A monad that can perform concurrent or parallel IO operations. Streams that can be composed concurrently require the underlying monad to be MonadAsync.

An abstract type for specifying the configuration parameters of a Channel. Use Config -> Config modifier functions to modify the default configuration. See the individual modifier documentation for default values.

Specify the maximum number of threads that can be spawned by the channel. A value of 0 resets the thread limit to default, a negative value means there is no limit. The default value is 1500.

When the actions in a stream are IO bound, having blocking IO calls, this option can be used to control the maximum number of in-flight IO requests. When the actions are CPU bound this option can be used to control the amount of CPU used by the stream.

Specify the maximum size of the buffer for storing the results from concurrent computations. If the buffer becomes full we stop spawning more concurrent tasks until there is space in the buffer. A value of 0 resets the buffer size to default, a negative value means there is no limit. The default value is 1500.

CAUTION! using an unbounded maxBuffer value (i.e. a negative value) coupled with an unbounded maxThreads value is a recipe for disaster in presence of infinite streams, or very large streams. Especially, it must not be used when pure is used in ZipAsyncM streams as pure in applicative zip streams generates an infinite stream causing unbounded concurrent generation with no limit on the buffer or threads.

Specifies the stream yield rate in yields per second (Hertz). We keep accumulating yield credits at rateGoal. At any point of time we allow only as many yields as we have accumulated as per rateGoal since the start of time. If the consumer or the producer is slower or faster, the actual rate may fall behind or exceed rateGoal. We try to recover the gap between the two by increasing or decreasing the pull rate from the producer. However, if the gap becomes more than rateBuffer we try to recover only as much as rateBuffer.

rateLow puts a bound on how low the instantaneous rate can go when recovering the rate gap. In other words, it determines the maximum yield latency. Similarly, rateHigh puts a bound on how high the instantaneous rate can go when recovering the rate gap. In other words, it determines the minimum yield latency. We reduce the latency by increasing concurrency, therefore we can say that it puts an upper bound on concurrency.

If the rateGoal is 0 or negative the stream never yields a value. If the rateBuffer is 0 or negative we do not attempt to recover.

Specify the stream evaluation rate of a channel.

A Nothing value means there is no smart rate control, concurrent execution blocks only if maxThreads or maxBuffer is reached, or there are no more concurrent tasks to execute. This is the default.

When rate (throughput) is specified, concurrent production may be ramped up or down automatically to achieve the specified stream throughput. The specific behavior for different styles of Rate specifications is documented under Rate. The effective maximum production rate achieved by a channel is governed by:

• The maxThreads limit
• The maxBuffer limit
• The maximum rate that the stream producer can achieve
• The maximum rate that the stream consumer can achieve

Maximum production rate is given by:

\(rate = \frac{maxThreads}{latency}\)

If we know the average latency of the tasks we can set maxThreads accordingly.

Same as rate (Just $ Rate (r/2) r (2*r) maxBound)

Specifies the average production rate of a stream in number of yields per second (i.e. Hertz). Concurrent production is ramped up or down automatically to achieve the specified average yield rate. The rate can go down to half of the specified rate on the lower side and double of the specified rate on the higher side.

Same as rate (Just $ Rate r r (2*r) maxBound)

Specifies the minimum rate at which the stream should yield values. As far as possible the yield rate would never be allowed to go below the specified rate, even though it may possibly go above it at times, the upper limit is double of the specified rate.

Same as rate (Just $ Rate (r/2) r r maxBound)

Specifies the maximum rate at which the stream should yield values. As far as possible the yield rate would never be allowed to go above the specified rate, even though it may possibly go below it at times, the lower limit is half of the specified rate. This can be useful in applications where certain resource usage must not be allowed to go beyond certain limits.

Same as rate (Just $ Rate r r r 0)

Specifies a constant yield rate. If for some reason the actual rate goes above or below the specified rate we do not try to recover it by increasing or decreasing the rate in future. This can be useful in applications like graphics frame refresh where we need to maintain a constant refresh rate.

Specify when the Channel should stop.

Specify when the Channel should stop.

By default, processing of output from the worker threads is given priority over dispatching new workers. More workers are dispatched only when there is no output to process. When eager is set to True, workers are dispatched aggresively as long as there is more work to do irrespective of whether there is output pending to be processed by the stream consumer. However, dispatching may stop if maxThreads or maxBuffer is reached.

Note: This option has no effect when rate has been specified.

Note: Not supported with interleaved.

When enabled the streams may be evaluated cocnurrently but the results are produced in the same sequence as a serial evaluation would produce.

Note: Not supported with interleaved.

Interleave the streams fairly instead of prioritizing the left stream. This schedules all streams in a round robin fashion over limited number of threads.

Note: Can only be used on finite number of streams.

Note: Not supported with ordered.

