spring webflux返回application/stream+json

序

本文主要研究下spring webflux返回application/stream+json的實例

maven

<dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-webflux</artifactId>
        </dependency>

controller

/**
     * curl -i localhost:8080/stream
     * @return
     */
    @GetMapping(value = "/stream",produces = MediaType.APPLICATION_STREAM_JSON_VALUE)
    public Flux<Price> priceStream(){
        return Flux.interval(Duration.ofMillis(500))
                .map(l -> new Price(System.currentTimeMillis(),ThreadLocalRandom.current().nextInt(100, 125)))
                .log();
    }

注意這裏produces = MediaType.APPLICATION_STREAM_JSON_VALUE
若是不是application/stream+json則調用端沒法滾動獲得結果，將一直阻塞等待數據流結束或超時。

輸出

• 後臺輸出

2018-02-08 21:36:49.701  INFO 1270 --- [ctor-http-nio-2] reactor.Flux.Map.1                       : onSubscribe(FluxMap.MapSubscriber)
2018-02-08 21:36:49.702  INFO 1270 --- [ctor-http-nio-2] reactor.Flux.Map.1                       : request(1)
2018-02-08 21:36:50.208  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097010208, value=120.0))
2018-02-08 21:36:50.229  INFO 1270 --- [ctor-http-nio-2] reactor.Flux.Map.1                       : request(31)
2018-02-08 21:36:50.708  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097010708, value=124.0))
2018-02-08 21:36:51.208  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097011208, value=119.0))
2018-02-08 21:36:51.707  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097011707, value=120.0))
2018-02-08 21:36:52.207  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097012207, value=109.0))
2018-02-08 21:36:52.707  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097012707, value=101.0))
2018-02-08 21:36:53.208  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097013208, value=114.0))
2018-02-08 21:36:53.707  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097013707, value=113.0))
2018-02-08 21:36:54.206  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097014206, value=105.0))
2018-02-08 21:36:54.708  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097014708, value=103.0))
2018-02-08 21:36:55.208  INFO 1270 --- [     parallel-2] reactor.Flux.Map.1                       : onNext(Price(timestamp=1518097015207, value=123.0))
2018-02-08 21:36:55.212  INFO 1270 --- [ctor-http-nio-2] reactor.Flux.Map.1                       : cancel()

• 調用端輸出

curl -i localhost:8080/stream
HTTP/1.1 200 OK
transfer-encoding: chunked
Content-Type: application/stream+json;charset=UTF-8

{"timestamp":1518097010208,"value":120.0}
{"timestamp":1518097010708,"value":124.0}
{"timestamp":1518097011208,"value":119.0}
{"timestamp":1518097011707,"value":120.0}
{"timestamp":1518097012207,"value":109.0}
{"timestamp":1518097012707,"value":101.0}
{"timestamp":1518097013208,"value":114.0}
{"timestamp":1518097013707,"value":113.0}
{"timestamp":1518097014206,"value":105.0}
^C

能夠看到因爲使用了application/stream+json，返回的transfer-encoding是chunked，所以調用端能夠作到滾動輸出。

分頁

使用了webflux以後，可能好奇以前的分頁調用怎麼辦。reactive-streams是把數據當作數據流來用的，所以spring data reactive並不支持返回Page，可是調用參數能夠傳Pageable參數

public interface StocDao extends ReactiveCrudRepository<Stock, String> {

    Flux<Stock> findByName(String name,Pageable pageable);
}

注意這裏返回Flux<Stock>，而不是Page<Stock>
也就是至關於丟失了total count

小結

對於webflux返回的Flux的流數據，須要配合返回MediaType.APPLICATION_STREAM_JSON_VALUE，同時調用端也須要可以支持這種mediaType(WebClient支持)，這樣才能啓到reactive streams的效果。

doc

• Serving large datasets with Spring WebFlux
• mongo.reactive.repositories.usage
• Is there any way to implement pagination in spring webflux and spring data reactive