🗄️ Mini-DBMS: High-Performance Database Engine

A custom-built relational Database Management System (DBMS) written in Java. This project implements the core components of a database engine from scratch, focusing on low-level memory management, disk paging, and query execution operators.

📖 Project Overview

This engine simulates the internal behavior of systems like PostgreSQL or MySQL. Unlike high-level SQL applications, this project handles the physical storage of data, buffer pooling, and the algorithmic implementation of relational operators (Selection, Projection, Join).

Key Technical Features:

• Page-Oriented Storage: Data is stored in binary pages (fixed size) managed directly on the disk.
• Buffer Management: Implementation of a Frame Pool with replacement policies (LRU/Clock) to optimize I/O cost.
• Query Execution Engine: Implementation of Volcan-style iterators (IRecordIterator) for pipelined query processing.
• Heap Files: Management of free slots and record serialization.

---

🏗️ System Architecture & Code Structure

The codebase follows a strict layered architecture, separating physical storage from logical query processing.

graph TD;
    A["Client / CLI"] --> B["requete (DBManager)"];
    B --> C["relationnel (Relations & Records)"];
    C --> D["buffer (BufferManager)"];
    D --> E["espaceDisque (DiskManager)"];
    E --> F[("Disk Storage")];

Loading

📂 Repository Organization

Based on the src/main/java structure:

• espaceDisque/ (Disk Manager Layer)

  • Handles the physical allocation/deallocation of Pages (PageId, DiskManager).
  • Manages persistent storage in .db files.
  • Implements low-level iterators (PageDirectoryIterator).

• buffer/ (Buffer Management Layer)

  • BufferManager: Acts as the cache between Disk and Memory. It manages "Frames" and handles page pinning/unpinning to minimize expensive disk access.

• relationnel/ (Data Structure Layer)

  • Defines the schema and data representation.
  • Relation: Represents a table schema.
  • Record: Handles byte-level serialization/deserialization of tuples.

• requete/ (Query Processor Layer)

  • DBManager: The main entry point for executing queries.
  • Condition: Evaluates WHERE clauses (filters).
  • ProjectOperator / JoinOperator: Implementation of relational algebra operators.

• test/ (Unit Testing)

  • Comprehensive testing suite for each layer (DiskManagerTests, RelationTest, etc.) ensuring component reliability.

---

👥 Team & Credits

This project was developed as part of the Computer Science curriculum at Université Paris Cité.

• Ali Traore
• Valentin Ponoussamy
• Thomas Gourmelen
• Abdel Malik Djaffer

---

⚡ Supported Operations

The engine supports a subset of SQL-like commands handled by the DBManager:

CREATE TABLE

Create a table with defined columns.

CREATE TABLE NomTable (NomCol1:TypeCol1, NomCol2:TypeCol2, ..., NomColN:TypeColN(size))

Rules :

• Column names and types are separated by : with no spaces.

• Keywords and table names are separated by a single space.

• Example:

  CREATE TABLE Clients (ID:INT, Nom:STRING(50), Age:INT)

INSERT

Insert a tuple into a table.

INSERT INTO NomTable VALUES (val1, val2, ..., valn)

Rules :

• Strings must be surrounded by quotes ("value").
• Example:

  INSERT INTO Clients VALUES (1, "Alice", 25)

BULKINSERT

Insert data from a CSV file.

BULKINSERT INTO NomTable nomFichier.csv

SELECT

Retrieve data with or without conditions.

SELECT aliasRel.col1, aliasRel.col2, ... FROM NomTable aliasRel [WHERE condition]

Rules :

• Conditions follow: col1 OP col2, where OP is on of =, <, >, <=, >=, <>.

• Join example :

  SELECT c.Nom, o.Montant FROM Clients c, Commandes o WHERE c.ID = o.ClientID

  Additional Commands

• Create a database :

  CREATE DATABASE NomDB

• Select a database :

  SET DATABASE NomDB

• List tables in the current database :

  LIST TABLES

• Drop a table :

  DROP TABLE NomTable

• List all tables :

  LIST DATABASES

• Drop a database :

  DROP DATABASE NomDB

  Running the Program

This guide explains how to run the program on Linux or macOS. Make sure you are at the project root: projet-bdda.**

---

Requirements

1. Linux or macOS terminal

The script is compatible with these systems only.

2. Java installed

java -version

Script run.sh

Make run.sh executable (if needed)

chmod +x run.sh

---

Execution Modes

Le programme peut être lancé de deux manières différentes :

1. Run with the default config

In this mode, the program automatically uses file-config.json in the current directory.

./run.sh

2. Run with a custom config file

./run.sh path/to/config.json