acapulco-spl / acapulco

Eclipse-based tool for optimal configuration of software product lines

Acapulco

Acapulco is a tool to support Product Line configuration using genetic algorithms and consistency-preserving configuration operators.

Installation

For the moment Acapulco is to be launched from an Eclipse workspace and not as an update site or similar.

Requirements

• Java JDK 13+
• Eclipse 2020 (4.18.0)+

In your eclipse installation, install

• FeatureIDE framework for feature-oriented software development. Tested with version 3.8.1 and 3.9.3.
• Xtend framework. Some acapulco source code uses Xtend instead of Java, so once you will import the acapulco plugins in the workspace, the automatic build of the source code will create some needed Java classes in the xtend-gen package of the acapulco project. Tested with version 2.31

Clone or download the Eclipse Henshin repository Eclipse Henshin. Import the plugin org.eclipse.emf.henshin.model in the Eclipse workspace. File -> Import -> General -> Projects from Folder or Archive (import only the project and not the whole folder as a project). Tested with version 1.9.0.

Clone or download this repository. Import the acapulco plugins (plugins folder of this repository) in the Eclipse workspace. File -> Import -> General -> Projects from Folder or Archive (import only the project and not the whole folder as a project). Then right click the plugin and select Run As -> Eclipse Application. A new instance of Eclipse will be opened with the Acapulco functionality.

Getting started

• Right click a feature model (in the Package Explorer or Project Explorer) and click on Acapulco prepare. A set of resources will be generated with a message about how to perform the Acapulco launch. When launching, a window to configure the parameters for the optimization will appear. The pareto front configurations and a CSV file with their objectives' values will be generated at the end of the process.

The video below shows a demo of Acapulco.

Acapulco.Video.mp4

Optimization objectives and extensibility

• Objectives for the optimization can be contributed through the acapulco.objective extension point.
• Predefined generic objectives already included are: Total sum of numeric feature attributes in a CSV file, Number of features in the configuration, or Number of generated files after deriving the configuration through the Product Line.

Other resources

• A comparison of a version of Acapulco with other two tools is available in a replication package. The first version of Acapulco is a refactoring of this other repository.

acapulco's Issues

Provide a Java API to easily access the tool functionality programmatically

In Acapulco Preparation, FeatureIDE functionality can be used to handle Complex Constraints

Complex constraints are removed in Acapulco Preparation. However, in some cases, complex constraints can be transformed in simple constraints. FeatureIDE functionality can be reused for this. In Feature IDE UI, when you right click a feature model, you can...

• create a Product-Equivalent Model without Complex Constraints. The internal class is de.ovgu.featureide.fm.core.conversion.ComplexConstraintConverter.
• Export Feature Dependency Information that uses the class de.ovgu.featureide.fm.core.analysis.mig.ModalImplicationGraph. An example of usage to get implies and excludes relations can be found in de.ovgu.featureide.visualisation.ShowFeatureRelationsGraphCommandHandler . Related publication: Krieter et al. ICSE 2018 Propagating Configuration Decisions with Modal Implication Graphs.

If the user does not want to transform complex constraints, an evolution objective can be to minimize non-satisfied constraints

The user will need to configure the objective by providing the path to a feature model that includes these complex constraints (right now complex constraints are removed during Acapulco preparation).

Provide a Command-Line Interface

As suggested by SPLC 2022 reviewer 1, "CLI command for the tool can help developers to easily integrate the tool with others"

Create Tests plugin fragment with the tests we had in the repository acapulco_replication_package

Incorrect generated CPCOs for one case?

If we have this model

namespace MySPL

features
	MySPL {abstract}	
		optional
			A
			B
			C

constraints
	A & B => C

the Knuppel method to convert complex constraints obtain this model

namespace MySPL

features
	MySPL	
		optional
			A
			B
			C

		mandatory
			Subtree	
				mandatory
					OR0_0	
						or
							A_1_1
							B_2_1
							C_3_1

constraints
	!A_1_1 | !A
	!B_2_1 | !B
	!C_3_1 | C

Then, once the CPCOs are generated, these rules seem wrong as not(A) or not(B) is an excludes relation

Rule Act_A_1_1
Node A_1_1: false->true

and

Rule Act_A
Node A: false->true
Node A_1_1: true

It seems that the CPCO generation is interpreting !A_1_1 | !A as not(A) or A_1_1 which is A implies A_1_1. The same happens for the second constraint, while the rules for the features of the third seems correct.

Create a basic visualization for the pareto-front in case of multi-objective