io.reactivex.parallel

Class ParallelFlowable<T>

java.lang.Object

io.reactivex.parallel.ParallelFlowable<T>

Type Parameters:

T - the value type

public abstract class ParallelFlowable<T>
extends java.lang.Object

Abstract base class for Parallel publishers that take an array of Subscribers.

Use from() to start processing a regular Publisher in 'rails'. Use runOn() to introduce where each 'rail' shoud run on thread-vise. Use sequential() to merge the sources back into a single Flowable.

Since:

2.0.5 - experimental

Constructor Summary

Constructors

Constructor and Description
ParallelFlowable()

Method Summary

Methods

Modifier and Type	Method and Description
<C> ParallelFlowable<C>	collect(java.util.concurrent.Callable<? extends C> collectionSupplier, BiConsumer<? super C,? super T> collector) Collect the elements in each rail into a collection supplied via a collectionSupplier and collected into with a collector action, emitting the collection at the end.
<U> ParallelFlowable<U>	compose(Function<? super ParallelFlowable<T>,ParallelFlowable<U>> composer) Allows composing operators, in assembly time, on top of this ParallelFlowable and returns another ParallelFlowable with composed features.
<R> ParallelFlowable<R>	concatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper) Generates and concatenates Publishers on each 'rail', signalling errors immediately and generating 2 publishers upfront.
<R> ParallelFlowable<R>	concatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper, int prefetch) Generates and concatenates Publishers on each 'rail', signalling errors immediately and using the given prefetch amount for generating Publishers upfront.
<R> ParallelFlowable<R>	concatMapDelayError(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper, boolean tillTheEnd) Generates and concatenates Publishers on each 'rail', optionally delaying errors and generating 2 publishers upfront.
<R> ParallelFlowable<R>	concatMapDelayError(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper, int prefetch, boolean tillTheEnd) Generates and concatenates Publishers on each 'rail', optionally delaying errors and using the given prefetch amount for generating Publishers upfront.
ParallelFlowable<T>	doAfterNext(Consumer<? super T> onAfterNext) Call the specified consumer with the current element passing through any 'rail' after it has been delivered to downstream within the rail.
ParallelFlowable<T>	doAfterTerminated(Action onAfterTerminate) Run the specified Action when a 'rail' completes or signals an error.
ParallelFlowable<T>	doOnCancel(Action onCancel) Run the specified Action when a 'rail' receives a cancellation.
ParallelFlowable<T>	doOnComplete(Action onComplete) Run the specified Action when a 'rail' completes.
ParallelFlowable<T>	doOnError(Consumer<java.lang.Throwable> onError) Call the specified consumer with the exception passing through any 'rail'.
ParallelFlowable<T>	doOnNext(Consumer<? super T> onNext) Call the specified consumer with the current element passing through any 'rail'.
ParallelFlowable<T>	doOnRequest(LongConsumer onRequest) Call the specified consumer with the request amount if any rail receives a request.
ParallelFlowable<T>	doOnSubscribe(Consumer<? super org.reactivestreams.Subscription> onSubscribe) Call the specified callback when a 'rail' receives a Subscription from its upstream.
ParallelFlowable<T>	filter(Predicate<? super T> predicate) Filters the source values on each 'rail'.
<R> ParallelFlowable<R>	flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper) Generates and flattens Publishers on each 'rail'.
<R> ParallelFlowable<R>	flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper, boolean delayError) Generates and flattens Publishers on each 'rail', optionally delaying errors.
<R> ParallelFlowable<R>	flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper, boolean delayError, int maxConcurrency) Generates and flattens Publishers on each 'rail', optionally delaying errors and having a total number of simultaneous subscriptions to the inner Publishers.
