* eval.h (scm_t_trampoline_1, scm_t_trampoline_2): New types.
[bpt/guile.git] / libguile / sort.c
1 /* Copyright (C) 1999,2000,2001,2002 Free Software Foundation, Inc.
2 * This program is free software; you can redistribute it and/or modify
3 * it under the terms of the GNU General Public License as published by
4 * the Free Software Foundation; either version 2, or (at your option)
5 * any later version.
6 *
7 * This program is distributed in the hope that it will be useful,
8 * but WITHOUT ANY WARRANTY; without even the implied warranty of
9 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
10 * GNU General Public License for more details.
11 *
12 * You should have received a copy of the GNU General Public License
13 * along with this software; see the file COPYING. If not, write to
14 * the Free Software Foundation, Inc., 59 Temple Place, Suite 330,
15 * Boston, MA 02111-1307 USA
16 *
17 * As a special exception, the Free Software Foundation gives permission
18 * for additional uses of the text contained in its release of GUILE.
19 *
20 * The exception is that, if you link the GUILE library with other files
21 * to produce an executable, this does not by itself cause the
22 * resulting executable to be covered by the GNU General Public License.
23 * Your use of that executable is in no way restricted on account of
24 * linking the GUILE library code into it.
25 *
26 * This exception does not however invalidate any other reasons why
27 * the executable file might be covered by the GNU General Public License.
28 *
29 * This exception applies only to the code released by the
30 * Free Software Foundation under the name GUILE. If you copy
31 * code from other Free Software Foundation releases into a copy of
32 * GUILE, as the General Public License permits, the exception does
33 * not apply to the code that you add in this way. To avoid misleading
34 * anyone as to the status of such modified files, you must delete
35 * this exception notice from them.
36 *
37 * If you write modifications of your own for GUILE, it is your choice
38 * whether to permit this exception to apply to your modifications.
39 * If you do not wish that, delete this exception notice. */
40
41
42
43 /* Written in December 1998 by Roland Orre <orre@nada.kth.se>
44 * This implements the same sort interface as slib/sort.scm
45 * for lists and vectors where slib defines:
46 * sorted?, merge, merge!, sort, sort!
47 * For scsh compatibility sort-list and sort-list! are also defined.
48 * In cases where a stable-sort is required use stable-sort or
49 * stable-sort!. An additional feature is
50 * (restricted-vector-sort! vector less? startpos endpos)
51 * which allows you to sort part of a vector.
52 * Thanks to Aubrey Jaffer for the slib/sort.scm library.
53 * Thanks to Richard A. O'Keefe (based on Prolog code by D.H.D.Warren)
54 * for the merge sort inspiration.
55 * Thanks to Douglas C. Schmidt (schmidt@ics.uci.edu) for the
56 * quicksort code.
57 */
58
59 /* We need this to get the definitions for HAVE_ALLOCA_H, etc. */
60 #include "libguile/scmconfig.h"
61
62 /* AIX requires this to be the first thing in the file. The #pragma
63 directive is indented so pre-ANSI compilers will ignore it, rather
64 than choke on it. */
65 #ifndef __GNUC__
66 # if HAVE_ALLOCA_H
67 # include <alloca.h>
68 # else
69 # ifdef _AIX
70 #pragma alloca
71 # else
72 # ifndef alloca /* predefined by HP cc +Olibcalls */
73 char *alloca ();
74 # endif
75 # endif
76 # endif
77 #endif
78
79 #include <string.h>
80 #include "libguile/_scm.h"
81
82 #include "libguile/eval.h"
83 #include "libguile/unif.h"
84 #include "libguile/ramap.h"
85 #include "libguile/alist.h"
86 #include "libguile/feature.h"
87 #include "libguile/root.h"
88 #include "libguile/vectors.h"
89 #include "libguile/lang.h"
90
91 #include "libguile/validate.h"
92 #include "libguile/sort.h"
93
94 /* The routine quicksort was extracted from the GNU C Library qsort.c
95 written by Douglas C. Schmidt (schmidt@ics.uci.edu)
96 and adapted to guile by adding an extra pointer less
97 to quicksort by Roland Orre <orre@nada.kth.se>.
98
99 The reason to do this instead of using the library function qsort
100 was to avoid dependency of the ANSI-C extensions for local functions
101 and also to avoid obscure pool based solutions.
102
103 This sorting routine is not much more efficient than the stable
104 version but doesn't consume extra memory.
