1 This is ../info/internals.info, produced by makeinfo version 4.0b 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: The XEmacs Object System (Abstractly Speaking), Next: How Lisp Objects Are Represented in C, Prev: XEmacs From the Inside, Up: Top
43 The XEmacs Object System (Abstractly Speaking)
44 **********************************************
46 At the heart of the Lisp interpreter is its management of objects.
47 XEmacs Lisp contains many built-in objects, some of which are simple
48 and others of which can be very complex; and some of which are very
49 common, and others of which are rarely used or are only used
50 internally. (Since the Lisp allocation system, with its automatic
51 reclamation of unused storage, is so much more convenient than
52 `malloc()' and `free()', the C code makes extensive use of it in its
55 The basic Lisp objects are
58 28 or 31 bits of precision, or 60 or 63 bits on 64-bit machines;
59 the reason for this is described below when the internal Lisp
60 object representation is described.
63 Same precision as a double in C.
66 A simple container for two Lisp objects, used to implement lists
67 and most other data structures in Lisp.
70 An object representing a single character of text; chars behave
71 like integers in many ways but are logically considered text
72 rather than numbers and have a different read syntax. (the read
73 syntax for a char contains the char itself or some textual
74 encoding of it--for example, a Japanese Kanji character might be
75 encoded as `^[$(B#&^[(B' using the ISO-2022 encoding
76 standard--rather than the numerical representation of the char;
77 this way, if the mapping between chars and integers changes, which
78 is quite possible for Kanji characters and other extended
79 characters, the same character will still be created. Note that
80 some primitives confuse chars and integers. The worst culprit is
81 `eq', which makes a special exception and considers a char to be
82 `eq' to its integer equivalent, even though in no other case are
83 objects of two different types `eq'. The reason for this
84 monstrosity is compatibility with existing code; the separation of
85 char from integer came fairly recently.)
88 An object that contains Lisp objects and is referred to by name;
89 symbols are used to implement variables and named functions and to
90 provide the equivalent of preprocessor constants in C.
93 A one-dimensional array of Lisp objects providing constant-time
94 access to any of the objects; access to an arbitrary object in a
95 vector is faster than for lists, but the operations that can be
96 done on a vector are more limited.
99 Self-explanatory; behaves much like a vector of chars but has a
100 different read syntax and is stored and manipulated more compactly.
103 A vector of bits; similar to a string in spirit.
106 An object containing compiled Lisp code, known as "byte code".
109 A Lisp primitive, i.e. a Lisp-callable function implemented in C.
111 Note that there is no basic "function" type, as in more powerful
112 versions of Lisp (where it's called a "closure"). XEmacs Lisp does not
113 provide the closure semantics implemented by Common Lisp and Scheme.
114 The guts of a function in XEmacs Lisp are represented in one of four
115 ways: a symbol specifying another function (when one function is an
116 alias for another), a list (whose first element must be the symbol
117 `lambda') containing the function's source code, a compiled-function
118 object, or a subr object. (In other words, given a symbol specifying
119 the name of a function, calling `symbol-function' to retrieve the
120 contents of the symbol's function cell will return one of these types
123 XEmacs Lisp also contains numerous specialized objects used to
124 implement the editor:
127 Stores text like a string, but is optimized for insertion and
128 deletion and has certain other properties that can be set.
131 An object with various properties whose displayable representation
132 is a "window" in window-system parlance.
135 A section of a frame that displays the contents of a buffer; often
136 called a "pane" in window-system parlance.
138 `window-configuration'
139 An object that represents a saved configuration of windows in a
143 An object representing a screen on which frames can be displayed;
144 equivalent to a "display" in the X Window System and a "TTY" in
148 An object specifying the appearance of text or graphics; it has
149 properties such as font, foreground color, and background color.
152 An object that refers to a particular position in a buffer and
153 moves around as text is inserted and deleted to stay in the same
154 relative position to the text around it.
157 Similar to a marker but covers a range of text in a buffer; can
158 also specify properties of the text, such as a face in which the
159 text is to be displayed, whether the text is invisible or
163 Generated by calling `next-event' and contains information
164 describing a particular event happening in the system, such as the
165 user pressing a key or a process terminating.
168 An object that maps from events (described using lists, vectors,
169 and symbols rather than with an event object because the mapping
170 is for classes of events, rather than individual events) to
171 functions to execute or other events to recursively look up; the
172 functions are described by name, using a symbol, or using lists to
173 specify the function's code.
176 An object that describes the appearance of an image (e.g. pixmap)
177 on the screen; glyphs can be attached to the beginning or end of
178 extents and in some future version of XEmacs will be able to be
179 inserted directly into a buffer.
182 An object that describes a connection to an externally-running
185 There are some other, less-commonly-encountered general objects:
188 An object that maps from an arbitrary Lisp object to another
189 arbitrary Lisp object, using hashing for fast lookup.
192 A limited form of hash-table that maps from strings to symbols;
193 obarrays are used to look up a symbol given its name and are not
194 actually their own object type but are kludgily represented using
195 vectors with hidden fields (this representation derives from GNU
199 A complex object used to specify the value of a display property; a
200 default value is given and different values can be specified for
201 particular frames, buffers, windows, devices, or classes of device.
