*** empty log message ***

[emacs.git] / lispref / processes.texi

@c -*-texinfo-*-
@c This is part of the GNU Emacs Lisp Reference Manual.
@c Copyright (C) 1990, 1991, 1992, 1993, 1994, 1995, 1998, 1999, 2002, 2003,
@c   2004, 2005 Free Software Foundation, Inc.
@c See the file elisp.texi for copying conditions.
@setfilename ../info/processes
@node Processes, Display, Abbrevs, Top
@chapter Processes
@cindex child process
@cindex parent process
@cindex subprocess
@cindex process

  In the terminology of operating systems, a @dfn{process} is a space in
which a program can execute.  Emacs runs in a process.  Emacs Lisp
programs can invoke other programs in processes of their own.  These are
called @dfn{subprocesses} or @dfn{child processes} of the Emacs process,
which is their @dfn{parent process}.

  A subprocess of Emacs may be @dfn{synchronous} or @dfn{asynchronous},
depending on how it is created.  When you create a synchronous
subprocess, the Lisp program waits for the subprocess to terminate
before continuing execution.  When you create an asynchronous
subprocess, it can run in parallel with the Lisp program.  This kind of
subprocess is represented within Emacs by a Lisp object which is also
called a ``process''.  Lisp programs can use this object to communicate
with the subprocess or to control it.  For example, you can send
signals, obtain status information, receive output from the process, or
send input to it.

@defun processp object
This function returns @code{t} if @var{object} is a process,
@code{nil} otherwise.
@end defun

@menu
* Subprocess Creation::      Functions that start subprocesses.
* Shell Arguments::          Quoting an argument to pass it to a shell.
* Synchronous Processes::    Details of using synchronous subprocesses.
* Asynchronous Processes::   Starting up an asynchronous subprocess.
* Deleting Processes::       Eliminating an asynchronous subprocess.
* Process Information::      Accessing run-status and other attributes.
* Input to Processes::       Sending input to an asynchronous subprocess.
* Signals to Processes::     Stopping, continuing or interrupting
                               an asynchronous subprocess.
* Output from Processes::    Collecting output from an asynchronous subprocess.
* Sentinels::                Sentinels run when process run-status changes.
* Query Before Exit::        Whether to query if exiting will kill a process.
* Transaction Queues::       Transaction-based communication with subprocesses.
* Network::                  Opening network connections.
* Network Servers::          Network servers let Emacs accept net connections.
* Datagrams::                UDP network connections.
* Low-Level Network::        Lower-level but more general function
                               to create connections and servers.
* Misc Network::             Additional relevant functions for network connections.
* Byte Packing::             Using bindat to pack and unpack binary data.
@end menu

@node Subprocess Creation
@section Functions that Create Subprocesses

  There are three functions that create a new subprocess in which to run
a program.  One of them, @code{start-process}, creates an asynchronous
process and returns a process object (@pxref{Asynchronous Processes}).
The other two, @code{call-process} and @code{call-process-region},
create a synchronous process and do not return a process object
(@pxref{Synchronous Processes}).

  Synchronous and asynchronous processes are explained in the following
sections.  Since the three functions are all called in a similar
fashion, their common arguments are described here.

@cindex execute program
@cindex @code{PATH} environment variable
@cindex @code{HOME} environment variable
  In all cases, the function's @var{program} argument specifies the
program to be run.  An error is signaled if the file is not found or
cannot be executed.  If the file name is relative, the variable
@code{exec-path} contains a list of directories to search.  Emacs
initializes @code{exec-path} when it starts up, based on the value of
the environment variable @code{PATH}.  The standard file name
constructs, @samp{~}, @samp{.}, and @samp{..}, are interpreted as
usual in @code{exec-path}, but environment variable substitutions
(@samp{$HOME}, etc.) are not recognized; use
@code{substitute-in-file-name} to perform them (@pxref{File Name
Expansion}).  @code{nil} in this list refers to
@code{default-directory}.

  Executing a program can also try adding suffixes to the specified
name:

@defvar exec-suffixes
This variable is a list of suffixes (strings) to try adding to the
specified program file name.  The list should include @code{""} if you
want the name to be tried exactly as specified.  The default value is
system-dependent.
@end defvar

  @strong{Please note:} The argument @var{program} contains only the
name of the program; it may not contain any command-line arguments.  You
must use @var{args} to provide those.

  Each of the subprocess-creating functions has a @var{buffer-or-name}
argument which specifies where the standard output from the program will
go.  It should be a buffer or a buffer name; if it is a buffer name,
that will create the buffer if it does not already exist.  It can also
be @code{nil}, which says to discard the output unless a filter function
handles it.  (@xref{Filter Functions}, and @ref{Read and Print}.)
Normally, you should avoid having multiple processes send output to the
same buffer because their output would be intermixed randomly.

@cindex program arguments
  All three of the subprocess-creating functions have a @code{&rest}
argument, @var{args}.  The @var{args} must all be strings, and they are
supplied to @var{program} as separate command line arguments.  Wildcard
characters and other shell constructs have no special meanings in these
strings, since the strings are passed directly to the specified program.

  The subprocess gets its current directory from the value of
@code{default-directory} (@pxref{File Name Expansion}).

@cindex environment variables, subprocesses
  The subprocess inherits its environment from Emacs, but you can
specify overrides for it with @code{process-environment}.  @xref{System
Environment}.

@defvar exec-directory
@pindex movemail
The value of this variable is a string, the name of a directory that
contains programs that come with GNU Emacs, programs intended for Emacs
to invoke.  The program @code{movemail} is an example of such a program;
Rmail uses it to fetch new mail from an inbox.
@end defvar

@defopt exec-path
The value of this variable is a list of directories to search for
programs to run in subprocesses.  Each element is either the name of a
directory (i.e., a string), or @code{nil}, which stands for the default
directory (which is the value of @code{default-directory}).
@cindex program directories

The value of @code{exec-path} is used by @code{call-process} and
@code{start-process} when the @var{program} argument is not an absolute
file name.
@end defopt

@node Shell Arguments
@section Shell Arguments

  Lisp programs sometimes need to run a shell and give it a command
that contains file names that were specified by the user.  These
programs ought to be able to support any valid file name.  But the shell
gives special treatment to certain characters, and if these characters
occur in the file name, they will confuse the shell.  To handle these
characters, use the function @code{shell-quote-argument}:

@defun shell-quote-argument argument
This function returns a string which represents, in shell syntax,
an argument whose actual contents are @var{argument}.  It should
work reliably to concatenate the return value into a shell command
and then pass it to a shell for execution.

Precisely what this function does depends on your operating system.  The
function is designed to work with the syntax of your system's standard
shell; if you use an unusual shell, you will need to redefine this
function.

@example
;; @r{This example shows the behavior on GNU and Unix systems.}
(shell-quote-argument "foo > bar")
     @result{} "foo\\ \\>\\ bar"

;; @r{This example shows the behavior on MS-DOS and MS-Windows.}
(shell-quote-argument "foo > bar")
     @result{} "\"foo > bar\""
@end example

Here's an example of using @code{shell-quote-argument} to construct
a shell command:

@example
(concat "diff -c "
        (shell-quote-argument oldfile)
        " "
        (shell-quote-argument newfile))
@end example
@end defun

@node Synchronous Processes
@section Creating a Synchronous Process
@cindex synchronous subprocess

  After a @dfn{synchronous process} is created, Emacs waits for the
process to terminate before continuing.  Starting Dired on GNU or
Unix@footnote{On other systems, Emacs uses a Lisp emulation of
@code{ls}; see @ref{Contents of Directories}.} is an example of this: it
runs @code{ls} in a synchronous process, then modifies the output
slightly.  Because the process is synchronous, the entire directory
listing arrives in the buffer before Emacs tries to do anything with it.

  While Emacs waits for the synchronous subprocess to terminate, the
user can quit by typing @kbd{C-g}.  The first @kbd{C-g} tries to kill
the subprocess with a @code{SIGINT} signal; but it waits until the
subprocess actually terminates before quitting.  If during that time the
user types another @kbd{C-g}, that kills the subprocess instantly with
@code{SIGKILL} and quits immediately (except on MS-DOS, where killing
other processes doesn't work).  @xref{Quitting}.

  The synchronous subprocess functions return an indication of how the
process terminated.

  The output from a synchronous subprocess is generally decoded using a
coding system, much like text read from a file.  The input sent to a
subprocess by @code{call-process-region} is encoded using a coding
system, much like text written into a file.  @xref{Coding Systems}.

@defun call-process program &optional infile destination display &rest args
This function calls @var{program} in a separate process and waits for
it to finish.

The standard input for the process comes from file @var{infile} if
@var{infile} is not @code{nil}, and from the null device otherwise.
The argument @var{destination} says where to put the process output.
Here are the possibilities:

@table @asis
@item a buffer
Insert the output in that buffer, before point.  This includes both the
standard output stream and the standard error stream of the process.

@item a string
Insert the output in a buffer with that name, before point.

@item @code{t}
Insert the output in the current buffer, before point.

@item @code{nil}
Discard the output.

@item 0
Discard the output, and return @code{nil} immediately without waiting
for the subprocess to finish.

In this case, the process is not truly synchronous, since it can run in
parallel with Emacs; but you can think of it as synchronous in that
Emacs is essentially finished with the subprocess as soon as this
function returns.

MS-DOS doesn't support asynchronous subprocesses, so this option doesn't
work there.

@item @code{(@var{real-destination} @var{error-destination})}
Keep the standard output stream separate from the standard error stream;
deal with the ordinary output as specified by @var{real-destination},
and dispose of the error output according to @var{error-destination}.
If @var{error-destination} is @code{nil}, that means to discard the
error output, @code{t} means mix it with the ordinary output, and a
string specifies a file name to redirect error output into.

You can't directly specify a buffer to put the error output in; that is
too difficult to implement.  But you can achieve this result by sending
the error output to a temporary file and then inserting the file into a
buffer.
@end table

If @var{display} is non-@code{nil}, then @code{call-process} redisplays
the buffer as output is inserted.  (However, if the coding system chosen
for decoding output is @code{undecided}, meaning deduce the encoding
from the actual data, then redisplay sometimes cannot continue once
non-@acronym{ASCII} characters are encountered.  There are fundamental
reasons why it is hard to fix this; see @ref{Output from Processes}.)