<R> ParallelFlowable<R>	flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper, boolean delayError, int maxConcurrency, int prefetch) Generates and flattens Publishers on each 'rail', optionally delaying errors, having a total number of simultaneous subscriptions to the inner Publishers and using the given prefetch amount for the inner Publishers.
static <T> ParallelFlowable<T>	from(org.reactivestreams.Publisher<? extends T> source) Take a Publisher and prepare to consume it on multiple 'rails' (number of CPUs) in a round-robin fashion.
static <T> ParallelFlowable<T>	from(org.reactivestreams.Publisher<? extends T> source, int parallelism) Take a Publisher and prepare to consume it on parallallism number of 'rails' in a round-robin fashion.
static <T> ParallelFlowable<T>	from(org.reactivestreams.Publisher<? extends T> source, int parallelism, int prefetch) Take a Publisher and prepare to consume it on parallallism number of 'rails' , possibly ordered and round-robin fashion and use custom prefetch amount and queue for dealing with the source Publisher's values.
static <T> ParallelFlowable<T>	fromArray(org.reactivestreams.Publisher<T>... publishers) Wraps multiple Publishers into a ParallelFlowable which runs them in parallel and unordered.
<R> ParallelFlowable<R>	map(Function<? super T,? extends R> mapper) Maps the source values on each 'rail' to another value.
abstract int	parallelism() Returns the number of expected parallel Subscribers.
Flowable<T>	reduce(BiFunction<T,T,T> reducer) Reduces all values within a 'rail' and across 'rails' with a reducer function into a single sequential value.
<R> ParallelFlowable<R>	reduce(java.util.concurrent.Callable<R> initialSupplier, BiFunction<R,? super T,R> reducer) Reduces all values within a 'rail' to a single value (with a possibly different type) via a reducer function that is initialized on each rail from an initialSupplier value.
ParallelFlowable<T>	runOn(Scheduler scheduler) Specifies where each 'rail' will observe its incoming values with no work-stealing and default prefetch amount.
ParallelFlowable<T>	runOn(Scheduler scheduler, int prefetch) Specifies where each 'rail' will observe its incoming values with possibly work-stealing and a given prefetch amount.
Flowable<T>	sequential() Merges the values from each 'rail' in a round-robin or same-order fashion and exposes it as a regular Publisher sequence, running with a default prefetch value for the rails.
Flowable<T>	sequential(int prefetch) Merges the values from each 'rail' in a round-robin or same-order fashion and exposes it as a regular Publisher sequence, running with a give prefetch value for the rails.
Flowable<T>	sorted(java.util.Comparator<? super T> comparator) Sorts the 'rails' of this ParallelFlowable and returns a Publisher that sequentially picks the smallest next value from the rails.
Flowable<T>	sorted(java.util.Comparator<? super T> comparator, int capacityHint) Sorts the 'rails' of this ParallelFlowable and returns a Publisher that sequentially picks the smallest next value from the rails.
abstract void	subscribe(org.reactivestreams.Subscriber<? super T>[] subscribers) Subscribes an array of Subscribers to this ParallelFlowable and triggers the execution chain for all 'rails'.
<U> U	to(Function<? super ParallelFlowable<T>,U> converter) Perform a fluent transformation to a value via a converter function which receives this ParallelFlowable.
Flowable<java.util.List<T>>	toSortedList(java.util.Comparator<? super T> comparator) Sorts the 'rails' according to the comparator and returns a full sorted list as a Publisher.
Flowable<java.util.List<T>>	toSortedList(java.util.Comparator<? super T> comparator, int capacityHint) Sorts the 'rails' according to the comparator and returns a full sorted list as a Publisher.
protected boolean	validate(org.reactivestreams.Subscriber<?>[] subscribers) Validates the number of subscribers and returns true if their number matches the parallelism level of this ParallelFlowable.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

ParallelFlowable

public ParallelFlowable()

Method Detail

subscribe

public abstract void subscribe(org.reactivestreams.Subscriber<? super T>[] subscribers)

Subscribes an array of Subscribers to this ParallelFlowable and triggers the execution chain for all 'rails'.