204 An object that maps from chars or classes of chars to arbitrary
205 Lisp objects; internally char tables use a complex nested-vector
206 representation that is optimized to the way characters are
207 represented as integers.
210 An object that maps from ranges of integers to arbitrary Lisp
213 And some strange special-purpose objects:
217 Objects used when MULE, or multi-lingual/Asian-language, support is
223 An object that encapsulates a window-system resource; instances are
224 mostly used internally but are exposed on the Lisp level for
225 cleanness of the specifier model and because it's occasionally
226 useful for Lisp program to create or query the properties of
230 An object that encapsulate a "subwindow" resource, i.e. a
231 window-system child window that is drawn into by an external
232 process; this object should be integrated into the glyph system
233 but isn't yet, and may change form when this is done.
237 Objects that represent resources used in the ToolTalk interprocess
238 communication protocol.
241 An object used in conjunction with the toolbar.
243 And objects that are only used internally:
246 A generic object for encapsulating arbitrary memory; this allows
247 you the generality of `malloc()' and the convenience of the Lisp
251 A buffering I/O stream, used to provide a unified interface to
252 anything that can accept output or provide input, such as a file
253 descriptor, a stdio stream, a chunk of memory, a Lisp buffer, a
254 Lisp string, etc.; it's a Lisp object to make its memory
255 management more convenient.
258 Subsidiary objects in the internal char-table representation.
263 Various special-purpose objects that are basically just used to
264 encapsulate memory for particular subsystems, similar to the more
265 general "opaque" object.
267 `symbol-value-forward'
268 `symbol-value-buffer-local'
269 `symbol-value-varalias'
270 `symbol-value-lisp-magic'
271 Special internal-only objects that are placed in the value cell of
272 a symbol to indicate that there is something special with this
273 variable - e.g. it has no value, it mirrors another variable, or
274 it mirrors some C variable; there is really only one kind of
275 object, called a "symbol-value-magic", but it is sort-of halfway
276 kludged into semi-different object types.
278 Some types of objects are "permanent", meaning that once created,
279 they do not disappear until explicitly destroyed, using a function such
280 as `delete-buffer', `delete-window', `delete-frame', etc. Others will
281 disappear once they are not longer used, through the garbage collection
282 mechanism. Buffers, frames, windows, devices, and processes are among
283 the objects that are permanent. Note that some objects can go both
284 ways: Faces can be created either way; extents are normally permanent,
285 but detached extents (extents not referring to any text, as happens to
286 some extents when the text they are referring to is deleted) are
287 temporary. Note that some permanent objects, such as faces and coding
288 systems, cannot be deleted. Note also that windows are unique in that
289 they can be _undeleted_ after having previously been deleted. (This
290 happens as a result of restoring a window configuration.)
292 Note that many types of objects have a "read syntax", i.e. a way of
293 specifying an object of that type in Lisp code. When you load a Lisp
294 file, or type in code to be evaluated, what really happens is that the
295 function `read' is called, which reads some text and creates an object
296 based on the syntax of that text; then `eval' is called, which possibly
297 does something special; then this loop repeats until there's no more
298 text to read. (`eval' only actually does something special with
299 symbols, which causes the symbol's value to be returned, similar to
300 referencing a variable; and with conses [i.e. lists], which cause a
301 function invocation. All other values are returned unchanged.)
307 converts to an integer whose value is 17297.
311 converts to a float whose value is 1.983e-4, or .0001983.
315 converts to a char that represents the lowercase letter b.
319 (where `^[' actually is an `ESC' character) converts to a particular
320 Kanji character when using an ISO2022-based coding system for input.
321 (To decode this goo: `ESC' begins an escape sequence; `ESC $ (' is a
322 class of escape sequences meaning "switch to a 94x94 character set";
323 `ESC $ ( B' means "switch to Japanese Kanji"; `#' and `&' collectively
324 index into a 94-by-94 array of characters [subtract 33 from the ASCII
325 value of each character to get the corresponding index]; `ESC (' is a
326 class of escape sequences meaning "switch to a 94 character set"; `ESC
327 (B' means "switch to US ASCII". It is a coincidence that the letter
328 `B' is used to denote both Japanese Kanji and US ASCII. If the first
329 `B' were replaced with an `A', you'd be requesting a Chinese Hanzi
330 character from the GB2312 character set.)
334 converts to a string.
338 converts to a symbol whose name is `"foobar"'. This is done by
339 looking up the string equivalent in the global variable `obarray',
340 whose contents should be an obarray. If no symbol is found, a new
341 symbol with the name `"foobar"' is automatically created and added to
342 `obarray'; this process is called "interning" the symbol.
346 converts to a cons cell containing the symbols `foo' and `bar'.
350 converts to a three-element list containing the specified objects
351 (note that a list is actually a set of nested conses; see the XEmacs
356 converts to a three-element vector containing the specified objects.
360 converts to a compiled-function object (the actual contents are not
361 shown since they are not relevant here; look at a file that ends with
362 `.elc' for examples).
366 converts to a bit-vector.
