* Some more work to get rid of SCM_LENGTH
[bpt/guile.git] / libguile / tags.h
1 /* classes: h_files */
2
3 #ifndef TAGSH
4 #define TAGSH
5 /* Copyright (C) 1995, 96, 97, 98, 99, 2000 Free Software Foundation, Inc.
6 *
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2, or (at your option)
10 * any later version.
11 *
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with this software; see the file COPYING. If not, write to
19 * the Free Software Foundation, Inc., 59 Temple Place, Suite 330,
20 * Boston, MA 02111-1307 USA
21 *
22 * As a special exception, the Free Software Foundation gives permission
23 * for additional uses of the text contained in its release of GUILE.
24 *
25 * The exception is that, if you link the GUILE library with other files
26 * to produce an executable, this does not by itself cause the
27 * resulting executable to be covered by the GNU General Public License.
28 * Your use of that executable is in no way restricted on account of
29 * linking the GUILE library code into it.
30 *
31 * This exception does not however invalidate any other reasons why
32 * the executable file might be covered by the GNU General Public License.
33 *
34 * This exception applies only to the code released by the
35 * Free Software Foundation under the name GUILE. If you copy
36 * code from other Free Software Foundation releases into a copy of
37 * GUILE, as the General Public License permits, the exception does
38 * not apply to the code that you add in this way. To avoid misleading
39 * anyone as to the status of such modified files, you must delete
40 * this exception notice from them.
41 *
42 * If you write modifications of your own for GUILE, it is your choice
43 * whether to permit this exception to apply to your modifications.
44 * If you do not wish that, delete this exception notice. */
45
46 /* Software engineering face-lift by Greg J. Badros, 11-Dec-1999,
47 gjb@cs.washington.edu, http://www.cs.washington.edu/homes/gjb */
48
49 \f
50
51 /** This file defines the format of SCM values and cons pairs.
52 ** It is here that tag bits are assigned for various purposes.
53 **/
54
55 \f
56
57 /* #define SCM_VOIDP_TEST */
58
59 /* In the beginning was the Word:
60 */
61 typedef long scm_bits_t;
62
63 /* But as external interface, we use SCM, which may, according to the desired
64 * level of type checking, be defined in several ways:
65 */
66 #if (SCM_DEBUG_TYPING_STRICTNESS == 1)
67 typedef union { struct { scm_bits_t n; } n; } SCM;
68 static SCM scm_pack(scm_bits_t b) { SCM s; s.n.n = b; return s; }
69 # define SCM_UNPACK(x) ((x).n.n)
70 # define SCM_PACK(x) (scm_pack ((scm_bits_t) (x)))
71 #elif defined (SCM_VOIDP_TEST)
72 /* This is the default, which provides an intermediate level of compile time
73 * type checking while still resulting in very efficient code.
74 */
75 typedef void * SCM;
76 # define SCM_UNPACK(x) ((scm_bits_t) (x))
77 # define SCM_PACK(x) ((SCM) (x))
78 #else
79 /* This should be used as a fall back solution for machines on which casting
80 * to a pointer may lead to loss of bit information, e. g. in the three least
81 * significant bits.
82 */
83 typedef scm_bits_t SCM;
84 # define SCM_UNPACK(x) (x)
85 # define SCM_PACK(x) ((scm_bits_t) (x))
86 #endif
87
88
89 /* SCM values can not be compared by using the operator ==. Use the following
90 * macro instead, which is the equivalent of the scheme predicate 'eq?'.
91 */
92 #define SCM_EQ_P(x, y) (SCM_UNPACK (x) == SCM_UNPACK (y))
93
94 \f
95
96 /* SCM variables can contain:
97 *
98 * Non-objects -- meaning that the tag-related macros don't apply to them
99 * in the usual way.
100 *
101 * Immediates -- meaning that the variable contains an entire Scheme object.
102 *
103 * Non-immediates -- meaning that the variable holds a (possibly
104 * tagged) pointer into the cons pair heap.
105 *
106 * Non-objects are distinguished from other values by careful coding
107 * only (i.e., programmers must keep track of any SCM variables they
108 * create that don't contain ordinary scheme values).
