%**<title>The Haskell 98 Library Report: IO</title>
%**~header
\section{Input/Output}
\label{IO}
\index{input/output}
\index{I/O}
\indextt{IO}

% Gotta break the figure ...
\outline{
\inputHS{headers/IO}
}
\outline{
\inputHS{headers/IO1}
}

The monadic I/O system used in \Haskell{} is described by the
\Haskell{} language report.  Commonly used I/O functions such as
@print@ are part of the standard prelude and need not be explicitly
imported.  This library contain more advanced I/O features.
Some related operations on file systems are contained in the
@Directory@ library.

\subsection{I/O Errors}
\index{I/O errors}
\label{IOError}
Errors of type @IOError@ are used by the I/O monad.  This is an
abstract type; the library provides functions to interrogate and
construct values in @IOError@:\indextt{IOError}
\begin{itemize}
\item
@isAlreadyExistsError@\indextt{isAlreadyExistsError}
 -- the operation failed because one of its
arguments already exists.
\item
@isDoesNotExistError@\indextt{isDoesNotExistError}
 -- the operation failed because one of its
arguments does not exist.
\item
@isAlreadyInUseError@\indextt{isAlreadyInUseError}
 -- the operation failed because one of its
arguments is a single-use resource, which is already being used (for
example, opening the same file twice for writing might give this
error).
\item
@isFullError@\indextt{isFullError}
 -- the operation failed because the device is full.
\item
@isEOFError@\indextt{isEOFError}
 -- the operation failed because the end of file
has been reached.
\item
@isIllegalOperation@\indextt{isIllegalOperation}
 -- the operation is not possible.
\item
@isPermissionError@\indextt{isPermissionError}
 -- the operation failed because the user does not
have sufficient operating system privilege to perform that operation.
\item
@isUserError@\indextt{isUserError}
 -- a programmer-defined error value has been raised using
@fail@.\indextt{fail}
\end{itemize}
All these functions return a @Bool@, which is @True@ if
its argument is the corresponding kind of error, and @False@
otherwise.  

Any computation which returns an @IO@ result may fail with
@isIllegalOperation@.  Additional errors which could be raised by
an implementation are listed after the corresponding operation.  In
some cases, an implementation will not be able to distinguish between
the possible error causes.  In this case it should return
@isIllegalOperation@.

Three additional functions are provided to obtain information about an
error value.  These are @ioeGetHandle@\indextt{ioeGetHandle} which
returns @Just@~"hdl" if the error value refers to handle "hdl" and
@Nothing@ otherwise; @ioeGetFileName@\indextt{ioeGetFileName} which
returns @Just@~"name" if the error value refers to file "name", and
@Nothing@ otherwise; and @ioeGetErrorString@\indextt{ioeGetErrorString} which returns a
string.  For ``user'' errors (those which are raised using @fail@),
the string returned by @ioeGetErrorString@ is the argument that was passed to
@fail@; for all other errors, the string is implementation-dependent.

The @try@ function returns an error in a computation explicitly using
the @Either@ type.

The @bracket@ function 
captures a common allocate, compute, deallocate idiom in which the
deallocation step must occur even in the case of an error during
computation.  This is similar to try-catch-finally in Java.
% Inline the code here since there's no other functions in IO that
% are not primitive.



\subsection{Files and Handles}
\Haskell{} interfaces to the external world through an abstract {\em file
system}\index{file system}.  This file system is a collection of named {\em file
system objects}, which may be organised
in {\em directories}\index{directories} (see @Directory@).  
In some implementations, directories may themselves be file system
objects and could be entries in other directories.  For simplicity,
any non-directory file system object is termed a {\em file}\index{file},
although it could in fact be a communication channel, or any other
object recognised by the operating system.  {\em Physical
files}\index{physical file} are 
persistent, ordered files, and normally reside on disk.

File and directory names are values of type @String@, whose precise
meaning is operating system dependent.  Files can be opened, yielding
a handle which can then be used to operate on the contents of that
file.