368 #s(hash-table ... ...)
370 converts to a hash table (the actual contents are not shown).
372 #s(range-table ... ...)
374 converts to a range table (the actual contents are not shown).
376 #s(char-table ... ...)
378 converts to a char table (the actual contents are not shown).
380 Note that the `#s()' syntax is the general syntax for structures,
381 which are not really implemented in XEmacs Lisp but should be.
383 When an object is printed out (using `print' or a related function),
384 the read syntax is used, so that the same object can be read in again.
386 The other objects do not have read syntaxes, usually because it does
387 not really make sense to create them in this fashion (i.e. processes,
388 where it doesn't make sense to have a subprocess created as a side
389 effect of reading some Lisp code), or because they can't be created at
390 all (e.g. subrs). Permanent objects, as a rule, do not have a read
391 syntax; nor do most complex objects, which contain too much state to be
392 easily initialized through a read syntax.
395 File: internals.info, Node: How Lisp Objects Are Represented in C, Next: Rules When Writing New C Code, Prev: The XEmacs Object System (Abstractly Speaking), Up: Top
397 How Lisp Objects Are Represented in C
398 *************************************
400 Lisp objects are represented in C using a 32-bit or 64-bit machine
401 word (depending on the processor; i.e. DEC Alphas use 64-bit Lisp
402 objects and most other processors use 32-bit Lisp objects). The
403 representation stuffs a pointer together with a tag, as follows:
405 [ 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 ]
406 [ 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 ]
408 <---------------------------------------------------------> <->
409 a pointer to a structure, or an integer tag
411 A tag of 00 is used for all pointer object types, a tag of 10 is used
412 for characters, and the other two tags 01 and 11 are joined together to
413 form the integer object type. This representation gives us 31 bit
414 integers and 30 bit characters, while pointers are represented directly
415 without any bit masking or shifting. This representation, though,
416 assumes that pointers to structs are always aligned to multiples of 4,
417 so the lower 2 bits are always zero.
419 Lisp objects use the typedef `Lisp_Object', but the actual C type
420 used for the Lisp object can vary. It can be either a simple type
421 (`long' on the DEC Alpha, `int' on other machines) or a structure whose
422 fields are bit fields that line up properly (actually, a union of
423 structures is used). Generally the simple integral type is preferable
424 because it ensures that the compiler will actually use a machine word
425 to represent the object (some compilers will use more general and less
426 efficient code for unions and structs even if they can fit in a machine
427 word). The union type, however, has the advantage of stricter type
428 checking. If you accidentally pass an integer where a Lisp object is
429 desired, you get a compile error. The choice of which type to use is
430 determined by the preprocessor constant `USE_UNION_TYPE' which is
431 defined via the `--use-union-type' option to `configure'.
433 Various macros are used to convert between Lisp_Objects and the
434 corresponding C type. Macros of the form `XINT()', `XCHAR()',
435 `XSTRING()', `XSYMBOL()', do any required bit shifting and/or masking
436 and cast it to the appropriate type. `XINT()' needs to be a bit tricky
437 so that negative numbers are properly sign-extended. Since integers
438 are stored left-shifted, if the right-shift operator does an arithmetic
439 shift (i.e. it leaves the most-significant bit as-is rather than
440 shifting in a zero, so that it mimics a divide-by-two even for negative
441 numbers) the shift to remove the tag bit is enough. This is the case
442 on all the systems we support.
444 Note that when `ERROR_CHECK_TYPECHECK' is defined, the converter
445 macros become more complicated--they check the tag bits and/or the type
446 field in the first four bytes of a record type to ensure that the
447 object is really of the correct type. This is great for catching places
448 where an incorrect type is being dereferenced--this typically results
449 in a pointer being dereferenced as the wrong type of structure, with
450 unpredictable (and sometimes not easily traceable) results.
452 There are similar `XSETTYPE()' macros that construct a Lisp object.
453 These macros are of the form `XSETTYPE (LVALUE, RESULT)', i.e. they
454 have to be a statement rather than just used in an expression. The
455 reason for this is that standard C doesn't let you "construct" a
456 structure (but GCC does). Granted, this sometimes isn't too
457 convenient; for the case of integers, at least, you can use the
458 function `make_int()', which constructs and _returns_ an integer Lisp
459 object. Note that the `XSETTYPE()' macros are also affected by
460 `ERROR_CHECK_TYPECHECK' and make sure that the structure is of the
461 right type in the case of record types, where the type is contained in
464 The C programmer is responsible for *guaranteeing* that a
465 Lisp_Object is the correct type before using the `XTYPE' macros. This
466 is especially important in the case of lists. Use `XCAR' and `XCDR' if
467 a Lisp_Object is certainly a cons cell, else use `Fcar()' and `Fcdr()'.
468 Trust other C code, but not Lisp code. On the other hand, if XEmacs
469 has an internal logic error, it's better to crash immediately, so
470 sprinkle `assert()'s and "unreachable" `abort()'s liberally about the
471 source code. Where performance is an issue, use `type_checking_assert',
472 `bufpos_checking_assert', and `gc_checking_assert', which do nothing
473 unless the corresponding configure error checking flag was specified.
