[chise/xemacs-chise.git.1] / man / internals / internals.texi

\input texinfo  @c -*-texinfo-*-
@c %**start of header
@setfilename ../../info/internals.info
@settitle XEmacs Internals Manual
@c %**end of header

@ifinfo
@dircategory XEmacs Editor
@direntry
* Internals: (internals).       XEmacs Internals Manual.
@end direntry

Copyright @copyright{} 1992 - 1996 Ben Wing.
Copyright @copyright{} 1996, 1997 Sun Microsystems.
Copyright @copyright{} 1994 - 1998 Free Software Foundation.
Copyright @copyright{} 1994, 1995 Board of Trustees, University of Illinois.


Permission is granted to make and distribute verbatim copies of this
manual provided the copyright notice and this permission notice are
preserved on all copies.

@ignore
Permission is granted to process this file through TeX and print the
results, provided the printed document carries copying permission notice
identical to this one except for the removal of this paragraph (this
paragraph not being relevant to the printed manual).

@end ignore
Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided that the
entire resulting derived work is distributed under the terms of a
permission notice identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that this permission notice may be stated in a translation
approved by the Foundation.

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that the
section entitled ``GNU General Public License'' is included exactly as
in the original, and provided that the entire resulting derived work is
distributed under the terms of a permission notice identical to this
one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that the section entitled ``GNU General Public License'' may be
included in a translation approved by the Free Software Foundation
instead of in the original English.
@end ifinfo

@c Combine indices.
@synindex cp fn
@syncodeindex vr fn
@syncodeindex ky fn
@syncodeindex pg fn
@syncodeindex tp fn

@setchapternewpage odd
@finalout

@titlepage
@title XEmacs Internals Manual
@subtitle Version 1.3, August 1999

@author Ben Wing
@author Martin Buchholz
@author Hrvoje Niksic
@author Matthias Neubauer
@page
@vskip 0pt plus 1fill

@noindent
Copyright @copyright{} 1992 - 1996 Ben Wing. @*
Copyright @copyright{} 1996, 1997 Sun Microsystems, Inc. @*
Copyright @copyright{} 1994 - 1998 Free Software Foundation. @*
Copyright @copyright{} 1994, 1995 Board of Trustees, University of Illinois.

@sp 2
Version 1.3 @*
August 1999.@*

Permission is granted to make and distribute verbatim copies of this
manual provided the copyright notice and this permission notice are
preserved on all copies.

Permission is granted to copy and distribute modified versions of this
manual under the conditions for verbatim copying, provided also that the
section entitled ``GNU General Public License'' is included
exactly as in the original, and provided that the entire resulting
derived work is distributed under the terms of a permission notice
identical to this one.

Permission is granted to copy and distribute translations of this manual
into another language, under the above conditions for modified versions,
except that the section entitled ``GNU General Public License'' may be
included in a translation approved by the Free Software Foundation
instead of in the original English.
@end titlepage
@page

@node Top, A History of Emacs, (dir), (dir)

@ifinfo
This Info file contains v1.0 of the XEmacs Internals Manual.
@end ifinfo

@menu
* A History of Emacs::          Times, dates, important events.
* XEmacs From the Outside::     A broad conceptual overview.
* The Lisp Language::           An overview.
* XEmacs From the Perspective of Building::
* XEmacs From the Inside::
* The XEmacs Object System (Abstractly Speaking)::
* How Lisp Objects Are Represented in C::
* Rules When Writing New C Code::
* A Summary of the Various XEmacs Modules::
* Allocation of Objects in XEmacs Lisp::
* Events and the Event Loop::
* Evaluation; Stack Frames; Bindings::
* Symbols and Variables::
* Buffers and Textual Representation::
* MULE Character Sets and Encodings::
* The Lisp Reader and Compiler::
* Lstreams::
* Consoles; Devices; Frames; Windows::
* The Redisplay Mechanism::
* Extents::
* Faces::
* Glyphs::
* Specifiers::
* Menus::
* Subprocesses::
* Interface to X Windows::
* Index::                   Index including concepts, functions, variables,
                              and other terms.

      --- The Detailed Node Listing ---

Here are other nodes that are inferiors of those already listed,
mentioned here so you can get to them in one step:

A History of Emacs

* Through Version 18::          Unification prevails.
* Lucid Emacs::                 One version 19 Emacs.
* GNU Emacs 19::                The other version 19 Emacs.
* XEmacs::                      The continuation of Lucid Emacs.

Rules When Writing New C Code

* General Coding Rules::
* Writing Lisp Primitives::
* Adding Global Lisp Variables::
* Techniques for XEmacs Developers::

A Summary of the Various XEmacs Modules

* Low-Level Modules::
* Basic Lisp Modules::
* Modules for Standard Editing Operations::
* Editor-Level Control Flow Modules::
* Modules for the Basic Displayable Lisp Objects::
* Modules for other Display-Related Lisp Objects::
* Modules for the Redisplay Mechanism::
* Modules for Interfacing with the File System::
* Modules for Other Aspects of the Lisp Interpreter and Object System::
* Modules for Interfacing with the Operating System::
* Modules for Interfacing with X Windows::
* Modules for Internationalization::

Allocation of Objects in XEmacs Lisp

* Introduction to Allocation::
* Garbage Collection::
* GCPROing::
* Garbage Collection - Step by Step::
* Integers and Characters::
* Allocation from Frob Blocks::
* lrecords::
* Low-level allocation::
* Pure Space::
* Cons::
* Vector::
* Bit Vector::
* Symbol::
* Marker::
* String::
* Compiled Function::

Events and the Event Loop

* Introduction to Events::
* Main Loop::
* Specifics of the Event Gathering Mechanism::
* Specifics About the Emacs Event::
* The Event Stream Callback Routines::
* Other Event Loop Functions::
* Converting Events::
* Dispatching Events; The Command Builder::

Evaluation; Stack Frames; Bindings

* Evaluation::
* Dynamic Binding; The specbinding Stack; Unwind-Protects::
* Simple Special Forms::
* Catch and Throw::

Symbols and Variables

* Introduction to Symbols::
* Obarrays::
* Symbol Values::

Buffers and Textual Representation

* Introduction to Buffers::     A buffer holds a block of text such as a file.
* The Text in a Buffer::        Representation of the text in a buffer.
* Buffer Lists::                Keeping track of all buffers.
* Markers and Extents::         Tagging locations within a buffer.
* Bufbytes and Emchars::        Representation of individual characters.
* The Buffer Object::           The Lisp object corresponding to a buffer.

MULE Character Sets and Encodings

* Character Sets::
* Encodings::
* Internal Mule Encodings::

Encodings

* Japanese EUC (Extended Unix Code)::
* JIS7::

Internal Mule Encodings

* Internal String Encoding::
* Internal Character Encoding::

The Lisp Reader and Compiler

Lstreams

Consoles; Devices; Frames; Windows

* Introduction to Consoles; Devices; Frames; Windows::
* Point::
* Window Hierarchy::

The Redisplay Mechanism

* Critical Redisplay Sections::
* Line Start Cache::

Extents

* Introduction to Extents::     Extents are ranges over text, with properties.
* Extent Ordering::             How extents are ordered internally.
* Format of the Extent Info::   The extent information in a buffer or string.
* Zero-Length Extents::         A weird special case.
* Mathematics of Extent Ordering::      A rigorous foundation.
* Extent Fragments::            Cached information useful for redisplay.

Faces

Glyphs

Specifiers

Menus

Subprocesses

Interface to X Windows

@end menu

@node A History of Emacs, XEmacs From the Outside, Top, Top
@chapter A History of Emacs
@cindex history of Emacs
@cindex Hackers (Steven Levy)
@cindex Levy, Steven
@cindex ITS (Incompatible Timesharing System)
@cindex Stallman, Richard
@cindex RMS
@cindex MIT
@cindex TECO
@cindex FSF
@cindex Free Software Foundation

  XEmacs is a powerful, customizable text editor and development
environment.  It began as Lucid Emacs, which was in turn derived from
GNU Emacs, a program written by Richard Stallman of the Free Software
Foundation.  GNU Emacs dates back to the 1970's, and was modelled
after a package called ``Emacs'', written in 1976, that was a set of
macros on top of TECO, an old, old text editor written at MIT on the
DEC PDP 10 under one of the earliest time-sharing operating systems,
ITS (Incompatible Timesharing System). (ITS dates back well before
Unix.) ITS, TECO, and Emacs were products of a group of people at MIT
who called themselves ``hackers'', who shared an idealistic belief
system about the free exchange of information and were fanatical in
their devotion to and time spent with computers. (The hacker
subculture dates back to the late 1950's at MIT and is described in
detail in Steven Levy's book @cite{Hackers}.  This book also includes
a lot of information about Stallman himself and the development of
Lisp, a programming language developed at MIT that underlies Emacs.)

@menu
* Through Version 18::          Unification prevails.
* Lucid Emacs::                 One version 19 Emacs.
* GNU Emacs 19::                The other version 19 Emacs.
* GNU Emacs 20::                The other version 20 Emacs.
* XEmacs::                      The continuation of Lucid Emacs.
@end menu

@node Through Version 18
@section Through Version 18
@cindex Gosling, James
@cindex Great Usenet Renaming

  Although the history of the early versions of GNU Emacs is unclear,
the history is well-known from the middle of 1985.  A time line is:

@itemize @bullet
@item
GNU Emacs version 15 (15.34) was released sometime in 1984 or 1985 and
shared some code with a version of Emacs written by James Gosling (the
same James Gosling who later created the Java language).
@item
GNU Emacs version 16 (first released version was 16.56) was released on
July 15, 1985.  All Gosling code was removed due to potential copyright
problems with the code.
@item
version 16.57: released on September 16, 1985.
@item
versions 16.58, 16.59: released on September 17, 1985.
@item
version 16.60: released on September 19, 1985.  These later version 16's
incorporated patches from the net, esp. for getting Emacs to work under
System V.
@item
version 17.36 (first official v17 release) released on December 20,
1985.  Included a TeX-able user manual.  First official unpatched
version that worked on vanilla System V machines.
@item
version 17.43 (second official v17 release) released on January 25,
1986.
@item
version 17.45 released on January 30, 1986.
@item
version 17.46 released on February 4, 1986.
@item
version 17.48 released on February 10, 1986.
@item
version 17.49 released on February 12, 1986.
@item
version 17.55 released on March 18, 1986.
@item
version 17.57 released on March 27, 1986.
@item
version 17.58 released on April 4, 1986.
@item
version 17.61 released on April 12, 1986.
@item
version 17.63 released on May 7, 1986.
@item
version 17.64 released on May 12, 1986.
@item
version 18.24 (a beta version) released on October 2, 1986.
@item
version 18.30 (a beta version) released on November 15, 1986.
@item
version 18.31 (a beta version) released on November 23, 1986.
@item
version 18.32 (a beta version) released on December 7, 1986.
@item
version 18.33 (a beta version) released on December 12, 1986.
@item
version 18.35 (a beta version) released on January 5, 1987.
@item
version 18.36 (a beta version) released on January 21, 1987.
@item
January 27, 1987: The Great Usenet Renaming.  net.emacs is now
comp.emacs.
@item
version 18.37 (a beta version) released on February 12, 1987.
@item
version 18.38 (a beta version) released on March 3, 1987.
@item
version 18.39 (a beta version) released on March 14, 1987.
@item
version 18.40 (a beta version) released on March 18, 1987.
@item
version 18.41 (the first ``official'' release) released on March 22,
1987.
@item
version 18.45 released on June 2, 1987.
@item
version 18.46 released on June 9, 1987.
@item
version 18.47 released on June 18, 1987.
@item
version 18.48 released on September 3, 1987.
@item
version 18.49 released on September 18, 1987.
@item
version 18.50 released on February 13, 1988.
@item
version 18.51 released on May 7, 1988.
@item
version 18.52 released on September 1, 1988.
@item
version 18.53 released on February 24, 1989.
@item
version 18.54 released on April 26, 1989.
@item
version 18.55 released on August 23, 1989.  This is the earliest version
that is still available by FTP.
@item
version 18.56 released on January 17, 1991.
@item
version 18.57 released late January, 1991.
@item
version 18.58 released ?????.
@item
version 18.59 released October 31, 1992.
@end itemize

@node Lucid Emacs
@section Lucid Emacs
@cindex Lucid Emacs
@cindex Lucid Inc.
@cindex Energize
@cindex Epoch

  Lucid Emacs was developed by the (now-defunct) Lucid Inc., a maker of
C++ and Lisp development environments.  It began when Lucid decided they
wanted to use Emacs as the editor and cornerstone of their C++
development environment (called ``Energize'').  They needed many features
that were not available in the existing version of GNU Emacs (version
18.5something), in particular good and integrated support for GUI
elements such as mouse support, multiple fonts, multiple window-system
windows, etc.  A branch of GNU Emacs called Epoch, written at the
University of Illinois, existed that supplied many of these features;
however, Lucid needed more than what existed in Epoch.  At the time, the
Free Software Foundation was working on version 19 of Emacs (this was
sometime around 1991), which was planned to have similar features, and
so Lucid decided to work with the Free Software Foundation.  Their plan
was to add features that they needed, and coordinate with the FSF so
that the features would get included back into Emacs version 19.

  Delays in the release of version 19 occurred, however (resulting in it
finally being released more than a year after what was initially
planned), and Lucid encountered unexpected technical resistance in
getting their changes merged back into version 19, so they decided to
release their own version of Emacs, which became Lucid Emacs 19.0.

@cindex Zawinski, Jamie
@cindex Sexton, Harlan
@cindex Benson, Eric
@cindex Devin, Matthieu
  The initial authors of Lucid Emacs were Matthieu Devin, Harlan Sexton,
and Eric Benson, and the work was later taken over by Jamie Zawinski,
who became ``Mr. Lucid Emacs'' for many releases.

  A time line for Lucid Emacs/XEmacs is

@itemize @bullet
@item
version 19.0 shipped with Energize 1.0, April 1992.
@item
version 19.1 released June 4, 1992.
@item
version 19.2 released June 19, 1992.
@item
version 19.3 released September 9, 1992.
@item
version 19.4 released January 21, 1993.
@item
version 19.5 was a repackaging of 19.4 with a few bug fixes and
shipped with Energize 2.0.  Never released to the net.
@item
version 19.6 released April 9, 1993.
@item
version 19.7 was a repackaging of 19.6 with a few bug fixes and
shipped with Energize 2.1.  Never released to the net.
@item
version 19.8 released September 6, 1993.
@item
version 19.9 released January 12, 1994.
@item
version 19.10 released May 27, 1994.
@item
version 19.11 (first XEmacs) released September 13, 1994.
@item
version 19.12 released June 23, 1995.
@item
version 19.13 released September 1, 1995.
@item
version 19.14 released June 23, 1996.
@item
version 20.0 released February 9, 1997.
@item
version 19.15 released March 28, 1997.
@item
version 20.1 (not released to the net) April 15, 1997.
@item
version 20.2 released May 16, 1997.
@item
version 19.16 released October 31, 1997.
@item
version 20.3 (the first stable version of XEmacs 20.x) released November 30,
1997.
version 20.4 released February 28, 1998.
@end itemize

@node GNU Emacs 19
@section GNU Emacs 19
@cindex GNU Emacs 19
@cindex FSF Emacs

  About a year after the initial release of Lucid Emacs, the FSF
released a beta of their version of Emacs 19 (referred to here as ``GNU
Emacs'').  By this time, the current version of Lucid Emacs was
19.6. (Strangely, the first released beta from the FSF was GNU Emacs
19.7.) A time line for GNU Emacs version 19 is

@itemize @bullet
@item
version 19.8 (beta) released May 27, 1993.
@item
version 19.9 (beta) released May 27, 1993.
@item
version 19.10 (beta) released May 30, 1993.
@item
version 19.11 (beta) released June 1, 1993.
@item
version 19.12 (beta) released June 2, 1993.
@item
version 19.13 (beta) released June 8, 1993.
@item
version 19.14 (beta) released June 17, 1993.
@item
version 19.15 (beta) released June 19, 1993.
@item
version 19.16 (beta) released July 6, 1993.
@item
version 19.17 (beta) released late July, 1993.
@item
version 19.18 (beta) released August 9, 1993.
@item
version 19.19 (beta) released August 15, 1993.
@item
version 19.20 (beta) released November 17, 1993.
@item
version 19.21 (beta) released November 17, 1993.
@item
version 19.22 (beta) released November 28, 1993.
@item
version 19.23 (beta) released May 17, 1994.
@item
version 19.24 (beta) released May 16, 1994.
@item
version 19.25 (beta) released June 3, 1994.
@item
version 19.26 (beta) released September 11, 1994.
@item
version 19.27 (beta) released September 14, 1994.
@item
version 19.28 (first ``official'' release) released November 1, 1994.
@item
version 19.29 released June 21, 1995.
@item
version 19.30 released November 24, 1995.
@item
version 19.31 released May 25, 1996.
@item
version 19.32 released July 31, 1996.
@item
version 19.33 released August 11, 1996.
@item
version 19.34 released August 21, 1996.
@item
version 19.34b released September 6, 1996.
@end itemize

@cindex Mlynarik, Richard
  In some ways, GNU Emacs 19 was better than Lucid Emacs; in some ways,
worse.  Lucid soon began incorporating features from GNU Emacs 19 into
Lucid Emacs; the work was mostly done by Richard Mlynarik, who had been
working on and using GNU Emacs for a long time (back as far as version
16 or 17).

@node GNU Emacs 20
@section GNU Emacs 20
@cindex GNU Emacs 20
@cindex FSF Emacs

On February 2, 1997 work began on GNU Emacs to integrate Mule.  The first
release was made in September of that year.

A timeline for Emacs 20 is

@itemize @bullet
@item
version 20.1 released September 17, 1997.
@item
version 20.2 released September 20, 1997.
@item
version 20.3 released August 19, 1998.
@end itemize

@node XEmacs
@section XEmacs
@cindex XEmacs

@cindex Sun Microsystems
@cindex University of Illinois
@cindex Illinois, University of
@cindex SPARCWorks
@cindex Andreessen, Marc
@cindex Baur, Steve
@cindex Buchholz, Martin
@cindex Kaplan, Simon
@cindex Wing, Ben
@cindex Thompson, Chuck
@cindex Win-Emacs
@cindex Epoch
@cindex Amdahl Corporation
  Around the time that Lucid was developing Energize, Sun Microsystems
was developing their own development environment (called ``SPARCWorks'')
and also decided to use Emacs.  They joined forces with the Epoch team
at the University of Illinois and later with Lucid.  The maintainer of
the last-released version of Epoch was Marc Andreessen, but he dropped
out and the Epoch project, headed by Simon Kaplan, lured Chuck Thompson
away from a system administration job to become the primary Lucid Emacs
author for Epoch and Sun.  Chuck's area of specialty became the
redisplay engine (he replaced the old Lucid Emacs redisplay engine with
a ported version from Epoch and then later rewrote it from scratch).
Sun also hired Ben Wing (the author of Win-Emacs, a port of Lucid Emacs
to Microsoft Windows 3.1) in 1993, for what was initially a one-month
contract to fix some event problems but later became a many-year
involvement, punctuated by a six-month contract with Amdahl Corporation.

@cindex rename to XEmacs
  In 1994, Sun and Lucid agreed to rename Lucid Emacs to XEmacs (a name
not favorable to either company); the first release called XEmacs was
version 19.11.  In June 1994, Lucid folded and Jamie quit to work for
the newly formed Mosaic Communications Corp., later Netscape
Communications Corp. (co-founded by the same Marc Andreessen, who had
quit his Epoch job to work on a graphical browser for the World Wide
Web).  Chuck then become the primary maintainer of XEmacs, and put out
versions 19.11 through 19.14 in conjunction with Ben.  For 19.12 and
19.13, Chuck added the new redisplay and many other display improvements
and Ben added MULE support (support for Asian and other languages) and
redesigned most of the internal Lisp subsystems to better support the
MULE work and the various other features being added to XEmacs.  After
19.14 Chuck retired as primary maintainer and Steve Baur stepped in.

@cindex MULE merged XEmacs appears
  Soon after 19.13 was released, work began in earnest on the MULE
internationalization code and the source tree was divided into two
development paths.  The MULE version was initially called 19.20, but was
soon renamed to 20.0.  In 1996 Martin Buchholz of Sun Microsystems took
over the care and feeding of it and worked on it in parallel with the
19.14 development that was occurring at the same time.  After much work
by Martin, it was decided to release 20.0 ahead of 19.15 in February
1997.  The source tree remained divided until 20.2 when the version 19
source was finally retired at version 19.16.

@cindex Baur, Steve
@cindex Buchholz, Martin
@cindex Jones, Kyle
@cindex Niksic, Hrvoje
@cindex XEmacs goes it alone
  In 1997, Sun finally dropped all pretense of support for XEmacs and
Martin Buchholz left the company in November.  Since then, and mostly
for the previous year, because Steve Baur was never paid to work on
XEmacs, XEmacs has existed solely on the contributions of volunteers
from the Free Software Community.  Starting from 1997, Hrvoje Niksic and
Kyle Jones have figured prominently in XEmacs development.

@cindex merging attempts
  Many attempts have been made to merge XEmacs and GNU Emacs, but they
have consistently failed.

  A more detailed history is contained in the XEmacs About page.

@node XEmacs From the Outside, The Lisp Language, A History of Emacs, Top
@chapter XEmacs From the Outside
@cindex read-eval-print

  XEmacs appears to the outside world as an editor, but it is really a
Lisp environment.  At its heart is a Lisp interpreter; it also
``happens'' to contain many specialized object types (e.g. buffers,
windows, frames, events) that are useful for implementing an editor.
Some of these objects (in particular windows and frames) have
displayable representations, and XEmacs provides a function
@code{redisplay()} that ensures that the display of all such objects
matches their internal state.  Most of the time, a standard Lisp
environment is in a @dfn{read-eval-print} loop---i.e. ``read some Lisp
code, execute it, and print the results''.  XEmacs has a similar loop:

@itemize @bullet
@item
read an event
@item
dispatch the event (i.e. ``do it'')
@item
redisplay
@end itemize

  Reading an event is done using the Lisp function @code{next-event},
which waits for something to happen (typically, the user presses a key
or moves the mouse) and returns an event object describing this.
Dispatching an event is done using the Lisp function
@code{dispatch-event}, which looks up the event in a keymap object (a
particular kind of object that associates an event with a Lisp function)
and calls that function.  The function ``does'' what the user has
requested by changing the state of particular frame objects, buffer
objects, etc.  Finally, @code{redisplay()} is called, which updates the
display to reflect those changes just made.  Thus is an ``editor'' born.

@cindex bridge, playing
@cindex taxes, doing
@cindex pi, calculating
  Note that you do not have to use XEmacs as an editor; you could just
as well make it do your taxes, compute pi, play bridge, etc.  You'd just
have to write functions to do those operations in Lisp.

@node The Lisp Language, XEmacs From the Perspective of Building, XEmacs From the Outside, Top
@chapter The Lisp Language
@cindex Lisp vs. C
@cindex C vs. Lisp
@cindex Lisp vs. Java
@cindex Java vs. Lisp
@cindex dynamic scoping
@cindex scoping, dynamic
@cindex dynamic types
@cindex types, dynamic
@cindex Java
@cindex Common Lisp
@cindex Gosling, James

  Lisp is a general-purpose language that is higher-level than C and in
many ways more powerful than C.  Powerful dialects of Lisp such as
Common Lisp are probably much better languages for writing very large
applications than is C. (Unfortunately, for many non-technical
reasons C and its successor C++ have become the dominant languages for
application development.  These languages are both inadequate for
extremely large applications, which is evidenced by the fact that newer,
larger programs are becoming ever harder to write and are requiring ever
more programmers despite great increases in C development environments;
and by the fact that, although hardware speeds and reliability have been
growing at an exponential rate, most software is still generally
considered to be slow and buggy.)

  The new Java language holds promise as a better general-purpose
development language than C.  Java has many features in common with
Lisp that are not shared by C (this is not a coincidence, since
Java was designed by James Gosling, a former Lisp hacker).  This
will be discussed more later.

For those used to C, here is a summary of the basic differences between
C and Lisp:

@enumerate
@item
Lisp has an extremely regular syntax.  Every function, expression,
and control statement is written in the form

@example
   (@var{func} @var{arg1} @var{arg2} ...)
@end example

This is as opposed to C, which writes functions as

@example
   func(@var{arg1}, @var{arg2}, ...)
@end example

but writes expressions involving operators as (e.g.)

@example
   @var{arg1} + @var{arg2}
@end example

and writes control statements as (e.g.)

@example
   while (@var{expr}) @{ @var{statement1}; @var{statement2}; ... @}
@end example

Lisp equivalents of the latter two would be

@example
   (+ @var{arg1} @var{arg2} ...)
@end example

and

@example
   (while @var{expr} @var{statement1} @var{statement2} ...)
@end example

@item
Lisp is a safe language.  Assuming there are no bugs in the Lisp
interpreter/compiler, it is impossible to write a program that ``core
dumps'' or otherwise causes the machine to execute an illegal
instruction.  This is very different from C, where perhaps the most
common outcome of a bug is exactly such a crash.  A corollary of this is that
the C operation of casting a pointer is impossible (and unnecessary) in
Lisp, and that it is impossible to access memory outside the bounds of
an array.

@item
Programs and data are written in the same form.  The
parenthesis-enclosing form described above for statements is the same
form used for the most common data type in Lisp, the list.  Thus, it is
possible to represent any Lisp program using Lisp data types, and for
one program to construct Lisp statements and then dynamically
@dfn{evaluate} them, or cause them to execute.

@item
All objects are @dfn{dynamically typed}.  This means that part of every
object is an indication of what type it is.  A Lisp program can
manipulate an object without knowing what type it is, and can query an
object to determine its type.  This means that, correspondingly,
variables and function parameters can hold objects of any type and are
not normally declared as being of any particular type.  This is opposed
to the @dfn{static typing} of C, where variables can hold exactly one
type of object and must be declared as such, and objects do not contain
an indication of their type because it's implicit in the variables they
are stored in.  It is possible in C to have a variable hold different
types of objects (e.g. through the use of @code{void *} pointers or
variable-argument functions), but the type information must then be
passed explicitly in some other fashion, leading to additional program
complexity.

@item
Allocated memory is automatically reclaimed when it is no longer in use.
This operation is called @dfn{garbage collection} and involves looking
through all variables to see what memory is being pointed to, and
reclaiming any memory that is not pointed to and is thus
``inaccessible'' and out of use.  This is as opposed to C, in which
allocated memory must be explicitly reclaimed using @code{free()}.  If
you simply drop all pointers to memory without freeing it, it becomes
``leaked'' memory that still takes up space.  Over a long period of
time, this can cause your program to grow and grow until it runs out of
memory.

@item
Lisp has built-in facilities for handling errors and exceptions.  In C,
when an error occurs, usually either the program exits entirely or the
routine in which the error occurs returns a value indicating this.  If
an error occurs in a deeply-nested routine, then every routine currently
called must unwind itself normally and return an error value back up to
the next routine.  This means that every routine must explicitly check
for an error in all the routines it calls; if it does not do so,
unexpected and often random behavior results.  This is an extremely
common source of bugs in C programs.  An alternative would be to do a
non-local exit using @code{longjmp()}, but that is often very dangerous
because the routines that were exited past had no opportunity to clean
up after themselves and may leave things in an inconsistent state,
causing a crash shortly afterwards.

Lisp provides mechanisms to make such non-local exits safe.  When an
error occurs, a routine simply signals that an error of a particular
class has occurred, and a non-local exit takes place.  Any routine can
trap errors occurring in routines it calls by registering an error
handler for some or all classes of errors. (If no handler is registered,
a default handler, generally installed by the top-level event loop, is
executed; this prints out the error and continues.) Routines can also
specify cleanup code (called an @dfn{unwind-protect}) that will be
called when control exits from a block of code, no matter how that exit
occurs---i.e. even if a function deeply nested below it causes a
non-local exit back to the top level.

Note that this facility has appeared in some recent vintages of C, in
particular Visual C++ and other PC compilers written for the Microsoft
Win32 API.

@item
In Emacs Lisp, local variables are @dfn{dynamically scoped}.  This means
that if you declare a local variable in a particular function, and then
call another function, that subfunction can ``see'' the local variable
you declared.  This is actually considered a bug in Emacs Lisp and in
all other early dialects of Lisp, and was corrected in Common Lisp. (In
Common Lisp, you can still declare dynamically scoped variables if you
want to---they are sometimes useful---but variables by default are
@dfn{lexically scoped} as in C.)
@end enumerate

For those familiar with Lisp, Emacs Lisp is modelled after MacLisp, an
early dialect of Lisp developed at MIT (no relation to the Macintosh
computer).  There is a Common Lisp compatibility package available for
Emacs that provides many of the features of Common Lisp.

The Java language is derived in many ways from C, and shares a similar
syntax, but has the following features in common with Lisp (and different
from C):

@enumerate
@item
Java is a safe language, like Lisp.
@item
Java provides garbage collection, like Lisp.
@item
Java has built-in facilities for handling errors and exceptions, like
Lisp.
@item
Java has a type system that combines the best advantages of both static
and dynamic typing.  Objects (except very simple types) are explicitly
marked with their type, as in dynamic typing; but there is a hierarchy
of types and functions are declared to accept only certain types, thus
providing the increased compile-time error-checking of static typing.
@end enumerate

The Java language also has some negative attributes:

@enumerate
@item
Java uses the edit/compile/run model of software development.  This
makes it hard to use interactively.  For example, to use Java like
@code{bc} it is necessary to write a special purpose, albeit tiny,
application.  In Emacs Lisp, a calculator comes built-in without any
effort - one can always just type an expression in the @code{*scratch*}
buffer.
@item
Java tries too hard to enforce, not merely enable, portability, making
ordinary access to standard OS facilities painful.  Java has an
@dfn{agenda}.  I think this is why @code{chdir} is not part of standard
Java, which is inexcusable.
@end enumerate

Unfortunately, there is no perfect language.  Static typing allows a
compiler to catch programmer errors and produce more efficient code, but
makes programming more tedious and less fun.  For the forseeable future,
an Ideal Editing and Programming Environment (and that is what XEmacs
aspires to) will be programmable in multiple languages: high level ones
like Lisp for user customization and prototyping, and lower level ones
for infrastructure and industrial strength applications.  If I had my
way, XEmacs would be friendly towards the Python, Scheme, C++, ML,
etc... communities.  But there are serious technical difficulties to
achieving that goal.

The word @dfn{application} in the previous paragraph was used
intentionally.  XEmacs implements an API for programs written in Lisp
that makes it a full-fledged application platform, very much like an OS
inside the real OS.

@node XEmacs From the Perspective of Building, XEmacs From the Inside, The Lisp Language, Top
@chapter XEmacs From the Perspective of Building

The heart of XEmacs is the Lisp environment, which is written in C.
This is contained in the @file{src/} subdirectory.  Underneath
@file{src/} are two subdirectories of header files: @file{s/} (header
files for particular operating systems) and @file{m/} (header files for
particular machine types).  In practice the distinction between the two
types of header files is blurred.  These header files define or undefine
certain preprocessor constants and macros to indicate particular
characteristics of the associated machine or operating system.  As part
of the configure process, one @file{s/} file and one @file{m/} file is
identified for the particular environment in which XEmacs is being
built.

XEmacs also contains a great deal of Lisp code.  This implements the
operations that make XEmacs useful as an editor as well as just a Lisp
environment, and also contains many add-on packages that allow XEmacs to
browse directories, act as a mail and Usenet news reader, compile Lisp
code, etc.  There is actually more Lisp code than C code associated with
XEmacs, but much of the Lisp code is peripheral to the actual operation
of the editor.  The Lisp code all lies in subdirectories underneath the
@file{lisp/} directory.

The @file{lwlib/} directory contains C code that implements a
generalized interface onto different X widget toolkits and also
implements some widgets of its own that behave like Motif widgets but
are faster, free, and in some cases more powerful.  The code in this
directory compiles into a library and is mostly independent from XEmacs.

The @file{etc/} directory contains various data files associated with
XEmacs.  Some of them are actually read by XEmacs at startup; others
merely contain useful information of various sorts.

The @file{lib-src/} directory contains C code for various auxiliary
programs that are used in connection with XEmacs.  Some of them are used
during the build process; others are used to perform certain functions
that cannot conveniently be placed in the XEmacs executable (e.g. the
@file{movemail} program for fetching mail out of @file{/var/spool/mail},
which must be setgid to @file{mail} on many systems; and the
@file{gnuclient} program, which allows an external script to communicate
with a running XEmacs process).

The @file{man/} directory contains the sources for the XEmacs
documentation.  It is mostly in a form called Texinfo, which can be
converted into either a printed document (by passing it through @TeX{})
or into on-line documentation called @dfn{info files}.

The @file{info/} directory contains the results of formatting the XEmacs
documentation as @dfn{info files}, for on-line use.  These files are
used when you enter the Info system using @kbd{C-h i} or through the
Help menu.

The @file{dynodump/} directory contains auxiliary code used to build
XEmacs on Solaris platforms.

The other directories contain various miscellaneous code and information
that is not normally used or needed.

The first step of building involves running the @file{configure} program
and passing it various parameters to specify any optional features you
want and compiler arguments and such, as described in the @file{INSTALL}
file.  This determines what the build environment is, chooses the
appropriate @file{s/} and @file{m/} file, and runs a series of tests to
determine many details about your environment, such as which library
functions are available and exactly how they work.  The reason for
running these tests is that it allows XEmacs to be compiled on a much
wider variety of platforms than those that the XEmacs developers happen
to be familiar with, including various sorts of hybrid platforms.  This
is especially important now that many operating systems give you a great
deal of control over exactly what features you want installed, and allow
for easy upgrading of parts of a system without upgrading the rest.  It
would be impossible to pre-determine and pre-specify the information for
all possible configurations.

In fact, the @file{s/} and @file{m/} files are basically @emph{evil},
since they contain unmaintainable platform-specific hard-coded
information.  XEmacs has been moving in the direction of having all
system-specific information be determined dynamically by
@file{configure}.  Perhaps someday we can @code{rm -rf src/s src/m}.

When configure is done running, it generates @file{Makefile}s and
@file{GNUmakefile}s and the file @file{src/config.h} (which describes
the features of your system) from template files.  You then run
@file{make}, which compiles the auxiliary code and programs in
@file{lib-src/} and @file{lwlib/} and the main XEmacs executable in
@file{src/}.  The result of compiling and linking is an executable
called @file{temacs}, which is @emph{not} the final XEmacs executable.
@file{temacs} by itself is not intended to function as an editor or even
display any windows on the screen, and if you simply run it, it will
exit immediately.  The @file{Makefile} runs @file{temacs} with certain
options that cause it to initialize itself, read in a number of basic
Lisp files, and then dump itself out into a new executable called
@file{xemacs}.  This new executable has been pre-initialized and
contains pre-digested Lisp code that is necessary for the editor to
function (this includes most basic editing functions,
e.g. @code{kill-line}, that can be defined in terms of other Lisp
primitives; some initialization code that is called when certain
objects, such as frames, are created; and all of the standard
keybindings and code for the actions they result in).  This executable,
@file{xemacs}, is the executable that you run to use the XEmacs editor.

Although @file{temacs} is not intended to be run as an editor, it can,
by using the incantation @code{temacs -batch -l loadup.el run-temacs}.
This is useful when the dumping procedure described above is broken, or
when using certain program debugging tools such as Purify.  These tools
get mighty confused by the tricks played by the XEmacs build process,
such as allocation memory in one process, and freeing it in the next.

@node XEmacs From the Inside, The XEmacs Object System (Abstractly Speaking), XEmacs From the Perspective of Building, Top
@chapter XEmacs From the Inside

Internally, XEmacs is quite complex, and can be very confusing.  To
simplify things, it can be useful to think of XEmacs as containing an
event loop that ``drives'' everything, and a number of other subsystems,
such as a Lisp engine and a redisplay mechanism.  Each of these other
subsystems exists simultaneously in XEmacs, and each has a certain
state.  The flow of control continually passes in and out of these
different subsystems in the course of normal operation of the editor.

It is important to keep in mind that, most of the time, the editor is
``driven'' by the event loop.  Except during initialization and batch
mode, all subsystems are entered directly or indirectly through the
event loop, and ultimately, control exits out of all subsystems back up
to the event loop.  This cycle of entering a subsystem, exiting back out
to the event loop, and starting another iteration of the event loop
occurs once each keystroke, mouse motion, etc.

If you're trying to understand a particular subsystem (other than the
event loop), think of it as a ``daemon'' process or ``servant'' that is
responsible for one particular aspect of a larger system, and
periodically receives commands or environment changes that cause it to
do something.  Ultimately, these commands and environment changes are
always triggered by the event loop.  For example:

@itemize @bullet
@item
The window and frame mechanism is responsible for keeping track of what
windows and frames exist, what buffers are in them, etc.  It is
periodically given commands (usually from the user) to make a change to
the current window/frame state: i.e. create a new frame, delete a
window, etc.

@item
The buffer mechanism is responsible for keeping track of what buffers
exist and what text is in them.  It is periodically given commands
(usually from the user) to insert or delete text, create a buffer, etc.
When it receives a text-change command, it notifies the redisplay
mechanism.

@item
The redisplay mechanism is responsible for making sure that windows and
frames are displayed correctly.  It is periodically told (by the event
loop) to actually ``do its job'', i.e. snoop around and see what the
current state of the environment (mostly of the currently-existing
windows, frames, and buffers) is, and make sure that that state matches
what's actually displayed.  It keeps lots and lots of information around
(such as what is actually being displayed currently, and what the
environment was last time it checked) so that it can minimize the work
it has to do.  It is also helped along in that whenever a relevant
change to the environment occurs, the redisplay mechanism is told about
this, so it has a pretty good idea of where it has to look to find
possible changes and doesn't have to look everywhere.

