NovaDB

NovaDB is a lightweight relational database engine written entirely in Core Java 21, designed without any third-party database engines. It supports a subset of SQL, offering persistent storage, indexing, caching, basic transactions, and role-based security.

Architecture

NovaDB is built adhering to SOLID principles and a clean architecture, modularized into distinct packages:

com.novadb
├── backup        # CSV export and ZIP-based automatic backups
├── cache         # In-memory LRU caching for frequently accessed tables
├── cli           # Interactive command-line shell (REPL)
├── exception     # Custom domain exceptions (SyntaxException, StorageException, etc.)
├── executor      # Query execution engine and Expression Evaluator
├── index         # HashMap-based equality indexes
├── logging       # Execution history and query timing logs
├── metadata      # CatalogManager managing database schemas and table metadata
├── model         # Core records/classes: Row, Column, TableInfo, DataType, Result
├── parser        # AST-based recursive descent SQL Parser
├── security      # Basic user authentication and role-based access control (Admin/ReadOnly)
├── storage       # Binary row serialization/deserialization and StorageManager
├── tokenizer     # Lexical analysis converting SQL strings into Tokens
└── util          # CLI result table formatting

Storage Format

NovaDB uses a custom binary format for persisting data.

• Metadata (catalog.meta): Binary encoded mapping of Databases -> Tables -> Columns.
• Table Data (dbName_tableName.dat): Sequential binary log of rows.
  • Each row starts with a boolean (1 byte) indicating if it is active or deleted (for MVCC / deferred cleanup).
  • Followed by an int (4 bytes) denoting the length of the serialized row payload.
  • Followed by the binary serialized payload (custom DataOutputStream encoding per data type).

Supported SQL Syntax

DDL (Data Definition Language)

• CREATE DATABASE <name>;
• DROP DATABASE <name>;
• USE DATABASE <name>; (or USE <name>;)
• CREATE TABLE <name> (col1 INT PRIMARY KEY, col2 STRING NOT NULL, col3 BOOLEAN DEFAULT true);
• DROP TABLE <name>;
• SHOW TABLES;
• DESCRIBE TABLE <name>; (or DESCRIBE <name>;)

DML (Data Manipulation Language)

• INSERT INTO <name> (col1, col2) VALUES (val1, 'val2');
• SELECT col1, col2 FROM <name> WHERE col1 = 1 ORDER BY col2 DESC LIMIT 10; (Supports *, =, !=, >, <, >=, <=, AND, OR, LIKE)
• UPDATE <name> SET col1 = val1 WHERE col2 = val2;
• DELETE FROM <name> WHERE col1 = val1;

Transactions

• BEGIN;
• COMMIT;
• ROLLBACK;

Building and Running

Ensure you have Java 21 and Maven installed.

# Compile the project
mvn clean install

# Run the CLI
mvn exec:java

When prompted, login with the default credentials:

• Username: admin
• Password: admin123

Limitations

• This is a lightweight educational engine. It does not implement B-Trees for range indexing; equality indexes use HashMaps.
• The SQL Parser uses recursive descent and supports a limited subset of SQL (no JOINs or subqueries yet).
• Concurrency uses a global ReadWriteLock for simplicity.
• The transaction model uses deferred execution instead of a full redo/undo Write-Ahead Log.

Future Enhancements

• Add B-Tree indexing for efficient range queries (>, <).
• Implement nested queries and JOIN operations.
• Introduce an LRU Page Cache at the storage level instead of a row-based cache.
• Implement a true Write-Ahead Log (WAL) with ARIES recovery.