476 File: internals.info, Node: Rules When Writing New C Code, Next: A Summary of the Various XEmacs Modules, Prev: How Lisp Objects Are Represented in C, Up: Top
478 Rules When Writing New C Code
479 *****************************
481 The XEmacs C Code is extremely complex and intricate, and there are
482 many rules that are more or less consistently followed throughout the
483 code. Many of these rules are not obvious, so they are explained here.
484 It is of the utmost importance that you follow them. If you don't,
485 you may get something that appears to work, but which will crash in odd
486 situations, often in code far away from where the actual breakage is.
490 * General Coding Rules::
491 * Writing Lisp Primitives::
492 * Adding Global Lisp Variables::
494 * Techniques for XEmacs Developers::
497 File: internals.info, Node: General Coding Rules, Next: Writing Lisp Primitives, Prev: Rules When Writing New C Code, Up: Rules When Writing New C Code
502 The C code is actually written in a dialect of C called "Clean C",
503 meaning that it can be compiled, mostly warning-free, with either a C or
504 C++ compiler. Coding in Clean C has several advantages over plain C.
505 C++ compilers are more nit-picking, and a number of coding errors have
506 been found by compiling with C++. The ability to use both C and C++
507 tools means that a greater variety of development tools are available to
510 Every module includes `<config.h>' (angle brackets so that
511 `--srcdir' works correctly; `config.h' may or may not be in the same
512 directory as the C sources) and `lisp.h'. `config.h' must always be
513 included before any other header files (including system header files)
514 to ensure that certain tricks played by various `s/' and `m/' files
517 When including header files, always use angle brackets, not double
518 quotes, except when the file to be included is always in the same
519 directory as the including file. If either file is a generated file,
520 then that is not likely to be the case. In order to understand why we
521 have this rule, imagine what happens when you do a build in the source
522 directory using `./configure' and another build in another directory
523 using `../work/configure'. There will be two different `config.h'
524 files. Which one will be used if you `#include "config.h"'?
526 Almost every module contains a `syms_of_*()' function and a
527 `vars_of_*()' function. The former declares any Lisp primitives you
528 have defined and defines any symbols you will be using. The latter
529 declares any global Lisp variables you have added and initializes global
530 C variables in the module. *Important*: There are stringent
531 requirements on exactly what can go into these functions. See the
532 comment in `emacs.c'. The reason for this is to avoid obscure unwanted
533 interactions during initialization. If you don't follow these rules,
534 you'll be sorry! If you want to do anything that isn't allowed, create
535 a `complex_vars_of_*()' function for it. Doing this is tricky, though:
536 you have to make sure your function is called at the right time so that
537 all the initialization dependencies work out.
539 Declare each function of these kinds in `symsinit.h'. Make sure
540 it's called in the appropriate place in `emacs.c'. You never need to
541 include `symsinit.h' directly, because it is included by `lisp.h'.
543 *All global and static variables that are to be modifiable must be
544 declared uninitialized.* This means that you may not use the "declare
545 with initializer" form for these variables, such as `int some_variable
546 = 0;'. The reason for this has to do with some kludges done during the
547 dumping process: If possible, the initialized data segment is re-mapped
548 so that it becomes part of the (unmodifiable) code segment in the
549 dumped executable. This allows this memory to be shared among multiple
550 running XEmacs processes. XEmacs is careful to place as much constant
551 data as possible into initialized variables during the `temacs' phase.
553 *Please note:* This kludge only works on a few systems nowadays, and
554 is rapidly becoming irrelevant because most modern operating systems
555 provide "copy-on-write" semantics. All data is initially shared
556 between processes, and a private copy is automatically made (on a
557 page-by-page basis) when a process first attempts to write to a page of
560 Formerly, there was a requirement that static variables not be
561 declared inside of functions. This had to do with another hack along
562 the same vein as what was just described: old USG systems put
563 statically-declared variables in the initialized data space, so those
564 header files had a `#define static' declaration. (That way, the
565 data-segment remapping described above could still work.) This fails
566 badly on static variables inside of functions, which suddenly become
567 automatic variables; therefore, you weren't supposed to have any of
568 them. This awful kludge has been removed in XEmacs because
570 1. almost all of the systems that used this kludge ended up having to
571 disable the data-segment remapping anyway;
573 2. the only systems that didn't were extremely outdated ones;
575 3. this hack completely messed up inline functions.
577 The C source code makes heavy use of C preprocessor macros. One
578 popular macro style is:
580 #define FOO(var, value) do { \
581 Lisp_Object FOO_value = (value); \
582 ... /* compute using FOO_value */ \
586 The `do {...} while (0)' is a standard trick to allow FOO to have
587 statement semantics, so that it can safely be used within an `if'
588 statement in C, for example. Multiple evaluation is prevented by
589 copying a supplied argument into a local variable, so that
590 `FOO(var,fun(1))' only calls `fun' once.
592 Lisp lists are popular data structures in the C code as well as in
593 Elisp. There are two sets of macros that iterate over lists.
594 `EXTERNAL_LIST_LOOP_N' should be used when the list has been supplied
595 by the user, and cannot be trusted to be acyclic and `nil'-terminated.
