PRISM Artifact: Experiment Outputs

This archive is primarily intended for those who wish to explore the raw outputs of each APCC tool. If your goal is simply to regenerate the experimental results in our paper, please refer to the CSV-formatted summary data we separately provide. After extraction, this `outputs/` directory will contain the following structure: How to Use 1. Download the archive from the link above. 2. Extract it into this `outputs/` directory: tar -xvzf outputs.tar.gz 3. After extraction, this directory will contain subdirectories in the following format: outputs/ ├── Chart1b/ │ └── 642cd9b0/ │ ├── correctness.txt │ ├── formatted.patch │ ├── patches/ │ ├── patch_mthds.results │ ├── test_info.json │ ├── patchsim/ │ ├── ods/ │ ├── shibboleth/ │ └── prism/ ... Directory Structure and Contents Each patch result is stored under a two-level directory path: `outputs/[project_id]/[patch_id]/`. Each resulting directory contains the following metadata: correctness.txt: Ground-truth label indicating whether the patch is correct or incorrect. formatted.patch: Syntactically normalized patch. This is used to represent all patches that are syntactically equivalent. patches: Contains all original benchmark patches that are syntactically equivalent to the current one. patch_mthds.results: Target method informations which are patched (it is used for running shibboleth). test_info.json: Test execution results for the patch using the provided developer test suite. Each patch directory also includes outputs from the following APCC techniques: PatchSim patchsim/result.json: Main classification result. patchsim/classify/result_{threshold}.txt: Classification decisions at different similarity thresholds. ODS ods/features_{project}_{bug_id}_FeatureAnalyzer.json: Raw syntactic feature vectors. ods/feature.json: Subset of selected features used for classification. Shibboleth shibboleth/score.csv: Score vector produced by Shibboleth's static analysis and test impact assessment. PRISM prism/trace-out/: Dynamically collected execution traces from instrumented test runs. It is used for reinforcing PRISM's static analyzer. prism/result.json: Extracted semantic feature vectors for patch classification.