Otherwise the function @code{call-process} does no redisplay, and the
results become visible on the screen only when Emacs redisplays that
buffer in the normal course of events.

The remaining arguments, @var{args}, are strings that specify command
line arguments for the program.

The value returned by @code{call-process} (unless you told it not to
wait) indicates the reason for process termination.  A number gives the
exit status of the subprocess; 0 means success, and any other value
means failure.  If the process terminated with a signal,
@code{call-process} returns a string describing the signal.

In the examples below, the buffer @samp{foo} is current.

@smallexample
@group
(call-process "pwd" nil t)
     @result{} 0

---------- Buffer: foo ----------
/usr/user/lewis/manual
---------- Buffer: foo ----------
@end group

@group
(call-process "grep" nil "bar" nil "lewis" "/etc/passwd")
     @result{} 0

---------- Buffer: bar ----------
lewis:5LTsHm66CSWKg:398:21:Bil Lewis:/user/lewis:/bin/csh

---------- Buffer: bar ----------
@end group
@end smallexample

Here is a good example of the use of @code{call-process}, which used to
be found in the definition of @code{insert-directory}:

@smallexample
@group
(call-process insert-directory-program nil t nil @var{switches}
              (if full-directory-p
                  (concat (file-name-as-directory file) ".")
                file))
@end group
@end smallexample
@end defun

@defun process-file program &optional infile buffer display &rest args
This function processes files synchronously in a separate process.  It
is similar to @code{call-process} but may invoke a file handler based
on the value of the variable @code{default-directory}.  The current
working directory of the subprocess is @code{default-directory}.

The arguments are handled in almost the same way as for
@code{call-process}, with the following differences:

Some file handlers may not support all combinations and forms of the
arguments @var{infile}, @var{buffer}, and @var{display}.  For example,
some file handlers might behave as if @var{display} were @code{nil},
regardless of the value actually passed.  As another example, some
file handlers might not support separating standard output and error
output by way of the @var{buffer} argument.

If a file handler is invoked, it determines the program to run based
on the first argument @var{program}.  For instance, consider that a
handler for remote files is invoked.  Then the path that is used for
searching the program might be different than @code{exec-path}.

The second argument @var{infile} may invoke a file handler.  The file
handler could be different from the handler chosen for the
@code{process-file} function itself.  (For example,
@code{default-directory} could be on a remote host, whereas
@var{infile} is on another remote host.  Or @code{default-directory}
could be non-special, whereas @var{infile} is on a remote host.)

If @var{buffer} has the form @code{(@var{real-destination}
@var{error-destination})}, and @var{error-destination} names a file,
then the same remarks as for @var{infile} apply.

The remaining arguments (@var{args}) will be passed to the process
verbatim.  Emacs is not involved in processing file names that are
present in @var{args}.  To avoid confusion, it may be best to avoid
absolute file names in @var{args}, but rather to specify all file
names as relative to @code{default-directory}.  The function
@code{file-relative-name} is useful for constructing such relative
file names.
@end defun

@defun call-process-region start end program &optional delete destination display &rest args
This function sends the text from @var{start} to @var{end} as
standard input to a process running @var{program}.  It deletes the text
sent if @var{delete} is non-@code{nil}; this is useful when
@var{destination} is @code{t}, to insert the output in the current
buffer in place of the input.

The arguments @var{destination} and @var{display} control what to do
with the output from the subprocess, and whether to update the display
as it comes in.  For details, see the description of
@code{call-process}, above.  If @var{destination} is the integer 0,
@code{call-process-region} discards the output and returns @code{nil}
immediately, without waiting for the subprocess to finish (this only
works if asynchronous subprocesses are supported).

The remaining arguments, @var{args}, are strings that specify command
line arguments for the program.

The return value of @code{call-process-region} is just like that of
@code{call-process}: @code{nil} if you told it to return without
waiting; otherwise, a number or string which indicates how the
subprocess terminated.

In the following example, we use @code{call-process-region} to run the
@code{cat} utility, with standard input being the first five characters
in buffer @samp{foo} (the word @samp{input}).  @code{cat} copies its
standard input into its standard output.  Since the argument
@var{destination} is @code{t}, this output is inserted in the current
buffer.

@smallexample
@group
---------- Buffer: foo ----------
input@point{}
---------- Buffer: foo ----------
@end group

@group
(call-process-region 1 6 "cat" nil t)
     @result{} 0

---------- Buffer: foo ----------
inputinput@point{}
---------- Buffer: foo ----------
@end group
@end smallexample

  The @code{shell-command-on-region} command uses
@code{call-process-region} like this:

@smallexample
@group
(call-process-region
 start end
 shell-file-name      ; @r{Name of program.}
 nil                  ; @r{Do not delete region.}
 buffer               ; @r{Send output to @code{buffer}.}
 nil                  ; @r{No redisplay during output.}
 "-c" command)        ; @r{Arguments for the shell.}
@end group
@end smallexample
@end defun

@defun call-process-shell-command command &optional infile destination display &rest args
This function executes the shell command @var{command} synchronously
in a separate process.  The final arguments @var{args} are additional
arguments to add at the end of @var{command}.  The other arguments
are handled as in @code{call-process}.
@end defun

@defun shell-command-to-string command
This function executes @var{command} (a string) as a shell command,
then returns the command's output as a string.
@end defun

@node Asynchronous Processes
@section Creating an Asynchronous Process
@cindex asynchronous subprocess

  After an @dfn{asynchronous process} is created, Emacs and the subprocess
both continue running immediately.  The process thereafter runs
in parallel with Emacs, and the two can communicate with each other
using the functions described in the following sections.  However,
communication is only partially asynchronous: Emacs sends data to the
process only when certain functions are called, and Emacs accepts data
from the process only when Emacs is waiting for input or for a time
delay.

  Here we describe how to create an asynchronous process.

@defun start-process name buffer-or-name program &rest args
This function creates a new asynchronous subprocess and starts the
program @var{program} running in it.  It returns a process object that
stands for the new subprocess in Lisp.  The argument @var{name}
specifies the name for the process object; if a process with this name
already exists, then @var{name} is modified (by appending @samp{<1>},
etc.) to be unique.  The buffer @var{buffer-or-name} is the buffer to
associate with the process.

The remaining arguments, @var{args}, are strings that specify command
line arguments for the program.

In the example below, the first process is started and runs (rather,
sleeps) for 100 seconds.  Meanwhile, the second process is started, and
given the name @samp{my-process<1>} for the sake of uniqueness.  It
inserts the directory listing at the end of the buffer @samp{foo},
before the first process finishes.  Then it finishes, and a message to
that effect is inserted in the buffer.  Much later, the first process
finishes, and another message is inserted in the buffer for it.

@smallexample
@group
(start-process "my-process" "foo" "sleep" "100")
     @result{} #<process my-process>
@end group

@group
(start-process "my-process" "foo" "ls" "-l" "/user/lewis/bin")
     @result{} #<process my-process<1>>

---------- Buffer: foo ----------
total 2
lrwxrwxrwx  1 lewis     14 Jul 22 10:12 gnuemacs --> /emacs
-rwxrwxrwx  1 lewis     19 Jul 30 21:02 lemon

Process my-process<1> finished

Process my-process finished
---------- Buffer: foo ----------
@end group
@end smallexample
@end defun

@defun start-process-shell-command name buffer-or-name command &rest command-args
This function is like @code{start-process} except that it uses a shell
to execute the specified command.  The argument @var{command} is a shell
command name, and @var{command-args} are the arguments for the shell
command.  The variable @code{shell-file-name} specifies which shell to
use.

The point of running a program through the shell, rather than directly
with @code{start-process}, is so that you can employ shell features such
as wildcards in the arguments.  It follows that if you include an
arbitrary user-specified arguments in the command, you should quote it
with @code{shell-quote-argument} first, so that any special shell
characters do @emph{not} have their special shell meanings.  @xref{Shell
Arguments}.
@end defun

@defvar process-connection-type
@cindex pipes
@cindex @acronym{PTY}s
This variable controls the type of device used to communicate with
asynchronous subprocesses.  If it is non-@code{nil}, then @acronym{PTY}s are
used, when available.  Otherwise, pipes are used.

@acronym{PTY}s are usually preferable for processes visible to the user, as
in Shell mode, because they allow job control (@kbd{C-c}, @kbd{C-z},
etc.) to work between the process and its children, whereas pipes do
not.  For subprocesses used for internal purposes by programs, it is
often better to use a pipe, because they are more efficient.  In
addition, the total number of @acronym{PTY}s is limited on many systems and
it is good not to waste them.

The value of @code{process-connection-type} takes effect when
@code{start-process} is called.  So you can specify how to communicate
with one subprocess by binding the variable around the call to
@code{start-process}.

@smallexample
@group
(let ((process-connection-type nil))  ; @r{Use a pipe.}
  (start-process @dots{}))
@end group
@end smallexample

To determine whether a given subprocess actually got a pipe or a
@acronym{PTY}, use the function @code{process-tty-name} (@pxref{Process
Information}).
@end defvar

@node Deleting Processes
@section Deleting Processes
@cindex deleting processes

  @dfn{Deleting a process} disconnects Emacs immediately from the
subprocess.  Processes are deleted automatically after they terminate,
but not necessarily right away.  You can delete a process explicitly
at any time.  If you delete a terminated process explicitly before it
is deleted automatically, no harm results.  Deleting a running
process sends a signal to terminate it (and its child processes if
any), and calls the process sentinel if it has one.  @xref{Sentinels}.

  When a process is deleted, the process object itself continues to
exist as long as other Lisp objects point to it.  All the Lisp
primitives that work on process objects accept deleted processes, but
those that do I/O or send signals will report an error.  The process
mark continues to point to the same place as before, usually into a
buffer where output from the process was being inserted.