105 */
106
107 /* Byte-wise swap two items of size SIZE. */
108 #define SWAP(a, b, size) \
109 do \
110 { \
111 register size_t __size = (size); \
112 register char *__a = (a), *__b = (b); \
113 do \
114 { \
115 char __tmp = *__a; \
116 *__a++ = *__b; \
117 *__b++ = __tmp; \
118 } while (--__size > 0); \
119 } while (0)
120
121 /* Discontinue quicksort algorithm when partition gets below this size.
122 This particular magic number was chosen to work best on a Sun 4/260. */
123 #define MAX_THRESH 4
124
125 /* Stack node declarations used to store unfulfilled partition obligations. */
126 typedef struct
127 {
128 char *lo;
129 char *hi;
130 }
131 stack_node;
132
133 /* The next 4 #defines implement a very fast in-line stack abstraction. */
134 #define STACK_SIZE (8 * sizeof(unsigned long int))
135 #define PUSH(low, high) ((void) ((top->lo = (low)), (top->hi = (high)), ++top))
136 #define POP(low, high) ((void) (--top, (low = top->lo), (high = top->hi)))
137 #define STACK_NOT_EMPTY (stack < top)
138
139
140 /* Order size using quicksort. This implementation incorporates
141 four optimizations discussed in Sedgewick:
142
143 1. Non-recursive, using an explicit stack of pointer that store the
144 next array partition to sort. To save time, this maximum amount
145 of space required to store an array of MAX_INT is allocated on the
146 stack. Assuming a 32-bit integer, this needs only 32 *
147 sizeof(stack_node) == 136 bits. Pretty cheap, actually.
148
149 2. Chose the pivot element using a median-of-three decision tree.
150 This reduces the probability of selecting a bad pivot value and
151 eliminates certain extraneous comparisons.
152
153 3. Only quicksorts TOTAL_ELEMS / MAX_THRESH partitions, leaving
154 insertion sort to order the MAX_THRESH items within each partition.
155 This is a big win, since insertion sort is faster for small, mostly
156 sorted array segments.
157
158 4. The larger of the two sub-partitions is always pushed onto the
159 stack first, with the algorithm then concentrating on the
160 smaller partition. This *guarantees* no more than log (n)
161 stack size is needed (actually O(1) in this case)! */
162
163 typedef int (*cmp_fun_t) (SCM less,
164 const void*,
165 const void*);
166
167 static const char s_buggy_less[] = "buggy less predicate used when sorting";
168
169 static void
170 quicksort (void *const pbase,
171 size_t total_elems,
172 size_t size,
173 cmp_fun_t cmp,
174 SCM less)
175 {
176 register char *base_ptr = (char *) pbase;
177
178 /* Allocating SIZE bytes for a pivot buffer facilitates a better
179 algorithm below since we can do comparisons directly on the pivot. */
180 char *pivot_buffer = (char *) alloca (size);
181 const size_t max_thresh = MAX_THRESH * size;
182
183 if (total_elems == 0)
184 /* Avoid lossage with unsigned arithmetic below. */
185 return;
186
187 if (total_elems > MAX_THRESH)
188 {
189 char *lo = base_ptr;
190 char *hi = &lo[size * (total_elems - 1)];
191 /* Largest size needed for 32-bit int!!! */
192 stack_node stack[STACK_SIZE];
193 stack_node *top = stack + 1;
194
195 while (STACK_NOT_EMPTY)
196 {
197 char *left_ptr;
198 char *right_ptr;
199
200 char *pivot = pivot_buffer;
201
202 /* Select median value from among LO, MID, and HI. Rearrange
203 LO and HI so the three values are sorted. This lowers the
204 probability of picking a pathological pivot value and
205 skips a comparison for both the LEFT_PTR and RIGHT_PTR. */
206
207 char *mid = lo + size * ((hi - lo) / size >> 1);
208
209 if ((*cmp) (less, (void *) mid, (void *) lo))
210 SWAP (mid, lo, size);
211 if ((*cmp) (less, (void *) hi, (void *) mid))
212 SWAP (mid, hi, size);
213 else
214 goto jump_over;
215 if ((*cmp) (less, (void *) mid, (void *) lo))
216 SWAP (mid, lo, size);
217 jump_over:;
218 memcpy (pivot, mid, size);
219 pivot = pivot_buffer;
220
221 left_ptr = lo + size;
222 right_ptr = hi - size;
223
224 /* Here's the famous ``collapse the walls'' section of quicksort.