109 *
110 * All immediates and non-immediates must have a 0 in bit 0. Only
111 * non-object values can have a 1 in bit 0. In some cases, bit 0 of a
112 * word in the heap is used for the GC tag so during garbage
113 * collection, that bit might be 1 even in an immediate or
114 * non-immediate value. In other cases, bit 0 of a word in the heap
115 * is used to tag a pointer to a GLOC (VM global variable address) or
116 * the header of a struct. But whenever an SCM variable holds a
117 * normal Scheme value, bit 0 is 0.
118 *
119 * Immediates and non-immediates are distinguished by bits two and four.
120 * Immediate values must have a 1 in at least one of those bits. Does
121 * this (or any other detail of tagging) seem arbitrary? Try changing it!
122 * (Not always impossible but it is fair to say that many details of tags
123 * are mutually dependent). */
124
125 #define SCM_IMP(x) (6 & SCM_UNPACK (x))
126 #define SCM_NIMP(x) (!SCM_IMP (x))
127
128 /* Here is a summary of tagging in SCM values as they might occur in
129 * SCM variables or in the heap.
130 *
131 * low bits meaning
132 *
133 *
134 * 0 Most objects except...
135 * 1 ...glocs and structs (this tag valid only in a SCM_CAR or
136 * in the header of a struct's data).
137 *
138 * 00 heap addresses and many immediates (not integers)
139 * 01 glocs/structs, some tc7_ codes
140 * 10 immediate integers
141 * 11 various tc7_ codes including, tc16_ codes.
142 *
143 *
144 * 000 heap address
145 * 001 glocs/structs
146 * 010 integer
147 * 011 closure
148 * 100 immediates
149 * 101 tc7_
150 * 110 integer
151 * 111 tc7_
152 *
153 *
154 * 100 --- IMMEDIATES
155 *
156 * Looking at the seven final bits of an immediate:
157 *
158 * 0000-100 short instruction
159 * 0001-100 short instruction
160 * 0010-100 short instruction
161 * 0011-100 short instruction
162 * 0100-100 short instruction
163 * 0101-100 short instruction
164 * 0110-100 various immediates and long instructions
165 * 0111-100 short instruction
166 * 1000-100 short instruction
167 * 1001-100 short instruction
168 * 1010-100 short instruction
169 * 1011-100 short instruction
170 * 1100-100 short instruction
171 * 1101-100 short instruction
172 * 1110-100 immediate characters
173 * 1111-100 ilocs
174 *
175 * Some of the 0110100 immediates are long instructions (they dispatch
176 * in two steps compared to one step for a short instruction).
177 * The two steps are, (1) dispatch on 7 bits to the long instruction
178 * handler, (2) dispatch on 7 additional bits.
179 *
180 * One way to think of it is that there are 128 short instructions,
181 * with the 13 immediates above being some of the most interesting.
182 *
183 * Also noteworthy are the groups of 16 7-bit instructions implied by
184 * some of the 3-bit tags. For example, closure references consist
185 * of an 8-bit aligned address tagged with 011. There are 16 identical 7-bit
186 * instructions, all ending 011, which are invoked by evaluating closures.
187 *
188 * In other words, if you hand the evaluator a closure, the evaluator
189 * treats the closure as a graph of virtual machine instructions.
190 * A closure is a pair with a pointer to the body of the procedure
191 * in the CDR and a pointer to the environment of the closure in the CAR.
192 * The environment pointer is tagged 011 which implies that the least
193 * significant 7 bits of the environment pointer also happen to be
194 * a virtual machine instruction we could call "SELF" (for self-evaluating
195 * object).
196 *
197 * A less trivial example are the 16 instructions ending 000. If those
198 * bits tag the CAR of a pair, then evidently the pair is an ordinary
199 * cons pair and should be evaluated as a procedure application. The sixteen,
200 * 7-bit 000 instructions are all "NORMAL-APPLY" (Things get trickier.