\label{Handles}
\index{handles}

\Haskell{} defines operations to read and write characters from and to files,
represented by values of type @Handle@.  Each value of this type is a {\em
handle}: a record used by the \Haskell{} run-time system to {\em manage} I/O
with file system objects.  A handle has at least the following properties:

\begin{itemize}
\item
 whether it manages input or output or both;
\item
 whether it is {\em open}, {\em closed} or {\em semi-closed};
\item
 whether the object is seekable;
\item
 whether buffering is disabled, or enabled on a line or block basis;
\item
 a buffer (whose length may be zero).
\end{itemize}

Most handles will also have a current I/O position indicating where the next
input or output operation will occur.  A handle is {\em readable} if it
manages only input or both input and output; likewise, it is {\em writable} if
it manages only output or both input and output.  A handle is {\em open} when
first allocated.  Once it is closed it can no longer be used for either input
or output, though an implementation cannot re-use its storage while references
remain to it.  Handles are in the @Show@ and @Eq@ classes.  The string
produced by showing a handle is system dependent; it should include 
enough information to identify the handle for debugging.  A handle is
equal according to @==@ only to itself; no attempt
is made to compare the internal state of different handles for equality.

\subsubsection{Standard Handles}
\label{StandardHandles}
\index{standard handles}

Three handles are allocated during program initialisation.  The first
two (@stdin@\indextt{stdin} and @stdout@\indextt{stdout}) manage input or output from the \Haskell{}
program's standard input or output channel respectively.  The third
(@stderr@\indextt{stderr}) manages output to the standard error channel.  These
handles are initially open.

\subsubsection{Semi-Closed Handles}
\label{SemiClosed}
\index{semi-closed handles}

The operation "@hGetContents@ hdl"\indextt{hGetContents} (Section~\ref{hGetContents})
puts a handle "hdl" into an intermediate
state, {\em semi-closed}.  In this state, "hdl" is effectively closed,
but items are read from "hdl" on demand and accumulated in a special
list returned by @hGetContents@~"hdl".
 
Any operation that fails because a handle is
closed, also fails if a handle is semi-closed. The only exception is @hClose@.
A semi-closed handle becomes closed:
\begin{itemize}
\item
if  @hClose@ is applied to it;
\item
if an I/O error occurs when reading an item from the handle;
\item
or once the entire contents of the handle has been read.
\end{itemize}

Once a semi-closed handle becomes closed, the contents of the
associated list becomes fixed.  The contents of this final list is
only partially specified: it will contain at least all the items of
the stream that were evaluated prior to the handle becoming closed.

Any I/O errors encountered while a handle is semi-closed are simply
discarded.
 
\subsubsection{File locking}

Implementations should enforce as far as possible, at least locally to the
\Haskell{} process, multiple-reader single-writer locking on files.
That is, {\em there may either be many handles on the same file which manage
input, or just one handle on the file which manages output}.  If any
open or semi-closed handle is managing a file for output, no new
handle can be allocated for that file.  If any open or semi-closed
handle is managing a file for input, new handles can only be allocated
if they do not manage output.  Whether two files are the same is
implementation-dependent, but they should normally be the same if they
have the same absolute path name and neither has been renamed, for
example.

{\em Warning}: the @readFile@ operation (Section 7.1 of the Haskell Language Report)
holds a semi-closed handle on the file until the entire contents of the file have been
consumed.  It follows that an attempt to write to a file (using @writeFile@, for example)
that was earlier opened by @readFile@ will usually result in 
failure with @isAlreadyInUseError@.
\indextt{readFile}
\indextt{writeFile}


\subsection{Opening and Closing Files}
\label{OpeningClosing}

\subsubsection{Opening Files}
\label{Opening}
\index{opening a file}
\index{creating a file}

Computation @openFile@~"file"~"mode"\indextt{openFile} allocates and
returns a new, open handle to manage the file "file".
% I don't believe this footnote is technically correct -- functions
% are never computations IIRC: the computation is the action
% that occurs when the function is applied to a state token -- KH
% \footnote{We use
% the term "computation" instead of "function" here to separate
% functions which denote actions in the I/O monad from those outside the monad.}
It manages
input if "mode"\indextycon{IOMode} is @ReadMode@\indextt{ReadMode},
output if "mode" is @WriteMode@\indextt{WriteMode} or
@AppendMode@,\indextt{AppendMode} and both input and output if mode is
@ReadWriteMode@.\indextt{ReadWriteMode}

If the file does not exist and it is opened for output, it should be created
as a new file.  If "mode" is @WriteMode@ and the file already exists, then it
should be truncated to zero length.  Some operating systems delete empty
files, so there is no guarantee that the file will exist following an
@openFile@ with "mode" @WriteMode@ unless it is subsequently written to
successfully.  The handle is positioned at the end of the file if "mode" is
@AppendMode@, and otherwise at the beginning (in which case its internal I/O
position is 0).  The initial buffer mode is implementation-dependent.