@defopt delete-exited-processes
This variable controls automatic deletion of processes that have
terminated (due to calling @code{exit} or to a signal).  If it is
@code{nil}, then they continue to exist until the user runs
@code{list-processes}.  Otherwise, they are deleted immediately after
they exit.
@end defopt

@defun delete-process process
This function deletes a process, killing it with a @code{SIGKILL}
signal.  The argument may be a process, the name of a process, a
buffer, or the name of a buffer.  (A buffer or buffer-name stands for
the process that @code{get-buffer-process} returns.)  Calling
@code{delete-process} on a running process terminates it, updates the
process status, and runs the sentinel (if any) immediately.  If the
process has already terminated, calling @code{delete-process} has no
effect on its status, or on the running of its sentinel (which will
happen sooner or later).

@smallexample
@group
(delete-process "*shell*")
     @result{} nil
@end group
@end smallexample
@end defun

@node Process Information
@section Process Information

  Several functions return information about processes.
@code{list-processes} is provided for interactive use.

@deffn Command list-processes &optional query-only
This command displays a listing of all living processes.  In addition,
it finally deletes any process whose status was @samp{Exited} or
@samp{Signaled}.  It returns @code{nil}.

If @var{query-only} is non-@code{nil} then it lists only processes
whose query flag is non-@code{nil}.  @xref{Query Before Exit}.
@end deffn

@defun process-list
This function returns a list of all processes that have not been deleted.

@smallexample
@group
(process-list)
     @result{} (#<process display-time> #<process shell>)
@end group
@end smallexample
@end defun

@defun get-process name
This function returns the process named @var{name}, or @code{nil} if
there is none.  An error is signaled if @var{name} is not a string.

@smallexample
@group
(get-process "shell")
     @result{} #<process shell>
@end group
@end smallexample
@end defun

@defun process-command process
This function returns the command that was executed to start
@var{process}.  This is a list of strings, the first string being the
program executed and the rest of the strings being the arguments that
were given to the program.

@smallexample
@group
(process-command (get-process "shell"))
     @result{} ("/bin/csh" "-i")
@end group
@end smallexample
@end defun

@defun process-id process
This function returns the @acronym{PID} of @var{process}.  This is an
integer that distinguishes the process @var{process} from all other
processes running on the same computer at the current time.  The
@acronym{PID} of a process is chosen by the operating system kernel when the
process is started and remains constant as long as the process exists.
@end defun

@defun process-name process
This function returns the name of @var{process}.
@end defun

@defun process-status process-name
This function returns the status of @var{process-name} as a symbol.
The argument @var{process-name} must be a process, a buffer, a
process name (string) or a buffer name (string).

The possible values for an actual subprocess are:

@table @code
@item run
for a process that is running.
@item stop
for a process that is stopped but continuable.
@item exit
for a process that has exited.
@item signal
for a process that has received a fatal signal.
@item open
for a network connection that is open.
@item closed
for a network connection that is closed.  Once a connection
is closed, you cannot reopen it, though you might be able to open
a new connection to the same place.
@item connect
for a non-blocking connection that is waiting to complete.
@item failed
for a non-blocking connection that has failed to complete.
@item listen
for a network server that is listening.
@item nil
if @var{process-name} is not the name of an existing process.
@end table

@smallexample
@group
(process-status "shell")
     @result{} run
@end group
@group
(process-status (get-buffer "*shell*"))
     @result{} run
@end group
@group
x
     @result{} #<process xx<1>>
(process-status x)
     @result{} exit
@end group
@end smallexample

For a network connection, @code{process-status} returns one of the symbols
@code{open} or @code{closed}.  The latter means that the other side
closed the connection, or Emacs did @code{delete-process}.
@end defun

@defun process-exit-status process
This function returns the exit status of @var{process} or the signal
number that killed it.  (Use the result of @code{process-status} to
determine which of those it is.)  If @var{process} has not yet
terminated, the value is 0.
@end defun

@defun process-tty-name process
This function returns the terminal name that @var{process} is using for
its communication with Emacs---or @code{nil} if it is using pipes
instead of a terminal (see @code{process-connection-type} in
@ref{Asynchronous Processes}).
@end defun

@defun process-coding-system process
@anchor{Coding systems for a subprocess}
This function returns a cons cell describing the coding systems in use
for decoding output from @var{process} and for encoding input to
@var{process} (@pxref{Coding Systems}).  The value has this form:

@example
(@var{coding-system-for-decoding} . @var{coding-system-for-encoding})
@end example
@end defun

@defun set-process-coding-system process &optional decoding-system encoding-system
This function specifies the coding systems to use for subsequent output
from and input to @var{process}.  It will use @var{decoding-system} to
decode subprocess output, and @var{encoding-system} to encode subprocess
input.
@end defun

  Every process also has a property list that you can use to store
miscellaneous values associated with the process.

@defun process-get process propname
This function returns the value of the @var{propname} property
of @var{process}.
@end defun

@defun process-put process propname value
This function sets the value of the @var{propname} property
of @var{process} to @var{value}.
@end defun

@defun process-plist process
This function returns the process plist of @var{process}.
@end defun

@defun set-process-plist process plist
This function sets the process plist of @var{process} to @var{plist}.
@end defun

@node Input to Processes
@section Sending Input to Processes
@cindex process input

  Asynchronous subprocesses receive input when it is sent to them by
Emacs, which is done with the functions in this section.  You must
specify the process to send input to, and the input data to send.  The
data appears on the ``standard input'' of the subprocess.

  Some operating systems have limited space for buffered input in a
@acronym{PTY}.  On these systems, Emacs sends an @acronym{EOF}
periodically amidst the other characters, to force them through.  For
most programs, these @acronym{EOF}s do no harm.

  Subprocess input is normally encoded using a coding system before the
subprocess receives it, much like text written into a file.  You can use
@code{set-process-coding-system} to specify which coding system to use
(@pxref{Process Information}).  Otherwise, the coding system comes from
@code{coding-system-for-write}, if that is non-@code{nil}; or else from
the defaulting mechanism (@pxref{Default Coding Systems}).

  Sometimes the system is unable to accept input for that process,
because the input buffer is full.  When this happens, the send functions
wait a short while, accepting output from subprocesses, and then try
again.  This gives the subprocess a chance to read more of its pending
input and make space in the buffer.  It also allows filters, sentinels
and timers to run---so take account of that in writing your code.

  In these functions, the @var{process} argument can be a process or
the name of a process, or a buffer or buffer name (which stands
for a process via @code{get-buffer-process}).  @code{nil} means
the current buffer's process.

@defun process-send-string process string
This function sends @var{process} the contents of @var{string} as
standard input.  If it is @code{nil}, the current buffer's process is used.

  The function returns @code{nil}.

@smallexample
@group
(process-send-string "shell<1>" "ls\n")
     @result{} nil
@end group


@group
---------- Buffer: *shell* ----------
...
introduction.texi               syntax-tables.texi~
introduction.texi~              text.texi
introduction.txt                text.texi~
...
---------- Buffer: *shell* ----------
@end group
@end smallexample
@end defun

@defun process-send-region process start end
This function sends the text in the region defined by @var{start} and
@var{end} as standard input to @var{process}.

An error is signaled unless both @var{start} and @var{end} are
integers or markers that indicate positions in the current buffer.  (It
is unimportant which number is larger.)
@end defun

@defun process-send-eof &optional process
This function makes @var{process} see an end-of-file in its
input.  The @acronym{EOF} comes after any text already sent to it.

The function returns @var{process}.

@smallexample
@group
(process-send-eof "shell")
     @result{} "shell"
@end group
@end smallexample
@end defun

@defun process-running-child-p process
@tindex process-running-child-p process
This function will tell you whether a subprocess has given control of
its terminal to its own child process.  The value is @code{t} if this is
true, or if Emacs cannot tell; it is @code{nil} if Emacs can be certain
that this is not so.
@end defun

@node Signals to Processes
@section Sending Signals to Processes
@cindex process signals
@cindex sending signals
@cindex signals

  @dfn{Sending a signal} to a subprocess is a way of interrupting its
activities.  There are several different signals, each with its own
meaning.  The set of signals and their names is defined by the operating
system.  For example, the signal @code{SIGINT} means that the user has
typed @kbd{C-c}, or that some analogous thing has happened.

  Each signal has a standard effect on the subprocess.  Most signals
kill the subprocess, but some stop or resume execution instead.  Most
signals can optionally be handled by programs; if the program handles
the signal, then we can say nothing in general about its effects.

  You can send signals explicitly by calling the functions in this
section.  Emacs also sends signals automatically at certain times:
killing a buffer sends a @code{SIGHUP} signal to all its associated
processes; killing Emacs sends a @code{SIGHUP} signal to all remaining
processes.  (@code{SIGHUP} is a signal that usually indicates that the
user hung up the phone.)

  Each of the signal-sending functions takes two optional arguments:
@var{process} and @var{current-group}.

  The argument @var{process} must be either a process, a process
name, a buffer, a buffer name, or @code{nil}.  A buffer or buffer name
stands for a process through @code{get-buffer-process}.  @code{nil}
stands for the process associated with the current buffer.  An error
is signaled if @var{process} does not identify a process.

  The argument @var{current-group} is a flag that makes a difference
when you are running a job-control shell as an Emacs subprocess.  If it
is non-@code{nil}, then the signal is sent to the current process-group
of the terminal that Emacs uses to communicate with the subprocess.  If
the process is a job-control shell, this means the shell's current
subjob.  If it is @code{nil}, the signal is sent to the process group of
the immediate subprocess of Emacs.  If the subprocess is a job-control
shell, this is the shell itself.

  The flag @var{current-group} has no effect when a pipe is used to
communicate with the subprocess, because the operating system does not
support the distinction in the case of pipes.  For the same reason,
job-control shells won't work when a pipe is used.  See
@code{process-connection-type} in @ref{Asynchronous Processes}.

