Automated Extraction and Refactoring of NoSQL Schemas from Application Code

This repository contains the implementation of a code analysis framework to infer NoSQL database schemas and generate refactoring plans from application source code. Produces logical schemas conforming to the U-Schema metamodel.

Overview

uschema-code-analysis analyzes the source code of data-intensive applications built with JavaScript and MongoDB to extract:

• CRUD operations and their target containers.
• Implicit references between data entities.
• Logical schema models (U-Schema).
• Join query removal candidates via duplication plans.

The approach is designed to support schema discovery in schemaless NoSQL databases, improve query performance, and assist in database refactoring through code and data rewriting.

Features

• Code metamodel.
• Control flow metamodel.
• Database operation and Structure metamodel.
• Extraction of U-Schema logical schemas from code.
• Generation of join query removal plans.
• Automated schema, data, and code updates.
• Graph visualization for control flow analysis using Neo4j.

U-Schema Code Transformations Prototype

This repository contains the prototype implementation described in the paper.
It provides tools for analyzing application code, extracting schemas, and applying refactorings.

Requirements

• Java 11+
• Maven 3+
• Eclipse Modeling distribution (including EMF libraries)

Running the Project

To execute the prototype and reproduce the results described in the paper, please follow these steps:

1. Clone and build the project
   Make sure all submodules are properly initialized and run:

   mvn install

   This will download and configure all required dependencies.

2. Open in Eclipse Modeling
   Import the project into an Eclipse installation with the Modeling distribution, which includes the EMF libraries.

3. Locate the launcher project
   Open the project:

   es.um.uschema.code.transfs.launcher
   

4. Run the main class
   Execute the class:

   Launcher.java
   

   This will automatically generate an outputs/ folder containing all inferred models and the corresponding generated code.

5. Input application
   The example application used in the paper is provided in the inputs/ folder.
   You can replace this with other projects if desired.

License

This project is distributed for academic and research purposes.

Dependencies

This project relies on the following tools and libraries:

• Esprima – A high-performance ECMAScript parser used to generate the Abstract Syntax Tree (AST) from JavaScript source code.

This project depend on the U-Schema core projects that are found in:

• U-Schema repository - U-Schema Metamodel and Utils.

$ git clone https://github.com/modelum/uschema

List of related publications:

• Carlos J. Fernández-Candel, Anthony Cleve, Jesús García-Molina, Automated Extraction and Refactoring of NoSQL Schemas from Application Code. arXiv. — The APP used to validate the approach (music-app) in the paper can be found in "es.um.uschema.code.transfs.launcher" under the folder "inputs".