225 Gotta like those tight inner loops! They are the main reason
226 that this algorithm runs much faster than others. */
227 do
228 {
229 while ((*cmp) (less, (void *) left_ptr, (void *) pivot))
230 {
231 left_ptr += size;
232 /* The comparison predicate may be buggy */
233 if (left_ptr > hi)
234 scm_misc_error (NULL, s_buggy_less, SCM_EOL);
235 }
236
237 while ((*cmp) (less, (void *) pivot, (void *) right_ptr))
238 {
239 right_ptr -= size;
240 /* The comparison predicate may be buggy */
241 if (right_ptr < lo)
242 scm_misc_error (NULL, s_buggy_less, SCM_EOL);
243 }
244
245 if (left_ptr < right_ptr)
246 {
247 SWAP (left_ptr, right_ptr, size);
248 left_ptr += size;
249 right_ptr -= size;
250 }
251 else if (left_ptr == right_ptr)
252 {
253 left_ptr += size;
254 right_ptr -= size;
255 break;
256 }
257 }
258 while (left_ptr <= right_ptr);
259
260 /* Set up pointers for next iteration. First determine whether
261 left and right partitions are below the threshold size. If so,
262 ignore one or both. Otherwise, push the larger partition's
263 bounds on the stack and continue sorting the smaller one. */
264
265 if ((size_t) (right_ptr - lo) <= max_thresh)
266 {
267 if ((size_t) (hi - left_ptr) <= max_thresh)
268 /* Ignore both small partitions. */
269 POP (lo, hi);
270 else
271 /* Ignore small left partition. */
272 lo = left_ptr;
273 }
274 else if ((size_t) (hi - left_ptr) <= max_thresh)
275 /* Ignore small right partition. */
276 hi = right_ptr;
277 else if ((right_ptr - lo) > (hi - left_ptr))
278 {
279 /* Push larger left partition indices. */
280 PUSH (lo, right_ptr);
281 lo = left_ptr;
282 }
283 else
284 {
285 /* Push larger right partition indices. */
286 PUSH (left_ptr, hi);
287 hi = right_ptr;
288 }
289 }
290 }
291
292 /* Once the BASE_PTR array is partially sorted by quicksort the rest
293 is completely sorted using insertion sort, since this is efficient
294 for partitions below MAX_THRESH size. BASE_PTR points to the beginning
295 of the array to sort, and END_PTR points at the very last element in
296 the array (*not* one beyond it!). */
297
298 {
299 char *const end_ptr = &base_ptr[size * (total_elems - 1)];
300 char *tmp_ptr = base_ptr;
301 char *thresh = min (end_ptr, base_ptr + max_thresh);
302 register char *run_ptr;
303
304 /* Find smallest element in first threshold and place it at the
305 array's beginning. This is the smallest array element,
306 and the operation speeds up insertion sort's inner loop. */
307
308 for (run_ptr = tmp_ptr + size; run_ptr <= thresh; run_ptr += size)
309 if ((*cmp) (less, (void *) run_ptr, (void *) tmp_ptr))
310 tmp_ptr = run_ptr;
311
312 if (tmp_ptr != base_ptr)
313 SWAP (tmp_ptr, base_ptr, size);
314
315 /* Insertion sort, running from left-hand-side up to right-hand-side. */
316
317 run_ptr = base_ptr + size;
318 while ((run_ptr += size) <= end_ptr)
319 {
320 tmp_ptr = run_ptr - size;
321 while ((*cmp) (less, (void *) run_ptr, (void *) tmp_ptr))
322 {
323 tmp_ptr -= size;
324 /* The comparison predicate may be buggy */