Print debug information about the Channel when the stream ends.

parEval evaluates a stream as a whole asynchronously with respect to the consumer of the stream. A worker thread evaluates multiple elements of the stream ahead of time and buffers the results; the consumer of the stream runs in another thread consuming the elements from the buffer, thus decoupling the production and consumption of the stream. parEval can be used to run different stages of a pipeline concurrently.

It is important to note that parEval does not evaluate individual actions in the stream concurrently with respect to each other, it merely evaluates the stream serially but in a different thread than the consumer thread, thus the consumer and producer can run concurrently. See parMapM and parSequence to evaluate actions in the stream concurrently.

The evaluation requires only one thread as only one stream needs to be evaluated. Therefore, the concurrency options that are relevant to multiple streams do not apply here e.g. maxThreads, eager, interleaved, ordered, stopWhen options do not have any effect on parEval.

Useful idioms:

>>> parUnfoldrM step = Stream.parEval id . Stream.unfoldrM step
>>> parIterateM step = Stream.parEval id . Stream.iterateM step

Definition:

>>> parRepeatM cfg = Stream.parSequence cfg . Stream.repeat

Generate a stream by repeatedly executing a monadic action forever.

Generate a stream by concurrently performing a monadic action n times.

Definition:

>>> parReplicateM cfg n = Stream.parSequence cfg . Stream.replicate n

Example, parReplicateM in the following example executes all the replicated actions concurrently, thus taking only 1 second:

>>> Stream.fold Fold.drain $ Stream.parReplicateM id 10 $ delay 1
...

fromCallback f creates an entangled pair of a callback and a stream i.e. whenever the callback is called a value appears in the stream. The function f is invoked with the callback as argument, and the stream is returned. f would store the callback for calling it later for generating values in the stream.

The callback queues a value to a concurrent channel associated with the stream. The stream can be evaluated safely in any thread.

Pre-release

Definition:

>>> parMapM modifier f = Stream.parConcatMap modifier (Stream.fromEffect . f)

For example, the following finishes in 3 seconds (as opposed to 6 seconds) because all actions run in parallel. Even though results are available out of order they are ordered due to the config option:

>>> f x = delay x >> return x
>>> Stream.fold Fold.toList $ Stream.parMapM (Stream.ordered True) f $ Stream.fromList [3,2,1]
1 sec
2 sec
3 sec
[3,2,1]

Definition:

>>> parSequence modifier = Stream.parMapM modifier id

Useful idioms:

>>> parFromListM = Stream.parSequence id . Stream.fromList
>>> parFromFoldableM = Stream.parSequence id . StreamK.toStream . StreamK.fromFoldable

Evaluates the streams being zipped in separate threads than the consumer. The zip function is evaluated in the consumer thread.

>>> parZipWithM cfg f m1 m2 = Stream.zipWithM f (Stream.parEval cfg m1) (Stream.parEval cfg m2)

Multi-stream concurrency options won't apply here, see the notes in parEval.

If you want to evaluate the zip function as well in a separate thread, you can use a parEval on parZipWithM.

>>> parZipWith cfg f = Stream.parZipWithM cfg (\a b -> return $ f a b)

>>> m1 = Stream.fromList [1,2,3]
>>> m2 = Stream.fromList [4,5,6]
>>> Stream.fold Fold.toList $ Stream.parZipWith id (,) m1 m2
[(1,4),(2,5),(3,6)]

Like mergeByM but evaluates both the streams concurrently.

Definition:

>>> parMergeByM cfg f m1 m2 = Stream.mergeByM f (Stream.parEval cfg m1) (Stream.parEval cfg m2)

Like mergeBy but evaluates both the streams concurrently.

Definition:

>>> parMergeBy cfg f = Stream.parMergeByM cfg (\a b -> return $ f a b)

Like parConcat but works on a list of streams.

>>> parList modifier = Stream.parConcat modifier . Stream.fromList

Apply an argument stream to a function stream concurrently. Uses a shared channel for all individual applications within a stream application.

Evaluate the streams in the input stream concurrently and combine them.

>>> parConcat modifier = Stream.parConcatMap modifier id

Map each element of the input to a stream and then concurrently evaluate and concatenate the resulting streams. Multiple streams may be evaluated concurrently but earlier streams are perferred. Output from the streams are used as they arrive.