@defun interrupt-process &optional process current-group
This function interrupts the process @var{process} by sending the
signal @code{SIGINT}.  Outside of Emacs, typing the ``interrupt
character'' (normally @kbd{C-c} on some systems, and @code{DEL} on
others) sends this signal.  When the argument @var{current-group} is
non-@code{nil}, you can think of this function as ``typing @kbd{C-c}''
on the terminal by which Emacs talks to the subprocess.
@end defun

@defun kill-process &optional process current-group
This function kills the process @var{process} by sending the
signal @code{SIGKILL}.  This signal kills the subprocess immediately,
and cannot be handled by the subprocess.
@end defun

@defun quit-process &optional process current-group
This function sends the signal @code{SIGQUIT} to the process
@var{process}.  This signal is the one sent by the ``quit
character'' (usually @kbd{C-b} or @kbd{C-\}) when you are not inside
Emacs.
@end defun

@defun stop-process &optional process current-group
This function stops the process @var{process} by sending the
signal @code{SIGTSTP}.  Use @code{continue-process} to resume its
execution.

Outside of Emacs, on systems with job control, the ``stop character''
(usually @kbd{C-z}) normally sends this signal.  When
@var{current-group} is non-@code{nil}, you can think of this function as
``typing @kbd{C-z}'' on the terminal Emacs uses to communicate with the
subprocess.
@end defun

@defun continue-process &optional process current-group
This function resumes execution of the process @var{process} by sending
it the signal @code{SIGCONT}.  This presumes that @var{process} was
stopped previously.
@end defun

@c Emacs 19 feature
@defun signal-process process signal
This function sends a signal to process @var{process}.  The argument
@var{signal} specifies which signal to send; it should be an integer.

The @var{process} argument can be a system process @acronym{ID}; that
allows you to send signals to processes that are not children of
Emacs.
@end defun

@node Output from Processes
@section Receiving Output from Processes
@cindex process output
@cindex output from processes

  There are two ways to receive the output that a subprocess writes to
its standard output stream.  The output can be inserted in a buffer,
which is called the associated buffer of the process, or a function
called the @dfn{filter function} can be called to act on the output.  If
the process has no buffer and no filter function, its output is
discarded.

  When a subprocess terminates, Emacs reads any pending output,
then stops reading output from that subprocess.  Therefore, if the
subprocess has children that are still live and still producing
output, Emacs won't receive that output.

  Output from a subprocess can arrive only while Emacs is waiting: when
reading terminal input, in @code{sit-for} and @code{sleep-for}
(@pxref{Waiting}), and in @code{accept-process-output} (@pxref{Accepting
Output}).  This minimizes the problem of timing errors that usually
plague parallel programming.  For example, you can safely create a
process and only then specify its buffer or filter function; no output
can arrive before you finish, if the code in between does not call any
primitive that waits.

@defvar process-adaptive-read-buffering
On some systems, when Emacs reads the output from a subprocess, the
output data is read in very small blocks, potentially resulting in
very poor performance.  This behavior can be remedied to some extent
by setting the variable @var{process-adaptive-read-buffering} to a
non-@code{nil} value (the default), as it will automatically delay reading
from such processes, thus allowing them to produce more output before
Emacs tries to read it.
@end defvar

  It is impossible to separate the standard output and standard error
streams of the subprocess, because Emacs normally spawns the subprocess
inside a pseudo-TTY, and a pseudo-TTY has only one output channel.  If
you want to keep the output to those streams separate, you should
redirect one of them to a file---for example, by using an appropriate
shell command.

@menu
* Process Buffers::         If no filter, output is put in a buffer.
* Filter Functions::        Filter functions accept output from the process.
* Decoding Output::         Filters can get unibyte or multibyte strings.
* Accepting Output::        How to wait until process output arrives.
@end menu

@node Process Buffers
@subsection Process Buffers

  A process can (and usually does) have an @dfn{associated buffer},
which is an ordinary Emacs buffer that is used for two purposes: storing
the output from the process, and deciding when to kill the process.  You
can also use the buffer to identify a process to operate on, since in
normal practice only one process is associated with any given buffer.
Many applications of processes also use the buffer for editing input to
be sent to the process, but this is not built into Emacs Lisp.

  Unless the process has a filter function (@pxref{Filter Functions}),
its output is inserted in the associated buffer.  The position to insert
the output is determined by the @code{process-mark}, which is then
updated to point to the end of the text just inserted.  Usually, but not
always, the @code{process-mark} is at the end of the buffer.

@defun process-buffer process
This function returns the associated buffer of the process
@var{process}.

@smallexample
@group
(process-buffer (get-process "shell"))
     @result{} #<buffer *shell*>
@end group
@end smallexample
@end defun

@defun process-mark process
This function returns the process marker for @var{process}, which is the
marker that says where to insert output from the process.

If @var{process} does not have a buffer, @code{process-mark} returns a
marker that points nowhere.

Insertion of process output in a buffer uses this marker to decide where
to insert, and updates it to point after the inserted text.  That is why
successive batches of output are inserted consecutively.

Filter functions normally should use this marker in the same fashion
as is done by direct insertion of output in the buffer.  A good
example of a filter function that uses @code{process-mark} is found at
the end of the following section.

When the user is expected to enter input in the process buffer for
transmission to the process, the process marker separates the new input
from previous output.
@end defun

@defun set-process-buffer process buffer
This function sets the buffer associated with @var{process} to
@var{buffer}.  If @var{buffer} is @code{nil}, the process becomes
associated with no buffer.
@end defun

@defun get-buffer-process buffer-or-name
This function returns a nondeleted process associated with the buffer
specified by @var{buffer-or-name}.  If there are several processes
associated with it, this function chooses one (currently, the one most
recently created, but don't count on that).  Deletion of a process
(see @code{delete-process}) makes it ineligible for this function to
return.

It is usually a bad idea to have more than one process associated with
the same buffer.

@smallexample
@group
(get-buffer-process "*shell*")
     @result{} #<process shell>
@end group
@end smallexample

Killing the process's buffer deletes the process, which kills the
subprocess with a @code{SIGHUP} signal (@pxref{Signals to Processes}).
@end defun

@node Filter Functions
@subsection Process Filter Functions
@cindex filter function
@cindex process filter

  A process @dfn{filter function} is a function that receives the
standard output from the associated process.  If a process has a filter,
then @emph{all} output from that process is passed to the filter.  The
process buffer is used directly for output from the process only when
there is no filter.

  The filter function can only be called when Emacs is waiting for
something, because process output arrives only at such times.  Emacs
waits when reading terminal input, in @code{sit-for} and
@code{sleep-for} (@pxref{Waiting}), and in @code{accept-process-output}
(@pxref{Accepting Output}).

  A filter function must accept two arguments: the associated process
and a string, which is output just received from it.  The function is
then free to do whatever it chooses with the output.

  Quitting is normally inhibited within a filter function---otherwise,
the effect of typing @kbd{C-g} at command level or to quit a user
command would be unpredictable.  If you want to permit quitting inside
a filter function, bind @code{inhibit-quit} to @code{nil}.  In most
cases, the right way to do this is with the macro
@code{with-local-quit}.  @xref{Quitting}.

  If an error happens during execution of a filter function, it is
caught automatically, so that it doesn't stop the execution of whatever
program was running when the filter function was started.  However, if
@code{debug-on-error} is non-@code{nil}, the error-catching is turned
off.  This makes it possible to use the Lisp debugger to debug the
filter function.  @xref{Debugger}.

  Many filter functions sometimes or always insert the text in the
process's buffer, mimicking the actions of Emacs when there is no
filter.  Such filter functions need to use @code{set-buffer} in order to
be sure to insert in that buffer.  To avoid setting the current buffer
semipermanently, these filter functions must save and restore the
current buffer.  They should also update the process marker, and in some
cases update the value of point.  Here is how to do these things:

@smallexample
@group
(defun ordinary-insertion-filter (proc string)
  (with-current-buffer (process-buffer proc)
    (let ((moving (= (point) (process-mark proc))))
@end group
@group
      (save-excursion
        ;; @r{Insert the text, advancing the process marker.}
        (goto-char (process-mark proc))
        (insert string)
        (set-marker (process-mark proc) (point)))
      (if moving (goto-char (process-mark proc))))))
@end group
@end smallexample

@noindent
The reason to use @code{with-current-buffer}, rather than using
@code{save-excursion} to save and restore the current buffer, is so as
to preserve the change in point made by the second call to
@code{goto-char}.

  To make the filter force the process buffer to be visible whenever new
text arrives, insert the following line just before the
@code{with-current-buffer} construct:

@smallexample
(display-buffer (process-buffer proc))
@end smallexample

  To force point to the end of the new output, no matter where it was
previously, eliminate the variable @code{moving} and call
@code{goto-char} unconditionally.

  In earlier Emacs versions, every filter function that did regular
expression searching or matching had to explicitly save and restore the
match data.  Now Emacs does this automatically for filter functions;
they never need to do it explicitly.  @xref{Match Data}.

  A filter function that writes the output into the buffer of the
process should check whether the buffer is still alive.  If it tries to
insert into a dead buffer, it will get an error.  The expression
@code{(buffer-name (process-buffer @var{process}))} returns @code{nil}
if the buffer is dead.

  The output to the function may come in chunks of any size.  A program
that produces the same output twice in a row may send it as one batch of
200 characters one time, and five batches of 40 characters the next.  If
the filter looks for certain text strings in the subprocess output, make
sure to handle the case where one of these strings is split across two
or more batches of output.

@defun set-process-filter process filter
This function gives @var{process} the filter function @var{filter}.  If
@var{filter} is @code{nil}, it gives the process no filter.
@end defun

@defun process-filter process
This function returns the filter function of @var{process}, or @code{nil}
if it has none.
@end defun

  Here is an example of use of a filter function:

@smallexample
@group
(defun keep-output (process output)
   (setq kept (cons output kept)))
     @result{} keep-output
@end group
@group
(setq kept nil)
     @result{} nil
@end group
@group
(set-process-filter (get-process "shell") 'keep-output)
     @result{} keep-output
@end group
@group
(process-send-string "shell" "ls ~/other\n")
     @result{} nil
kept
     @result{} ("lewis@@slug[8] % "
@end group
@group
"FINAL-W87-SHORT.MSS    backup.otl              kolstad.mss~
address.txt             backup.psf              kolstad.psf
backup.bib~             david.mss               resume-Dec-86.mss~
backup.err              david.psf               resume-Dec.psf
backup.mss              dland                   syllabus.mss
"
"#backups.mss#          backup.mss~             kolstad.mss
")
@end group
@end smallexample

@ignore   @c The code in this example doesn't show the right way to do things.
Here is another, more realistic example, which demonstrates how to use
the process mark to do insertion in the same fashion as is done when
there is no filter function:

@smallexample
@group
;; @r{Insert input in the buffer specified by @code{my-shell-buffer}}
;;   @r{and make sure that buffer is shown in some window.}
(defun my-process-filter (proc str)
  (let ((cur (selected-window))
        (pop-up-windows t))
    (pop-to-buffer my-shell-buffer)
@end group
@group
    (goto-char (point-max))
    (insert str)
    (set-marker (process-mark proc) (point-max))
    (select-window cur)))
@end group
@end smallexample
@end ignore

@node Decoding Output
@subsection Decoding Process Output

  When Emacs writes process output directly into a multibyte buffer,
it decodes the output according to the process output coding system.
If the coding system is @code{raw-text} or @code{no-conversion}, Emacs
converts the unibyte output to multibyte using
@code{string-to-multibyte}, and inserts the resulting multibyte text.