325 if (tmp_ptr < base_ptr)
326 scm_misc_error (NULL, s_buggy_less, SCM_EOL);
327 }
328
329 tmp_ptr += size;
330 if (tmp_ptr != run_ptr)
331 {
332 char *trav;
333
334 trav = run_ptr + size;
335 while (--trav >= run_ptr)
336 {
337 char c = *trav;
338 char *hi, *lo;
339
340 for (hi = lo = trav; (lo -= size) >= tmp_ptr; hi = lo)
341 *hi = *lo;
342 *hi = c;
343 }
344 }
345 }
346 }
347 } /* quicksort */
348
349
350 /* comparison routines */
351
352 static int
353 subr2less (SCM less, const void *a, const void *b)
354 {
355 return SCM_NFALSEP (SCM_SUBRF (less) (*(SCM *) a, *(SCM *) b));
356 } /* subr2less */
357
358 static int
359 lsubrless (SCM less, const void *a, const void *b)
360 {
361 return SCM_NFALSEP (SCM_SUBRF (less)
362 (scm_cons (*(SCM *) a,
363 scm_cons (*(SCM *) b, SCM_EOL))));
364 } /* lsubrless */
365
366 static int
367 closureless (SCM code, const void *a, const void *b)
368 {
369 SCM env = SCM_EXTEND_ENV (SCM_CLOSURE_FORMALS (code),
370 scm_cons (*(SCM *) a,
371 scm_cons (*(SCM *) b, SCM_EOL)),
372 SCM_ENV (code));
373 /* Evaluate the closure body */
374 return !SCM_FALSEP (scm_eval_body (SCM_CLOSURE_BODY (code), env));
375 } /* closureless */
376
377 static int
378 applyless (SCM less, const void *a, const void *b)
379 {
380 return SCM_NFALSEP (scm_call_2 (less, *(SCM *) a, *(SCM *) b));
381 } /* applyless */
382
383 static cmp_fun_t
384 scm_cmp_function (SCM p)
385 {
386 switch (SCM_TYP7 (p))
387 {
388 case scm_tc7_subr_2:
389 case scm_tc7_subr_2o:
390 case scm_tc7_rpsubr:
391 case scm_tc7_asubr:
392 return subr2less;
393 case scm_tc7_lsubr:
394 return lsubrless;
395 case scm_tcs_closures:
396 return closureless;
397 default:
398 return applyless;
399 }
400 } /* scm_cmp_function */
401
402
403 /* Question: Is there any need to make this a more general array sort?
404 It is probably enough to manage the vector type. */
405 /* endpos equal as for substring, i.e. endpos is not included. */
406 /* More natural with length? */
407
408 SCM_DEFINE (scm_restricted_vector_sort_x, "restricted-vector-sort!", 4, 0, 0,
409 (SCM vec, SCM less, SCM startpos, SCM endpos),
410 "Sort the vector @var{vec}, using @var{less} for comparing\n"
411 "the vector elements. @var{startpos} and @var{endpos} delimit\n"
412 "the range of the vector which gets sorted. The return value\n"
413 "is not specified.")
414 #define FUNC_NAME s_scm_restricted_vector_sort_x
415 {
416 size_t vlen, spos, len, size = sizeof (SCM);
417 SCM *vp;
418
419 SCM_VALIDATE_VECTOR (1, vec);
420 SCM_VALIDATE_NIM (2, less);
421
422 vp = SCM_WRITABLE_VELTS (vec); /* vector pointer */
423 vlen = SCM_VECTOR_LENGTH (vec);
424
425 SCM_VALIDATE_INUM_MIN_COPY (3, startpos, 0, spos);
426 SCM_ASSERT_RANGE (3, startpos, spos <= vlen);
427 SCM_VALIDATE_INUM_RANGE (4, endpos,0, vlen+1);
428 len = SCM_INUM (endpos) - spos;
429
430 quicksort (&vp[spos], len, size, scm_cmp_function (less), less);
431
432 return scm_return_first (SCM_UNSPECIFIED, vec);
433 /* return vec; */
434 }
435 #undef FUNC_NAME
436
437 /* (sorted? sequence less?)
