System.IO.Streams.Internal

Internal implementation of the io-streams library, intended for library writers

Library users should use the interface provided by System.IO.Streams

Types

data SP a b

A strict pair type.

Constructors

SP !a !b

Instances

BufferedIO (StreamPair ByteString)

RawIO (StreamPair ByteString)

IODevice (StreamPair ByteString)

Typeable (* -> * -> *) SP

type StreamPair a = SP (InputStream a) (OutputStream a)

Internal convenience synonym for a pair of input/output streams.

About pushback

Users can push a value back into an input stream using the unRead function. Usually this will use the default pushback mechanism which provides a buffer for the stream. Some stream transformers, like takeBytes, produce streams that send pushed-back values back to the streams that they wrap. A function like map cannot do this because the types don't match up:

map :: (a -> b) -> InputStream a -> IO (InputStream b)

A function will usually document if its pushback behaviour differs from the default. No matter what the case, input streams should obey the following law:

Streams.unRead c stream >> Streams.read stream === return (Just c)

Input and output streams

data InputStream a

An InputStream generates values of type c in the IO monad.

Two primitive operations are defined on InputStream:

• read :: InputStream c -> IO (Maybe c) reads a value from the stream, where "end of stream" is signaled by read returning Nothing.
• unRead :: c -> InputStream c -> IO () "pushes back" a value to the stream.

It is intended that InputStreams obey the following law:

unRead c stream >> read stream === return (Just c)

Constructors

InputStream

Fields

_read :: IO (Maybe a)

_unRead :: a -> IO ()

Instances

BufferedIO (StreamPair ByteString)

BufferedIO (InputStream ByteString)

RawIO (StreamPair ByteString)

RawIO (InputStream ByteString)

IODevice (StreamPair ByteString)

IODevice (InputStream ByteString)

Typeable (* -> *) InputStream

data OutputStream a

An OutputStream consumes values of type c in the IO monad. The only primitive operation defined on OutputStream is:

• write :: Maybe c -> OutputStream c -> IO ()

Values of type c are written in an OutputStream by wrapping them in Just, and the end of the stream is indicated by supplying Nothing.

If you supply a value after a Nothing, the behavior is defined by the implementer of the given OutputStream. (All OutputStream definitions in this library will simply discard the extra input.)

Constructors

OutputStream

Fields

_write :: Maybe a -> IO ()

Instances

BufferedIO (StreamPair ByteString)

BufferedIO (OutputStream ByteString)

RawIO (StreamPair ByteString)

RawIO (OutputStream ByteString)

IODevice (StreamPair ByteString)

IODevice (OutputStream ByteString)

Typeable (* -> *) OutputStream

Primitive stream operations

read :: InputStream a -> IO (Maybe a)

Reads one value from an InputStream.

Returns either a value wrapped in a Just, or Nothing if the end of the stream is reached.

unRead :: a -> InputStream a -> IO ()

Pushes a value back onto an input stream. read and unRead should satisfy the following law, with the possible exception of side effects:

Streams.unRead c stream >> Streams.read stream === return (Just c)

Note that this could be used to add values back to the stream that were not originally drawn from the stream.

peek :: InputStream a -> IO (Maybe a)

Observes the first value from an InputStream without consuming it.

Returns Nothing if the InputStream is empty. peek satisfies the following law:

Streams.peek stream >> Streams.read stream === Streams.read stream

write :: Maybe a -> OutputStream a -> IO ()

Feeds a value to an OutputStream. Values of type c are written in an OutputStream by wrapping them in Just, and the end of the stream is indicated by supplying Nothing.

atEOF :: InputStream a -> IO Bool

Checks if an InputStream is at end-of-stream.

Building streams

makeInputStream :: IO (Maybe a) -> IO (InputStream a)

Creates an InputStream from a value-producing action.

(makeInputStream m) calls the action m each time you request a value from the InputStream. The given action is extended with the default pushback mechanism (see System.IO.Streams.Internal).

makeOutputStream :: (Maybe a -> IO ()) -> IO (OutputStream a)

Creates an OutputStream from a value-consuming action.

