Haskell Virtual I/O -- a system to increase the flexibility of input and
output in Haskell
Copyright (c) 2004-2005 John Goerzen, firstname.lastname@example.org
HVIO provides the following general features:
- The ability to use a single set of functions on various different
types of objects, including standard Handles, in-memory buffers,
compressed files, network data streams, etc.
- The ability to transparently add filters to the I/O process.
These filters could include things such as character set conversions,
compression or decompression of a data stream, and more.
- The ability to define new objects that have the properties
of I/O objects and can be used interchangably with them.
- Specification compatibility with, and complete support for,
existing I/O on Handles.
- Provide easier unit testing capabilities for I/O actions
HVIO defines several basic type classes that you can use. You will mostly
be interested in HVIO.
It's trivial to adapt old code to work with HVIO. For instance, consider
this example of old and new code:
printMsg :: Handle -> String -> IO ()
printMsg h msg = hPutStr h ("msg: " ++ msg)
And now, the new way:
printMsg :: HVIO h => h -> String -> IO ()
printMsg h msg = vPutStr h ("msg: " ++ msg)
There are several points to note about this conversion:
- The new method can still accept a Handle in exactly the same way as
the old method. Changing your functions to use HVIO will require no
changes from functions that call them with Handles.
- Most "h" functions have equivolent "v" functions that operate
on HVIO classes instead of the more specific Handle. The "v" functions
behave identically to the "h" functions whenever possible.
- There is no equivolent of "openFile" in any HVIO class. You must
create your Handle (or other HVIO object) using normal means.
This is because the creation is so different that it cannot be standardized.
In addition to Handle, there are several pre-defined classes for your use.
StreamReader is a particularly interesting one. At creation time, you pass
it a String. Its contents are read lazily whenever a read call is made. It
can be used, therefore, to implement filters (simply initialize it with the
result from, say, a map over hGetContents from another HVIO object), codecs,
and simple I/O testing. Because it is lazy, it need not hold the entire
string in memory. You can create a StreamReader with a call to
MemoryBuffer is a similar class, but with a different purpose. It provides
a full interface like Handle (it implements HVIOReader, HVIOWriter,
and HVIOSeeker). However, it maintains an in-memory buffer with the
contents of the file, rather than an actual on-disk file. You can access
the entire contents of this buffer at any time. This can be quite useful
for testing I/O code, or for cases where existing APIs use I/O, but you
prefer a String representation. You can create a MemoryBuffer with a call
Finally, there are pipes. These pipes are analogous to the Unix
pipes that are available from System.Posix, but don't require Unix and work
only in Haskell. When you create a pipe, you actually get two HVIO objects:
a PipeReader and a PipeWriter. You must use the PipeWriter in one
thread and the PipeReader in another thread. Data that's written to the
PipeWriter will then be available for reading with the PipeReader. The
pipes are implemented completely with existing Haskell threading primitives,
and require no special operating system support. Unlike Unix pipes, these
pipes cannot be used across a fork(). Also unlike Unix pipes, these pipes
are portable and interact well with Haskell threads. A new pipe can be created
with a call to newHVIOPipe.
Together with System.IO.HVFS, this module is part of a complete
virtual filesystem solution.
This is the generic I/O support class. All objects that are to be used
in the HVIO system must provide an instance of HVIO.
Functions in this class provide an interface with the same specification as
the similar functions in System.IO. Please refer to that documentation
for a more complete specification than is provided here.
Instances of HVIO must provide vClose, vIsEOF, and either
vIsOpen or vIsClosed.
Implementators of readable objects must provide at least vGetChar
An implementation of vGetContents is also highly suggested, since
the default cannot implement proper partial closing semantics.
Implementators of writable objects must provide at least vPutChar and
Implementators of seekable objects must provide at least
vIsSeekable, vTell, and vSeek.
|vClose :: a -> IO ()|
|Close a file
|vIsOpen :: a -> IO Bool|
|Test if a file is open
|vIsClosed :: a -> IO Bool|
|Test if a file is closed
|vTestOpen :: a -> IO ()|
|Raise an error if the file is not open.
This is a new HVIO function and is implemented in terms of
|vIsEOF :: a -> IO Bool|
|Whether or not we're at EOF. This may raise on exception
on some items, most notably write-only Handles such as stdout.
In general, this is most reliable on items opened for reading.
vIsEOF implementations must implicitly call vTestOpen.
|vShow :: a -> IO String|
|Detailed show output.
|vMkIOError :: a -> IOErrorType -> String -> Maybe FilePath -> IOError|
|Make an IOError.
|vThrow :: a -> IOErrorType -> IO b|
|Throw an IOError.
|vGetFP :: a -> IO (Maybe FilePath)|
|Get the filename/object/whatever that this corresponds to.
May be Nothing.
|vTestEOF :: a -> IO ()|
|Throw an isEOFError if we're at EOF; returns nothing otherwise.
If an implementation overrides the default, make sure that it
calls vTestOpen at some point. The default implementation is
a wrapper around a call to vIsEOF.
|vGetChar :: a -> IO Char|
|Read one character
|vGetLine :: a -> IO String|
|Read one line
|vGetContents :: a -> IO String|
Get the remaining contents. Please note that as a user of this
function, the same partial-closing semantics as are used in the
standard hGetContents are encouraged from implementators,
but are not required. That means that, for instance,
a vGetChar after a vGetContents may return some undefined
result instead of the error you would normally get. You should
use caution to make sure your code doesn't fall into that trap,
or make sure to test your code with Handle or one of the
default instances defined in this module. Also, some implementations
may essentially provide a complete close after a call to vGetContents.
The bottom line: after a call to vGetContents, you should do nothing
else with the object save closing it with vClose.
For implementators, you are highly encouraged to provide a correct
|vReady :: a -> IO Bool|
|Indicate whether at least one item is ready for reading.
This will always be True for a great many implementations.
|vIsReadable :: a -> IO Bool|
|Indicate whether a particular item is available for reading.
|vPutChar :: a -> Char -> IO ()|
|Write one character
|vPutStr :: a -> String -> IO ()|
|Write a string
|vPutStrLn :: a -> String -> IO ()|
|Write a string with newline character after it
|vPrint :: Show b => a -> b -> IO ()|
|Write a string representation of the argument, plus a newline.
|vFlush :: a -> IO ()|
|Flush any output buffers.
Note: implementations should assure that a vFlush is automatically
on file close, if necessary to ensure all data sent is written.
|vIsWritable :: a -> IO Bool|
|Indicate whether or not this particular object supports writing.
|vSeek :: a -> SeekMode -> Integer -> IO ()|
|Seek to a specific location.
|vTell :: a -> IO Integer|
|Get the current position.
|vRewind :: a -> IO ()|
|Convenience function to reset the file pointer to the beginning
of the file. A call to vRewind h is the
same as vSeek h AbsoluteSeek 0.
|vIsSeekable :: a -> IO Bool|
|Indicate whether this instance supports seeking.
|vSetBuffering :: a -> BufferMode -> IO ()|
|Set buffering; the default action is a no-op.
|vGetBuffering :: a -> IO BufferMode|
|Get buffering; the default action always returns NoBuffering.
|vPutBuf :: a -> Ptr b -> Int -> IO ()|
|Binary output: write the specified number of octets from the specified
|vGetBuf :: a -> Ptr b -> Int -> IO Int|
|Binary input: read the specified number of octets from the
specified buffer location, continuing to read
until it either consumes that much data or EOF is encountered.
Returns the number of octets actually read. EOF errors are never
raised; fewer bytes than requested are returned on EOF.