concurrent-output-1.1.0: Ungarble output from several threads

Copyright2013 Joey Hess <id@joeyh.name>
LicenseBSD-2-clause
Safe HaskellNone
LanguageHaskell98

System.Console.Concurrent

Contents

Description

Concurrent output handling.

 import Control.Concurrent.Async
 import System.Console.Concurrent

 main = withConcurrentOutput $
 	outputConcurrent "washed the car\n"
 		`concurrently`
	outputConcurrent "walked the dog\n"
		`concurrently`
 	createProcessConcurrent (proc "ls" [])

Synopsis

Concurrent output

withConcurrentOutput :: (MonadIO m, MonadMask m) => m a -> m a Source

Use this around any actions that use outputConcurrent or createProcessConcurrent

This is necessary to ensure that buffered concurrent output actually gets displayed before the program exits.

class Outputable v where Source

Values that can be output.

Methods

toOutput :: v -> ByteString Source

outputConcurrent :: Outputable v => v -> IO () Source

Displays a value to stdout.

No newline is appended to the value, so if you want a newline, be sure to include it yourself.

Uses locking to ensure that the whole output occurs atomically even when other threads are concurrently generating output.

When something else is writing to the console at the same time, this does not block. It buffers the value, so it will be displayed once the other writer is done.

createProcessConcurrent :: CreateProcess -> IO (Maybe Handle, Maybe Handle, Maybe Handle, ProcessHandle) Source

Wrapper around createProcess that prevents multiple processes that are running concurrently from writing to stdout/stderr at the same time.

If the process does not output to stdout or stderr, it's run by createProcess entirely as usual. Only processes that can generate output are handled specially:

A process is allowed to write to stdout and stderr in the usual way, assuming it can successfully take the output lock.

When the output lock is held (ie, by another concurrent process, or because outputConcurrent is being called at the same time), the process is instead run with its stdout and stderr redirected to a buffer. The buffered output will be displayed as soon as the output lock becomes free, or after the command has finished.

waitForProcessConcurrent :: ProcessHandle -> IO ExitCode Source

This must be used to wait for processes started with createProcessConcurrent.

This is necessary because waitForProcess has a race condition when two threads check the same process. If the race is triggered, one thread will successfully wait, but the other throws a DoesNotExist exception.

flushConcurrentOutput :: IO () Source

Blocks until any processes started by createProcessConcurrent have finished, and any buffered output is displayed.

withConcurrentOutput calls this at the end; you can call it anytime you want to flush output.

lockOutput :: (MonadIO m, MonadMask m) => m a -> m a Source

Holds a lock while performing an action. This allows the action to perform its own output to the console, without using functions from this module.

While this is running, other threads that try to lockOutput will block. Any calls to outputConcurrent and createProcessConcurrent will not block, but the output will be buffered and displayed only once the action is done.

Low level access to the output buffer

data OutputBuffer Source

Buffered output.

data StdHandle Source

Constructors

StdOut 
StdErr 

bufferOutputSTM :: Outputable v => StdHandle -> v -> STM () Source

Adds a value to the output buffer for later display.

Note that buffering large quantities of data this way will keep it resident in memory until it can be displayed. While outputConcurrent uses temp files if the buffer gets too big, this STM function cannot do so.

outputBufferWaiterSTM :: (OutputBuffer -> (OutputBuffer, OutputBuffer)) -> STM [(StdHandle, OutputBuffer)] Source

A STM action that waits for some buffered output to become available, and returns it.

The function can select a subset of output when only some is desired; the fst part is returned and the snd is left in the buffer.

This will prevent it from being displayed in the usual way, so you'll need to use emitOutputBuffer to display it yourself.

waitCompleteLines :: OutputBuffer -> (OutputBuffer, OutputBuffer) Source

Use with outputBufferWaiterSTM to make it only return buffered output that ends with a newline. Anything buffered without a newline is left in the buffer.

emitOutputBuffer :: StdHandle -> OutputBuffer -> IO () Source

Emits the content of the OutputBuffer to the Handle

If you use this, you should use lockOutput to ensure you're the only thread writing to the console.