Parameters:

subscribers - the subscribers array to run in parallel, the number of items must be equal to the parallelism level of this ParallelFlowable

See Also:

parallelism()

parallelism

public abstract int parallelism()

Returns the number of expected parallel Subscribers.

Returns:

the number of expected parallel Subscribers

validate

protected final boolean validate(org.reactivestreams.Subscriber<?>[] subscribers)

Validates the number of subscribers and returns true if their number matches the parallelism level of this ParallelFlowable.

Parameters:

subscribers - the array of Subscribers

Returns:

true if the number of subscribers equals to the parallelism level

from

@CheckReturnValue
public static <T> ParallelFlowable<T> from(org.reactivestreams.Publisher<? extends T> source)

Take a Publisher and prepare to consume it on multiple 'rails' (number of CPUs) in a round-robin fashion.

Type Parameters:

T - the value type

Parameters:

source - the source Publisher

Returns:

the ParallelFlowable instance

from

@CheckReturnValue
public static <T> ParallelFlowable<T> from(org.reactivestreams.Publisher<? extends T> source,
                                            int parallelism)

Take a Publisher and prepare to consume it on parallallism number of 'rails' in a round-robin fashion.

Type Parameters:

T - the value type

Parameters:

source - the source Publisher

parallelism - the number of parallel rails

Returns:

the new ParallelFlowable instance

from

@CheckReturnValue
public static <T> ParallelFlowable<T> from(org.reactivestreams.Publisher<? extends T> source,
                                            int parallelism,
                                            int prefetch)

Take a Publisher and prepare to consume it on parallallism number of 'rails' , possibly ordered and round-robin fashion and use custom prefetch amount and queue for dealing with the source Publisher's values.

Type Parameters:

T - the value type

Parameters:

source - the source Publisher

parallelism - the number of parallel rails

prefetch - the number of values to prefetch from the source the source until there is a rail ready to process it.

Returns:

the new ParallelFlowable instance

map

@CheckReturnValue
public final <R> ParallelFlowable<R> map(Function<? super T,? extends R> mapper)

Maps the source values on each 'rail' to another value.

Note that the same mapper function may be called from multiple threads concurrently.

Type Parameters:

R - the output value type

Parameters:

mapper - the mapper function turning Ts into Us.

Returns:

the new ParallelFlowable instance

filter

@CheckReturnValue
public final ParallelFlowable<T> filter(Predicate<? super T> predicate)

Filters the source values on each 'rail'.

Note that the same predicate may be called from multiple threads concurrently.

Parameters:

predicate - the function returning true to keep a value or false to drop a value

Returns:

the new ParallelFlowable instance

runOn

@CheckReturnValue
public final ParallelFlowable<T> runOn(Scheduler scheduler)

Specifies where each 'rail' will observe its incoming values with no work-stealing and default prefetch amount.

This operator uses the default prefetch size returned by Flowable.bufferSize().

The operator will call Scheduler.createWorker() as many times as this ParallelFlowable's parallelism level is.

No assumptions are made about the Scheduler's parallelism level, if the Scheduler's parallelism level is lower than the ParallelFlowable's, some rails may end up on the same thread/worker.

This operator doesn't require the Scheduler to be trampolining as it does its own built-in trampolining logic.

Parameters:

scheduler - the scheduler to use

Returns:

the new ParallelFlowable instance

runOn

@CheckReturnValue
public final ParallelFlowable<T> runOn(Scheduler scheduler,
                                         int prefetch)

Specifies where each 'rail' will observe its incoming values with possibly work-stealing and a given prefetch amount.

This operator uses the default prefetch size returned by Flowable.bufferSize().

The operator will call Scheduler.createWorker() as many times as this ParallelFlowable's parallelism level is.

No assumptions are made about the Scheduler's parallelism level, if the Scheduler's parallelism level is lower than the ParallelFlowable's, some rails may end up on the same thread/worker.

