📦 MiniSQL — Database System Implementation in C++

🧠 A full-fledged relational database engine from scratch
📘 Inspired by: Database Systems: The Complete Book by Garcia-Molina, Ullman, Widom

---

📝 Overview

MiniSQL is a complete educational database system implemented entirely in C++, developed over the period of 5 months, graduate semester long project. It simulates the core internals of a relational database engine, including query parsing, compilation, relational algebra execution, indexing via B+ Trees, and binary-level storage.

This project was developed in multiple phases, each representing a major subsystem of a real-world DBMS. From catalog management to join algorithms, from scan and filter operations to query plan visualization — everything was built by hand, with a focus on learning and adhering to theory from foundational texts.

---

✨ Features

✅ Custom SQL-like parser using Flex/Bison
✅ Execution of relational algebra operations: Scan, Select, Join, Project, Aggregate, GroupBy, Duplicate Removal
✅ Join algorithms implemented:

• Nested-Loop Join
• Hash Join
• Symmetric Hash Join

✅ Custom-built B+ Tree Indexing (with disk persistence and binary serialization)
✅ File-based table storage using custom heap files (.dat format)
✅ Metadata stored via lightweight SQLite-backed catalog
✅ Graphviz-based query plan visualizer
✅ Interactive shell with support for:

• .help
• .tables
• .schema
• .quit ✅ One-command build system via Makefile

---

🖥️ CLI in Action

Example of entering a query and getting tabular output + plan summary

---

🧠 Query Plan Visualization (Graphviz)

Exported DOT graph rendered for a join query

---

📦 B+ Tree Structure (Create Index)

Shows the tree building from insertions during index creation

---

🧮 Disk-backed Heap File (Visual Debug)

Illustrating table storage in tpc-h data

---

🗂️ Project Structure

MiniSQL/
├── code/
│   ├── headers/              # All header files (Schema.h, Catalog.h, DBFile.h, etc.)
│   ├── source/               # All source files (RelOp.cc, File.cc, QueryCompiler.cc, etc.)
│   ├── project/              # Phase entry points (e.g., test-phase-1.cc, main.cc)
│   ├── execs/                # Output binaries (e.g., main.out, test-phase-*.out)
│   └── makefile              # Master Makefile for compilation
├── data/
│   ├── *.tbl                 # TPC-H formatted raw data files
│   └── *.schema              # Optional schema descriptors
├── catalog.sqlite            # SQLite-backed metadata catalog
├── Dockerfile                # 🐳 (optional) Docker container setup
├── screenshots/              # Folder for screenshots (for README)
└── README.md                 # 📘 This file

---

⚙️ Build & Run

📥 Requirements

• Linux (or WSL/macOS)
• g++, make, flex, bison, sqlite3, libsqlite3-dev, graphviz

🛠️ Compile

cd code
make init
make all

💻 Start the CLI

./execs/main.out

🧪 Sample SQL Queries

-- 1. Simple selection

SELECT * FROM nation WHERE n_regionkey = 1;

-- 2. Join with filter

SELECT c.c_name, o.o_orderkey 
FROM customer c, orders o 
WHERE c.c_custkey = o.o_custkey AND c.c_name = 'Customer#000000011';

-- 3. Aggregate with GROUP BY

SELECT c_nationkey, SUM(c_acctbal) 
FROM customer 
GROUP BY c_nationkey;

-- 4. Create Index

CREATE INDEX idx_cname ON customer(c_name);

💡 After creating an index, MiniSQL will automatically use it for eligible queries via ScanIndex.

🕹️ CLI Commands

Command	Description
.help	List available commands
.tables	Show all tables
.schema	Show schema of a table
.quit	Exit the CLI

Each SQL query must end with a semicolon ;. Multi-line queries are supported.

📚 References 📘 Database Systems: The Complete Book, 2nd ed. — Garcia-Molina, Ullman, Widom

🧪 TPC-H Benchmark data

🧰 SQLite3 (for catalog metadata)

🧠 Graphviz (for query plan visualizations)

🙏 Acknowledgments

This project is the result of deep effort and iterative implementation over an entire semester. Grateful thanks to:

🧑‍🏫 Course instructors & TAs for the structure and feedback

🧠 The authors of the database textbook that made concepts click

🧰 Tools like SQLite, Graphviz, and Flex/Bison

🙌 Everyone who reviewed my code, debugged with me, or supported this journey

🎯 Conclusion Building a database engine from scratch is one of the most enlightening journeys in systems programming and data engineering. This project taught me how the database engines we use daily are engineered layer by layer — from bit-level file management to high-level SQL interpretation.

This project is now a powerful sandbox for exploring ideas like query optimization, indexing strategies, and operator design.

Made with ❤️, C++, and coffee. Happy querying! ☕