Semantic Code Search Engine

Overview

PyCodeSearch is a semantic code search engine allows both Code-To-Code and Text-To-Code search with fine-tuned UniXcoder model.

Prerequisites & Installation

In your virtual environment, run:

pip install -r requirements.txt

Or you can directly install in the search engine notebook.

The above commands will install cpu-only version of the PyTorch package. Please refer to PyTorch website for instructions on how to install other versions of PyTorch.

Directory Structure

PyCodeSearch:.
│  README.md
│  requirements.txt
│
├─notebook
│      CodeSearchEngine.ipynb
│      SearchEngine_with_examples.ipynb
│
└─search_engine
    │  data_prepare.py
    │  model.py
    │  unixcoder.py
    │  __init__.py
    │
    ├─Code2Code
    │  ├─Bi-Encoders
    │  │      Code2code_CodeT5_summarization.ipynb
    │  │      Code2code_UniXcoder.ipynb
    │  │
    │  └─Cross-Encoders
    │          Code2code_GraphCodeBERT.ipynb
    │
    ├─Evaluation
    │      Models_C4_Evaluation.ipynb
    │
    └─Text2Code
            Text2code_CodeT5_summarization.ipynb
            Text2code_GraphCodeBERT.ipynb
            Text2code_Unxicoder.ipynb

Usage

The Code Search Engine User Interface CodeSearchEngine.ipynb can be find in folder notebook, where you can find a demo with 20 search examples for both Code-To-Code and Text-To-Code paradigms.

In folder search_engine, you can explore details implementations about both search paradigms in sub-folders. The Python file here is the library for PyCodeSearch, providing necessary functions APIs.

Notice: our demo takes "https://github.com/keno/algorithms" as an example repository database to search, thus the searching code snippets results all come from the source code in this repository. So, the searching result maybe irrelevant whenever there is no target code snippet in the database.