1 This is ../info/internals.info, produced by makeinfo version 4.0 from
2 internals/internals.texi.
4 INFO-DIR-SECTION XEmacs Editor
6 * Internals: (internals). XEmacs Internals Manual.
9 Copyright (C) 1992 - 1996 Ben Wing. Copyright (C) 1996, 1997 Sun
10 Microsystems. Copyright (C) 1994 - 1998 Free Software Foundation.
11 Copyright (C) 1994, 1995 Board of Trustees, University of Illinois.
13 Permission is granted to make and distribute verbatim copies of this
14 manual provided the copyright notice and this permission notice are
15 preserved on all copies.
17 Permission is granted to copy and distribute modified versions of
18 this manual under the conditions for verbatim copying, provided that the
19 entire resulting derived work is distributed under the terms of a
20 permission notice identical to this one.
22 Permission is granted to copy and distribute translations of this
23 manual into another language, under the above conditions for modified
24 versions, except that this permission notice may be stated in a
25 translation approved by the Foundation.
27 Permission is granted to copy and distribute modified versions of
28 this manual under the conditions for verbatim copying, provided also
29 that the section entitled "GNU General Public License" is included
30 exactly as in the original, and provided that the entire resulting
31 derived work is distributed under the terms of a permission notice
32 identical to this one.
34 Permission is granted to copy and distribute translations of this
35 manual into another language, under the above conditions for modified
36 versions, except that the section entitled "GNU General Public License"
37 may be included in a translation approved by the Free Software
38 Foundation instead of in the original English.
41 File: internals.info, Node: Top, Next: A History of Emacs, Prev: (dir), Up: (dir)
43 This Info file contains v1.0 of the XEmacs Internals Manual.
47 * A History of Emacs:: Times, dates, important events.
48 * XEmacs From the Outside:: A broad conceptual overview.
49 * The Lisp Language:: An overview.
50 * XEmacs From the Perspective of Building::
51 * XEmacs From the Inside::
52 * The XEmacs Object System (Abstractly Speaking)::
53 * How Lisp Objects Are Represented in C::
54 * Rules When Writing New C Code::
55 * A Summary of the Various XEmacs Modules::
56 * Allocation of Objects in XEmacs Lisp::
57 * Events and the Event Loop::
58 * Evaluation; Stack Frames; Bindings::
59 * Symbols and Variables::
60 * Buffers and Textual Representation::
61 * MULE Character Sets and Encodings::
62 * The Lisp Reader and Compiler::
64 * Consoles; Devices; Frames; Windows::
65 * The Redisplay Mechanism::
72 * Interface to X Windows::
73 * Index:: Index including concepts, functions, variables,
76 --- The Detailed Node Listing ---
78 Here are other nodes that are inferiors of those already listed,
79 mentioned here so you can get to them in one step:
83 * Through Version 18:: Unification prevails.
84 * Lucid Emacs:: One version 19 Emacs.
85 * GNU Emacs 19:: The other version 19 Emacs.
86 * XEmacs:: The continuation of Lucid Emacs.
88 Rules When Writing New C Code
90 * General Coding Rules::
91 * Writing Lisp Primitives::
92 * Adding Global Lisp Variables::
93 * Techniques for XEmacs Developers::
95 A Summary of the Various XEmacs Modules
98 * Basic Lisp Modules::
99 * Modules for Standard Editing Operations::
100 * Editor-Level Control Flow Modules::
101 * Modules for the Basic Displayable Lisp Objects::
102 * Modules for other Display-Related Lisp Objects::
103 * Modules for the Redisplay Mechanism::
104 * Modules for Interfacing with the File System::
105 * Modules for Other Aspects of the Lisp Interpreter and Object System::
106 * Modules for Interfacing with the Operating System::
107 * Modules for Interfacing with X Windows::
108 * Modules for Internationalization::
110 Allocation of Objects in XEmacs Lisp
112 * Introduction to Allocation::
113 * Garbage Collection::
115 * Garbage Collection - Step by Step::
116 * Integers and Characters::
117 * Allocation from Frob Blocks::
119 * Low-level allocation::
127 * Compiled Function::
129 Events and the Event Loop
131 * Introduction to Events::
133 * Specifics of the Event Gathering Mechanism::
134 * Specifics About the Emacs Event::
135 * The Event Stream Callback Routines::
136 * Other Event Loop Functions::
137 * Converting Events::
138 * Dispatching Events; The Command Builder::
140 Evaluation; Stack Frames; Bindings
143 * Dynamic Binding; The specbinding Stack; Unwind-Protects::
144 * Simple Special Forms::
147 Symbols and Variables
149 * Introduction to Symbols::
153 Buffers and Textual Representation
155 * Introduction to Buffers:: A buffer holds a block of text such as a file.
156 * The Text in a Buffer:: Representation of the text in a buffer.
157 * Buffer Lists:: Keeping track of all buffers.
158 * Markers and Extents:: Tagging locations within a buffer.
159 * Bufbytes and Emchars:: Representation of individual characters.
160 * The Buffer Object:: The Lisp object corresponding to a buffer.