If @openFile@ fails on a file opened for output, the file may still
have been created if it did not already exist.

{\em Error reporting}: the @openFile@ computation may fail with
@isAlreadyInUseError@ if the file is already open and cannot be
reopened;
@isDoesNotExistError@ if the file does not exist; or
@isPermissionError@ if the user does not have
permission to open the file.
\indextt{isAlreadyInUseError}
\indextt{isDoesNotExistError}
\indextt{isPermissionError}

\subsubsection{Closing Files}
\label{Closing}
\index{closing a file}

Computation @hClose@~"hdl"\indextt{hClose} makes handle "hdl" closed.  Before the
computation finishes, if "hdl" is writable its buffer is flushed as
for @hFlush@.
Performing @hClose@ on a handle that has already been closed has no effect; 
doing so not an error.  All other operations on a closed handle will fail.
If @hClose@ fails for any reason, any further operations (apart from @hClose@) on the 
handle will still fail as if "hdl" had been successfully closed.

\subsection{Determining the Size of a File}
\label{FileSize}
\index{size of file}

For a handle "hdl" which is attached to a physical file, @hFileSize@\indextt{hFileSize}
"hdl" returns the size of that file in 8-bit bytes ("\geq" 0).

\subsubsection{Detecting the End of Input}
\label{EOF}
\index{end of file}

For a readable handle "hdl", computation @hIsEOF@~"hdl"\indextt{hIsEOF} returns @True@
if no further input can be taken from "hdl"; for a handle attached to a 
physical file this means that the current I/O position is equal to the length of the file.
Otherwise, it returns @False@.  The computation @isEOF@\indextt{isEOF} is identical,
except that it works only on @stdin@.

% The computation may fail with:
% \begin{itemize}
% \item
% @HardwareFault@
% A physical I/O error has occurred.
% [@EIO@]
% \item
% @ResourceExhausted@
% Insufficient resources are available to perform the operation.
% [@ENOMEM@]
% \item
% @IllegalOperation@
% The handle is not open for reading.
% \end{itemize}

\subsubsection{Buffering Operations}
\label{Buffering}
\index{file buffering}

Three kinds of buffering are supported: line-buffering, 
block-buffering or no-buffering.  These modes have the following effects.
For output, items are written out, or {\em flushed}, from the internal buffer 
according to the buffer mode:
\begin{itemize}
\item
{\bf line-buffering:}
the entire buffer is flushed whenever a newline is output, the
buffer overflows, a @hFlush@ is issued, or the handle is closed.
\item
{\bf block-buffering:}
the entire buffer is written out whenever it overflows, a @hFlush@ is
issued, or the handle is closed.
\item
{\bf no-buffering:}
output is written immediately, and never stored in the buffer.
\end{itemize}
An implementation is free to flush the buffer more frequently, but not 
less frequently, than
specified above.  The buffer is emptied as soon as it has been written out.

Similarly, input occurs according to the buffer mode for handle "hdl".
\begin{itemize}
\item
{\bf line-buffering:}
when the buffer for "hdl" is not empty, the next item is obtained from
the buffer; otherwise, when the buffer is empty, characters are read into
the buffer until the next newline character is encountered or the buffer is full.  No
characters are available until the newline character is available or the buffer is full.
\item
{\bf block-buffering:} 
when the buffer for "hdl" becomes empty, the
next block of data is read into the buffer.
\item
{\bf no-buffering:} 
the next input item is read and returned.  The @hLookAhead@\indextt{hLookAhead} 
operation (Section~\ref{hLookAhead}) implies that
even a no-buffered handle may require a one-character buffer.
\end{itemize}

For most implementations, physical files will normally be block-buffered 
and terminals will normally be line-buffered.

Computation @hSetBuffering@~"hdl"~"mode"\indextt{hSetBuffering} sets the
mode of buffering for handle "hdl" on subsequent reads and writes.
\begin{itemize}
\item
If "mode" is @LineBuffering@, line-buffering is
enabled if possible.
\item
If "mode" is @BlockBuffering@~"size", then block-buffering
is enabled if possible.  The size of the buffer is "n" items
if "size" is @Just @"n" and is otherwise implementation-dependent.
\item
If "mode" is @NoBuffering@, then buffering is disabled if possible.
\end{itemize}

If the buffer mode is changed from @BlockBuffering@ or
@LineBuffering@ to @NoBuffering@, then 
\begin{itemize}
\item
if "hdl" is writable, the buffer is flushed as for 
@hFlush@;
\item
if "hdl" is not writable, the contents of the buffer is discarded.
\end{itemize}

{\em Error reporting}: the @hSetBuffering@ computation may fail with
@isPermissionError@ if
the handle has already been used for reading or writing
and the implementation does not allow the buffering mode to
be changed.