596 A `malformed-list' or `circular-list' error will be generated if the
597 list being iterated over is not entirely kosher. `LIST_LOOP_N', on the
598 other hand, is faster and less safe, and can be used only on trusted
601 Related macros are `GET_EXTERNAL_LIST_LENGTH' and `GET_LIST_LENGTH',
602 which calculate the length of a list, and in the case of
603 `GET_EXTERNAL_LIST_LENGTH', validating the properness of the list. The
604 macros `EXTERNAL_LIST_LOOP_DELETE_IF' and `LIST_LOOP_DELETE_IF' delete
605 elements from a lisp list satisfying some predicate.
608 File: internals.info, Node: Writing Lisp Primitives, Next: Adding Global Lisp Variables, Prev: General Coding Rules, Up: Rules When Writing New C Code
610 Writing Lisp Primitives
611 =======================
613 Lisp primitives are Lisp functions implemented in C. The details of
614 interfacing the C function so that Lisp can call it are handled by a few
615 C macros. The only way to really understand how to write new C code is
616 to read the source, but we can explain some things here.
618 An example of a special form is the definition of `prog1', from
619 `eval.c'. (An ordinary function would have the same general
622 DEFUN ("prog1", Fprog1, 1, UNEVALLED, 0, /*
623 Similar to `progn', but the value of the first form is returned.
624 \(prog1 FIRST BODY...): All the arguments are evaluated sequentially.
625 The value of FIRST is saved during evaluation of the remaining args,
626 whose values are discarded.
630 /* This function can GC */
631 REGISTER Lisp_Object val, form, tail;
634 val = Feval (XCAR (args));
638 LIST_LOOP_3 (form, XCDR (args), tail)
645 Let's start with a precise explanation of the arguments to the
646 `DEFUN' macro. Here is a template for them:
648 DEFUN (LNAME, FNAME, MIN_ARGS, MAX_ARGS, INTERACTIVE, /*
654 This string is the name of the Lisp symbol to define as the
655 function name; in the example above, it is `"prog1"'.
658 This is the C function name for this function. This is the name
659 that is used in C code for calling the function. The name is, by
660 convention, `F' prepended to the Lisp name, with all dashes (`-')
661 in the Lisp name changed to underscores. Thus, to call this
662 function from C code, call `Fprog1'. Remember that the arguments
663 are of type `Lisp_Object'; various macros and functions for
664 creating values of type `Lisp_Object' are declared in the file
667 Primitives whose names are special characters (e.g. `+' or `<')
668 are named by spelling out, in some fashion, the special character:
669 e.g. `Fplus()' or `Flss()'. Primitives whose names begin with
670 normal alphanumeric characters but also contain special characters
671 are spelled out in some creative way, e.g. `let*' becomes
674 Each function also has an associated structure that holds the data
675 for the subr object that represents the function in Lisp. This
676 structure conveys the Lisp symbol name to the initialization
677 routine that will create the symbol and store the subr object as
678 its definition. The C variable name of this structure is always
679 `S' prepended to the FNAME. You hardly ever need to be aware of
680 the existence of this structure, since `DEFUN' plus `DEFSUBR'
681 takes care of all the details.
684 This is the minimum number of arguments that the function
685 requires. The function `prog1' allows a minimum of one argument.
688 This is the maximum number of arguments that the function accepts,
689 if there is a fixed maximum. Alternatively, it can be `UNEVALLED',
690 indicating a special form that receives unevaluated arguments, or
691 `MANY', indicating an unlimited number of evaluated arguments (the
692 C equivalent of `&rest'). Both `UNEVALLED' and `MANY' are macros.
693 If MAX_ARGS is a number, it may not be less than MIN_ARGS and it
694 may not be greater than 8. (If you need to add a function with
695 more than 8 arguments, use the `MANY' form. Resist the urge to
696 edit the definition of `DEFUN' in `lisp.h'. If you do it anyways,
697 make sure to also add another clause to the switch statement in
698 `primitive_funcall().')
701 This is an interactive specification, a string such as might be
702 used as the argument of `interactive' in a Lisp function. In the
703 case of `prog1', it is 0 (a null pointer), indicating that `prog1'
704 cannot be called interactively. A value of `""' indicates a
705 function that should receive no arguments when called
709 This is the documentation string. It is written just like a
710 documentation string for a function defined in Lisp; in
711 particular, the first line should be a single sentence. Note how
712 the documentation string is enclosed in a comment, none of the
713 documentation is placed on the same lines as the comment-start and
714 comment-end characters, and the comment-start characters are on
715 the same line as the interactive specification. `make-docfile',
716 which scans the C files for documentation strings, is very
717 particular about what it looks for, and will not properly extract
718 the doc string if it's not in this exact format.
720 In order to make both `etags' and `make-docfile' happy, make sure
721 that the `DEFUN' line contains the LNAME and FNAME, and that the
722 comment-start characters for the doc string are on the same line
723 as the interactive specification, and put a newline directly after
724 them (and before the comment-end characters).