162 MULE Character Sets and Encodings
166 * Internal Mule Encodings::
170 * Japanese EUC (Extended Unix Code)::
173 Internal Mule Encodings
175 * Internal String Encoding::
176 * Internal Character Encoding::
178 The Lisp Reader and Compiler
182 Consoles; Devices; Frames; Windows
184 * Introduction to Consoles; Devices; Frames; Windows::
188 The Redisplay Mechanism
190 * Critical Redisplay Sections::
195 * Introduction to Extents:: Extents are ranges over text, with properties.
196 * Extent Ordering:: How extents are ordered internally.
197 * Format of the Extent Info:: The extent information in a buffer or string.
198 * Zero-Length Extents:: A weird special case.
199 * Mathematics of Extent Ordering:: A rigorous foundation.
200 * Extent Fragments:: Cached information useful for redisplay.
212 Interface to X Windows
215 File: internals.info, Node: A History of Emacs, Next: XEmacs From the Outside, Prev: Top, Up: Top
220 XEmacs is a powerful, customizable text editor and development
221 environment. It began as Lucid Emacs, which was in turn derived from
222 GNU Emacs, a program written by Richard Stallman of the Free Software
223 Foundation. GNU Emacs dates back to the 1970's, and was modelled after
224 a package called "Emacs", written in 1976, that was a set of macros on
225 top of TECO, an old, old text editor written at MIT on the DEC PDP 10
226 under one of the earliest time-sharing operating systems, ITS
227 (Incompatible Timesharing System). (ITS dates back well before Unix.)
228 ITS, TECO, and Emacs were products of a group of people at MIT who
229 called themselves "hackers", who shared an idealistic belief system
230 about the free exchange of information and were fanatical in their
231 devotion to and time spent with computers. (The hacker subculture dates
232 back to the late 1950's at MIT and is described in detail in Steven
233 Levy's book `Hackers'. This book also includes a lot of information
234 about Stallman himself and the development of Lisp, a programming
235 language developed at MIT that underlies Emacs.)
239 * Through Version 18:: Unification prevails.
240 * Lucid Emacs:: One version 19 Emacs.
241 * GNU Emacs 19:: The other version 19 Emacs.
242 * GNU Emacs 20:: The other version 20 Emacs.
243 * XEmacs:: The continuation of Lucid Emacs.
246 File: internals.info, Node: Through Version 18, Next: Lucid Emacs, Up: A History of Emacs
251 Although the history of the early versions of GNU Emacs is unclear,
252 the history is well-known from the middle of 1985. A time line is:
254 * GNU Emacs version 15 (15.34) was released sometime in 1984 or 1985
255 and shared some code with a version of Emacs written by James
256 Gosling (the same James Gosling who later created the Java
259 * GNU Emacs version 16 (first released version was 16.56) was
260 released on July 15, 1985. All Gosling code was removed due to
261 potential copyright problems with the code.
263 * version 16.57: released on September 16, 1985.
265 * versions 16.58, 16.59: released on September 17, 1985.
267 * version 16.60: released on September 19, 1985. These later
268 version 16's incorporated patches from the net, esp. for getting
269 Emacs to work under System V.
271 * version 17.36 (first official v17 release) released on December 20,
272 1985. Included a TeX-able user manual. First official unpatched
273 version that worked on vanilla System V machines.
275 * version 17.43 (second official v17 release) released on January 25,
278 * version 17.45 released on January 30, 1986.
280 * version 17.46 released on February 4, 1986.
282 * version 17.48 released on February 10, 1986.
284 * version 17.49 released on February 12, 1986.
286 * version 17.55 released on March 18, 1986.
288 * version 17.57 released on March 27, 1986.
290 * version 17.58 released on April 4, 1986.
292 * version 17.61 released on April 12, 1986.
294 * version 17.63 released on May 7, 1986.
296 * version 17.64 released on May 12, 1986.
298 * version 18.24 (a beta version) released on October 2, 1986.
300 * version 18.30 (a beta version) released on November 15, 1986.
302 * version 18.31 (a beta version) released on November 23, 1986.
304 * version 18.32 (a beta version) released on December 7, 1986.
306 * version 18.33 (a beta version) released on December 12, 1986.
308 * version 18.35 (a beta version) released on January 5, 1987.
310 * version 18.36 (a beta version) released on January 21, 1987.
312 * January 27, 1987: The Great Usenet Renaming. net.emacs is now
315 * version 18.37 (a beta version) released on February 12, 1987.
317 * version 18.38 (a beta version) released on March 3, 1987.
319 * version 18.39 (a beta version) released on March 14, 1987.
321 * version 18.40 (a beta version) released on March 18, 1987.
323 * version 18.41 (the first "official" release) released on March 22,
326 * version 18.45 released on June 2, 1987.
328 * version 18.46 released on June 9, 1987.
330 * version 18.47 released on June 18, 1987.
332 * version 18.48 released on September 3, 1987.
334 * version 18.49 released on September 18, 1987.
336 * version 18.50 released on February 13, 1988.
338 * version 18.51 released on May 7, 1988.
340 * version 18.52 released on September 1, 1988.
342 * version 18.53 released on February 24, 1989.
