Online Appendix of "An Extensive Study of Smell-Aware Bug Localization"

Prerequisites

• Python 3 + pandas, scipy, numpy, matplotlib
• JupyterLab or Jupyter Notebook

Files Overview

• `smells/${PROJECT}-${VERSION}.xml`: Results of code smell detection by inFusion
• `exp/${GROUP}/${PROJECT}/${BLT}_${PROJECT}_${PROJECT}_${VERSION}/recommended/${NUMBER}.txt`: Bug localization result rankings when using `${BLT}` in Bench4BL
• `exp/${GROUP}/${PROJECT}/${BLT}_${PROJECT}_${PROJECT}_${VERSION}_output.txt`: Gold files in the rankings.
• `result/${CONFIGURATION}/${BLT}/_${METRIC}.csv`: Summary of smell-aware bug localization results using `$CONFIGURATION` and `$BLT`, evaluated by `$METRIC`, which can be generated in Step 3
  • bid: Target bug ID
  • vid: Target version ID
  • base: Metric value of the baseline approach
  • value: Metric value when using the following alpha
  • alpha: Best (but smallest) alpha value
  • alphas: Best alpha values that maximizing the `$METRIC` (colon-separated)
  • total: Metric value before divided by the number of bugs
• `cache/projects.csv.bz2`: A list of projects to be used
  • group: Project group name in Bench4BL
  • project: Project name
  • nbugs: Number of bug reports
  • nversions: Number of versions of the project
  • nsources: Average number of source files (among versions)
• `cache/versions.csv.bz2`: A list of project versions (systems)
  • vid: Version ID (`${PROJECT}-${VERSION}`)
  • group: Project group name in Bench4BL
  • project: Project name
  • version: Version number
  • nsources: Number of source files
  • nbugs: Number of bug reports
  • nsmells: Number of smell instances
• `cache/bugs.csv.bz2`: A list of bug reports to be used
  • bid: Bug ID (`${PROJECT}-${NUMBER}`)
  • group: Project group name
  • project: Project name
  • vid: Version ID to which the bug belongs
  • version: Version number
  • number: Bug number
• `cache/smells.csv.bz2`: A list of smell instances, generated from the XML files in `smell/`
  • vid: Version ID
  • granularity: Granularity of the smell (`class` or `method`)
  • package: Package of the target module
  • class: Class of the target module
  • method: Target method if class-level smell, otherwise empty
  • file: Source filename
  • type: Type of the smell
  • severity: Severity of the smell
• `cache/confs.csv.bz2`: A list of considered configurations, which can be generated in Step 3
  • name: Configuration name (basically `${GRANULARITY}_${AGGREGATOR}_${SELECTOR}`)
  • granularity: Granularity parameter (g)
  • aggregator: Aggregator parameter (a)
  • selector: Type selector parameter (s)
  • level: `basic` (1-150), `individual` (individual smells), `opt` (ideal (0))
• `Bench4BL-patch/`: Patches for Bench4BL. See Step 1.

Since the files generated from Steps 1-3 that are necessary to run the scripts in Step 4 are contained in the archive, you can start from Step 4 directly.

Step 1. Running Bench4BL

• Follow the instruction of repository cloning and archive downloading in https://github.com/exatoa/Bench4BL. We used the commit of Aug 8, 2019.
• Add `Bench4BL-patch/VSM.jar` and `Bench4BL-patch/rVSM.jar` into `techniques/releases/` in Bench4BL.
• Overwrite `scripts/launcher_Tool.py` with `Bench4BL-patch/launcher_Tool.jar` so that the added two techniques become available.
• To run Bench4BL, the `Bench4BL-patch/Dockerfile` may be useful:
  • % docker run -it -v /path/to/Bench4BL:/mnt/exp/Bench4BL bench4bl /bin/bash
• Put the generated results to `exp/`.

Step 2. Preprocessing ranking files

• Run `preprocess.ipynb` so that the files will be generated in `preprocessed/`.
• `preprocessed/${BLT}/${PROJECT}_${VERSION}/${NUMBER}.csv`: Bug localization result rankings attached with additional oracle and smell information, which can be generated in Step 2
  • rank: Rank in the ranking
  • gold: Whether the file is gold or not
  • file_name: Source filename
  • nscore: Similarity score obtained from the bug localization technique
  • class_smells: List of class-level smells with their severity (comma-separated)
  • method_smells: List of method-level smells with their severity (comma-separated)

Step 3. Applying smell-aware bug localization

• Run `sabl.ipynb` so that the files in `result/` will be generated.

Step 4. Analysis

• Run `analysis-rq1.ipynb` for RQ1, RQ3, and RQ4.
• Run `analysis-rq2.ipynb` for RQ2. Files generated by Step 2 are needed, which are not included in the archive.
• Run `analysis-rq3.ipynb` for RQ3.