727 This is the comma-separated list of arguments to the C function.
728 For a function with a fixed maximum number of arguments, provide a
729 C argument for each Lisp argument. In this case, unlike regular C
730 functions, the types of the arguments are not declared; they are
731 simply always of type `Lisp_Object'.
733 The names of the C arguments will be used as the names of the
734 arguments to the Lisp primitive as displayed in its documentation,
735 modulo the same concerns described above for `F...' names (in
736 particular, underscores in the C arguments become dashes in the
739 There is one additional kludge: A trailing `_' on the C argument is
740 discarded when forming the Lisp argument. This allows C language
741 reserved words (like `default') or global symbols (like `dirname')
742 to be used as argument names without compiler warnings or errors.
744 A Lisp function with MAX_ARGS = `UNEVALLED' is a "special form";
745 its arguments are not evaluated. Instead it receives one argument
746 of type `Lisp_Object', a (Lisp) list of the unevaluated arguments,
747 conventionally named `(args)'.
749 When a Lisp function has no upper limit on the number of arguments,
750 specify MAX_ARGS = `MANY'. In this case its implementation in C
751 actually receives exactly two arguments: the number of Lisp
752 arguments (an `int') and the address of a block containing their
753 values (a `Lisp_Object *'). In this case only are the C types
754 specified in the ARGLIST: `(int nargs, Lisp_Object *args)'.
756 Within the function `Fprog1' itself, note the use of the macros
757 `GCPRO1' and `UNGCPRO'. `GCPRO1' is used to "protect" a variable from
758 garbage collection--to inform the garbage collector that it must look
759 in that variable and regard the object pointed at by its contents as an
760 accessible object. This is necessary whenever you call `Feval' or
761 anything that can directly or indirectly call `Feval' (this includes
762 the `QUIT' macro!). At such a time, any Lisp object that you intend to
763 refer to again must be protected somehow. `UNGCPRO' cancels the
764 protection of the variables that are protected in the current function.
765 It is necessary to do this explicitly.
767 The macro `GCPRO1' protects just one local variable. If you want to
768 protect two, use `GCPRO2' instead; repeating `GCPRO1' will not work.
769 Macros `GCPRO3' and `GCPRO4' also exist.
771 These macros implicitly use local variables such as `gcpro1'; you
772 must declare these explicitly, with type `struct gcpro'. Thus, if you
773 use `GCPRO2', you must declare `gcpro1' and `gcpro2'.
775 Note also that the general rule is "caller-protects"; i.e. you are
776 only responsible for protecting those Lisp objects that you create. Any
777 objects passed to you as arguments should have been protected by whoever
778 created them, so you don't in general have to protect them.
780 In particular, the arguments to any Lisp primitive are always
781 automatically `GCPRO'ed, when called "normally" from Lisp code or
782 bytecode. So only a few Lisp primitives that are called frequently from
783 C code, such as `Fprogn' protect their arguments as a service to their
784 caller. You don't need to protect your arguments when writing a new
787 `GCPRO'ing is perhaps the trickiest and most error-prone part of
788 XEmacs coding. It is *extremely* important that you get this right and
789 use a great deal of discipline when writing this code. *Note
790 `GCPRO'ing: GCPROing, for full details on how to do this.
792 What `DEFUN' actually does is declare a global structure of type
793 `Lisp_Subr' whose name begins with capital `SF' and which contains
794 information about the primitive (e.g. a pointer to the function, its
795 minimum and maximum allowed arguments, a string describing its Lisp
796 name); `DEFUN' then begins a normal C function declaration using the
797 `F...' name. The Lisp subr object that is the function definition of a
798 primitive (i.e. the object in the function slot of the symbol that
799 names the primitive) actually points to this `SF' structure; when
800 `Feval' encounters a subr, it looks in the structure to find out how to
803 Defining the C function is not enough to make a Lisp primitive
804 available; you must also create the Lisp symbol for the primitive (the
805 symbol is "interned"; *note Obarrays::) and store a suitable subr
806 object in its function cell. (If you don't do this, the primitive won't
807 be seen by Lisp code.) The code looks like this:
811 Here FNAME is the same name you used as the second argument to `DEFUN'.
813 This call to `DEFSUBR' should go in the `syms_of_*()' function at
814 the end of the module. If no such function exists, create it and make
815 sure to also declare it in `symsinit.h' and call it from the
816 appropriate spot in `main()'. *Note General Coding Rules::.
818 Note that C code cannot call functions by name unless they are
819 defined in C. The way to call a function written in Lisp from C is to
820 use `Ffuncall', which embodies the Lisp function `funcall'. Since the
821 Lisp function `funcall' accepts an unlimited number of arguments, in C
822 it takes two: the number of Lisp-level arguments, and a one-dimensional
823 array containing their values. The first Lisp-level argument is the
824 Lisp function to call, and the rest are the arguments to pass to it.
825 Since `Ffuncall' can call the evaluator, you must protect pointers from
826 garbage collection around the call to `Ffuncall'. (However, `Ffuncall'
827 explicitly protects all of its parameters, so you don't have to protect
828 any pointers passed as parameters to it.)
830 The C functions `call0', `call1', `call2', and so on, provide handy
831 ways to call a Lisp function conveniently with a fixed number of
832 arguments. They work by calling `Ffuncall'.
