为什么在此代码中不调用onCompleted？

Question

我在viewmodel中有这个主题:

private PublishSubject<String> articleSubject;

public Observable<Article> newArticleSubject() {
    articleSubject = PublishSubject.create();
    return articleSubject.flatMap(new Func1<String, Observable<Article>>() {         
        @Override
        public Observable<Article> call(String articleId) {
            return dataModel.getArticleById(articleId).subscribeOn(Schedulers.newThread());
        }
    });
}

我像这样绑定视图:

viewModel.newArticleSubject()
    .observeOn(AndroidSchedulers.mainThread())
    .subscribe(new ArticleSubscriber());

这是我的订阅者:

 private class ArticleSubscriber extends Subscriber<Article> {

        @Override
        public void onCompleted() {
            Log.d("test", "This is not ever printed");
        }

        @Override
        public void onError(Throwable e) {
            //do stuff
        }

        @Override
        public void onNext(Article article) {
            //do stuff
        }
}

onCompleted永远不会被调用,但onError并onNext是.为什么？在里面dataModel.getArticleById(articleId),subscriber.onCompleted();被称为.

Answer 1

从PublishSubject文档中可用的图像可以得出结论:

PublishSubject以及其他*Subjects将生活,直到有人为来电onError或onCompleted.事实上,他们在这件事上与其他任何事物都没有什么不同Observable.

在newArticleSubject创建新的PublishSubject和平面地图发射到不同的可观察到的每一个项目.如果getArticleById结束有错误,则flatMap结束返回的observable 也是如此.但是,当getArticleById正常发射和完成时,flatMap将继续等待新项目发出articleSubject.如果你需要明确地结束articleSubject只是打电话articleSubject.onCompleted()