Definition:

>>> parConcatMap modifier f stream = Stream.parConcat modifier $ fmap f stream

Examples:

>>> f cfg xs = Stream.fold Fold.toList $ Stream.parConcatMap cfg id $ Stream.fromList xs

The following streams finish in 4 seconds:

>>> stream1 = Stream.fromEffect (delay 4)
>>> stream2 = Stream.fromEffect (delay 2)
>>> stream3 = Stream.fromEffect (delay 1)
>>> f id [stream1, stream2, stream3]
1 sec
2 sec
4 sec
[1,2,4]

Limiting threads to 2 schedules the third stream only after one of the first two has finished, releasing a thread:

>>> f (Stream.maxThreads 2) [stream1, stream2, stream3]
...
[2,1,4]

When used with a Single thread it behaves like serial concatMap:

>>> f (Stream.maxThreads 1) [stream1, stream2, stream3]
...
[4,2,1]

>>> stream1 = Stream.fromList [1,2,3]
>>> stream2 = Stream.fromList [4,5,6]
>>> f (Stream.maxThreads 1) [stream1, stream2]
[1,2,3,4,5,6]

Schedule all streams in a round robin fashion over the available threads:

>>> f cfg xs = Stream.fold Fold.toList $ Stream.parConcatMap (Stream.interleaved True . cfg) id $ Stream.fromList xs

>>> stream1 = Stream.fromList [1,2,3]
>>> stream2 = Stream.fromList [4,5,6]
>>> f (Stream.maxThreads 1) [stream1, stream2]
[1,4,2,5,3,6]

Same as concatIterate but concurrent.

Pre-release

parTapCount predicate fold stream taps the count of those elements in the stream that pass the predicate. The resulting count stream is sent to a fold running concurrently in another thread.

For example, to print the count of elements processed every second:

>>> rate = Stream.rollingMap2 (flip (-)) . Stream.delayPost 1
>>> report = Stream.fold (Fold.drainMapM print) . rate
>>> tap = Stream.parTapCount (const True) report
>>> go = Stream.fold Fold.drain $ tap $ Stream.enumerateFrom 0

Note: This may not work correctly on 32-bit machines because of Int overflow.

Pre-release

Intersperse a monadic action into the input stream after every n seconds.

Definition:

>>> interject n f xs = Stream.parListEagerFst [xs, Stream.periodic f n]

Example:

>>> s = Stream.fromList "hello"
>>> input = Stream.mapM (\x -> threadDelay 1000000 >> putChar x) s
>>> Stream.fold Fold.drain $ Stream.interject (putChar ',') 1.05 input
h,e,l,l,o

takeInterval interval runs the stream only upto the specified time interval in seconds.

The interval starts when the stream is evaluated for the first time.

dropInterval interval drops all the stream elements that are generated before the specified interval in seconds has passed.

The interval begins when the stream is evaluated for the first time.

Group the input stream into windows of n second each and then fold each group using the provided fold function.

>>> twoPerSec = Stream.parEval (Stream.constRate 2) $ Stream.enumerateFrom 1
>>> intervals = Stream.intervalsOf 1 Fold.toList twoPerSec
>>> Stream.fold Fold.toList $ Stream.take 2 intervals
[...,...]

Continuously evaluate the input stream and sample the last event in each time window of n seconds.

This is also known as throttle in some libraries.

>>> sampleIntervalEnd n = Stream.catMaybes . Stream.intervalsOf n Fold.latest

Like sampleInterval but samples at the beginning of the time window.

>>> sampleIntervalStart n = Stream.catMaybes . Stream.intervalsOf n Fold.one

Sample one event at the end of each burst of events. A burst is a group of events close together in time, it ends when an event is spaced by more than the specified time interval (in seconds) from the previous event.

This is known as debounce in some libraries.

The clock granularity is 10 ms.

Like sampleBurstEnd but samples the event at the beginning of the burst instead of at the end of it.

Run the action m b whenever the stream Stream m a stops normally, or if it is garbage collected after a partial lazy evaluation.

The semantics of the action m b are similar to the semantics of cleanup action in bracket.

See also after_

Run the alloc action IO b with async exceptions disabled but keeping blocking operations interruptible (see mask). Use the output b of the IO action as input to the function b -> Stream m a to generate an output stream.

b is usually a resource under the IO monad, e.g. a file handle, that requires a cleanup after use. The cleanup action b -> m c, runs whenever (1) the stream ends normally, (2) due to a sync or async exception or, (3) if it gets garbage collected after a partial lazy evaluation. The exception is not caught, it is rethrown.

bracket only guarantees that the cleanup action runs, and it runs with async exceptions enabled. The action must ensure that it can successfully cleanup the resource in the face of sync or async exceptions.

When the stream ends normally or on a sync exception, cleanup action runs immediately in the current thread context, whereas in other cases it runs in the GC context, therefore, cleanup may be delayed until the GC gets to run.

See also: bracket_

Inhibits stream fusion

Run the action m b whenever the stream Stream m a stops normally, aborts due to an exception or if it is garbage collected after a partial lazy evaluation.

The semantics of running the action m b are similar to the cleanup action semantics described in bracket.

>>> finally action xs = Stream.bracket (return ()) (const action) (const xs)

See also finally_

Inhibits stream fusion

Same as parTapCount. Deprecated.