Copyright	(c) The University of Glasgow 2004-2008
License	BSD-style (see the file libraries/base/LICENSE)
Maintainer	[email protected]
Stability	experimental
Portability	non-portable (requires concurrency)
Safe Haskell	Safe
Language	Haskell2010

Running sub-processes

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

This is the most general way to spawn an external process. The process can be a command line to be executed by a shell or a raw command with a list of arguments. The stdin, stdout, and stderr streams of the new process may individually be attached to new pipes, to existing Handles, or just inherited from the parent (the default.)

The details of how to create the process are passed in the CreateProcess record. To make it easier to construct a CreateProcess, the functions proc and shell are supplied that fill in the fields with default values which can be overriden as needed.

createProcess returns (mb_stdin_hdl, mb_stdout_hdl, mb_stderr_hdl, ph), where

• if std_in == CreatePipe, then mb_stdin_hdl will be Just h, where h is the write end of the pipe connected to the child process's stdin.
• otherwise, mb_stdin_hdl == Nothing

Similarly for mb_stdout_hdl and mb_stderr_hdl.

For example, to execute a simple ls command:

  r <- createProcess (proc "ls" [])

To create a pipe from which to read the output of ls:

  (_, Just hout, _, _) <-
      createProcess (proc "ls" []){ std_out = CreatePipe }

To also set the directory in which to run ls:

  (_, Just hout, _, _) <-
      createProcess (proc "ls" []){ cwd = Just "\home\bob",
                                    std_out = CreatePipe }

Note that Handles provided for std_in, std_out, or std_err via the UseHandle constructor will be closed by calling this function. This is not always the desired behavior. In cases where you would like to leave the Handle open after spawning the child process, please use createProcess_ instead.

createProcess_ Source

This function is almost identical to createProcess. The only differences are:

• Handles provided via UseHandle are not closed automatically.
• This function takes an extra String argument to be used in creating error messages.

This function has been available from the System.Process.Internals module for some time, and is part of the System.Process module since version 1.2.1.0.

Since: 1.2.1.0

shell :: String -> CreateProcess Source

Construct a CreateProcess record for passing to createProcess, representing a command to be passed to the shell.

proc :: FilePath -> [String] -> CreateProcess Source

Construct a CreateProcess record for passing to createProcess, representing a raw command with arguments.

See RawCommand for precise semantics of the specified FilePath.

data CreateProcess Source

data CmdSpec Source

data StdStream Source

data ProcessHandle Source

Simpler functions for common tasks

callProcess :: FilePath -> [String] -> IO () Source

Creates a new process to run the specified command with the given arguments, and wait for it to finish. If the command returns a non-zero exit code, an exception is raised.

If an asynchronous exception is thrown to the thread executing callProcess. The forked process will be terminated and callProcess will wait (block) until the process has been terminated.

Since: 1.2.0.0

callCommand :: String -> IO () Source

Creates a new process to run the specified shell command. If the command returns a non-zero exit code, an exception is raised.

If an asynchronous exception is thrown to the thread executing callCommand. The forked process will be terminated and callCommand will wait (block) until the process has been terminated.

Since: 1.2.0.0

spawnProcess :: FilePath -> [String] -> IO ProcessHandle Source

Creates a new process to run the specified raw command with the given arguments. It does not wait for the program to finish, but returns the ProcessHandle.

Since: 1.2.0.0

spawnCommand :: String -> IO ProcessHandle Source

Creates a new process to run the specified shell command. It does not wait for the program to finish, but returns the ProcessHandle.

Since: 1.2.0.0

readCreateProcess Source

readCreateProcess works exactly like readProcess except that it lets you pass CreateProcess giving better flexibility.

 > readCreateProcess (shell "pwd" { cwd = "/etc/" }) ""
 "/etc\n"

Note that Handles provided for std_in or std_out via the CreateProcess record will be ignored. Since: 1.2.3.0

readProcess Source

readProcess forks an external process, reads its standard output strictly, blocking until the process terminates, and returns the output string. The external process inherits the standard error.

If an asynchronous exception is thrown to the thread executing readProcess, the forked process will be terminated and readProcess will wait (block) until the process has been terminated.

Output is returned strictly, so this is not suitable for interactive applications.

This function throws an IOError if the process ExitCode is anything other than ExitSuccess. If instead you want to get the ExitCode then use readProcessWithExitCode.

Users of this function should compile with -threaded if they want other Haskell threads to keep running while waiting on the result of readProcess.

 > readProcess "date" [] []
 "Thu Feb  7 10:03:39 PST 2008\n"

The arguments are:

• The command to run, which must be in the $PATH, or an absolute or relative path
• A list of separate command line arguments to the program
• A string to pass on standard input to the forked process.

readCreateProcessWithExitCode Source

readCreateProcessWithExitCode works exactly like readProcessWithExitCode except that it lets you pass CreateProcess giving better flexibility.

Note that Handles provided for std_in, std_out, or std_err via the CreateProcess record will be ignored.

Since: 1.2.3.0

readProcessWithExitCode Source

readProcessWithExitCode is like readProcess but with two differences:

• it returns the ExitCode of the process, and does not throw any exception if the code is not ExitSuccess.
• it reads and returns the output from process' standard error handle, rather than the process inheriting the standard error handle.

On Unix systems, see waitForProcess for the meaning of exit codes when the process died as the result of a signal.

Related utilities

showCommandForUser :: FilePath -> [String] -> String Source

Given a program p and arguments args, showCommandForUser p args returns a string suitable for pasting into /bin/sh (on Unix systems) or CMD.EXE (on Windows).