  You can use @code{set-process-coding-system} to specify which coding
system to use (@pxref{Process Information}).  Otherwise, the coding
system comes from @code{coding-system-for-read}, if that is
non-@code{nil}; or else from the defaulting mechanism (@pxref{Default
Coding Systems}).

  @strong{Warning:} Coding systems such as @code{undecided} which
determine the coding system from the data do not work entirely
reliably with asynchronous subprocess output.  This is because Emacs
has to process asynchronous subprocess output in batches, as it
arrives.  Emacs must try to detect the proper coding system from one
batch at a time, and this does not always work.  Therefore, if at all
possible, specify a coding system that determines both the character
code conversion and the end of line conversion---that is, one like
@code{latin-1-unix}, rather than @code{undecided} or @code{latin-1}.

@cindex filter multibyte flag, of process
@cindex process filter multibyte flag
  When Emacs calls a process filter function, it provides the process
output as a multibyte string or as a unibyte string according to the
process's filter multibyte flag.  If the flag is non-@code{nil}, Emacs
decodes the output according to the process output coding system to
produce a multibyte string, and passes that to the process.  If the
flag is @code{nil}, Emacs puts the output into a unibyte string, with
no decoding, and passes that.

  When you create a process, the filter multibyte flag takes its
initial value from @code{default-enable-multibyte-characters}.  If you
want to change the flag later on, use
@code{set-process-filter-multibyte}.

@defun set-process-filter-multibyte process multibyte
This function sets the filter multibyte flag of @var{process}
to @var{multibyte}.
@end defun

@defun process-filter-multibyte-p process
This function returns the filter multibyte flag of @var{process}.
@end defun

@node Accepting Output
@subsection Accepting Output from Processes

  Output from asynchronous subprocesses normally arrives only while
Emacs is waiting for some sort of external event, such as elapsed time
or terminal input.  Occasionally it is useful in a Lisp program to
explicitly permit output to arrive at a specific point, or even to wait
until output arrives from a process.

@defun accept-process-output &optional process seconds millisec just-this-one
This function allows Emacs to read pending output from processes.  The
output is inserted in the associated buffers or given to their filter
functions.  If @var{process} is non-@code{nil} then this function does
not return until some output has been received from @var{process}.

@c Emacs 19 feature
The arguments @var{seconds} and @var{millisec} let you specify timeout
periods.  The former specifies a period measured in seconds and the
latter specifies one measured in milliseconds.  The two time periods
thus specified are added together, and @code{accept-process-output}
returns after that much time whether or not there has been any
subprocess output.

The argument @var{seconds} need not be an integer.  If it is a floating
point number, this function waits for a fractional number of seconds.
Some systems support only a whole number of seconds; on these systems,
@var{seconds} is rounded down.

Not all operating systems support waiting periods other than multiples
of a second; on those that do not, you get an error if you specify
nonzero @var{millisec}.

@c Emacs 22.1 feature
If @var{process} is a process, and the argument @var{just-this-one} is
non-@code{nil}, only output from that process is handled, suspending output
from other processes until some output has been received from that
process or the timeout expires.  If @var{just-this-one} is an integer,
also inhibit running timers.  This feature is generally not
recommended, but may be necessary for specific applications, such as
speech synthesis.

The function @code{accept-process-output} returns non-@code{nil} if it
did get some output, or @code{nil} if the timeout expired before output
arrived.
@end defun

@node Sentinels
@section Sentinels: Detecting Process Status Changes
@cindex process sentinel
@cindex sentinel

  A @dfn{process sentinel} is a function that is called whenever the
associated process changes status for any reason, including signals
(whether sent by Emacs or caused by the process's own actions) that
terminate, stop, or continue the process.  The process sentinel is
also called if the process exits.  The sentinel receives two
arguments: the process for which the event occurred, and a string
describing the type of event.

  The string describing the event looks like one of the following:

@itemize @bullet
@item
@code{"finished\n"}.

@item
@code{"exited abnormally with code @var{exitcode}\n"}.

@item
@code{"@var{name-of-signal}\n"}.

@item
@code{"@var{name-of-signal} (core dumped)\n"}.
@end itemize

  A sentinel runs only while Emacs is waiting (e.g., for terminal
input, or for time to elapse, or for process output).  This avoids the
timing errors that could result from running them at random places in
the middle of other Lisp programs.  A program can wait, so that
sentinels will run, by calling @code{sit-for} or @code{sleep-for}
(@pxref{Waiting}), or @code{accept-process-output} (@pxref{Accepting
Output}).  Emacs also allows sentinels to run when the command loop is
reading input.  @code{delete-process} calls the sentinel when it
terminates a running process.

  Emacs does not keep a queue of multiple reasons to call the sentinel
of one process; it records just the current status and the fact that
there has been a change.  Therefore two changes in status, coming in
quick succession, can call the sentinel just once.  However, process
termination will always run the sentinel exactly once.  This is
because the process status can't change again after termination.

  Emacs explicitly checks for output from the process before running
the process sentinel.  Once the sentinel runs due to process
termination, no further output can arrive from the process.

  A sentinel that writes the output into the buffer of the process
should check whether the buffer is still alive.  If it tries to insert
into a dead buffer, it will get an error.  If the buffer is dead,
@code{(buffer-name (process-buffer @var{process}))} returns @code{nil}.

  Quitting is normally inhibited within a sentinel---otherwise, the
effect of typing @kbd{C-g} at command level or to quit a user command
would be unpredictable.  If you want to permit quitting inside a
sentinel, bind @code{inhibit-quit} to @code{nil}.  In most cases, the
right way to do this is with the macro @code{with-local-quit}.
@xref{Quitting}.

  If an error happens during execution of a sentinel, it is caught
automatically, so that it doesn't stop the execution of whatever
programs was running when the sentinel was started.  However, if
@code{debug-on-error} is non-@code{nil}, the error-catching is turned
off.  This makes it possible to use the Lisp debugger to debug the
sentinel.  @xref{Debugger}.

  While a sentinel is running, the process sentinel is temporarily
set to @code{nil} so that the sentinel won't run recursively.
For this reason it is not possible for a sentinel to specify
a new sentinel.

  In earlier Emacs versions, every sentinel that did regular expression
searching or matching had to explicitly save and restore the match data.
Now Emacs does this automatically for sentinels; they never need to do
it explicitly.  @xref{Match Data}.

@defun set-process-sentinel process sentinel
This function associates @var{sentinel} with @var{process}.  If
@var{sentinel} is @code{nil}, then the process will have no sentinel.
The default behavior when there is no sentinel is to insert a message in
the process's buffer when the process status changes.

Changes in process sentinel take effect immediately---if the sentinel
is slated to be run but has not been called yet, and you specify a new
sentinel, the eventual call to the sentinel will use the new one.

@smallexample
@group
(defun msg-me (process event)
   (princ
     (format "Process: %s had the event `%s'" process event)))
(set-process-sentinel (get-process "shell") 'msg-me)
     @result{} msg-me
@end group
@group
(kill-process (get-process "shell"))
     @print{} Process: #<process shell> had the event `killed'
     @result{} #<process shell>
@end group
@end smallexample
@end defun

@defun process-sentinel process
This function returns the sentinel of @var{process}, or @code{nil} if it
has none.
@end defun

@defun waiting-for-user-input-p
While a sentinel or filter function is running, this function returns
non-@code{nil} if Emacs was waiting for keyboard input from the user at
the time the sentinel or filter function was called, @code{nil} if it
was not.
@end defun

@node Query Before Exit
@section Querying Before Exit

  When Emacs exits, it terminates all its subprocesses by sending them
the @code{SIGHUP} signal.  Because subprocesses may be doing
valuable work, Emacs normally asks the user to confirm that it is ok
to terminate them.  Each process has a query flag which, if
non-@code{nil}, says that Emacs should ask for confirmation before
exiting and thus killing that process.  The default for the query flag
is @code{t}, meaning @emph{do} query.

@tindex process-query-on-exit-flag
@defun process-query-on-exit-flag process
This returns the query flag of @var{process}.
@end defun

@tindex set-process-query-on-exit-flag
@defun set-process-query-on-exit-flag process flag
This function sets the query flag of @var{process} to @var{flag}.  It
returns @var{flag}.

@smallexample
@group
;; @r{Don't query about the shell process}
(set-process-query-on-exit-flag (get-process "shell") nil)
     @result{} t
@end group
@end smallexample
@end defun

@defun process-kill-without-query process &optional do-query
This function clears the query flag of @var{process}, so that
Emacs will not query the user on account of that process.

Actually, the function does more than that: it returns the old value of
the process's query flag, and sets the query flag to @var{do-query}.
Please don't use this function to do those things any more---please
use the newer, cleaner functions @code{process-query-on-exit-flag} and
@code{set-process-query-on-exit-flag} in all but the simplest cases.
The only way you should use @code{process-kill-without-query} nowadays
is like this:

@smallexample
@group
;; @r{Don't query about the shell process}
(process-kill-without-query (get-process "shell"))
@end group
@end smallexample
@end defun

@node Transaction Queues
@section Transaction Queues
@cindex transaction queue

You can use a @dfn{transaction queue} to communicate with a subprocess
using transactions.  First use @code{tq-create} to create a transaction
queue communicating with a specified process.  Then you can call
@code{tq-enqueue} to send a transaction.