834 `eval.c' is a very good file to look through for examples; `lisp.h'
835 contains the definitions for important macros and functions.
838 File: internals.info, Node: Adding Global Lisp Variables, Next: Coding for Mule, Prev: Writing Lisp Primitives, Up: Rules When Writing New C Code
840 Adding Global Lisp Variables
841 ============================
843 Global variables whose names begin with `Q' are constants whose
844 value is a symbol of a particular name. The name of the variable should
845 be derived from the name of the symbol using the same rules as for Lisp
846 primitives. These variables are initialized using a call to
847 `defsymbol()' in the `syms_of_*()' function. (This call interns a
848 symbol, sets the C variable to the resulting Lisp object, and calls
849 `staticpro()' on the C variable to tell the garbage-collection
850 mechanism about this variable. What `staticpro()' does is add a
851 pointer to the variable to a large global array; when
852 garbage-collection happens, all pointers listed in the array are used
853 as starting points for marking Lisp objects. This is important because
854 it's quite possible that the only current reference to the object is
855 the C variable. In the case of symbols, the `staticpro()' doesn't
856 matter all that much because the symbol is contained in `obarray',
857 which is itself `staticpro()'ed. However, it's possible that a naughty
858 user could do something like uninterning the symbol out of `obarray' or
859 even setting `obarray' to a different value [although this is likely to
860 make XEmacs crash!].)
862 *Please note:* It is potentially deadly if you declare a `Q...'
863 variable in two different modules. The two calls to `defsymbol()' are
864 no problem, but some linkers will complain about multiply-defined
865 symbols. The most insidious aspect of this is that often the link will
866 succeed anyway, but then the resulting executable will sometimes crash
867 in obscure ways during certain operations! To avoid this problem,
868 declare any symbols with common names (such as `text') that are not
869 obviously associated with this particular module in the module
872 Global variables whose names begin with `V' are variables that
873 contain Lisp objects. The convention here is that all global variables
874 of type `Lisp_Object' begin with `V', and all others don't (including
875 integer and boolean variables that have Lisp equivalents). Most of the
876 time, these variables have equivalents in Lisp, but some don't. Those
877 that do are declared this way by a call to `DEFVAR_LISP()' in the
878 `vars_of_*()' initializer for the module. What this does is create a
879 special "symbol-value-forward" Lisp object that contains a pointer to
880 the C variable, intern a symbol whose name is as specified in the call
881 to `DEFVAR_LISP()', and set its value to the symbol-value-forward Lisp
882 object; it also calls `staticpro()' on the C variable to tell the
883 garbage-collection mechanism about the variable. When `eval' (or
884 actually `symbol-value') encounters this special object in the process
885 of retrieving a variable's value, it follows the indirection to the C
886 variable and gets its value. `setq' does similar things so that the C
887 variable gets changed.
889 Whether or not you `DEFVAR_LISP()' a variable, you need to
890 initialize it in the `vars_of_*()' function; otherwise it will end up
891 as all zeroes, which is the integer 0 (_not_ `nil'), and this is
892 probably not what you want. Also, if the variable is not
893 `DEFVAR_LISP()'ed, *you must call* `staticpro()' on the C variable in
894 the `vars_of_*()' function. Otherwise, the garbage-collection
895 mechanism won't know that the object in this variable is in use, and
896 will happily collect it and reuse its storage for another Lisp object,
897 and you will be the one who's unhappy when you can't figure out how
898 your variable got overwritten.
901 File: internals.info, Node: Coding for Mule, Next: Techniques for XEmacs Developers, Prev: Adding Global Lisp Variables, Up: Rules When Writing New C Code
906 Although Mule support is not compiled by default in XEmacs, many
907 people are using it, and we consider it crucial that new code works
908 correctly with multibyte characters. This is not hard; it is only a
909 matter of following several simple user-interface guidelines. Even if
910 you never compile with Mule, with a little practice you will find it
911 quite easy to code Mule-correctly.
913 Note that these guidelines are not necessarily tied to the current
914 Mule implementation; they are also a good idea to follow on the grounds
915 of code generalization for future I18N work.
919 * Character-Related Data Types::
920 * Working With Character and Byte Positions::
921 * Conversion to and from External Data::
922 * General Guidelines for Writing Mule-Aware Code::
923 * An Example of Mule-Aware Code::
926 File: internals.info, Node: Character-Related Data Types, Next: Working With Character and Byte Positions, Prev: Coding for Mule, Up: Coding for Mule
928 Character-Related Data Types
929 ----------------------------
931 First, let's review the basic character-related datatypes used by
932 XEmacs. Note that the separate `typedef's are not mandatory in the
933 current implementation (all of them boil down to `unsigned char' or
934 `int'), but they improve clarity of code a great deal, because one
935 glance at the declaration can tell the intended use of the variable.
938 An `Emchar' holds a single Emacs character.
940 Obviously, the equality between characters and bytes is lost in
941 the Mule world. Characters can be represented by one or more
942 bytes in the buffer, and `Emchar' is the C type large enough to
945 Without Mule support, an `Emchar' is equivalent to an `unsigned
949 The data representing the text in a buffer or string is logically
952 XEmacs does not work with the same character formats all the time;
953 when reading characters from the outside, it decodes them to an
954 internal format, and likewise encodes them when writing.