(makeOutputStream f) runs the computation f on each value fed to it.

Since version 1.2.0.0, makeOutputStream also ensures that output streams no longer receive data once EOF is received (i.e. you can now assume that makeOutputStream will feed your function Nothing at most once.)

appendInputStream :: InputStream a -> InputStream a -> IO (InputStream a)

appendInputStream concatenates two InputStreams, analogous to (++) for lists.

The second InputStream continues where the first InputStream ends.

Note: values pushed back to appendInputStream are not propagated to either wrapped InputStream.

concatInputStreams :: [InputStream a] -> IO (InputStream a)

concatInputStreams concatenates a list of InputStreams, analogous to (++) for lists.

Subsequent InputStreams continue where the previous one InputStream ends.

Note: values pushed back to the InputStream returned by concatInputStreams are not propagated to any of the source InputStreams.

Connecting streams

connect :: InputStream a -> OutputStream a -> IO ()

Connects an InputStream and OutputStream, supplying values from the InputStream to the OutputStream, and propagating the end-of-stream message from the InputStream through to the OutputStream.

The connection ends when the InputStream yields a Nothing.

connectTo :: OutputStream a -> InputStream a -> IO ()

The connectTo function is just flip connect.

Useful for writing expressions like fromList [1,2,3] >>= connectTo foo.

supply :: InputStream a -> OutputStream a -> IO ()

Connects an InputStream to an OutputStream without passing the end-of-stream notification through to the OutputStream.

Use this to supply an OutputStream with multiple InputStreams and use connect for the final InputStream to finalize the OutputStream, like so:

do Streams.supply  input1 output
   Streams.supply  input2 output
   Streams.connect input3 output

supplyTo :: OutputStream a -> InputStream a -> IO ()

supply with the arguments flipped.

Thread safety

lockingInputStream :: InputStream a -> IO (InputStream a)

Converts an InputStream into a thread-safe InputStream, at a slight performance penalty.

For performance reasons, this library provides non-thread-safe streams by default. Use the locking functions to convert these streams into slightly slower, but thread-safe, equivalents.

lockingOutputStream :: OutputStream a -> IO (OutputStream a)

Converts an OutputStream into a thread-safe OutputStream, at a slight performance penalty.

For performance reasons, this library provides non-thread-safe streams by default. Use the locking functions to convert these streams into slightly slower, but thread-safe, equivalents.

Utility streams

nullInput :: IO (InputStream a)

An empty InputStream that yields Nothing immediately.

nullOutput :: IO (OutputStream a)

An empty OutputStream that discards any input fed to it.

Generator monad

data Generator r a

A Generator is a coroutine monad that can be used to define complex InputStreams. You can cause a value of type Just r to appear when the InputStream is read by calling yield:

g :: Generator Int ()
g = do
    Streams.yield 1
    Streams.yield 2
    Streams.yield 3

A Generator can be turned into an InputStream by calling fromGenerator:

m :: IO [Int]
m = Streams.fromGenerator g >>= Streams.toList     -- value returned is [1,2,3]

You can perform IO by calling liftIO, and turn a Generator into an InputStream with fromGenerator.

As a general rule, you should not acquire resources that need to be freed from a Generator, because there is no guarantee the coroutine continuation will ever be called, nor can you catch an exception from within a Generator.

Instances

Monad (Generator r)

Functor (Generator r)

Applicative (Generator r)

MonadIO (Generator r)

Typeable (* -> * -> *) Generator

fromGenerator :: Generator r a -> IO (InputStream r)

Turns a Generator into an InputStream.

yield :: r -> Generator r ()

Calling yield x causes the value Just x to appear on the input when this generator is converted to an InputStream. The rest of the computation after the call to yield is resumed later when the InputStream is read again.

Consumer monad

data Consumer c a

Instances

Monad (Consumer c)

Functor (Consumer r)

Applicative (Consumer r)

MonadIO (Consumer c)

Typeable (* -> * -> *) Consumer

fromConsumer :: Consumer r a -> IO (OutputStream r)

await :: Consumer r (Maybe r)