@defun tq-create process
This function creates and returns a transaction queue communicating with
@var{process}.  The argument @var{process} should be a subprocess
capable of sending and receiving streams of bytes.  It may be a child
process, or it may be a TCP connection to a server, possibly on another
machine.
@end defun

@defun tq-enqueue queue question regexp closure fn
This function sends a transaction to queue @var{queue}.  Specifying the
queue has the effect of specifying the subprocess to talk to.

The argument @var{question} is the outgoing message that starts the
transaction.  The argument @var{fn} is the function to call when the
corresponding answer comes back; it is called with two arguments:
@var{closure}, and the answer received.

The argument @var{regexp} is a regular expression that should match
text at the end of the entire answer, but nothing before; that's how
@code{tq-enqueue} determines where the answer ends.

The return value of @code{tq-enqueue} itself is not meaningful.
@end defun

@defun tq-close queue
Shut down transaction queue @var{queue}, waiting for all pending transactions
to complete, and then terminate the connection or child process.
@end defun

Transaction queues are implemented by means of a filter function.
@xref{Filter Functions}.

@node Network
@section Network Connections
@cindex network connection
@cindex TCP
@cindex UDP

  Emacs Lisp programs can open stream (TCP) and datagram (UDP) network
connections to other processes on the same machine or other machines.
A network connection is handled by Lisp much like a subprocess, and is
represented by a process object.  However, the process you are
communicating with is not a child of the Emacs process, so it has no
process @acronym{ID}, and you can't kill it or send it signals.  All you
can do is send and receive data.  @code{delete-process} closes the
connection, but does not kill the program at the other end; that
program must decide what to do about closure of the connection.

  Lisp programs can listen for connections by creating network
servers.  A network server is also represented by a kind of process
object, but unlike a network connection, the network server never
transfers data itself.  When it receives a connection request, it
creates a new network connection to represent the connection just
made.  (The network connection inherits certain information, including
the process plist, from the server.)  The network server then goes
back to listening for more connection requests.

  Network connections and servers are created by calling
@code{make-network-process} with an argument list consisting of
keyword/argument pairs, for example @code{:server t} to create a
server process, or @code{:type 'datagram} to create a datagram
connection.  @xref{Low-Level Network}, for details.  You can also use
the @code{open-network-stream} function described below.

  You can distinguish process objects representing network connections
and servers from those representing subprocesses with the
@code{process-status} function.  The possible status values for
network connections are @code{open}, @code{closed}, @code{connect},
and @code{failed}.  For a network server, the status is always
@code{listen}.  None of those values is possible for a real
subprocess.  @xref{Process Information}.

  You can stop and resume operation of a network process by calling
@code{stop-process} and @code{continue-process}.  For a server
process, being stopped means not accepting new connections.  (Up to 5
connection requests will be queued for when you resume the server; you
can increase this limit, unless it is imposed by the operating
system.)  For a network stream connection, being stopped means not
processing input (any arriving input waits until you resume the
connection).  For a datagram connection, some number of packets may be
queued but input may be lost.  You can use the function
@code{process-command} to determine whether a network connection or
server is stopped; a non-@code{nil} value means yes.

@defun open-network-stream name buffer-or-name host service
This function opens a TCP connection, and returns a process object
that represents the connection.

The @var{name} argument specifies the name for the process object.  It
is modified as necessary to make it unique.

The @var{buffer-or-name} argument is the buffer to associate with the
connection.  Output from the connection is inserted in the buffer,
unless you specify a filter function to handle the output.  If
@var{buffer-or-name} is @code{nil}, it means that the connection is not
associated with any buffer.

The arguments @var{host} and @var{service} specify where to connect to;
@var{host} is the host name (a string), and @var{service} is the name of
a defined network service (a string) or a port number (an integer).
@end defun

@defun process-contact process &optional key
This function returns information about how a network process was set
up.  For a connection, when @var{key} is @code{nil}, it returns
@code{(@var{hostname} @var{service})} which specifies what you
connected to.

If @var{key} is @code{t}, the value is the complete status information
for the connection or server; that is, the list of keywords and values
specified in @code{make-network-process}, except that some of the
values represent the current status instead of what you specified:

@table @code
@item :buffer
The associated value is the process buffer.
@item :filter
The associated value is the process filter function.
@item :sentinel
The associated value is the process sentinel function.
@item :remote
In a connection, this is the address in internal format of the remote peer.
@item :local
The local address, in internal format.
@item :service
In a server, if you specified @code{t} for @var{service},
this value is the actual port number.
@end table

@code{:local} and @code{:remote} are included even if they were not
specified explicitly in @code{make-network-process}.

If @var{key} is a keyword, the function returns the value corresponding
to that keyword.

For an ordinary child process, this function always returns @code{t}.
@end defun

@node Network Servers
@section Network Servers

  You create a server by calling @code{make-network-process} with
@code{:server t}.  The server will listen for connection requests from
clients.  When it accepts a client connection request, that creates a
new network connection, itself a process object, with the following
parameters:

@itemize @bullet
@item
The connection's process name is constructed by concatenating the
server process' @var{name} with a client identification string.  The
client identification string for an IPv4 connection looks like
@samp{<@var{a}.@var{b}.@var{c}.@var{d}:@var{p}>}.  Otherwise, it is a
unique number in brackets, as in @samp{<@var{nnn}>}.  The number
is unique for each connection in the Emacs session.

@item
If the server's filter is non-@code{nil}, the connection process does
not get a separate process buffer; otherwise, Emacs creates a new
buffer for the purpose.  The buffer name is the server's buffer name
or process name, concatenated with the client identification string.

The server's process buffer value is never used directly by Emacs, but
it is passed to the log function, which can log connections by
inserting text there.

@item
The communication type and the process filter and sentinel are
inherited from those of the server.  The server never directly
uses its filter and sentinel; their sole purpose is to initialize
connections made to the server.

@item
The connection's process contact info is set according to the client's
addressing information (typically an IP address and a port number).
This information is associated with the @code{process-contact}
keywords @code{:host}, @code{:service}, @code{:remote}.

@item
The connection's local address is set up according to the port
number used for the connection.

@item
The client process' plist is initialized from the server's plist.
@end itemize

@node Datagrams
@section Datagrams
@cindex datagrams

  A datagram connection communicates with individual packets rather
than streams of data.  Each call to @code{process-send} sends one
datagram packet (@pxref{Input to Processes}), and each datagram
received results in one call to the filter function.

  The datagram connection doesn't have to talk with the same remote
peer all the time.  It has a @dfn{remote peer address} which specifies
where to send datagrams to.  Each time an incoming datagram is passed
to the filter function, the peer address is set to the address that
datagram came from; that way, if the filter function sends a datagram,
it will go back to that place.  You can specify the remote peer
address when you create the datagram connection using the
@code{:remote} keyword.  You can change it later on by calling
@code{set-process-datagram-address}.

@defun process-datagram-address process
If @var{process} is a datagram connection or server, this function
returns its remote peer address.
@end defun

@defun set-process-datagram-address process address
If @var{process} is a datagram connection or server, this function
sets its remote peer address to @var{address}.
@end defun

@node Low-Level Network
@section Low-Level Network Access

  You can also create network connections by operating at a lower
level that that of @code{open-network-stream}, using
@code{make-network-process}.

@menu
* Make Network::             Using @code{make-network-process}.
* Network Options::          Further control over network connections.
* Network Feature Testing::  Determining which network features work on
                               the machine you are using.
@end menu

@node Make Network
@subsection @code{make-network-process}

   The basic function for creating network connections and network
servers is @code{make-network-process}.  It can do either of those
jobs, depending on the arguments you give it.

@defun make-network-process &rest args
This function creates a network connection or server and returns the
process object that represents it.  The arguments @var{args} are a
list of keyword/argument pairs.  Omitting a keyword is always
equivalent to specifying it with value @code{nil}, except for
@code{:coding}, @code{:filter-multibyte}, and @code{:reuseaddr}.  Here
are the meaningful keywords:

@table @asis
@item :name @var{name}
Use the string @var{name} as the process name.  It is modified if
necessary to make it unique.

@item :type @var{type}
Specify the communication type.  A value of @code{nil} specifies a
stream connection (the default); @code{datagram} specifies a datagram
connection.  Both connections and servers can be of either type.

@item :server @var{server-flag}
If @var{server-flag} is non-@code{nil}, create a server.  Otherwise,
create a connection.  For a stream type server, @var{server-flag} may
be an integer which then specifies the length of the queue of pending
connections to the server.  The default queue length is 5.

@item :host @var{host}
Specify the host to connect to.  @var{host} should be a host name or
internet address, as a string, or the symbol @code{local} to specify
the local host.  If you specify @var{host} for a server, it must
specify a valid address for the local host, and only clients
connecting to that address will be accepted.

@item :service @var{service}
@var{service} specifies a port number to connect to, or, for a server,
the port number to listen on.  It should be a service name that
translates to a port number, or an integer specifying the port number
directly.  For a server, it can also be @code{t}, which means to let
the system select an unused port number.

@item :family @var{family}
@var{family} specifies the address (and protocol) family for
communication.  @code{nil} stands for IPv4.  @code{local} specifies a
Unix socket, in which case @var{host} is ignored.

@item :local @var{local-address}
For a server process, @var{local-address} is the address to listen on.
It overrides @var{family}, @var{host} and @var{service}, and you
may as well not specify them.

@item :remote @var{remote-address}
For a connection, @var{remote-address} is the address to connect to.
It overrides @var{family}, @var{host} and @var{service}, and you
may as well not specify them.

For a datagram server, @var{remote-address} specifies the initial
setting of the remote datagram address.

The format of @var{local-address} or @var{remote-address} depends on
the address family:

@itemize -
@item
An IPv4 address is represented as a vector of integers @code{[@var{a}
@var{b} @var{c} @var{d} @var{p}]} corresponding to numeric IP address
@var{a}.@var{b}.@var{c}.@var{d} and port number @var{p}.

@item
A local address is represented as a string which specifies the address
in the local address space.

