Skip to content

TUTORIAL.md

Latest commit

聽

History

History
54 lines (46 loc) 路 2.33 KB

TUTORIAL.md

File metadata and controls

54 lines (46 loc) 路 2.33 KB
Raw

How to Build an Enterprise Payment Gateway (The PayLink Way)

Introduction

This tutorial dissects PayLink, a production-ready payment orchestration backend. You will learn how to architect a system that handles money safely, using Go, C++, and Docker.

Prerequisites

  • Linux (Ubuntu 20.04+)
  • Go 1.22+
  • Docker & Docker Compose
  • Basic C++ knowledge

Chapter 1: The Architecture

We don't just dump code; we design systems.

  1. The Core (Go): Handles HTTP requests. It's fast, concurrent, and safe.
  2. The Muscle (C++): Handles crypto. When you need to verify 10,000 webhook signatures/sec, Go's GC can get in the way. We use a C++ shared library.
  3. The Memory (Redis): Handles idempotency. Replaying a $100 charge request shouldn't charge the user $200. Redis keys with expiration prevent this.

Chapter 2: The C++ Interop (cgo)

Look at crypto_cpp/paycrypto.cpp. We expose ComputeHMAC_SHA256 as an extern "C" function. In Go (internal/crypto/wrapper.go), we use:

// #cgo LDFLAGS: -lpaycrypto
import "C"

This allows us to call C++ code as if it were a Go function. Warning: Don't do this for everything. Use it for batch operations or heavy compute to offset the Context Switch cost.

Chapter 3: The Adapter Pattern

We define a single interface:

type ProviderAdapter interface {
    CreatePayment(...)
}

Midtrans, Xendit, and Stripe implementations satisfy this. This means api/handler.go doesn't care which provider is used. It just calls .CreatePayment(). This is Dependency Inversion, key for maintainable software.

Chapter 4: Running It

  1. Setup: 
    make build-cpp # Compiles the .so library
docker compose up --build
  2. Test: 
    curl -X POST http://localhost:8080/v1/checkout -d '{"amount": 10000, "provider_preference": "midtrans"}'

Chapter 5: Best Practices Used

  • Idempotency: We check if an Event ID has been processed in HandleWebhook.
  • Graceful Shutdown: The worker listens for SIGTERM to finish current jobs before quitting.
  • Structural Logging: JSON logs for Datadog/Splunk integration.
  • Clean Architecture: internal/ prevents external imports, keeping the core logic protected.

Conclusion

You now possess the blueprint for a fintech-grade backend. Study the internal/adapters to see how we abstract external APIs.