3-- Copyright : (c) 2017 Harendra Kumar
6-- Maintainer : streamly@composewell.com
7-- Stability : experimental
10-- Streamly is a general purpose programming framework using cocnurrent data
11-- flow programming paradigm. It can be considered as a generalization of
12-- Haskell lists to monadic streaming with concurrent composition capability.
13-- The serial stream type in streamly @SerialT m a@ is like the list type @[a]@
14-- parameterized by the monad @m@. For example, @SerialT IO a@ is a moral
15-- equivalent of @[a]@ in the IO monad. Streams are constructed very much like
16-- lists, except that they use 'nil' and 'cons' instead of '[]' and ':'.
20-- > import "Streamly.Prelude" (cons, consM)
21-- > import qualified "Streamly.Prelude" as S
23-- > S.toList $ 1 \`cons` 2 \`cons` 3 \`cons` nil
27-- Unlike lists, streams can be constructed from monadic effects:
30-- > S.'toList' $ 'getLine' \`consM` 'getLine' \`consM` S.'nil'
36-- Streams are processed just like lists, with list like combinators, except
37-- that they are monadic and work in a streaming fashion. Here is a simple
38-- console echo program example:
41-- > S.drain $ S.repeatM getLine & S.mapM putStrLn
44-- @SerialT Identity a@ is a moral equivalent of pure lists. Streamly utilizes
45-- fusion for high performance, therefore, we can represent and process strings
46-- as streams of 'Char', encode and decode the streams to/from UTF8 and
47-- serialize them to @Array Word8@ obviating the need for special purpose
48-- libraries like @bytestring@ and @text@.
51-- please see the "Streamly.Tutorial" module and the examples directory in this
55{-# LANGUAGE FlexibleContexts #-}
56{-# LANGUAGE MultiParamTypeClasses #-}
58#if __GLASGOW_HASKELL__ >= 800
59{-# OPTIONS_GHC -Wno-orphans #-}
66 -- -- * Concepts Overview
82 -- * Stream transformers
83 -- | A stream represents a sequence of pure or effectful actions. The
84 -- `cons` and `consM` operations and the corresponding operators '.:' and
85 -- '|:' can be used to join pure values or effectful actions in a sequence.
86 -- The effects in the stream can be executed in many different ways
87 -- depending on the type of stream. In other words, the behavior of 'consM'
88 -- depends on the type of the stream.
90 -- There are three high level categories of streams, /spatially ordered/
91 -- streams, /speculative/ streams and /time ordered/ streams. Spatially
92 -- ordered streams, 'SerialT' and 'WSerialT', execute the effects in serial
93 -- order i.e. one at a time and present the outputs of those effects to the
94 -- consumer in the same order. Speculative streams, 'AheadT', may execute
95 -- many effects concurrently but present the outputs to the consumer in the
96 -- specified spatial order. Time ordered streams, 'AsyncT', 'WAsyncT' and
97 -- 'ParallelT', may execute many effects concurrently and present the
98 -- outputs of those effects to the consumer in time order i.e. as soon as
99 -- the output is generated.
101 -- We described above how the effects in a sequence are executed for
102 -- different types of streams. The behvavior of the 'Semigroup' and 'Monad'
103 -- instances follow the behavior of 'consM'. Stream generation operations
104 -- like 'repeatM' also execute the effects differently for different
105 -- streams, providing a concurrent generation capability when used with
106 -- stream types that execute effects concurrently. Similarly, effectful
107 -- transformation operations like 'mapM' also execute the transforming
108 -- effects differently for different types of streams.
115 -- ** Speculative Streams
119 -- ** Asynchronous Streams
125 -- ** Zipping Streams
130 -- * Parallel Function Application
147 -- * Concurrency Control
160 -- * Stream Type Adapters
184 -- ** Folding Containers of Streams
185 -- | These are variants of standard 'Foldable' fold functions that use a
186 -- polymorphic stream sum operation (e.g. 'async' or 'wSerial') to fold a
187 -- finite container of streams. Note that these are just special cases of
188 -- the more general 'concatMapWith' operation.
217 -- * Deprecated/Moved
218 -- | These APIs have been moved to other modules
226import Data.Semigroup (Semigroup(..))
227import Streamly.Internal.Data.SVar (MonadAsync, Rate(..))