@item
An ``unsupported family'' address is represented by a cons
@code{(@var{f} . @var{av})}, where @var{f} is the family number and
@var{av} is a vector specifying the socket address using one element
per address data byte.  Do not rely on this format in portable code,
as it may depend on implementation defined constants, data sizes, and
data structure alignment.
@end itemize

@item :nowait @var{bool}
If @var{bool} is non-@code{nil} for a stream connection, return
without waiting for the connection to complete.  When the connection
succeeds or fails, Emacs will call the sentinel function, with a
second argument matching @code{"open"} (if successful) or
@code{"failed"}.  The default is to block, so that
@code{make-network-process} does not return until the connection
has succeeded or failed.

@item :stop @var{stopped}
Start the network connection or server in the `stopped' state if
@var{stopped} is non-@code{nil}.

@item :buffer @var{buffer}
Use @var{buffer} as the process buffer.

@item :coding @var{coding}
Use @var{coding} as the coding system for this process.  To specify
different coding systems for decoding data from the connection and for
encoding data sent to it, specify @code{(@var{decoding} .
@var{encoding})} for @var{coding}.

If you don't specify this keyword at all, the default
is to determine the coding systems from the data.

@item :noquery @var{query-flag}
Initialize the process query flag to @var{query-flag}.
@xref{Query Before Exit}.

@item :filter @var{filter}
Initialize the process filter to @var{filter}.

@item :filter-multibyte @var{bool}
If @var{bool} is non-@code{nil}, strings given to the process filter
are multibyte, otherwise they are unibyte.  If you don't specify this
keyword at all, the default is that the strings are multibyte if
@code{default-enable-multibyte-characters} is non-@code{nil}.

@item :sentinel @var{sentinel}
Initialize the process sentinel to @var{sentinel}.

@item :log @var{log}
Initialize the log function of a server process to @var{log}.  The log
function is called each time the server accepts a network connection
from a client.  The arguments passed to the log function are
@var{server}, @var{connection}, and @var{message}, where @var{server}
is the server process, @var{connection} is the new process for the
connection, and @var{message} is a string describing what has
happened.

@item :plist @var{plist}
Initialize the process plist to @var{plist}.
@end table

The original argument list, modified with the actual connection
information, is available via the @code{process-contact} function.
@end defun

@node Network Options
@subsection Network Options

  The following network options can be specified when you create a
network process.  Except for @code{:reuseaddr}, you can also set or
modify these options later, using @code{set-network-process-option}.

  For a server process, the options specified with
@code{make-network-process} are not inherited by the client
connections, so you will need to set the necessary options for each
child connection as it is created.

@table @asis
@item :bindtodevice @var{device-name}
If @var{device-name} is a non-empty string identifying a network
interface name (see @code{network-interface-list}), only handle
packets received on that interface.  If @var{device-name} is @code{nil}
(the default), handle packets received on any interface.

Using this option may require special privileges on some systems.

@item :broadcast @var{broadcast-flag}
If @var{broadcast-flag} is non-@code{nil} for a datagram process, the
process will receive datagram packet sent to a broadcast address, and
be able to send packets to a broadcast address.  Ignored for a stream
connection.

@item :dontroute @var{dontroute-flag}
If @var{dontroute-flag} is non-@code{nil}, the process can only send
to hosts on the same network as the local host.

@item :keepalive @var{keepalive-flag}
If @var{keepalive-flag} is non-@code{nil} for a stream connection,
enable exchange of low-level keep-alive messages.

@item :linger @var{linger-arg}
If @var{linger-arg} is non-@code{nil}, wait for successful
transmission of all queued packets on the connection before it is
deleted (see @code{delete-process}).  If @var{linger-arg} is an
integer, it specifies the maximum time in seconds to wait for queued
packets to be sent before closing the connection.  Default is
@code{nil} which means to discard unsent queued packets when the
process is deleted.

@item :oobinline @var{oobinline-flag}
If @var{oobinline-flag} is non-@code{nil} for a stream connection,
receive out-of-band data in the normal data stream.  Otherwise, ignore
out-of-band data.

@item :priority @var{priority}
Set the priority for packets sent on this connection to the integer
@var{priority}.  The interpretation of this number is protocol
specific, such as setting the TOS (type of service) field on IP
packets sent on this connection.  It may also have system dependent
effects, such as selecting a specific output queue on the network
interface.

@item :reuseaddr @var{reuseaddr-flag}
If @var{reuseaddr-flag} is non-@code{nil} (the default) for a stream
server process, allow this server to reuse a specific port number (see
@code{:service}) unless another process on this host is already
listening on that port.  If @var{reuseaddr-flag} is @code{nil}, there
may be a period of time after the last use of that port (by any
process on the host), where it is not possible to make a new server on
that port.
@end table

@defun set-network-process-option process option value
This function sets or modifies a network option for network process
@var{process}.  See @code{make-network-process} for details of options
@var{option} and their corresponding values @var{value}.

The current setting of an option is available via the
@code{process-contact} function.
@end defun

@node Network Feature Testing
@subsection Testing Availability of Network Features

  To test for the availability of a given network feature, use
@code{featurep} like this:

@example
(featurep 'make-network-process '(@var{keyword} @var{value}))
@end example

@noindent
The result of the first form is @code{t} if it works to specify
@var{keyword} with value @var{value} in @code{make-network-process}.
The result of the second form is @code{t} if @var{keyword} is
supported by @code{make-network-process}.  Here are some of the
@var{keyword}---@var{value} pairs you can test in
this way.

@table @code
@item (:nowait t)
Non-@code{nil} if non-blocking connect is supported.
@item (:type datagram)
Non-@code{nil} if datagrams are supported.
@item (:family local)
Non-@code{nil} if local (aka ``UNIX domain'') sockets are supported.
@item (:service t)
Non-@code{nil} if the system can select the port for a server.
@end table

  To test for the availability of a given network option, use
@code{featurep} like this:

@example
(featurep 'make-network-process '@var{keyword})
@end example

@noindent
Here are some of the options you can test in this way.

@table @code
@item :bindtodevice
@itemx :broadcast
@itemx :dontroute
@itemx :keepalive
@itemx :linger
@itemx :oobinline
@itemx :priority
@itemx :reuseaddr
That particular network option is supported by
@code{make-network-process} and @code{set-network-process-option}.
@end table

@node Misc Network
@section Misc Network Facilities

  These additional functions are useful for creating and operating
on network connections.

@defun network-interface-list
This function returns a list describing the network interfaces
of the machine you are using.  The value is an alist whose
elements have the form @code{(@var{name} . @var{address})}.
@var{address} has the same form as the @var{local-address}
and @var{remote-address} arguments to @code{make-network-process}.
@end defun

@defun network-interface-info ifname
This function returns information about the network interface named
@var{ifname}.  The value is a list of the form
@code{(@var{addr} @var{bcast} @var{netmask} @var{hwaddr} @var{flags})}.

@table @var
@item addr
The internet protocol address.
@item bcast
The broadcast address.
@item netmask
The network mask.
@item hwaddr
The layer 2 address (Ethernet MAC address, for instance).
@item flags
The current flags of the interface.
@end table
@end defun

@defun format-network-address address &optional omit-port
This function converts the Lisp representation of a network address to
a string.  For example, a five-element vector @code{[@var{a} @var{b}
@var{c} @var{d} @var{p}]} represents an IP address
@var{a}.@var{b}.@var{c}.@var{d} and port number @var{p}.
@code{format-network-address} converts that to the string
@code{"@var{a}.@var{b}.@var{c}.@var{d}:@var{p}"}.

If @var{omit-port} is non-@code{nil}, the value does not include
the port number.
@end defun

@node Byte Packing
@section Packing and Unpacking Byte Arrays

  This section describes how to pack and unpack arrays of bytes,
usually for binary network protocols.  These functions convert byte arrays
to alists, and vice versa.  The byte array can be represented as a
unibyte string or as a vector of integers, while the alist associates
symbols either with fixed-size objects or with recursive sub-alists.

@cindex serializing
@cindex deserializing
@cindex packing
@cindex unpacking
  Conversion from byte arrays to nested alists is also known as
@dfn{deserializing} or @dfn{unpacking}, while going in the opposite
direction is also known as @dfn{serializing} or @dfn{packing}.

@menu
* Bindat Spec::         Describing data layout.
* Bindat Functions::    Doing the unpacking and packing.
* Bindat Examples::     Samples of what bindat.el can do for you!
@end menu

@node Bindat Spec
@subsection Describing Data Layout

  To control unpacking and packing, you write a @dfn{data layout
specification}, a special nested list describing named and typed
@dfn{fields}.  This specification controls length of each field to be
processed, and how to pack or unpack it.

@cindex endianness
@cindex big endian
@cindex little endian
@cindex network byte ordering
  A field's @dfn{type} describes the size (in bytes) of the object
that the field represents and, in the case of multibyte fields, how
the bytes are ordered within the field.  The two possible orderings
are ``big endian'' (also known as ``network byte ordering'') and
``little endian''.  For instance, the number @code{#x23cd} (decimal
9165) in big endian would be the two bytes @code{#x23} @code{#xcd};
and in little endian, @code{#xcd} @code{#x23}.  Here are the possible
type values:

@table @code
@item u8
@itemx byte
Unsigned byte, with length 1.

@item u16
@itemx word
@itemx short
Unsigned integer in network byte order, with length 2.

@item u24
Unsigned integer in network byte order, with length 3.

@item u32
@itemx dword
@itemx long
Unsigned integer in network byte order, with length 4.
Note: These values may be limited by Emacs' integer implementation limits.

@item u16r
@itemx u24r
@itemx u32r
Unsigned integer in little endian order, with length 2, 3 and 4, respectively.

@item str @var{len}
String of length @var{len}.

@item strz @var{len}
Zero-terminated string of length @var{len}.

@item vec @var{len}
Vector of @var{len} bytes.

@item ip
Four-byte vector representing an Internet address.  For example:
@code{[127 0 0 1]} for localhost.