@item
The Lisp engine is responsible for executing the Lisp code in which most
user commands are written.  It is entered through a call to @code{eval}
or @code{funcall}, which occurs as a result of dispatching an event from
the event loop.  The functions it calls issue commands to the buffer
mechanism, the window/frame subsystem, etc.

@item
The Lisp allocation subsystem is responsible for keeping track of Lisp
objects.  It is given commands from the Lisp engine to allocate objects,
garbage collect, etc.
@end itemize

etc.

  The important idea here is that there are a number of independent
subsystems each with its own responsibility and persistent state, just
like different employees in a company, and each subsystem is
periodically given commands from other subsystems.  Commands can flow
from any one subsystem to any other, but there is usually some sort of
hierarchy, with all commands originating from the event subsystem.

  XEmacs is entered in @code{main()}, which is in @file{emacs.c}.  When
this is called the first time (in a properly-invoked @file{temacs}), it
does the following:

@enumerate
@item
It does some very basic environment initializations, such as determining
where it and its directories (e.g. @file{lisp/} and @file{etc/}) reside
and setting up signal handlers.
@item
It initializes the entire Lisp interpreter.
@item
It sets the initial values of many built-in variables (including many
variables that are visible to Lisp programs), such as the global keymap
object and the built-in faces (a face is an object that describes the
display characteristics of text).  This involves creating Lisp objects
and thus is dependent on step (2).
@item
It performs various other initializations that are relevant to the
particular environment it is running in, such as retrieving environment
variables, determining the current date and the user who is running the
program, examining its standard input, creating any necessary file
descriptors, etc.
@item
At this point, the C initialization is complete.  A Lisp program that
was specified on the command line (usually @file{loadup.el}) is called
(temacs is normally invoked as @code{temacs -batch -l loadup.el dump}).
@file{loadup.el} loads all of the other Lisp files that are needed for
the operation of the editor, calls the @code{dump-emacs} function to
write out @file{xemacs}, and then kills the temacs process.
@end enumerate

  When @file{xemacs} is then run, it only redoes steps (1) and (4)
above; all variables already contain the values they were set to when
the executable was dumped, and all memory that was allocated with
@code{malloc()} is still around. (XEmacs knows whether it is being run
as @file{xemacs} or @file{temacs} because it sets the global variable
@code{initialized} to 1 after step (4) above.) At this point,
@file{xemacs} calls a Lisp function to do any further initialization,
which includes parsing the command-line (the C code can only do limited
command-line parsing, which includes looking for the @samp{-batch} and
@samp{-l} flags and a few other flags that it needs to know about before
initialization is complete), creating the first frame (or @dfn{window}
in standard window-system parlance), running the user's init file
(usually the file @file{.emacs} in the user's home directory), etc.  The
function to do this is usually called @code{normal-top-level};
@file{loadup.el} tells the C code about this function by setting its
name as the value of the Lisp variable @code{top-level}.

  When the Lisp initialization code is done, the C code enters the event
loop, and stays there for the duration of the XEmacs process.  The code
for the event loop is contained in @file{keyboard.c}, and is called
@code{Fcommand_loop_1()}.  Note that this event loop could very well be
written in Lisp, and in fact a Lisp version exists; but apparently,
doing this makes XEmacs run noticeably slower.

  Notice how much of the initialization is done in Lisp, not in C.
In general, XEmacs tries to move as much code as is possible
into Lisp.  Code that remains in C is code that implements the
Lisp interpreter itself, or code that needs to be very fast, or
code that needs to do system calls or other such stuff that
needs to be done in C, or code that needs to have access to
``forbidden'' structures. (One conscious aspect of the design of
Lisp under XEmacs is a clean separation between the external
interface to a Lisp object's functionality and its internal
implementation.  Part of this design is that Lisp programs
are forbidden from accessing the contents of the object other
than through using a standard API.  In this respect, XEmacs Lisp
is similar to modern Lisp dialects but differs from GNU Emacs,
which tends to expose the implementation and allow Lisp
programs to look at it directly.  The major advantage of
hiding the implementation is that it allows the implementation
to be redesigned without affecting any Lisp programs, including
those that might want to be ``clever'' by looking directly at
the object's contents and possibly manipulating them.)

  Moving code into Lisp makes the code easier to debug and maintain and
makes it much easier for people who are not XEmacs developers to
customize XEmacs, because they can make a change with much less chance
of obscure and unwanted interactions occurring than if they were to
change the C code.

@node The XEmacs Object System (Abstractly Speaking), How Lisp Objects Are Represented in C, XEmacs From the Inside, Top
@chapter The XEmacs Object System (Abstractly Speaking)

  At the heart of the Lisp interpreter is its management of objects.
XEmacs Lisp contains many built-in objects, some of which are
simple and others of which can be very complex; and some of which
are very common, and others of which are rarely used or are only
used internally. (Since the Lisp allocation system, with its
automatic reclamation of unused storage, is so much more convenient
than @code{malloc()} and @code{free()}, the C code makes extensive use of it
in its internal operations.)

  The basic Lisp objects are

@table @code
@item integer
28 or 31 bits of precision, or 60 or 63 bits on 64-bit machines; the
reason for this is described below when the internal Lisp object
representation is described.
@item float
Same precision as a double in C.
@item cons
A simple container for two Lisp objects, used to implement lists and
most other data structures in Lisp.
@item char
An object representing a single character of text; chars behave like
integers in many ways but are logically considered text rather than
numbers and have a different read syntax. (the read syntax for a char
contains the char itself or some textual encoding of it---for example,
a Japanese Kanji character might be encoded as @samp{^[$(B#&^[(B} using the
ISO-2022 encoding standard---rather than the numerical representation
of the char; this way, if the mapping between chars and integers
changes, which is quite possible for Kanji characters and other extended
characters, the same character will still be created.  Note that some
primitives confuse chars and integers.  The worst culprit is @code{eq},
which makes a special exception and considers a char to be @code{eq} to
its integer equivalent, even though in no other case are objects of two
different types @code{eq}.  The reason for this monstrosity is
compatibility with existing code; the separation of char from integer
came fairly recently.)
@item symbol
An object that contains Lisp objects and is referred to by name;
symbols are used to implement variables and named functions
and to provide the equivalent of preprocessor constants in C.
@item vector
A one-dimensional array of Lisp objects providing constant-time access
to any of the objects; access to an arbitrary object in a vector is
faster than for lists, but the operations that can be done on a vector
are more limited.
@item string
Self-explanatory; behaves much like a vector of chars
but has a different read syntax and is stored and manipulated
more compactly.
@item bit-vector
A vector of bits; similar to a string in spirit.
@item compiled-function
An object containing compiled Lisp code, known as @dfn{byte code}.
@item subr
A Lisp primitive, i.e. a Lisp-callable function implemented in C.
@end table

@cindex closure
Note that there is no basic ``function'' type, as in more powerful
versions of Lisp (where it's called a @dfn{closure}).  XEmacs Lisp does
not provide the closure semantics implemented by Common Lisp and Scheme.
The guts of a function in XEmacs Lisp are represented in one of four
ways: a symbol specifying another function (when one function is an
alias for another), a list (whose first element must be the symbol
@code{lambda}) containing the function's source code, a
compiled-function object, or a subr object. (In other words, given a
symbol specifying the name of a function, calling @code{symbol-function}
to retrieve the contents of the symbol's function cell will return one
of these types of objects.)

XEmacs Lisp also contains numerous specialized objects used to implement
the editor:

@table @code
@item buffer
Stores text like a string, but is optimized for insertion and deletion
and has certain other properties that can be set.
@item frame
An object with various properties whose displayable representation is a
@dfn{window} in window-system parlance.
@item window
A section of a frame that displays the contents of a buffer;
often called a @dfn{pane} in window-system parlance.
@item window-configuration
An object that represents a saved configuration of windows in a frame.
@item device
An object representing a screen on which frames can be displayed;
equivalent to a @dfn{display} in the X Window System and a @dfn{TTY} in
character mode.
@item face
An object specifying the appearance of text or graphics; it has
properties such as font, foreground color, and background color.
@item marker
An object that refers to a particular position in a buffer and moves
around as text is inserted and deleted to stay in the same relative
position to the text around it.
@item extent
Similar to a marker but covers a range of text in a buffer; can also
specify properties of the text, such as a face in which the text is to
be displayed, whether the text is invisible or unmodifiable, etc.
@item event
Generated by calling @code{next-event} and contains information
describing a particular event happening in the system, such as the user
pressing a key or a process terminating.
@item keymap
An object that maps from events (described using lists, vectors, and
symbols rather than with an event object because the mapping is for
classes of events, rather than individual events) to functions to
execute or other events to recursively look up; the functions are
described by name, using a symbol, or using lists to specify the
function's code.
@item glyph
An object that describes the appearance of an image (e.g.  pixmap) on
the screen; glyphs can be attached to the beginning or end of extents
and in some future version of XEmacs will be able to be inserted
directly into a buffer.
@item process
An object that describes a connection to an externally-running process.
@end table

  There are some other, less-commonly-encountered general objects:

@table @code
@item hash-table
An object that maps from an arbitrary Lisp object to another arbitrary
Lisp object, using hashing for fast lookup.
@item obarray
A limited form of hash-table that maps from strings to symbols; obarrays
are used to look up a symbol given its name and are not actually their
own object type but are kludgily represented using vectors with hidden
fields (this representation derives from GNU Emacs).
@item specifier
A complex object used to specify the value of a display property; a
default value is given and different values can be specified for
particular frames, buffers, windows, devices, or classes of device.
@item char-table
An object that maps from chars or classes of chars to arbitrary Lisp
objects; internally char tables use a complex nested-vector
representation that is optimized to the way characters are represented
as integers.
@item range-table
An object that maps from ranges of integers to arbitrary Lisp objects.
@end table

  And some strange special-purpose objects:

@table @code
@item charset
@itemx coding-system
Objects used when MULE, or multi-lingual/Asian-language, support is
enabled.
@item color-instance
@itemx font-instance
@itemx image-instance
An object that encapsulates a window-system resource; instances are
mostly used internally but are exposed on the Lisp level for cleanness
of the specifier model and because it's occasionally useful for Lisp
program to create or query the properties of instances.
@item subwindow
An object that encapsulate a @dfn{subwindow} resource, i.e. a
window-system child window that is drawn into by an external process;
this object should be integrated into the glyph system but isn't yet,
and may change form when this is done.
@item tooltalk-message
@itemx tooltalk-pattern
Objects that represent resources used in the ToolTalk interprocess
communication protocol.
@item toolbar-button
An object used in conjunction with the toolbar.
@end table

  And objects that are only used internally:

@table @code
@item opaque
A generic object for encapsulating arbitrary memory; this allows you the
generality of @code{malloc()} and the convenience of the Lisp object
system.
@item lstream
A buffering I/O stream, used to provide a unified interface to anything
that can accept output or provide input, such as a file descriptor, a
stdio stream, a chunk of memory, a Lisp buffer, a Lisp string, etc.;
it's a Lisp object to make its memory management more convenient.
@item char-table-entry
Subsidiary objects in the internal char-table representation.
@item extent-auxiliary
@itemx menubar-data
@itemx toolbar-data
Various special-purpose objects that are basically just used to
encapsulate memory for particular subsystems, similar to the more
general ``opaque'' object.
@item symbol-value-forward
@itemx symbol-value-buffer-local
@itemx symbol-value-varalias
@itemx symbol-value-lisp-magic
Special internal-only objects that are placed in the value cell of a
symbol to indicate that there is something special with this variable --
e.g. it has no value, it mirrors another variable, or it mirrors some C
variable; there is really only one kind of object, called a
@dfn{symbol-value-magic}, but it is sort-of halfway kludged into
semi-different object types.
@end table

@cindex permanent objects
@cindex temporary objects
  Some types of objects are @dfn{permanent}, meaning that once created,
they do not disappear until explicitly destroyed, using a function such
as @code{delete-buffer}, @code{delete-window}, @code{delete-frame}, etc.
Others will disappear once they are not longer used, through the garbage
collection mechanism.  Buffers, frames, windows, devices, and processes
are among the objects that are permanent.  Note that some objects can go
both ways: Faces can be created either way; extents are normally
permanent, but detached extents (extents not referring to any text, as
happens to some extents when the text they are referring to is deleted)
are temporary.  Note that some permanent objects, such as faces and
coding systems, cannot be deleted.  Note also that windows are unique in
that they can be @emph{undeleted} after having previously been
deleted. (This happens as a result of restoring a window configuration.)

@cindex read syntax
  Note that many types of objects have a @dfn{read syntax}, i.e. a way of
specifying an object of that type in Lisp code.  When you load a Lisp
file, or type in code to be evaluated, what really happens is that the
function @code{read} is called, which reads some text and creates an object
based on the syntax of that text; then @code{eval} is called, which
possibly does something special; then this loop repeats until there's
no more text to read. (@code{eval} only actually does something special
with symbols, which causes the symbol's value to be returned,
similar to referencing a variable; and with conses [i.e. lists],
which cause a function invocation.  All other values are returned
unchanged.)

  The read syntax

@example
17297
@end example

converts to an integer whose value is 17297.

@example
1.983e-4
@end example

converts to a float whose value is 1.983e-4, or .0001983.

@example
?b
@end example

converts to a char that represents the lowercase letter b.

@example
?^[$(B#&^[(B
@end example

(where @samp{^[} actually is an @samp{ESC} character) converts to a
particular Kanji character when using an ISO2022-based coding system for
input. (To decode this goo: @samp{ESC} begins an escape sequence;
@samp{ESC $ (} is a class of escape sequences meaning ``switch to a
94x94 character set''; @samp{ESC $ ( B} means ``switch to Japanese
Kanji''; @samp{#} and @samp{&} collectively index into a 94-by-94 array
of characters [subtract 33 from the ASCII value of each character to get
the corresponding index]; @samp{ESC (} is a class of escape sequences
meaning ``switch to a 94 character set''; @samp{ESC (B} means ``switch
to US ASCII''.  It is a coincidence that the letter @samp{B} is used to
denote both Japanese Kanji and US ASCII.  If the first @samp{B} were
replaced with an @samp{A}, you'd be requesting a Chinese Hanzi character
from the GB2312 character set.)

@example
"foobar"
@end example

converts to a string.

@example
foobar
@end example

converts to a symbol whose name is @code{"foobar"}.  This is done by
looking up the string equivalent in the global variable
@code{obarray}, whose contents should be an obarray.  If no symbol
is found, a new symbol with the name @code{"foobar"} is automatically
created and added to @code{obarray}; this process is called
@dfn{interning} the symbol.
@cindex interning

@example
(foo . bar)
@end example

converts to a cons cell containing the symbols @code{foo} and @code{bar}.

@example
(1 a 2.5)
@end example

converts to a three-element list containing the specified objects
(note that a list is actually a set of nested conses; see the
XEmacs Lisp Reference).

@example
[1 a 2.5]
@end example

converts to a three-element vector containing the specified objects.

@example
#[... ... ... ...]
@end example

converts to a compiled-function object (the actual contents are not
shown since they are not relevant here; look at a file that ends with
@file{.elc} for examples).

@example
#*01110110
@end example

converts to a bit-vector.

@example
#s(hash-table ... ...)
@end example

converts to a hash table (the actual contents are not shown).

@example
#s(range-table ... ...)
@end example

converts to a range table (the actual contents are not shown).

@example
#s(char-table ... ...)
@end example

converts to a char table (the actual contents are not shown).

Note that the @code{#s()} syntax is the general syntax for structures,
which are not really implemented in XEmacs Lisp but should be.

When an object is printed out (using @code{print} or a related
function), the read syntax is used, so that the same object can be read
in again.

The other objects do not have read syntaxes, usually because it does not
really make sense to create them in this fashion (i.e.  processes, where
it doesn't make sense to have a subprocess created as a side effect of
reading some Lisp code), or because they can't be created at all
(e.g. subrs).  Permanent objects, as a rule, do not have a read syntax;
nor do most complex objects, which contain too much state to be easily
initialized through a read syntax.

@node How Lisp Objects Are Represented in C, Rules When Writing New C Code, The XEmacs Object System (Abstractly Speaking), Top
@chapter How Lisp Objects Are Represented in C

Lisp objects are represented in C using a 32-bit or 64-bit machine word
(depending on the processor; i.e. DEC Alphas use 64-bit Lisp objects and
most other processors use 32-bit Lisp objects).  The representation
stuffs a pointer together with a tag, as follows:

@example
 [ 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 ]
 [ 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 ]

   <---> ^ <------------------------------------------------------>
    tag  |       a pointer to a structure, or an integer
         |
       mark bit
@end example

The tag describes the type of the Lisp object.  For integers and chars,
the lower 28 bits contain the value of the integer or char; for all
others, the lower 28 bits contain a pointer.  The mark bit is used
during garbage-collection, and is always 0 when garbage collection is
not happening. (The way that garbage collection works, basically, is that it
loops over all places where Lisp objects could exist---this includes
all global variables in C that contain Lisp objects [including
@code{Vobarray}, the C equivalent of @code{obarray}; through this, all
Lisp variables will get marked], plus various other places---and
recursively scans through the Lisp objects, marking each object it finds
by setting the mark bit.  Then it goes through the lists of all objects
allocated, freeing the ones that are not marked and turning off the mark
bit of the ones that are marked.)

Lisp objects use the typedef @code{Lisp_Object}, but the actual C type
used for the Lisp object can vary.  It can be either a simple type
(@code{long} on the DEC Alpha, @code{int} on other machines) or a
structure whose fields are bit fields that line up properly (actually, a
union of structures is used).  Generally the simple integral type is
preferable because it ensures that the compiler will actually use a
machine word to represent the object (some compilers will use more
general and less efficient code for unions and structs even if they can
fit in a machine word).  The union type, however, has the advantage of
stricter type checking (if you accidentally pass an integer where a Lisp
object is desired, you get a compile error), and it makes it easier to
decode Lisp objects when debugging.  The choice of which type to use is
determined by the preprocessor constant @code{USE_UNION_TYPE} which is
defined via the @code{--use-union-type} option to @code{configure}.

@cindex record type

Note that there are only eight types that the tag can represent, but
many more actual types than this.  This is handled by having one of the
tag types specify a meta-type called a @dfn{record}; for all such
objects, the first four bytes of the pointed-to structure indicate what
the actual type is.

Note also that having 28 bits for pointers and integers restricts a lot
of things to 256 megabytes of memory. (Basically, enough pointers and
indices and whatnot get stuffed into Lisp objects that the total amount
of memory used by XEmacs can't grow above 256 megabytes.  In older
versions of XEmacs and GNU Emacs, the tag was 5 bits wide, allowing for
32 types, which was more than the actual number of types that existed at
the time, and no ``record'' type was necessary.  However, this limited
the editor to 64 megabytes total, which some users who edited large
files might conceivably exceed.)

Also, note that there is an implicit assumption here that all pointers
are low enough that the top bits are all zero and can just be chopped
off.  On standard machines that allocate memory from the bottom up (and
give each process its own address space), this works fine.  Some
machines, however, put the data space somewhere else in memory
(e.g. beginning at 0x80000000).  Those machines cope by defining
@code{DATA_SEG_BITS} in the corresponding @file{m/} or @file{s/} file to
the proper mask.  Then, pointers retrieved from Lisp objects are
automatically OR'ed with this value prior to being used.

A corollary of the previous paragraph is that @strong{(pointers to)
stack-allocated structures cannot be put into Lisp objects}.  The stack
is generally located near the top of memory; if you put such a pointer
into a Lisp object, it will get its top bits chopped off, and you will
lose.

Actually, there's an alternative representation of a @code{Lisp_Object},
invented by Kyle Jones, that is used when the
@code{--use-minimal-tagbits} option to @code{configure} is used.  In
this case the 2 lower bits are used for the tag bits.  This
representation assumes that pointers to structs are always aligned to
multiples of 4, so the lower 2 bits are always zero.

@example
 [ 3 3 2 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 ]
 [ 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 9 8 7 6 5 4 3 2 1 0 ]

   <---------------------------------------------------------> <->
            a pointer to a structure, or an integer            tag
@end example

A tag of 00 is used for all pointer object types, a tag of 10 is used
for characters, and the other two tags 01 and 11 are joined together to
form the integer object type.  The markbit is moved to part of the
structure being pointed at (integers and chars do not need to be marked,
since no memory is allocated).  This representation has these
advantages:

@enumerate
@item
31 bits can be used for Lisp Integers.
@item
@emph{Any} pointer can be represented directly, and no bit masking
operations are necessary.
@end enumerate

The disadvantages are:

@enumerate
@item
An extra level of indirection is needed when accessing the object types
that were not record types.  So checking whether a Lisp object is a cons
cell becomes a slower operation.
@item
Mark bits can no longer be stored directly in Lisp objects, so another
place for them must be found.  This means that a cons cell requires more
memory than merely room for 2 lisp objects, leading to extra memory use.
@end enumerate

Various macros are used to construct Lisp objects and extract the
components.  Macros of the form @code{XINT()}, @code{XCHAR()},
@code{XSTRING()}, @code{XSYMBOL()}, etc. mask out the pointer/integer
field and cast it to the appropriate type.  All of the macros that
construct pointers will @code{OR} with @code{DATA_SEG_BITS} if
necessary.  @code{XINT()} needs to be a bit tricky so that negative
numbers are properly sign-extended: Usually it does this by shifting the
number four bits to the left and then four bits to the right.  This
assumes that the right-shift operator does an arithmetic shift (i.e. it
leaves the most-significant bit as-is rather than shifting in a zero, so
that it mimics a divide-by-two even for negative numbers).  Not all
machines/compilers do this, and on the ones that don't, a more
complicated definition is selected by defining
@code{EXPLICIT_SIGN_EXTEND}.

Note that when @code{ERROR_CHECK_TYPECHECK} is defined, the extractor
macros become more complicated---they check the tag bits and/or the
type field in the first four bytes of a record type to ensure that the
object is really of the correct type.  This is great for catching places
where an incorrect type is being dereferenced---this typically results
in a pointer being dereferenced as the wrong type of structure, with
unpredictable (and sometimes not easily traceable) results.

There are similar @code{XSET@var{TYPE}()} macros that construct a Lisp
object.  These macros are of the form @code{XSET@var{TYPE}
(@var{lvalue}, @var{result})},
i.e. they have to be a statement rather than just used in an expression.
The reason for this is that standard C doesn't let you ``construct'' a
structure (but GCC does).  Granted, this sometimes isn't too convenient;
for the case of integers, at least, you can use the function
@code{make_int()}, which constructs and @emph{returns} an integer
Lisp object.  Note that the @code{XSET@var{TYPE}()} macros are also
affected by @code{ERROR_CHECK_TYPECHECK} and make sure that the
structure is of the right type in the case of record types, where the
type is contained in the structure.

The C programmer is responsible for @strong{guaranteeing} that a
Lisp_Object is is the correct type before using the @code{X@var{TYPE}}
macros.  This is especially important in the case of lists.  Use
@code{XCAR} and @code{XCDR} if a Lisp_Object is certainly a cons cell,
else use @code{Fcar()} and @code{Fcdr()}.  Trust other C code, but not
Lisp code.  On the other hand, if XEmacs has an internal logic error,
it's better to crash immediately, so sprinkle ``unreachable''
@code{abort()}s liberally about the source code.

@node Rules When Writing New C Code, A Summary of the Various XEmacs Modules, How Lisp Objects Are Represented in C, Top
@chapter Rules When Writing New C Code

The XEmacs C Code is extremely complex and intricate, and there are many
rules that are more or less consistently followed throughout the code.
Many of these rules are not obvious, so they are explained here.  It is
of the utmost importance that you follow them.  If you don't, you may
get something that appears to work, but which will crash in odd
situations, often in code far away from where the actual breakage is.

@menu
* General Coding Rules::
* Writing Lisp Primitives::
* Adding Global Lisp Variables::
* Coding for Mule::
* Techniques for XEmacs Developers::
@end menu

@node General Coding Rules
@section General Coding Rules

The C code is actually written in a dialect of C called @dfn{Clean C},
meaning that it can be compiled, mostly warning-free, with either a C or
C++ compiler.  Coding in Clean C has several advantages over plain C.
C++ compilers are more nit-picking, and a number of coding errors have
been found by compiling with C++.  The ability to use both C and C++
tools means that a greater variety of development tools are available to
the developer.

Almost every module contains a @code{syms_of_*()} function and a
@code{vars_of_*()} function.  The former declares any Lisp primitives
you have defined and defines any symbols you will be using.  The latter
declares any global Lisp variables you have added and initializes global
C variables in the module.  For each such function, declare it in
@file{symsinit.h} and make sure it's called in the appropriate place in
@file{emacs.c}.  @strong{Important}: There are stringent requirements on
exactly what can go into these functions.  See the comment in
@file{emacs.c}.  The reason for this is to avoid obscure unwanted
interactions during initialization.  If you don't follow these rules,
you'll be sorry!  If you want to do anything that isn't allowed, create
a @code{complex_vars_of_*()} function for it.  Doing this is tricky,
though: You have to make sure your function is called at the right time
so that all the initialization dependencies work out.

Every module includes @file{<config.h>} (angle brackets so that
@samp{--srcdir} works correctly; @file{config.h} may or may not be in
the same directory as the C sources) and @file{lisp.h}.  @file{config.h}
must always be included before any other header files (including
system header files) to ensure that certain tricks played by various
@file{s/} and @file{m/} files work out correctly.

When including header files, always use angle brackets, not double
quotes, except when the file to be included is in the same directory as
the including file.  If either file is a generated file, then that is
not likely to be the case.  In order to understand why we have this
rule, imagine what happens when you do a build in the source directory
using @samp{./configure} and another build in another directory using
@samp{../work/configure}.  There will be two different @file{config.h}
files.  Which one will be used if you @samp{#include "config.h"}?

@strong{All global and static variables that are to be modifiable must
be declared uninitialized.}  This means that you may not use the
``declare with initializer'' form for these variables, such as @code{int
some_variable = 0;}.  The reason for this has to do with some kludges
done during the dumping process: If possible, the initialized data
segment is re-mapped so that it becomes part of the (unmodifiable) code
segment in the dumped executable.  This allows this memory to be shared
among multiple running XEmacs processes.  XEmacs is careful to place as
much constant data as possible into initialized variables (in
particular, into what's called the @dfn{pure space}---see below) during
the @file{temacs} phase.

@cindex copy-on-write
@strong{Please note:} This kludge only works on a few systems nowadays,
and is rapidly becoming irrelevant because most modern operating systems
provide @dfn{copy-on-write} semantics.  All data is initially shared
between processes, and a private copy is automatically made (on a
page-by-page basis) when a process first attempts to write to a page of
memory.

Formerly, there was a requirement that static variables not be declared
inside of functions.  This had to do with another hack along the same
vein as what was just described: old USG systems put statically-declared
variables in the initialized data space, so those header files had a
@code{#define static} declaration. (That way, the data-segment remapping
described above could still work.) This fails badly on static variables
inside of functions, which suddenly become automatic variables;
therefore, you weren't supposed to have any of them.  This awful kludge
has been removed in XEmacs because

@enumerate
@item
almost all of the systems that used this kludge ended up having
to disable the data-segment remapping anyway;
@item
the only systems that didn't were extremely outdated ones;
@item
this hack completely messed up inline functions.
@end enumerate

The C source code makes heavy use of C preprocessor macros.  One popular
macro style is:

@example
#define FOO(var, value) do @{           \
  Lisp_Object FOO_value = (value);      \
  ... /* compute using FOO_value */     \
  (var) = bar;                          \
@} while (0)
@end example

The @code{do @{...@} while (0)} is a standard trick to allow FOO to have
statement semantics, so that it can safely be used within an @code{if}
statement in C, for example.  Multiple evaluation is prevented by
copying a supplied argument into a local variable, so that
@code{FOO(var,fun(1))} only calls @code{fun} once.

Lisp lists are popular data structures in the C code as well as in
Elisp.  There are two sets of macros that iterate over lists.
@code{EXTERNAL_LIST_LOOP_@var{n}} should be used when the list has been
supplied by the user, and cannot be trusted to be acyclic and
nil-terminated.  A @code{malformed-list} or @code{circular-list} error
will be generated if the list being iterated over is not entirely
kosher.  @code{LIST_LOOP_@var{n}}, on the other hand, is faster and less
safe, and can be used only on trusted lists.

Related macros are @code{GET_EXTERNAL_LIST_LENGTH} and
@code{GET_LIST_LENGTH}, which calculate the length of a list, and in the
case of @code{GET_EXTERNAL_LIST_LENGTH}, validating the properness of
the list.  The macros @code{EXTERNAL_LIST_LOOP_DELETE_IF} and
@code{LIST_LOOP_DELETE_IF} delete elements from a lisp list satisfying some
predicate.

@node Writing Lisp Primitives
@section Writing Lisp Primitives

Lisp primitives are Lisp functions implemented in C.  The details of
interfacing the C function so that Lisp can call it are handled by a few
C macros.  The only way to really understand how to write new C code is
to read the source, but we can explain some things here.

An example of a special form is the definition of @code{prog1}, from
@file{eval.c}.  (An ordinary function would have the same general
appearance.)

@cindex garbage collection protection
@smallexample
@group
DEFUN ("prog1", Fprog1, 1, UNEVALLED, 0, /*
Similar to `progn', but the value of the first form is returned.
\(prog1 FIRST BODY...): All the arguments are evaluated sequentially.
The value of FIRST is saved during evaluation of the remaining args,
whose values are discarded.
*/
       (args))
@{
  /* This function can GC */
  REGISTER Lisp_Object val, form, tail;
  struct gcpro gcpro1;

  val = Feval (XCAR (args));

  GCPRO1 (val);

  LIST_LOOP_3 (form, XCDR (args), tail)
    Feval (form);

  UNGCPRO;
  return val;
@}
@end group
@end smallexample

  Let's start with a precise explanation of the arguments to the
@code{DEFUN} macro.  Here is a template for them:

@example
@group
DEFUN (@var{lname}, @var{fname}, @var{min_args}, @var{max_args}, @var{interactive}, /*
@var{docstring}
*/
   (@var{arglist}))
@end group
@end example

@table @var
@item lname
This string is the name of the Lisp symbol to define as the function
name; in the example above, it is @code{"prog1"}.

@item fname
This is the C function name for this function.  This is the name that is
used in C code for calling the function.  The name is, by convention,
@samp{F} prepended to the Lisp name, with all dashes (@samp{-}) in the
Lisp name changed to underscores.  Thus, to call this function from C
code, call @code{Fprog1}.  Remember that the arguments are of type
@code{Lisp_Object}; various macros and functions for creating values of
type @code{Lisp_Object} are declared in the file @file{lisp.h}.

Primitives whose names are special characters (e.g. @code{+} or
@code{<}) are named by spelling out, in some fashion, the special
character: e.g. @code{Fplus()} or @code{Flss()}.  Primitives whose names
begin with normal alphanumeric characters but also contain special
characters are spelled out in some creative way, e.g. @code{let*}
becomes @code{FletX()}.

Each function also has an associated structure that holds the data for
the subr object that represents the function in Lisp.  This structure
conveys the Lisp symbol name to the initialization routine that will
create the symbol and store the subr object as its definition.  The C
variable name of this structure is always @samp{S} prepended to the
@var{fname}.  You hardly ever need to be aware of the existence of this
structure, since @code{DEFUN} plus @code{DEFSUBR} takes care of all the
details.

@item min_args
This is the minimum number of arguments that the function requires.  The
function @code{prog1} allows a minimum of one argument.

@item max_args
This is the maximum number of arguments that the function accepts, if
there is a fixed maximum.  Alternatively, it can be @code{UNEVALLED},
indicating a special form that receives unevaluated arguments, or
@code{MANY}, indicating an unlimited number of evaluated arguments (the
C equivalent of @code{&rest}).  Both @code{UNEVALLED} and @code{MANY}
are macros.  If @var{max_args} is a number, it may not be less than
@var{min_args} and it may not be greater than 8. (If you need to add a
function with more than 8 arguments, use the @code{MANY} form.  Resist
the urge to edit the definition of @code{DEFUN} in @file{lisp.h}.  If
you do it anyways, make sure to also add another clause to the switch
statement in @code{primitive_funcall().})

@item interactive
This is an interactive specification, a string such as might be used as
the argument of @code{interactive} in a Lisp function.  In the case of
@code{prog1}, it is 0 (a null pointer), indicating that @code{prog1}
cannot be called interactively.  A value of @code{""} indicates a
function that should receive no arguments when called interactively.

@item docstring
This is the documentation string.  It is written just like a
documentation string for a function defined in Lisp; in particular, the
first line should be a single sentence.  Note how the documentation
string is enclosed in a comment, none of the documentation is placed on
the same lines as the comment-start and comment-end characters, and the
comment-start characters are on the same line as the interactive
specification.  @file{make-docfile}, which scans the C files for
documentation strings, is very particular about what it looks for, and
will not properly extract the doc string if it's not in this exact format.

In order to make both @file{etags} and @file{make-docfile} happy, make
sure that the @code{DEFUN} line contains the @var{lname} and
@var{fname}, and that the comment-start characters for the doc string
are on the same line as the interactive specification, and put a newline
directly after them (and before the comment-end characters).

@item arglist
This is the comma-separated list of arguments to the C function.  For a
function with a fixed maximum number of arguments, provide a C argument
for each Lisp argument.  In this case, unlike regular C functions, the
types of the arguments are not declared; they are simply always of type
@code{Lisp_Object}.

The names of the C arguments will be used as the names of the arguments
to the Lisp primitive as displayed in its documentation, modulo the same
concerns described above for @code{F...} names (in particular,
underscores in the C arguments become dashes in the Lisp arguments).

There is one additional kludge: A trailing `_' on the C argument is
discarded when forming the Lisp argument.  This allows C language
reserved words (like @code{default}) or global symbols (like
@code{dirname}) to be used as argument names without compiler warnings
or errors.

A Lisp function with @w{@var{max_args} = @code{UNEVALLED}} is a
@w{@dfn{special form}}; its arguments are not evaluated.  Instead it
receives one argument of type @code{Lisp_Object}, a (Lisp) list of the
unevaluated arguments, conventionally named @code{(args)}.

When a Lisp function has no upper limit on the number of arguments,
specify @w{@var{max_args} = @code{MANY}}.  In this case its implementation in
C actually receives exactly two arguments: the number of Lisp arguments
(an @code{int}) and the address of a block containing their values (a
@w{@code{Lisp_Object *}}).  In this case only are the C types specified
in the @var{arglist}: @w{@code{(int nargs, Lisp_Object *args)}}.

@end table

Within the function @code{Fprog1} itself, note the use of the macros
@code{GCPRO1} and @code{UNGCPRO}.  @code{GCPRO1} is used to ``protect''
a variable from garbage collection---to inform the garbage collector
that it must look in that variable and regard the object pointed at by
its contents as an accessible object.  This is necessary whenever you
call @code{Feval} or anything that can directly or indirectly call
@code{Feval} (this includes the @code{QUIT} macro!).  At such a time,
any Lisp object that you intend to refer to again must be protected
somehow.  @code{UNGCPRO} cancels the protection of the variables that
are protected in the current function.  It is necessary to do this
explicitly.

The macro @code{GCPRO1} protects just one local variable.  If you want
to protect two, use @code{GCPRO2} instead; repeating @code{GCPRO1} will
not work.  Macros @code{GCPRO3} and @code{GCPRO4} also exist.

These macros implicitly use local variables such as @code{gcpro1}; you
must declare these explicitly, with type @code{struct gcpro}.  Thus, if
you use @code{GCPRO2}, you must declare @code{gcpro1} and @code{gcpro2}.

@cindex caller-protects (@code{GCPRO} rule)
Note also that the general rule is @dfn{caller-protects}; i.e. you are
only responsible for protecting those Lisp objects that you create.  Any
objects passed to you as arguments should have been protected by whoever
created them, so you don't in general have to protect them.

In particular, the arguments to any Lisp primitive are always
automatically @code{GCPRO}ed, when called ``normally'' from Lisp code or
bytecode.  So only a few Lisp primitives that are called frequently from
C code, such as @code{Fprogn} protect their arguments as a service to
their caller.  You don't need to protect your arguments when writing a
new @code{DEFUN}.

@code{GCPRO}ing is perhaps the trickiest and most error-prone part of
XEmacs coding.  It is @strong{extremely} important that you get this
right and use a great deal of discipline when writing this code.
@xref{GCPROing, ,@code{GCPRO}ing}, for full details on how to do this.

What @code{DEFUN} actually does is declare a global structure of type
@code{Lisp_Subr} whose name begins with capital @samp{SF} and which
contains information about the primitive (e.g. a pointer to the
function, its minimum and maximum allowed arguments, a string describing
its Lisp name); @code{DEFUN} then begins a normal C function declaration
using the @code{F...} name.  The Lisp subr object that is the function
definition of a primitive (i.e. the object in the function slot of the
symbol that names the primitive) actually points to this @samp{SF}
structure; when @code{Feval} encounters a subr, it looks in the
structure to find out how to call the C function.

Defining the C function is not enough to make a Lisp primitive
available; you must also create the Lisp symbol for the primitive (the
symbol is @dfn{interned}; @pxref{Obarrays}) and store a suitable subr
object in its function cell. (If you don't do this, the primitive won't
be seen by Lisp code.) The code looks like this:

@example
DEFSUBR (@var{fname});
@end example

@noindent
Here @var{fname} is the same name you used as the second argument to
@code{DEFUN}.

This call to @code{DEFSUBR} should go in the @code{syms_of_*()} function
at the end of the module.  If no such function exists, create it and
make sure to also declare it in @file{symsinit.h} and call it from the
appropriate spot in @code{main()}.  @xref{General Coding Rules}.