Computation @hGetBuffering@~"hdl"\indextt{hGetBuffering} returns the current buffering mode
for "hdl".

The default buffering mode when a handle is opened is
implementation-dependent and may depend on the file system object which is
attached to that handle.

\subsubsection{Flushing Buffers}
\label{Flushing}
\index{flushing a file buffer}

Computation @hFlush@~"hdl"\indextt{hFlush} causes any items buffered for output in
handle "hdl" to be sent immediately to the operating system.

{\em Error reporting}: the @hFlush@ computation may fail with:
@isFullError@ if the device is full; @isPermissionError@ if a
system resource limit would be exceeded.  It is unspecified whether the
characters in the buffer are discarded or retained under these circumstances.

\subsection{Repositioning Handles}
\label{Seeking}
\index{random access files}
\index{seeking a file}

\subsubsection{Revisiting an I/O Position}

Computation @hGetPosn@~"hdl"\indextt{hGetPosn} returns the current I/O position of "hdl" as a
value of the abstract type @HandlePosn@.  If a call to "@hGetPosn@~h" returns a position "p",
then computation @hSetPosn@~"p"\indextt{hSetPosn} sets the
position of "h" to the position it held at the time of the call to @hGetPosn@.


{\em Error reporting}: the @hSetPosn@ computation may fail with:
@isPermissionError@ if a system resource limit would be exceeded.

\subsubsection{Seeking to a new Position}

Computation @hSeek@~"hdl"~"mode"~"i"\indextt{hSeek} sets the position of handle
"hdl" depending on "mode".\indextycon{SeekMode}  If "mode" is:
\begin{itemize}
\item
@AbsoluteSeek@:\indextt{AbsoluteSeek} the position of "hdl" is set to "i".
\item
@RelativeSeek@:\indextt{RelativeSeek} the position of "hdl" is set to offset "i" from
the current position.
\item
@SeekFromEnd@:\indextt{SeekFromEnd} the position of "hdl" is set to offset "i" from
the end of the file.
\end{itemize}
The offset is given in terms of 8-bit bytes.

If "hdl" is block- or line-buffered, then seeking to a position which is not
in the current buffer will first cause any items in the output buffer to be
written to the device, and then cause the input buffer to be discarded.  Some
handles may not be seekable (see @hIsSeekable@), or only support a subset of
the possible positioning operations (for instance, it may only be possible to
seek to the end of a tape, or to a positive offset from the beginning or
current position).  It is not possible to set a negative I/O position, or for
a physical file, an I/O position beyond the current end-of-file.

{\em Error reporting}:
the @hSeek@ computation may fail with:
@isPermissionError@ if a system resource limit would be exceeded.

\subsection{Handle Properties}
\label{Query}

The functions 
@hIsOpen@\indextt{hIsOpen}, 
@hIsClosed@\indextt{hIsClosed},
@hIsReadable@\indextt{hIsReadable},
@hIsWritable@\indextt{hIsWritable} and
@hIsSeekable@\indextt{hIsSeekable}
return information about the properties of a handle.
Each of these returns @True@ if the handle has the specified property, and
@False@ otherwise.


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% Haskell 1.3 Text Input: LibReadTextIO
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection{Text Input and Output}
\index{reading from a file}
%\indextt{LibReadTextIO}

Here we define a standard set of input operations for reading
characters and strings from text files, using handles.  Many of these
functions are generalizations of Prelude functions.  I/O in the
Prelude generally uses @stdin@ and @stdout@; here, handles are explicitly
specified by the I/O operation.

\subsubsection{Checking for Input}
\label{hReady}
\label{hWaitForInput}
\index{polling a handle for input}

Computation @hWaitForInput@~"hdl"~"t"\indextt{hWaitForInput} 
waits until input is available on handle "hdl".
It returns @True@ as soon as input is available on "hdl", or @False@ if no input is available
within "t" milliseconds.

Computation @hReady@~"hdl"\indextt{hReady} indicates whether at least one item is
available for input from handle "hdl".

{\em Error reporting}.
The @hWaitForInput@ and @hReady@ computations fail with
@isEOFError@ if the end of file has been reached.

