166 phases. 1221 tests. 5 network protocols. Zero dependencies.
A complete relational database engine with ACID transactions,
PostgreSQL Wire Protocol, GraphQL API, CDC, WebSocket, WASM, Multi-Tenant, HA Sentinel, and adaptive query optimizer.
Built phase-by-phase from the ground up — every feature is hand-crafted in pure C++17 without a single external library.
Full MVCC with Write-Ahead Logging, SAVEPOINT, ROLLBACK TO SAVEPOINT, and SELECT FOR UPDATE row-level locking — battle-tested durability.
Learn more →Pure C++17 standard library only. No Boost, no OpenSSL, no third-party headers. Clone, compile, run — on Linux, Windows, or macOS.
View source →TCP(4406) + MySQL Wire(4407) + PostgreSQL Wire(5433) + REST/HTTP(8080) + GraphQL(8081). Connect with psql, mysql CLI, curl, or any HTTP client.
Get started →CREATE COLUMN TABLE for OLAP workloads. Columnar SUM/AVG/COUNT/MIN/MAX scans, GROUP BY fallback, and SHOW STORAGE FORMAT to compare row vs column store.
See example →Binary-log master/slave replication with auto-sync, PLUS logical replication via CREATE PUBLICATION / CREATE SUBSCRIPTION with append-only change log.
Learn more →ENABLE/DISABLE RLS per table, CREATE POLICY with USING expressions, CURRENT_USER_ID() predicate substitution, and persistent policy storage.
Learn more →FULLTEXT INDEX with MATCH … AGAINST syntax and TF-based relevance scoring across multiple columns — no external search engine required.
See example →MySQL-compatible CREATE EVENT with EVERY n SECOND/MINUTE/HOUR/DAY schedules and AT timestamp one-shot events — background job execution built in.
Learn more →Automatic query rewriting (removes WHERE 1=1, redundant conditions), access statistics tracking, and automatic index suggestions via SHOW INDEX SUGGESTIONS.
See example →ALTER TABLE … ENABLE CDC tracks every INSERT/UPDATE/DELETE as a change event. Query CDC streams with AFTER SEQUENCE N for real-time data pipelines.
Learn more →Built-in GraphQL server on port 8081. Query any table with auto-generated schema, field filtering, and an in-browser GraphQL playground — no config needed.
Get started →LZ4, RLE, Dictionary, and ZSTD compression engines for page-level storage. Transparent to queries — compresses cold data automatically.
Learn more →time_bucket() aggregation, retention policies, and partitioned time-series tables. Efficient range scans and automatic partition pruning for metrics workloads.
See example →MilanSQL speaks standard SQL — and adds table inheritance, column stores, pub/sub, and more.
-- Phase 78: Table Inheritance CREATE TABLE person (id INT PRIMARY KEY, name TEXT); CREATE TABLE employee (salary REAL) INHERITS (person); INSERT INTO person VALUES (1, 'Alice'); INSERT INTO employee VALUES (2, 'Bob', 75000); SELECT * FROM person; -- returns Alice + Bob (inherited) SELECT * FROM ONLY person; -- returns only Alice -- Phase 75: Row-Level Security ENABLE ROW LEVEL SECURITY ON employee; CREATE POLICY own_rows ON employee FOR SELECT TO alice USING (id = CURRENT_USER_ID()); -- Phase 82: Adaptive Optimizer SET QUERY_REWRITE = ON; EXPLAIN REWRITTEN SELECT * FROM employee WHERE 1=1; -- Rewrite: removed always-true condition WHERE 1=1 SHOW INDEX SUGGESTIONS; -- CREATE INDEX idx_employee_salary ON employee(salary) -- 78% -- Phase 76: LISTEN / NOTIFY pub/sub LISTEN updates; NOTIFY updates, 'new_employee_added';
# DB-API 2.0 compatible Python client (stdlib only) import milansql # Connect via DSN (Phase 79) conn = milansql.connect("milansql://root:pass@localhost:4406/mydb") cursor = conn.cursor() # Execute SQL cursor.execute(""" CREATE TABLE orders ( id INT PRIMARY KEY AUTO_INCREMENT, item TEXT NOT NULL, price REAL ) """) cursor.execute("INSERT INTO orders VALUES (NULL, 'Laptop', 999.99)") conn.commit() # Fetch results cursor.execute("SELECT * FROM orders WHERE price > %s", (100,)) rows = cursor.fetchall() for row in rows: print(row) conn.close()
# Start REST API server $ ./build/milansql.exe --http --port 8080 # Execute a query via HTTP POST $ curl -s -X POST http://localhost:8080/query \ -H 'Content-Type: application/json' \ -d '{"sql": "SELECT * FROM orders LIMIT 10"}' # Response (JSON) { "columns": ["id", "item", "price"], "rows": [ ["1", "Laptop", "999.99"] ], "rowCount": 1, "elapsed_ms": 0.42 } # Web Dashboard available at: $ open http://localhost:8080/ # List tables $ curl http://localhost:8080/tables # Health check $ curl http://localhost:8080/health "OK"
Every component lives in a single .hpp file. No build magic — just CMake and a C++17 compiler.
Comprehensive guides, SQL reference, and architecture docs — all in one place.
Complete reference for all SQL commands — DDL, DML, transactions, window functions, CTEs, and MilanSQL-specific extensions.
Open reference →Step-by-step tutorial from first query to advanced features — joins, transactions, replication, and client drivers.
Start tutorial →Deep-dive into the query pipeline, storage format, optimizer layers, and concurrency model of MilanSQL.
Explore architecture →No package manager, no Docker, no config files. Just a C++17 compiler and CMake.
git clone https://github.com/ haidari9819-lang/milansql cd milansql cmake -B build -G Ninja cmake --build build
# HTTP + Web Dashboard ./build/milansql.exe \ --http --port 8080 # TCP Server ./build/milansql.exe \ --server --port 4406 # Interactive REPL ./build/milansql.exe
# Via REST API curl -X POST \ http://localhost:8080/query \ -H 'Content-Type: application/json'\ -d '{"sql":"SHOW TABLES"}' # Via DSN (Python) import milansql conn = milansql.connect( "milansql://root@localhost/db" )