344 * version 18.54 released on April 26, 1989.
346 * version 18.55 released on August 23, 1989. This is the earliest
347 version that is still available by FTP.
349 * version 18.56 released on January 17, 1991.
351 * version 18.57 released late January, 1991.
353 * version 18.58 released ?????.
355 * version 18.59 released October 31, 1992.
358 File: internals.info, Node: Lucid Emacs, Next: GNU Emacs 19, Prev: Through Version 18, Up: A History of Emacs
363 Lucid Emacs was developed by the (now-defunct) Lucid Inc., a maker of
364 C++ and Lisp development environments. It began when Lucid decided they
365 wanted to use Emacs as the editor and cornerstone of their C++
366 development environment (called "Energize"). They needed many features
367 that were not available in the existing version of GNU Emacs (version
368 18.5something), in particular good and integrated support for GUI
369 elements such as mouse support, multiple fonts, multiple window-system
370 windows, etc. A branch of GNU Emacs called Epoch, written at the
371 University of Illinois, existed that supplied many of these features;
372 however, Lucid needed more than what existed in Epoch. At the time, the
373 Free Software Foundation was working on version 19 of Emacs (this was
374 sometime around 1991), which was planned to have similar features, and
375 so Lucid decided to work with the Free Software Foundation. Their plan
376 was to add features that they needed, and coordinate with the FSF so
377 that the features would get included back into Emacs version 19.
379 Delays in the release of version 19 occurred, however (resulting in
380 it finally being released more than a year after what was initially
381 planned), and Lucid encountered unexpected technical resistance in
382 getting their changes merged back into version 19, so they decided to
383 release their own version of Emacs, which became Lucid Emacs 19.0.
385 The initial authors of Lucid Emacs were Matthieu Devin, Harlan
386 Sexton, and Eric Benson, and the work was later taken over by Jamie
387 Zawinski, who became "Mr. Lucid Emacs" for many releases.
389 A time line for Lucid Emacs/XEmacs is
391 * version 19.0 shipped with Energize 1.0, April 1992.
393 * version 19.1 released June 4, 1992.
395 * version 19.2 released June 19, 1992.
397 * version 19.3 released September 9, 1992.
399 * version 19.4 released January 21, 1993.
401 * version 19.5 was a repackaging of 19.4 with a few bug fixes and
402 shipped with Energize 2.0. Never released to the net.
404 * version 19.6 released April 9, 1993.
406 * version 19.7 was a repackaging of 19.6 with a few bug fixes and
407 shipped with Energize 2.1. Never released to the net.
409 * version 19.8 released September 6, 1993.
411 * version 19.9 released January 12, 1994.
413 * version 19.10 released May 27, 1994.
415 * version 19.11 (first XEmacs) released September 13, 1994.
417 * version 19.12 released June 23, 1995.
419 * version 19.13 released September 1, 1995.
421 * version 19.14 released June 23, 1996.
423 * version 20.0 released February 9, 1997.
425 * version 19.15 released March 28, 1997.
427 * version 20.1 (not released to the net) April 15, 1997.
429 * version 20.2 released May 16, 1997.
431 * version 19.16 released October 31, 1997.
433 * version 20.3 (the first stable version of XEmacs 20.x) released
434 November 30, 1997. version 20.4 released February 28, 1998.
437 File: internals.info, Node: GNU Emacs 19, Next: GNU Emacs 20, Prev: Lucid Emacs, Up: A History of Emacs
442 About a year after the initial release of Lucid Emacs, the FSF
443 released a beta of their version of Emacs 19 (referred to here as "GNU
444 Emacs"). By this time, the current version of Lucid Emacs was 19.6.
445 (Strangely, the first released beta from the FSF was GNU Emacs 19.7.) A
446 time line for GNU Emacs version 19 is
448 * version 19.8 (beta) released May 27, 1993.
450 * version 19.9 (beta) released May 27, 1993.
452 * version 19.10 (beta) released May 30, 1993.
454 * version 19.11 (beta) released June 1, 1993.
456 * version 19.12 (beta) released June 2, 1993.
458 * version 19.13 (beta) released June 8, 1993.
460 * version 19.14 (beta) released June 17, 1993.
462 * version 19.15 (beta) released June 19, 1993.
464 * version 19.16 (beta) released July 6, 1993.
466 * version 19.17 (beta) released late July, 1993.
468 * version 19.18 (beta) released August 9, 1993.
470 * version 19.19 (beta) released August 15, 1993.
472 * version 19.20 (beta) released November 17, 1993.
474 * version 19.21 (beta) released November 17, 1993.
476 * version 19.22 (beta) released November 28, 1993.
478 * version 19.23 (beta) released May 17, 1994.
480 * version 19.24 (beta) released May 16, 1994.
482 * version 19.25 (beta) released June 3, 1994.
484 * version 19.26 (beta) released September 11, 1994.
486 * version 19.27 (beta) released September 14, 1994.
488 * version 19.28 (first "official" release) released November 1, 1994.
490 * version 19.29 released June 21, 1995.
492 * version 19.30 released November 24, 1995.
494 * version 19.31 released May 25, 1996.
