Educational Project: MilanSQL is built to understand database internals from scratch. It uses Claude Code as an AI pair programmer. Not a production MySQL/PostgreSQL replacement.
Live Browser Demo — Run SQL queries directly in your browser, no installation needed!
Features: Full SQL, table browser, example queries, query sharing, dark theme.
v8.2.0 — 152 phases, pure C++17 — Production Ready
A complete relational database engine built from scratch in C++17.
152 development phases. Zero external dependencies. 5 network protocols.
Multi-Tenant, High Availability, Load Balancer, pgvector AI/ML, 10k Fuzz-Tested.
http://178.105.206.36:8080/webui — Live server running on Hetzner, no installation needed!
| Guide | Description |
|---|---|
| SQL Reference | Complete SQL command reference — all 152 phases |
| Tutorial | Step-by-step getting started guide |
| Architecture | Internal architecture deep-dive |
| Live Demo | Try MilanSQL in your browser |
| WASM Build | Build for WebAssembly / Emscripten |
| Language | Install | Protocol |
|---|---|---|
| Python | pip install milansql |
DB-API 2.0, TCP :4406 |
| Node.js | npm install milansql |
TCP :4406 |
| Java | com.milansql:milansql-jdbc:5.0.0 |
JDBC over TCP |
| Rust | cargo add milansql |
TCP :4406 |
| .NET | dotnet add package MilanSQL.Data |
ADO.NET over TCP |
# Build
cmake -B build -G Ninja && ninja -C build
# Start interactive REPL
./build/milansql.exe
# Start all servers
./build/milansql.exe --http --port 8080 --mysql --mysql-port 4407 --pg --pg-port 5433 --graphql --graphql-port 8081| Category | Features | Status |
|---|---|---|
| SQL Core | SELECT/INSERT/UPDATE/DELETE/TRUNCATE | ✅ |
| JOINs | INNER/LEFT/RIGHT/FULL, Hash/Merge/Nested Loop | ✅ |
| Subqueries | IN/EXISTS/Correlated/Scalar/Recursive CTE | ✅ |
| Window Functions | ROW_NUMBER/RANK/DENSE_RANK/SUM/AVG OVER PARTITION | ✅ |
| Transactions | MVCC/WAL/Savepoints/SELECT FOR UPDATE | ✅ |
| Constraints | PK/FK/NOT NULL/UNIQUE/CHECK/DEFAULT/AUTO_INCREMENT | ✅ |
| Data Types | INT/TEXT/REAL/DATE/TIME/DATETIME/JSON/POINT/Array | ✅ |
| Indexes | B-Tree/Composite/Full-Text/Spatial | ✅ |
| DDL | CREATE/DROP/ALTER TABLE/VIEW/INDEX/TRIGGER/PROCEDURE | ✅ |
| Procedures | Stored Procedures with Cursor/Loop/IF/SIGNAL | ✅ |
| Triggers | BEFORE/AFTER INSERT/UPDATE/DELETE (ROW + STATEMENT) | ✅ |
| Views | Regular Views + Materialized Views | ✅ |
| Partitioning | RANGE/LIST/HASH + Partition Pruning | ✅ |
| Replication | Physical Master/Slave + Logical Pub/Sub | ✅ |
| Protocols | TCP(4406) + MySQL(4407) + PG Wire(5433) + HTTP(8080) + GraphQL(8081) | ✅ |
| Security | Users/GRANT/REVOKE/Row-Level Security/Policies | ✅ |
| Performance | Hash Join/Buffer Pool/Query Cache/Parallel/Column Store | ✅ |
| Analytics | Window Functions/Column Store/Time-Series/time_bucket | ✅ |
| Admin | Backup/Restore/VACUUM/Checkpoint/CDC/Event Scheduler | ✅ |
| Extensions | milansql_math/crypto/uuid/text + Extension System | ✅ |
| Clients | Python DB-API 2.0 / Node.js / MySQL / psql | ✅ |
| Storage | Page-based I/O + Compression (LZ4/RLE/Dictionary) | ✅ |
| Catalog | INFORMATION_SCHEMA + pg_catalog (9 tables) | ✅ |
| Testing | 636 automated tests + 10k Fuzz Hardening | ✅ |
| Statistics | ANALYZE TABLE / CREATE STATISTICS / Optimizer Hints | ✅ |
| UDF | CREATE FUNCTION / CREATE PROCEDURE / CALL / SHOW FUNCTIONS | ✅ |
| Protocol | Port | Compatible With |
|---|---|---|
