GenProgJS: a Baseline System for Test-based Automated Repair of JavaScript Programs
GenProgJS was evaluated on the BugsJS dataset. The random nature of the genetic algorithm makes it harder to validate and reproduce our results, thus, for the sake of generalization our experiments were carried out on 5 independent runs. On this page we gathered and systematized the data for all the generated patches and display them.
Online appendix components
- GenProgJS homepage
- Docker environment
- Experiment data (this page)
Data
In the table below one can find the experiment data for all the five independent runs. We created diffs for every generated patch and created an html file for each run (see column 3 - Diffs). In the data folder (column 2) there is a subfolder for each bug following the [project]_[bugNumber] naming convention (eg. the first bug in the Eslint project would have the folder name eslint_1). Inside of these subfolders one can find 3 types of files:
general_info.log: log messages containing general information about the fitness values, running time and population throughout the search process.[project]_[bugNumber]_[index].info: contains detailed information about the[index]th generated patch. This information includes the generation in which the patch was generated, the elapsed time, test execution information and the list of applied operators (and their corresponding modifications).[project]_[bugNumber]_[index].js: the patched version of the original program.
| Run | Data folder | Diffs | Avg. runtime (30 gens.) | Avg. #candidate | #operators |
|---|---|---|---|---|---|
| #1 | GenProgJS/experiments/#1 | #1/ | 19.54 h | 5.92 | 622 |
| #2 | GenProgJS/experiments/#2 | #2/ | 16.32 h | 6.9 | 764 |
| #3 | GenProgJS/experiments/#3 | #3/ | 16.34 h | 6 | 434 |
| #4 | GenProgJS/experiments/#4 | #4/ | 14.12 h | 3.9 | 453 |
| #5 | GenProgJS/experiments/#5 | #5/ | 17.39 h | 9 | 954 |