การสร้างบริการเรียลไทม์เชิงรีแอกทีฟ

Welcome! In this lesson, we'll build end-to-end reactive real-time services using Spring WebFlux and Project Reactor.

Reactive services are excellent for handling many concurrent connections efficiently. They offer better scalability and responsiveness compared to traditional blocking approaches.

At the heart of reactive programming in Spring is Project Reactor. It provides two key types for handling data streams:

• Flux: Represents a stream of 0 to N items. Think of it as a publisher that can emit multiple values over time.
• Mono: Represents a stream of 0 to 1 item. Useful for operations that return a single result or no result (like void).

These types allow us to compose asynchronous operations in a clear and non-blocking way.

Spring WebFlux uses the WebSocketHandler interface to manage WebSocket connections. Its main method, handle(), takes a WebSocketSession and returns a Mono.

This Mono signifies that the handling process is complete once the reactive stream it represents finishes. We can use Flux inside to send continuous messages.

To build a real-time service, we need a source of data. Let's create a simple Flux that emits a message periodically. This simulates a real-time data feed, like a stock ticker or a sensor reading.

We'll use Flux.interval() to generate events and map() to transform them into useful messages.

Here's a basic WebSocketHandler that sends a 'tick' message every second. It uses the Flux.interval() we discussed.

The session.send() method takes a Flux to push data to the client.

To make our TimeTickerHandler runnable, we need a Spring Boot application. This example sets up the WebFlux server and registers our handler.

Access this via ws://localhost:8080/ticker in a WebSocket client (like Postman or a browser's DevTools console) to see it in action.

Our previous ticker only sent data. To make it truly interactive, we can also process messages coming from the client.

The session.receive() method returns a Flux that represents incoming messages. You can subscribe to this Flux to react to client input, for example, by filtering, transforming, or using the data to control the output stream.

Errors can occur in any part of a reactive pipeline. Project Reactor provides operators to handle these gracefully, preventing your application from crashing:

• onErrorResume(): Recovers from an error by switching to an alternative publisher.
• doOnError(): Performs a side-effect (like logging) when an error occurs, then re-throws it or completes.
• retry(): Retries the sequence if an error occurs.

Using these helps build robust real-time services that can recover from transient issues.

Backpressure is crucial for reactive systems. It's a mechanism where a consumer can signal to a producer that it's receiving data too quickly and needs the producer to slow down.

Project Reactor handles backpressure automatically. When a client can't keep up, the WebSocket connection might buffer messages or eventually close, but the server-side Flux won't overwhelm itself or the network.

We've explored how to build reactive real-time services using Spring WebFlux and Project Reactor.

• We saw how Flux can generate continuous data streams.
• We implemented a WebSocketHandler to push these streams to clients.
• We configured a basic Spring Boot application to host our reactive WebSocket endpoint.
• We touched upon error handling and backpressure, vital for robust systems.

These principles enable highly scalable and responsive real-time applications.