438 * is true when sequence is a list (x0 x1 ... xm) or a vector #(x0 ... xm)
439 * such that for all 1 <= i <= m,
440 * (not (less? (list-ref list i) (list-ref list (- i 1)))). */
441 SCM_DEFINE (scm_sorted_p, "sorted?", 2, 0, 0,
442 (SCM items, SCM less),
443 "Return @code{#t} iff @var{items} is a list or a vector such that\n"
444 "for all 1 <= i <= m, the predicate @var{less} returns true when\n"
445 "applied to all elements i - 1 and i")
446 #define FUNC_NAME s_scm_sorted_p
447 {
448 long len, j; /* list/vector length, temp j */
449 SCM item, rest; /* rest of items loop variable */
450 SCM const *vp;
451 cmp_fun_t cmp = scm_cmp_function (less);
452
453 if (SCM_NULL_OR_NIL_P (items))
454 return SCM_BOOL_T;
455
456 SCM_VALIDATE_NIM (2, less);
457
458 if (SCM_CONSP (items))
459 {
460 len = scm_ilength (items); /* also checks that it's a pure list */
461 SCM_ASSERT_RANGE (1, items, len >= 0);
462 if (len <= 1)
463 return SCM_BOOL_T;
464
465 item = SCM_CAR (items);
466 rest = SCM_CDR (items);
467 j = len - 1;
468 while (j > 0)
469 {
470 if ((*cmp) (less, SCM_CARLOC(rest), &item))
471 return SCM_BOOL_F;
472 else
473 {
474 item = SCM_CAR (rest);
475 rest = SCM_CDR (rest);
476 j--;
477 }
478 }
479 return SCM_BOOL_T;
480 }
481 else
482 {
483 SCM_VALIDATE_VECTOR (1, items);
484
485 vp = SCM_VELTS (items); /* vector pointer */
486 len = SCM_VECTOR_LENGTH (items);
487 j = len - 1;
488 while (j > 0)
489 {
490 if ((*cmp) (less, &vp[1], vp))
491 return SCM_BOOL_F;
492 else
493 {
494 vp++;
495 j--;
496 }
497 }
498 return SCM_BOOL_T;
499 }
500
501 return SCM_BOOL_F;
502 }
503 #undef FUNC_NAME
504
505 /* (merge a b less?)
506 takes two lists a and b such that (sorted? a less?) and (sorted? b less?)
507 and returns a new list in which the elements of a and b have been stably
508 interleaved so that (sorted? (merge a b less?) less?).
509 Note: this does _not_ accept vectors. */
510 SCM_DEFINE (scm_merge, "merge", 3, 0, 0,
511 (SCM alist, SCM blist, SCM less),
512 "Merge two already sorted lists into one.\n"
513 "Given two lists @var{alist} and @var{blist}, such that\n"
514 "@code{(sorted? alist less?)} and @code{(sorted? blist less?)},\n"
515 "return a new list in which the elements of @var{alist} and\n"
516 "@var{blist} have been stably interleaved so that\n"
517 "@code{(sorted? (merge alist blist less?) less?)}.\n"
518 "Note: this does _not_ accept vectors.")
519 #define FUNC_NAME s_scm_merge
520 {
521 long alen, blen; /* list lengths */
522 SCM build, last;
523 cmp_fun_t cmp = scm_cmp_function (less);
524 SCM_VALIDATE_NIM (3, less);
525
526 if (SCM_NULL_OR_NIL_P (alist))
527 return blist;
528 else if (SCM_NULL_OR_NIL_P (blist))
529 return alist;
530 else
531 {
532 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1, alist, alen);
533 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2, blist, blen);
534 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
535 {
536 build = scm_cons (SCM_CAR (blist), SCM_EOL);
537 blist = SCM_CDR (blist);
538 blen--;
539 }
540 else
541 {
542 build = scm_cons (SCM_CAR (alist), SCM_EOL);
543 alist = SCM_CDR (alist);
544 alen--;
545 }
546 last = build;
547 while ((alen > 0) && (blen > 0))
548 {
549 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
550 {
551 SCM_SETCDR (last, scm_cons (SCM_CAR (blist), SCM_EOL));
552 blist = SCM_CDR (blist);
553 blen--;
554 }
555 else
556 {
557 SCM_SETCDR (last, scm_cons (SCM_CAR (alist), SCM_EOL));
558 alist = SCM_CDR (alist);
559 alen--;
560 }
561 last = SCM_CDR (last);
562 }
563 if ((alen > 0) && (blen == 0))
564 SCM_SETCDR (last, alist);
565 else if ((alen == 0) && (blen > 0))
566 SCM_SETCDR (last, blist);
