SPLC2020-FeatureRevisionLocation

Artifacts for the paper "Locating Feature Revisions in Software Systems Evolving in Space and Time" submitted as full paper to the research track of SPLC 2020.

Each target system folder contains:

• the ground truth variants used as input for our feature revision location technique Input_variants.
• the ground truth variants with random configurations Random_variants.
• the results Retrieved_results containing csv file for each ground truth variant comparison (compare_input_variants and compare_random_variants) and the composed variants by the traces computed (input_config_composed_variants and random_config_composed_variants).
• an illustration of the introduced and changed features over the commits analyzed Evolution_in_space_and_time.
• a file configurations.csv containing the configuration of input variants.
• a file features_report_each_project_commit.csv containing the number of features introduces and changed for each commit analyzed.
• a file randomconfigurations.csv containing the configuration of random variants.
• a file runtime.csv containing the recorded runtime to extract traces for each input variant in milliseconds.

The implementation of the results shown in this paper is available in two repositories:

• ECCO repository contains the feature revision location technique.
• git-ecco repository contains the implementation for mining ground truth variants.

To reproduce the results of this paper you can clone this repository (which contains the artifacts used and the results presented in the paper) to your local directory and the git-ecco repository, which contains the ECCO repository as a lib. Then you are able to run the process to create the ground truth variants and the process of feature revision location.

The steps to reproduce the results of the feature revision location are explained below:

• It is necessary to use the input variants (which are in the folder Input_variants) and the configurations.csv and runtime.csv files inside each target system folder (LibSSH, Marlin, SQLite). In case of reproducing the results of the random configurations you must use as input variants the variants from the folder Random_variants and the configuration of the file randomconfigurations.csv. The input variants are the ground truth variants used in this paper. However, if you desire to create variants as a ground truth instead of using the ones available in this repository, you can follow the steps in the git-ecco repository.
• In the git-ecco repository repository there is a module translation with a class FeatureRevisionLocationTest.java, which contains four necessary tests to obtain the results of this paper (note: you probably will need to edit the VM options in each test configuration, suggested VM options: "-Xss1000m -Xmx10g"):
  • TestEccoCommit: the test for the feature revision location.
  • TestEccoCheckout: the test for composing the variants with the mapping between feature revisions and artifacts.
  • TestCompareVariants the test for comparison of the ground truth variants with the composed variants.
  • GetCSVInformationTotalTest the test for computing the metrics to compute the results.
• For each target system folder (LibSSH, Marlin, SQLite) extract the files of the folder Input_variants (if it is compacted) and uses its directory string in the variable resultsCSVs_path.
• Use the string of the directory containing the file configurations.csv in the variable configuration_path.
• In the same folder of the target system create a new folder (named as variant_results) and use its directory string in the variable resultMetrics_path. In this new folder (variant_results) also create a new folder named as checkout.
• Each directory containing the folder of the artifacts must be defined in the variables corresponding to them marlinFolder, libsshFolder and sqliteFolder.
• The directory where will be stored the results of the metrics computed in the paper (precision, recall, f1 score and runtime) must be defined in the variable metricsResultFolder.
• Now, run the first three tests in the ordered mentioned above for each target system (TestEccoCommit, TestEccoCheckout, TestCompareVariants).
• Lastly, you can run the test to compute metrics and the results (GetCSVInformationTotalTest).

In case that you only want to compute the metrics (run the test GetCSVInformationTotalTest) you can skip the first three tests mentioned above and just copy the csv files (inside the system folder Retrieved_results/compare_input_variants) in the root folder of each target system cloned from this repository.