201 * For example, if the CAR of a procedure application is a symbol, the NORMAL-APPLY
202 * instruction will, as a side effect, overwrite that CAR with a new instruction
203 * that contains a cached address for the variable named by the symbol.)
204 *
205 * Here is a summary of tags in the CAR of a non-immediate:
206 *
207 * HEAP CELL: G=gc_mark; 1 during mark, 0 other times.
208 *
209 * cons ..........SCM car..............0 ...........SCM cdr.............G
210 * gloc ..........SCM vcell..........001 ...........SCM cdr.............G
211 * struct ..........void * type........001 ...........void * data.........G
212 * closure ..........SCM code...........011 ...........SCM env.............G
213 * tc7 ......24.bits of data...Gxxxx1S1 ..........void *data............
214 *
215 *
216 *
217 * 101 & 111 --- tc7_ types
218 *
219 * tc7_tags are 7 bit tags ending in 1x1. These tags
220 * occur only in the CAR of heap cells, and have the
221 * handy property that all bits of the CAR above the
222 * bottom eight can be used to store some data, thus
223 * saving a word in the body itself. Thus, we use them
224 * for strings and vectors (among other things).
225 *
226 * TYP7(X) returns bits 0...6 of CELL_TYPE (X)
227 *
228 * Sometimes we choose the bottom seven bits carefully,
229 * so that the 2-valued bit (called S bit) can be masked
230 * off to reveal a common type.
231 *
232 * TYP7S(X) returns TYP7, but masking out the option bit S.
233 *
234 * For example, all strings have 0010 in the 'xxxx' bits
235 * in the diagram above, the S bit says whether it's a
236 * substring.
237 *
238 * for example:
239 * S
240 * scm_tc7_string = G0010101
241 * scm_tc7_substring = G0010111
242 *
243 * TYP7S turns both string tags into tc7_string; thus,
244 * testing TYP7S against tc7_string is a quick way to
245 * test for any kind of string, shared or unshared.
246 *
247 * Some TC7 types are subdivided into 256 subtypes giving
248 * rise to the macros:
249 *
250 * TYP16
251 * TYP16S
252 * GCTYP16
253 *
254 * TYP16S functions similarly wrt to TYP16 as TYP7S to TYP7,
255 * but a different option bit is used (bit 2 for TYP7S,
256 * bit 8 for TYP16S).
257 * */
258
259
260
261 \f
262 /* {Non-immediate values.}
263 *
264 * If X is non-immediate, it is necessary to look at SCM_CAR (X) to
265 * figure out Xs type. X may be a cons pair, in which case the value
266 * SCM_CAR (x) will be either an immediate or non-immediate value. X
267 * may be something other than a cons pair, in which case the value
268 * SCM_CAR (x) will be a non-object value.
269 *
270 * All immediates and non-immediates have a 0 in bit 0. We
271 * additionally preserve the invariant that all non-object values
272 * stored in the SCM_CAR of a non-immediate object have a 1 in bit 1:
273 */
274
275 #define SCM_SLOPPY_CONSP(x) ((1 & SCM_CELL_TYPE (x)) == 0)
276 #define SCM_SLOPPY_NCONSP(x) (!SCM_SLOPPY_CONSP(x))
277
278 #define SCM_CONSP(x) (!SCM_IMP (x) && SCM_SLOPPY_CONSP (x))
279 #define SCM_NCONSP(x) (!SCM_CONSP (x))
280
281
282 /* SCM_ECONSP should be used instead of SCM_CONSP at places where GLOCS
283 * can be expected to occur.
284 */
285 #define SCM_ECONSP(x) \
286 (!SCM_IMP (x) \
287 && (SCM_SLOPPY_CONSP (x) \
288 || (SCM_TYP3 (x) == 1 \
289 && (SCM_STRUCT_VTABLE_DATA (x)[scm_vtable_index_vcell] != 0))))
290 #define SCM_NECONSP(x) (!SCM_ECONSP (x))
291
292 \f
293
294 #define SCM_CELLP(x) (((sizeof (scm_cell) - 1) & SCM_UNPACK (x)) == 0)
295 #define SCM_NCELLP(x) (!SCM_CELLP (x))
296
297 #define SCM_DOUBLE_CELLP(x) (((2 * sizeof (scm_cell) - 1) & SCM_UNPACK (x)) == 0)
298
299 /* See numbers.h for macros relating to immediate integers.