This operator doesn't require the Scheduler to be trampolining as it does its own built-in trampolining logic.

Parameters:

scheduler - the scheduler to use that rail's worker has run out of work.

prefetch - the number of values to request on each 'rail' from the source

Returns:

the new ParallelFlowable instance

reduce

@CheckReturnValue
public final Flowable<T> reduce(BiFunction<T,T,T> reducer)

Reduces all values within a 'rail' and across 'rails' with a reducer function into a single sequential value.

Note that the same reducer function may be called from multiple threads concurrently.

Parameters:

reducer - the function to reduce two values into one.

Returns:

the new Flowable instance emitting the reduced value or empty if the ParallelFlowable was empty

reduce

@CheckReturnValue
public final <R> ParallelFlowable<R> reduce(java.util.concurrent.Callable<R> initialSupplier,
                                              BiFunction<R,? super T,R> reducer)

Reduces all values within a 'rail' to a single value (with a possibly different type) via a reducer function that is initialized on each rail from an initialSupplier value.

Note that the same mapper function may be called from multiple threads concurrently.

Type Parameters:

R - the reduced output type

Parameters:

initialSupplier - the supplier for the initial value

reducer - the function to reduce a previous output of reduce (or the initial value supplied) with a current source value.

Returns:

the new ParallelFlowable instance

sequential

@BackpressureSupport(value=FULL)
@SchedulerSupport(value="none")
@CheckReturnValue
public final Flowable<T> sequential()

Merges the values from each 'rail' in a round-robin or same-order fashion and exposes it as a regular Publisher sequence, running with a default prefetch value for the rails.

This operator uses the default prefetch size returned by Flowable.bufferSize().

Backpressure:

The operator honors backpressure.

Scheduler:

sequential does not operate by default on a particular Scheduler.

Returns:

the new Flowable instance

See Also:

sequential(int)

sequential

@BackpressureSupport(value=FULL)
@SchedulerSupport(value="none")
@CheckReturnValue
public final Flowable<T> sequential(int prefetch)

Merges the values from each 'rail' in a round-robin or same-order fashion and exposes it as a regular Publisher sequence, running with a give prefetch value for the rails.

Backpressure:

The operator honors backpressure.

Scheduler:

sequential does not operate by default on a particular Scheduler.

Parameters:

prefetch - the prefetch amount to use for each rail

Returns:

the new Flowable instance

See Also:

sequential()

sorted

@CheckReturnValue
public final Flowable<T> sorted(java.util.Comparator<? super T> comparator)

Sorts the 'rails' of this ParallelFlowable and returns a Publisher that sequentially picks the smallest next value from the rails.

This operator requires a finite source ParallelFlowable.

Parameters:

comparator - the comparator to use

Returns:

the new Flowable instance

sorted

@CheckReturnValue
public final Flowable<T> sorted(java.util.Comparator<? super T> comparator,
                                  int capacityHint)

Sorts the 'rails' of this ParallelFlowable and returns a Publisher that sequentially picks the smallest next value from the rails.

This operator requires a finite source ParallelFlowable.

Parameters:

comparator - the comparator to use

capacityHint - the expected number of total elements

Returns:

the new Flowable instance

toSortedList

@CheckReturnValue
public final Flowable<java.util.List<T>> toSortedList(java.util.Comparator<? super T> comparator)

Sorts the 'rails' according to the comparator and returns a full sorted list as a Publisher.

This operator requires a finite source ParallelFlowable.

Parameters:

comparator - the comparator to compare elements

Returns:

the new Px instannce

toSortedList

@CheckReturnValue
public final Flowable<java.util.List<T>> toSortedList(java.util.Comparator<? super T> comparator,
                                                        int capacityHint)

Sorts the 'rails' according to the comparator and returns a full sorted list as a Publisher.

This operator requires a finite source ParallelFlowable.

