如何在不使用reduce()或last()折叠的情况下获取Flux的最后一项

时间:2019-11-03 22:03:03

标签: java project-reactor reactive-streams

如何在不使用reduce()或last()折叠的情况下获取Flux的最后一项?这是我的用例:

1)我有一个根据状态产生Flux<T>的生成器。 2)内部Flux完成后,它将更改影响我在生成器中发出的下一个Flux对象的状态。

示意图看起来像这样

static class State {
    int secret = 2;
    int iteration = 0;
}

Random rand = new Random(1024);
Flux<Integer> stream = Flux.<Flux<Integer>, State>generate(State::new, (state, sink) -> {

    System.out.println(String.format("Generate: %d", state.secret));
    Flux<Integer> inner = Flux.range(1, rand.nextInt(10));

    sink.next(inner.doOnComplete(() -> {
        // How do I get last item of `inner` here ?
        // For example I'd like to decrement `state.secret` by last value of `inner`
    }));

    return state;
}).flatMap(Function.identity());

UPD:我未标记我的答案,因为事实证明该黑客是不可靠的。可能会在完全消耗之前的.generate()之前调用Flux,从而使last中的值不正确。

1 个答案:

答案 0 :(得分:0)

第一个版本不可靠。我砍了另一个:

static <T> Flux<T> expandOnLastItem(Supplier<Flux<T>> seed, Function<T, Flux<T>> generator) {
    return Flux.just(new AtomicReference<T>())
            .flatMap(last -> Flux.just(seed.get().materialize())
                    .flatMap(Function.identity())
                    .expand(v -> {
                        if (v.hasValue()) {
                            last.set(v.get());
                        } else if (v.isOnComplete() && last.get() != null) {
                            Flux<T> res = generator.apply(last.get());
                            last.set(null);
                            return res.materialize();
                        }
                        return Flux.empty();
                    })
                    .filter(s -> !s.isOnComplete())
                    .dematerialize());
}

可以用作

static Flux<Integer> getPage(int pageId, int size) {
    return Flux.defer(() -> {
        if (pageId < 3) {
            System.out.println("Returning data for pageId: " + pageId);
            return Flux.range(pageId * 100, size);
        } else {
            System.out.println("Returning empty for pageId: " + pageId);
            return Flux.empty();
        }
    });
}

expandOnLastItem(
        () -> getPage(0, 5),
        lastId -> {
            System.out.println("  Expanding. Last item: " + lastId);
            int curPage = lastId / 100;
            return getPage(curPage + 1, 5);
        })
        .reduce(0L, (count, value) -> {
            System.out.println("==> " + value);
            return count + 1;
        })
        .block();

所以我通过在generator中改变状态变量来破解它。它可以工作,但不是很实用。如果其他人可以提出其他建议,我将不胜感激。

Random rand = new Random(1024);
Flux.<Flux<String>, State>generate(State::new, (state, sink) -> {

    if (state.iteration < 4) {
        final int count = rand.nextInt(10) + 1;
        System.out.println(String.format("*** Generate %d: start %d (count %d)", state.iteration, state.secret, count));
        Flux<Integer> inner = Flux.range(state.secret, count);

        final int[] last = {Integer.MIN_VALUE};
        sink.next(
                inner
                        .doOnNext(value -> {
                            last[0] = value;
                        })
                        .map(value -> String.format("Iter %d value %d", state.iteration, value))
                        .doOnComplete(() -> {
                            System.out.println(String.format("Inner complete (last item was %d)", last[0]));
                            state.secret = last[0];
                            state.iteration += 1;
                        }));
    } else {
        System.out.println("Generate complete");
        sink.complete();
    }

    return state;
})
        .flatMap(Function.identity())
        .map(value -> {
            System.out.println(String.format("Ext map: %s", value));
            return value;
        })
        .buffer(5)
        .flatMapIterable(Function.identity())
        .subscribe(value -> System.out.println(String.format("  ---> %s", value)));

System.out.println("Exiting");