About

A Java-based Implementation of an RDF Triple Store built on dynamic K2-Trees (DK2) and BMatrix indexing. Designed to bridge compact data structure theory with practical software engineering, this system uses RoaringBitmaps for compressed Bitstring Representations and Apache Jena for RDF Serialization. It features full API Abstractions accompanied by a Web-Application built with Spring Boot and Vue.js to demonstrate it in a real-world Context and to provide Interaction wiht it.

The Implementation is based on the Works of N. R. Brisaboa Et al. as described in the following two Papers:

• N. R. Brisaboa, A. Cerdeira-Pena, G. De Bernardo, and A. Fariña, "Revisiting Compact RDF Stores Based on k2-Trees," 2020 Data Compression Conference (DCC), 2020. DOI: 10.1109/DCC47342.2020.00020
• N. R. Brisaboa, A. Cerdeira-Pena, D. B. Guillermo, and G. Navarro, "Compressed Representation of Dynamic Binary Relations with Applications," arXiv.org, 2017. arXiv: 1707.02769

---

Key Features

• CRUD: Supports Create, Read, Update and Delete Operations.
• Full Pattern Matching: Read Operations supports all 8 Subject-Predicate-Object (SPO) query patterns (SPO, SP_, _PO, _P_, etc.).
• Persistence & Serialization: Integrates Apache Jena ARQ serialization mechanisms alongside transaction logging for session Persistence.
• Decoupled Architecture: Features clear API Boundaries, allowing the store to be embedded cleanly into external applications.
• Accompanying Web-App: Features a Full-Stack-Interface with a Spring Boot Backend and a Vue.js Frontend for querying, store management, and visualization.
• Expanded examplary Functionality: Features a Set of additional APIs performing compounded Actions relevant for the provided Web-App.

---

Tech Stack & Dependencies

• Backend Core: Java
• Web Framework: Spring Boot
• Frontend: Vue.js
• RDF Serialization: Apache Jena ARQ
• Bitstring Compression: RoaringBitmap

---

API Documentation

The Triple Store exposes RESTful Endpoints for CRUD operations, query execution, and session management.

• Complete Specs: See docs/api.md for Details, Request/Response Payloads and Query Parameters.

---

Web Application

The Repository includes a Web-Application designed to demonstrate the Store's Capabilities in a practical Environment. The Applications models the territorial Restructuring of the german Territories at the Beginning of the 19th Century.

Features

• Interactive Querying: Execute Pattern-based Triple Queries.
• Live Store Management: Add/Delete or modify individual Triples.
• State Management: View State Changes and trigger State Mediatizations, State Foundings or Ruler Changes.
• Action History: View a History of past State Actions.

For instructions on running the Web-Application, see the Startup Guide.

---

Known Limitations & Problems

While the current Implementation proves the feasibility of dynamic K2-Trees embedded into the BMatrix-Indexing-Strategy in a native Java environment, several Problems are still present:

• Unbound Predicate Scans (_P_): Because Triples are currrently unordered in their respective Matrices, Predicates lack a global spatial Index, leading to unbound Predicate-Queries requirying a full Matrix-Scan.
• Object-Bloat: Managing the Tree-and Traversal-Structures as individual Java Objects creates a significant Pointer Overhead on the heap for large graphs (>=1M triples).