Note that C code cannot call functions by name unless they are defined
in C.  The way to call a function written in Lisp from C is to use
@code{Ffuncall}, which embodies the Lisp function @code{funcall}.  Since
the Lisp function @code{funcall} accepts an unlimited number of
arguments, in C it takes two: the number of Lisp-level arguments, and a
one-dimensional array containing their values.  The first Lisp-level
argument is the Lisp function to call, and the rest are the arguments to
pass to it.  Since @code{Ffuncall} can call the evaluator, you must
protect pointers from garbage collection around the call to
@code{Ffuncall}. (However, @code{Ffuncall} explicitly protects all of
its parameters, so you don't have to protect any pointers passed as
parameters to it.)

The C functions @code{call0}, @code{call1}, @code{call2}, and so on,
provide handy ways to call a Lisp function conveniently with a fixed
number of arguments.  They work by calling @code{Ffuncall}.

@file{eval.c} is a very good file to look through for examples;
@file{lisp.h} contains the definitions for important macros and
functions.

@node Adding Global Lisp Variables
@section Adding Global Lisp Variables

Global variables whose names begin with @samp{Q} are constants whose
value is a symbol of a particular name.  The name of the variable should
be derived from the name of the symbol using the same rules as for Lisp
primitives.  These variables are initialized using a call to
@code{defsymbol()} in the @code{syms_of_*()} function. (This call
interns a symbol, sets the C variable to the resulting Lisp object, and
calls @code{staticpro()} on the C variable to tell the
garbage-collection mechanism about this variable.  What
@code{staticpro()} does is add a pointer to the variable to a large
global array; when garbage-collection happens, all pointers listed in
the array are used as starting points for marking Lisp objects.  This is
important because it's quite possible that the only current reference to
the object is the C variable.  In the case of symbols, the
@code{staticpro()} doesn't matter all that much because the symbol is
contained in @code{obarray}, which is itself @code{staticpro()}ed.
However, it's possible that a naughty user could do something like
uninterning the symbol out of @code{obarray} or even setting
@code{obarray} to a different value [although this is likely to make
XEmacs crash!].)

  @strong{Please note:} It is potentially deadly if you declare a
@samp{Q...}  variable in two different modules.  The two calls to
@code{defsymbol()} are no problem, but some linkers will complain about
multiply-defined symbols.  The most insidious aspect of this is that
often the link will succeed anyway, but then the resulting executable
will sometimes crash in obscure ways during certain operations!  To
avoid this problem, declare any symbols with common names (such as
@code{text}) that are not obviously associated with this particular
module in the module @file{general.c}.

  Global variables whose names begin with @samp{V} are variables that
contain Lisp objects.  The convention here is that all global variables
of type @code{Lisp_Object} begin with @samp{V}, and all others don't
(including integer and boolean variables that have Lisp
equivalents). Most of the time, these variables have equivalents in
Lisp, but some don't.  Those that do are declared this way by a call to
@code{DEFVAR_LISP()} in the @code{vars_of_*()} initializer for the
module.  What this does is create a special @dfn{symbol-value-forward}
Lisp object that contains a pointer to the C variable, intern a symbol
whose name is as specified in the call to @code{DEFVAR_LISP()}, and set
its value to the symbol-value-forward Lisp object; it also calls
@code{staticpro()} on the C variable to tell the garbage-collection
mechanism about the variable.  When @code{eval} (or actually
@code{symbol-value}) encounters this special object in the process of
retrieving a variable's value, it follows the indirection to the C
variable and gets its value.  @code{setq} does similar things so that
the C variable gets changed.

  Whether or not you @code{DEFVAR_LISP()} a variable, you need to
initialize it in the @code{vars_of_*()} function; otherwise it will end
up as all zeroes, which is the integer 0 (@emph{not} @code{nil}), and
this is probably not what you want.  Also, if the variable is not
@code{DEFVAR_LISP()}ed, @strong{you must call} @code{staticpro()} on the
C variable in the @code{vars_of_*()} function.  Otherwise, the
garbage-collection mechanism won't know that the object in this variable
is in use, and will happily collect it and reuse its storage for another
Lisp object, and you will be the one who's unhappy when you can't figure
out how your variable got overwritten.

@node Coding for Mule
@section Coding for Mule
@cindex Coding for Mule

Although Mule support is not compiled by default in XEmacs, many people
are using it, and we consider it crucial that new code works correctly
with multibyte characters.  This is not hard; it is only a matter of
following several simple user-interface guidelines.  Even if you never
compile with Mule, with a little practice you will find it quite easy
to code Mule-correctly.

Note that these guidelines are not necessarily tied to the current Mule
implementation; they are also a good idea to follow on the grounds of
code generalization for future I18N work.

@menu
* Character-Related Data Types::
* Working With Character and Byte Positions::
* Conversion to and from External Data::
* General Guidelines for Writing Mule-Aware Code::
* An Example of Mule-Aware Code::
@end menu

@node Character-Related Data Types
@subsection Character-Related Data Types

First, let's review the basic character-related datatypes used by
XEmacs.  Note that the separate @code{typedef}s are not mandatory in the
current implementation (all of them boil down to @code{unsigned char} or
@code{int}), but they improve clarity of code a great deal, because one
glance at the declaration can tell the intended use of the variable.

@table @code
@item Emchar
@cindex Emchar
An @code{Emchar} holds a single Emacs character.

Obviously, the equality between characters and bytes is lost in the Mule
world.  Characters can be represented by one or more bytes in the
buffer, and @code{Emchar} is the C type large enough to hold any
character.

Without Mule support, an @code{Emchar} is equivalent to an
@code{unsigned char}.

@item Bufbyte
@cindex Bufbyte
The data representing the text in a buffer or string is logically a set
of @code{Bufbyte}s.

XEmacs does not work with character formats all the time; when reading
characters from the outside, it decodes them to an internal format, and
likewise encodes them when writing.  @code{Bufbyte} (in fact
@code{unsigned char}) is the basic unit of XEmacs internal buffers and
strings format.

One character can correspond to one or more @code{Bufbyte}s.  In the
current implementation, an ASCII character is represented by the same
@code{Bufbyte}, and extended characters are represented by a sequence of
@code{Bufbyte}s.

Without Mule support, a @code{Bufbyte} is equivalent to an
@code{Emchar}.

@item Bufpos
@itemx Charcount
@cindex Bufpos
@cindex Charcount
A @code{Bufpos} represents a character position in a buffer or string.
A @code{Charcount} represents a number (count) of characters.
Logically, subtracting two @code{Bufpos} values yields a
@code{Charcount} value.  Although all of these are @code{typedef}ed to
@code{int}, we use them in preference to @code{int} to make it clear
what sort of position is being used.

@code{Bufpos} and @code{Charcount} values are the only ones that are
ever visible to Lisp.

@item Bytind
@itemx Bytecount
@cindex Bytind
@cindex Bytecount
A @code{Bytind} represents a byte position in a buffer or string.  A
@code{Bytecount} represents the distance between two positions in bytes.
The relationship between @code{Bytind} and @code{Bytecount} is the same
as the relationship between @code{Bufpos} and @code{Charcount}.

@item Extbyte
@itemx Extcount
@cindex Extbyte
@cindex Extcount
When dealing with the outside world, XEmacs works with @code{Extbyte}s,
which are equivalent to @code{unsigned char}.  Obviously, an
@code{Extcount} is the distance between two @code{Extbyte}s.  Extbytes
and Extcounts are not all that frequent in XEmacs code.
@end table

@node Working With Character and Byte Positions
@subsection Working With Character and Byte Positions

Now that we have defined the basic character-related types, we can look
at the macros and functions designed for work with them and for
conversion between them.  Most of these macros are defined in
@file{buffer.h}, and we don't discuss all of them here, but only the
most important ones.  Examining the existing code is the best way to
learn about them.

@table @code
@item MAX_EMCHAR_LEN
@cindex MAX_EMCHAR_LEN
This preprocessor constant is the maximum number of buffer bytes per
Emacs character, i.e. the byte length of an @code{Emchar}.  It is useful
when allocating temporary strings to keep a known number of characters.
For instance:

@example
@group
@{
  Charcount cclen;
  ...
  @{
    /* Allocate place for @var{cclen} characters. */
    Bufbyte *buf = (Bufbyte *)alloca (cclen * MAX_EMCHAR_LEN);
...
@end group
@end example

If you followed the previous section, you can guess that, logically,
multiplying a @code{Charcount} value with @code{MAX_EMCHAR_LEN} produces
a @code{Bytecount} value.

In the current Mule implementation, @code{MAX_EMCHAR_LEN} equals 4.
Without Mule, it is 1.

@item charptr_emchar
@itemx set_charptr_emchar
@cindex charptr_emchar
@cindex set_charptr_emchar
The @code{charptr_emchar} macro takes a @code{Bufbyte} pointer and
returns the @code{Emchar} stored at that position.  If it were a
function, its prototype would be:

@example
Emchar charptr_emchar (Bufbyte *p);
@end example

@code{set_charptr_emchar} stores an @code{Emchar} to the specified byte
position.  It returns the number of bytes stored:

@example
Bytecount set_charptr_emchar (Bufbyte *p, Emchar c);
@end example

It is important to note that @code{set_charptr_emchar} is safe only for
appending a character at the end of a buffer, not for overwriting a
character in the middle.  This is because the width of characters
varies, and @code{set_charptr_emchar} cannot resize the string if it
writes, say, a two-byte character where a single-byte character used to
reside.

A typical use of @code{set_charptr_emchar} can be demonstrated by this
example, which copies characters from buffer @var{buf} to a temporary
string of Bufbytes.

@example
@group
@{
  Bufpos pos;
  for (pos = beg; pos < end; pos++)
    @{
      Emchar c = BUF_FETCH_CHAR (buf, pos);
      p += set_charptr_emchar (buf, c);
    @}
@}
@end group
@end example

Note how @code{set_charptr_emchar} is used to store the @code{Emchar}
and increment the counter, at the same time.

@item INC_CHARPTR
@itemx DEC_CHARPTR
@cindex INC_CHARPTR
@cindex DEC_CHARPTR
These two macros increment and decrement a @code{Bufbyte} pointer,
respectively.  They will adjust the pointer by the appropriate number of
bytes according to the byte length of the character stored there.  Both
macros assume that the memory address is located at the beginning of a
valid character.

Without Mule support, @code{INC_CHARPTR (p)} and @code{DEC_CHARPTR (p)}
simply expand to @code{p++} and @code{p--}, respectively.

@item bytecount_to_charcount
@cindex bytecount_to_charcount
Given a pointer to a text string and a length in bytes, return the
equivalent length in characters.

@example
Charcount bytecount_to_charcount (Bufbyte *p, Bytecount bc);
@end example

@item charcount_to_bytecount
@cindex charcount_to_bytecount
Given a pointer to a text string and a length in characters, return the
equivalent length in bytes.

@example
Bytecount charcount_to_bytecount (Bufbyte *p, Charcount cc);
@end example

@item charptr_n_addr
@cindex charptr_n_addr
Return a pointer to the beginning of the character offset @var{cc} (in
characters) from @var{p}.

@example
Bufbyte *charptr_n_addr (Bufbyte *p, Charcount cc);
@end example
@end table

@node Conversion to and from External Data
@subsection Conversion to and from External Data

When an external function, such as a C library function, returns a
@code{char} pointer, you should almost never treat it as @code{Bufbyte}.
This is because these returned strings may contain 8bit characters which
can be misinterpreted by XEmacs, and cause a crash.  Likewise, when
exporting a piece of internal text to the outside world, you should
always convert it to an appropriate external encoding, lest the internal
stuff (such as the infamous \201 characters) leak out.

The interface to conversion between the internal and external
representations of text are the numerous conversion macros defined in
@file{buffer.h}.  Before looking at them, we'll look at the external
formats supported by these macros.

Currently meaningful formats are @code{FORMAT_BINARY},
@code{FORMAT_FILENAME}, @code{FORMAT_OS}, and @code{FORMAT_CTEXT}.  Here
is a description of these.

@table @code
@item FORMAT_BINARY
Binary format.  This is the simplest format and is what we use in the
absence of a more appropriate format.  This converts according to the
@code{binary} coding system:

@enumerate a
@item
On input, bytes 0--255 are converted into characters 0--255.
@item
On output, characters 0--255 are converted into bytes 0--255 and other
characters are converted into `X'.
@end enumerate

@item FORMAT_FILENAME
Format used for filenames.  In the original Mule, this is user-definable
with the @code{pathname-coding-system} variable.  For the moment, we
just use the @code{binary} coding system.

@item FORMAT_OS
Format used for the external Unix environment---@code{argv[]}, stuff
from @code{getenv()}, stuff from the @file{/etc/passwd} file, etc.

Perhaps should be the same as FORMAT_FILENAME.

@item FORMAT_CTEXT
Compound--text format.  This is the standard X format used for data
stored in properties, selections, and the like.  This is an 8-bit
no-lock-shift ISO2022 coding system.
@end table

The macros to convert between these formats and the internal format, and
vice versa, follow.

@table @code
@item GET_CHARPTR_INT_DATA_ALLOCA
@itemx GET_CHARPTR_EXT_DATA_ALLOCA
These two are the most basic conversion macros.
@code{GET_CHARPTR_INT_DATA_ALLOCA} converts external data to internal
format, and @code{GET_CHARPTR_EXT_DATA_ALLOCA} converts the other way
around.  The arguments each of these receives are @var{ptr} (pointer to
the text in external format), @var{len} (length of texts in bytes),
@var{fmt} (format of the external text), @var{ptr_out} (lvalue to which
new text should be copied), and @var{len_out} (lvalue which will be
assigned the length of the internal text in bytes).  The resulting text
is stored to a stack-allocated buffer.  If the text doesn't need
changing, these macros will do nothing, except for setting
@var{len_out}.

The macros above take many arguments which makes them unwieldy.  For
this reason, a number of convenience macros are defined with obvious
functionality, but accepting less arguments.  The general rule is that
macros with @samp{INT} in their name convert text to internal Emacs
representation, whereas the @samp{EXT} macros convert to external
representation.

@item GET_C_CHARPTR_INT_DATA_ALLOCA
@itemx GET_C_CHARPTR_EXT_DATA_ALLOCA
As their names imply, these macros work on C char pointers, which are
zero-terminated, and thus do not need @var{len} or @var{len_out}
parameters.

@item GET_STRING_EXT_DATA_ALLOCA
@itemx GET_C_STRING_EXT_DATA_ALLOCA
These two macros convert a Lisp string into an external representation.
The difference between them is that @code{GET_STRING_EXT_DATA_ALLOCA}
stores its output to a generic string, providing @var{len_out}, the
length of the resulting external string.  On the other hand,
@code{GET_C_STRING_EXT_DATA_ALLOCA} assumes that the caller will be
satisfied with output string being zero-terminated.

Note that for Lisp strings only one conversion direction makes sense.

@item GET_C_CHARPTR_EXT_BINARY_DATA_ALLOCA
@itemx GET_CHARPTR_EXT_BINARY_DATA_ALLOCA
@itemx GET_STRING_BINARY_DATA_ALLOCA
@itemx GET_C_STRING_BINARY_DATA_ALLOCA
@itemx GET_C_CHARPTR_EXT_FILENAME_DATA_ALLOCA
@itemx ...
These macros convert internal text to a specific external
representation, with the external format being encoded into the name of
the macro.  Note that the @code{GET_STRING_...} and
@code{GET_C_STRING...}  macros lack the @samp{EXT} tag, because they
only make sense in that direction.

@item GET_C_CHARPTR_INT_BINARY_DATA_ALLOCA
@itemx GET_CHARPTR_INT_BINARY_DATA_ALLOCA
@itemx GET_C_CHARPTR_INT_FILENAME_DATA_ALLOCA
@itemx ...
These macros convert external text of a specific format to its internal
representation, with the external format being incoded into the name of
the macro.
@end table

@node General Guidelines for Writing Mule-Aware Code
@subsection General Guidelines for Writing Mule-Aware Code

This section contains some general guidance on how to write Mule-aware
code, as well as some pitfalls you should avoid.

@table @emph
@item Never use @code{char} and @code{char *}.
In XEmacs, the use of @code{char} and @code{char *} is almost always a
mistake.  If you want to manipulate an Emacs character from ``C'', use
@code{Emchar}.  If you want to examine a specific octet in the internal
format, use @code{Bufbyte}.  If you want a Lisp-visible character, use a
@code{Lisp_Object} and @code{make_char}.  If you want a pointer to move
through the internal text, use @code{Bufbyte *}.  Also note that you
almost certainly do not need @code{Emchar *}.

@item Be careful not to confuse @code{Charcount}, @code{Bytecount}, and @code{Bufpos}.
The whole point of using different types is to avoid confusion about the
use of certain variables.  Lest this effect be nullified, you need to be
careful about using the right types.

@item Always convert external data
It is extremely important to always convert external data, because
XEmacs can crash if unexpected 8bit sequences are copied to its internal
buffers literally.

This means that when a system function, such as @code{readdir}, returns
a string, you need to convert it using one of the conversion macros
described in the previous chapter, before passing it further to Lisp.
In the case of @code{readdir}, you would use the
@code{GET_C_CHARPTR_INT_FILENAME_DATA_ALLOCA} macro.

Also note that many internal functions, such as @code{make_string},
accept Bufbytes, which removes the need for them to convert the data
they receive.  This increases efficiency because that way external data
needs to be decoded only once, when it is read.  After that, it is
passed around in internal format.
@end table

@node An Example of Mule-Aware Code
@subsection An Example of Mule-Aware Code

As an example of Mule-aware code, we shall will analyze the
@code{string} function, which conses up a Lisp string from the character
arguments it receives.  Here is the definition, pasted from
@code{alloc.c}:

@example
@group
DEFUN ("string", Fstring, 0, MANY, 0, /*
Concatenate all the argument characters and make the result a string.
*/
       (int nargs, Lisp_Object *args))
@{
  Bufbyte *storage = alloca_array (Bufbyte, nargs * MAX_EMCHAR_LEN);
  Bufbyte *p = storage;

  for (; nargs; nargs--, args++)
    @{
      Lisp_Object lisp_char = *args;
      CHECK_CHAR_COERCE_INT (lisp_char);
      p += set_charptr_emchar (p, XCHAR (lisp_char));
    @}
  return make_string (storage, p - storage);
@}
@end group
@end example

Now we can analyze the source line by line.

Obviously, string will be as long as there are arguments to the
function.  This is why we allocate @code{MAX_EMCHAR_LEN} * @var{nargs}
bytes on the stack, i.e. the worst-case number of bytes for @var{nargs}
@code{Emchar}s to fit in the string.

Then, the loop checks that each element is a character, converting
integers in the process.  Like many other functions in XEmacs, this
function silently accepts integers where characters are expected, for
historical and compatibility reasons.  Unless you know what you are
doing, @code{CHECK_CHAR} will also suffice.  @code{XCHAR (lisp_char)}
extracts the @code{Emchar} from the @code{Lisp_Object}, and
@code{set_charptr_emchar} stores it to storage, increasing @code{p} in
the process.

Other instructive examples of correct coding under Mule can be found all
over the XEmacs code.  For starters, I recommend
@code{Fnormalize_menu_item_name} in @file{menubar.c}.  After you have
understood this section of the manual and studied the examples, you can
proceed writing new Mule-aware code.

@node Techniques for XEmacs Developers
@section Techniques for XEmacs Developers

To make a quantified XEmacs, do: @code{make quantmacs}.

You simply can't dump Quantified and Purified images.  Run the image
like so:  @code{quantmacs -batch -l loadup.el run-temacs @var{xemacs-args...}}.

Before you go through the trouble, are you compiling with all
debugging and error-checking off?  If not try that first.  Be warned
that while Quantify is directly responsible for quite a few
optimizations which have been made to XEmacs, doing a run which
generates results which can be acted upon is not necessarily a trivial
task.

Also, if you're still willing to do some runs make sure you configure
with the @samp{--quantify} flag.  That will keep Quantify from starting
to record data until after the loadup is completed and will shut off
recording right before it shuts down (which generates enough bogus data
to throw most results off).  It also enables three additional elisp
commands: @code{quantify-start-recording-data},
@code{quantify-stop-recording-data} and @code{quantify-clear-data}.

If you want to make XEmacs faster, target your favorite slow benchmark,
run a profiler like Quantify, @code{gprof}, or @code{tcov}, and figure
out where the cycles are going.  Specific projects:

@itemize @bullet
@item
Make the garbage collector faster.  Figure out how to write an
incremental garbage collector.
@item
Write a compiler that takes bytecode and spits out C code.
Unfortunately, you will then need a C compiler and a more fully
developed module system.
@item
Speed up redisplay.
@item
Speed up syntax highlighting.  Maybe moving some of the syntax
highlighting capabilities into C would make a difference.
@item
Implement tail recursion in Emacs Lisp (hard!).
@end itemize

Unfortunately, Emacs Lisp is slow, and is going to stay slow.  Function
calls in elisp are especially expensive.  Iterating over a long list is
going to be 30 times faster implemented in C than in Elisp.

To get started debugging XEmacs, take a look at the @file{.gdbinit} and
@file{.dbxrc} files in the @file{src} directory.
@xref{Q2.1.15 - How to Debug an XEmacs problem with a debugger,,,
xemacs-faq, XEmacs FAQ}.

After making source code changes, run @code{make check} to ensure that
you haven't introduced any regressions.  If you're feeling ambitious,
you can try to improve the test suite in @file{tests/automated}.

Here are things to know when you create a new source file:

@itemize @bullet
@item
All @file{.c} files should @code{#include <config.h>} first.  Almost all
@file{.c} files should @code{#include "lisp.h"} second.

@item
Generated header files should be included using the @code{#include <...>} syntax,
not the @code{#include "..."} syntax.  The generated headers are:

@file{config.h puresize-adjust.h sheap-adjust.h paths.h Emacs.ad.h}

The basic rule is that you should assume builds using @code{--srcdir}
and the @code{#include <...>} syntax needs to be used when the
to-be-included generated file is in a potentially different directory
@emph{at compile time}.  The non-obvious C rule is that @code{#include "..."}
means to search for the included file in the same directory as the
including file, @emph{not} in the current directory.

@item
Header files should @emph{not} include @code{<config.h>} and
@code{"lisp.h"}.  It is the responsibility of the @file{.c} files that
use it to do so.

@item
If the header uses @code{INLINE}, either directly or through
@code{DECLARE_LRECORD}, then it must be added to @file{inline.c}'s
includes.

@item
Try compiling at least once with

@example
gcc --with-mule --with-union-type --error-checking=all
@end example

@item
Did I mention that you should run the test suite?
@example
make check
@end example
@end itemize


@node A Summary of the Various XEmacs Modules, Allocation of Objects in XEmacs Lisp, Rules When Writing New C Code, Top
@chapter A Summary of the Various XEmacs Modules

  This is accurate as of XEmacs 20.0.

@menu
* Low-Level Modules::
* Basic Lisp Modules::
* Modules for Standard Editing Operations::
* Editor-Level Control Flow Modules::
* Modules for the Basic Displayable Lisp Objects::
* Modules for other Display-Related Lisp Objects::
* Modules for the Redisplay Mechanism::
* Modules for Interfacing with the File System::
* Modules for Other Aspects of the Lisp Interpreter and Object System::
* Modules for Interfacing with the Operating System::
* Modules for Interfacing with X Windows::
* Modules for Internationalization::
@end menu

@node Low-Level Modules
@section Low-Level Modules

@example
config.h
@end example

This is automatically generated from @file{config.h.in} based on the
results of configure tests and user-selected optional features and
contains preprocessor definitions specifying the nature of the
environment in which XEmacs is being compiled.



@example
paths.h
@end example

This is automatically generated from @file{paths.h.in} based on supplied
configure values, and allows for non-standard installed configurations
of the XEmacs directories.  It's currently broken, though.



@example
emacs.c
signal.c
@end example

@file{emacs.c} contains @code{main()} and other code that performs the most
basic environment initializations and handles shutting down the XEmacs
process (this includes @code{kill-emacs}, the normal way that XEmacs is
exited; @code{dump-emacs}, which is used during the build process to
write out the XEmacs executable; @code{run-emacs-from-temacs}, which can
be used to start XEmacs directly when temacs has finished loading all
the Lisp code; and emergency code to handle crashes [XEmacs tries to
auto-save all files before it crashes]).

Low-level code that directly interacts with the Unix signal mechanism,
however, is in @file{signal.c}.  Note that this code does not handle system
dependencies in interfacing to signals; that is handled using the
@file{syssignal.h} header file, described in section J below.



@example
unexaix.c
unexalpha.c
unexapollo.c
unexconvex.c
unexec.c
unexelf.c
unexelfsgi.c
unexencap.c
unexenix.c
unexfreebsd.c
unexfx2800.c
unexhp9k3.c
unexhp9k800.c
unexmips.c
unexnext.c
unexsol2.c
unexsunos4.c
@end example

These modules contain code dumping out the XEmacs executable on various
different systems. (This process is highly machine-specific and
requires intimate knowledge of the executable format and the memory map
of the process.) Only one of these modules is actually used; this is
chosen by @file{configure}.



@example
crt0.c
lastfile.c
pre-crt0.c
@end example

These modules are used in conjunction with the dump mechanism.  On some
systems, an alternative version of the C startup code (the actual code
that receives control from the operating system when the process is
started, and which calls @code{main()}) is required so that the dumping
process works properly; @file{crt0.c} provides this.

@file{pre-crt0.c} and @file{lastfile.c} should be the very first and
very last file linked, respectively. (Actually, this is not really true.
@file{lastfile.c} should be after all Emacs modules whose initialized
data should be made constant, and before all other Emacs files and all
libraries.  In particular, the allocation modules @file{gmalloc.c},
@file{alloca.c}, etc. are normally placed past @file{lastfile.c}, and
all of the files that implement Xt widget classes @emph{must} be placed
after @file{lastfile.c} because they contain various structures that
must be statically initialized and into which Xt writes at various
times.) @file{pre-crt0.c} and @file{lastfile.c} contain exported symbols
that are used to determine the start and end of XEmacs' initialized
data space when dumping.



@example
alloca.c
free-hook.c
getpagesize.h
gmalloc.c
malloc.c
mem-limits.h
ralloc.c
vm-limit.c
@end example

These handle basic C allocation of memory.  @file{alloca.c} is an emulation of
the stack allocation function @code{alloca()} on machines that lack
this. (XEmacs makes extensive use of @code{alloca()} in its code.)

@file{gmalloc.c} and @file{malloc.c} are two implementations of the standard C
functions @code{malloc()}, @code{realloc()} and @code{free()}.  They are
often used in place of the standard system-provided @code{malloc()}
because they usually provide a much faster implementation, at the
expense of additional memory use.  @file{gmalloc.c} is a newer implementation
that is much more memory-efficient for large allocations than @file{malloc.c},
and should always be preferred if it works. (At one point, @file{gmalloc.c}
didn't work on some systems where @file{malloc.c} worked; but this should be
fixed now.)

@cindex relocating allocator
@file{ralloc.c} is the @dfn{relocating allocator}.  It provides
functions similar to @code{malloc()}, @code{realloc()} and @code{free()}
that allocate memory that can be dynamically relocated in memory.  The
advantage of this is that allocated memory can be shuffled around to
place all the free memory at the end of the heap, and the heap can then
be shrunk, releasing the memory back to the operating system.  The use
of this can be controlled with the configure option @code{--rel-alloc};
if enabled, memory allocated for buffers will be relocatable, so that if
a very large file is visited and the buffer is later killed, the memory
can be released to the operating system.  (The disadvantage of this
mechanism is that it can be very slow.  On systems with the
@code{mmap()} system call, the XEmacs version of @file{ralloc.c} uses
this to move memory around without actually having to block-copy it,
which can speed things up; but it can still cause noticeable performance
degradation.)

@file{free-hook.c} contains some debugging functions for checking for invalid
arguments to @code{free()}.

@file{vm-limit.c} contains some functions that warn the user when memory is
getting low.  These are callback functions that are called by @file{gmalloc.c}
and @file{malloc.c} at appropriate times.

@file{getpagesize.h} provides a uniform interface for retrieving the size of a
page in virtual memory.  @file{mem-limits.h} provides a uniform interface for
retrieving the total amount of available virtual memory.  Both are
similar in spirit to the @file{sys*.h} files described in section J, below.



@example
blocktype.c
blocktype.h
dynarr.c
@end example

These implement a couple of basic C data types to facilitate memory
allocation.  The @code{Blocktype} type efficiently manages the
allocation of fixed-size blocks by minimizing the number of times that
@code{malloc()} and @code{free()} are called.  It allocates memory in
large chunks, subdivides the chunks into blocks of the proper size, and
returns the blocks as requested.  When blocks are freed, they are placed
onto a linked list, so they can be efficiently reused.  This data type
is not much used in XEmacs currently, because it's a fairly new
addition.

@cindex dynamic array
The @code{Dynarr} type implements a @dfn{dynamic array}, which is
similar to a standard C array but has no fixed limit on the number of
elements it can contain.  Dynamic arrays can hold elements of any type,
and when you add a new element, the array automatically resizes itself
if it isn't big enough.  Dynarrs are extensively used in the redisplay
mechanism.



@example
inline.c
@end example

This module is used in connection with inline functions (available in
some compilers).  Often, inline functions need to have a corresponding
non-inline function that does the same thing.  This module is where they
reside.  It contains no actual code, but defines some special flags that
cause inline functions defined in header files to be rendered as actual
functions.  It then includes all header files that contain any inline
function definitions, so that each one gets a real function equivalent.



@example
debug.c
debug.h
@end example

These functions provide a system for doing internal consistency checks
during code development.  This system is not currently used; instead the
simpler @code{assert()} macro is used along with the various checks
provided by the @samp{--error-check-*} configuration options.



@example
prefix-args.c
@end example

This is actually the source for a small, self-contained program
used during building.


@example
universe.h
@end example

This is not currently used.



@node Basic Lisp Modules
@section Basic Lisp Modules

@example
emacsfns.h
lisp-disunion.h
lisp-union.h
lisp.h
lrecord.h
symsinit.h
@end example

These are the basic header files for all XEmacs modules.  Each module
includes @file{lisp.h}, which brings the other header files in.
@file{lisp.h} contains the definitions of the structures and extractor
and constructor macros for the basic Lisp objects and various other
basic definitions for the Lisp environment, as well as some
general-purpose definitions (e.g. @code{min()} and @code{max()}).
@file{lisp.h} includes either @file{lisp-disunion.h} or
@file{lisp-union.h}, depending on whether @code{USE_UNION_TYPE} is
defined.  These files define the typedef of the Lisp object itself (as
described above) and the low-level macros that hide the actual
implementation of the Lisp object.  All extractor and constructor macros
for particular types of Lisp objects are defined in terms of these
low-level macros.

As a general rule, all typedefs should go into the typedefs section of
@file{lisp.h} rather than into a module-specific header file even if the
structure is defined elsewhere.  This allows function prototypes that
use the typedef to be placed into other header files.  Forward structure
declarations (i.e. a simple declaration like @code{struct foo;} where
the structure itself is defined elsewhere) should be placed into the
typedefs section as necessary.

@file{lrecord.h} contains the basic structures and macros that implement
all record-type Lisp objects---i.e. all objects whose type is a field
in their C structure, which includes all objects except the few most
basic ones.

@file{lisp.h} contains prototypes for most of the exported functions in
the various modules.  Lisp primitives defined using @code{DEFUN} that
need to be called by C code should be declared using @code{EXFUN}.
Other function prototypes should be placed either into the appropriate
section of @code{lisp.h}, or into a module-specific header file,
depending on how general-purpose the function is and whether it has
special-purpose argument types requiring definitions not in
@file{lisp.h}.)  All initialization functions are prototyped in
@file{symsinit.h}.



@example
alloc.c
pure.c
puresize.h
@end example

The large module @file{alloc.c} implements all of the basic allocation and
garbage collection for Lisp objects.  The most commonly used Lisp
objects are allocated in chunks, similar to the Blocktype data type
described above; others are allocated in individually @code{malloc()}ed
blocks.  This module provides the foundation on which all other aspects
of the Lisp environment sit, and is the first module initialized at
startup.

Note that @file{alloc.c} provides a series of generic functions that are
not dependent on any particular object type, and interfaces to
particular types of objects using a standardized interface of
type-specific methods.  This scheme is a fundamental principle of
object-oriented programming and is heavily used throughout XEmacs.  The
great advantage of this is that it allows for a clean separation of
functionality into different modules---new classes of Lisp objects, new
event interfaces, new device types, new stream interfaces, etc. can be
added transparently without affecting code anywhere else in XEmacs.
Because the different subsystems are divided into general and specific
code, adding a new subtype within a subsystem will in general not
require changes to the generic subsystem code or affect any of the other
subtypes in the subsystem; this provides a great deal of robustness to
the XEmacs code.

@cindex pure space
@file{pure.c} contains the declaration of the @dfn{purespace} array.
Pure space is a hack used to place some constant Lisp data into the code
segment of the XEmacs executable, even though the data needs to be
initialized through function calls.  (See above in section VIII for more
info about this.)  During startup, certain sorts of data is
automatically copied into pure space, and other data is copied manually
in some of the basic Lisp files by calling the function @code{purecopy},
which copies the object if possible (this only works in temacs, of
course) and returns the new object.  In particular, while temacs is
executing, the Lisp reader automatically copies all compiled-function
objects that it reads into pure space.  Since compiled-function objects
are large, are never modified, and typically comprise the majority of
the contents of a compiled-Lisp file, this works well.  While XEmacs is
running, any attempt to modify an object that resides in pure space
causes an error.  Objects in pure space are never garbage collected --
almost all of the time, they're intended to be permanent, and in any
case you can't write into pure space to set the mark bits.

@file{puresize.h} contains the declaration of the size of the pure space
array.  This depends on the optional features that are compiled in, any
extra purespace requested by the user at compile time, and certain other
factors (e.g. 64-bit machines need more pure space because their Lisp
objects are larger).  The smallest size that suffices should be used, so
that there's no wasted space.  If there's not enough pure space, you
will get an error during the build process, specifying how much more
pure space is needed.



@example
eval.c
backtrace.h
@end example

This module contains all of the functions to handle the flow of control.
This includes the mechanisms of defining functions, calling functions,
traversing stack frames, and binding variables; the control primitives
and other special forms such as @code{while}, @code{if}, @code{eval},
@code{let}, @code{and}, @code{or}, @code{progn}, etc.; handling of
non-local exits, unwind-protects, and exception handlers; entering the
debugger; methods for the subr Lisp object type; etc.  It does
@emph{not} include the @code{read} function, the @code{print} function,
or the handling of symbols and obarrays.

@file{backtrace.h} contains some structures related to stack frames and the
flow of control.



@example
lread.c
@end example

This module implements the Lisp reader and the @code{read} function,
which converts text into Lisp objects, according to the read syntax of
the objects, as described above.  This is similar to the parser that is
a part of all compilers.



@example
print.c
@end example

This module implements the Lisp print mechanism and the @code{print}
function and related functions.  This is the inverse of the Lisp reader
-- it converts Lisp objects to a printed, textual representation.
(Hopefully something that can be read back in using @code{read} to get
an equivalent object.)



@example
general.c
symbols.c
symeval.h
@end example

@file{symbols.c} implements the handling of symbols, obarrays, and
retrieving the values of symbols.  Much of the code is devoted to
handling the special @dfn{symbol-value-magic} objects that define
special types of variables---this includes buffer-local variables,
variable aliases, variables that forward into C variables, etc.  This
module is initialized extremely early (right after @file{alloc.c}),
because it is here that the basic symbols @code{t} and @code{nil} are
created, and those symbols are used everywhere throughout XEmacs.

@file{symeval.h} contains the definitions of symbol structures and the
@code{DEFVAR_LISP()} and related macros for declaring variables.



@example
data.c
floatfns.c
fns.c
@end example

These modules implement the methods and standard Lisp primitives for all
the basic Lisp object types other than symbols (which are described
above).  @file{data.c} contains all the predicates (primitives that return
whether an object is of a particular type); the integer arithmetic
functions; and the basic accessor and mutator primitives for the various
object types.  @file{fns.c} contains all the standard predicates for working
with sequences (where, abstractly speaking, a sequence is an ordered set
of objects, and can be represented by a list, string, vector, or
bit-vector); it also contains @code{equal}, perhaps on the grounds that
bulk of the operation of @code{equal} is comparing sequences.
@file{floatfns.c} contains methods and primitives for floats and floating-point
arithmetic.



@example
bytecode.c
bytecode.h
@end example

@file{bytecode.c} implements the byte-code interpreter and
compiled-function objects, and @file{bytecode.h} contains associated
structures.  Note that the byte-code @emph{compiler} is written in Lisp.




@node Modules for Standard Editing Operations
@section Modules for Standard Editing Operations

@example
buffer.c
buffer.h
bufslots.h
@end example

@file{buffer.c} implements the @dfn{buffer} Lisp object type.  This
includes functions that create and destroy buffers; retrieve buffers by
name or by other properties; manipulate lists of buffers (remember that
buffers are permanent objects and stored in various ordered lists);
retrieve or change buffer properties; etc.  It also contains the
definitions of all the built-in buffer-local variables (which can be
viewed as buffer properties).  It does @emph{not} contain code to
manipulate buffer-local variables (that's in @file{symbols.c}, described
above); or code to manipulate the text in a buffer.

@file{buffer.h} defines the structures associated with a buffer and the various
macros for retrieving text from a buffer and special buffer positions
(e.g. @code{point}, the default location for text insertion).  It also
contains macros for working with buffer positions and converting between
their representations as character offsets and as byte offsets (under
MULE, they are different, because characters can be multi-byte).  It is
one of the largest header files.

@file{bufslots.h} defines the fields in the buffer structure that correspond to
the built-in buffer-local variables.  It is its own header file because
it is included many times in @file{buffer.c}, as a way of iterating over all
the built-in buffer-local variables.



@example
insdel.c
insdel.h
@end example

@file{insdel.c} contains low-level functions for inserting and deleting text in
a buffer, keeping track of changed regions for use by redisplay, and
calling any before-change and after-change functions that may have been
registered for the buffer.  It also contains the actual functions that
convert between byte offsets and character offsets.