300 */
301
302 #define SCM_ITAG3(x) (7 & SCM_UNPACK (x))
303 #define SCM_TYP3(x) (7 & SCM_CELL_TYPE (x))
304 #define scm_tc3_cons 0
305 #define scm_tc3_cons_gloc 1
306 #define scm_tc3_int_1 2
307 #define scm_tc3_closure 3
308 #define scm_tc3_imm24 4
309 #define scm_tc3_tc7_1 5
310 #define scm_tc3_int_2 6
311 #define scm_tc3_tc7_2 7
312
313
314 /*
315 * Do not change the three bit tags.
316 */
317
318
319 #define SCM_ITAG7(x) (127 & SCM_UNPACK (x))
320 #define SCM_TYP7(x) (0x7f & SCM_CELL_TYPE (x))
321 #define SCM_TYP7S(x) ((0x7f & ~2) & SCM_CELL_TYPE (x))
322
323
324 #define SCM_TYP16(x) (0xffff & SCM_CELL_TYPE (x))
325 #define SCM_TYP16S(x) (0xfeff & SCM_CELL_TYPE (x))
326
327 \f
328
329 #define scm_tc7_symbol 5
330 /* free 7 */
331
332 /* couple */
333 #define scm_tc7_vector 13
334 #define scm_tc7_wvect 15
335
336 /* couple */
337 #define scm_tc7_string 21
338 #define scm_tc7_substring 23
339
340 /* Many of the following should be turned
341 * into structs or smobs. We need back some
342 * of these 7 bit tags!
343 */
344 #define scm_tc7_pws 31
345
346 #ifdef HAVE_ARRAYS
347 #define scm_tc7_llvect 29
348 #define scm_tc7_uvect 37
349 /* free 39 */
350 #define scm_tc7_fvect 45
351 #define scm_tc7_dvect 47
352 #define scm_tc7_cvect 53
353 #define scm_tc7_svect 55
354 #define scm_tc7_bvect 71
355 #define scm_tc7_byvect 77
356 #define scm_tc7_ivect 79
357 #endif
358
359 #define scm_tc7_contin 61
360 #define scm_tc7_cclo 63
361 #define scm_tc7_rpsubr 69
362 #define scm_tc7_subr_0 85
363 #define scm_tc7_subr_1 87
364 #define scm_tc7_cxr 93
365 #define scm_tc7_subr_3 95
366 #define scm_tc7_subr_2 101
367 #define scm_tc7_asubr 103
368 #define scm_tc7_subr_1o 109
369 #define scm_tc7_subr_2o 111
370 #define scm_tc7_lsubr_2 117
371 #define scm_tc7_lsubr 119
372
373
374 /* There are 256 port subtypes. Here are the first few.
375 * These must agree with the init function in ports.c
376 */
377 #define scm_tc7_port 125
378
379 #define scm_tc16_fport (scm_tc7_port + 0 * 256L)
380 /* scm_tc16_pipe was here. */
381 #define scm_tc16_strport (scm_tc7_port + 2 * 256L)
382 #define scm_tc16_sfport (scm_tc7_port + 3 * 256L)
383
384
385 /* There are 256 smob subtypes. Here are the first four.
386 */
387
388 #define scm_tc7_smob 127 /* DO NOT CHANGE [**] */
389
390 /* [**] If you change scm_tc7_smob, you must also change
391 * the places it is hard coded in this file and possibly others.
392 */
393
394
395 /* scm_tc_free_cell is also the 0th smob type. We place this
396 * in free cells to tell the conservative marker not to trace it.
