[2] Linux Variability Analysis Tools
[3] SPLAR - Software Product Lines Automated Reasoning library
[4] Sat4j: the boolean satisfaction and optimization library for Java
[5] JCommander: a Java framework that makes it trivial to parse command line parameters
[6] SPL Covering Array Tool
About the feature models
- Browse or download (262KB) the feature models generated with SPLOT [1] and the valid pairs files used in the paper
- Browse or download (1.3MB) the real feature models [1], [2] and the valid pairs files used in the paper
- Browse or download (175MB) all the feature models generated with SPLOT [1] and the valid pairs files used in the paper
- Browse or download (224MB) all the real feature models [1], [2] and the valid pairs files used in the paper
About the source code and the implementation
- The implementation is licensed under the GNU General Public License 3.0 - Copyright 2012 University of Luxembourg - Interdisciplinary Centre for Security Reliability and Trust (SnT)
- The implementation uses the SPLAR [3], Sat4j [4] and JCommander [5] libraries. Other libraries which are dependencies of SPLAR are also present.
- A release version with a GUI is under development: PLEDGE
- Download (2.6Mo) the developed source code only (development / last version)
- Download (21KB) the developed source code only (old version)
- Download (1.1MB) the whole application (old version)
- There are two versions of the Sat4j [4] library which are present in the lib folder of the whole application to enable the unpredictable products selection from the solver. Such a behavior has been implemented as follows:
- Get Sat4j official release 2_0_5 source code via the SVN: svn checkout svn://svn.forge.objectweb.org/svnroot/sat4j/maven/tags/2_0_5
- Get Sat4j HEAD development version (trunk) via the SVN: svn checkout svn://svn.forge.objectweb.org/svnroot/sat4j/maven/trunk
- Replace Heap.java of the 2_0_5 version by the Heap.java from the trunk version
- Modify the line private static final Random rand = new Random(123456789); of 2_0_5/org.sat4j.core/src/main/java/org/sat4j/minisat/orders/RandomWalkDecorator.java into private static final Random rand = new Random(); (The fixed seed is replaced)
- Build the two versions with ant
- Link both the 2_0_5 and trunk versions to the application
- Define the random ordering in the application that uses the solver with ((Solver)reasonerSAT.getSolver()).setOrder(new RandomWalkDecorator(new VarOrderHeap(new RandomLiteralSelectionStrategy()),1));
About the use of the implementation (old version)
- Usage:
java -jar SPL.jar [options] [command] [command options] Commands: prio_splcat_products Computes the values of the prioritization techniques and fitness function using the set of valid products from SPLCAT. Usage: prio_splcat_products [options] Options: * -csv CSV file output from SPLCAT tool * -fm Feature model (SPLOT format) * -o Output directory -p Valid pairs file -runs Number of runs to average the random technique Default: 10 prio_solver_products Computes the values of the prioritization techniques and fitness function using a set of products from the solver. The number of products generated will be equal to #features/2 Usage: prio_solver_products [options] Options: * -fm Feature model (SPLOT format) * -o Output directory -p Valid pairs file -predictable Use the predictable products for the solver Default: false -runs Number of runs to average Default: 10 prio_splcat_solver_products Computes the values of the prioritization techniques and fitness function using the set of valid products from SPLCAT + additional products from the solver. Usage: prio_splcat_solver_products [options] Options: * -csv CSV file output from SPLCAT tool * -fm Feature model (SPLOT format) * -o Output directory -p Valid pairs file -predictable Use the predictable products for the solver Default: false -runs Number of runs to average Default: 10 average_data_files Compute the average between the data files specified in the directory Usage: average_data_files [options] Options: * -i Directory containing .dat files -noNorm Don't work on normalized data files. The minimum length data file will be used Default: false -o Output directory Default: Same as input normalize_data_files Compute the normalized version of the data files present specified in the directory Usage: normalize_data_files [options] Options: * -i Directory containing data file to normalize generate_prods_ga Generate products using the GA Usage: generate_prods_ga [options] Options: -csv CSV file output from SPLCAT tool. This argument is optional, see -nbProds option -dimacs Specify if the FM is a dimacs one Default: false * -fm Feature model (SPLOT format) -nbProds Number of products to generate. If negative, then the same number as the SPLCAT output will be used. In this case, the -csv option is required Default: 10 -noCoverage Do not compute the coverage (only generate the products) Default: false * -o Output directory -onlyGA Run only the GA approach Default: false -p Valid pairs file -runs Number of runs to average Default: 10 -timeAllowedMS Time allowed for the GA in ms Default: 60000 compute_valid_pairs Compute the valid pairs to a file for a XML FM Usage: compute_valid_pairs [options] Options: -dimacs Specify if the FM is a dimacs one Default: false * -fm Feature model (Dimacs or SPLOT format) -part Specify which part should be considered. Only if nbParts > 1 Default: 1 -parts Specify the number of parts Default: 1 compute_stats Compute some statistics for a FM Usage: compute_stats [options] Options: -dimacs Specify if the FM is a dimacs one Default: false * -fm Feature model (Dimacs or SPLOT format)
About the empirical study data
- The paper compares some results with SPLCAT [6]
- Browse or download (1.9MB) the SPLCAT [6] results on the generated FMs
- Browse or download (15KB) the SPLCAT [6] results on the real FMs
- Browse or download (5.6MB) the prioritization data results for the generated FMs (10 runs average)
- Browse or download (167KB) the prioritization data results for the real FMs (10 runs average)
- Browse or download (1MB) the search-based products generation data results for the generated FMs (10 runs average)
- Browse or download (81KB) the search-based products generation data results for the real FMs (10 runs average)
- Browse or download (43MB) the generated products files of the 4 big FMs for the 10 runs discussed in the paper
About the similarity heuristic
- Additional elements showing how the similarity distance and fitness function are relevant to appreciate the t-wise coverage of sets of products.