496 * version 19.32 released July 31, 1996.
498 * version 19.33 released August 11, 1996.
500 * version 19.34 released August 21, 1996.
502 * version 19.34b released September 6, 1996.
504 In some ways, GNU Emacs 19 was better than Lucid Emacs; in some ways,
505 worse. Lucid soon began incorporating features from GNU Emacs 19 into
506 Lucid Emacs; the work was mostly done by Richard Mlynarik, who had been
507 working on and using GNU Emacs for a long time (back as far as version
511 File: internals.info, Node: GNU Emacs 20, Next: XEmacs, Prev: GNU Emacs 19, Up: A History of Emacs
516 On February 2, 1997 work began on GNU Emacs to integrate Mule. The
517 first release was made in September of that year.
519 A timeline for Emacs 20 is
521 * version 20.1 released September 17, 1997.
523 * version 20.2 released September 20, 1997.
525 * version 20.3 released August 19, 1998.
528 File: internals.info, Node: XEmacs, Prev: GNU Emacs 20, Up: A History of Emacs
533 Around the time that Lucid was developing Energize, Sun Microsystems
534 was developing their own development environment (called "SPARCWorks")
535 and also decided to use Emacs. They joined forces with the Epoch team
536 at the University of Illinois and later with Lucid. The maintainer of
537 the last-released version of Epoch was Marc Andreessen, but he dropped
538 out and the Epoch project, headed by Simon Kaplan, lured Chuck Thompson
539 away from a system administration job to become the primary Lucid Emacs
540 author for Epoch and Sun. Chuck's area of specialty became the
541 redisplay engine (he replaced the old Lucid Emacs redisplay engine with
542 a ported version from Epoch and then later rewrote it from scratch).
543 Sun also hired Ben Wing (the author of Win-Emacs, a port of Lucid Emacs
544 to Microsoft Windows 3.1) in 1993, for what was initially a one-month
545 contract to fix some event problems but later became a many-year
546 involvement, punctuated by a six-month contract with Amdahl Corporation.
548 In 1994, Sun and Lucid agreed to rename Lucid Emacs to XEmacs (a name
549 not favorable to either company); the first release called XEmacs was
550 version 19.11. In June 1994, Lucid folded and Jamie quit to work for
551 the newly formed Mosaic Communications Corp., later Netscape
552 Communications Corp. (co-founded by the same Marc Andreessen, who had
553 quit his Epoch job to work on a graphical browser for the World Wide
554 Web). Chuck then become the primary maintainer of XEmacs, and put out
555 versions 19.11 through 19.14 in conjunction with Ben. For 19.12 and
556 19.13, Chuck added the new redisplay and many other display improvements
557 and Ben added MULE support (support for Asian and other languages) and
558 redesigned most of the internal Lisp subsystems to better support the
559 MULE work and the various other features being added to XEmacs. After
560 19.14 Chuck retired as primary maintainer and Steve Baur stepped in.
562 Soon after 19.13 was released, work began in earnest on the MULE
563 internationalization code and the source tree was divided into two
564 development paths. The MULE version was initially called 19.20, but was
565 soon renamed to 20.0. In 1996 Martin Buchholz of Sun Microsystems took
566 over the care and feeding of it and worked on it in parallel with the
567 19.14 development that was occurring at the same time. After much work
568 by Martin, it was decided to release 20.0 ahead of 19.15 in February
569 1997. The source tree remained divided until 20.2 when the version 19
570 source was finally retired at version 19.16.
572 In 1997, Sun finally dropped all pretense of support for XEmacs and
573 Martin Buchholz left the company in November. Since then, and mostly
574 for the previous year, because Steve Baur was never paid to work on
575 XEmacs, XEmacs has existed solely on the contributions of volunteers
576 from the Free Software Community. Starting from 1997, Hrvoje Niksic and
577 Kyle Jones have figured prominently in XEmacs development.
579 Many attempts have been made to merge XEmacs and GNU Emacs, but they
580 have consistently failed.
582 A more detailed history is contained in the XEmacs About page.
585 File: internals.info, Node: XEmacs From the Outside, Next: The Lisp Language, Prev: A History of Emacs, Up: Top
587 XEmacs From the Outside
588 ***********************
590 XEmacs appears to the outside world as an editor, but it is really a
591 Lisp environment. At its heart is a Lisp interpreter; it also
592 "happens" to contain many specialized object types (e.g. buffers,
593 windows, frames, events) that are useful for implementing an editor.
594 Some of these objects (in particular windows and frames) have
595 displayable representations, and XEmacs provides a function
596 `redisplay()' that ensures that the display of all such objects matches
597 their internal state. Most of the time, a standard Lisp environment is
598 in a "read-eval-print" loop - i.e. "read some Lisp code, execute it,
599 and print the results". XEmacs has a similar loop:
603 * dispatch the event (i.e. "do it")
607 Reading an event is done using the Lisp function `next-event', which
608 waits for something to happen (typically, the user presses a key or
609 moves the mouse) and returns an event object describing this.