397 */
398 #define scm_tc_free_cell 0x007f
399
400 /* Smob type 1 (note the dependency on the predicate SCM_NUMP)
401 */
402 #define scm_tc16_big 0x017f
403
404 /* Smob types 2 and 3:
405 */
406 #define scm_tc16_real 0x027f
407 #define scm_tc16_complex 0x037f
408
409 \f
410 /* {Immediate Values}
411 */
412
413 enum scm_tags
414 {
415 scm_tc8_char = 0xf4,
416 scm_tc8_iloc = 0xfc
417 };
418
419 #define SCM_ITAG8(X) (SCM_UNPACK (X) & 0xff)
420 #define SCM_MAKE_ITAG8(X, TAG) SCM_PACK (((X) << 8) + TAG)
421 #define SCM_ITAG8_DATA(X) (SCM_UNPACK (X) >> 8)
422
423
424 \f
425 /* Immediate Symbols, Special Symbols, Flags (various constants).
426 */
427
428 /* SCM_ISYMP tests for ISPCSYM and ISYM */
429 #define SCM_ISYMP(n) ((0x187 & SCM_UNPACK (n)) == 4)
430
431 /* SCM_IFLAGP tests for ISPCSYM, ISYM and IFLAG */
432 #define SCM_IFLAGP(n) ((0x87 & SCM_UNPACK (n)) == 4)
433 #define SCM_ISYMNUM(n) (SCM_UNPACK (n) >> 9)
434 #define SCM_ISYMCHARS(n) (scm_isymnames[SCM_ISYMNUM (n)])
435 #define SCM_MAKSPCSYM(n) SCM_PACK (((n) << 9) + ((n) << 3) + 4L)
436 #define SCM_MAKISYM(n) SCM_PACK (((n) << 9) + 0x74L)
437 #define SCM_MAKIFLAG(n) SCM_PACK (((n) << 9) + 0x174L)
438
439 extern char *scm_isymnames[]; /* defined in print.c */
440
441 /* This table must agree with the declarations
442 * in repl.c: {Names of immediate symbols}.
443 *
444 * These are used only in eval but their values
445 * have to be allocated here.
446 *
447 */
448
449 #define SCM_IM_AND SCM_MAKSPCSYM (0)
450 #define SCM_IM_BEGIN SCM_MAKSPCSYM (1)
451 #define SCM_IM_CASE SCM_MAKSPCSYM (2)
452 #define SCM_IM_COND SCM_MAKSPCSYM (3)
453 #define SCM_IM_DO SCM_MAKSPCSYM (4)
454 #define SCM_IM_IF SCM_MAKSPCSYM (5)
455 #define SCM_IM_LAMBDA SCM_MAKSPCSYM (6)
456 #define SCM_IM_LET SCM_MAKSPCSYM (7)
457 #define SCM_IM_LETSTAR SCM_MAKSPCSYM (8)
458 #define SCM_IM_LETREC SCM_MAKSPCSYM (9)
459 #define SCM_IM_OR SCM_MAKSPCSYM (10)
460 #define SCM_IM_QUOTE SCM_MAKSPCSYM (11)
461 #define SCM_IM_SET_X SCM_MAKSPCSYM (12)
462 #define SCM_IM_DEFINE SCM_MAKSPCSYM (13)
463 #define SCM_IM_APPLY SCM_MAKISYM (14)
464 #define SCM_IM_CONT SCM_MAKISYM (15)
465 #define SCM_BOOL_F SCM_MAKIFLAG (16)
466 #define SCM_BOOL_T SCM_MAKIFLAG (17)
467 #define SCM_UNDEFINED SCM_MAKIFLAG (18)
468 #define SCM_EOF_VAL SCM_MAKIFLAG (19)
469 #define SCM_EOL SCM_MAKIFLAG (20)
470 #define SCM_UNSPECIFIED SCM_MAKIFLAG (21)
471 #define SCM_IM_DISPATCH SCM_MAKISYM (22)
472 #define SCM_IM_SLOT_REF SCM_MAKISYM (23)
473 #define SCM_IM_SLOT_SET_X SCM_MAKISYM (24)
474
475 /* Multi-language support */
476
477 #define SCM_IM_NIL_COND SCM_MAKISYM (25)
478 #define SCM_IM_NIL_IFY SCM_MAKISYM (26)
479 #define SCM_IM_T_IFY SCM_MAKISYM (27)
480 #define SCM_IM_0_COND SCM_MAKISYM (28)
481 #define SCM_IM_0_IFY SCM_MAKISYM (29)
482 #define SCM_IM_1_IFY SCM_MAKISYM (30)
483 #define SCM_IM_BIND SCM_MAKISYM (31)
484
485 #define SCM_IM_DELAY SCM_MAKISYM (32)
486
487 /* When a variable is unbound this is marked by the SCM_UNDEFINED
488 * value. The following is an unbound value which can be handled on
489 * the Scheme level, i.e., it can be stored in and retrieved from a
490 * Scheme variable. This value is only intended to mark an unbound
491 * slot in GOOPS. It is needed now, but we should probably rewrite
492 * the code which handles this value in C so that SCM_UNDEFINED can be
493 * used instead. It is not ideal to let this kind of unique and
494 * strange values loose on the Scheme level.