@file{insdel.h} contains associated headers.



@example
marker.c
@end example

This module implements the @dfn{marker} Lisp object type, which
conceptually is a pointer to a text position in a buffer that moves
around as text is inserted and deleted, so as to remain in the same
relative position.  This module doesn't actually move the markers around
-- that's handled in @file{insdel.c}.  This module just creates them and
implements the primitives for working with them.  As markers are simple
objects, this does not entail much.

Note that the standard arithmetic primitives (e.g. @code{+}) accept
markers in place of integers and automatically substitute the value of
@code{marker-position} for the marker, i.e. an integer describing the
current buffer position of the marker.



@example
extents.c
extents.h
@end example

This module implements the @dfn{extent} Lisp object type, which is like
a marker that works over a range of text rather than a single position.
Extents are also much more complex and powerful than markers and have a
more efficient (and more algorithmically complex) implementation.  The
implementation is described in detail in comments in @file{extents.c}.

The code in @file{extents.c} works closely with @file{insdel.c} so that
extents are properly moved around as text is inserted and deleted.
There is also code in @file{extents.c} that provides information needed
by the redisplay mechanism for efficient operation. (Remember that
extents can have display properties that affect [sometimes drastically,
as in the @code{invisible} property] the display of the text they
cover.)



@example
editfns.c
@end example

@file{editfns.c} contains the standard Lisp primitives for working with
a buffer's text, and calls the low-level functions in @file{insdel.c}.
It also contains primitives for working with @code{point} (the default
buffer insertion location).

@file{editfns.c} also contains functions for retrieving various
characteristics from the external environment: the current time, the
process ID of the running XEmacs process, the name of the user who ran
this XEmacs process, etc.  It's not clear why this code is in
@file{editfns.c}.



@example
callint.c
cmds.c
commands.h
@end example

@cindex interactive
These modules implement the basic @dfn{interactive} commands,
i.e. user-callable functions.  Commands, as opposed to other functions,
have special ways of getting their parameters interactively (by querying
the user), as opposed to having them passed in a normal function
invocation.  Many commands are not really meant to be called from other
Lisp functions, because they modify global state in a way that's often
undesired as part of other Lisp functions.

@file{callint.c} implements the mechanism for querying the user for
parameters and calling interactive commands.  The bulk of this module is
code that parses the interactive spec that is supplied with an
interactive command.

@file{cmds.c} implements the basic, most commonly used editing commands:
commands to move around the current buffer and insert and delete
characters.  These commands are implemented using the Lisp primitives
defined in @file{editfns.c}.

@file{commands.h} contains associated structure definitions and prototypes.



@example
regex.c
regex.h
search.c
@end example

@file{search.c} implements the Lisp primitives for searching for text in
a buffer, and some of the low-level algorithms for doing this.  In
particular, the fast fixed-string Boyer-Moore search algorithm is
implemented in @file{search.c}.  The low-level algorithms for doing
regular-expression searching, however, are implemented in @file{regex.c}
and @file{regex.h}.  These two modules are largely independent of
XEmacs, and are similar to (and based upon) the regular-expression
routines used in @file{grep} and other GNU utilities.



@example
doprnt.c
@end example

@file{doprnt.c} implements formatted-string processing, similar to
@code{printf()} command in C.



@example
undo.c
@end example

This module implements the undo mechanism for tracking buffer changes.
Most of this could be implemented in Lisp.



@node Editor-Level Control Flow Modules
@section Editor-Level Control Flow Modules

@example
event-Xt.c
event-stream.c
event-tty.c
events.c
events.h
@end example

These implement the handling of events (user input and other system
notifications).

@file{events.c} and @file{events.h} define the @dfn{event} Lisp object
type and primitives for manipulating it.

@file{event-stream.c} implements the basic functions for working with
event queues, dispatching an event by looking it up in relevant keymaps
and such, and handling timeouts; this includes the primitives
@code{next-event} and @code{dispatch-event}, as well as related
primitives such as @code{sit-for}, @code{sleep-for}, and
@code{accept-process-output}. (@file{event-stream.c} is one of the
hairiest and trickiest modules in XEmacs.  Beware!  You can easily mess
things up here.)

@file{event-Xt.c} and @file{event-tty.c} implement the low-level
interfaces onto retrieving events from Xt (the X toolkit) and from TTY's
(using @code{read()} and @code{select()}), respectively.  The event
interface enforces a clean separation between the specific code for
interfacing with the operating system and the generic code for working
with events, by defining an API of basic, low-level event methods;
@file{event-Xt.c} and @file{event-tty.c} are two different
implementations of this API.  To add support for a new operating system
(e.g. NeXTstep), one merely needs to provide another implementation of
those API functions.

Note that the choice of whether to use @file{event-Xt.c} or
@file{event-tty.c} is made at compile time!  Or at the very latest, it
is made at startup time.  @file{event-Xt.c} handles events for
@emph{both} X and TTY frames; @file{event-tty.c} is only used when X
support is not compiled into XEmacs.  The reason for this is that there
is only one event loop in XEmacs: thus, it needs to be able to receive
events from all different kinds of frames.



@example
keymap.c
keymap.h
@end example

@file{keymap.c} and @file{keymap.h} define the @dfn{keymap} Lisp object
type and associated methods and primitives. (Remember that keymaps are
objects that associate event descriptions with functions to be called to
``execute'' those events; @code{dispatch-event} looks up events in the
relevant keymaps.)



@example
keyboard.c
@end example

@file{keyboard.c} contains functions that implement the actual editor
command loop---i.e. the event loop that cyclically retrieves and
dispatches events.  This code is also rather tricky, just like
@file{event-stream.c}.



@example
macros.c
macros.h
@end example

These two modules contain the basic code for defining keyboard macros.
These functions don't actually do much; most of the code that handles keyboard
macros is mixed in with the event-handling code in @file{event-stream.c}.



@example
minibuf.c
@end example

This contains some miscellaneous code related to the minibuffer (most of
the minibuffer code was moved into Lisp by Richard Mlynarik).  This
includes the primitives for completion (although filename completion is
in @file{dired.c}), the lowest-level interface to the minibuffer (if the
command loop were cleaned up, this too could be in Lisp), and code for
dealing with the echo area (this, too, was mostly moved into Lisp, and
the only code remaining is code to call out to Lisp or provide simple
bootstrapping implementations early in temacs, before the echo-area Lisp
code is loaded).



@node Modules for the Basic Displayable Lisp Objects
@section Modules for the Basic Displayable Lisp Objects

@example
device-ns.h
device-stream.c
device-stream.h
device-tty.c
device-tty.h
device-x.c
device-x.h
device.c
device.h
@end example

These modules implement the @dfn{device} Lisp object type.  This
abstracts a particular screen or connection on which frames are
displayed.  As with Lisp objects, event interfaces, and other
subsystems, the device code is separated into a generic component that
contains a standardized interface (in the form of a set of methods) onto
particular device types.

The device subsystem defines all the methods and provides method
services for not only device operations but also for the frame, window,
menubar, scrollbar, toolbar, and other displayable-object subsystems.
The reason for this is that all of these subsystems have the same
subtypes (X, TTY, NeXTstep, Microsoft Windows, etc.) as devices do.



@example
frame-ns.h
frame-tty.c
frame-x.c
frame-x.h
frame.c
frame.h
@end example

Each device contains one or more frames in which objects (e.g. text) are
displayed.  A frame corresponds to a window in the window system;
usually this is a top-level window but it could potentially be one of a
number of overlapping child windows within a top-level window, using the
MDI (Multiple Document Interface) protocol in Microsoft Windows or a
similar scheme.

The @file{frame-*} files implement the @dfn{frame} Lisp object type and
provide the generic and device-type-specific operations on frames
(e.g. raising, lowering, resizing, moving, etc.).



@example
window.c
window.h
@end example

@cindex window (in Emacs)
@cindex pane
Each frame consists of one or more non-overlapping @dfn{windows} (better
known as @dfn{panes} in standard window-system terminology) in which a
buffer's text can be displayed.  Windows can also have scrollbars
displayed around their edges.

@file{window.c} and @file{window.h} implement the @dfn{window} Lisp
object type and provide code to manage windows.  Since windows have no
associated resources in the window system (the window system knows only
about the frame; no child windows or anything are used for XEmacs
windows), there is no device-type-specific code here; all of that code
is part of the redisplay mechanism or the code for particular object
types such as scrollbars.



@node Modules for other Display-Related Lisp Objects
@section Modules for other Display-Related Lisp Objects

@example
faces.c
faces.h
@end example



@example
bitmaps.h
glyphs-ns.h
glyphs-x.c
glyphs-x.h
glyphs.c
glyphs.h
@end example



@example
objects-ns.h
objects-tty.c
objects-tty.h
objects-x.c
objects-x.h
objects.c
objects.h
@end example



@example
menubar-x.c
menubar.c
@end example



@example
scrollbar-x.c
scrollbar-x.h
scrollbar.c
scrollbar.h
@end example



@example
toolbar-x.c
toolbar.c
toolbar.h
@end example



@example
font-lock.c
@end example

This file provides C support for syntax highlighting---i.e.
highlighting different syntactic constructs of a source file in
different colors, for easy reading.  The C support is provided so that
this is fast.



@example
dgif_lib.c
gif_err.c
gif_lib.h
gifalloc.c
@end example

These modules decode GIF-format image files, for use with glyphs.



@node Modules for the Redisplay Mechanism
@section Modules for the Redisplay Mechanism

@example
redisplay-output.c
redisplay-tty.c
redisplay-x.c
redisplay.c
redisplay.h
@end example

These files provide the redisplay mechanism.  As with many other
subsystems in XEmacs, there is a clean separation between the general
and device-specific support.

@file{redisplay.c} contains the bulk of the redisplay engine.  These
functions update the redisplay structures (which describe how the screen
is to appear) to reflect any changes made to the state of any
displayable objects (buffer, frame, window, etc.) since the last time
that redisplay was called.  These functions are highly optimized to
avoid doing more work than necessary (since redisplay is called
extremely often and is potentially a huge time sink), and depend heavily
on notifications from the objects themselves that changes have occurred,
so that redisplay doesn't explicitly have to check each possible object.
The redisplay mechanism also contains a great deal of caching to further
speed things up; some of this caching is contained within the various
displayable objects.

@file{redisplay-output.c} goes through the redisplay structures and converts
them into calls to device-specific methods to actually output the screen
changes.

@file{redisplay-x.c} and @file{redisplay-tty.c} are two implementations
of these redisplay output methods, for X frames and TTY frames,
respectively.



@example
indent.c
@end example

This module contains various functions and Lisp primitives for
converting between buffer positions and screen positions.  These
functions call the redisplay mechanism to do most of the work, and then
examine the redisplay structures to get the necessary information.  This
module needs work.



@example
termcap.c
terminfo.c
tparam.c
@end example

These files contain functions for working with the termcap (BSD-style)
and terminfo (System V style) databases of terminal capabilities and
escape sequences, used when XEmacs is displaying in a TTY.



@example
cm.c
cm.h
@end example

These files provide some miscellaneous TTY-output functions and should
probably be merged into @file{redisplay-tty.c}.



@node Modules for Interfacing with the File System
@section Modules for Interfacing with the File System

@example
lstream.c
lstream.h
@end example

These modules implement the @dfn{stream} Lisp object type.  This is an
internal-only Lisp object that implements a generic buffering stream.
The idea is to provide a uniform interface onto all sources and sinks of
data, including file descriptors, stdio streams, chunks of memory, Lisp
buffers, Lisp strings, etc.  That way, I/O functions can be written to
the stream interface and can transparently handle all possible sources
and sinks.  (For example, the @code{read} function can read data from a
file, a string, a buffer, or even a function that is called repeatedly
to return data, without worrying about where the data is coming from or
what-size chunks it is returned in.)

@cindex lstream
Note that in the C code, streams are called @dfn{lstreams} (for ``Lisp
streams'') to distinguish them from other kinds of streams, e.g. stdio
streams and C++ I/O streams.

Similar to other subsystems in XEmacs, lstreams are separated into
generic functions and a set of methods for the different types of
lstreams.  @file{lstream.c} provides implementations of many different
types of streams; others are provided, e.g., in @file{mule-coding.c}.



@example
fileio.c
@end example

This implements the basic primitives for interfacing with the file
system.  This includes primitives for reading files into buffers,
writing buffers into files, checking for the presence or accessibility
of files, canonicalizing file names, etc.  Note that these primitives
are usually not invoked directly by the user: There is a great deal of
higher-level Lisp code that implements the user commands such as
@code{find-file} and @code{save-buffer}.  This is similar to the
distinction between the lower-level primitives in @file{editfns.c} and
the higher-level user commands in @file{commands.c} and
@file{simple.el}.



@example
filelock.c
@end example

This file provides functions for detecting clashes between different
processes (e.g. XEmacs and some external process, or two different
XEmacs processes) modifying the same file.  (XEmacs can optionally use
the @file{lock/} subdirectory to provide a form of ``locking'' between
different XEmacs processes.)  This module is also used by the low-level
functions in @file{insdel.c} to ensure that, if the first modification
is being made to a buffer whose corresponding file has been externally
modified, the user is made aware of this so that the buffer can be
synched up with the external changes if necessary.


@example
filemode.c
@end example

This file provides some miscellaneous functions that construct a
@samp{rwxr-xr-x}-type permissions string (as might appear in an
@file{ls}-style directory listing) given the information returned by the
@code{stat()} system call.



@example
dired.c
ndir.h
@end example

These files implement the XEmacs interface to directory searching.  This
includes a number of primitives for determining the files in a directory
and for doing filename completion. (Remember that generic completion is
handled by a different mechanism, in @file{minibuf.c}.)

@file{ndir.h} is a header file used for the directory-searching
emulation functions provided in @file{sysdep.c} (see section J below),
for systems that don't provide any directory-searching functions. (On
those systems, directories can be read directly as files, and parsed.)



@example
realpath.c
@end example

This file provides an implementation of the @code{realpath()} function
for expanding symbolic links, on systems that don't implement it or have
a broken implementation.



@node Modules for Other Aspects of the Lisp Interpreter and Object System
@section Modules for Other Aspects of the Lisp Interpreter and Object System

@example
elhash.c
elhash.h
hash.c
hash.h
@end example

These files provide two implementations of hash tables.  Files
@file{hash.c} and @file{hash.h} provide a generic C implementation of
hash tables which can stand independently of XEmacs.  Files
@file{elhash.c} and @file{elhash.h} provide a separate implementation of
hash tables that can store only Lisp objects, and knows about Lispy
things like garbage collection, and implement the @dfn{hash-table} Lisp
object type.


@example
specifier.c
specifier.h
@end example

This module implements the @dfn{specifier} Lisp object type.  This is
primarily used for displayable properties, and allows for values that
are specific to a particular buffer, window, frame, device, or device
class, as well as a default value existing.  This is used, for example,
to control the height of the horizontal scrollbar or the appearance of
the @code{default}, @code{bold}, or other faces.  The specifier object
consists of a number of specifications, each of which maps from a
buffer, window, etc. to a value.  The function @code{specifier-instance}
looks up a value given a window (from which a buffer, frame, and device
can be derived).


@example
chartab.c
chartab.h
casetab.c
@end example

@file{chartab.c} and @file{chartab.h} implement the @dfn{char table}
Lisp object type, which maps from characters or certain sorts of
character ranges to Lisp objects.  The implementation of this object
type is optimized for the internal representation of characters.  Char
tables come in different types, which affect the allowed object types to
which a character can be mapped and also dictate certain other
properties of the char table.

@cindex case table
@file{casetab.c} implements one sort of char table, the @dfn{case
table}, which maps characters to other characters of possibly different
case.  These are used by XEmacs to implement case-changing primitives
and to do case-insensitive searching.



@example
syntax.c
syntax.h
@end example

@cindex scanner
This module implements @dfn{syntax tables}, another sort of char table
that maps characters into syntax classes that define the syntax of these
characters (e.g. a parenthesis belongs to a class of @samp{open}
characters that have corresponding @samp{close} characters and can be
nested).  This module also implements the Lisp @dfn{scanner}, a set of
primitives for scanning over text based on syntax tables.  This is used,
for example, to find the matching parenthesis in a command such as
@code{forward-sexp}, and by @file{font-lock.c} to locate quoted strings,
comments, etc.



@example
casefiddle.c
@end example

This module implements various Lisp primitives for upcasing, downcasing
and capitalizing strings or regions of buffers.



@example
rangetab.c
@end example

This module implements the @dfn{range table} Lisp object type, which
provides for a mapping from ranges of integers to arbitrary Lisp
objects.



@example
opaque.c
opaque.h
@end example

This module implements the @dfn{opaque} Lisp object type, an
internal-only Lisp object that encapsulates an arbitrary block of memory
so that it can be managed by the Lisp allocation system.  To create an
opaque object, you call @code{make_opaque()}, passing a pointer to a
block of memory.  An object is created that is big enough to hold the
memory, which is copied into the object's storage.  The object will then
stick around as long as you keep pointers to it, after which it will be
automatically reclaimed.

@cindex mark method
Opaque objects can also have an arbitrary @dfn{mark method} associated
with them, in case the block of memory contains other Lisp objects that
need to be marked for garbage-collection purposes. (If you need other
object methods, such as a finalize method, you should just go ahead and
create a new Lisp object type---it's not hard.)



@example
abbrev.c
@end example

This function provides a few primitives for doing dynamic abbreviation
expansion.  In XEmacs, most of the code for this has been moved into
Lisp.  Some C code remains for speed and because the primitive
@code{self-insert-command} (which is executed for all self-inserting
characters) hooks into the abbrev mechanism. (@code{self-insert-command}
is itself in C only for speed.)



@example
doc.c
@end example

This function provides primitives for retrieving the documentation
strings of functions and variables.  These documentation strings contain
certain special markers that get dynamically expanded (e.g. a
reverse-lookup is performed on some named functions to retrieve their
current key bindings).  Some documentation strings (in particular, for
the built-in primitives and pre-loaded Lisp functions) are stored
externally in a file @file{DOC} in the @file{lib-src/} directory and
need to be fetched from that file. (Part of the build stage involves
building this file, and another part involves constructing an index for
this file and embedding it into the executable, so that the functions in
@file{doc.c} do not have to search the entire @file{DOC} file to find
the appropriate documentation string.)



@example
md5.c
@end example

This function provides a Lisp primitive that implements the MD5 secure
hashing scheme, used to create a large hash value of a string of data such that
the data cannot be derived from the hash value.  This is used for
various security applications on the Internet.




@node Modules for Interfacing with the Operating System
@section Modules for Interfacing with the Operating System

@example
callproc.c
process.c
process.h
@end example

These modules allow XEmacs to spawn and communicate with subprocesses
and network connections.

@cindex synchronous subprocesses
@cindex subprocesses, synchronous
  @file{callproc.c} implements (through the @code{call-process}
primitive) what are called @dfn{synchronous subprocesses}.  This means
that XEmacs runs a program, waits till it's done, and retrieves its
output.  A typical example might be calling the @file{ls} program to get
a directory listing.

@cindex asynchronous subprocesses
@cindex subprocesses, asynchronous
  @file{process.c} and @file{process.h} implement @dfn{asynchronous
subprocesses}.  This means that XEmacs starts a program and then
continues normally, not waiting for the process to finish.  Data can be
sent to the process or retrieved from it as it's running.  This is used
for the @code{shell} command (which provides a front end onto a shell
program such as @file{csh}), the mail and news readers implemented in
XEmacs, etc.  The result of calling @code{start-process} to start a
subprocess is a process object, a particular kind of object used to
communicate with the subprocess.  You can send data to the process by
passing the process object and the data to @code{send-process}, and you
can specify what happens to data retrieved from the process by setting
properties of the process object. (When the process sends data, XEmacs
receives a process event, which says that there is data ready.  When
@code{dispatch-event} is called on this event, it reads the data from
the process and does something with it, as specified by the process
object's properties.  Typically, this means inserting the data into a
buffer or calling a function.) Another property of the process object is
called the @dfn{sentinel}, which is a function that is called when the
process terminates.

@cindex network connections
  Process objects are also used for network connections (connections to a
process running on another machine).  Network connections are started
with @code{open-network-stream} but otherwise work just like
subprocesses.



@example
sysdep.c
sysdep.h
@end example

  These modules implement most of the low-level, messy operating-system
interface code.  This includes various device control (ioctl) operations
for file descriptors, TTY's, pseudo-terminals, etc. (usually this stuff
is fairly system-dependent; thus the name of this module), and emulation
of standard library functions and system calls on systems that don't
provide them or have broken versions.



@example
sysdir.h
sysfile.h
sysfloat.h
sysproc.h
syspwd.h
syssignal.h
systime.h
systty.h
syswait.h
@end example

These header files provide consistent interfaces onto system-dependent
header files and system calls.  The idea is that, instead of including a
standard header file like @file{<sys/param.h>} (which may or may not
exist on various systems) or having to worry about whether all system
provide a particular preprocessor constant, or having to deal with the
four different paradigms for manipulating signals, you just include the
appropriate @file{sys*.h} header file, which includes all the right
system header files, defines and missing preprocessor constants,
provides a uniform interface onto system calls, etc.

@file{sysdir.h} provides a uniform interface onto directory-querying
functions. (In some cases, this is in conjunction with emulation
functions in @file{sysdep.c}.)

@file{sysfile.h} includes all the necessary header files for standard
system calls (e.g. @code{read()}), ensures that all necessary
@code{open()} and @code{stat()} preprocessor constants are defined, and
possibly (usually) substitutes sugared versions of @code{read()},
@code{write()}, etc. that automatically restart interrupted I/O
operations.

@file{sysfloat.h} includes the necessary header files for floating-point
operations.

@file{sysproc.h} includes the necessary header files for calling
@code{select()}, @code{fork()}, @code{execve()}, socket operations, and
the like, and ensures that the @code{FD_*()} macros for descriptor-set
manipulations are available.

@file{syspwd.h} includes the necessary header files for obtaining
information from @file{/etc/passwd} (the functions are emulated under
VMS).

@file{syssignal.h} includes the necessary header files for
signal-handling and provides a uniform interface onto the different
signal-handling and signal-blocking paradigms.

@file{systime.h} includes the necessary header files and provides
uniform interfaces for retrieving the time of day, setting file
access/modification times, getting the amount of time used by the XEmacs
process, etc.

@file{systty.h} buffers against the infinitude of different ways of
controlling TTY's.

@file{syswait.h} provides a uniform way of retrieving the exit status
from a @code{wait()}ed-on process (some systems use a union, others use
an int).



@example
hpplay.c
libsst.c
libsst.h
libst.h
linuxplay.c
nas.c
sgiplay.c
sound.c
sunplay.c
@end example

These files implement the ability to play various sounds on some types
of computers.  You have to configure your XEmacs with sound support in
order to get this capability.

@file{sound.c} provides the generic interface.  It implements various
Lisp primitives and variables that let you specify which sounds should
be played in certain conditions. (The conditions are identified by
symbols, which are passed to @code{ding} to make a sound.  Various
standard functions call this function at certain times; if sound support
does not exist, a simple beep results.

@cindex native sound
@cindex sound, native
@file{sgiplay.c}, @file{sunplay.c}, @file{hpplay.c}, and
@file{linuxplay.c} interface to the machine's speaker for various
different kind of machines.  This is called @dfn{native} sound.

@cindex sound, network
@cindex network sound
@cindex NAS
@file{nas.c} interfaces to a computer somewhere else on the network
using the NAS (Network Audio Server) protocol, playing sounds on that
machine.  This allows you to run XEmacs on a remote machine, with its
display set to your local machine, and have the sounds be made on your
local machine, provided that you have a NAS server running on your local
machine.

@file{libsst.c}, @file{libsst.h}, and @file{libst.h} provide some
additional functions for playing sound on a Sun SPARC but are not
currently in use.



@example
tooltalk.c
tooltalk.h
@end example

These two modules implement an interface to the ToolTalk protocol, which
is an interprocess communication protocol implemented on some versions
of Unix.  ToolTalk is a high-level protocol that allows processes to
register themselves as providers of particular services; other processes
can then request a service without knowing or caring exactly who is
providing the service.  It is similar in spirit to the DDE protocol
provided under Microsoft Windows.  ToolTalk is a part of the new CDE
(Common Desktop Environment) specification and is used to connect the
parts of the SPARCWorks development environment.



@example
getloadavg.c
@end example

This module provides the ability to retrieve the system's current load
average. (The way to do this is highly system-specific, unfortunately,
and requires a lot of special-case code.)



@example
sunpro.c
@end example

This module provides a small amount of code used internally at Sun to
keep statistics on the usage of XEmacs.



@example
broken-sun.h
strcmp.c
strcpy.c
sunOS-fix.c
@end example

These files provide replacement functions and prototypes to fix numerous
bugs in early releases of SunOS 4.1.



@example
hftctl.c
@end example

This module provides some terminal-control code necessary on versions of
AIX prior to 4.1.



@example
msdos.c
msdos.h
@end example

These modules are used for MS-DOS support, which does not work in
XEmacs.



@node Modules for Interfacing with X Windows
@section Modules for Interfacing with X Windows

@example
Emacs.ad.h
@end example

A file generated from @file{Emacs.ad}, which contains XEmacs-supplied
fallback resources (so that XEmacs has pretty defaults).



@example
EmacsFrame.c
EmacsFrame.h
EmacsFrameP.h
@end example

These modules implement an Xt widget class that encapsulates a frame.
This is for ease in integrating with Xt.  The EmacsFrame widget covers
the entire X window except for the menubar; the scrollbars are
positioned on top of the EmacsFrame widget.

@strong{Warning:} Abandon hope, all ye who enter here.  This code took
an ungodly amount of time to get right, and is likely to fall apart
mercilessly at the slightest change.  Such is life under Xt.



@example
EmacsManager.c
EmacsManager.h
EmacsManagerP.h
@end example

These modules implement a simple Xt manager (i.e. composite) widget
class that simply lets its children set whatever geometry they want.
It's amazing that Xt doesn't provide this standardly, but on second
thought, it makes sense, considering how amazingly broken Xt is.


@example
EmacsShell-sub.c
EmacsShell.c
EmacsShell.h
EmacsShellP.h
@end example

These modules implement two Xt widget classes that are subclasses of
the TopLevelShell and TransientShell classes.  This is necessary to deal
with more brokenness that Xt has sadistically thrust onto the backs of
developers.



@example
xgccache.c
xgccache.h
@end example

These modules provide functions for maintenance and caching of GC's
(graphics contexts) under the X Window System.  This code is junky and
needs to be rewritten.



@example
xselect.c
@end example

@cindex selections
  This module provides an interface to the X Window System's concept of
@dfn{selections}, the standard way for X applications to communicate
with each other.



@example
xintrinsic.h
xintrinsicp.h
xmmanagerp.h
xmprimitivep.h
@end example

These header files are similar in spirit to the @file{sys*.h} files and buffer
against different implementations of Xt and Motif.

@itemize @bullet
@item
@file{xintrinsic.h} should be included in place of @file{<Intrinsic.h>}.
@item
@file{xintrinsicp.h} should be included in place of @file{<IntrinsicP.h>}.
@item
@file{xmmanagerp.h} should be included in place of @file{<XmManagerP.h>}.
@item
@file{xmprimitivep.h} should be included in place of @file{<XmPrimitiveP.h>}.
@end itemize



@example
xmu.c
xmu.h
@end example

These files provide an emulation of the Xmu library for those systems
(i.e. HPUX) that don't provide it as a standard part of X.



@example
ExternalClient-Xlib.c
ExternalClient.c
ExternalClient.h
ExternalClientP.h
ExternalShell.c
ExternalShell.h
ExternalShellP.h
extw-Xlib.c
extw-Xlib.h
extw-Xt.c
extw-Xt.h
@end example

@cindex external widget
  These files provide the @dfn{external widget} interface, which allows an
XEmacs frame to appear as a widget in another application.  To do this,
you have to configure with @samp{--external-widget}.

@file{ExternalShell*} provides the server (XEmacs) side of the
connection.

@file{ExternalClient*} provides the client (other application) side of
the connection.  These files are not compiled into XEmacs but are
compiled into libraries that are then linked into your application.

@file{extw-*} is common code that is used for both the client and server.

Don't touch this code; something is liable to break if you do.



@node Modules for Internationalization
@section Modules for Internationalization

@example
mule-canna.c
mule-ccl.c
mule-charset.c
mule-charset.h
mule-coding.c
mule-coding.h
mule-mcpath.c
mule-mcpath.h
mule-wnnfns.c
mule.c
@end example

These files implement the MULE (Asian-language) support.  Note that MULE
actually provides a general interface for all sorts of languages, not
just Asian languages (although they are generally the most complicated
to support).  This code is still in beta.

@file{mule-charset.*} and @file{mule-coding.*} provide the heart of the
XEmacs MULE support.  @file{mule-charset.*} implements the @dfn{charset}
Lisp object type, which encapsulates a character set (an ordered one- or
two-dimensional set of characters, such as US ASCII or JISX0208 Japanese
Kanji).

@file{mule-coding.*} implements the @dfn{coding-system} Lisp object
type, which encapsulates a method of converting between different
encodings.  An encoding is a representation of a stream of characters,
possibly from multiple character sets, using a stream of bytes or words,
and defines (e.g.) which escape sequences are used to specify particular
character sets, how the indices for a character are converted into bytes
(sometimes this involves setting the high bit; sometimes complicated
rearranging of the values takes place, as in the Shift-JIS encoding),
etc.

@file{mule-ccl.c} provides the CCL (Code Conversion Language)
interpreter.  CCL is similar in spirit to Lisp byte code and is used to
implement converters for custom encodings.

@file{mule-canna.c} and @file{mule-wnnfns.c} implement interfaces to
external programs used to implement the Canna and WNN input methods,
respectively.  This is currently in beta.

@file{mule-mcpath.c} provides some functions to allow for pathnames
containing extended characters.  This code is fragmentary, obsolete, and
completely non-working.  Instead, @var{pathname-coding-system} is used
to specify conversions of names of files and directories.  The standard
C I/O functions like @samp{open()} are wrapped so that conversion occurs
automatically.

@file{mule.c} provides a few miscellaneous things that should probably
be elsewhere.



@example
intl.c
@end example

This provides some miscellaneous internationalization code for
implementing message translation and interfacing to the Ximp input
method.  None of this code is currently working.



@example
iso-wide.h
@end example

This contains leftover code from an earlier implementation of
Asian-language support, and is not currently used.




@node Allocation of Objects in XEmacs Lisp, Events and the Event Loop, A Summary of the Various XEmacs Modules, Top
@chapter Allocation of Objects in XEmacs Lisp

@menu
* Introduction to Allocation::
* Garbage Collection::
* GCPROing::
* Garbage Collection - Step by Step::
* Integers and Characters::
* Allocation from Frob Blocks::
* lrecords::
* Low-level allocation::
* Pure Space::
* Cons::
* Vector::
* Bit Vector::
* Symbol::
* Marker::
* String::
* Compiled Function::
@end menu

@node Introduction to Allocation
@section Introduction to Allocation

  Emacs Lisp, like all Lisps, has garbage collection.  This means that
the programmer never has to explicitly free (destroy) an object; it
happens automatically when the object becomes inaccessible.  Most
experts agree that garbage collection is a necessity in a modern,
high-level language.  Its omission from C stems from the fact that C was
originally designed to be a nice abstract layer on top of assembly
language, for writing kernels and basic system utilities rather than
large applications.

  Lisp objects can be created by any of a number of Lisp primitives.
Most object types have one or a small number of basic primitives
for creating objects.  For conses, the basic primitive is @code{cons};
for vectors, the primitives are @code{make-vector} and @code{vector}; for
symbols, the primitives are @code{make-symbol} and @code{intern}; etc.
Some Lisp objects, especially those that are primarily used internally,
have no corresponding Lisp primitives.  Every Lisp object, though,
has at least one C primitive for creating it.

  Recall from section (VII) that a Lisp object, as stored in a 32-bit
or 64-bit word, has a mark bit, a few tag bits, and a ``value'' that
occupies the remainder of the bits.  We can separate the different
Lisp object types into four broad categories:

@itemize @bullet
@item
(a) Those for whom the value directly represents the contents of the
Lisp object.  Only two types are in this category: integers and
characters.  No special allocation or garbage collection is necessary
for such objects.  Lisp objects of these types do not need to be
@code{GCPRO}ed.
@end itemize

  In the remaining three categories, the value is a pointer to a
structure.

@itemize @bullet
@item
@cindex frob block
(b) Those for whom the tag directly specifies the type.  Recall that
there are only three tag bits; this means that at most five types can be
specified this way.  The most commonly-used types are stored in this
format; this includes conses, strings, vectors, and sometimes symbols.
With the exception of vectors, objects in this category are allocated in
@dfn{frob blocks}, i.e. large blocks of memory that are subdivided into
individual objects.  This saves a lot on malloc overhead, since there
are typically quite a lot of these objects around, and the objects are
small.  (A cons, for example, occupies 8 bytes on 32-bit machines---4
bytes for each of the two objects it contains.) Vectors are individually
@code{malloc()}ed since they are of variable size.  (It would be
possible, and desirable, to allocate vectors of certain small sizes out
of frob blocks, but it isn't currently done.) Strings are handled
specially: Each string is allocated in two parts, a fixed size structure
containing a length and a data pointer, and the actual data of the
string.  The former structure is allocated in frob blocks as usual, and
the latter data is stored in @dfn{string chars blocks} and is relocated
during garbage collection to eliminate holes.
@end itemize

  In the remaining two categories, the type is stored in the object
itself.  The tag for all such objects is the generic @dfn{lrecord}
(Lisp_Record) tag.  The first four bytes (or eight, for 64-bit machines)
of the object's structure are a pointer to a structure that describes
the object's type, which includes method pointers and a pointer to a
string naming the type.  Note that it's possible to save some space by
using a one- or two-byte tag, rather than a four- or eight-byte pointer
to store the type, but it's not clear it's worth making the change.

@itemize @bullet
@item
(c) Those lrecords that are allocated in frob blocks (see above).  This
includes the objects that are most common and relatively small, and
includes floats, compiled functions, symbols (when not in category (b)),
extents, events, and markers.  With the cleanup of frob blocks done in
19.12, it's not terribly hard to add more objects to this category, but
it's a bit trickier than adding an object type to type (d) (esp. if the
object needs a finalization method), and is not likely to save much
space unless the object is small and there are many of them. (In fact,
if there are very few of them, it might actually waste space.)
@item
(d) Those lrecords that are individually @code{malloc()}ed.  These are
called @dfn{lcrecords}.  All other types are in this category.  Adding a
new type to this category is comparatively easy, and all types added
since 19.8 (when the current allocation scheme was devised, by Richard
Mlynarik), with the exception of the character type, have been in this
category.
@end itemize

  Note that bit vectors are a bit of a special case.  They are
simple lrecords as in category (c), but are individually @code{malloc()}ed
like vectors.  You can basically view them as exactly like vectors
except that their type is stored in lrecord fashion rather than
in directly-tagged fashion.

  Note that FSF Emacs redesigned their object system in 19.29 to follow
a similar scheme.  However, given RMS's expressed dislike for data
abstraction, the FSF scheme is not nearly as clean or as easy to
extend. (FSF calls items of type (c) @code{Lisp_Misc} and items of type
(d) @code{Lisp_Vectorlike}, with separate tags for each, although
@code{Lisp_Vectorlike} is also used for vectors.)

@node Garbage Collection
@section Garbage Collection
@cindex garbage collection

@cindex mark and sweep
  Garbage collection is simple in theory but tricky to implement.
Emacs Lisp uses the oldest garbage collection method, called
@dfn{mark and sweep}.  Garbage collection begins by starting with
all accessible locations (i.e. all variables and other slots where
Lisp objects might occur) and recursively traversing all objects
accessible from those slots, marking each one that is found.
We then go through all of memory and free each object that is
not marked, and unmarking each object that is marked.  Note
that ``all of memory'' means all currently allocated objects.
Traversing all these objects means traversing all frob blocks,
all vectors (which are chained in one big list), and all
lcrecords (which are likewise chained).

  Note that, when an object is marked, the mark has to occur
inside of the object's structure, rather than in the 32-bit
@code{Lisp_Object} holding the object's pointer; i.e. you can't just
set the pointer's mark bit.  This is because there may be many
pointers to the same object.  This means that the method of
marking an object can differ depending on the type.  The
different marking methods are approximately as follows:

@enumerate
@item
For conses, the mark bit of the car is set.
@item
For strings, the mark bit of the string's plist is set.
@item
For symbols when not lrecords, the mark bit of the
symbol's plist is set.
@item
For vectors, the length is negated after adding 1.
@item
For lrecords, the pointer to the structure describing
the type is changed (see below).
@item
Integers and characters do not need to be marked, since
no allocation occurs for them.
@end enumerate

  The details of this are in the @code{mark_object()} function.

  Note that any code that operates during garbage collection has
to be especially careful because of the fact that some objects
may be marked and as such may not look like they normally do.
In particular:

@itemize @bullet
Some object pointers may have their mark bit set.  This will make
@code{FOOBARP()} predicates fail.  Use @code{GC_FOOBARP()} to deal with
this.
@item
Even if you clear the mark bit, @code{FOOBARP()} will still fail
for lrecords because the implementation pointer has been
changed (see below).  @code{GC_FOOBARP()} will correctly deal with
this.
@item
Vectors have their size field munged, so anything that
looks at this field will fail.
@item
Note that @code{XFOOBAR()} macros @emph{will} work correctly on object
pointers with their mark bit set, because the logical shift operations
that remove the tag also remove the mark bit.
@end itemize

  Finally, note that garbage collection can be invoked explicitly
by calling @code{garbage-collect} but is also called automatically
by @code{eval}, once a certain amount of memory has been allocated
since the last garbage collection (according to @code{gc-cons-threshold}).