610 Dispatching an event is done using the Lisp function `dispatch-event',
611 which looks up the event in a keymap object (a particular kind of
612 object that associates an event with a Lisp function) and calls that
613 function. The function "does" what the user has requested by changing
614 the state of particular frame objects, buffer objects, etc. Finally,
615 `redisplay()' is called, which updates the display to reflect those
616 changes just made. Thus is an "editor" born.
618 Note that you do not have to use XEmacs as an editor; you could just
619 as well make it do your taxes, compute pi, play bridge, etc. You'd just
620 have to write functions to do those operations in Lisp.
623 File: internals.info, Node: The Lisp Language, Next: XEmacs From the Perspective of Building, Prev: XEmacs From the Outside, Up: Top
628 Lisp is a general-purpose language that is higher-level than C and in
629 many ways more powerful than C. Powerful dialects of Lisp such as
630 Common Lisp are probably much better languages for writing very large
631 applications than is C. (Unfortunately, for many non-technical reasons
632 C and its successor C++ have become the dominant languages for
633 application development. These languages are both inadequate for
634 extremely large applications, which is evidenced by the fact that newer,
635 larger programs are becoming ever harder to write and are requiring ever
636 more programmers despite great increases in C development environments;
637 and by the fact that, although hardware speeds and reliability have been
638 growing at an exponential rate, most software is still generally
639 considered to be slow and buggy.)
641 The new Java language holds promise as a better general-purpose
642 development language than C. Java has many features in common with
643 Lisp that are not shared by C (this is not a coincidence, since Java
644 was designed by James Gosling, a former Lisp hacker). This will be
645 discussed more later.
647 For those used to C, here is a summary of the basic differences
650 1. Lisp has an extremely regular syntax. Every function, expression,
651 and control statement is written in the form
655 This is as opposed to C, which writes functions as
657 func(ARG1, ARG2, ...)
659 but writes expressions involving operators as (e.g.)
663 and writes control statements as (e.g.)
665 while (EXPR) { STATEMENT1; STATEMENT2; ... }
667 Lisp equivalents of the latter two would be
673 (while EXPR STATEMENT1 STATEMENT2 ...)
675 2. Lisp is a safe language. Assuming there are no bugs in the Lisp
676 interpreter/compiler, it is impossible to write a program that
677 "core dumps" or otherwise causes the machine to execute an illegal
678 instruction. This is very different from C, where perhaps the most
679 common outcome of a bug is exactly such a crash. A corollary of
680 this is that the C operation of casting a pointer is impossible
681 (and unnecessary) in Lisp, and that it is impossible to access
682 memory outside the bounds of an array.
684 3. Programs and data are written in the same form. The
685 parenthesis-enclosing form described above for statements is the
686 same form used for the most common data type in Lisp, the list.
687 Thus, it is possible to represent any Lisp program using Lisp data
688 types, and for one program to construct Lisp statements and then
689 dynamically "evaluate" them, or cause them to execute.
691 4. All objects are "dynamically typed". This means that part of every
692 object is an indication of what type it is. A Lisp program can
693 manipulate an object without knowing what type it is, and can
694 query an object to determine its type. This means that,
695 correspondingly, variables and function parameters can hold
696 objects of any type and are not normally declared as being of any
697 particular type. This is opposed to the "static typing" of C,
698 where variables can hold exactly one type of object and must be
699 declared as such, and objects do not contain an indication of
700 their type because it's implicit in the variables they are stored
701 in. It is possible in C to have a variable hold different types
702 of objects (e.g. through the use of `void *' pointers or
703 variable-argument functions), but the type information must then be
704 passed explicitly in some other fashion, leading to additional
707 5. Allocated memory is automatically reclaimed when it is no longer
708 in use. This operation is called "garbage collection" and
709 involves looking through all variables to see what memory is being
710 pointed to, and reclaiming any memory that is not pointed to and
711 is thus "inaccessible" and out of use. This is as opposed to C,
712 in which allocated memory must be explicitly reclaimed using
713 `free()'. If you simply drop all pointers to memory without
714 freeing it, it becomes "leaked" memory that still takes up space.
715 Over a long period of time, this can cause your program to grow
716 and grow until it runs out of memory.
718 6. Lisp has built-in facilities for handling errors and exceptions.
719 In C, when an error occurs, usually either the program exits
720 entirely or the routine in which the error occurs returns a value
721 indicating this. If an error occurs in a deeply-nested routine,
722 then every routine currently called must unwind itself normally
723 and return an error value back up to the next routine. This means
724 that every routine must explicitly check for an error in all the
725 routines it calls; if it does not do so, unexpected and often
726 random behavior results. This is an extremely common source of
727 bugs in C programs. An alternative would be to do a non-local
728 exit using `longjmp()', but that is often very dangerous because
729 the routines that were exited past had no opportunity to clean up
730 after themselves and may leave things in an inconsistent state,
731 causing a crash shortly afterwards.
733 Lisp provides mechanisms to make such non-local exits safe. When
734 an error occurs, a routine simply signals that an error of a
735 particular class has occurred, and a non-local exit takes place.
736 Any routine can trap errors occurring in routines it calls by
737 registering an error handler for some or all classes of errors.