| Native TCP | 4406 | MilanSQL clients |
| MySQL Wire | 4407 | mysql CLI, Python mysql-connector, Node mysql2 |
| PostgreSQL Wire | 5433 | psql, libpq, psycopg2 |
| REST/HTTP API | 8080 | curl, any HTTP client, web browser |
| GraphQL | 8081 | GraphQL clients, web browser playground |
-- Window functions
SELECT name, salary,
RANK() OVER (PARTITION BY dept ORDER BY salary DESC) AS rank
FROM employees;
-- Recursive CTE (Fibonacci)
WITH RECURSIVE fib(n, a, b) AS (
SELECT 0, 0, 1
UNION ALL
SELECT n+1, b, a+b FROM fib WHERE n < 10
)
SELECT n, a FROM fib;
-- Array operations
SELECT array_agg(name) AS names FROM employees GROUP BY dept;
SELECT * FROM employees WHERE array_contains(skills, 'SQL');
-- Time-series
SELECT time_bucket(1 DAY, ts) AS day, AVG(value)
FROM metrics GROUP BY day ORDER BY day;
-- Change Data Capture
ALTER TABLE orders ENABLE CDC;
SELECT * FROM cdc.orders AFTER SEQUENCE 5;
-- Extensions
CREATE EXTENSION milansql_crypto;
SELECT md5('hello'); -- 5d41402abc4b2a76b9719d911017c592
-- pg_catalog
SELECT * FROM pg_catalog.pg_tables;
SELECT * FROM information_schema.columns WHERE table_name = 'employees';src/
├── engine/engine.hpp # Core engine: Table/Row/Column, all DML/DDL
├── parser/parser.hpp # SQL tokenizer + parser → ParsedCommand
├── dispatch.hpp # SQL command dispatcher
├── storage/storage.hpp # Binary format v9 serializer
├── storage/column_store.hpp # Column store engine (OLAP)
├── server/http_server.hpp # REST API + Web Dashboard
├── server/pg_server.hpp # PostgreSQL wire protocol v3
├── server/mysql_server.hpp # MySQL wire protocol
├── api/graphql_server.hpp # GraphQL API server
├── cdc/cdc_manager.hpp # Change Data Capture
├── compression/ # LZ4/RLE/Dictionary/ZSTD compressors
├── timeseries/ # Time-series manager
├── extensions/ # Extension system + built-ins
├── pool/connection_pool.hpp # Connection pool multiplexer
├── types/array_type.hpp # Array type utilities
├── fdw/ # Foreign Data Wrapper (CSV + HTTP)
├── copy/copy_manager.hpp # COPY FROM/TO bulk import/export
├── cache/statement_cache.hpp # LRU prepared statement cache
├── optimizer/ # Query rewriter + adaptive stats
├── replication/ # Logical replication pub/sub
├── parallel/ # Parallel query executor
├── pubsub/ # LISTEN/NOTIFY pub/sub
├── locking/ # Lock manager (SELECT FOR UPDATE)
└── main.cpp # REPL + CLI
Requirements: CMake 3.16+, C++17 compiler, MSYS2/UCRT64 (Windows) or GCC 10+ (Linux)
# Windows (MSYS2 UCRT64)
export PATH="/c/msys64/ucrt64/bin:$PATH"
cmake -B build -G Ninja -DCMAKE_BUILD_TYPE=Release
ninja -C build
# Linux
cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build
# Run tests
./build/milansql_tests.exe # 223 tests
./build/milansql_stress.exe # stress tests# Python
from milansql_client import connect
conn = connect(host='localhost', port=4406)
cur = conn.cursor()
cur.execute("SELECT * FROM users")
print(cur.fetchall())// Node.js
const { connect } = require('./clients/nodejs/milansql_client');
const conn = await connect({ host: 'localhost', port: 4406 });
const rows = await conn.query('SELECT * FROM users');| Metric | MilanSQL v5.8.0 |
|---|---|
| INSERT throughput | ~98,000 rows/sec |
| SELECT scan | ~1M rows/sec |
| Index lookup | O(log n) B-Tree |
| Q1 (GROUP BY) | <10 ms |
| Q3 (3-table JOIN) | <20 ms |
| Q5 (5-table JOIN) | <30 ms |
| Column store vs row | 5–10× faster aggregation |
| Test suite | 223 tests, 0 failures |
MIT License — see LICENSE