228import Streamly.Internal.Data.Stream.Ahead
229import Streamly.Internal.Data.Stream.Async hiding (mkAsync)
230import Streamly.Internal.Data.Stream.Combinators
231import Streamly.Internal.Data.Stream.Parallel
232import Streamly.Internal.Data.Stream.Serial
233import Streamly.Internal.Data.Stream.StreamK hiding (serial)
234import Streamly.Internal.Data.Stream.Zip
236import qualified Streamly.Prelude as P
237import qualified Streamly.Internal.Prelude as IP
238import qualified Streamly.Internal.Data.Stream.StreamK as K
239import qualified Streamly.Internal.Data.Stream.Async as Async
241-- XXX provide good succinct examples of pipelining, merging, splitting etc.
246-- A program is expressed as a network of streams and folds. A stream is a
247-- source or generator of data elements and a fold is a consumer of data elements that
248-- reduces multiple input elements to a single value.
250-- In the following example, a 'Word8' stream is generated by using
251-- 'Streamly.FileSystem.Handle.read' on a file handle, then the
252-- 'Streamly.Prelude.splitBySuffix' transformation splits the stream on
253-- newlines (ascii value 10); it uses the 'Streamly.Data.Fold.drain' fold to reduce
254-- the resulting lines to unit values (@()@), 'Streamly.Prelude.length' fold
255-- then counts the unit elements in the resulting stream which gives us the
256-- number of lines in the file:
258-- > S.length $ S.splitOnSuffix FL.drain 10 $ FH.read fh
260-- The following example folds the lines to arrays of 'Word8' using the
261-- 'Streamly.Memory.Array.writeF' fold and then wraps the lines in square
262-- brackets before writing them to standard output using
263-- 'Streamly.FileSystem.Handle.write':
265-- > wrapLine ln = S.fromList "[" <> A.read ln <> S.fromList "]\n"
266-- > readLines = S.splitOnSuffix A.writeF 10
267-- > FH.write stdout $ S.concatMap wrapLine $ readLines fh1
269-- One stream can be appended after another:
271-- > FH.write stdout $ S.concatMap wrapLine $ readLines fh1 <> readLines fh2
273-- The following example reads two files concurrently, merges the lines from
274-- the two streams and writes the resulting stream to another file:
276-- > FH.write stdout $ S.concatMap wrapLine $ readLines fh1 `parallel` readLines fh2
278-- There are many ways to generate, merge, zip, transform and fold data
279-- streams. Many transformations can be chained in a stream pipeline. See
280-- "Streamly.Prelude" module for combinators to manipulate streams.
284-- The way stream types in this module like 'SerialT' represent data sources,
285-- the same way the 'Fold' type from "Streamly.Data.Fold" represents data sinks or
286-- reducers of streams. Reducers can be combined to consume a stream source in
287-- many ways. The simplest is to reduce a stream source using a fold e.g.:
289-- > S.runFold FL.length $ S.enumerateTo 100
291-- Folds are consumers of streams and can be used to split a stream into
292-- multiple independent flows. Grouping transforms a stream by applying a fold
293-- on segments of a stream, distributing applies multiple folds in parallel on
294-- the same stream and combines them, partitioning sends different elements of
295-- a stream to different folds, unzipping divides the elements of a stream into
296-- parts and sends them through different folds. Parsers are nothing but a
297-- particular type of folds. Transformations can be applied contravariantly on
298-- the input of a fold.
302-- |---transform----Fold m a b--------|
303-- ---stream m a-->transform-->| |---f b c ...
304-- |---transform----Fold m a c--------|
312-- Streamly arrays (See "Streamly.Memory.Array") complement streams to provide an
313-- efficient computing paradigm. Streams are suitable for immutable
314-- transformations of /potentially infinite/ data using /sequential access/ and
315-- pipelined transformations whereas arrays are suitable for in-place
316-- transformations of /necessarily finite/ data using /random access/. Streams
317-- are synonymous with /sequential pipelined processing/ whereas arrays are
318-- synonymous with /efficient buffering and random access/.
320-- In general, a data processing pipeline reads data from some IO device, does
321-- some processing on it and finally writes the output to another IO device.
322-- Streams provide the overall framework of sequential processing pipeline in
323-- which arrays are used as buffering elements in the middle. In addition to
324-- buffering in the middle, arrays can also be used at the boundaries of the
325-- pipeline to efficiently interface with external storage systems like memory,
326-- files and network. If streams are the pipes in a water pipeline network
327-- then arrays are like the storage tanks in the middle. On the input side,
328-- think of arrays as buckets to fetch water to feed the pipeline and on the
329-- output side buckets to remove the processed water.