738 (If no handler is registered, a default handler, generally
739 installed by the top-level event loop, is executed; this prints
740 out the error and continues.) Routines can also specify cleanup
741 code (called an "unwind-protect") that will be called when control
742 exits from a block of code, no matter how that exit occurs - i.e.
743 even if a function deeply nested below it causes a non-local exit
744 back to the top level.
746 Note that this facility has appeared in some recent vintages of C,
747 in particular Visual C++ and other PC compilers written for the
750 7. In Emacs Lisp, local variables are "dynamically scoped". This
751 means that if you declare a local variable in a particular
752 function, and then call another function, that subfunction can
753 "see" the local variable you declared. This is actually
754 considered a bug in Emacs Lisp and in all other early dialects of
755 Lisp, and was corrected in Common Lisp. (In Common Lisp, you can
756 still declare dynamically scoped variables if you want to - they
757 are sometimes useful - but variables by default are "lexically
760 For those familiar with Lisp, Emacs Lisp is modelled after MacLisp,
761 an early dialect of Lisp developed at MIT (no relation to the Macintosh
762 computer). There is a Common Lisp compatibility package available for
763 Emacs that provides many of the features of Common Lisp.
765 The Java language is derived in many ways from C, and shares a
766 similar syntax, but has the following features in common with Lisp (and
769 1. Java is a safe language, like Lisp.
771 2. Java provides garbage collection, like Lisp.
773 3. Java has built-in facilities for handling errors and exceptions,
776 4. Java has a type system that combines the best advantages of both
777 static and dynamic typing. Objects (except very simple types) are
778 explicitly marked with their type, as in dynamic typing; but there
779 is a hierarchy of types and functions are declared to accept only
780 certain types, thus providing the increased compile-time
781 error-checking of static typing.
783 The Java language also has some negative attributes:
785 1. Java uses the edit/compile/run model of software development. This
786 makes it hard to use interactively. For example, to use Java like
787 `bc' it is necessary to write a special purpose, albeit tiny,
788 application. In Emacs Lisp, a calculator comes built-in without
789 any effort - one can always just type an expression in the
792 2. Java tries too hard to enforce, not merely enable, portability,
793 making ordinary access to standard OS facilities painful. Java
794 has an "agenda". I think this is why `chdir' is not part of
795 standard Java, which is inexcusable.
797 Unfortunately, there is no perfect language. Static typing allows a
798 compiler to catch programmer errors and produce more efficient code, but
799 makes programming more tedious and less fun. For the forseeable future,
800 an Ideal Editing and Programming Environment (and that is what XEmacs
801 aspires to) will be programmable in multiple languages: high level ones
802 like Lisp for user customization and prototyping, and lower level ones
803 for infrastructure and industrial strength applications. If I had my
804 way, XEmacs would be friendly towards the Python, Scheme, C++, ML,
805 etc... communities. But there are serious technical difficulties to
808 The word "application" in the previous paragraph was used
809 intentionally. XEmacs implements an API for programs written in Lisp
810 that makes it a full-fledged application platform, very much like an OS
814 File: internals.info, Node: XEmacs From the Perspective of Building, Next: XEmacs From the Inside, Prev: The Lisp Language, Up: Top
816 XEmacs From the Perspective of Building
817 ***************************************
819 The heart of XEmacs is the Lisp environment, which is written in C.
820 This is contained in the `src/' subdirectory. Underneath `src/' are
821 two subdirectories of header files: `s/' (header files for particular
822 operating systems) and `m/' (header files for particular machine
823 types). In practice the distinction between the two types of header
824 files is blurred. These header files define or undefine certain
825 preprocessor constants and macros to indicate particular
826 characteristics of the associated machine or operating system. As part
827 of the configure process, one `s/' file and one `m/' file is identified
828 for the particular environment in which XEmacs is being built.
830 XEmacs also contains a great deal of Lisp code. This implements the
831 operations that make XEmacs useful as an editor as well as just a Lisp
832 environment, and also contains many add-on packages that allow XEmacs to
833 browse directories, act as a mail and Usenet news reader, compile Lisp
834 code, etc. There is actually more Lisp code than C code associated with
835 XEmacs, but much of the Lisp code is peripheral to the actual operation
836 of the editor. The Lisp code all lies in subdirectories underneath the
839 The `lwlib/' directory contains C code that implements a generalized
840 interface onto different X widget toolkits and also implements some
841 widgets of its own that behave like Motif widgets but are faster, free,
842 and in some cases more powerful. The code in this directory compiles
843 into a library and is mostly independent from XEmacs.
845 The `etc/' directory contains various data files associated with
846 XEmacs. Some of them are actually read by XEmacs at startup; others
847 merely contain useful information of various sorts.
849 The `lib-src/' directory contains C code for various auxiliary
850 programs that are used in connection with XEmacs. Some of them are used
851 during the build process; others are used to perform certain functions
852 that cannot conveniently be placed in the XEmacs executable (e.g. the
853 `movemail' program for fetching mail out of `/var/spool/mail', which
854 must be setgid to `mail' on many systems; and the `gnuclient' program,
855 which allows an external script to communicate with a running XEmacs
858 The `man/' directory contains the sources for the XEmacs
859 documentation. It is mostly in a form called Texinfo, which can be
860 converted into either a printed document (by passing it through TeX) or
861 into on-line documentation called "info files".