Parameters:

comparator - the comparator to compare elements

capacityHint - the expected number of total elements

Returns:

the new Px instannce

doOnNext

@CheckReturnValue
public final ParallelFlowable<T> doOnNext(Consumer<? super T> onNext)

Call the specified consumer with the current element passing through any 'rail'.

Parameters:

onNext - the callback

Returns:

the new ParallelFlowable instance

doAfterNext

@CheckReturnValue
public final ParallelFlowable<T> doAfterNext(Consumer<? super T> onAfterNext)

Call the specified consumer with the current element passing through any 'rail' after it has been delivered to downstream within the rail.

Parameters:

onAfterNext - the callback

Returns:

the new ParallelFlowable instance

doOnError

@CheckReturnValue
public final ParallelFlowable<T> doOnError(Consumer<java.lang.Throwable> onError)

Call the specified consumer with the exception passing through any 'rail'.

Parameters:

onError - the callback

Returns:

the new ParallelFlowable instance

doOnComplete

@CheckReturnValue
public final ParallelFlowable<T> doOnComplete(Action onComplete)

Run the specified Action when a 'rail' completes.

Parameters:

onComplete - the callback

Returns:

the new ParallelFlowable instance

doAfterTerminated

@CheckReturnValue
public final ParallelFlowable<T> doAfterTerminated(Action onAfterTerminate)

Run the specified Action when a 'rail' completes or signals an error.

Parameters:

onAfterTerminate - the callback

Returns:

the new ParallelFlowable instance

doOnSubscribe

@CheckReturnValue
public final ParallelFlowable<T> doOnSubscribe(Consumer<? super org.reactivestreams.Subscription> onSubscribe)

Call the specified callback when a 'rail' receives a Subscription from its upstream.

Parameters:

onSubscribe - the callback

Returns:

the new ParallelFlowable instance

doOnRequest

@CheckReturnValue
public final ParallelFlowable<T> doOnRequest(LongConsumer onRequest)

Call the specified consumer with the request amount if any rail receives a request.

Parameters:

onRequest - the callback

Returns:

the new ParallelFlowable instance

doOnCancel

@CheckReturnValue
public final ParallelFlowable<T> doOnCancel(Action onCancel)

Run the specified Action when a 'rail' receives a cancellation.

Parameters:

onCancel - the callback

Returns:

the new ParallelFlowable instance

collect

@CheckReturnValue
public final <C> ParallelFlowable<C> collect(java.util.concurrent.Callable<? extends C> collectionSupplier,
                                               BiConsumer<? super C,? super T> collector)

Collect the elements in each rail into a collection supplied via a collectionSupplier and collected into with a collector action, emitting the collection at the end.

Type Parameters:

C - the collection type

Parameters:

collectionSupplier - the supplier of the collection in each rail

collector - the collector, taking the per-rali collection and the current item

Returns:

the new ParallelFlowable instance

fromArray

@CheckReturnValue
public static <T> ParallelFlowable<T> fromArray(org.reactivestreams.Publisher<T>... publishers)

Wraps multiple Publishers into a ParallelFlowable which runs them in parallel and unordered.

Type Parameters:

T - the value type

Parameters:

publishers - the array of publishers

Returns:

the new ParallelFlowable instance

to

@CheckReturnValue
public final <U> U to(Function<? super ParallelFlowable<T>,U> converter)

Perform a fluent transformation to a value via a converter function which receives this ParallelFlowable.

Type Parameters:

U - the output value type

Parameters:

converter - the converter function from ParallelFlowable to some type

Returns:

the value returned by the converter function

compose

@CheckReturnValue
public final <U> ParallelFlowable<U> compose(Function<? super ParallelFlowable<T>,ParallelFlowable<U>> composer)

Allows composing operators, in assembly time, on top of this ParallelFlowable and returns another ParallelFlowable with composed features.