331-- 'ByteString' data type from the 'bytestring' package and the 'Text' data
332-- type from the 'text' package are special cases of arrays. 'ByteString' is
333-- like @Array Word8@ and 'Text' is like @utf16@ encoded @Array Word8@.
334-- Streamly arrays can be transformed as efficiently as @bytestring@ or @text@
335-- by using stream operations on them.
337-- Streams and arrays are equally important in computing. They are computing
338-- duals of each other.
341-- The basic stream type is 'Serial', it represents a sequence of IO actions,
342-- and is a 'Monad'. The type 'SerialT' is a monad transformer that can
343-- represent a sequence of actions in an arbitrary monad. The type 'Serial' is
344-- in fact a synonym for @SerialT IO@. There are a few more types similar to
345-- 'SerialT', all of them represent a stream and differ only in the
346-- 'Semigroup', 'Applicative' and 'Monad' compositions of the stream. 'Serial'
347-- and 'WSerial' types compose serially whereas 'Async' and 'WAsync'
348-- types compose concurrently. All these types can be freely inter-converted
349-- using type combinators without any cost. You can freely switch to any type
350-- of composition at any point in the program. When no type annotation or
351-- explicit stream type combinators are used, the default stream type is
352-- inferred as 'Serial'.
354-- This module exports stream types, instances and combinators for:
356-- * converting between different stream types
357-- * appending and concurrently merging streams
358-- * Concurrency control
359-- * Concurrent function application
360-- * Stream rate control
362-- This module is designed to be imported unqualified:
368-- See the "Streamly.Prelude" module for APIs for construction,
369-- generation, elimination and transformation of streams.
371------------------------------------------------------------------------------
372-- Eliminating a stream
373------------------------------------------------------------------------------
375-- | Same as 'runStream'
378{-# DEPRECATED runStreaming "Please use runStream instead." #-}
379runStreaming :: (Monad m, IsStream t) => t m a -> m ()
380runStreaming = P.drain . K.adapt
382-- | Same as @runStream@.
385{-# DEPRECATED runStreamT "Please use runStream instead." #-}
386runStreamT :: Monad m => SerialT m a -> m ()
389-- | Same as "Streamly.Prelude.runStream".
391{-# DEPRECATED runStream "Please use Streamly.Prelude.drain instead." #-}
392runStream :: Monad m => SerialT m a -> m ()
395-- | Same as @runStream . wSerially@.
398{-# DEPRECATED runInterleavedT "Please use 'runStream . interleaving' instead." #-}
399runInterleavedT :: Monad m => WSerialT m a -> m ()
400runInterleavedT = P.drain . K.adapt
402-- | Same as @runStream . parallely@.
405{-# DEPRECATED runParallelT "Please use 'runStream . parallely' instead." #-}
406runParallelT :: Monad m => ParallelT m a -> m ()
407runParallelT = P.drain . K.adapt
409-- | Same as @runStream . asyncly@.
412{-# DEPRECATED runAsyncT "Please use 'runStream . asyncly' instead." #-}
413runAsyncT :: Monad m => AsyncT m a -> m ()
414runAsyncT = P.drain . K.adapt
416-- | Same as @runStream . zipping@.
419{-# DEPRECATED runZipStream "Please use 'runStream . zipSerially instead." #-}
420runZipStream :: Monad m => ZipSerialM m a -> m ()
421runZipStream = P.drain . K.adapt
423-- | Same as @runStream . zippingAsync@.
426{-# DEPRECATED runZipAsync "Please use 'runStream . zipAsyncly instead." #-}
427runZipAsync :: Monad m => ZipAsyncM m a -> m ()
428runZipAsync = P.drain . K.adapt
431-- | Same as "Streamly.Prelude.foldWith".
433{-# DEPRECATED foldWith "Please use Streamly.Prelude.foldWith instead." #-}
434{-# INLINABLE foldWith #-}
435foldWith :: (IsStream t, Foldable f)
436 => (t m a -> t m a -> t m a) -> f (t m a) -> t m a
439-- | Same as "Streamly.Prelude.foldMapWith".
