TSE2025SemArc

This repository contains dataset for the paper "Software Architecture Recovery Augmented with Semantics". The folders is organized as follows:

└─ground-truth
    ├─collected
    └─labeled
└─semantic_analysis
└─SemArc

ground-truth

The dataset consists of eight ground-truth datasets sourced from existing research and seven additional datasets created by us. These datasets cover three different programming languages and vary in sizes, providing a comprehensive set of real-world examples for software architecture recovery tasks.

• collected: This folder contains datasets sourced from existing research, providing a reliable foundation for evaluating architecture recovery methods.
• labeled: This folder includes the datasets created by us, where the ground-truth architecture has been manually labeled and validated. These datasets are designed to cover a wide range of system sizes and programming languages, ensuring a diverse set of examples.

semantic_analysis

This module utilizes Large Language Models (LLMs) to identify both code semantics and architectural semantics within the project. Before use this you should setup API_KEY and choose LLM_MODEL in config.py. The analysis results are automatically saved into two JSON files, which are named after the project as follows:

• [project name]_ArchSem.json: This file contains the architectural semantics identified in the project.
• [project name]_CodeSem.json: This file contains the code semantics identified in the project.

Usage

To run the semantic analysis, use the following command:

python semantic_analysis.py [project folder]

SemArc

Usage

python SemArc.py [-h] [-g  [...]] [-o] [--cache_dir] [-s  [...]] [-a  [...]] [-c  [...]] [-r] [-n] datapath [datapath ...]

positional arguments:

  datapath              path to the input project folder

options:
  -h, --help            show this help message and exit
  -g  [ ...], --gt  [ ...]
                        path to the ground truth json file
  -o , --out_dir        path to the result folder
  --cache_dir           cache path
  -s  [ ...], --stopword_file  [ ...]
                        paths to external stopword lists
  -a  [ ...], --archsem_file  [ ...]
                        paths to architecture semantic file
  -c  [ ...], --codesem_file  [ ...]
                        paths to code semantic file
  -r , --resolution     resolution parameter, affecting the final cluster size.
  -n, --no_fig          prevent figure generation

Example

We have provided a demo folder that contains the source code, ground-truth architecture and semantic files of bash-4.2. You can run the architecture recovery process on this example project using the following command:

python .\SemArc.py .\demo\bash-4.2 -s .\stopwords.txt -a .\semantic_analysis\bash-4.2_ArchSem.json -c .\semantic_analysis\bash-4.2_CodeSem.json -g .\demo\bash-4.2-GT.json