correctness { all c : GPL | c.holds != GPL + Algorithm + Connected + Search + BFS + Graph + Undirected + Weighted }
Checking the source of unsatisfiability using “core extension”
Checking the solvability of feature models
Analyzing semantic equivalence of feature models
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

13. Checking & analyzing feature models
Semantic equivalence checking
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

14. Conclusion & future work
Defined a formal semantics for the feature modeling language in Alloy
Provide a precise and rigorous formal basis for the feature diagram
Identify basic relations in feature diagram and showed that other complex relations can be expressed as a combination of the two
Checking feature models using Alloy analyzer.
Dynamic feature model re-configurations
Feature interaction problem
My research interests are in improving programmer productivity. This requires being able to measure productivity, which in turns needs some way to evaluate the quality

15. Jing Sun Computer Science j.sun@cs.auckland.ac.nz
Questions?
Jing Sun
Computer Science
This might seem like an odd title for a talk by someone who claims to do metrics research, but hopefully it will become clear why I chose it.
Roughly what I’m going to do is discuss some of the metrics you might read about, and evaluate how useful they are. To do this, I’ll have to explain a bit of metrics theory, but hopefully not too much.