\subsubsection{Reading Input}

Computation @hGetChar@~"hdl"\indextt{hGetChar} reads a character from
the file or channel managed by "hdl".

Computation @hGetLine@~"hdl"\indextt{hGetLine} reads a line from the file or
channel managed by "hdl". The Prelude's @getLine@ is a shorthand
for @hGetLine stdin@.
 
{\em Error reporting}.
The @hGetChar@ computation fails with
@isEOFError@ if the end of file has been reached.
The @hGetLine@ fails with @isEOFError@ if the end of file is encountered
when reading the {\em first} character of the line. If @hGetLine@ encounters
end-of-file at any other point while reading in a line, it is treated as
a line terminator and the (partial) line is returned.


\subsubsection{Reading Ahead}
\label{hLookAhead}
\index{lookahead}

Computation @hLookAhead@~"hdl"\indextt{hLookAhead} returns the next character from handle
"hdl" without removing it from the input buffer, blocking until a
character is available.

{\em Error reporting}:
the @hLookAhead@ computation may fail with:
@isEOFError@ if the end of file has been reached.

\subsubsection{Reading The Entire Input}
\label{hGetContents}
\index{get the contents of a file}

Computation @hGetContents@~"hdl"\indextt{hGetContents} returns the list of
characters corresponding to the unread portion of the channel or file managed
by "hdl", which is made semi-closed.

{\em Error reporting}:
the @hGetContents@ computation may fail with:
@isEOFError@ if the end of file has been reached.

\subsubsection{Text Output}

Computation @hPutChar@~"hdl"~"c"\indextt{hPutChar} writes the character "c" to the file
or channel managed by "hdl".  Characters may be buffered if buffering
is enabled for "hdl".

Computation @hPutStr@~"hdl"~"s"\indextt{hPutStr} writes the string
"s" to the file or channel managed by "hdl".

Computation @hPrint@~"hdl"~"t"\indextt{hPrint} writes the string representation of "t"
given by the @shows@ function to the file or channel managed by "hdl" and appends a newline.

{\em Error reporting}:
the @hPutChar@, @hPutStr@ and @hPrint@ computations may fail with:
@isFull@-@Error@ if the device is full;
or @isPermissionError@ if another system resource limit would be exceeded.



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%% Haskell 1.3 Examples
%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

\subsection{Examples}
\index{input/output examples}

Here are some simple examples to illustrate \Haskell{} I/O.

\subsubsection{Summing Two Numbers}

This program reads and sums two @Integer@s.
\bprog
@
import IO

main = do
         hSetBuffering stdout NoBuffering            
         putStr   "Enter an integer: "        
         x1 <- readNum 
         putStr   "Enter another integer: "          
         x2 <- readNum                          
         putStr  ("Their sum is " ++ show (x1+x2) ++ "\n")
       where readNum :: IO Integer
		-- Providing a type signature avoids reliance on
		-- the defaulting rule to fix the type of x1,x2
             readNum = readLn
@
\eprog

\subsubsection{Copying Files}

A simple program to create a copy of a file, with all lower-case
characters translated to upper-case.  This program will not allow a
file to be copied to itself.  This version uses character-level I/O.
Note that exactly two arguments must be supplied to the program.
\bprog
@
import IO
import System
import Char( toUpper )

main = do 
         [f1,f2] <- getArgs
         h1 <- openFile f1 ReadMode     
         h2 <- openFile f2 WriteMode 
         copyFile h1 h2            
         hClose h1                  
         hClose h2

copyFile h1 h2 = do
                   eof <- hIsEOF h1
                   if eof then return () else
                      do
                        c <- hGetChar h1
                        hPutChar h2 (toUpper c)   
                        copyFile h1 h2
@
\eprog

An equivalent but much shorter version, using string I/O is:
\bprog
@
import System
import Char( toUpper )

main = do
         [f1,f2] <- getArgs
         s <- readFile f1
         writeFile f2 (map toUpper s)
@
\eprog

%	Not any more in Haskell 98!
% The @~@ used in the patterns above is a necessary consequence of the
% @do@-notation which has been used for the I/O operations.  In general,
% if the pattern on the left of any @<-@ fails to match, the value of
% the entire @do@-expression is defined to be the ``zero'' in the
% underlying monad.  However, since the @IO@ monad has no zero, the @~@
% is required in order to force the pattern to be irrefutable.  Without
% the @~@, a class error would occur because there is no instance of
% @IO@ for class @MonadZero@.

\subsection{Library @IO@}

\inputHS{code/IO}

%**~footer