@node GCPROing
@section @code{GCPRO}ing

@code{GCPRO}ing is one of the ugliest and trickiest parts of Emacs
internals.  The basic idea is that whenever garbage collection
occurs, all in-use objects must be reachable somehow or
other from one of the roots of accessibility.  The roots
of accessibility are:

@enumerate
@item
All objects that have been @code{staticpro()}d.  This is used for
any global C variables that hold Lisp objects.  A call to
@code{staticpro()} happens implicitly as a result of any symbols
declared with @code{defsymbol()} and any variables declared with
@code{DEFVAR_FOO()}.  You need to explicitly call @code{staticpro()}
(in the @code{vars_of_foo()} method of a module) for other global
C variables holding Lisp objects. (This typically includes
internal lists and such things.)

Note that @code{obarray} is one of the @code{staticpro()}d things.
Therefore, all functions and variables get marked through this.
@item
Any shadowed bindings that are sitting on the @code{specpdl} stack.
@item
Any objects sitting in currently active (Lisp) stack frames,
catches, and condition cases.
@item
A couple of special-case places where active objects are
located.
@item
Anything currently marked with @code{GCPRO}.
@end enumerate

  Marking with @code{GCPRO} is necessary because some C functions (quite
a lot, in fact), allocate objects during their operation.  Quite
frequently, there will be no other pointer to the object while the
function is running, and if a garbage collection occurs and the object
needs to be referenced again, bad things will happen.  The solution is
to mark those objects with @code{GCPRO}.  Unfortunately this is easy to
forget, and there is basically no way around this problem.  Here are
some rules, though:

@enumerate
@item
For every @code{GCPRO@var{n}}, there have to be declarations of
@code{struct gcpro gcpro1, gcpro2}, etc.

@item
You @emph{must} @code{UNGCPRO} anything that's @code{GCPRO}ed, and you
@emph{must not} @code{UNGCPRO} if you haven't @code{GCPRO}ed.  Getting
either of these wrong will lead to crashes, often in completely random
places unrelated to where the problem lies.

@item
The way this actually works is that all currently active @code{GCPRO}s
are chained through the @code{struct gcpro} local variables, with the
variable @samp{gcprolist} pointing to the head of the list and the nth
local @code{gcpro} variable pointing to the first @code{gcpro} variable
in the next enclosing stack frame.  Each @code{GCPRO}ed thing is an
lvalue, and the @code{struct gcpro} local variable contains a pointer to
this lvalue.  This is why things will mess up badly if you don't pair up
the @code{GCPRO}s and @code{UNGCPRO}s---you will end up with
@code{gcprolist}s containing pointers to @code{struct gcpro}s or local
@code{Lisp_Object} variables in no-longer-active stack frames.

@item
It is actually possible for a single @code{struct gcpro} to
protect a contiguous array of any number of values, rather than
just a single lvalue.  To effect this, call @code{GCPRO@var{n}} as usual on
the first object in the array and then set @code{gcpro@var{n}.nvars}.

@item
@strong{Strings are relocated.}  What this means in practice is that the
pointer obtained using @code{XSTRING_DATA()} is liable to change at any
time, and you should never keep it around past any function call, or
pass it as an argument to any function that might cause a garbage
collection.  This is why a number of functions accept either a
``non-relocatable'' @code{char *} pointer or a relocatable Lisp string,
and only access the Lisp string's data at the very last minute.  In some
cases, you may end up having to @code{alloca()} some space and copy the
string's data into it.

@item
By convention, if you have to nest @code{GCPRO}'s, use @code{NGCPRO@var{n}}
(along with @code{struct gcpro ngcpro1, ngcpro2}, etc.), @code{NNGCPRO@var{n}},
etc.  This avoids compiler warnings about shadowed locals.

@item
It is @emph{always} better to err on the side of extra @code{GCPRO}s
rather than too few.  The extra cycles spent on this are
almost never going to make a whit of difference in the
speed of anything.

@item
The general rule to follow is that caller, not callee, @code{GCPRO}s.
That is, you should not have to explicitly @code{GCPRO} any Lisp objects
that are passed in as parameters.

One exception from this rule is if you ever plan to change the parameter
value, and store a new object in it.  In that case, you @emph{must}
@code{GCPRO} the parameter, because otherwise the new object will not be
protected.

So, if you create any Lisp objects (remember, this happens in all sorts
of circumstances, e.g. with @code{Fcons()}, etc.), you are responsible
for @code{GCPRO}ing them, unless you are @emph{absolutely sure} that
there's no possibility that a garbage-collection can occur while you
need to use the object.  Even then, consider @code{GCPRO}ing.

@item
A garbage collection can occur whenever anything calls @code{Feval}, or
whenever a QUIT can occur where execution can continue past
this. (Remember, this is almost anywhere.)

@item
If you have the @emph{least smidgeon of doubt} about whether
you need to @code{GCPRO}, you should @code{GCPRO}.

@item
Beware of @code{GCPRO}ing something that is uninitialized.  If you have
any shade of doubt about this, initialize all your variables to @code{Qnil}.

@item
Be careful of traps, like calling @code{Fcons()} in the argument to
another function.  By the ``caller protects'' law, you should be
@code{GCPRO}ing the newly-created cons, but you aren't.  A certain
number of functions that are commonly called on freshly created stuff
(e.g. @code{nconc2()}, @code{Fsignal()}), break the ``caller protects''
law and go ahead and @code{GCPRO} their arguments so as to simplify
things, but make sure and check if it's OK whenever doing something like
this.

@item
Once again, remember to @code{GCPRO}!  Bugs resulting from insufficient
@code{GCPRO}ing are intermittent and extremely difficult to track down,
often showing up in crashes inside of @code{garbage-collect} or in
weirdly corrupted objects or even in incorrect values in a totally
different section of code.
@end enumerate

@cindex garbage collection, conservative
@cindex conservative garbage collection
  Given the extremely error-prone nature of the @code{GCPRO} scheme, and
the difficulties in tracking down, it should be considered a deficiency
in the XEmacs code.  A solution to this problem would involve
implementing so-called @dfn{conservative} garbage collection for the C
stack.  That involves looking through all of stack memory and treating
anything that looks like a reference to an object as a reference.  This
will result in a few objects not getting collected when they should, but
it obviates the need for @code{GCPRO}ing, and allows garbage collection
to happen at any point at all, such as during object allocation.

@node Garbage Collection - Step by Step
@section Garbage Collection - Step by Step
@cindex garbage collection step by step

@menu
* Invocation::
* garbage_collect_1::
* mark_object::
* gc_sweep::
* sweep_lcrecords_1::
* compact_string_chars::
* sweep_strings::
* sweep_bit_vectors_1::
@end menu

@node Invocation
@subsection Invocation
@cindex garbage collection, invocation

The first thing that anyone should know about garbage collection is:
when and how the garbage collector is invoked. One might think that this 
could happen every time new memory is allocated, e.g. new objects are
created, but this is @emph{not} the case. Instead, we have the following
situation:

The entry point of any process of garbage collection is an invocation
of the function @code{garbage_collect_1} in file @code{alloc.c}. The
invocation can occur @emph{explicitly} by calling the function
@code{Fgarbage_collect} (in addition this function provides information
about the freed memory), or can occur @emph{implicitly} in four different 
situations:
@enumerate
@item
In function @code{main_1} in file @code{emacs.c}. This function is called
at each startup of xemacs. The garbage collection is invoked after all
initial creations are completed, but only if a special internal error
checking-constant @code{ERROR_CHECK_GC} is defined.
@item
In function @code{disksave_object_finalization} in file
@code{alloc.c}. The only purpose of this function is to clear the
objects from memory which need not be stored with xemacs when we dump out 
an executable. This is only done by @code{Fdump_emacs} or by
@code{Fdump_emacs_data} respectively (both in @code{emacs.c}). The
actual clearing is accomplished by making these objects unreachable and
starting a garbage collection. The function is only used while building
xemacs.
@item
In function @code{Feval / eval} in file @code{eval.c}. Each time the
well known and often used function eval is called to evaluate a form,
one of the first things that could happen, is a potential call of
@code{garbage_collect_1}. There exist three global variables,
@code{consing_since_gc} (counts the created cons-cells since the last
garbage collection), @code{gc_cons_threshold} (a specified threshold
after which a garbage collection occurs) and @code{always_gc}. If
@code{always_gc} is set or if the threshold is exceeded, the garbage
collection will start.
@item
In function @code{Ffuncall / funcall} in file @code{eval.c}. This
function evaluates calls of elisp functions and works according to
@code{Feval}.
@end enumerate

The upshot is that garbage collection can basically occur everywhere
@code{Feval}, respectively @code{Ffuncall}, is used - either directly or
through another function. Since calls to these two functions are
hidden in various other functions, many calls to
@code{garabge_collect_1} are not obviously foreseeable, and therefore
unexpected. Instances where they are used that are worth remembering are
various elisp commands, as for example @code{or},
@code{and}, @code{if}, @code{cond}, @code{while}, @code{setq}, etc.,
miscellaneous @code{gui_item_...} functions, everything related to
@code{eval} (@code{Feval_buffer}, @code{call0}, ...) and inside
@code{Fsignal}. The latter is used to handle signals, as for example the
ones raised by every @code{QUITE}-macro triggered after pressing Ctrl-g.

@node garbage_collect_1
@subsection @code{garbage_collect_1}
@cindex @code{garbage_collect_1}

We can now describe exactly what happens after the invocation takes
place.
@enumerate
@item
There are several cases in which the garbage collector is left immediately: 
when we are already garbage collecting (@code{gc_in_progress}), when
the garbage collection is somehow forbidden
(@code{gc_currently_forbidden}), when we are currently displaying something
(@code{in_display}) or when we are preparing for the armageddon of the
whole system (@code{preparing_for_armageddon}).
@item
Next the correct frame in which to put
all the output occurring during garbage collecting is determined. In
order to be able to restore the old display's state after displaying the
message, some data about the current cursor position has to be
saved. The variables @code{pre_gc_curser} and @code{cursor_changed} take
care of that.
@item
The state of @code{gc_currently_forbidden} must be restored after
the garbage collection, no matter what happens during the process. We
accomplish this by @code{record_unwind_protect}ing the suitable function
@code{restore_gc_inhibit} together with the current value of
@code{gc_currently_forbidden}. 
@item
If we are concurrently running an interactive xemacs session, the next step
is simply to show the garbage collector's cursor/message.
@item
The following steps are the intrinsic steps of the garbage collector,
therefore @code{gc_in_progress} is set.
@item
For debugging purposes, it is possible to copy the current C stack
frame. However, this seems to be a currently unused feature.
@item
Before actually starting to go over all live objects, references to
objects that are no longer used are pruned. We only have to do this for events
(@code{clear_event_resource}) and for specifiers
(@code{cleanup_specifiers}). 
@item
Now the mark phase begins and marks all accessible elements. In order to
start from
all slots that serve as roots of accessibility, the function
@code{mark_object} is called for each root individually to go out from
there to mark all reachable objects. All roots that are traversed are
shown in their processed order:
@itemize @bullet
@item
all constant symbols and static variables that are registered via
@code{staticpro}@ in the array @code{staticvec}.
@xref{Adding Global Lisp Variables}. 
@item
all Lisp objects that are created in C functions and that must be
protected from freeing them. They are registered in the global
list @code{gcprolist}.
@xref{GCPROing}.
@item 
all local variables (i.e. their name fields @code{symbol} and old
values @code{old_values}) that are bound during the evaluation by the Lisp
engine. They are stored in @code{specbinding} structs pushed on a stack
called @code{specpdl}.
@xref{Dynamic Binding; The specbinding Stack; Unwind-Protects}.
@item
all catch blocks that the Lisp engine encounters during the evaluation
cause the creation of structs @code{catchtag} inserted in the list
@code{catchlist}. Their tag (@code{tag}) and value (@code{val} fields
are freshly created objects and therefore have to be marked.
@xref{Catch and Throw}.
@item
every function application pushes new structs @code{backtrace} 
on the call stack of the Lisp engine (@code{backtrace_list}). The unique 
parts that have to be marked are the fields for each function
(@code{function}) and all their arguments (@code{args}).
@xref{Evaluation}.
@item
all objects that are used by the redisplay engine that must not be freed 
are marked by a special function called @code{mark_redisplay} (in
@code{redisplay.c}).
@item
all objects created for profiling purposes are allocated by C functions
instead of using the lisp allocation mechanisms. In order to receive the
right ones during the sweep phase, they also have to be marked
manually. That is done by the function @code{mark_profiling_info}
@end itemize
@item
Hash tables in XEmacs belong to a kind of special objects that
make use of a concept often called 'weak pointers'.
To make a long story short, these kind of pointers are not followed
during the estimation of the live objects during garbage collection.
Any object referenced only by weak pointers is collected
anyway, and the reference to it is cleared. In hash tables there are
different usage patterns of them, manifesting in different types of hash
tables, namely 'non-weak', 'weak', 'key-weak' and 'value-weak'
(internally also 'key-car-weak' and 'value-car-weak') hash tables, each 
clearing entries depending on different conditions. More information can 
be found in the documentation to the function @code{make-hash-table}.

Because there are complicated dependency rules about when and what to
mark while processing weak hash tables, the standard @code{marker}
method is only active if it is marking non-weak hash tables. As soon as
a weak component is in the table, the hash table entries are ignored
while marking. Instead their marking is done each separately by the
function @code{finish_marking_weak_hash_tables}. This function iterates
over each hash table entry @code{hentries} for each weak hash table in
@code{Vall_weak_hash_tables}. Depending on the type of a table, the
appropriate action is performed. 
If a table is acting as @code{HASH_TABLE_KEY_WEAK}, and a key already marked,
everything reachable from the @code{value} component is marked. If it is 
acting as a @code{HASH_TABLE_VALUE_WEAK} and the value component is
already marked, the marking starts beginning only from the 
@code{key} component.
If it is a @code{HASH_TABLE_KEY_CAR_WEAK} and the car 
of the key entry is already marked, we mark both the @code{key} and
@code{value} components.
Finally, if the table is of the type @code{HASH_TABLE_VALUE_CAR_WEAK}
and the car of the value components is already marked, again both the
@code{key} and the @code{value} components get marked.

Again, there are lists with comparable properties called weak
lists. There exist different peculiarities of their types called
@code{simple}, @code{assoc}, @code{key-assoc} and
@code{value-assoc}. You can find further details about them in the
description to the function @code{make-weak-list}. The scheme of their
marking is similar: all weak lists are listed in @code{Qall_weak_lists}, 
therefore we iterate over them. The marking is advanced until we hit an
already marked pair. Then we know that during a former run all 
the rest has been marked completely. Again, depending on the special
type of the weak list, our jobs differ. If it is a @code{WEAK_LIST_SIMPLE}
and the elem is marked, we mark the @code{cons} part. If it is a
@code{WEAK_LIST_ASSOC} and not a pair or a pair with both marked car and
cdr, we mark the @code{cons} and the @code{elem}. If it is a
@code{WEAK_LIST_KEY_ASSOC} and not a pair or a pair with a marked car of
the elem, we mark the @code{cons} and the @code{elem}. Finally, if it is
a @code{WEAK_LIST_VALUE_ASSOC} and not a pair or a pair with a marked
cdr of the elem, we mark both the @code{cons} and the @code{elem}.

Since, by marking objects in reach from weak hash tables and weak lists,
other objects could get marked, this perhaps implies further marking of
other weak objects, both finishing functions are redone as long as 
yet unmarked objects get freshly marked.

@item
After completing the special marking for the weak hash tables and for the weak
lists, all entries that point to objects that are going to be swept in
the further process are useless, and therefore have to be removed from
the table or the list.

The function @code{prune_weak_hash_tables} does the job for weak hash
tables. Totally unmarked hash tables are removed from the list
@code{Vall_weak_hash_tables}. The other ones are treated more carefully
by scanning over all entries and removing one as soon as one of 
the components @code{key} and @code{value} is unmarked.

The same idea applies to the weak lists. It is accomplished by
@code{prune_weak_lists}: An unmarked list is pruned from
@code{Vall_weak_lists} immediately. A marked list is treated more
carefully by going over it and removing just the unmarked pairs.

@item
The function @code{prune_specifiers} checks all listed specifiers held
in @code{Vall_speficiers} and removes the ones from the lists that are 
unmarked.

@item
All syntax tables are stored in a list called
@code{Vall_syntax_tables}. The function @code{prune_syntax_tables} walks 
through it and unlinks the tables that are unmarked.

@item
Next, we will attack the complete sweeping - the function
@code{gc_sweep} which holds the predominance.
@item
First, all the variables with respect to garbage collection are
reset. @code{consing_since_gc} - the counter of the created cells since 
the last garbage collection - is set back to 0, and
@code{gc_in_progress} is not @code{true} anymore.
@item
In case the session is interactive, the displayed cursor and message are 
removed again.
@item
The state of @code{gc_inhibit} is restored to the former value by
unwinding the stack.
@item
A small memory reserve is always held back that can be reached by
@code{breathing_space}. If nothing more is left, we create a new reserve
and exit. 
@end enumerate

@node mark_object
@subsection @code{mark_object}
@cindex @code{mark_object}

The first thing that is checked while marking an object is whether the
object is a real Lisp object @code{Lisp_Type_Record} or just an integer
or a character. Integers and characters are the only two types that are
stored directly - without another level of indirection, and therefore they
don't have to be marked and collected. 
@xref{How Lisp Objects Are Represented in C}.

The second case is the one we have to handle. It is the one when we are
dealing with a pointer to a Lisp object. But, there exist also three
possibilities, that prevent us from doing anything while marking: The
object is read only which prevents it from being garbage collected,
i.e. marked (@code{C_READONLY_RECORD_HEADER}). The object in question is
already marked, and need not be marked for the second time (checked by
@code{MARKED_RECORD_HEADER_P}). If it is a special, unmarkable object
(@code{UNMARKABLE_RECORD_HEADER_P}, apparently, these are objects that
sit in some CONST space, and can therefore not be marked, see
@code{this_one_is_unmarkable} in @code{alloc.c}).

Now, the actual marking is feasible. We do so by once using the macro
@code{MARK_RECORD_HEADER} to mark the object itself (actually the
special flag in the lrecord header), and calling its special marker
"method" @code{marker} if available. The marker method marks every
other object that is in reach from our current object. Note, that these 
marker methods should not call @code{mark_object} recursively, but
instead should return the next object from where further marking has to
be performed.

In case another object was returned, as mentioned before, we reiterate
the whole @code{mark_object} process beginning with this next object.

@node gc_sweep
@subsection @code{gc_sweep}
@cindex @code{gc_sweep}

The job of this function is to free all unmarked records from memory. As 
we know, there are different types of objects implemented and managed, and
consequently different ways to free them from memory.
@xref{Introduction to Allocation}.

We start with all objects stored through @code{lcrecords}. All
bulkier objects are allocated and handled using that scheme of
@code{lcrecords}. Each object is @code{malloc}ed separately
instead of placing it in one of the contiguous frob blocks. All types
that are currently stored 
using @code{lcrecords}'s  @code{alloc_lcrecord} and
@code{make_lcrecord_list} are the types: vectors, buffers,
char-table, char-table-entry, console, weak-list, database, device,
ldap, hash-table, command-builder, extent-auxiliary, extent-info, face,
coding-system, frame, image-instance, glyph, popup-data, gui-item,
keymap, charset, color_instance, font_instance, opaque, opaque-list,
process, range-table, specifier, symbol-value-buffer-local,
symbol-value-lisp-magic, symbol-value-varalias, toolbar-button,
tooltalk-message, tooltalk-pattern, window, and window-configuration. We
take care of them in the fist place
in order to be able to handle and to finalize items stored in them more
easily. The function @code{sweep_lcrecords_1} as described below is
doing the whole job for us.
For a description about the internals: @xref{lrecords}.

Our next candidates are the other objects that behave quite differently
than everything else: the strings. They consists of two parts, a
fixed-size portion (@code{struct Lisp_string}) holding the string's
length, its property list and a pointer to the second part, and the
actual string data, which is stored in string-chars blocks comparable to
frob blocks. In this block, the data is not only freed, but also a
compression of holes is made, i.e. all strings are relocated together.
@xref{String}. This compacting phase is performed by the function
@code{compact_string_chars}, the actual sweeping by the function
@code{sweep_strings} is described below.

After that, the other types are swept step by step using functions
@code{sweep_conses}, @code{sweep_bit_vectors_1},
@code{sweep_compiled_functions}, @code{sweep_floats},
@code{sweep_symbols}, @code{sweep_extents}, @code{sweep_markers} and
@code{sweep_extents}.  They are the fixed-size types cons, floats,
compiled-functions, symbol, marker, extent, and event stored in
so-called "frob blocks", and therefore we can basically do the same on
every type objects, using the same macros, especially defined only to
handle everything with respect to fixed-size blocks. The only fixed-size 
type that is not handled here are the fixed-size portion of strings,
because we took special care of them earlier.

The only big exceptions are bit vectors stored differently and
therefore treated differently by the function @code{sweep_bit_vectors_1} 
described later.

At first, we need some brief information about how
these fixed-size types are managed in general, in order to understand
how the sweeping is done. They have all a fixed size, and are therefore
stored in big blocks of memory - allocated at once - that can hold a
certain amount of objects of one type. The macro
@code{DECLARE_FIXED_TYPE_ALLOC} creates the suitable structures for
every type. More precisely, we have the block struct 
(holding a pointer to the previous block @code{prev} and the
objects in @code{block[]}), a pointer to current block
(@code{current_..._block)}) and its last index
(@code{current_..._block_index}), and a pointer to the free list that
will be created. Also a macro @code{FIXED_TYPE_FROM_BLOCK} plus some
related macros exists that are used to obtain a new object, either from
the free list @code{ALLOCATE_FIXED_TYPE_1} if there is an unused object
of that type stored or by allocating a completely new block using
@code{ALLOCATE_FIXED_TYPE_FROM_BLOCK}.

The rest works as follows: all of them define a
macro @code{UNMARK_...} that is used to unmark the object. They define a
macro @code{ADDITIONAL_FREE_...} that defines additional work that has
to be done when converting an object from in use to not in use (so far,
only markers use it in order to unchain them). Then, they all call
the macro @code{SWEEP_FIXED_TYPE_BLOCK} instantiated with their type name 
and their struct name.

This call in particular does the following: we go over all blocks
starting with the current moving towards the oldest.
For each block, we look at every object in it. If the object already
freed (checked with @code{FREE_STRUCT_P} using the first pointer of the
object), or if it is 
set to read only (@code{C_READONLY_RECORD_HEADER_P}, nothing must be
done. If it is unmarked (checked with @code{MARKED_RECORD_HEADER_P}), it
is put in the free list and set free (using the macro
@code{FREE_FIXED_TYPE}, otherwise it stays in the block, but is unmarked 
(by @code{UNMARK_...}). While going through one block, we note if the
whole block is empty. If so, the whole block is freed (using
@code{xfree}) and the free list state is set to the state it had before
handling this block.

@node sweep_lcrecords_1
@subsection @code{sweep_lcrecords_1}
@cindex @code{sweep_lcrecords_1}

After nullifying the complete lcrecord statistics, we go over all
lcrecords two separate times. They are all chained together in a list with 
a head called @code{all_lcrecords}. 

The first loop calls for each object its @code{finalizer} method, but only 
in the case that it is not read only
(@code{C_READONLY_RECORD_HEADER_P)}, it is not already marked
(@code{MARKED_RECORD_HEADER_P}), it is not already in a free list (list of
freed objects, field @code{free}) and finally it owns a finalizer
method.
 
The second loop actually frees the appropriate objects again by iterating 
through the whole list. In case an object is read only or marked, it 
has to persist, otherwise it is manually freed by calling
@code{xfree}. During this loop, the lcrecord statistics are kept up to
date by calling @code{tick_lcrecord_stats} with the right arguments, 

@node compact_string_chars
@subsection @code{compact_string_chars}
@cindex @code{compact_string_chars}

The purpose of this function is to compact all the data parts of the
strings that are held in so-called @code{string_chars_block}, i.e. the
strings that do not exceed a certain maximal length.

The procedure with which this is done is as follows. We are keeping two
positions in the @code{string_chars_block}s using two pointer/integer
pairs, namely @code{from_sb}/@code{from_pos} and
@code{to_sb}/@code{to_pos}. They stand for the actual positions, from
where to where, to copy the actually handled string. 

While going over all chained @code{string_char_block}s and their held
strings, staring at @code{first_string_chars_block}, both pointers
are advanced and eventually a string is copied from @code{from_sb} to
@code{to_sb}, depending on the status of the pointed at strings.

More precisely, we can distinguish between the following actions.
@itemize @bullet
@item
The string at @code{from_sb}'s position could be marked as free, which
is indicated by an invalid pointer to the pointer that should point back 
to the fixed size string object, and which is checked by
@code{FREE_STRUCT_P}. In this case, the @code{from_sb}/@code{from_pos}
is advanced to the next string, and nothing has to be copied.
@item
Also, if a string object itself is unmarked, nothing has to be
copied. We likewise advance the @code{from_sb}/@code{from_pos}
pair as described above.
@item
In all other cases, we have a marked string at hand. The string data 
must be moved from the from-position to the to-position. In case
there is not enough space in the actual @code{to_sb}-block, we advance
this pointer to the beginning of the next block before copying. In case the
from and to positions are different, we perform the
actual copying using the library function @code{memmove}.
@end itemize

After compacting, the pointer to the current
@code{string_chars_block}, sitting in @code{current_string_chars_block},
is reset on the last block to which we moved a string,
i.e. @code{to_block}, and all remaining blocks (we know that they just
carry garbage) are explicitly @code{xfree}d.

@node sweep_strings
@subsection @code{sweep_strings}
@cindex @code{sweep_strings}

The sweeping for the fixed sized string objects is essentially exactly
the same as it is for all other fixed size types. As before, the freeing
into the suitable free list is done by using the macro
@code{SWEEP_FIXED_SIZE_BLOCK} after defining the right macros
@code{UNMARK_string} and @code{ADDITIONAL_FREE_string}. These two
definitions are a little bit special compared to the ones used
for the other fixed size types.

@code{UNMARK_string} is defined the same way except some additional code 
used for updating the bookkeeping information.

For strings, @code{ADDITIONAL_FREE_string} has to do something in
addition: in case, the string was not allocated in a
@code{string_chars_block} because it exceeded the maximal length, and
therefore it was @code{malloc}ed separately, we know also @code{xfree}
it explicitly.

@node sweep_bit_vectors_1
@subsection @code{sweep_bit_vectors_1}
@cindex @code{sweep_bit_vectors_1}

Bit vectors are also one of the rare types that are @code{malloc}ed
individually. Consequently, while sweeping, all further needless
bit vectors must be freed by hand. This is done, as one might imagine,
the expected way: since they are all registered in a list called
@code{all_bit_vectors}, all elements of that list are traversed,
all unmarked bit vectors are unlinked by calling @code{xfree} and all of 
them become unmarked.
In addition, the bookkeeping information used for garbage 
collector's output purposes is updated.

@node Integers and Characters
@section Integers and Characters

  Integer and character Lisp objects are created from integers using the
macros @code{XSETINT()} and @code{XSETCHAR()} or the equivalent
functions @code{make_int()} and @code{make_char()}. (These are actually
macros on most systems.)  These functions basically just do some moving
of bits around, since the integral value of the object is stored
directly in the @code{Lisp_Object}.

  @code{XSETINT()} and the like will truncate values given to them that
are too big; i.e. you won't get the value you expected but the tag bits
will at least be correct.

@node Allocation from Frob Blocks
@section Allocation from Frob Blocks

The uninitialized memory required by a @code{Lisp_Object} of a particular type
is allocated using
@code{ALLOCATE_FIXED_TYPE()}.  This only occurs inside of the
lowest-level object-creating functions in @file{alloc.c}:
@code{Fcons()}, @code{make_float()}, @code{Fmake_byte_code()},
@code{Fmake_symbol()}, @code{allocate_extent()},
@code{allocate_event()}, @code{Fmake_marker()}, and
@code{make_uninit_string()}.  The idea is that, for each type, there are
a number of frob blocks (each 2K in size); each frob block is divided up
into object-sized chunks.  Each frob block will have some of these
chunks that are currently assigned to objects, and perhaps some that are
free. (If a frob block has nothing but free chunks, it is freed at the
end of the garbage collection cycle.)  The free chunks are stored in a
free list, which is chained by storing a pointer in the first four bytes
of the chunk. (Except for the free chunks at the end of the last frob
block, which are handled using an index which points past the end of the
last-allocated chunk in the last frob block.)
@code{ALLOCATE_FIXED_TYPE()} first tries to retrieve a chunk from the
free list; if that fails, it calls
@code{ALLOCATE_FIXED_TYPE_FROM_BLOCK()}, which looks at the end of the
last frob block for space, and creates a new frob block if there is
none. (There are actually two versions of these macros, one of which is
more defensive but less efficient and is used for error-checking.)

@node lrecords
@section lrecords

  [see @file{lrecord.h}]

  All lrecords have at the beginning of their structure a @code{struct
lrecord_header}.  This just contains a pointer to a @code{struct
lrecord_implementation}, which is a structure containing method pointers
and such.  There is one of these for each type, and it is a global,
constant, statically-declared structure that is declared in the
@code{DEFINE_LRECORD_IMPLEMENTATION()} macro. (This macro actually
declares an array of two @code{struct lrecord_implementation}
structures.  The first one contains all the standard method pointers,
and is used in all normal circumstances.  During garbage collection,
however, the lrecord is @dfn{marked} by bumping its implementation
pointer by one, so that it points to the second structure in the array.
This structure contains a special indication in it that it's a
@dfn{marked-object} structure: the finalize method is the special
function @code{this_marks_a_marked_record()}, and all other methods are
null pointers.  At the end of garbage collection, all lrecords will
either be reclaimed or unmarked by decrementing their implementation
pointers, so this second structure pointer will never remain past
garbage collection.

  Simple lrecords (of type (c) above) just have a @code{struct
lrecord_header} at their beginning.  lcrecords, however, actually have a
@code{struct lcrecord_header}.  This, in turn, has a @code{struct
lrecord_header} at its beginning, so sanity is preserved; but it also
has a pointer used to chain all lcrecords together, and a special ID
field used to distinguish one lcrecord from another. (This field is used
only for debugging and could be removed, but the space gain is not
significant.)

  Simple lrecords are created using @code{ALLOCATE_FIXED_TYPE()}, just
like for other frob blocks.  The only change is that the implementation
pointer must be initialized correctly. (The implementation structure for
an lrecord, or rather the pointer to it, is named @code{lrecord_float},
@code{lrecord_extent}, @code{lrecord_buffer}, etc.)

  lcrecords are created using @code{alloc_lcrecord()}.  This takes a
size to allocate and an implementation pointer. (The size needs to be
passed because some lcrecords, such as window configurations, are of
variable size.) This basically just @code{malloc()}s the storage,
initializes the @code{struct lcrecord_header}, and chains the lcrecord
onto the head of the list of all lcrecords, which is stored in the
variable @code{all_lcrecords}.  The calls to @code{alloc_lcrecord()}
generally occur in the lowest-level allocation function for each lrecord
type.

Whenever you create an lrecord, you need to call either
@code{DEFINE_LRECORD_IMPLEMENTATION()} or
@code{DEFINE_LRECORD_SEQUENCE_IMPLEMENTATION()}.  This needs to be
specified in a C file, at the top level.  What this actually does is
define and initialize the implementation structure for the lrecord. (And
possibly declares a function @code{error_check_foo()} that implements
the @code{XFOO()} macro when error-checking is enabled.)  The arguments
to the macros are the actual type name (this is used to construct the C
variable name of the lrecord implementation structure and related
structures using the @samp{##} macro concatenation operator), a string
that names the type on the Lisp level (this may not be the same as the C
type name; typically, the C type name has underscores, while the Lisp
string has dashes), various method pointers, and the name of the C
structure that contains the object.  The methods are used to encapsulate
type-specific information about the object, such as how to print it or
mark it for garbage collection, so that it's easy to add new object
types without having to add a specific case for each new type in a bunch
of different places.

  The difference between @code{DEFINE_LRECORD_IMPLEMENTATION()} and
@code{DEFINE_LRECORD_SEQUENCE_IMPLEMENTATION()} is that the former is
used for fixed-size object types and the latter is for variable-size
object types.  Most object types are fixed-size; some complex
types, however (e.g. window configurations), are variable-size.
Variable-size object types have an extra method, which is called
to determine the actual size of a particular object of that type.
(Currently this is only used for keeping allocation statistics.)

  For the purpose of keeping allocation statistics, the allocation
engine keeps a list of all the different types that exist.  Note that,
since @code{DEFINE_LRECORD_IMPLEMENTATION()} is a macro that is
specified at top-level, there is no way for it to add to the list of all
existing types.  What happens instead is that each implementation
structure contains in it a dynamically assigned number that is
particular to that type. (Or rather, it contains a pointer to another
structure that contains this number.  This evasiveness is done so that
the implementation structure can be declared const.) In the sweep stage
of garbage collection, each lrecord is examined to see if its
implementation structure has its dynamically-assigned number set.  If
not, it must be a new type, and it is added to the list of known types
and a new number assigned.  The number is used to index into an array
holding the number of objects of each type and the total memory
allocated for objects of that type.  The statistics in this array are
also computed during the sweep stage.  These statistics are returned by
the call to @code{garbage-collect} and are printed out at the end of the
loadup phase.

  Note that for every type defined with a @code{DEFINE_LRECORD_*()}
macro, there needs to be a @code{DECLARE_LRECORD_IMPLEMENTATION()}
somewhere in a @file{.h} file, and this @file{.h} file needs to be
included by @file{inline.c}.

  Furthermore, there should generally be a set of @code{XFOOBAR()},
@code{FOOBARP()}, etc. macros in a @file{.h} (or occasionally @file{.c})
file.  To create one of these, copy an existing model and modify as
necessary.

  The various methods in the lrecord implementation structure are:

@enumerate
@item
@cindex mark method
A @dfn{mark} method.  This is called during the marking stage and passed
a function pointer (usually the @code{mark_object()} function), which is
used to mark an object.  All Lisp objects that are contained within the
object need to be marked by applying this function to them.  The mark
method should also return a Lisp object, which should be either nil or
an object to mark. (This can be used in lieu of calling
@code{mark_object()} on the object, to reduce the recursion depth, and
consequently should be the most heavily nested sub-object, such as a
long list.)

@strong{Please note:} When the mark method is called, garbage collection
is in progress, and special precautions need to be taken when accessing
objects; see section (B) above.

If your mark method does not need to do anything, it can be
@code{NULL}.

@item
A @dfn{print} method.  This is called to create a printed representation
of the object, whenever @code{princ}, @code{prin1}, or the like is
called.  It is passed the object, a stream to which the output is to be
directed, and an @code{escapeflag} which indicates whether the object's
printed representation should be @dfn{escaped} so that it is
readable. (This corresponds to the difference between @code{princ} and
@code{prin1}.) Basically, @dfn{escaped} means that strings will have
quotes around them and confusing characters in the strings such as
quotes, backslashes, and newlines will be backslashed; and that special
care will be taken to make symbols print in a readable fashion
(e.g. symbols that look like numbers will be backslashed).  Other
readable objects should perhaps pass @code{escapeflag} on when
sub-objects are printed, so that readability is preserved when necessary
(or if not, always pass in a 1 for @code{escapeflag}).  Non-readable
objects should in general ignore @code{escapeflag}, except that some use
it as an indication that more verbose output should be given.

Sub-objects are printed using @code{print_internal()}, which takes
exactly the same arguments as are passed to the print method.

Literal C strings should be printed using @code{write_c_string()},
or @code{write_string_1()} for non-null-terminated strings.

Functions that do not have a readable representation should check the
@code{print_readably} flag and signal an error if it is set.

If you specify NULL for the print method, the
@code{default_object_printer()} will be used.

@item
A @dfn{finalize} method.  This is called at the beginning of the sweep
stage on lcrecords that are about to be freed, and should be used to
perform any extra object cleanup.  This typically involves freeing any
extra @code{malloc()}ed memory associated with the object, releasing any
operating-system and window-system resources associated with the object
(e.g. pixmaps, fonts), etc.

The finalize method can be NULL if nothing needs to be done.

WARNING #1: The finalize method is also called at the end of the dump
phase; this time with the for_disksave parameter set to non-zero.  The
object is @emph{not} about to disappear, so you have to make sure to
@emph{not} free any extra @code{malloc()}ed memory if you're going to
need it later.  (Also, signal an error if there are any operating-system
and window-system resources here, because they can't be dumped.)

Finalize methods should, as a rule, set to zero any pointers after
they've been freed, and check to make sure pointers are not zero before
freeing.  Although I'm pretty sure that finalize methods are not called
twice on the same object (except for the @code{for_disksave} proviso),
we've gotten nastily burned in some cases by not doing this.

WARNING #2: The finalize method is @emph{only} called for
lcrecords, @emph{not} for simply lrecords.  If you need a
finalize method for simple lrecords, you have to stick
it in the @code{ADDITIONAL_FREE_foo()} macro in @file{alloc.c}.

WARNING #3: Things are in an @emph{extremely} bizarre state
when @code{ADDITIONAL_FREE_foo()} is called, so you have to
be incredibly careful when writing one of these functions.
See the comment in @code{gc_sweep()}.  If you ever have to add
one of these, consider using an lcrecord or dealing with
the problem in a different fashion.

@item
An @dfn{equal} method.  This compares the two objects for similarity,
when @code{equal} is called.  It should compare the contents of the
objects in some reasonable fashion.  It is passed the two objects and a
@dfn{depth} value, which is used to catch circular objects.  To compare
sub-Lisp-objects, call @code{internal_equal()} and bump the depth value
by one.  If this value gets too high, a @code{circular-object} error
will be signaled.