Type Parameters:

U - the output value type

Parameters:

composer - the composer function from ParallelFlowable (this) to another ParallelFlowable

Returns:

the ParallelFlowable returned by the function

flatMap

@CheckReturnValue
public final <R> ParallelFlowable<R> flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper)

Generates and flattens Publishers on each 'rail'.

Errors are not delayed and uses unbounded concurrency along with default inner prefetch.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher

Returns:

the new ParallelFlowable instance

flatMap

@CheckReturnValue
public final <R> ParallelFlowable<R> flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper,
                                               boolean delayError)

Generates and flattens Publishers on each 'rail', optionally delaying errors.

It uses unbounded concurrency along with default inner prefetch.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher

delayError - should the errors from the main and the inner sources delayed till everybody terminates?

Returns:

the new ParallelFlowable instance

flatMap

@CheckReturnValue
public final <R> ParallelFlowable<R> flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper,
                                               boolean delayError,
                                               int maxConcurrency)

Generates and flattens Publishers on each 'rail', optionally delaying errors and having a total number of simultaneous subscriptions to the inner Publishers.

It uses a default inner prefetch.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher

delayError - should the errors from the main and the inner sources delayed till everybody terminates?

maxConcurrency - the maximum number of simultaneous subscriptions to the generated inner Publishers

Returns:

the new ParallelFlowable instance

flatMap

@CheckReturnValue
public final <R> ParallelFlowable<R> flatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper,
                                               boolean delayError,
                                               int maxConcurrency,
                                               int prefetch)

Generates and flattens Publishers on each 'rail', optionally delaying errors, having a total number of simultaneous subscriptions to the inner Publishers and using the given prefetch amount for the inner Publishers.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher

delayError - should the errors from the main and the inner sources delayed till everybody terminates?

maxConcurrency - the maximum number of simultaneous subscriptions to the generated inner Publishers

prefetch - the number of items to prefetch from each inner Publisher

Returns:

the new ParallelFlowable instance

concatMap

@CheckReturnValue
public final <R> ParallelFlowable<R> concatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper)

Generates and concatenates Publishers on each 'rail', signalling errors immediately and generating 2 publishers upfront.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher source and the inner Publishers (immediate, boundary, end)

Returns:

the new ParallelFlowable instance

concatMap

@CheckReturnValue
public final <R> ParallelFlowable<R> concatMap(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper,
                                                 int prefetch)

Generates and concatenates Publishers on each 'rail', signalling errors immediately and using the given prefetch amount for generating Publishers upfront.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher

prefetch - the number of items to prefetch from each inner Publisher source and the inner Publishers (immediate, boundary, end)

Returns:

the new ParallelFlowable instance

concatMapDelayError

@CheckReturnValue
public final <R> ParallelFlowable<R> concatMapDelayError(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper,
                                                           boolean tillTheEnd)

Generates and concatenates Publishers on each 'rail', optionally delaying errors and generating 2 publishers upfront.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher

tillTheEnd - if true all errors from the upstream and inner Publishers are delayed till all of them terminate, if false, the error is emitted when an inner Publisher terminates. source and the inner Publishers (immediate, boundary, end)

Returns:

the new ParallelFlowable instance

concatMapDelayError

@CheckReturnValue
public final <R> ParallelFlowable<R> concatMapDelayError(Function<? super T,? extends org.reactivestreams.Publisher<? extends R>> mapper,
                                                           int prefetch,
                                                           boolean tillTheEnd)

Generates and concatenates Publishers on each 'rail', optionally delaying errors and using the given prefetch amount for generating Publishers upfront.

Type Parameters:

R - the result type

Parameters:

mapper - the function to map each rail's value into a Publisher

prefetch - the number of items to prefetch from each inner Publisher

tillTheEnd - if true all errors from the upstream and inner Publishers are delayed till all of them terminate, if false, the error is emitted when an inner Publisher terminates.

Returns:

the new ParallelFlowable instance