Add more operators to RxScala

[zsxwing]

The commits explain this PR.

In addition, the reason that I changed window(closings: () => Observable[Any]) to window(boundary: => Observable[Any]) is window(boundary: => Observable[Any]) can support both two overloads window(Observable[U]) and window(Func0[_ <: Observable[_ <: TClosing]]) in RxJava.

[zsxwing]

/cc @samuelgruetter, @headinthebox

[cloudbees-pull-request-builder]

RxJava-pull-requests #1160 FAILURE
Looks like there's a problem with this pull request

[headinthebox]

Thanks @zsxwing.

Probably a good idea to look for more places where we can use =>{} and curry it as the last parameter; but I guess there are not that many places to do that if anay..

[headinthebox]

The code looks fine to me. Can't see quickly howthe cloudbees failure connected to the code in this pull request, though.

[samuelgruetter]

Currently, there's no such thing as o1 amb o2. You could add this. And then, maybe no varargs, but Iterable for the one in the companion object, because the typical use case of 2 Observables would be covered already.

[zsxwing]

Already have this instance method: def amb[U >: T](that: Observable[U]): Observable[U] in RxScala

[samuelgruetter]

The two Java methods

public final <TClosing> Observable<Observable<T>> window(Func0<? extends Observable<? extends TClosing>> closingSelector)
public final <U> Observable<Observable<T>> window(Observable<U> boundary)

have very different semantics, and both of them are useful and should be present in the RxScala. Since we cannot have an overload which takes an Observable[Any] and one which takes an => Observable[Any], we should use () => Observable[Any] (or rename one of the functions?).

The same applies to buffer, and we should keep buffer and window consistent.

[headinthebox]

Of course, main pitfall of =>T

[zsxwing]

have very different semantics

In RxJava, it's

    public OperatorWindowWithObservable(Func0<? extends Observable<? extends U>> otherFactory) {
        this.otherFactory = otherFactory;
    }
    public OperatorWindowWithObservable(final Observable<U> other) {
        this.otherFactory = new Func0<Observable<U>>() {

            @Override
            public Observable<U> call() {
                return other;
            }

        };
    }

Looks => can cover these two use cases. Any difference I miss?

[zsxwing]

I just noticed that this window method does not what it should do: When I tried to use it, I got a "java.lang.ClassCastException: rx.operators.BufferUntilSubscriber cannot be cast to java.lang.Long". Can you please rewrite this method from scratch? No idea what I was thinking when I wrote this method ;-)
And can you also add an example to RxScalaDemo?

OK.

[cloudbees-pull-request-builder]

RxJava-pull-requests #1162 SUCCESS
This pull request looks good

[samuelgruetter]

Looks => can cover these two use cases. Any difference I miss?

I think the javadoc does not correspond to the implementation...

[zsxwing]

Probably a good idea to look for more places where we can use =>{} and curry it as the last parameter; but I guess there are not that many places to do that if anay..

The only place I found is "buffer". Already updated it.

[headinthebox]

Cool.

[cloudbees-pull-request-builder]

RxJava-pull-requests #1173 SUCCESS
This pull request looks good

[zsxwing]

@DavidMGross, could you help improve the javadoc for buffer(Func0<? extends Observable<? extends TClosing>> bufferClosingSelector) and window(Func0<? extends Observable<? extends TClosing>> closingSelector)? The current docs are misleading. Func0 is actually only called once to generate an Observable which is used to determined the boundary of each window/buffer. Also the marble diagrams:) Thank you.

[DavidMGross]

Will do.

On Wed, May 28, 2014 at 8:43 AM, Shixiong Zhu [email protected]:

@DavidMGross https://github.com/DavidMGross, could you help improve the
javadoc for buffer(Func0<? extends Observable<? extends TClosing>>
bufferClosingSelector) and window(Func0<? extends Observable<? extends
TClosing>> closingSelector)? The current docs are misleading. Func0 is
actually only called once to generate an Observable which is used to
determined the boundary of each window/buffer.

—
Reply to this email directly or view it on GitHubhttps://github.com//pull/1265#issuecomment-44424293
.

David M. Gross
PLP Consulting

[benjchristensen]

Is this ready for merging?

[headinthebox]

Looks good to go to me.

[samuelgruetter]

IMHO, it makes more sense to say that the javadoc is correct, but the implementation of window(Func0) is wrong. For future reference, here's the current javadoc that I think is correct:

/**
 * Returns an Observable that emits windows of items it collects from the source Observable. The resulting
 * Observable emits connected, non-overlapping windows. It emits the current window and opens a new one when
 * the Observable produced by the specified {@code closingSelector} emits an item. The {@code closingSelector} 
 * then creates a new Observable to generate the closer of the next window.
 * <p>
 * <img width="640" src="https://raw.github.com/wiki/Netflix/RxJava/images/rx-operators/window1.png">
 * 
 * @param closingSelector
 *            a {@link Func0} that produces an Observable for every window created. When this Observable
 *            emits an item, {@code window()} emits the associated window and begins a new one.
 * @return an Observable that emits connected, non-overlapping windows of items from the source Observable
 *         when {@code closingSelector} emits an item
 * @see <a href="https://github.com/Netflix/RxJava/wiki/Transforming-Observables#wiki-window">RxJava Wiki: window()</a>
 */
public final <TClosing> Observable<Observable<T>> window(Func0<? extends Observable<? extends TClosing>> closingSelector)

/**
 * Returns an Observable that emits non-overlapping windows of items it collects from the source Observable
 * where the boundary of each window is determined by the items emitted from a specified boundary-governing
 * Observable.
 * <p>
 * <img width="640" src="https://raw.github.com/wiki/Netflix/RxJava/images/rx-operators/window8.png">
 * 
 * @param <U>
 *            the window element type (ignored)
 * @param boundary
 *            an Observable whose emitted items close and open windows
 * @return an Observable that emits non-overlapping windows of items it collects from the source Observable
 *         where the boundary of each window is determined by the items emitted from the {@code boundary} Observable
 */
public final <U> Observable<Observable<T>> window(Observable<U> boundary)

@headinthebox @benjchristensen what do you think?

[zsxwing]

I think

public final <A, B> Observable<Observable<T>> window(Func0<? extends Observable<? extends A>> firstClosingSelector, Func1<? super T, ? extends Observable<? extends B>> closingSelector)

is better.

However, from MSDN: http://msdn.microsoft.com/en-us/library/hh229909(v=vs.103).aspx , the current implementation in RxJava is same as Rx.Net.