Control-C handling on Unix

When running an interactive console process (such as a shell, console-based text editor or ghci), we typically want that process to be allowed to handle Ctl-C keyboard interrupts how it sees fit. For example, while most programs simply quit on a Ctl-C, some handle it specially. To allow this to happen, use the delegate_ctlc = True option in the CreateProcess options.

The gory details:

By default Ctl-C will generate a SIGINT signal, causing a UserInterrupt exception to be sent to the main Haskell thread of your program, which if not specially handled will terminate the program. Normally, this is exactly what is wanted: an orderly shutdown of the program in response to Ctl-C.

Of course when running another interactive program in the console then we want to let that program handle Ctl-C. Under Unix however, Ctl-C sends SIGINT to every process using the console. The standard solution is that while running an interactive program, ignore SIGINT in the parent, and let it be handled in the child process. If that process then terminates due to the SIGINT signal, then at that point treat it as if we had recieved the SIGINT ourselves and begin an orderly shutdown.

This behaviour is implemented by createProcess (and waitForProcess / getProcessExitCode) when the delegate_ctlc = True option is set. In particular, the SIGINT signal will be ignored until waitForProcess returns (or getProcessExitCode returns a non-Nothing result), so it becomes especially important to use waitForProcess for every processes created.

In addition, in delegate_ctlc mode, waitForProcess and getProcessExitCode will throw a UserInterrupt exception if the process terminated with ExitFailure (-SIGINT). Typically you will not want to catch this exception, but let it propagate, giving a normal orderly shutdown. One detail to be aware of is that the UserInterrupt exception is thrown synchronously in the thread that calls waitForProcess, whereas normally SIGINT causes the exception to be thrown asynchronously to the main thread.

For even more detail on this topic, see "Proper handling of SIGINT/SIGQUIT".

Process completion

waitForProcess :: ProcessHandle -> IO ExitCode Source

Waits for the specified process to terminate, and returns its exit code.

GHC Note: in order to call waitForProcess without blocking all the other threads in the system, you must compile the program with -threaded.

(Since: 1.2.0.0) On Unix systems, a negative value ExitFailure -signum indicates that the child was terminated by signal signum. The signal numbers are platform-specific, so to test for a specific signal use the constants provided by System.Posix.Signals in the unix package. Note: core dumps are not reported, use System.Posix.Process if you need this detail.

getProcessExitCode :: ProcessHandle -> IO (Maybe ExitCode) Source

This is a non-blocking version of waitForProcess. If the process is still running, Nothing is returned. If the process has exited, then Just e is returned where e is the exit code of the process.

On Unix systems, see waitForProcess for the meaning of exit codes when the process died as the result of a signal.

terminateProcess :: ProcessHandle -> IO () Source

Attempts to terminate the specified process. This function should not be used under normal circumstances - no guarantees are given regarding how cleanly the process is terminated. To check whether the process has indeed terminated, use getProcessExitCode.

On Unix systems, terminateProcess sends the process the SIGTERM signal. On Windows systems, the Win32 TerminateProcess function is called, passing an exit code of 1.

Note: on Windows, if the process was a shell command created by createProcess with shell, or created by runCommand or runInteractiveCommand, then terminateProcess will only terminate the shell, not the command itself. On Unix systems, both processes are in a process group and will be terminated together.

interruptProcessGroupOf Source

Sends an interrupt signal to the process group of the given process.

On Unix systems, it sends the group the SIGINT signal.

On Windows systems, it generates a CTRL_BREAK_EVENT and will only work for processes created using createProcess and setting the create_group flag

createPipe :: IO (Handle, Handle) Source

Create a pipe for interprocess communication and return a (readEnd, writeEnd) Handle pair.

Since: 1.2.1.0

Old deprecated functions

These functions pre-date createProcess which is much more flexible.

runProcess Source

Runs a raw command, optionally specifying Handles from which to take the stdin, stdout and stderr channels for the new process (otherwise these handles are inherited from the current process).

Any Handles passed to runProcess are placed immediately in the closed state.

Note: consider using the more general createProcess instead of runProcess.

runCommand :: String -> IO ProcessHandle Source

Runs a command using the shell.

runInteractiveProcess Source

Runs a raw command, and returns Handles that may be used to communicate with the process via its stdin, stdout and stderr respectively.

For example, to start a process and feed a string to its stdin:

  (inp,out,err,pid) <- runInteractiveProcess "..."
  forkIO (hPutStr inp str)

The Handles are initially in binary mode; if you need them to be in text mode then use hSetBinaryMode.

runInteractiveCommand :: String -> IO (Handle, Handle, Handle, ProcessHandle) Source

Runs a command using the shell, and returns Handles that may be used to communicate with the process via its stdin, stdout, and stderr respectively. The Handles are initially in binary mode; if you need them to be in text mode then use hSetBinaryMode.

system :: String -> IO ExitCode Source

Computation system cmd returns the exit code produced when the operating system runs the shell command cmd.

This computation may fail with one of the following IOErrorType exceptions:

PermissionDenied

The process has insufficient privileges to perform the operation.

ResourceExhausted

Insufficient resources are available to perform the operation.

UnsupportedOperation

The implementation does not support system calls.

On Windows, system passes the command to the Windows command interpreter (CMD.EXE or COMMAND.COM), hence Unixy shell tricks will not work.

On Unix systems, see waitForProcess for the meaning of exit codes when the process died as the result of a signal.

rawSystem :: String -> [String] -> IO ExitCode Source

The computation rawSystem cmd args runs the operating system command cmd in such a way that it receives as arguments the args strings exactly as given, with no funny escaping or shell meta-syntax expansion. It will therefore behave more portably between operating systems than system.

The return codes and possible failures are the same as for system.