495 */
496 #define SCM_UNBOUND SCM_MAKIFLAG (33)
497
498 #define SCM_UNBNDP(x) (SCM_EQ_P ((x), SCM_UNDEFINED))
499
500 \f
501
502 /* Dispatching aids: */
503
504
505 /* For cons pairs with immediate values in the CAR
506 */
507
508 #define scm_tcs_cons_imcar 2:case 4:case 6:case 10:\
509 case 12:case 14:case 18:case 20:\
510 case 22:case 26:case 28:case 30:\
511 case 34:case 36:case 38:case 42:\
512 case 44:case 46:case 50:case 52:\
513 case 54:case 58:case 60:case 62:\
514 case 66:case 68:case 70:case 74:\
515 case 76:case 78:case 82:case 84:\
516 case 86:case 90:case 92:case 94:\
517 case 98:case 100:case 102:case 106:\
518 case 108:case 110:case 114:case 116:\
519 case 118:case 122:case 124:case 126
520
521 /* For cons pairs with non-immediate values in the SCM_CAR
522 */
523 #define scm_tcs_cons_nimcar 0:case 8:case 16:case 24:\
524 case 32:case 40:case 48:case 56:\
525 case 64:case 72:case 80:case 88:\
526 case 96:case 104:case 112:case 120
527
528 /* A CONS_GLOC occurs in code. It's CAR is a pointer to the
529 * CDR of a variable. The low order bits of the CAR are 001.
530 * The CDR of the gloc is the code continuation.
531 */
532 #define scm_tcs_cons_gloc 1:case 9:case 17:case 25:\
533 case 33:case 41:case 49:case 57:\
534 case 65:case 73:case 81:case 89:\
535 case 97:case 105:case 113:case 121
536
537 #define scm_tcs_closures 3:case 11:case 19:case 27:\
538 case 35:case 43:case 51:case 59:\
539 case 67:case 75:case 83:case 91:\
540 case 99:case 107:case 115:case 123
541
542 #define scm_tcs_subrs scm_tc7_asubr:case scm_tc7_subr_0:case scm_tc7_subr_1:case scm_tc7_cxr:\
543 case scm_tc7_subr_3:case scm_tc7_subr_2:case scm_tc7_rpsubr:case scm_tc7_subr_1o:\
544 case scm_tc7_subr_2o:case scm_tc7_lsubr_2:case scm_tc7_lsubr
545
546 \f
547
548 #if (SCM_DEBUG_DEPRECATED == 0)
549
550 #define scm_tc7_ssymbol scm_tc7_symbol
551 #define scm_tc7_msymbol scm_tc7_symbol
552 #define scm_tcs_symbols scm_tc7_symbol
553
554 #define scm_tc16_flo scm_tc16_real
555 #define scm_tc_flo 0x017fL
556 #define scm_tc_dblr scm_tc16_real
557 #define scm_tc_dblc scm_tc16_complex
558
559 #endif /* SCM_DEBUG_DEPRECATED == 0 */
560
561 #endif /* TAGSH */
562
563 /*
564 Local Variables:
565 c-file-style: "gnu"
566 End:
567 */