If this is NULL, objects are @code{equal} only when they are @code{eq},
i.e. identical.

@item
A @dfn{hash} method.  This is used to hash objects when they are to be
compared with @code{equal}.  The rule here is that if two objects are
@code{equal}, they @emph{must} hash to the same value; i.e. your hash
function should use some subset of the sub-fields of the object that are
compared in the ``equal'' method.  If you specify this method as
@code{NULL}, the object's pointer will be used as the hash, which will
@emph{fail} if the object has an @code{equal} method, so don't do this.

To hash a sub-Lisp-object, call @code{internal_hash()}.  Bump the
depth by one, just like in the ``equal'' method.

To convert a Lisp object directly into a hash value (using
its pointer), use @code{LISP_HASH()}.  This is what happens when
the hash method is NULL.

To hash two or more values together into a single value, use
@code{HASH2()}, @code{HASH3()}, @code{HASH4()}, etc.

@item
@dfn{getprop}, @dfn{putprop}, @dfn{remprop}, and @dfn{plist} methods.
These are used for object types that have properties.  I don't feel like
documenting them here.  If you create one of these objects, you have to
use different macros to define them,
i.e. @code{DEFINE_LRECORD_IMPLEMENTATION_WITH_PROPS()} or
@code{DEFINE_LRECORD_SEQUENCE_IMPLEMENTATION_WITH_PROPS()}.

@item
A @dfn{size_in_bytes} method, when the object is of variable-size.
(i.e. declared with a @code{_SEQUENCE_IMPLEMENTATION} macro.)  This should
simply return the object's size in bytes, exactly as you might expect.
For an example, see the methods for window configurations and opaques.
@end enumerate

@node Low-level allocation
@section Low-level allocation

  Memory that you want to allocate directly should be allocated using
@code{xmalloc()} rather than @code{malloc()}.  This implements
error-checking on the return value, and once upon a time did some more
vital stuff (i.e. @code{BLOCK_INPUT}, which is no longer necessary).
Free using @code{xfree()}, and realloc using @code{xrealloc()}.  Note
that @code{xmalloc()} will do a non-local exit if the memory can't be
allocated. (Many functions, however, do not expect this, and thus XEmacs
will likely crash if this happens.  @strong{This is a bug.}  If you can,
you should strive to make your function handle this OK.  However, it's
difficult in the general circumstance, perhaps requiring extra
unwind-protects and such.)

  Note that XEmacs provides two separate replacements for the standard
@code{malloc()} library function.  These are called @dfn{old GNU malloc}
(@file{malloc.c}) and @dfn{new GNU malloc} (@file{gmalloc.c}),
respectively.  New GNU malloc is better in pretty much every way than
old GNU malloc, and should be used if possible.  (It used to be that on
some systems, the old one worked but the new one didn't.  I think this
was due specifically to a bug in SunOS, which the new one now works
around; so I don't think the old one ever has to be used any more.) The
primary difference between both of these mallocs and the standard system
malloc is that they are much faster, at the expense of increased space.
The basic idea is that memory is allocated in fixed chunks of powers of
two.  This allows for basically constant malloc time, since the various
chunks can just be kept on a number of free lists. (The standard system
malloc typically allocates arbitrary-sized chunks and has to spend some
time, sometimes a significant amount of time, walking the heap looking
for a free block to use and cleaning things up.)  The new GNU malloc
improves on things by allocating large objects in chunks of 4096 bytes
rather than in ever larger powers of two, which results in ever larger
wastage.  There is a slight speed loss here, but it's of doubtful
significance.

  NOTE: Apparently there is a third-generation GNU malloc that is
significantly better than the new GNU malloc, and should probably
be included in XEmacs.

  There is also the relocating allocator, @file{ralloc.c}.  This actually
moves blocks of memory around so that the @code{sbrk()} pointer shrunk
and virtual memory released back to the system.  On some systems,
this is a big win.  On all systems, it causes a noticeable (and
sometimes huge) speed penalty, so I turn it off by default.
@file{ralloc.c} only works with the new GNU malloc in @file{gmalloc.c}.
There are also two versions of @file{ralloc.c}, one that uses @code{mmap()}
rather than block copies to move data around.  This purports to
be faster, although that depends on the amount of data that would
have had to be block copied and the system-call overhead for
@code{mmap()}.  I don't know exactly how this works, except that the
relocating-allocation routines are pretty much used only for
the memory allocated for a buffer, which is the biggest consumer
of space, esp. of space that may get freed later.

  Note that the GNU mallocs have some ``memory warning'' facilities.
XEmacs taps into them and issues a warning through the standard
warning system, when memory gets to 75%, 85%, and 95% full.
(On some systems, the memory warnings are not functional.)

  Allocated memory that is going to be used to make a Lisp object
is created using @code{allocate_lisp_storage()}.  This calls @code{xmalloc()}
but also verifies that the pointer to the memory can fit into
a Lisp word (remember that some bits are taken away for a type
tag and a mark bit).  If not, an error is issued through @code{memory_full()}.
@code{allocate_lisp_storage()} is called by @code{alloc_lcrecord()},
@code{ALLOCATE_FIXED_TYPE()}, and the vector and bit-vector creation
routines.  These routines also call @code{INCREMENT_CONS_COUNTER()} at the
appropriate times; this keeps statistics on how much memory is
allocated, so that garbage-collection can be invoked when the
threshold is reached.

@node Pure Space
@section Pure Space

  Not yet documented.

@node Cons
@section Cons

  Conses are allocated in standard frob blocks.  The only thing to
note is that conses can be explicitly freed using @code{free_cons()}
and associated functions @code{free_list()} and @code{free_alist()}.  This
immediately puts the conses onto the cons free list, and decrements
the statistics on memory allocation appropriately.  This is used
to good effect by some extremely commonly-used code, to avoid
generating extra objects and thereby triggering GC sooner.
However, you have to be @emph{extremely} careful when doing this.
If you mess this up, you will get BADLY BURNED, and it has happened
before.

@node Vector
@section Vector

  As mentioned above, each vector is @code{malloc()}ed individually, and
all are threaded through the variable @code{all_vectors}.  Vectors are
marked strangely during garbage collection, by kludging the size field.
Note that the @code{struct Lisp_Vector} is declared with its
@code{contents} field being a @emph{stretchy} array of one element.  It
is actually @code{malloc()}ed with the right size, however, and access
to any element through the @code{contents} array works fine.

@node Bit Vector
@section Bit Vector

  Bit vectors work exactly like vectors, except for more complicated
code to access an individual bit, and except for the fact that bit
vectors are lrecords while vectors are not. (The only difference here is
that there's an lrecord implementation pointer at the beginning and the
tag field in bit vector Lisp words is ``lrecord'' rather than
``vector''.)

@node Symbol
@section Symbol

  Symbols are also allocated in frob blocks.  Note that the code
exists for symbols to be either lrecords (category (c) above)
or simple types (category (b) above), and are lrecords by
default (I think), although there is no good reason for this.

  Note that symbols in the awful horrible obarray structure are
chained through their @code{next} field.

Remember that @code{intern} looks up a symbol in an obarray, creating
one if necessary.

@node Marker
@section Marker

  Markers are allocated in frob blocks, as usual.  They are kept
in a buffer unordered, but in a doubly-linked list so that they
can easily be removed. (Formerly this was a singly-linked list,
but in some cases garbage collection took an extraordinarily
long time due to the O(N^2) time required to remove lots of
markers from a buffer.) Markers are removed from a buffer in
the finalize stage, in @code{ADDITIONAL_FREE_marker()}.

@node String
@section String

  As mentioned above, strings are a special case.  A string is logically
two parts, a fixed-size object (containing the length, property list,
and a pointer to the actual data), and the actual data in the string.
The fixed-size object is a @code{struct Lisp_String} and is allocated in
frob blocks, as usual.  The actual data is stored in special
@dfn{string-chars blocks}, which are 8K blocks of memory.
Currently-allocated strings are simply laid end to end in these
string-chars blocks, with a pointer back to the @code{struct Lisp_String}
stored before each string in the string-chars block.  When a new string
needs to be allocated, the remaining space at the end of the last
string-chars block is used if there's enough, and a new string-chars
block is created otherwise.

  There are never any holes in the string-chars blocks due to the string
compaction and relocation that happens at the end of garbage collection.
During the sweep stage of garbage collection, when objects are
reclaimed, the garbage collector goes through all string-chars blocks,
looking for unused strings.  Each chunk of string data is preceded by a
pointer to the corresponding @code{struct Lisp_String}, which indicates
both whether the string is used and how big the string is, i.e. how to
get to the next chunk of string data.  Holes are compressed by
block-copying the next string into the empty space and relocating the
pointer stored in the corresponding @code{struct Lisp_String}.
@strong{This means you have to be careful with strings in your code.}
See the section above on @code{GCPRO}ing.

  Note that there is one situation not handled: a string that is too big
to fit into a string-chars block.  Such strings, called @dfn{big
strings}, are all @code{malloc()}ed as their own block. (#### Although it
would make more sense for the threshold for big strings to be somewhat
lower, e.g. 1/2 or 1/4 the size of a string-chars block.  It seems that
this was indeed the case formerly---indeed, the threshold was set at
1/8---but Mly forgot about this when rewriting things for 19.8.)

Note also that the string data in string-chars blocks is padded as
necessary so that proper alignment constraints on the @code{struct
Lisp_String} back pointers are maintained.

  Finally, strings can be resized.  This happens in Mule when a
character is substituted with a different-length character, or during
modeline frobbing. (You could also export this to Lisp, but it's not
done so currently.) Resizing a string is a potentially tricky process.
If the change is small enough that the padding can absorb it, nothing
other than a simple memory move needs to be done.  Keep in mind,
however, that the string can't shrink too much because the offset to the
next string in the string-chars block is computed by looking at the
length and rounding to the nearest multiple of four or eight.  If the
string would shrink or expand beyond the correct padding, new string
data needs to be allocated at the end of the last string-chars block and
the data moved appropriately.  This leaves some dead string data, which
is marked by putting a special marker of 0xFFFFFFFF in the @code{struct
Lisp_String} pointer before the data (there's no real @code{struct
Lisp_String} to point to and relocate), and storing the size of the dead
string data (which would normally be obtained from the now-non-existent
@code{struct Lisp_String}) at the beginning of the dead string data gap.
The string compactor recognizes this special 0xFFFFFFFF marker and
handles it correctly.

@node Compiled Function
@section Compiled Function

  Not yet documented.

@node Events and the Event Loop, Evaluation; Stack Frames; Bindings, Allocation of Objects in XEmacs Lisp, Top
@chapter Events and the Event Loop

@menu
* Introduction to Events::
* Main Loop::
* Specifics of the Event Gathering Mechanism::
* Specifics About the Emacs Event::
* The Event Stream Callback Routines::
* Other Event Loop Functions::
* Converting Events::
* Dispatching Events; The Command Builder::
@end menu

@node Introduction to Events
@section Introduction to Events

  An event is an object that encapsulates information about an
interesting occurrence in the operating system.  Events are
generated either by user action, direct (e.g. typing on the
keyboard or moving the mouse) or indirect (moving another
window, thereby generating an expose event on an Emacs frame),
or as a result of some other typically asynchronous action happening,
such as output from a subprocess being ready or a timer expiring.
Events come into the system in an asynchronous fashion (typically
through a callback being called) and are converted into a
synchronous event queue (first-in, first-out) in a process that
we will call @dfn{collection}.

  Note that each application has its own event queue. (It is
immaterial whether the collection process directly puts the
events in the proper application's queue, or puts them into
a single system queue, which is later split up.)

  The most basic level of event collection is done by the
operating system or window system.  Typically, XEmacs does
its own event collection as well.  Often there are multiple
layers of collection in XEmacs, with events from various
sources being collected into a queue, which is then combined
with other sources to go into another queue (i.e. a second
level of collection), with perhaps another level on top of
this, etc.

  XEmacs has its own types of events (called @dfn{Emacs events}),
which provides an abstract layer on top of the system-dependent
nature of the most basic events that are received.  Part of the
complex nature of the XEmacs event collection process involves
converting from the operating-system events into the proper
Emacs events---there may not be a one-to-one correspondence.

  Emacs events are documented in @file{events.h}; I'll discuss them
later.

@node Main Loop
@section Main Loop

  The @dfn{command loop} is the top-level loop that the editor is always
running.  It loops endlessly, calling @code{next-event} to retrieve an
event and @code{dispatch-event} to execute it. @code{dispatch-event} does
the appropriate thing with non-user events (process, timeout,
magic, eval, mouse motion); this involves calling a Lisp handler
function, redrawing a newly-exposed part of a frame, reading
subprocess output, etc.  For user events, @code{dispatch-event}
looks up the event in relevant keymaps or menubars; when a
full key sequence or menubar selection is reached, the appropriate
function is executed. @code{dispatch-event} may have to keep state
across calls; this is done in the ``command-builder'' structure
associated with each console (remember, there's usually only
one console), and the engine that looks up keystrokes and
constructs full key sequences is called the @dfn{command builder}.
This is documented elsewhere.

  The guts of the command loop are in @code{command_loop_1()}.  This
function doesn't catch errors, though---that's the job of
@code{command_loop_2()}, which is a condition-case (i.e. error-trapping)
wrapper around @code{command_loop_1()}.  @code{command_loop_1()} never
returns, but may get thrown out of.

  When an error occurs, @code{cmd_error()} is called, which usually
invokes the Lisp error handler in @code{command-error}; however, a
default error handler is provided if @code{command-error} is @code{nil}
(e.g. during startup).  The purpose of the error handler is simply to
display the error message and do associated cleanup; it does not need to
throw anywhere.  When the error handler finishes, the condition-case in
@code{command_loop_2()} will finish and @code{command_loop_2()} will
reinvoke @code{command_loop_1()}.

  @code{command_loop_2()} is invoked from three places: from
@code{initial_command_loop()} (called from @code{main()} at the end of
internal initialization), from the Lisp function @code{recursive-edit},
and from @code{call_command_loop()}.

  @code{call_command_loop()} is called when a macro is started and when
the minibuffer is entered; normal termination of the macro or minibuffer
causes a throw out of the recursive command loop. (To
@code{execute-kbd-macro} for macros and @code{exit} for minibuffers.
Note also that the low-level minibuffer-entering function,
@code{read-minibuffer-internal}, provides its own error handling and
does not need @code{command_loop_2()}'s error encapsulation; so it tells
@code{call_command_loop()} to invoke @code{command_loop_1()} directly.)

  Note that both read-minibuffer-internal and recursive-edit set up a
catch for @code{exit}; this is why @code{abort-recursive-edit}, which
throws to this catch, exits out of either one.

  @code{initial_command_loop()}, called from @code{main()}, sets up a
catch for @code{top-level} when invoking @code{command_loop_2()},
allowing functions to throw all the way to the top level if they really
need to.  Before invoking @code{command_loop_2()},
@code{initial_command_loop()} calls @code{top_level_1()}, which handles
all of the startup stuff (creating the initial frame, handling the
command-line options, loading the user's @file{.emacs} file, etc.).  The
function that actually does this is in Lisp and is pointed to by the
variable @code{top-level}; normally this function is
@code{normal-top-level}.  @code{top_level_1()} is just an error-handling
wrapper similar to @code{command_loop_2()}.  Note also that
@code{initial_command_loop()} sets up a catch for @code{top-level} when
invoking @code{top_level_1()}, just like when it invokes
@code{command_loop_2()}.

@node Specifics of the Event Gathering Mechanism
@section Specifics of the Event Gathering Mechanism

  Here is an approximate diagram of the collection processes
at work in XEmacs, under TTY's (TTY's are simpler than X
so we'll look at this first):

@noindent
@example
 asynch.      asynch.    asynch.   asynch.             [Collectors in
kbd events  kbd events   process   process                the OS]
      |         |         output    output
      |         |           |         |
      |         |           |         |      SIGINT,   [signal handlers
      |         |           |         |      SIGQUIT,     in XEmacs]
      V         V           V         V      SIGWINCH,
     file      file        file      file    SIGALRM
     desc.     desc.       desc.     desc.     |
     (TTY)     (TTY)       (pipe)    (pipe)    |
      |          |          |         |      fake    timeouts
      |          |          |         |      file        |
      |          |          |         |      desc.       |
      |          |          |         |      (pipe)      |
      |          |          |         |        |         |
      |          |          |         |        |         |
      |          |          |         |        |         |
      V          V          V         V        V         V
      ------>-----------<----------------<----------------
                  |
                  |
                  | [collected using select() in emacs_tty_next_event()
                  |  and converted to the appropriate Emacs event]
                  |
                  |
                  V          (above this line is TTY-specific)
                Emacs -----------------------------------------------
                event (below this line is the generic event mechanism)
                  |
                  |
was there     if not, call
a SIGINT?  emacs_tty_next_event()
    |             |
    |             |
    |             |
    V             V
    --->------<----
           |
           |     [collected in event_stream_next_event();
           |      SIGINT is converted using maybe_read_quit_event()]
           V
         Emacs
         event
           |
           \---->------>----- maybe_kbd_translate() ---->---\
                                                            |
                                                            |
                                                            |
     command event queue                                    |
                                               if not from command
  (contains events that were                   event queue, call
  read earlier but not processed,              event_stream_next_event()
  typically when waiting in a                               |
  sit-for, sleep-for, etc. for                              |
 a particular event to be received)                         |
               |                                            |
               |                                            |
               V                                            V
               ---->------------------------------------<----
                                               |
                                               | [collected in
                                               |  next_event_internal()]
                                               |
 unread-     unread-       event from          |
 command-    command-       keyboard       else, call
 events      event           macro      next_event_internal()
   |           |               |               |
   |           |               |               |
   |           |               |               |
   V           V               V               V
   --------->----------------------<------------
                     |
                     |      [collected in `next-event', which may loop
                     |       more than once if the event it gets is on
                     |       a dead frame, device, etc.]
                     |
                     |
                     V
            feed into top-level event loop,
            which repeatedly calls `next-event'
            and then dispatches the event
            using `dispatch-event'
@end example

Notice the separation between TTY-specific and generic event mechanism.
When using the Xt-based event loop, the TTY-specific stuff is replaced
but the rest stays the same.

It's also important to realize that only one different kind of
system-specific event loop can be operating at a time, and must be able
to receive all kinds of events simultaneously.  For the two existing
event loops (implemented in @file{event-tty.c} and @file{event-Xt.c},
respectively), the TTY event loop @emph{only} handles TTY consoles,
while the Xt event loop handles @emph{both} TTY and X consoles.  This
situation is different from all of the output handlers, where you simply
have one per console type.

  Here's the Xt Event Loop Diagram (notice that below a certain point,
it's the same as the above diagram):

@example
asynch. asynch. asynch. asynch.                 [Collectors in
 kbd     kbd    process process                    the OS]
events  events  output  output
  |       |       |       |
  |       |       |       |     asynch. asynch. [Collectors in the
  |       |       |       |       X        X     OS and X Window System]
  |       |       |       |     events  events
  |       |       |       |       |        |
  |       |       |       |       |        |
  |       |       |       |       |        |    SIGINT, [signal handlers
  |       |       |       |       |        |    SIGQUIT,   in XEmacs]
  |       |       |       |       |        |    SIGWINCH,
  |       |       |       |       |        |    SIGALRM
  |       |       |       |       |        |       |
  |       |       |       |       |        |       |
  |       |       |       |       |        |       |      timeouts
  |       |       |       |       |        |       |          |
  |       |       |       |       |        |       |          |
  |       |       |       |       |        |       V          |
  V       V       V       V       V        V      fake        |
 file    file    file    file    file     file    file        |
 desc.   desc.   desc.   desc.   desc.    desc.   desc.       |
 (TTY)   (TTY)   (pipe)  (pipe) (socket) (socket) (pipe)      |
  |       |       |       |       |        |       |          |
  |       |       |       |       |        |       |          |
  |       |       |       |       |        |       |          |
  V       V       V       V       V        V       V          V
  --->----------------------------------------<---------<------
       |              |               |
       |              |               |[collected using select() in
       |              |               | _XtWaitForSomething(), called
       |              |               | from XtAppProcessEvent(), called
       |              |               | in emacs_Xt_next_event();
       |              |               | dispatched to various callbacks]
       |              |               |
       |              |               |
  emacs_Xt_        p_s_callback(),    | [popup_selection_callback]
  event_handler()  x_u_v_s_callback(),| [x_update_vertical_scrollbar_
       |           x_u_h_s_callback(),|  callback]
       |           search_callback()  | [x_update_horizontal_scrollbar_
       |              |               |  callback]
       |              |               |
       |              |               |
  enqueue_Xt_       signal_special_   |
  dispatch_event()  Xt_user_event()   |
  [maybe multiple     |               |
   times, maybe 0     |               |
   times]             |               |
       |            enqueue_Xt_       |
       |            dispatch_event()  |
       |              |               |
       |              |               |
       V              V               |
       -->----------<--               |
              |                       |
              |                       |
           dispatch             Xt_what_callback()
           event                  sets flags
           queue                      |
              |                       |
              |                       |
              |                       |
              |                       |
              ---->-----------<--------
                   |
                   |
                   |     [collected and converted as appropriate in
                   |            emacs_Xt_next_event()]
                   |
                   |
                   V          (above this line is Xt-specific)
                 Emacs ------------------------------------------------
                 event (below this line is the generic event mechanism)
                   |
                   |
was there      if not, call
a SIGINT?   emacs_Xt_next_event()
    |              |
    |              |
    |              |
    V              V
    --->-------<----
           |
           |        [collected in event_stream_next_event();
           |         SIGINT is converted using maybe_read_quit_event()]
           V
         Emacs
         event
           |
           \---->------>----- maybe_kbd_translate() -->-----\
                                                            |
                                                            |
                                                            |
     command event queue                                    |
                                              if not from command
  (contains events that were                  event queue, call
  read earlier but not processed,             event_stream_next_event()
  typically when waiting in a                               |
  sit-for, sleep-for, etc. for                              |
 a particular event to be received)                         |
               |                                            |
               |                                            |
               V                                            V
               ---->----------------------------------<------
                                               |
                                               | [collected in
                                               |  next_event_internal()]
                                               |
 unread-     unread-       event from          |
 command-    command-       keyboard       else, call
 events      event           macro      next_event_internal()
   |           |               |               |
   |           |               |               |
   |           |               |               |
   V           V               V               V
   --------->----------------------<------------
                     |
                     |      [collected in `next-event', which may loop
                     |       more than once if the event it gets is on
                     |       a dead frame, device, etc.]
                     |
                     |
                     V
            feed into top-level event loop,
            which repeatedly calls `next-event'
            and then dispatches the event
            using `dispatch-event'
@end example

@node Specifics About the Emacs Event
@section Specifics About the Emacs Event

@node The Event Stream Callback Routines
@section The Event Stream Callback Routines

@node Other Event Loop Functions
@section Other Event Loop Functions

  @code{detect_input_pending()} and @code{input-pending-p} look for
input by calling @code{event_stream->event_pending_p} and looking in
@code{[V]unread-command-event} and the @code{command_event_queue} (they
do not check for an executing keyboard macro, though).

  @code{discard-input} cancels any command events pending (and any
keyboard macros currently executing), and puts the others onto the
@code{command_event_queue}.  There is a comment about a ``race
condition'', which is not a good sign.

  @code{next-command-event} and @code{read-char} are higher-level
interfaces to @code{next-event}.  @code{next-command-event} gets the
next @dfn{command} event (i.e.  keypress, mouse event, menu selection,
or scrollbar action), calling @code{dispatch-event} on any others.
@code{read-char} calls @code{next-command-event} and uses
@code{event_to_character()} to return the character equivalent.  With
the right kind of input method support, it is possible for (read-char)
to return a Kanji character.

@node Converting Events
@section Converting Events

  @code{character_to_event()}, @code{event_to_character()},
@code{event-to-character}, and @code{character-to-event} convert between
characters and keypress events corresponding to the characters.  If the
event was not a keypress, @code{event_to_character()} returns -1 and
@code{event-to-character} returns @code{nil}.  These functions convert
between character representation and the split-up event representation
(keysym plus mod keys).

@node Dispatching Events; The Command Builder
@section Dispatching Events; The Command Builder

Not yet documented.

@node Evaluation; Stack Frames; Bindings, Symbols and Variables, Events and the Event Loop, Top
@chapter Evaluation; Stack Frames; Bindings

@menu
* Evaluation::
* Dynamic Binding; The specbinding Stack; Unwind-Protects::
* Simple Special Forms::
* Catch and Throw::
@end menu

@node Evaluation
@section Evaluation

  @code{Feval()} evaluates the form (a Lisp object) that is passed to
it.  Note that evaluation is only non-trivial for two types of objects:
symbols and conses.  A symbol is evaluated simply by calling
@code{symbol-value} on it and returning the value.

  Evaluating a cons means calling a function.  First, @code{eval} checks
to see if garbage-collection is necessary, and calls
@code{garbage_collect_1()} if so.  It then increases the evaluation
depth by 1 (@code{lisp_eval_depth}, which is always less than
@code{max_lisp_eval_depth}) and adds an element to the linked list of
@code{struct backtrace}'s (@code{backtrace_list}).  Each such structure
contains a pointer to the function being called plus a list of the
function's arguments.  Originally these values are stored unevalled, and
as they are evaluated, the backtrace structure is updated.  Garbage
collection pays attention to the objects pointed to in the backtrace
structures (garbage collection might happen while a function is being
called or while an argument is being evaluated, and there could easily
be no other references to the arguments in the argument list; once an
argument is evaluated, however, the unevalled version is not needed by
eval, and so the backtrace structure is changed).

At this point, the function to be called is determined by looking at
the car of the cons (if this is a symbol, its function definition is
retrieved and the process repeated).  The function should then consist
of either a @code{Lisp_Subr} (built-in function written in C), a
@code{Lisp_Compiled_Function} object, or a cons whose car is one of the
symbols @code{autoload}, @code{macro} or @code{lambda}.

If the function is a @code{Lisp_Subr}, the lisp object points to a
@code{struct Lisp_Subr} (created by @code{DEFUN()}), which contains a
pointer to the C function, a minimum and maximum number of arguments
(or possibly the special constants @code{MANY} or @code{UNEVALLED}), a
pointer to the symbol referring to that subr, and a couple of other
things.  If the subr wants its arguments @code{UNEVALLED}, they are
passed raw as a list.  Otherwise, an array of evaluated arguments is
created and put into the backtrace structure, and either passed whole
(@code{MANY}) or each argument is passed as a C argument.

If the function is a @code{Lisp_Compiled_Function},
@code{funcall_compiled_function()} is called.  If the function is a
lambda list, @code{funcall_lambda()} is called.  If the function is a
macro, [..... fill in] is done.  If the function is an autoload,
@code{do_autoload()} is called to load the definition and then eval
starts over [explain this more].

When @code{Feval()} exits, the evaluation depth is reduced by one, the
debugger is called if appropriate, and the current backtrace structure
is removed from the list.

Both @code{funcall_compiled_function()} and @code{funcall_lambda()} need
to go through the list of formal parameters to the function and bind
them to the actual arguments, checking for @code{&rest} and
@code{&optional} symbols in the formal parameters and making sure the
number of actual arguments is correct.
@code{funcall_compiled_function()} can do this a little more
efficiently, since the formal parameter list can be checked for sanity
when the compiled function object is created.

@code{funcall_lambda()} simply calls @code{Fprogn} to execute the code
in the lambda list.

@code{funcall_compiled_function()} calls the real byte-code interpreter
@code{execute_optimized_program()} on the byte-code instructions, which
are converted into an internal form for faster execution.

When a compiled function is executed for the first time by
@code{funcall_compiled_function()}, or when it is @code{Fpurecopy()}ed
during the dump phase of building XEmacs, the byte-code instructions are
converted from a @code{Lisp_String} (which is inefficient to access,
especially in the presence of MULE) into a @code{Lisp_Opaque} object
containing an array of unsigned char, which can be directly executed by
the byte-code interpreter.  At this time the byte code is also analyzed
for validity and transformed into a more optimized form, so that
@code{execute_optimized_program()} can really fly.

Here are some of the optimizations performed by the internal byte-code
transformer:
@enumerate
@item
References to the @code{constants} array are checked for out-of-range
indices, so that the byte interpreter doesn't have to.
@item
References to the @code{constants} array that will be used as a Lisp
variable are checked for being correct non-constant (i.e. not @code{t},
@code{nil}, or @code{keywordp}) symbols, so that the byte interpreter
doesn't have to.
@item
The maxiumum number of variable bindings in the byte-code is
pre-computed, so that space on the @code{specpdl} stack can be
pre-reserved once for the whole function execution.
@item
All byte-code jumps are relative to the current program counter instead
of the start of the program, thereby saving a register.
@item
One-byte relative jumps are converted from the byte-code form of unsigned
chars offset by 127 to machine-friendly signed chars.
@end enumerate

Of course, this transformation of the @code{instructions} should not be
visible to the user, so @code{Fcompiled_function_instructions()} needs
to know how to convert the optimized opaque object back into a Lisp
string that is identical to the original string from the @file{.elc}
file.  (Actually, the resulting string may (rarely) contain slightly
different, yet equivalent, byte code.)

@code{Ffuncall()} implements Lisp @code{funcall}.  @code{(funcall fun
x1 x2 x3 ...)} is equivalent to @code{(eval (list fun (quote x1) (quote
x2) (quote x3) ...))}.  @code{Ffuncall()} contains its own code to do
the evaluation, however, and is very similar to @code{Feval()}.

From the performance point of view, it is worth knowing that most of the
time in Lisp evaluation is spent executing @code{Lisp_Subr} and
@code{Lisp_Compiled_Function} objects via @code{Ffuncall()} (not
@code{Feval()}).

@code{Fapply()} implements Lisp @code{apply}, which is very similar to
@code{funcall} except that if the last argument is a list, the result is the
same as if each of the arguments in the list had been passed separately.
@code{Fapply()} does some business to expand the last argument if it's a
list, then calls @code{Ffuncall()} to do the work.

@code{apply1()}, @code{call0()}, @code{call1()}, @code{call2()}, and
@code{call3()} call a function, passing it the argument(s) given (the
arguments are given as separate C arguments rather than being passed as
an array).  @code{apply1()} uses @code{Fapply()} while the others use
@code{Ffuncall()} to do the real work.

@node Dynamic Binding; The specbinding Stack; Unwind-Protects
@section Dynamic Binding; The specbinding Stack; Unwind-Protects

@example
struct specbinding
@{
  Lisp_Object symbol;
  Lisp_Object old_value;
  Lisp_Object (*func) (Lisp_Object); /* for unwind-protect */
@};
@end example

  @code{struct specbinding} is used for local-variable bindings and
unwind-protects.  @code{specpdl} holds an array of @code{struct specbinding}'s,
@code{specpdl_ptr} points to the beginning of the free bindings in the
array, @code{specpdl_size} specifies the total number of binding slots
in the array, and @code{max_specpdl_size} specifies the maximum number
of bindings the array can be expanded to hold.  @code{grow_specpdl()}
increases the size of the @code{specpdl} array, multiplying its size by
2 but never exceeding @code{max_specpdl_size} (except that if this
number is less than 400, it is first set to 400).

  @code{specbind()} binds a symbol to a value and is used for local
variables and @code{let} forms.  The symbol and its old value (which
might be @code{Qunbound}, indicating no prior value) are recorded in the
specpdl array, and @code{specpdl_size} is increased by 1.

  @code{record_unwind_protect()} implements an @dfn{unwind-protect},
which, when placed around a section of code, ensures that some specified
cleanup routine will be executed even if the code exits abnormally
(e.g. through a @code{throw} or quit).  @code{record_unwind_protect()}
simply adds a new specbinding to the @code{specpdl} array and stores the
appropriate information in it.  The cleanup routine can either be a C
function, which is stored in the @code{func} field, or a @code{progn}
form, which is stored in the @code{old_value} field.

  @code{unbind_to()} removes specbindings from the @code{specpdl} array
until the specified position is reached.  Each specbinding can be one of
three types:

@enumerate
@item
an unwind-protect with a C cleanup function (@code{func} is not 0, and
@code{old_value} holds an argument to be passed to the function);
@item
an unwind-protect with a Lisp form (@code{func} is 0, @code{symbol}
is @code{nil}, and @code{old_value} holds the form to be executed with
@code{Fprogn()}); or
@item
a local-variable binding (@code{func} is 0, @code{symbol} is not
@code{nil}, and @code{old_value} holds the old value, which is stored as
the symbol's value).
@end enumerate

@node Simple Special Forms
@section Simple Special Forms

@code{or}, @code{and}, @code{if}, @code{cond}, @code{progn},
@code{prog1}, @code{prog2}, @code{setq}, @code{quote}, @code{function},
@code{let*}, @code{let}, @code{while}

All of these are very simple and work as expected, calling
@code{Feval()} or @code{Fprogn()} as necessary and (in the case of
@code{let} and @code{let*}) using @code{specbind()} to create bindings
and @code{unbind_to()} to undo the bindings when finished.

Note that, with the exeption of @code{Fprogn}, these functions are
typically called in real life only in interpreted code, since the byte
compiler knows how to convert calls to these functions directly into
byte code.

@node Catch and Throw
@section Catch and Throw

@example
struct catchtag
@{
  Lisp_Object tag;
  Lisp_Object val;
  struct catchtag *next;
  struct gcpro *gcpro;
  jmp_buf jmp;
  struct backtrace *backlist;
  int lisp_eval_depth;
  int pdlcount;
@};
@end example

  @code{catch} is a Lisp function that places a catch around a body of
code.  A catch is a means of non-local exit from the code.  When a catch
is created, a tag is specified, and executing a @code{throw} to this tag
will exit from the body of code caught with this tag, and its value will
be the value given in the call to @code{throw}.  If there is no such
call, the code will be executed normally.

  Information pertaining to a catch is held in a @code{struct catchtag},
which is placed at the head of a linked list pointed to by
@code{catchlist}.  @code{internal_catch()} is passed a C function to
call (@code{Fprogn()} when Lisp @code{catch} is called) and arguments to
give it, and places a catch around the function.  Each @code{struct
catchtag} is held in the stack frame of the @code{internal_catch()}
instance that created the catch.

  @code{internal_catch()} is fairly straightforward.  It stores into the
@code{struct catchtag} the tag name and the current values of
@code{backtrace_list}, @code{lisp_eval_depth}, @code{gcprolist}, and the
offset into the @code{specpdl} array, sets a jump point with @code{_setjmp()}
(storing the jump point into the @code{struct catchtag}), and calls the
function.  Control will return to @code{internal_catch()} either when
the function exits normally or through a @code{_longjmp()} to this jump
point.  In the latter case, @code{throw} will store the value to be
returned into the @code{struct catchtag} before jumping.  When it's
done, @code{internal_catch()} removes the @code{struct catchtag} from
the catchlist and returns the proper value.

  @code{Fthrow()} goes up through the catchlist until it finds one with
a matching tag.  It then calls @code{unbind_catch()} to restore
everything to what it was when the appropriate catch was set, stores the
return value in the @code{struct catchtag}, and jumps (with
@code{_longjmp()}) to its jump point.

  @code{unbind_catch()} removes all catches from the catchlist until it
finds the correct one.  Some of the catches might have been placed for
error-trapping, and if so, the appropriate entries on the handlerlist
must be removed (see ``errors'').  @code{unbind_catch()} also restores
the values of @code{gcprolist}, @code{backtrace_list}, and
@code{lisp_eval}, and calls @code{unbind_to()} to undo any specbindings
created since the catch.


@node Symbols and Variables, Buffers and Textual Representation, Evaluation; Stack Frames; Bindings, Top
@chapter Symbols and Variables

@menu
* Introduction to Symbols::
* Obarrays::
* Symbol Values::
@end menu

@node Introduction to Symbols
@section Introduction to Symbols

  A symbol is basically just an object with four fields: a name (a
string), a value (some Lisp object), a function (some Lisp object), and
a property list (usually a list of alternating keyword/value pairs).
What makes symbols special is that there is usually only one symbol with
a given name, and the symbol is referred to by name.  This makes a
symbol a convenient way of calling up data by name, i.e. of implementing
variables. (The variable's value is stored in the @dfn{value slot}.)
Similarly, functions are referenced by name, and the definition of the
function is stored in a symbol's @dfn{function slot}.  This means that
there can be a distinct function and variable with the same name.  The
property list is used as a more general mechanism of associating
additional values with particular names, and once again the namespace is
independent of the function and variable namespaces.

@node Obarrays
@section Obarrays

  The identity of symbols with their names is accomplished through a
structure called an obarray, which is just a poorly-implemented hash
table mapping from strings to symbols whose name is that string. (I say
``poorly implemented'' because an obarray appears in Lisp as a vector
with some hidden fields rather than as its own opaque type.  This is an
Emacs Lisp artifact that should be fixed.)

  Obarrays are implemented as a vector of some fixed size (which should
be a prime for best results), where each ``bucket'' of the vector
contains one or more symbols, threaded through a hidden @code{next}
field in the symbol.  Lookup of a symbol in an obarray, and adding a
symbol to an obarray, is accomplished through standard hash-table
techniques.

  The standard Lisp function for working with symbols and obarrays is
@code{intern}.  This looks up a symbol in an obarray given its name; if
it's not found, a new symbol is automatically created with the specified
name, added to the obarray, and returned.  This is what happens when the
Lisp reader encounters a symbol (or more precisely, encounters the name
of a symbol) in some text that it is reading.  There is a standard
obarray called @code{obarray} that is used for this purpose, although
the Lisp programmer is free to create his own obarrays and @code{intern}
symbols in them.

  Note that, once a symbol is in an obarray, it stays there until
something is done about it, and the standard obarray @code{obarray}
always stays around, so once you use any particular variable name, a
corresponding symbol will stay around in @code{obarray} until you exit
XEmacs.

  Note that @code{obarray} itself is a variable, and as such there is a
symbol in @code{obarray} whose name is @code{"obarray"} and which
contains @code{obarray} as its value.

