Remove covariance of scan/reduce

[benjchristensen]

Remove covariance of scan/reduce since consumer/producer are the same <T, T, T>

See #360 (comment)

[cloudbees-pull-request-builder]

RxJava-pull-requests #263 SUCCESS
This pull request looks good

[abersnaze]

This method and friends
public <R> Observable<R> reduce(R initialValue, Func2<? super R, ? super T, ? extends R> accumulator) {
also needs to changed to
public <R> Observable<R> reduce(R initialValue, Func2<R, ? super T, R> accumulator) {

T2 should remain "? super T" because it is only consumed.

[benjchristensen]

I didn't change those yet as I wasn't sure if they behave differently because they consume and produce different types. I wasn't able to get an example that proved or disproved anything so left it until I can.

Do you have example code that shows that method doesn't benefit from ? super T?

And what T2 are you referring to? There isn't anything with a type T2 in reduce.

[abersnaze]

Func2<T1, T2, R>

[benjchristensen]

(George and I chatted in person)

We looked at JDK 8 to try and get a final representation of others doing this and it does it like this:

<U> U reduce(U identity,
             BiFunction<U,? super T,U> accumulator,
             BinaryOperator<U> combiner)

This is at: http://download.java.net/jdk8/docs/api/java/util/stream/Stream.html#reduce(U,%20java.util.function.BiFunction,%20java.util.function.BinaryOperator)

Optional<T> reduce(BinaryOperator<T> accumulator)

The BinaryOperator is this:

public interface BinaryOperator<T>
extends BiFunction<T,T,T>

Thus this seems to confirm the decision to remove ? super in reduce without a seed and in the one with a seed value do what George recommended:

reduce(R initialValue, Func2<R, ? super T, R> accumulator)

[cloudbees-pull-request-builder]

RxJava-pull-requests #267 SUCCESS
This pull request looks good