863 The `info/' directory contains the results of formatting the XEmacs
864 documentation as "info files", for on-line use. These files are used
865 when you enter the Info system using `C-h i' or through the Help menu.
867 The `dynodump/' directory contains auxiliary code used to build
868 XEmacs on Solaris platforms.
870 The other directories contain various miscellaneous code and
871 information that is not normally used or needed.
873 The first step of building involves running the `configure' program
874 and passing it various parameters to specify any optional features you
875 want and compiler arguments and such, as described in the `INSTALL'
876 file. This determines what the build environment is, chooses the
877 appropriate `s/' and `m/' file, and runs a series of tests to determine
878 many details about your environment, such as which library functions
879 are available and exactly how they work. The reason for running these
880 tests is that it allows XEmacs to be compiled on a much wider variety
881 of platforms than those that the XEmacs developers happen to be
882 familiar with, including various sorts of hybrid platforms. This is
883 especially important now that many operating systems give you a great
884 deal of control over exactly what features you want installed, and allow
885 for easy upgrading of parts of a system without upgrading the rest. It
886 would be impossible to pre-determine and pre-specify the information for
887 all possible configurations.
889 In fact, the `s/' and `m/' files are basically _evil_, since they
890 contain unmaintainable platform-specific hard-coded information.
891 XEmacs has been moving in the direction of having all system-specific
892 information be determined dynamically by `configure'. Perhaps someday
893 we can `rm -rf src/s src/m'.
895 When configure is done running, it generates `Makefile's and
896 `GNUmakefile's and the file `src/config.h' (which describes the
897 features of your system) from template files. You then run `make',
898 which compiles the auxiliary code and programs in `lib-src/' and
899 `lwlib/' and the main XEmacs executable in `src/'. The result of
900 compiling and linking is an executable called `temacs', which is _not_
901 the final XEmacs executable. `temacs' by itself is not intended to
902 function as an editor or even display any windows on the screen, and if
903 you simply run it, it will exit immediately. The `Makefile' runs
904 `temacs' with certain options that cause it to initialize itself, read
905 in a number of basic Lisp files, and then dump itself out into a new
906 executable called `xemacs'. This new executable has been
907 pre-initialized and contains pre-digested Lisp code that is necessary
908 for the editor to function (this includes most basic editing functions,
909 e.g. `kill-line', that can be defined in terms of other Lisp
910 primitives; some initialization code that is called when certain
911 objects, such as frames, are created; and all of the standard
912 keybindings and code for the actions they result in). This executable,
913 `xemacs', is the executable that you run to use the XEmacs editor.
915 Although `temacs' is not intended to be run as an editor, it can, by
916 using the incantation `temacs -batch -l loadup.el run-temacs'. This is
917 useful when the dumping procedure described above is broken, or when
918 using certain program debugging tools such as Purify. These tools get
919 mighty confused by the tricks played by the XEmacs build process, such
920 as allocation memory in one process, and freeing it in the next.
923 File: internals.info, Node: XEmacs From the Inside, Next: The XEmacs Object System (Abstractly Speaking), Prev: XEmacs From the Perspective of Building, Up: Top
925 XEmacs From the Inside
926 **********************
928 Internally, XEmacs is quite complex, and can be very confusing. To
929 simplify things, it can be useful to think of XEmacs as containing an
930 event loop that "drives" everything, and a number of other subsystems,
931 such as a Lisp engine and a redisplay mechanism. Each of these other
932 subsystems exists simultaneously in XEmacs, and each has a certain
933 state. The flow of control continually passes in and out of these
934 different subsystems in the course of normal operation of the editor.
936 It is important to keep in mind that, most of the time, the editor is
937 "driven" by the event loop. Except during initialization and batch
938 mode, all subsystems are entered directly or indirectly through the
939 event loop, and ultimately, control exits out of all subsystems back up
940 to the event loop. This cycle of entering a subsystem, exiting back out
941 to the event loop, and starting another iteration of the event loop
942 occurs once each keystroke, mouse motion, etc.
944 If you're trying to understand a particular subsystem (other than the
945 event loop), think of it as a "daemon" process or "servant" that is
946 responsible for one particular aspect of a larger system, and
947 periodically receives commands or environment changes that cause it to
948 do something. Ultimately, these commands and environment changes are
949 always triggered by the event loop. For example:
951 * The window and frame mechanism is responsible for keeping track of
952 what windows and frames exist, what buffers are in them, etc. It
953 is periodically given commands (usually from the user) to make a
954 change to the current window/frame state: i.e. create a new frame,
955 delete a window, etc.
957 * The buffer mechanism is responsible for keeping track of what
958 buffers exist and what text is in them. It is periodically given
959 commands (usually from the user) to insert or delete text, create
960 a buffer, etc. When it receives a text-change command, it
961 notifies the redisplay mechanism.