  Note also that this call to @code{intern} occurs only when in the Lisp
reader, not when the code is executed (at which point the symbol is
already around, stored as such in the definition of the function).

  You can create your own obarray using @code{make-vector} (this is
horrible but is an artifact) and intern symbols into that obarray.
Doing that will result in two or more symbols with the same name.
However, at most one of these symbols is in the standard @code{obarray}:
You cannot have two symbols of the same name in any particular obarray.
Note that you cannot add a symbol to an obarray in any fashion other
than using @code{intern}: i.e. you can't take an existing symbol and put
it in an existing obarray.  Nor can you change the name of an existing
symbol. (Since obarrays are vectors, you can violate the consistency of
things by storing directly into the vector, but let's ignore that
possibility.)

  Usually symbols are created by @code{intern}, but if you really want,
you can explicitly create a symbol using @code{make-symbol}, giving it
some name.  The resulting symbol is not in any obarray (i.e. it is
@dfn{uninterned}), and you can't add it to any obarray.  Therefore its
primary purpose is as a symbol to use in macros to avoid namespace
pollution.  It can also be used as a carrier of information, but cons
cells could probably be used just as well.

  You can also use @code{intern-soft} to look up a symbol but not create
a new one, and @code{unintern} to remove a symbol from an obarray.  This
returns the removed symbol. (Remember: You can't put the symbol back
into any obarray.) Finally, @code{mapatoms} maps over all of the symbols
in an obarray.

@node Symbol Values
@section Symbol Values

  The value field of a symbol normally contains a Lisp object.  However,
a symbol can be @dfn{unbound}, meaning that it logically has no value.
This is internally indicated by storing a special Lisp object, called
@dfn{the unbound marker} and stored in the global variable
@code{Qunbound}.  The unbound marker is of a special Lisp object type
called @dfn{symbol-value-magic}.  It is impossible for the Lisp
programmer to directly create or access any object of this type.

  @strong{You must not let any ``symbol-value-magic'' object escape to
the Lisp level.}  Printing any of these objects will cause the message
@samp{INTERNAL EMACS BUG} to appear as part of the print representation.
(You may see this normally when you call @code{debug_print()} from the
debugger on a Lisp object.) If you let one of these objects escape to
the Lisp level, you will violate a number of assumptions contained in
the C code and make the unbound marker not function right.

  When a symbol is created, its value field (and function field) are set
to @code{Qunbound}.  The Lisp programmer can restore these conditions
later using @code{makunbound} or @code{fmakunbound}, and can query to
see whether the value of function fields are @dfn{bound} (i.e. have a
value other than @code{Qunbound}) using @code{boundp} and
@code{fboundp}.  The fields are set to a normal Lisp object using
@code{set} (or @code{setq}) and @code{fset}.

  Other symbol-value-magic objects are used as special markers to
indicate variables that have non-normal properties.  This includes any
variables that are tied into C variables (setting the variable magically
sets some global variable in the C code, and likewise for retrieving the
variable's value), variables that magically tie into slots in the
current buffer, variables that are buffer-local, etc.  The
symbol-value-magic object is stored in the value cell in place of
a normal object, and the code to retrieve a symbol's value
(i.e. @code{symbol-value}) knows how to do special things with them.
This means that you should not just fetch the value cell directly if you
want a symbol's value.

  The exact workings of this are rather complex and involved and are
well-documented in comments in @file{buffer.c}, @file{symbols.c}, and
@file{lisp.h}.

@node Buffers and Textual Representation, MULE Character Sets and Encodings, Symbols and Variables, Top
@chapter Buffers and Textual Representation

@menu
* Introduction to Buffers::     A buffer holds a block of text such as a file.
* The Text in a Buffer::        Representation of the text in a buffer.
* Buffer Lists::                Keeping track of all buffers.
* Markers and Extents::         Tagging locations within a buffer.
* Bufbytes and Emchars::        Representation of individual characters.
* The Buffer Object::           The Lisp object corresponding to a buffer.
@end menu

@node Introduction to Buffers
@section Introduction to Buffers

  A buffer is logically just a Lisp object that holds some text.
In this, it is like a string, but a buffer is optimized for
frequent insertion and deletion, while a string is not.  Furthermore:

@enumerate
@item
Buffers are @dfn{permanent} objects, i.e. once you create them, they
remain around, and need to be explicitly deleted before they go away.
@item
Each buffer has a unique name, which is a string.  Buffers are
normally referred to by name.  In this respect, they are like
symbols.
@item
Buffers have a default insertion position, called @dfn{point}.
Inserting text (unless you explicitly give a position) goes at point,
and moves point forward past the text.  This is what is going on when
you type text into Emacs.
@item
Buffers have lots of extra properties associated with them.
@item
Buffers can be @dfn{displayed}.  What this means is that there
exist a number of @dfn{windows}, which are objects that correspond
to some visible section of your display, and each window has
an associated buffer, and the current contents of the buffer
are shown in that section of the display.  The redisplay mechanism
(which takes care of doing this) knows how to look at the
text of a buffer and come up with some reasonable way of displaying
this.  Many of the properties of a buffer control how the
buffer's text is displayed.
@item
One buffer is distinguished and called the @dfn{current buffer}.  It is
stored in the variable @code{current_buffer}.  Buffer operations operate
on this buffer by default.  When you are typing text into a buffer, the
buffer you are typing into is always @code{current_buffer}.  Switching
to a different window changes the current buffer.  Note that Lisp code
can temporarily change the current buffer using @code{set-buffer} (often
enclosed in a @code{save-excursion} so that the former current buffer
gets restored when the code is finished).  However, calling
@code{set-buffer} will NOT cause a permanent change in the current
buffer.  The reason for this is that the top-level event loop sets
@code{current_buffer} to the buffer of the selected window, each time
it finishes executing a user command.
@end enumerate

  Make sure you understand the distinction between @dfn{current buffer}
and @dfn{buffer of the selected window}, and the distinction between
@dfn{point} of the current buffer and @dfn{window-point} of the selected
window. (This latter distinction is explained in detail in the section
on windows.)

@node The Text in a Buffer
@section The Text in a Buffer

  The text in a buffer consists of a sequence of zero or more
characters.  A @dfn{character} is an integer that logically represents
a letter, number, space, or other unit of text.  Most of the characters
that you will typically encounter belong to the ASCII set of characters,
but there are also characters for various sorts of accented letters,
special symbols, Chinese and Japanese ideograms (i.e. Kanji, Katakana,
etc.), Cyrillic and Greek letters, etc.  The actual number of possible
characters is quite large.

  For now, we can view a character as some non-negative integer that
has some shape that defines how it typically appears (e.g. as an
uppercase A). (The exact way in which a character appears depends on the
font used to display the character.) The internal type of characters in
the C code is an @code{Emchar}; this is just an @code{int}, but using a
symbolic type makes the code clearer.

  Between every character in a buffer is a @dfn{buffer position} or
@dfn{character position}.  We can speak of the character before or after
a particular buffer position, and when you insert a character at a
particular position, all characters after that position end up at new
positions.  When we speak of the character @dfn{at} a position, we
really mean the character after the position.  (This schizophrenia
between a buffer position being ``between'' a character and ``on'' a
character is rampant in Emacs.)

  Buffer positions are numbered starting at 1.  This means that
position 1 is before the first character, and position 0 is not
valid.  If there are N characters in a buffer, then buffer
position N+1 is after the last one, and position N+2 is not valid.

  The internal makeup of the Emchar integer varies depending on whether
we have compiled with MULE support.  If not, the Emchar integer is an
8-bit integer with possible values from 0 - 255.  0 - 127 are the
standard ASCII characters, while 128 - 255 are the characters from the
ISO-8859-1 character set.  If we have compiled with MULE support, an
Emchar is a 19-bit integer, with the various bits having meanings
according to a complex scheme that will be detailed later.  The
characters numbered 0 - 255 still have the same meanings as for the
non-MULE case, though.

  Internally, the text in a buffer is represented in a fairly simple
fashion: as a contiguous array of bytes, with a @dfn{gap} of some size
in the middle.  Although the gap is of some substantial size in bytes,
there is no text contained within it: From the perspective of the text
in the buffer, it does not exist.  The gap logically sits at some buffer
position, between two characters (or possibly at the beginning or end of
the buffer).  Insertion of text in a buffer at a particular position is
always accomplished by first moving the gap to that position
(i.e. through some block moving of text), then writing the text into the
beginning of the gap, thereby shrinking the gap.  If the gap shrinks
down to nothing, a new gap is created. (What actually happens is that a
new gap is ``created'' at the end of the buffer's text, which requires
nothing more than changing a couple of indices; then the gap is
``moved'' to the position where the insertion needs to take place by
moving up in memory all the text after that position.)  Similarly,
deletion occurs by moving the gap to the place where the text is to be
deleted, and then simply expanding the gap to include the deleted text.
(@dfn{Expanding} and @dfn{shrinking} the gap as just described means
just that the internal indices that keep track of where the gap is
located are changed.)

  Note that the total amount of memory allocated for a buffer text never
decreases while the buffer is live.  Therefore, if you load up a
20-megabyte file and then delete all but one character, there will be a
20-megabyte gap, which won't get any smaller (except by inserting
characters back again).  Once the buffer is killed, the memory allocated
for the buffer text will be freed, but it will still be sitting on the
heap, taking up virtual memory, and will not be released back to the
operating system. (However, if you have compiled XEmacs with rel-alloc,
the situation is different.  In this case, the space @emph{will} be
released back to the operating system.  However, this tends to result in a
noticeable speed penalty.)

  Astute readers may notice that the text in a buffer is represented as
an array of @emph{bytes}, while (at least in the MULE case) an Emchar is
a 19-bit integer, which clearly cannot fit in a byte.  This means (of
course) that the text in a buffer uses a different representation from
an Emchar: specifically, the 19-bit Emchar becomes a series of one to
four bytes.  The conversion between these two representations is complex
and will be described later.

  In the non-MULE case, everything is very simple: An Emchar
is an 8-bit value, which fits neatly into one byte.

  If we are given a buffer position and want to retrieve the
character at that position, we need to follow these steps:

@enumerate
@item
Pretend there's no gap, and convert the buffer position into a @dfn{byte
index} that indexes to the appropriate byte in the buffer's stream of
textual bytes.  By convention, byte indices begin at 1, just like buffer
positions.  In the non-MULE case, byte indices and buffer positions are
identical, since one character equals one byte.
@item
Convert the byte index into a @dfn{memory index}, which takes the gap
into account.  The memory index is a direct index into the block of
memory that stores the text of a buffer.  This basically just involves
checking to see if the byte index is past the gap, and if so, adding the
size of the gap to it.  By convention, memory indices begin at 1, just
like buffer positions and byte indices, and when referring to the
position that is @dfn{at} the gap, we always use the memory position at
the @emph{beginning}, not at the end, of the gap.
@item
Fetch the appropriate bytes at the determined memory position.
@item
Convert these bytes into an Emchar.
@end enumerate

  In the non-Mule case, (3) and (4) boil down to a simple one-byte
memory access.

  Note that we have defined three types of positions in a buffer:

@enumerate
@item
@dfn{buffer positions} or @dfn{character positions}, typedef @code{Bufpos}
@item
@dfn{byte indices}, typedef @code{Bytind}
@item
@dfn{memory indices}, typedef @code{Memind}
@end enumerate

  All three typedefs are just @code{int}s, but defining them this way makes
things a lot clearer.

  Most code works with buffer positions.  In particular, all Lisp code
that refers to text in a buffer uses buffer positions.  Lisp code does
not know that byte indices or memory indices exist.

  Finally, we have a typedef for the bytes in a buffer.  This is a
@code{Bufbyte}, which is an unsigned char.  Referring to them as
Bufbytes underscores the fact that we are working with a string of bytes
in the internal Emacs buffer representation rather than in one of a
number of possible alternative representations (e.g. EUC-encoded text,
etc.).

@node Buffer Lists
@section Buffer Lists

  Recall earlier that buffers are @dfn{permanent} objects, i.e.  that
they remain around until explicitly deleted.  This entails that there is
a list of all the buffers in existence.  This list is actually an
assoc-list (mapping from the buffer's name to the buffer) and is stored
in the global variable @code{Vbuffer_alist}.

  The order of the buffers in the list is important: the buffers are
ordered approximately from most-recently-used to least-recently-used.
Switching to a buffer using @code{switch-to-buffer},
@code{pop-to-buffer}, etc. and switching windows using
@code{other-window}, etc.  usually brings the new current buffer to the
front of the list.  @code{switch-to-buffer}, @code{other-buffer},
etc. look at the beginning of the list to find an alternative buffer to
suggest.  You can also explicitly move a buffer to the end of the list
using @code{bury-buffer}.

  In addition to the global ordering in @code{Vbuffer_alist}, each frame
has its own ordering of the list.  These lists always contain the same
elements as in @code{Vbuffer_alist} although possibly in a different
order.  @code{buffer-list} normally returns the list for the selected
frame.  This allows you to work in separate frames without things
interfering with each other.

  The standard way to look up a buffer given a name is
@code{get-buffer}, and the standard way to create a new buffer is
@code{get-buffer-create}, which looks up a buffer with a given name,
creating a new one if necessary.  These operations correspond exactly
with the symbol operations @code{intern-soft} and @code{intern},
respectively.  You can also force a new buffer to be created using
@code{generate-new-buffer}, which takes a name and (if necessary) makes
a unique name from this by appending a number, and then creates the
buffer.  This is basically like the symbol operation @code{gensym}.

@node Markers and Extents
@section Markers and Extents

  Among the things associated with a buffer are things that are
logically attached to certain buffer positions.  This can be used to
keep track of a buffer position when text is inserted and deleted, so
that it remains at the same spot relative to the text around it; to
assign properties to particular sections of text; etc.  There are two
such objects that are useful in this regard: they are @dfn{markers} and
@dfn{extents}.

  A @dfn{marker} is simply a flag placed at a particular buffer
position, which is moved around as text is inserted and deleted.
Markers are used for all sorts of purposes, such as the @code{mark} that
is the other end of textual regions to be cut, copied, etc.

  An @dfn{extent} is similar to two markers plus some associated
properties, and is used to keep track of regions in a buffer as text is
inserted and deleted, and to add properties (e.g. fonts) to particular
regions of text.  The external interface of extents is explained
elsewhere.

  The important thing here is that markers and extents simply contain
buffer positions in them as integers, and every time text is inserted or
deleted, these positions must be updated.  In order to minimize the
amount of shuffling that needs to be done, the positions in markers and
extents (there's one per marker, two per extent) and stored in Meminds.
This means that they only need to be moved when the text is physically
moved in memory; since the gap structure tries to minimize this, it also
minimizes the number of marker and extent indices that need to be
adjusted.  Look in @file{insdel.c} for the details of how this works.

  One other important distinction is that markers are @dfn{temporary}
while extents are @dfn{permanent}.  This means that markers disappear as
soon as there are no more pointers to them, and correspondingly, there
is no way to determine what markers are in a buffer if you are just
given the buffer.  Extents remain in a buffer until they are detached
(which could happen as a result of text being deleted) or the buffer is
deleted, and primitives do exist to enumerate the extents in a buffer.

@node Bufbytes and Emchars
@section Bufbytes and Emchars

  Not yet documented.

@node The Buffer Object
@section The Buffer Object

  Buffers contain fields not directly accessible by the Lisp programmer.
We describe them here, naming them by the names used in the C code.
Many are accessible indirectly in Lisp programs via Lisp primitives.

@table @code
@item name
The buffer name is a string that names the buffer.  It is guaranteed to
be unique.  @xref{Buffer Names,,, lispref, XEmacs Lisp Programmer's
Manual}.

@item save_modified
This field contains the time when the buffer was last saved, as an
integer.  @xref{Buffer Modification,,, lispref, XEmacs Lisp Programmer's
Manual}.

@item modtime
This field contains the modification time of the visited file.  It is
set when the file is written or read.  Every time the buffer is written
to the file, this field is compared to the modification time of the
file.  @xref{Buffer Modification,,, lispref, XEmacs Lisp Programmer's
Manual}.

@item auto_save_modified
This field contains the time when the buffer was last auto-saved.

@item last_window_start
This field contains the @code{window-start} position in the buffer as of
the last time the buffer was displayed in a window.

@item undo_list
This field points to the buffer's undo list.  @xref{Undo,,, lispref,
XEmacs Lisp Programmer's Manual}.

@item syntax_table_v
This field contains the syntax table for the buffer.  @xref{Syntax
Tables,,, lispref, XEmacs Lisp Programmer's Manual}.

@item downcase_table
This field contains the conversion table for converting text to lower
case.  @xref{Case Tables,,, lispref, XEmacs Lisp Programmer's Manual}.

@item upcase_table
This field contains the conversion table for converting text to upper
case.  @xref{Case Tables,,, lispref, XEmacs Lisp Programmer's Manual}.

@item case_canon_table
This field contains the conversion table for canonicalizing text for
case-folding search.  @xref{Case Tables,,, lispref, XEmacs Lisp
Programmer's Manual}.

@item case_eqv_table
This field contains the equivalence table for case-folding search.
@xref{Case Tables,,, lispref, XEmacs Lisp Programmer's Manual}.

@item display_table
This field contains the buffer's display table, or @code{nil} if it
doesn't have one.  @xref{Display Tables,,, lispref, XEmacs Lisp
Programmer's Manual}.

@item markers
This field contains the chain of all markers that currently point into
the buffer.  Deletion of text in the buffer, and motion of the buffer's
gap, must check each of these markers and perhaps update it.
@xref{Markers,,, lispref, XEmacs Lisp Programmer's Manual}.

@item backed_up
This field is a flag that tells whether a backup file has been made for
the visited file of this buffer.

@item mark
This field contains the mark for the buffer.  The mark is a marker,
hence it is also included on the list @code{markers}.  @xref{The Mark,,,
lispref, XEmacs Lisp Programmer's Manual}.

@item mark_active
This field is non-@code{nil} if the buffer's mark is active.

@item local_var_alist
This field contains the association list describing the variables local
in this buffer, and their values, with the exception of local variables
that have special slots in the buffer object.  (Those slots are omitted
from this table.)  @xref{Buffer-Local Variables,,, lispref, XEmacs Lisp
Programmer's Manual}.

@item modeline_format
This field contains a Lisp object which controls how to display the mode
line for this buffer.  @xref{Modeline Format,,, lispref, XEmacs Lisp
Programmer's Manual}.

@item base_buffer
This field holds the buffer's base buffer (if it is an indirect buffer),
or @code{nil}.
@end table

@node MULE Character Sets and Encodings, The Lisp Reader and Compiler, Buffers and Textual Representation, Top
@chapter MULE Character Sets and Encodings

  Recall that there are two primary ways that text is represented in
XEmacs.  The @dfn{buffer} representation sees the text as a series of
bytes (Bufbytes), with a variable number of bytes used per character.
The @dfn{character} representation sees the text as a series of integers
(Emchars), one per character.  The character representation is a cleaner
representation from a theoretical standpoint, and is thus used in many
cases when lots of manipulations on a string need to be done.  However,
the buffer representation is the standard representation used in both
Lisp strings and buffers, and because of this, it is the ``default''
representation that text comes in.  The reason for using this
representation is that it's compact and is compatible with ASCII.

@menu
* Character Sets::
* Encodings::
* Internal Mule Encodings::
* CCL::
@end menu

@node Character Sets
@section Character Sets

  A character set (or @dfn{charset}) is an ordered set of characters.  A
particular character in a charset is indexed using one or more
@dfn{position codes}, which are non-negative integers.  The number of
position codes needed to identify a particular character in a charset is
called the @dfn{dimension} of the charset.  In XEmacs/Mule, all charsets
have dimension 1 or 2, and the size of all charsets (except for a few
special cases) is either 94, 96, 94 by 94, or 96 by 96.  The range of
position codes used to index characters from any of these types of
character sets is as follows:

@example
Charset type            Position code 1         Position code 2
------------------------------------------------------------
94                      33 - 126                N/A
96                      32 - 127                N/A
94x94                   33 - 126                33 - 126
96x96                   32 - 127                32 - 127
@end example

  Note that in the above cases position codes do not start at an
expected value such as 0 or 1.  The reason for this will become clear
later.

  For example, Latin-1 is a 96-character charset, and JISX0208 (the
Japanese national character set) is a 94x94-character charset.

  [Note that, although the ranges above define the @emph{valid} position
codes for a charset, some of the slots in a particular charset may in
fact be empty.  This is the case for JISX0208, for example, where (e.g.)
all the slots whose first position code is in the range 118 - 127 are
empty.]

  There are three charsets that do not follow the above rules.  All of
them have one dimension, and have ranges of position codes as follows:

@example
Charset name            Position code 1
------------------------------------
ASCII                   0 - 127
Control-1               0 - 31
Composite               0 - some large number
@end example

  (The upper bound of the position code for composite characters has not
yet been determined, but it will probably be at least 16,383).

  ASCII is the union of two subsidiary character sets: Printing-ASCII
(the printing ASCII character set, consisting of position codes 33 -
126, like for a standard 94-character charset) and Control-ASCII (the
non-printing characters that would appear in a binary file with codes 0
- 32 and 127).

  Control-1 contains the non-printing characters that would appear in a
binary file with codes 128 - 159.

  Composite contains characters that are generated by overstriking one
or more characters from other charsets.

  Note that some characters in ASCII, and all characters in Control-1,
are @dfn{control} (non-printing) characters.  These have no printed
representation but instead control some other function of the printing
(e.g. TAB or 8 moves the current character position to the next tab
stop).  All other characters in all charsets are @dfn{graphic}
(printing) characters.

  When a binary file is read in, the bytes in the file are assigned to
character sets as follows:

@example
Bytes           Character set           Range
--------------------------------------------------
0 - 127         ASCII                   0 - 127
128 - 159       Control-1               0 - 31
160 - 255       Latin-1                 32 - 127
@end example

  This is a bit ad-hoc but gets the job done.

@node Encodings
@section Encodings

  An @dfn{encoding} is a way of numerically representing characters from
one or more character sets.  If an encoding only encompasses one
character set, then the position codes for the characters in that
character set could be used directly.  This is not possible, however, if
more than one character set is to be used in the encoding.

  For example, the conversion detailed above between bytes in a binary
file and characters is effectively an encoding that encompasses the
three character sets ASCII, Control-1, and Latin-1 in a stream of 8-bit
bytes.

  Thus, an encoding can be viewed as a way of encoding characters from a
specified group of character sets using a stream of bytes, each of which
contains a fixed number of bits (but not necessarily 8, as in the common
usage of ``byte'').

  Here are descriptions of a couple of common
encodings:

@menu
* Japanese EUC (Extended Unix Code)::
* JIS7::
@end menu

@node Japanese EUC (Extended Unix Code)
@subsection Japanese EUC (Extended Unix Code)

This encompasses the character sets Printing-ASCII, Japanese-JISX0201,
and Japanese-JISX0208-Kana (half-width katakana, the right half of
JISX0201).  It uses 8-bit bytes.

Note that Printing-ASCII and Japanese-JISX0201-Kana are 94-character
charsets, while Japanese-JISX0208 is a 94x94-character charset.

The encoding is as follows:

@example
Character set            Representation (PC=position-code)
-------------            --------------
Printing-ASCII           PC1
Japanese-JISX0201-Kana   0x8E       | PC1 + 0x80
Japanese-JISX0208        PC1 + 0x80 | PC2 + 0x80
Japanese-JISX0212        PC1 + 0x80 | PC2 + 0x80
@end example


@node JIS7
@subsection JIS7

This encompasses the character sets Printing-ASCII,
Japanese-JISX0201-Roman (the left half of JISX0201; this character set
is very similar to Printing-ASCII and is a 94-character charset),
Japanese-JISX0208, and Japanese-JISX0201-Kana.  It uses 7-bit bytes.

Unlike Japanese EUC, this is a @dfn{modal} encoding, which
means that there are multiple states that the encoding can
be in, which affect how the bytes are to be interpreted.
Special sequences of bytes (called @dfn{escape sequences})
are used to change states.

  The encoding is as follows:

@example
Character set              Representation (PC=position-code)
-------------              --------------
Printing-ASCII             PC1
Japanese-JISX0201-Roman    PC1
Japanese-JISX0201-Kana     PC1
Japanese-JISX0208          PC1 PC2


Escape sequence   ASCII equivalent   Meaning
---------------   ----------------   -------
0x1B 0x28 0x4A    ESC ( J            invoke Japanese-JISX0201-Roman
0x1B 0x28 0x49    ESC ( I            invoke Japanese-JISX0201-Kana
0x1B 0x24 0x42    ESC $ B            invoke Japanese-JISX0208
0x1B 0x28 0x42    ESC ( B            invoke Printing-ASCII
@end example

  Initially, Printing-ASCII is invoked.

@node Internal Mule Encodings
@section Internal Mule Encodings

In XEmacs/Mule, each character set is assigned a unique number, called a
@dfn{leading byte}.  This is used in the encodings of a character.
Leading bytes are in the range 0x80 - 0xFF (except for ASCII, which has
a leading byte of 0), although some leading bytes are reserved.

Charsets whose leading byte is in the range 0x80 - 0x9F are called
@dfn{official} and are used for built-in charsets.  Other charsets are
called @dfn{private} and have leading bytes in the range 0xA0 - 0xFF;
these are user-defined charsets.

  More specifically:

@example
Character set           Leading byte
-------------           ------------
ASCII                   0
Composite               0x80
Dimension-1 Official    0x81 - 0x8D
                          (0x8E is free)
Control-1               0x8F
Dimension-2 Official    0x90 - 0x99
                          (0x9A - 0x9D are free;
                           0x9E and 0x9F are reserved)
Dimension-1 Private     0xA0 - 0xEF
Dimension-2 Private     0xF0 - 0xFF
@end example

There are two internal encodings for characters in XEmacs/Mule.  One is
called @dfn{string encoding} and is an 8-bit encoding that is used for
representing characters in a buffer or string.  It uses 1 to 4 bytes per
character.  The other is called @dfn{character encoding} and is a 19-bit
encoding that is used for representing characters individually in a
variable.

(In the following descriptions, we'll ignore composite characters for
the moment.  We also give a general (structural) overview first,
followed later by the exact details.)

@menu
* Internal String Encoding::
* Internal Character Encoding::
@end menu

@node Internal String Encoding
@subsection Internal String Encoding

ASCII characters are encoded using their position code directly.  Other
characters are encoded using their leading byte followed by their
position code(s) with the high bit set.  Characters in private character
sets have their leading byte prefixed with a @dfn{leading byte prefix},
which is either 0x9E or 0x9F. (No character sets are ever assigned these
leading bytes.) Specifically:

@example
Character set           Encoding (PC=position-code, LB=leading-byte)
-------------           --------
ASCII                   PC-1 |
Control-1               LB   |  PC1 + 0xA0 |
Dimension-1 official    LB   |  PC1 + 0x80 |
Dimension-1 private     0x9E |  LB         | PC1 + 0x80 |
Dimension-2 official    LB   |  PC1 + 0x80 | PC2 + 0x80 |
Dimension-2 private     0x9F |  LB         | PC1 + 0x80 | PC2 + 0x80
@end example

  The basic characteristic of this encoding is that the first byte
of all characters is in the range 0x00 - 0x9F, and the second and
following bytes of all characters is in the range 0xA0 - 0xFF.
This means that it is impossible to get out of sync, or more
specifically:

@enumerate
@item
Given any byte position, the beginning of the character it is
within can be determined in constant time.
@item
Given any byte position at the beginning of a character, the
beginning of the next character can be determined in constant
time.
@item
Given any byte position at the beginning of a character, the
beginning of the previous character can be determined in constant
time.
@item
Textual searches can simply treat encoded strings as if they
were encoded in a one-byte-per-character fashion rather than
the actual multi-byte encoding.
@end enumerate

  None of the standard non-modal encodings meet all of these
conditions.  For example, EUC satisfies only (2) and (3), while
Shift-JIS and Big5 (not yet described) satisfy only (2). (All
non-modal encodings must satisfy (2), in order to be unambiguous.)

@node Internal Character Encoding
@subsection Internal Character Encoding

  One 19-bit word represents a single character.  The word is
separated into three fields:

@example
Bit number:     18 17 16 15 14 13 12 11 10 09 08 07 06 05 04 03 02 01 00
                <------------> <------------------> <------------------>
Field:                1                  2                    3
@end example

  Note that fields 2 and 3 hold 7 bits each, while field 1 holds 5 bits.

@example
Character set           Field 1         Field 2         Field 3
-------------           -------         -------         -------
ASCII                      0               0              PC1
   range:                                                   (00 - 7F)
Control-1                  0               1              PC1
   range:                                                   (00 - 1F)
Dimension-1 official       0            LB - 0x80         PC1
   range:                                    (01 - 0D)      (20 - 7F)
Dimension-1 private        0            LB - 0x80         PC1
   range:                                    (20 - 6F)      (20 - 7F)
Dimension-2 official    LB - 0x8F         PC1             PC2
   range:                    (01 - 0A)       (20 - 7F)      (20 - 7F)
Dimension-2 private     LB - 0xE1         PC1             PC2
   range:                    (0F - 1E)       (20 - 7F)      (20 - 7F)
Composite                 0x1F             ?               ?
@end example

  Note that character codes 0 - 255 are the same as the ``binary encoding''
described above.

@node CCL
@section CCL

@example
CCL PROGRAM SYNTAX:
     CCL_PROGRAM := (CCL_MAIN_BLOCK
                     [ CCL_EOF_BLOCK ])

     CCL_MAIN_BLOCK := CCL_BLOCK
     CCL_EOF_BLOCK := CCL_BLOCK

     CCL_BLOCK := STATEMENT | (STATEMENT [STATEMENT ...])
     STATEMENT :=
             SET | IF | BRANCH | LOOP | REPEAT | BREAK
             | READ | WRITE

     SET := (REG = EXPRESSION) | (REG SELF_OP EXPRESSION)
            | INT-OR-CHAR

     EXPRESSION := ARG | (EXPRESSION OP ARG)

     IF := (if EXPRESSION CCL_BLOCK CCL_BLOCK)
     BRANCH := (branch EXPRESSION CCL_BLOCK [CCL_BLOCK ...])
     LOOP := (loop STATEMENT [STATEMENT ...])
     BREAK := (break)
     REPEAT := (repeat)
             | (write-repeat [REG | INT-OR-CHAR | string])
             | (write-read-repeat REG [INT-OR-CHAR | string | ARRAY]?)
     READ := (read REG) | (read REG REG)
             | (read-if REG ARITH_OP ARG CCL_BLOCK CCL_BLOCK)
             | (read-branch REG CCL_BLOCK [CCL_BLOCK ...])
     WRITE := (write REG) | (write REG REG)
             | (write INT-OR-CHAR) | (write STRING) | STRING
             | (write REG ARRAY)
     END := (end)

     REG := r0 | r1 | r2 | r3 | r4 | r5 | r6 | r7
     ARG := REG | INT-OR-CHAR
     OP :=   + | - | * | / | % | & | '|' | ^ | << | >> | <8 | >8 | //
             | < | > | == | <= | >= | !=
     SELF_OP :=
             += | -= | *= | /= | %= | &= | '|=' | ^= | <<= | >>=
     ARRAY := '[' INT-OR-CHAR ... ']'
     INT-OR-CHAR := INT | CHAR

MACHINE CODE:

The machine code consists of a vector of 32-bit words.
The first such word specifies the start of the EOF section of the code;
this is the code executed to handle any stuff that needs to be done
(e.g. designating back to ASCII and left-to-right mode) after all
other encoded/decoded data has been written out.  This is not used for
charset CCL programs.

REGISTER: 0..7  -- refered by RRR or rrr

OPERATOR BIT FIELD (27-bit): XXXXXXXXXXXXXXX RRR TTTTT
        TTTTT (5-bit): operator type
        RRR (3-bit): register number
        XXXXXXXXXXXXXXXX (15-bit):
                CCCCCCCCCCCCCCC: constant or address
                000000000000rrr: register number

AAAA:   00000 +
        00001 -
        00010 *
        00011 /
        00100 %
        00101 &
        00110 |
        00111 ~

        01000 <<
        01001 >>
        01010 <8
        01011 >8
        01100 //
        01101 not used
        01110 not used
        01111 not used

        10000 <
        10001 >
        10010 ==
        10011 <=
        10100 >=
        10101 !=

OPERATORS:      TTTTT RRR XX..

SetCS:          00000 RRR C...C      RRR = C...C
SetCL:          00001 RRR .....      RRR = c...c
                c.............c
SetR:           00010 RRR ..rrr      RRR = rrr
SetA:           00011 RRR ..rrr      RRR = array[rrr]
                C.............C      size of array = C...C
                c.............c      contents = c...c

Jump:           00100 000 c...c      jump to c...c
JumpCond:       00101 RRR c...c      if (!RRR) jump to c...c
WriteJump:      00110 RRR c...c      Write1 RRR, jump to c...c
WriteReadJump:  00111 RRR c...c      Write1, Read1 RRR, jump to c...c
WriteCJump:     01000 000 c...c      Write1 C...C, jump to c...c
                C...C
WriteCReadJump: 01001 RRR c...c      Write1 C...C, Read1 RRR,
                C.............C      and jump to c...c
WriteSJump:     01010 000 c...c      WriteS, jump to c...c
                C.............C
                S.............S
                ...
WriteSReadJump: 01011 RRR c...c      WriteS, Read1 RRR, jump to c...c
                C.............C
                S.............S
                ...
WriteAReadJump: 01100 RRR c...c      WriteA, Read1 RRR, jump to c...c
                C.............C      size of array = C...C
                c.............c      contents = c...c
                ...
Branch:         01101 RRR C...C      if (RRR >= 0 && RRR < C..)
                c.............c      branch to (RRR+1)th address
Read1:          01110 RRR ...        read 1-byte to RRR
Read2:          01111 RRR ..rrr      read 2-byte to RRR and rrr
ReadBranch:     10000 RRR C...C      Read1 and Branch
                c.............c
                ...
Write1:         10001 RRR .....      write 1-byte RRR
Write2:         10010 RRR ..rrr      write 2-byte RRR and rrr
WriteC:         10011 000 .....      write 1-char C...CC
                C.............C
WriteS:         10100 000 .....      write C..-byte of string
                C.............C
                S.............S
                ...
WriteA:         10101 RRR .....      write array[RRR]
                C.............C      size of array = C...C
                c.............c      contents = c...c
                ...
End:            10110 000 .....      terminate the execution

SetSelfCS:      10111 RRR C...C      RRR AAAAA= C...C
                ..........AAAAA
SetSelfCL:      11000 RRR .....      RRR AAAAA= c...c
                c.............c
                ..........AAAAA
SetSelfR:       11001 RRR ..Rrr      RRR AAAAA= rrr
                ..........AAAAA
SetExprCL:      11010 RRR ..Rrr      RRR = rrr AAAAA c...c
                c.............c
                ..........AAAAA
SetExprR:       11011 RRR ..rrr      RRR = rrr AAAAA Rrr
                ............Rrr
                ..........AAAAA
JumpCondC:      11100 RRR c...c      if !(RRR AAAAA C..) jump to c...c
                C.............C
                ..........AAAAA
JumpCondR:      11101 RRR c...c      if !(RRR AAAAA rrr) jump to c...c
                ............rrr
                ..........AAAAA
ReadJumpCondC:  11110 RRR c...c      Read1 and JumpCondC
                C.............C
                ..........AAAAA
ReadJumpCondR:  11111 RRR c...c      Read1 and JumpCondR
                ............rrr
                ..........AAAAA
@end example

@node The Lisp Reader and Compiler, Lstreams, MULE Character Sets and Encodings, Top
@chapter The Lisp Reader and Compiler

Not yet documented.

@node Lstreams, Consoles; Devices; Frames; Windows, The Lisp Reader and Compiler, Top
@chapter Lstreams

  An @dfn{lstream} is an internal Lisp object that provides a generic
buffering stream implementation.  Conceptually, you send data to the
stream or read data from the stream, not caring what's on the other end
of the stream.  The other end could be another stream, a file
descriptor, a stdio stream, a fixed block of memory, a reallocating
block of memory, etc.  The main purpose of the stream is to provide a
standard interface and to do buffering.  Macros are defined to read or
write characters, so the calling functions do not have to worry about
blocking data together in order to achieve efficiency.

@menu
* Creating an Lstream::         Creating an lstream object.
* Lstream Types::               Different sorts of things that are streamed.
* Lstream Functions::           Functions for working with lstreams.
* Lstream Methods::             Creating new lstream types.
@end menu

@node Creating an Lstream
@section Creating an Lstream

Lstreams come in different types, depending on what is being interfaced
to.  Although the primitive for creating new lstreams is
@code{Lstream_new()}, generally you do not call this directly.  Instead,
you call some type-specific creation function, which creates the lstream
and initializes it as appropriate for the particular type.

All lstream creation functions take a @var{mode} argument, specifying
what mode the lstream should be opened as.  This controls whether the
lstream is for input and output, and optionally whether data should be
blocked up in units of MULE characters.  Note that some types of
lstreams can only be opened for input; others only for output; and
others can be opened either way.  #### Richard Mlynarik thinks that
there should be a strict separation between input and output streams,
and he's probably right.

