What is PCBench?

PCBench is a large-scale benchmark with 47,478 test cases, each labeled with compatibility information, covering 844 parameter-changed APIs from 33 popular Python third-party libraries. This large-scale dataset serves as a benchmark for evaluating the detection and repair tools of Python API parameter compatibility issues.

Directory Structure

├── Benchmark
    ├── Library
        ├── API@V1-V2
            ├── APIV1.yml
            ├── APIV2.yml
            ├── test_case#1YY
                ├── test_case.py
                ├── test_case.json
            ├── test_case#2YY
                ├── test_case.py
                ├── test_case.json
            ...
            └──
        ├── API@V1-V2
            ...
        └──
    ...
    └──

• Library: Python third-party library name.
• API@V1-V2: An API with parameter changes between version V1 and V2.
• APIV1.yml: An Aanconda virtual environment configuration file containing the library version V1.
• APIV2.yml: An Anaconda virtual environment configuration file containing the library version V2.
• test_case#1YY: The first test case generated by the API, where the first 'Y' means compatible, where the second 'Y' means runnable.
• test_case.py: Code snippet.
• test_case.json: Configuration file for PCART.

How to Install Virtual Environments?

Method 1:

Run the following command to install the virtual environment.

conda env create -f environment.yml

Method 2:

Download the .tar.gz packages from PCBenchVenvs or PCBenchVenvs (faster in China) and install the virtual environment by executing the following commands:

cd /home/usr/anaconda3/envs
mkdir envName
tar -xzvf envName.tar.gz -C envName

Note: Some envs may still be missing system libraries (e.g., libopenblas.so.0), which should be installed using apt.

Example of Parameter Mutation on foo for Generating Test Cases

Usage

Prerequisites

• Python 3.9
• openpyxl 3.1.2
• anaconda 23.5.2

Run Command

python run.py

How Does PCART Use It?

Each test case in PCBench can be regarded as an independent test project. To use PCART to detect and repair compatibility issues in PCBench, you need to create the virtual environments for the test case's current version and the target version based on the .yml files, and then create a JSON configuration file as follows:

Write Configuration Files Manually

{
    "projPath": "/home/usr/Benchmark/flask/[email protected]/flask.json.load#1YY",
    "runCommand": "python loadYY.py",
    "runFilePath": "",
    "libName": "flask",
    "currentVersion": "2.2.5",
    "targetVersion": "2.3.0",
    "currentEnv": "/home/usr/anaconda3/envs/flask2.2.5",
    "targetEnv": "/home/usr/anaconda3/envs/flask2.3.0"
} 

Generate Configuration Files Automatically

python generateJson.py

Then, copy the configuration file to PCART/Configure and run command as follows:

python main.py -cfg config.json