963 * The redisplay mechanism is responsible for making sure that
964 windows and frames are displayed correctly. It is periodically
965 told (by the event loop) to actually "do its job", i.e. snoop
966 around and see what the current state of the environment (mostly
967 of the currently-existing windows, frames, and buffers) is, and
968 make sure that that state matches what's actually displayed. It
969 keeps lots and lots of information around (such as what is
970 actually being displayed currently, and what the environment was
971 last time it checked) so that it can minimize the work it has to
972 do. It is also helped along in that whenever a relevant change to
973 the environment occurs, the redisplay mechanism is told about
974 this, so it has a pretty good idea of where it has to look to find
975 possible changes and doesn't have to look everywhere.
977 * The Lisp engine is responsible for executing the Lisp code in
978 which most user commands are written. It is entered through a
979 call to `eval' or `funcall', which occurs as a result of
980 dispatching an event from the event loop. The functions it calls
981 issue commands to the buffer mechanism, the window/frame
984 * The Lisp allocation subsystem is responsible for keeping track of
985 Lisp objects. It is given commands from the Lisp engine to
986 allocate objects, garbage collect, etc.
990 The important idea here is that there are a number of independent
991 subsystems each with its own responsibility and persistent state, just
992 like different employees in a company, and each subsystem is
993 periodically given commands from other subsystems. Commands can flow
994 from any one subsystem to any other, but there is usually some sort of
995 hierarchy, with all commands originating from the event subsystem.
997 XEmacs is entered in `main()', which is in `emacs.c'. When this is
998 called the first time (in a properly-invoked `temacs'), it does the
1001 1. It does some very basic environment initializations, such as
1002 determining where it and its directories (e.g. `lisp/' and `etc/')
1003 reside and setting up signal handlers.
1005 2. It initializes the entire Lisp interpreter.
1007 3. It sets the initial values of many built-in variables (including
1008 many variables that are visible to Lisp programs), such as the
1009 global keymap object and the built-in faces (a face is an object
1010 that describes the display characteristics of text). This
1011 involves creating Lisp objects and thus is dependent on step (2).
1013 4. It performs various other initializations that are relevant to the
1014 particular environment it is running in, such as retrieving
1015 environment variables, determining the current date and the user
1016 who is running the program, examining its standard input, creating
1017 any necessary file descriptors, etc.
1019 5. At this point, the C initialization is complete. A Lisp program
1020 that was specified on the command line (usually `loadup.el') is
1021 called (temacs is normally invoked as `temacs -batch -l loadup.el
1022 dump'). `loadup.el' loads all of the other Lisp files that are
1023 needed for the operation of the editor, calls the `dump-emacs'
1024 function to write out `xemacs', and then kills the temacs process.
1026 When `xemacs' is then run, it only redoes steps (1) and (4) above;
1027 all variables already contain the values they were set to when the
1028 executable was dumped, and all memory that was allocated with
1029 `malloc()' is still around. (XEmacs knows whether it is being run as
1030 `xemacs' or `temacs' because it sets the global variable `initialized'
1031 to 1 after step (4) above.) At this point, `xemacs' calls a Lisp
1032 function to do any further initialization, which includes parsing the
1033 command-line (the C code can only do limited command-line parsing,
1034 which includes looking for the `-batch' and `-l' flags and a few other
1035 flags that it needs to know about before initialization is complete),
1036 creating the first frame (or "window" in standard window-system
1037 parlance), running the user's init file (usually the file `.emacs' in
1038 the user's home directory), etc. The function to do this is usually
1039 called `normal-top-level'; `loadup.el' tells the C code about this
1040 function by setting its name as the value of the Lisp variable
1043 When the Lisp initialization code is done, the C code enters the
1044 event loop, and stays there for the duration of the XEmacs process.
1045 The code for the event loop is contained in `keyboard.c', and is called
1046 `Fcommand_loop_1()'. Note that this event loop could very well be
1047 written in Lisp, and in fact a Lisp version exists; but apparently,
1048 doing this makes XEmacs run noticeably slower.
1050 Notice how much of the initialization is done in Lisp, not in C. In
1051 general, XEmacs tries to move as much code as is possible into Lisp.
1052 Code that remains in C is code that implements the Lisp interpreter
1053 itself, or code that needs to be very fast, or code that needs to do
1054 system calls or other such stuff that needs to be done in C, or code
1055 that needs to have access to "forbidden" structures. (One conscious
1056 aspect of the design of Lisp under XEmacs is a clean separation between
1057 the external interface to a Lisp object's functionality and its internal
1058 implementation. Part of this design is that Lisp programs are
1059 forbidden from accessing the contents of the object other than through
1060 using a standard API. In this respect, XEmacs Lisp is similar to
1061 modern Lisp dialects but differs from GNU Emacs, which tends to expose
1062 the implementation and allow Lisp programs to look at it directly. The
1063 major advantage of hiding the implementation is that it allows the
1064 implementation to be redesigned without affecting any Lisp programs,
1065 including those that might want to be "clever" by looking directly at
1066 the object's contents and possibly manipulating them.)
1068 Moving code into Lisp makes the code easier to debug and maintain and
1069 makes it much easier for people who are not XEmacs developers to
1070 customize XEmacs, because they can make a change with much less chance
1071 of obscure and unwanted interactions occurring than if they were to