Robust Lazy Conflict Detection via Multi-Conflict Extraction and Genetic Diversity Control

A Java tool that identifies conflicts in configuration knowledge bases and feature models using a genetic algorithm.

This repository shows the implementation and the evaluation of the Robust LCD algorithm, presented at the AAAI 2026 in the paper entitled Robust Lazy Conflict Detection via Multi-Conflict Extraction and Genetic Diversity Control [1]. The research community can fully exploit this repository to reproduce the work described in our paper.

Repository structure

• pom.xml — Maven project file (Java 21, assembly plugin builds the fat JAR)
• src/main/java/ — source code
  • Main class: at.tugraz.ist.ase.conflict.fm.GeneticConflictForFM
• src/test/java/ — tests (JUnit 5)
• conf/ — sample configuration files
• data/ — sample datasets and results
• lib/ — provided dependency JAR(s) installed during mvn validate
• test-scripts/ — Python evaluation scripts (see test-scripts/aaai/README.md)
• target/ — build outputs (created after packaging)
• Dockerfile — Docker build for JAR compilation
• Dockerfile.eval — Docker build for running evaluations
• .env.example — template for environment variables

Requirements and dependencies

• Language: Java (JDK 21) (set JAVA_HOME accordingly)
• Build/Package manager: Apache Maven 3.9+ (earlier versions may work, but were not verified)
• Packaging: JAR (fat JAR with dependencies via maven-assembly-plugin)
• Key dependencies:
  • at.tugraz.ist.ase.hiconfit:configurator (installed from local lib/ JAR during Maven validate)
  • io.github.cdimascio:dotenv-java (environment variable / .env file support)
  • ch.qos.logback:logback-classic (logging)
  • org.junit.jupiter:junit-jupiter (tests)

See pom.xml for full details.

Setup

This project ships a required library JAR in lib/ which is installed into your local Maven repository during the Maven validate phase. The JAR is from the HiConfiT project (source code).

1. Install the provided library (happens automatically on validate):

mvn validate

This will install lib/configurator-1.0.1-alpha-40.jar into your local Maven repository under the coordinates: at.tugraz.ist.ase.hiconfit:configurator:${hiconfit.version}.

2. Build the runnable JAR:

mvn clean package

Artifact is created under target/:

• gc_seeker-jar-with-dependencies.jar

Running

The application reads options from a configuration file. Use the -cfg option to provide a path.

java -jar target/gc_seeker-jar-with-dependencies.jar -cfg ./conf/gc_seeker.cfg

Notes:

• If Java preview features are required at runtime, add --enable-preview after java in the above commands.
• When no -cfg is provided, a default config path from code may be used (see ConfigManager.defaultConfigFile_GeneticConflictSeeker).

Running with Docker

This project provides two Dockerfiles for different use cases:

Dockerfile	Purpose	Image Size
Dockerfile	Build JAR files only	~800MB (Maven+JDK)
Dockerfile.eval	Run evaluations (JAR + Python)	~600MB (JRE+Python)

For Developers: Build JAR Files

Use Dockerfile to compile and extract JAR files without installing Java/Maven locally.

# Build the image
docker build -t gc-seeker-build .

# Extract JAR files to your local target/ directory
docker run --rm -v "$PWD/target":/output gc-seeker-build

Output:

JAR files copied to /output:
- gc_seeker-jar-with-dependencies.jar

For Researchers: Run Evaluations

Use Dockerfile.eval to run experiments with the full environment (Java + Python).

# Build the image
docker build -f Dockerfile.eval -t gc-seeker-eval .

Run the JAR directly:

docker run --rm \
  -v "$PWD/conf":/app/conf \
  -v "$PWD/data":/app/data \
  gc-seeker-eval \
  java -jar /app/target/gc_seeker-jar-with-dependencies.jar -cfg /app/conf/gc_seeker.cfg

Run Python evaluation scripts:

docker run --rm \
  -v "$PWD/test-scripts":/app/test-scripts \
  -v "$PWD/data/kb":/app/data/kb \
  gc-seeker-eval \
  python3 test-scripts/aaai/eval_script.py -cfg test-scripts/aaai/conf/busybox_eval_script.toml

Generate summary and visualizations:

# Summary statistics
docker run --rm \
  -v "$PWD/test-scripts":/app/test-scripts \
  gc-seeker-eval \
  python3 test-scripts/aaai/summary.py -cfg test-scripts/aaai/conf/summary_conf.toml

# Visualizations
docker run --rm \
  -v "$PWD/test-scripts":/app/test-scripts \
  gc-seeker-eval \
  python3 test-scripts/aaai/viz_aaai.py -cfg test-scripts/aaai/conf/viz_conf.toml

Reproducibility Tips

• Mount data/ and test-scripts/ so outputs persist on your host machine
• Keep experiment configs (TOML files) in version control
• Use the same Docker image version across all runs for consistency
• Results appear in test-scripts/**/results/ directories

Configuration

Sample configuration file: conf/gc_seeker.cfg

Common/observed keys (see code for authoritative list):

• Paths and input
  • nameKB and kbPath (FM): combined to locate the feature model file.
  • resultPath: file to write log/status messages during search.
  • allCSPath: output file collecting identified conflict sets.
  • allCSWithoutCFPath: output file collecting conflict sets without CF constraints.
  • existingCSPath: optional file with known conflicts to seed the run.
  • existingCSWithoutCFPath: optional file with known conflicts excluding CF constraints.
  • statisticsPath, summaryPath (FM/APM): CSV outputs for statistics and summary.
• Algorithm parameters
  • maxNumConflicts, cfInConflicts
  • populationSize, maxNumGenerations
  • noPreferenceProbability, mutationProbability, maxFeaturesInUR
  • stopAfterXTimesNoConflict
  • Legacy/advanced: limitParentsToResolved, extinctAfterXTimesNoConflict, stopAfterXExtinctions, weightedConflicts, weightedCrossover, weightedCrossoverFactor, avoidSameOriginalConflict
• Notifications (optional)
  • emailAfterEachConf (yes/no), emailAddress, emailPass

Environment Variable Overrides

Sensitive configuration values can be set via environment variables or a .env file instead of the config file:

Config Key	Environment Variable
emailAddress	GC_EMAIL_ADDRESS
emailPass	GC_EMAIL_PASS

Option 1: Using a .env file (recommended)

Copy the example file and fill in your values:

cp .env.example .env
# Edit .env with your values

Then run normally:

java -jar target/gc_seeker-jar-with-dependencies.jar -cfg ./conf/gc_seeker.cfg

Option 2: Using system environment variables

export GC_EMAIL_PASS="your-secret-password"
java -jar target/gc_seeker-jar-with-dependencies.jar -cfg ./conf/gc_seeker.cfg

Priority order: .env file > system environment variable > config file > default value.

Security note: Add .env to .gitignore to avoid committing sensitive values.

Scripts and examples

• See test-scripts/aaai/README.md for an overview of all experiment scripts, how to run them, and the directory layout.
• Detailed AAAI experiment workflow: test-scripts/aaai/README.md
• Example TOML config/scripts and sample results are under test-scripts/ (e.g., test-scripts/aaai/...). These are for experiments/evaluations and not wired into Maven.
• Example data inputs/outputs are under data/.

License

This project is licensed under the MIT License — see LICENSE for details.

References

[1] V.M. Le, L. A. Feldgrill, A. Felfernig. Robust Lazy Conflict Detection via Multi-Conflict Extraction and Genetic Diversity Control. In 40th AAAI Conference on Artificial Intelligence. AAAI’26, Singapore. 2026. [TBD]