955 `Bufbyte' (in fact `unsigned char') is the basic unit of XEmacs
956 internal buffers and strings format. A `Bufbyte *' is the type
957 that points at text encoded in the variable-width internal
960 One character can correspond to one or more `Bufbyte's. In the
961 current Mule implementation, an ASCII character is represented by
962 the same `Bufbyte', and other characters are represented by a
963 sequence of two or more `Bufbyte's.
965 Without Mule support, there are exactly 256 characters, implicitly
966 Latin-1, and each character is represented using one `Bufbyte', and
967 there is a one-to-one correspondence between `Bufbyte's and
972 A `Bufpos' represents a character position in a buffer or string.
973 A `Charcount' represents a number (count) of characters.
974 Logically, subtracting two `Bufpos' values yields a `Charcount'
975 value. Although all of these are `typedef'ed to `EMACS_INT', we
976 use them in preference to `EMACS_INT' to make it clear what sort
977 of position is being used.
979 `Bufpos' and `Charcount' values are the only ones that are ever
984 A `Bytind' represents a byte position in a buffer or string. A
985 `Bytecount' represents the distance between two positions, in
986 bytes. The relationship between `Bytind' and `Bytecount' is the
987 same as the relationship between `Bufpos' and `Charcount'.
991 When dealing with the outside world, XEmacs works with `Extbyte's,
992 which are equivalent to `unsigned char'. Obviously, an `Extcount'
993 is the distance between two `Extbyte's. Extbytes and Extcounts
994 are not all that frequent in XEmacs code.
997 File: internals.info, Node: Working With Character and Byte Positions, Next: Conversion to and from External Data, Prev: Character-Related Data Types, Up: Coding for Mule
999 Working With Character and Byte Positions
1000 -----------------------------------------
1002 Now that we have defined the basic character-related types, we can
1003 look at the macros and functions designed for work with them and for
1004 conversion between them. Most of these macros are defined in
1005 `buffer.h', and we don't discuss all of them here, but only the most
1006 important ones. Examining the existing code is the best way to learn
1010 This preprocessor constant is the maximum number of buffer bytes to
1011 represent an Emacs character in the variable width internal
1012 encoding. It is useful when allocating temporary strings to keep
1013 a known number of characters. For instance:
1019 /* Allocate place for CCLEN characters. */
1020 Bufbyte *buf = (Bufbyte *)alloca (cclen * MAX_EMCHAR_LEN);
1023 If you followed the previous section, you can guess that,
1024 logically, multiplying a `Charcount' value with `MAX_EMCHAR_LEN'
1025 produces a `Bytecount' value.
1027 In the current Mule implementation, `MAX_EMCHAR_LEN' equals 4.
1028 Without Mule, it is 1.
1031 `set_charptr_emchar'
1032 The `charptr_emchar' macro takes a `Bufbyte' pointer and returns
1033 the `Emchar' stored at that position. If it were a function, its
1036 Emchar charptr_emchar (Bufbyte *p);
1038 `set_charptr_emchar' stores an `Emchar' to the specified byte
1039 position. It returns the number of bytes stored:
1041 Bytecount set_charptr_emchar (Bufbyte *p, Emchar c);
1043 It is important to note that `set_charptr_emchar' is safe only for
1044 appending a character at the end of a buffer, not for overwriting a
1045 character in the middle. This is because the width of characters
1046 varies, and `set_charptr_emchar' cannot resize the string if it
1047 writes, say, a two-byte character where a single-byte character
1050 A typical use of `set_charptr_emchar' can be demonstrated by this
1051 example, which copies characters from buffer BUF to a temporary
1056 for (pos = beg; pos < end; pos++)
1058 Emchar c = BUF_FETCH_CHAR (buf, pos);
1059 p += set_charptr_emchar (buf, c);
1063 Note how `set_charptr_emchar' is used to store the `Emchar' and
1064 increment the counter, at the same time.
1068 These two macros increment and decrement a `Bufbyte' pointer,
1069 respectively. They will adjust the pointer by the appropriate
1070 number of bytes according to the byte length of the character
1071 stored there. Both macros assume that the memory address is
1072 located at the beginning of a valid character.
1074 Without Mule support, `INC_CHARPTR (p)' and `DEC_CHARPTR (p)'
1075 simply expand to `p++' and `p--', respectively.
1077 `bytecount_to_charcount'
1078 Given a pointer to a text string and a length in bytes, return the
1079 equivalent length in characters.
1081 Charcount bytecount_to_charcount (Bufbyte *p, Bytecount bc);
1083 `charcount_to_bytecount'
1084 Given a pointer to a text string and a length in characters,
1085 return the equivalent length in bytes.
1087 Bytecount charcount_to_bytecount (Bufbyte *p, Charcount cc);
1090 Return a pointer to the beginning of the character offset CC (in
1093 Bufbyte *charptr_n_addr (Bufbyte *p, Charcount cc);