441{-# DEPRECATED foldMapWith "Please use Streamly.Prelude.foldMapWith instead." #-}
442{-# INLINABLE foldMapWith #-}
443foldMapWith :: (IsStream t, Foldable f)
444 => (t m b -> t m b -> t m b) -> (a -> t m b) -> f a -> t m b
445foldMapWith = P.foldMapWith
447-- | Same as "Streamly.Prelude.forEachWith".
449{-# DEPRECATED forEachWith "Please use Streamly.Prelude.forEachWith instead." #-}
450{-# INLINABLE forEachWith #-}
451forEachWith :: (IsStream t, Foldable f)
452 => (t m b -> t m b -> t m b) -> f a -> (a -> t m b) -> t m b
453forEachWith = P.forEachWith
456-- XXX Deprecate it in 0.8.0
458-- | Make a stream asynchronous, triggers the computation and returns a stream
459-- in the underlying monad representing the output generated by the original
460-- computation. The returned action is exhaustible and must be drained once. If
461-- not drained fully we may have a thread blocked forever and once exhausted it
462-- will always return 'empty'.
465{-# INLINABLE mkAsync #-}
466mkAsync :: (IsStream t, MonadAsync m) => t m a -> m (t m a)
467mkAsync = return . Async.mkAsync
469------------------------------------------------------------------------------
471------------------------------------------------------------------------------
475-- When a stream consumer demands an element from a serial stream constructed
476-- as @a \`consM` b \`consM` ... nil@, the action @a@ at the head of the stream
477-- sequence is executed and the result is supplied to the consumer. When the
478-- next element is demanded, the action @b@ is executed and its result is
479-- supplied. Thus, the effects are performed and results are consumed strictly
480-- in a serial order. Serial streams can be considered as /spatially ordered/
481-- streams as the order of execution and consumption is the same as the spatial
482-- order in which the actions are composed by the programmer.
484-- Serial streams enforce the side effects as well as the results of the
485-- actions to be in the same order in which the actions are added to the
486-- stream. Therefore, the semigroup operation for serial streams is not
490-- a <> b is not the same as b <> a
493-- There are two serial stream types 'SerialT' and 'WSerialT'. The stream
494-- evaluation of both the variants works in the same way as described above,
495-- they differ only in the 'Semigroup' and 'Monad' implementaitons.
499-- When a stream consumer demands an element from a speculative stream
500-- constructed as @a \`consM` b \`consM` ... nil@, the action @a@ at the head
501-- of the stream is executed and the output of the action is supplied to the
502-- consumer. However, in addition to the action at the head multiple actions
503-- following it may also be executed concurrently and the results buffered.
504-- When the next element is demanded it may be served from the buffer and we
505-- may execute the next action in the sequence to keep the buffer adequately
506-- filled. Thus, the actions are executed concurrently but results consumed in
507-- serial order just like serial streams. `consM` can be used to fold an
508-- infinite lazy container of effects, as the number of concurrent executions
511-- Similar to 'consM', the monadic stream generation (e.g. replicateM) and
512-- transformation operations (e.g. mapM) on speculative streams can execute
513-- multiple effects concurrently in a speculative manner.
515-- How many effects can be executed concurrently and how many results can be
516-- buffered are controlled by 'maxThreads' and 'maxBuffer' combinators
517-- respectively. The actual number of concurrent threads is adjusted according
518-- to the rate at which the consumer is consuming the stream. It may even
519-- execute actions serially in a single thread if that is enough to match the
522-- Speculative streams enforce ordering of the results of actions in the stream
523-- but the side effects are only partially ordered. Therefore, the semigroup
524-- operation for speculative streams is not commutative from the pure outputs
525-- perspective but commutative from side effects perspective.
529-- /Scheduling and execution:/ In an asynchronous stream @a \`consM` b \`consM`
530-- c ...@, the actions @a@, @b@, and @c@ are executed concurrently with the
531-- consumer of the stream. The actions are /scheduled/ for execution in the
532-- same order as they are specified in the stream. Multiple scheduled actions
533-- may be /executed/ concurrently in parallel threads of execution. The
534-- actions may be executed out of order and they may complete at arbitrary
535-- times. Therefore, the /effects/ of the actions may be observed out of
538-- /Buffering:/ The /results/ from multiple threads of execution are queued in
539-- a buffer as soon as they become available. The consumer of the stream is
540-- served from this buffer. Therefore, the consumer may observe the results to
541-- be out of order. In other words, an asynchronous stream is an unordered
542-- stream i.e. order does not matter.