  @var{mode} is a string, one of

@table @code
@item "r"
  Open for reading.
@item "w"
  Open for writing.
@item "rc"
  Open for reading, but ``read'' never returns partial MULE characters.
@item "wc"
  Open for writing, but never writes partial MULE characters.
@end table

@node Lstream Types
@section Lstream Types

@table @asis
@item stdio

@item filedesc

@item lisp-string

@item fixed-buffer

@item resizing-buffer

@item dynarr

@item lisp-buffer

@item print

@item decoding

@item encoding
@end table

@node Lstream Functions
@section Lstream Functions

@deftypefun {Lstream *} Lstream_new (Lstream_implementation *@var{imp}, CONST char *@var{mode})
Allocate and return a new Lstream.  This function is not really meant to
be called directly; rather, each stream type should provide its own
stream creation function, which creates the stream and does any other
necessary creation stuff (e.g. opening a file).
@end deftypefun

@deftypefun void Lstream_set_buffering (Lstream *@var{lstr}, Lstream_buffering @var{buffering}, int @var{buffering_size})
Change the buffering of a stream.  See @file{lstream.h}.  By default the
buffering is @code{STREAM_BLOCK_BUFFERED}.
@end deftypefun

@deftypefun int Lstream_flush (Lstream *@var{lstr})
Flush out any pending unwritten data in the stream.  Clear any buffered
input data.  Returns 0 on success, -1 on error.
@end deftypefun

@deftypefn Macro int Lstream_putc (Lstream *@var{stream}, int @var{c})
Write out one byte to the stream.  This is a macro and so it is very
efficient.  The @var{c} argument is only evaluated once but the @var{stream}
argument is evaluated more than once.  Returns 0 on success, -1 on
error.
@end deftypefn

@deftypefn Macro int Lstream_getc (Lstream *@var{stream})
Read one byte from the stream.  This is a macro and so it is very
efficient.  The @var{stream} argument is evaluated more than once.  Return
value is -1 for EOF or error.
@end deftypefn

@deftypefn Macro void Lstream_ungetc (Lstream *@var{stream}, int @var{c})
Push one byte back onto the input queue.  This will be the next byte
read from the stream.  Any number of bytes can be pushed back and will
be read in the reverse order they were pushed back---most recent
first. (This is necessary for consistency---if there are a number of
bytes that have been unread and I read and unread a byte, it needs to be
the first to be read again.) This is a macro and so it is very
efficient.  The @var{c} argument is only evaluated once but the @var{stream}
argument is evaluated more than once.
@end deftypefn

@deftypefun int Lstream_fputc (Lstream *@var{stream}, int @var{c})
@deftypefunx int Lstream_fgetc (Lstream *@var{stream})
@deftypefunx void Lstream_fungetc (Lstream *@var{stream}, int @var{c})
Function equivalents of the above macros.
@end deftypefun

@deftypefun ssize_t Lstream_read (Lstream *@var{stream}, void *@var{data}, size_t @var{size})
Read @var{size} bytes of @var{data} from the stream.  Return the number
of bytes read.  0 means EOF. -1 means an error occurred and no bytes
were read.
@end deftypefun

@deftypefun ssize_t Lstream_write (Lstream *@var{stream}, void *@var{data}, size_t @var{size})
Write @var{size} bytes of @var{data} to the stream.  Return the number
of bytes written.  -1 means an error occurred and no bytes were written.
@end deftypefun

@deftypefun void Lstream_unread (Lstream *@var{stream}, void *@var{data}, size_t @var{size})
Push back @var{size} bytes of @var{data} onto the input queue.  The next
call to @code{Lstream_read()} with the same size will read the same
bytes back.  Note that this will be the case even if there is other
pending unread data.
@end deftypefun

@deftypefun int Lstream_close (Lstream *@var{stream})
Close the stream.  All data will be flushed out.
@end deftypefun

@deftypefun void Lstream_reopen (Lstream *@var{stream})
Reopen a closed stream.  This enables I/O on it again.  This is not
meant to be called except from a wrapper routine that reinitializes
variables and such---the close routine may well have freed some
necessary storage structures, for example.
@end deftypefun

@deftypefun void Lstream_rewind (Lstream *@var{stream})
Rewind the stream to the beginning.
@end deftypefun

@node Lstream Methods
@section Lstream Methods

@deftypefn {Lstream Method} ssize_t reader (Lstream *@var{stream}, unsigned char *@var{data}, size_t @var{size})
Read some data from the stream's end and store it into @var{data}, which
can hold @var{size} bytes.  Return the number of bytes read.  A return
value of 0 means no bytes can be read at this time.  This may be because
of an EOF, or because there is a granularity greater than one byte that
the stream imposes on the returned data, and @var{size} is less than
this granularity. (This will happen frequently for streams that need to
return whole characters, because @code{Lstream_read()} calls the reader
function repeatedly until it has the number of bytes it wants or until 0
is returned.)  The lstream functions do not treat a 0 return as EOF or
do anything special; however, the calling function will interpret any 0
it gets back as EOF.  This will normally not happen unless the caller
calls @code{Lstream_read()} with a very small size.

This function can be @code{NULL} if the stream is output-only.
@end deftypefn

@deftypefn {Lstream Method} ssize_t writer (Lstream *@var{stream}, CONST unsigned char *@var{data}, size_t @var{size})
Send some data to the stream's end.  Data to be sent is in @var{data}
and is @var{size} bytes.  Return the number of bytes sent.  This
function can send and return fewer bytes than is passed in; in that
case, the function will just be called again until there is no data left
or 0 is returned.  A return value of 0 means that no more data can be
currently stored, but there is no error; the data will be squirreled
away until the writer can accept data. (This is useful, e.g., if you're
dealing with a non-blocking file descriptor and are getting
@code{EWOULDBLOCK} errors.)  This function can be @code{NULL} if the
stream is input-only.
@end deftypefn

@deftypefn {Lstream Method} int rewinder (Lstream *@var{stream})
Rewind the stream.  If this is @code{NULL}, the stream is not seekable.
@end deftypefn

@deftypefn {Lstream Method} int seekable_p (Lstream *@var{stream})
Indicate whether this stream is seekable---i.e. it can be rewound.
This method is ignored if the stream does not have a rewind method.  If
this method is not present, the result is determined by whether a rewind
method is present.
@end deftypefn

@deftypefn {Lstream Method} int flusher (Lstream *@var{stream})
Perform any additional operations necessary to flush the data in this
stream.
@end deftypefn

@deftypefn {Lstream Method} int pseudo_closer (Lstream *@var{stream})
@end deftypefn

@deftypefn {Lstream Method} int closer (Lstream *@var{stream})
Perform any additional operations necessary to close this stream down.
May be @code{NULL}.  This function is called when @code{Lstream_close()}
is called or when the stream is garbage-collected.  When this function
is called, all pending data in the stream will already have been written
out.
@end deftypefn

@deftypefn {Lstream Method} Lisp_Object marker (Lisp_Object @var{lstream}, void (*@var{markfun}) (Lisp_Object))
Mark this object for garbage collection.  Same semantics as a standard
@code{Lisp_Object} marker.  This function can be @code{NULL}.
@end deftypefn

@node Consoles; Devices; Frames; Windows, The Redisplay Mechanism, Lstreams, Top
@chapter Consoles; Devices; Frames; Windows

@menu
* Introduction to Consoles; Devices; Frames; Windows::
* Point::
* Window Hierarchy::
* The Window Object::
@end menu

@node Introduction to Consoles; Devices; Frames; Windows
@section Introduction to Consoles; Devices; Frames; Windows

A window-system window that you see on the screen is called a
@dfn{frame} in Emacs terminology.  Each frame is subdivided into one or
more non-overlapping panes, called (confusingly) @dfn{windows}.  Each
window displays the text of a buffer in it. (See above on Buffers.) Note
that buffers and windows are independent entities: Two or more windows
can be displaying the same buffer (potentially in different locations),
and a buffer can be displayed in no windows.

  A single display screen that contains one or more frames is called
a @dfn{display}.  Under most circumstances, there is only one display.
However, more than one display can exist, for example if you have
a @dfn{multi-headed} console, i.e. one with a single keyboard but
multiple displays. (Typically in such a situation, the various
displays act like one large display, in that the mouse is only
in one of them at a time, and moving the mouse off of one moves
it into another.) In some cases, the different displays will
have different characteristics, e.g. one color and one mono.

  XEmacs can display frames on multiple displays.  It can even deal
simultaneously with frames on multiple keyboards (called @dfn{consoles} in
XEmacs terminology).  Here is one case where this might be useful: You
are using XEmacs on your workstation at work, and leave it running.
Then you go home and dial in on a TTY line, and you can use the
already-running XEmacs process to display another frame on your local
TTY.

  Thus, there is a hierarchy console -> display -> frame -> window.
There is a separate Lisp object type for each of these four concepts.
Furthermore, there is logically a @dfn{selected console},
@dfn{selected display}, @dfn{selected frame}, and @dfn{selected window}.
Each of these objects is distinguished in various ways, such as being the
default object for various functions that act on objects of that type.
Note that every containing object rememembers the ``selected'' object
among the objects that it contains: e.g. not only is there a selected
window, but every frame remembers the last window in it that was
selected, and changing the selected frame causes the remembered window
within it to become the selected window.  Similar relationships apply
for consoles to devices and devices to frames.

@node Point
@section Point

  Recall that every buffer has a current insertion position, called
@dfn{point}.  Now, two or more windows may be displaying the same buffer,
and the text cursor in the two windows (i.e. @code{point}) can be in
two different places.  You may ask, how can that be, since each
buffer has only one value of @code{point}?  The answer is that each window
also has a value of @code{point} that is squirreled away in it.  There
is only one selected window, and the value of ``point'' in that buffer
corresponds to that window.  When the selected window is changed
from one window to another displaying the same buffer, the old
value of @code{point} is stored into the old window's ``point'' and the
value of @code{point} from the new window is retrieved and made the
value of @code{point} in the buffer.  This means that @code{window-point}
for the selected window is potentially inaccurate, and if you
want to retrieve the correct value of @code{point} for a window,
you must special-case on the selected window and retrieve the
buffer's point instead.  This is related to why @code{save-window-excursion}
does not save the selected window's value of @code{point}.

@node Window Hierarchy
@section Window Hierarchy
@cindex window hierarchy
@cindex hierarchy of windows

  If a frame contains multiple windows (panes), they are always created
by splitting an existing window along the horizontal or vertical axis.
Terminology is a bit confusing here: to @dfn{split a window
horizontally} means to create two side-by-side windows, i.e. to make a
@emph{vertical} cut in a window.  Likewise, to @dfn{split a window
vertically} means to create two windows, one above the other, by making
a @emph{horizontal} cut.

  If you split a window and then split again along the same axis, you
will end up with a number of panes all arranged along the same axis.
The precise way in which the splits were made should not be important,
and this is reflected internally.  Internally, all windows are arranged
in a tree, consisting of two types of windows, @dfn{combination} windows
(which have children, and are covered completely by those children) and
@dfn{leaf} windows, which have no children and are visible.  Every
combination window has two or more children, all arranged along the same
axis.  There are (logically) two subtypes of windows, depending on
whether their children are horizontally or vertically arrayed.  There is
always one root window, which is either a leaf window (if the frame
contains only one window) or a combination window (if the frame contains
more than one window).  In the latter case, the root window will have
two or more children, either horizontally or vertically arrayed, and
each of those children will be either a leaf window or another
combination window.

  Here are some rules:

@enumerate
@item
Horizontal combination windows can never have children that are
horizontal combination windows; same for vertical.

@item
Only leaf windows can be split (obviously) and this splitting does one
of two things: (a) turns the leaf window into a combination window and
creates two new leaf children, or (b) turns the leaf window into one of
the two new leaves and creates the other leaf.  Rule (1) dictates which
of these two outcomes happens.

@item
Every combination window must have at least two children.

@item
Leaf windows can never become combination windows.  They can be deleted,
however.  If this results in a violation of (3), the parent combination
window also gets deleted.

@item
All functions that accept windows must be prepared to accept combination
windows, and do something sane (e.g. signal an error if so).
Combination windows @emph{do} escape to the Lisp level.

@item
All windows have three fields governing their contents:
these are @dfn{hchild} (a list of horizontally-arrayed children),
@dfn{vchild} (a list of vertically-arrayed children), and @dfn{buffer}
(the buffer contained in a leaf window).  Exactly one of
these will be non-nil.  Remember that @dfn{horizontally-arrayed}
means ``side-by-side'' and @dfn{vertically-arrayed} means
@dfn{one above the other}.

@item
Leaf windows also have markers in their @code{start} (the
first buffer position displayed in the window) and @code{pointm}
(the window's stashed value of @code{point}---see above) fields,
while combination windows have nil in these fields.

@item
The list of children for a window is threaded through the
@code{next} and @code{prev} fields of each child window.

@item
@strong{Deleted windows can be undeleted}.  This happens as a result of
restoring a window configuration, and is unlike frames, displays, and
consoles, which, once deleted, can never be restored.  Deleting a window
does nothing except set a special @code{dead} bit to 1 and clear out the
@code{next}, @code{prev}, @code{hchild}, and @code{vchild} fields, for
GC purposes.

@item
Most frames actually have two top-level windows---one for the
minibuffer and one (the @dfn{root}) for everything else.  The modeline
(if present) separates these two.  The @code{next} field of the root
points to the minibuffer, and the @code{prev} field of the minibuffer
points to the root.  The other @code{next} and @code{prev} fields are
@code{nil}, and the frame points to both of these windows.
Minibuffer-less frames have no minibuffer window, and the @code{next}
and @code{prev} of the root window are @code{nil}.  Minibuffer-only
frames have no root window, and the @code{next} of the minibuffer window
is @code{nil} but the @code{prev} points to itself. (#### This is an
artifact that should be fixed.)
@end enumerate

@node The Window Object
@section The Window Object

  Windows have the following accessible fields:

@table @code
@item frame
The frame that this window is on.

@item mini_p
Non-@code{nil} if this window is a minibuffer window.

@item buffer
The buffer that the window is displaying.  This may change often during
the life of the window.

@item dedicated
Non-@code{nil} if this window is dedicated to its buffer.

@item pointm
@cindex window point internals
This is the value of point in the current buffer when this window is
selected; when it is not selected, it retains its previous value.

@item start
The position in the buffer that is the first character to be displayed
in the window.

@item force_start
If this flag is non-@code{nil}, it says that the window has been
scrolled explicitly by the Lisp program.  This affects what the next
redisplay does if point is off the screen: instead of scrolling the
window to show the text around point, it moves point to a location that
is on the screen.

@item last_modified
The @code{modified} field of the window's buffer, as of the last time
a redisplay completed in this window.

@item last_point
The buffer's value of point, as of the last time
a redisplay completed in this window.

@item left
This is the left-hand edge of the window, measured in columns.  (The
leftmost column on the screen is @w{column 0}.)

@item top
This is the top edge of the window, measured in lines.  (The top line on
the screen is @w{line 0}.)

@item height
The height of the window, measured in lines.

@item width
The width of the window, measured in columns.

@item next
This is the window that is the next in the chain of siblings.  It is
@code{nil} in a window that is the rightmost or bottommost of a group of
siblings.

@item prev
This is the window that is the previous in the chain of siblings.  It is
@code{nil} in a window that is the leftmost or topmost of a group of
siblings.

@item parent
Internally, XEmacs arranges windows in a tree; each group of siblings has
a parent window whose area includes all the siblings.  This field points
to a window's parent.

Parent windows do not display buffers, and play little role in display
except to shape their child windows.  Emacs Lisp programs usually have
no access to the parent windows; they operate on the windows at the
leaves of the tree, which actually display buffers.

@item hscroll
This is the number of columns that the display in the window is scrolled
horizontally to the left.  Normally, this is 0.

@item use_time
This is the last time that the window was selected.  The function
@code{get-lru-window} uses this field.

@item display_table
The window's display table, or @code{nil} if none is specified for it.

@item update_mode_line
Non-@code{nil} means this window's mode line needs to be updated.

@item base_line_number
The line number of a certain position in the buffer, or @code{nil}.
This is used for displaying the line number of point in the mode line.

@item base_line_pos
The position in the buffer for which the line number is known, or
@code{nil} meaning none is known.

@item region_showing
If the region (or part of it) is highlighted in this window, this field
holds the mark position that made one end of that region.  Otherwise,
this field is @code{nil}.
@end table

@node The Redisplay Mechanism, Extents, Consoles; Devices; Frames; Windows, Top
@chapter The Redisplay Mechanism

  The redisplay mechanism is one of the most complicated sections of
XEmacs, especially from a conceptual standpoint.  This is doubly so
because, unlike for the basic aspects of the Lisp interpreter, the
computer science theories of how to efficiently handle redisplay are not
well-developed.

  When working with the redisplay mechanism, remember the Golden Rules
of Redisplay:

@enumerate
@item
It Is Better To Be Correct Than Fast.
@item
Thou Shalt Not Run Elisp From Within Redisplay.
@item
It Is Better To Be Fast Than Not To Be.
@end enumerate

@menu
* Critical Redisplay Sections::
* Line Start Cache::
* Redisplay Piece by Piece::
@end menu

@node Critical Redisplay Sections
@section Critical Redisplay Sections
@cindex critical redisplay sections

Within this section, we are defenseless and assume that the
following cannot happen:

@enumerate
@item
garbage collection
@item
Lisp code evaluation
@item
frame size changes
@end enumerate

We ensure (3) by calling @code{hold_frame_size_changes()}, which
will cause any pending frame size changes to get put on hold
till after the end of the critical section.  (1) follows
automatically if (2) is met.  #### Unfortunately, there are
some places where Lisp code can be called within this section.
We need to remove them.

If @code{Fsignal()} is called during this critical section, we
will @code{abort()}.

If garbage collection is called during this critical section,
we simply return. #### We should abort instead.

#### If a frame-size change does occur we should probably
actually be preempting redisplay.

@node Line Start Cache
@section Line Start Cache
@cindex line start cache

  The traditional scrolling code in Emacs breaks in a variable height
world.  It depends on the key assumption that the number of lines that
can be displayed at any given time is fixed.  This led to a complete
separation of the scrolling code from the redisplay code.  In order to
fully support variable height lines, the scrolling code must actually be
tightly integrated with redisplay.  Only redisplay can determine how
many lines will be displayed on a screen for any given starting point.

  What is ideally wanted is a complete list of the starting buffer
position for every possible display line of a buffer along with the
height of that display line.  Maintaining such a full list would be very
expensive.  We settle for having it include information for all areas
which we happen to generate anyhow (i.e. the region currently being
displayed) and for those areas we need to work with.

  In order to ensure that the cache accurately represents what redisplay
would actually show, it is necessary to invalidate it in many
situations.  If the buffer changes, the starting positions may no longer
be correct.  If a face or an extent has changed then the line heights
may have altered.  These events happen frequently enough that the cache
can end up being constantly disabled.  With this potentially constant
invalidation when is the cache ever useful?

  Even if the cache is invalidated before every single usage, it is
necessary.  Scrolling often requires knowledge about display lines which
are actually above or below the visible region.  The cache provides a
convenient light-weight method of storing this information for multiple
display regions.  This knowledge is necessary for the scrolling code to
always obey the First Golden Rule of Redisplay.

  If the cache already contains all of the information that the scrolling
routines happen to need so that it doesn't have to go generate it, then
we are able to obey the Third Golden Rule of Redisplay.  The first thing
we do to help out the cache is to always add the displayed region.  This
region had to be generated anyway, so the cache ends up getting the
information basically for free.  In those cases where a user is simply
scrolling around viewing a buffer there is a high probability that this
is sufficient to always provide the needed information.  The second
thing we can do is be smart about invalidating the cache.

  TODO---Be smart about invalidating the cache.  Potential places:

@itemize @bullet
@item
Insertions at end-of-line which don't cause line-wraps do not alter the
starting positions of any display lines.  These types of buffer
modifications should not invalidate the cache.  This is actually a large
optimization for redisplay speed as well.
@item
Buffer modifications frequently only affect the display of lines at and
below where they occur.  In these situations we should only invalidate
the part of the cache starting at where the modification occurs.
@end itemize

  In case you're wondering, the Second Golden Rule of Redisplay is not
applicable.

@node Redisplay Piece by Piece
@section Redisplay Piece by Piece
@cindex Redisplay Piece by Piece

As you can begin to see redisplay is complex and also not well
documented. Chuck no longer works on XEmacs so this section is my take
on the workings of redisplay.

Redisplay happens in three phases:

@enumerate
@item
Determine desired display in area that needs redisplay.
Implemented by @code{redisplay.c}
@item
Compare desired display with current display
Implemented by @code{redisplay-output.c}
@item
Output changes Implemented by @code{redisplay-output.c},
@code{redisplay-x.c}, @code{redisplay-msw.c} and @code{redisplay-tty.c}
@end enumerate

Steps 1 and 2 are device-independant and relatively complex.  Step 3 is
mostly device-dependent.

Determining the desired display

Display attributes are stored in @code{display_line} structures. Each
@code{display_line} consists of a set of @code{display_block}'s and each
@code{display_block} contains a number of @code{rune}'s. Generally
dynarr's of @code{display_line}'s are held by each window representing
the current display and the desired display.

The @code{display_line} structures are tighly tied to buffers which
presents a problem for redisplay as this connection is bogus for the
modeline. Hence the @code{display_line} generation routines are
duplicated for generating the modeline. This means that the modeline
display code has many bugs that the standard redisplay code does not.

The guts of @code{display_line} generation are in
@code{create_text_block}, which creates a single display line for the
desired locale. This incrementally parses the characters on the current
line and generates redisplay structures for each. 

Gutter redisplay is different. Because the data to display is stored in
a string we cannot use @code{create_text_block}. Instead we use
@code{create_text_string_block} which performs the same function as
@code{create_text_block} but for strings. Many of the complexities of
@code{create_text_block} to do with cursor handling and selective
display have been removed.

@node Extents, Faces, The Redisplay Mechanism, Top
@chapter Extents

@menu
* Introduction to Extents::     Extents are ranges over text, with properties.
* Extent Ordering::             How extents are ordered internally.
* Format of the Extent Info::   The extent information in a buffer or string.
* Zero-Length Extents::         A weird special case.
* Mathematics of Extent Ordering::      A rigorous foundation.
* Extent Fragments::            Cached information useful for redisplay.
@end menu

@node Introduction to Extents
@section Introduction to Extents

  Extents are regions over a buffer, with a start and an end position
denoting the region of the buffer included in the extent.  In
addition, either end can be closed or open, meaning that the endpoint
is or is not logically included in the extent.  Insertion of a character
at a closed endpoint causes the character to go inside the extent;
insertion at an open endpoint causes the character to go outside.

  Extent endpoints are stored using memory indices (see @file{insdel.c}),
to minimize the amount of adjusting that needs to be done when
characters are inserted or deleted.

  (Formerly, extent endpoints at the gap could be either before or
after the gap, depending on the open/closedness of the endpoint.
The intent of this was to make it so that insertions would
automatically go inside or out of extents as necessary with no
further work needing to be done.  It didn't work out that way,
however, and just ended up complexifying and buggifying all the
rest of the code.)

@node Extent Ordering
@section Extent Ordering

  Extents are compared using memory indices.  There are two orderings
for extents and both orders are kept current at all times.  The normal
or @dfn{display} order is as follows:

@example
Extent A is ``less than'' extent B,
that is, earlier in the display order,
  if:    A-start < B-start,
  or if: A-start = B-start, and A-end > B-end
@end example

  So if two extents begin at the same position, the larger of them is the
earlier one in the display order (@code{EXTENT_LESS} is true).

  For the e-order, the same thing holds:

@example
Extent A is ``less than'' extent B in e-order,
that is, later in the buffer,
  if:    A-end < B-end,
  or if: A-end = B-end, and A-start > B-start
@end example

  So if two extents end at the same position, the smaller of them is the
earlier one in the e-order (@code{EXTENT_E_LESS} is true).

  The display order and the e-order are complementary orders: any
theorem about the display order also applies to the e-order if you swap
all occurrences of ``display order'' and ``e-order'', ``less than'' and
``greater than'', and ``extent start'' and ``extent end''.

@node Format of the Extent Info
@section Format of the Extent Info

  An extent-info structure consists of a list of the buffer or string's
extents and a @dfn{stack of extents} that lists all of the extents over
a particular position.  The stack-of-extents info is used for
optimization purposes---it basically caches some info that might
be expensive to compute.  Certain otherwise hard computations are easy
given the stack of extents over a particular position, and if the
stack of extents over a nearby position is known (because it was
calculated at some prior point in time), it's easy to move the stack
of extents to the proper position.

  Given that the stack of extents is an optimization, and given that
it requires memory, a string's stack of extents is wiped out each
time a garbage collection occurs.  Therefore, any time you retrieve
the stack of extents, it might not be there.  If you need it to
be there, use the @code{_force} version.

  Similarly, a string may or may not have an extent_info structure.
(Generally it won't if there haven't been any extents added to the
string.) So use the @code{_force} version if you need the extent_info
structure to be there.

  A list of extents is maintained as a double gap array: one gap array
is ordered by start index (the @dfn{display order}) and the other is
ordered by end index (the @dfn{e-order}).  Note that positions in an
extent list should logically be conceived of as referring @emph{to} a
particular extent (as is the norm in programs) rather than sitting
between two extents.  Note also that callers of these functions should
not be aware of the fact that the extent list is implemented as an
array, except for the fact that positions are integers (this should be
generalized to handle integers and linked list equally well).

@node Zero-Length Extents
@section Zero-Length Extents

  Extents can be zero-length, and will end up that way if their endpoints
are explicitly set that way or if their detachable property is nil
and all the text in the extent is deleted. (The exception is open-open
zero-length extents, which are barred from existing because there is
no sensible way to define their properties.  Deletion of the text in
an open-open extent causes it to be converted into a closed-open
extent.)  Zero-length extents are primarily used to represent
annotations, and behave as follows:

@enumerate
@item
Insertion at the position of a zero-length extent expands the extent
if both endpoints are closed; goes after the extent if it is closed-open;
and goes before the extent if it is open-closed.

@item
Deletion of a character on a side of a zero-length extent whose
corresponding endpoint is closed causes the extent to be detached if
it is detachable; if the extent is not detachable or the corresponding
endpoint is open, the extent remains in the buffer, moving as necessary.
@end enumerate

  Note that closed-open, non-detachable zero-length extents behave
exactly like markers and that open-closed, non-detachable zero-length
extents behave like the ``point-type'' marker in Mule.

@node Mathematics of Extent Ordering
@section Mathematics of Extent Ordering
@cindex extent mathematics
@cindex mathematics of extents
@cindex extent ordering

@cindex display order of extents
@cindex extents, display order
  The extents in a buffer are ordered by ``display order'' because that
is that order that the redisplay mechanism needs to process them in.
The e-order is an auxiliary ordering used to facilitate operations
over extents.  The operations that can be performed on the ordered
list of extents in a buffer are

@enumerate
@item
Locate where an extent would go if inserted into the list.
@item
Insert an extent into the list.
@item
Remove an extent from the list.
@item
Map over all the extents that overlap a range.
@end enumerate

  (4) requires being able to determine the first and last extents
that overlap a range.

  NOTE: @dfn{overlap} is used as follows:

@itemize @bullet
@item
two ranges overlap if they have at least one point in common.
Whether the endpoints are open or closed makes a difference here.
@item
a point overlaps a range if the point is contained within the
range; this is equivalent to treating a point @math{P} as the range
@math{[P, P]}.
@item
In the case of an @emph{extent} overlapping a point or range, the extent
is normally treated as having closed endpoints.  This applies
consistently in the discussion of stacks of extents and such below.
Note that this definition of overlap is not necessarily consistent with
the extents that @code{map-extents} maps over, since @code{map-extents}
sometimes pays attention to whether the endpoints of an extents are open
or closed.  But for our purposes, it greatly simplifies things to treat
all extents as having closed endpoints.
@end itemize

First, define @math{>}, @math{<}, @math{<=}, etc. as applied to extents
to mean comparison according to the display order.  Comparison between
an extent @math{E} and an index @math{I} means comparison between
@math{E} and the range @math{[I, I]}.

Also define @math{e>}, @math{e<}, @math{e<=}, etc. to mean comparison
according to the e-order.

For any range @math{R}, define @math{R(0)} to be the starting index of
the range and @math{R(1)} to be the ending index of the range.

For any extent @math{E}, define @math{E(next)} to be the extent directly
following @math{E}, and @math{E(prev)} to be the extent directly
preceding @math{E}.  Assume @math{E(next)} and @math{E(prev)} can be
determined from @math{E} in constant time.  (This is because we store
the extent list as a doubly linked list.)

Similarly, define @math{E(e-next)} and @math{E(e-prev)} to be the
extents directly following and preceding @math{E} in the e-order.

Now:

Let @math{R} be a range.
Let @math{F} be the first extent overlapping @math{R}.
Let @math{L} be the last extent overlapping @math{R}.

Theorem 1: @math{R(1)} lies between @math{L} and @math{L(next)},
i.e. @math{L <= R(1) < L(next)}.

  This follows easily from the definition of display order.  The
basic reason that this theorem applies is that the display order
sorts by increasing starting index.

  Therefore, we can determine @math{L} just by looking at where we would
insert @math{R(1)} into the list, and if we know @math{F} and are moving
forward over extents, we can easily determine when we've hit @math{L} by
comparing the extent we're at to @math{R(1)}.

@example
Theorem 2: @math{F(e-prev) e< [1, R(0)] e<= F}.
@end example

  This is the analog of Theorem 1, and applies because the e-order
sorts by increasing ending index.

  Therefore, @math{F} can be found in the same amount of time as
operation (1), i.e. the time that it takes to locate where an extent
would go if inserted into the e-order list.

  If the lists were stored as balanced binary trees, then operation (1)
would take logarithmic time, which is usually quite fast.  However,
currently they're stored as simple doubly-linked lists, and instead we
do some caching to try to speed things up.

  Define a @dfn{stack of extents} (or @dfn{SOE}) as the set of extents
(ordered in the display order) that overlap an index @math{I}, together
with the SOE's @dfn{previous} extent, which is an extent that precedes
@math{I} in the e-order. (Hopefully there will not be very many extents
between @math{I} and the previous extent.)

Now:

Let @math{I} be an index, let @math{S} be the stack of extents on
@math{I}, let @math{F} be the first extent in @math{S}, and let @math{P}
be @math{S}'s previous extent.

Theorem 3: The first extent in @math{S} is the first extent that overlaps
any range @math{[I, J]}.

Proof: Any extent that overlaps @math{[I, J]} but does not include
@math{I} must have a start index @math{> I}, and thus be greater than
any extent in @math{S}.

Therefore, finding the first extent that overlaps a range @math{R} is
the same as finding the first extent that overlaps @math{R(0)}.

Theorem 4: Let @math{I2} be an index such that @math{I2 > I}, and let
@math{F2} be the first extent that overlaps @math{I2}.  Then, either
@math{F2} is in @math{S} or @math{F2} is greater than any extent in
@math{S}.

Proof: If @math{F2} does not include @math{I} then its start index is
greater than @math{I} and thus it is greater than any extent in
@math{S}, including @math{F}.  Otherwise, @math{F2} includes @math{I}
and thus is in @math{S}, and thus @math{F2 >= F}.

@node Extent Fragments
@section Extent Fragments
@cindex extent fragment

  Imagine that the buffer is divided up into contiguous, non-overlapping
@dfn{runs} of text such that no extent starts or ends within a run
(extents that abut the run don't count).

  An extent fragment is a structure that holds data about the run that
contains a particular buffer position (if the buffer position is at the
junction of two runs, the run after the position is used)---the
beginning and end of the run, a list of all of the extents in that run,
the @dfn{merged face} that results from merging all of the faces
corresponding to those extents, the begin and end glyphs at the
beginning of the run, etc.  This is the information that redisplay needs
in order to display this run.

  Extent fragments have to be very quick to update to a new buffer
position when moving linearly through the buffer.  They rely on the
stack-of-extents code, which does the heavy-duty algorithmic work of
determining which extents overly a particular position.

@node Faces, Glyphs, Extents, Top
@chapter Faces

Not yet documented.

@node Glyphs, Specifiers, Faces, Top
@chapter Glyphs

Glyphs are graphical elements that can be displayed in XEmacs buffers or
gutters. We use the term graphical element here in the broadest possible
sense since glyphs can be as mundane as text to as arcane as a native
tab widget.

In XEmacs, glyphs represent the uninstantiated state of graphical
elements, i.e. they hold all the information necessary to produce an
image on-screen but the image does not exist at this stage.

Glyphs are lazily instantiated by calling one of the glyph
functions. This usually occurs within redisplay when
@code{Fglyph_height} is called. Instantiation causes an image-instance
to be created and cached. This cache is on a device basis for all glyphs
except glyph-widgets, and on a window basis for glyph widgets.  The
caching is done by @code{image_instantiate} and is necessary because it
is generally possible to display an image-instance in multiple
domains. For instance if we create a Pixmap, we can actually display
this on multiple windows - even though we only need a single Pixmap
instance to do this. If caching wasn't done then it would be necessary
to create image-instances for every displayable occurrance of a glyph -
and every usage - and this would be extremely memory and cpu intensive.

Widget-glyphs (a.k.a native widgets) are not cached in this way. This is
because widget-glyph image-instances on screen are toolkit windows, and
thus cannot be reused in multiple XEmacs domains. Thus widget-glyphs are
cached on a window basis.

Any action on a glyph first consults the cache before actually
instantiating a widget.

@section Widget-Glyphs in the MS-Windows Environment

To Do

@section Widget-Glyphs in the X Environment

Widget-glyphs under X make heavy use of lwlib for manipulating the
native toolkit objects. This is primarily so that different toolkits can
be supported for widget-glyphs, just as they are supported for features
such as menubars etc.

Lwlib is extremely poorly documented and quite hairy so here is my
understanding of what goes on.

Lwlib maintains a set of widget_instances which mirror the hierarchical
state of Xt widgets. I think this is so that widgets can be updated and
manipulated generically by the lwlib library. For instance
update_one_widget_instance can cope with multiple types of widget and
multiple types of toolkit. Each element in the widget hierarchy is updated
from its corresponding widget_instance by walking the widget_instance
tree recursively.

This has desirable properties such as lw_modify_all_widgets which is
called from glyphs-x.c and updates all the properties of a widget
without having to know what the widget is or what toolkit it is from.
Unfortunately this also has hairy properrties such as making the lwlib
code quite complex. And of course lwlib has to know at some level what
the widget is and how to set its properties.

@node Specifiers, Menus, Glyphs, Top
@chapter Specifiers

Not yet documented.

@node Menus, Subprocesses, Specifiers, Top
@chapter Menus

  A menu is set by setting the value of the variable
@code{current-menubar} (which may be buffer-local) and then calling
@code{set-menubar-dirty-flag} to signal a change.  This will cause the
menu to be redrawn at the next redisplay.  The format of the data in
@code{current-menubar} is described in @file{menubar.c}.

  Internally the data in current-menubar is parsed into a tree of
@code{widget_value's} (defined in @file{lwlib.h}); this is accomplished
by the recursive function @code{menu_item_descriptor_to_widget_value()},
called by @code{compute_menubar_data()}.  Such a tree is deallocated
using @code{free_widget_value()}.

  @code{update_screen_menubars()} is one of the external entry points.
This checks to see, for each screen, if that screen's menubar needs to
be updated.  This is the case if

@enumerate
@item
@code{set-menubar-dirty-flag} was called since the last redisplay.  (This
function sets the C variable menubar_has_changed.)
@item
The buffer displayed in the screen has changed.
@item
The screen has no menubar currently displayed.
@end enumerate

  @code{set_screen_menubar()} is called for each such screen.  This
function calls @code{compute_menubar_data()} to create the tree of
widget_value's, then calls @code{lw_create_widget()},
@code{lw_modify_all_widgets()}, and/or @code{lw_destroy_all_widgets()}
to create the X-Toolkit widget associated with the menu.

  @code{update_psheets()}, the other external entry point, actually
changes the menus being displayed.  It uses the widgets fixed by
@code{update_screen_menubars()} and calls various X functions to ensure
that the menus are displayed properly.

  The menubar widget is set up so that @code{pre_activate_callback()} is
called when the menu is first selected (i.e. mouse button goes down),
and @code{menubar_selection_callback()} is called when an item is
selected.  @code{pre_activate_callback()} calls the function in
activate-menubar-hook, which can change the menubar (this is described
in @file{menubar.c}).  If the menubar is changed,
@code{set_screen_menubars()} is called.
@code{menubar_selection_callback()} enqueues a menu event, putting in it
a function to call (either @code{eval} or @code{call-interactively}) and
its argument, which is the callback function or form given in the menu's
description.

@node Subprocesses, Interface to X Windows, Menus, Top
@chapter Subprocesses

  The fields of a process are:

@table @code
@item name
A string, the name of the process.

@item command
A list containing the command arguments that were used to start this
process.

@item filter
A function used to accept output from the process instead of a buffer,
or @code{nil}.

@item sentinel
A function called whenever the process receives a signal, or @code{nil}.

@item buffer
The associated buffer of the process.

@item pid
An integer, the Unix process @sc{id}.

@item childp
A flag, non-@code{nil} if this is really a child process.
It is @code{nil} for a network connection.

@item mark
A marker indicating the position of the end of the last output from this
process inserted into the buffer.  This is often but not always the end
of the buffer.

@item kill_without_query
If this is non-@code{nil}, killing XEmacs while this process is still
running does not ask for confirmation about killing the process.

@item raw_status_low
@itemx raw_status_high
These two fields record 16 bits each of the process status returned by
the @code{wait} system call.

@item status
The process status, as @code{process-status} should return it.

@item tick
@itemx update_tick
If these two fields are not equal, a change in the status of the process
needs to be reported, either by running the sentinel or by inserting a
message in the process buffer.

@item pty_flag
Non-@code{nil} if communication with the subprocess uses a @sc{pty};
@code{nil} if it uses a pipe.

@item infd
The file descriptor for input from the process.

@item outfd
The file descriptor for output to the process.

@item subtty
The file descriptor for the terminal that the subprocess is using.  (On
some systems, there is no need to record this, so the value is
@code{-1}.)

@item tty_name
The name of the terminal that the subprocess is using,
or @code{nil} if it is using pipes.
@end table

@node Interface to X Windows, Index, Subprocesses, Top
@chapter Interface to X Windows

Not yet documented.

@include index.texi

@c Print the tables of contents
@summarycontents
@contents
@c That's all

@bye