567 }
568 return build;
569 }
570 #undef FUNC_NAME
571
572
573 static SCM
574 scm_merge_list_x (SCM alist, SCM blist,
575 long alen, long blen,
576 cmp_fun_t cmp, SCM less)
577 {
578 SCM build, last;
579
580 if (SCM_NULL_OR_NIL_P (alist))
581 return blist;
582 else if (SCM_NULL_OR_NIL_P (blist))
583 return alist;
584 else
585 {
586 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
587 {
588 build = blist;
589 blist = SCM_CDR (blist);
590 blen--;
591 }
592 else
593 {
594 build = alist;
595 alist = SCM_CDR (alist);
596 alen--;
597 }
598 last = build;
599 while ((alen > 0) && (blen > 0))
600 {
601 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
602 {
603 SCM_SETCDR (last, blist);
604 blist = SCM_CDR (blist);
605 blen--;
606 }
607 else
608 {
609 SCM_SETCDR (last, alist);
610 alist = SCM_CDR (alist);
611 alen--;
612 }
613 last = SCM_CDR (last);
614 }
615 if ((alen > 0) && (blen == 0))
616 SCM_SETCDR (last, alist);
617 else if ((alen == 0) && (blen > 0))
618 SCM_SETCDR (last, blist);
619 }
620 return build;
621 } /* scm_merge_list_x */
622
623 SCM_DEFINE (scm_merge_x, "merge!", 3, 0, 0,
624 (SCM alist, SCM blist, SCM less),
625 "Takes two lists @var{alist} and @var{blist} such that\n"
626 "@code{(sorted? alist less?)} and @code{(sorted? blist less?)} and\n"
627 "returns a new list in which the elements of @var{alist} and\n"
628 "@var{blist} have been stably interleaved so that\n"
629 " @code{(sorted? (merge alist blist less?) less?)}.\n"
630 "This is the destructive variant of @code{merge}\n"
631 "Note: this does _not_ accept vectors.")
632 #define FUNC_NAME s_scm_merge_x
633 {
634 long alen, blen; /* list lengths */
635
636 SCM_VALIDATE_NIM (3, less);
637 if (SCM_NULL_OR_NIL_P (alist))
638 return blist;
639 else if (SCM_NULL_OR_NIL_P (blist))
640 return alist;
641 else
642 {
643 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1, alist, alen);
644 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2, blist, blen);
645 return scm_merge_list_x (alist, blist,
646 alen, blen,
647 scm_cmp_function (less),
648 less);
649 }
650 }
651 #undef FUNC_NAME
652
653 /* This merge sort algorithm is same as slib's by Richard A. O'Keefe.
654 The algorithm is stable. We also tried to use the algorithm used by
655 scsh's merge-sort but that algorithm showed to not be stable, even
656 though it claimed to be.
657 */
658 static SCM
659 scm_merge_list_step (SCM * seq,
660 cmp_fun_t cmp,
661 SCM less,
662 long n)
663 {
664 SCM a, b;
665
666 if (n > 2)
667 {
668 long mid = n / 2;
669 a = scm_merge_list_step (seq, cmp, less, mid);
670 b = scm_merge_list_step (seq, cmp, less, n - mid);
671 return scm_merge_list_x (a, b, mid, n - mid, cmp, less);
672 }
673 else if (n == 2)
674 {
675 SCM p = *seq;
676 SCM rest = SCM_CDR (*seq);
677 SCM x = SCM_CAR (*seq);
678 SCM y = SCM_CAR (SCM_CDR (*seq));
679 *seq = SCM_CDR (rest);
680 SCM_SETCDR (rest, SCM_EOL);
681 if ((*cmp) (less, &y, &x))
682 {
683 SCM_SETCAR (p, y);
684 SCM_SETCAR (rest, x);
685 }
686 return p;
687 }
688 else if (n == 1)
689 {
690 SCM p = *seq;
691 *seq = SCM_CDR (p);
692 SCM_SETCDR (p, SCM_EOL);
693 return p;
694 }
695 else
696 return SCM_EOL;
697 } /* scm_merge_list_step */
698
699
700 /* scm_sort_x manages lists and vectors, not stable sort */
701 SCM_DEFINE (scm_sort_x, "sort!", 2, 0, 0,
702 (SCM items, SCM less),
703 "Sort the sequence @var{items}, which may be a list or a\n"
704 "vector. @var{less} is used for comparing the sequence\n"
705 "elements. The sorting is destructive, that means that the\n"
706 "input sequence is modified to produce the sorted result.\n"
707 "This is not a stable sort.")