544-- /Concurrency control:/ Threads are suspended if the `maxBuffer` limit is
545-- reached, and resumed when the consumer makes space in the buffer. The
546-- maximum number of concurrent threads depends on `maxThreads`. Number of
547-- threads is increased or decreased based on the speed of the consumer.
549-- /Generation operations:/ Concurrent stream generation operations e.g.
550-- 'Streamly.Prelude.replicateM' when used in async style schedule and execute
551-- the stream generating actions in the manner described above. The generation
552-- actions run concurrently, effects and results of the actions as observed by
553-- the consumer of the stream may be out of order.
555-- /Transformation operations:/ Concurrent stream transformation operations
556-- e.g. 'Streamly.Prelude.mapM', when used in async style, schedule and
557-- execute transformation actions in the manner described above. Transformation
558-- actions run concurrently, effects and results of the actions may be
559-- observed by the consumer out of order.
561-- /Variants:/ There are two asynchronous stream types 'AsyncT' and 'WAsyncT'.
562-- They are identical with respect to single stream evaluation behavior. Their
563-- behaviors differ in how they combine multiple streams using 'Semigroup' or
564-- 'Monad' composition. Since the order of elements does not matter in
565-- asynchronous streams the 'Semigroup' operation is effectively commutative.
569-- 'ZipSerialM' and 'ZipAsyncM', provide 'Applicative' instances for zipping the
570-- corresponding elements of two streams together. Note that these types are
575-- Stream processing functions can be composed in a chain using function
576-- application with or without the '$' operator, or with reverse function
577-- application operator '&'. Streamly provides concurrent versions of these
578-- operators applying stream processing functions such that each stage of the
579-- stream can run in parallel. The operators start with a @|@; we can read '|$'
580-- as "@parallel dollar@" to remember that @|@ comes before '$'.
582-- Imports for the code snippets below:
586-- import qualified Streamly.Prelude as S
587-- import Control.Concurrent
591-- The 'Semigroup' operation '<>' of each stream type combines two streams in a
592-- type specific manner. This section provides polymorphic versions of '<>'
593-- which can be used to combine two streams in a predetermined way irrespective
596-- XXX An alternative design choice would be to let a control parameter affect
597-- the nearest SVar only and then it gets cleared. The benefit of the current
598-- choice is that it is simply just like global configuration, just like state
599-- behaves, so should be easy to comprehend. But it has the downside of leaking
600-- to undesired actions, that is we can forget to reset it.
604-- These combinators can be used at any point in a stream composition to set
605-- parameters to control the concurrency of the /argument stream/. A control
606-- parameter set at any point remains effective for any concurrent combinators
607-- used in the argument stream until it is reset by using the combinator again.
608-- These control parameters have no effect on non-concurrent combinators in the
609-- stream, or on non-concurrent streams.
611-- /Pitfall:/ Remember that 'maxBuffer' in the following example applies to
612-- 'mapM' and any other combinators that may follow it, and it does not apply
613-- to the combinators before it:
622-- If we use '&' instead of '$' the situation will reverse, in the following
623-- example, 'maxBuffer' does not apply to 'mapM', it applies to combinators
624-- that come before it, because those are the arguments to 'maxBuffer':
635-- You may want to use different stream composition styles at different points
636-- in your program. Stream types can be freely converted or adapted from one
637-- type to another. The 'IsStream' type class facilitates type conversion of
638-- one stream type to another. It is not used directly, instead the type
639-- combinators provided below are used for conversions.
641-- To adapt from one monomorphic type (e.g. 'AsyncT') to another monomorphic
642-- type (e.g. 'SerialT') use the 'adapt' combinator. To give a polymorphic code
643-- a specific interpretation or to adapt a specific type to a polymorphic type
644-- use the type specific combinators e.g. 'asyncly' or 'wSerially'. You
645-- cannot adapt polymorphic code to polymorphic code, as the compiler would not know
646-- which specific type you are converting from or to. If you see a an
647-- @ambiguous type variable@ error then most likely you are using 'adapt'
648-- unnecessarily on polymorphic code.
keysesc to close
- j k
- move down, up
- enter
- open
- esc
- back, or up a level
- 1 2 3 4
- readme, files, commits, refs
- /
- filter
- y
- copy the clone command
- c
- compare the chosen commit with another (or shift-click both)
- ?
- show or hide this list