@item bits @var{len}
List of set bits in @var{len} bytes.  The bytes are taken in big
endian order and the bits are numbered starting with @code{8 *
@var{len} @minus{} 1} and ending with zero.  For example: @code{bits
2} unpacks @code{#x28} @code{#x1c} to @code{(2 3 4 11 13)} and
@code{#x1c} @code{#x28} to @code{(3 5 10 11 12)}.

@item (eval @var{form})
@var{form} is a Lisp expression evaluated at the moment the field is
unpacked or packed.  The result of the evaluation should be one of the
above-listed type specifications.
@end table

A field specification generally has the form @code{([@var{name}]
@var{handler})}.  The square braces indicate that @var{name} is
optional.  (Don't use names that are symbols meaningful as type
specifications (above) or handler specifications (below), since that
would be ambiguous.)  @var{name} can be a symbol or the expression
@code{(eval @var{form})}, in which case @var{form} should evaluate to
a symbol.

@var{handler} describes how to unpack or pack the field and can be one
of the following:

@table @code
@item @var{type}
Unpack/pack this field according to the type specification @var{type}.

@item eval @var{form}
Evaluate @var{form}, a Lisp expression, for side-effect only.  If the
field name is specified, the value is bound to that field name.
@var{form} can access and update these dynamically bound variables:

@table @code
@item raw-data
The data as a byte array.

@item pos
Current position of the unpacking or packing operation.

@item struct
Alist.

@item last
Value of the last field processed.
@end table

@item fill @var{len}
Skip @var{len} bytes.  In packing, this leaves them unchanged,
which normally means they remain zero.  In unpacking, this means
they are ignored.

@item align @var{len}
Skip to the next multiple of @var{len} bytes.

@item struct @var{spec-name}
Process @var{spec-name} as a sub-specification.  This describes a
structure nested within another structure.

@item union @var{form} (@var{tag} @var{spec})@dots{}
@c ??? I don't see how one would actually  use this.
@c ??? what kind of expression would be useful for @var{form}?
Evaluate @var{form}, a Lisp expression, find the first @var{tag}
that matches it, and process its associated data layout specification
@var{spec}.  Matching can occur in one of three ways:

@itemize
@item
If a @var{tag} has the form @code{(eval @var{expr})}, evaluate
@var{expr} with the variable @code{tag} dynamically bound to the value
of @var{form}.  A non-@code{nil} result indicates a match.

@item
@var{tag} matches if it is @code{equal} to the value of @var{form}.

@item
@var{tag} matches unconditionally if it is @code{t}.
@end itemize

@item repeat @var{count} @var{field-spec}@dots{}
@var{count} may be an integer, or a list of one element naming a
previous field.  For correct operation, each @var{field-spec} must
include a name.
@c ??? What does it MEAN?
@end table

@node Bindat Functions
@subsection Functions to Unpack and Pack Bytes

  In the following documentation, @var{spec} refers to a data layout
specification, @code{raw-data} to a byte array, and @var{struct} to an
alist representing unpacked field data.

@defun bindat-unpack spec raw-data &optional pos
This function unpacks data from the byte array @code{raw-data}
according to @var{spec}.  Normally this starts unpacking at the
beginning of the byte array, but if @var{pos} is non-@code{nil}, it
specifies a zero-based starting position to use instead.

The value is an alist or nested alist in which each element describes
one unpacked field.
@end defun

@defun bindat-get-field struct &rest name
This function selects a field's data from the nested alist
@var{struct}.  Usually @var{struct} was returned by
@code{bindat-unpack}.  If @var{name} corresponds to just one argument,
that means to extract a top-level field value.  Multiple @var{name}
arguments specify repeated lookup of sub-structures.  An integer name
acts as an array index.

For example, if @var{name} is @code{(a b 2 c)}, that means to find
field @code{c} in the third element of subfield @code{b} of field
@code{a}.  (This corresponds to @code{struct.a.b[2].c} in C.)
@end defun

@defun bindat-length spec struct
@c ??? I don't understand this at all -- rms
This function returns the length in bytes of @var{struct}, according
to @var{spec}.
@end defun

@defun bindat-pack spec struct &optional raw-data pos
This function returns a byte array packed according to @var{spec} from
the data in the alist @var{struct}.  Normally it creates and fills a
new byte array starting at the beginning.  However, if @var{raw-data}
is non-@code{nil}, it specifies a pre-allocated string or vector to
pack into.  If @var{pos} is non-@code{nil}, it specifies the starting
offset for packing into @code{raw-data}.

@c ??? Isn't this a bug?  Shouldn't it always be unibyte?
Note: The result is a multibyte string; use @code{string-make-unibyte}
on it to make it unibyte if necessary.
@end defun

@defun bindat-ip-to-string ip
Convert the Internet address vector @var{ip} to a string in the usual
dotted notation.

@example
(bindat-ip-to-string [127 0 0 1])
     @result{} "127.0.0.1"
@end example
@end defun

@node Bindat Examples
@subsection Examples of Byte Unpacking and Packing

  Here is a complete example of byte unpacking and packing:

@lisp
(defvar fcookie-index-spec
  '((:version  u32)
    (:count    u32)
    (:longest  u32)
    (:shortest u32)
    (:flags    u32)
    (:delim    u8)
    (:ignored  fill 3)
    (:offset   repeat (:count)
               (:foo u32)))
  "Description of a fortune cookie index file's contents.")

(defun fcookie (cookies &optional index)
  "Display a random fortune cookie from file COOKIES.
Optional second arg INDEX specifies the associated index
filename, which is by default constructed by appending
\".dat\" to COOKIES.  Display cookie text in possibly
new buffer \"*Fortune Cookie: BASENAME*\" where BASENAME
is COOKIES without the directory part."
  (interactive "fCookies file: ")
  (let* ((info (with-temp-buffer
                 (insert-file-contents-literally
                  (or index (concat cookies ".dat")))
                 (bindat-unpack fcookie-index-spec
                                (buffer-string))))
         (sel (random (bindat-get-field info :count)))
         (beg (cdar (bindat-get-field info :offset sel)))
         (end (or (cdar (bindat-get-field info
                                          :offset (1+ sel)))
                  (nth 7 (file-attributes cookies)))))
    (switch-to-buffer
     (get-buffer-create
      (format "*Fortune Cookie: %s*"
              (file-name-nondirectory cookies))))
    (erase-buffer)
    (insert-file-contents-literally
     cookies nil beg (- end 3))))

(defun fcookie-create-index (cookies &optional index delim)
  "Scan file COOKIES, and write out its index file.
Optional second arg INDEX specifies the index filename,
which is by default constructed by appending \".dat\" to
COOKIES.  Optional third arg DELIM specifies the unibyte
character which, when found on a line of its own in
COOKIES, indicates the border between entries."
  (interactive "fCookies file: ")
  (setq delim (or delim ?%))
  (let ((delim-line (format "\n%c\n" delim))
        (count 0)
        (max 0)
        min p q len offsets)
    (unless (= 3 (string-bytes delim-line))
      (error "Delimiter cannot be represented in one byte"))
    (with-temp-buffer
      (insert-file-contents-literally cookies)
      (while (and (setq p (point))
                  (search-forward delim-line (point-max) t)
                  (setq len (- (point) 3 p)))
        (setq count (1+ count)
              max (max max len)
              min (min (or min max) len)
              offsets (cons (1- p) offsets))))
    (with-temp-buffer
      (set-buffer-multibyte nil)
      (insert
       (string-make-unibyte
        (bindat-pack
         fcookie-index-spec
         `((:version . 2)
           (:count . ,count)
           (:longest . ,max)
           (:shortest . ,min)
           (:flags . 0)
           (:delim . ,delim)
           (:offset . ,(mapcar (lambda (o)
                                 (list (cons :foo o)))
                               (nreverse offsets)))))))
      (let ((coding-system-for-write 'raw-text-unix))
        (write-file (or index (concat cookies ".dat")))))))
@end lisp

Following is an example of defining and unpacking a complex structure.
Consider the following C structures:

@example
struct header @{
    unsigned long    dest_ip;
    unsigned long    src_ip;
    unsigned short   dest_port;
    unsigned short   src_port;
@};

struct data @{
    unsigned char    type;
    unsigned char    opcode;
    unsigned long    length;  /* In little endian order */
    unsigned char    id[8];   /* null-terminated string  */
    unsigned char    data[/* (length + 3) & ~3 */];
@};

struct packet @{
    struct header    header;
    unsigned char    items;
    unsigned char    filler[3];
    struct data      item[/* items */];

@};
@end example

The corresponding data layout specification:

@lisp
(setq header-spec
      '((dest-ip   ip)
        (src-ip    ip)
        (dest-port u16)
        (src-port  u16)))

(setq data-spec
      '((type      u8)
        (opcode    u8)
        (length    u16r) ;; little endian order
        (id        strz 8)
        (data      vec (length))
        (align     4)))

(setq packet-spec
      '((header    struct header-spec)
        (items     u8)
        (fill      3)
        (item      repeat (items)
                   (struct data-spec))))
@end lisp

A binary data representation:

@lisp
(setq binary-data
      [ 192 168 1 100 192 168 1 101 01 28 21 32 2 0 0 0
        2 3 5 0 ?A ?B ?C ?D ?E ?F 0 0 1 2 3 4 5 0 0 0
        1 4 7 0 ?B ?C ?D ?E ?F ?G 0 0 6 7 8 9 10 11 12 0 ])
@end lisp

The corresponding decoded structure:

@lisp
(setq decoded (bindat-unpack packet-spec binary-data))
     @result{}
((header
  (dest-ip   . [192 168 1 100])
  (src-ip    . [192 168 1 101])
  (dest-port . 284)
  (src-port  . 5408))
 (items . 2)
 (item ((data . [1 2 3 4 5])
        (id . "ABCDEF")
        (length . 5)
        (opcode . 3)
        (type . 2))
       ((data . [6 7 8 9 10 11 12])
        (id . "BCDEFG")
        (length . 7)
        (opcode . 4)
        (type . 1))))
@end lisp

Fetching data from this structure:

@lisp
(bindat-get-field decoded 'item 1 'id)
     @result{} "BCDEFG"
@end lisp

@ignore
   arch-tag: ba9da253-e65f-4e7f-b727-08fba0a1df7a
@end ignore