708 #define FUNC_NAME s_scm_sort_x
709 {
710 long len; /* list/vector length */
711 if (SCM_NULL_OR_NIL_P (items))
712 return items;
713
714 SCM_VALIDATE_NIM (2, less);
715
716 if (SCM_CONSP (items))
717 {
718 SCM_VALIDATE_LIST_COPYLEN (1, items, len);
719 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
720 }
721 else if (SCM_VECTORP (items))
722 {
723 len = SCM_VECTOR_LENGTH (items);
724 scm_restricted_vector_sort_x (items,
725 less,
726 SCM_MAKINUM (0L),
727 SCM_MAKINUM (len));
728 return items;
729 }
730 else
731 SCM_WRONG_TYPE_ARG (1, items);
732 }
733 #undef FUNC_NAME
734
735 /* scm_sort manages lists and vectors, not stable sort */
736
737 SCM_DEFINE (scm_sort, "sort", 2, 0, 0,
738 (SCM items, SCM less),
739 "Sort the sequence @var{items}, which may be a list or a\n"
740 "vector. @var{less} is used for comparing the sequence\n"
741 "elements. This is not a stable sort.")
742 #define FUNC_NAME s_scm_sort
743 {
744 if (SCM_NULL_OR_NIL_P (items))
745 return items;
746
747 SCM_VALIDATE_NIM (2, less);
748 if (SCM_CONSP (items))
749 {
750 long len;
751
752 SCM_VALIDATE_LIST_COPYLEN (1, items, len);
753 items = scm_list_copy (items);
754 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
755 }
756 #ifdef HAVE_ARRAYS
757 /* support ordinary vectors even if arrays not available? */
758 else if (SCM_VECTORP (items))
759 {
760 long len = SCM_VECTOR_LENGTH (items);
761 SCM sortvec = scm_make_uve (len, scm_array_prototype (items));
762
763 scm_array_copy_x (items, sortvec);
764 scm_restricted_vector_sort_x (sortvec,
765 less,
766 SCM_MAKINUM (0L),
767 SCM_MAKINUM (len));
768 return sortvec;
769 }
770 #endif
771 else
772 SCM_WRONG_TYPE_ARG (1, items);
773 }
774 #undef FUNC_NAME
775
776 static void
777 scm_merge_vector_x (SCM vec,
778 SCM * temp,
779 cmp_fun_t cmp,
780 SCM less,
781 long low,
782 long mid,
783 long high)
784 {
785 long it; /* Index for temp vector */
786 long i1 = low; /* Index for lower vector segment */
787 long i2 = mid + 1; /* Index for upper vector segment */
788
789 /* Copy while both segments contain more characters */
790 for (it = low; (i1 <= mid) && (i2 <= high); ++it)
791 {
792 /*
793 Every call of LESS might invoke GC. For full correctness, we
794 should reset the generation of vecbase and tempbase between
795 every call of less.
796
797 */
798 register SCM *vp = SCM_WRITABLE_VELTS(vec);
799
800 if ((*cmp) (less, &vp[i2], &vp[i1]))
801 temp[it] = vp[i2++];
802 else
803 temp[it] = vp[i1++];
804 }
805
806 {
807 register SCM *vp = SCM_WRITABLE_VELTS(vec);
808
809 /* Copy while first segment contains more characters */
810 while (i1 <= mid)
811 temp[it++] = vp[i1++];
812
813 /* Copy while second segment contains more characters */
814 while (i2 <= high)
815 temp[it++] = vp[i2++];
816
817 /* Copy back from temp to vp */
818 for (it = low; it <= high; ++it)
819 vp[it] = temp[it];
820 }
821 } /* scm_merge_vector_x */
822
823 static void
824 scm_merge_vector_step (SCM vp,
825 SCM * temp,
826 cmp_fun_t cmp,
827 SCM less,
828 long low,
829 long high)
830 {
831 if (high > low)
832 {
833 long mid = (low + high) / 2;
834 scm_merge_vector_step (vp, temp, cmp, less, low, mid);
835 scm_merge_vector_step (vp, temp, cmp, less, mid+1, high);
836 scm_merge_vector_x (vp, temp, cmp, less, low, mid, high);
837 }
838 } /* scm_merge_vector_step */
839
840
841 /* stable-sort! manages lists and vectors */
842
843 SCM_DEFINE (scm_stable_sort_x, "stable-sort!", 2, 0, 0,
844 (SCM items, SCM less),
845 "Sort the sequence @var{items}, which may be a list or a\n"
846 "vector. @var{less} is used for comparing the sequence elements.\n"
847 "The sorting is destructive, that means that the input sequence\n"
848 "is modified to produce the sorted result.\n"
849 "This is a stable sort.")
850 #define FUNC_NAME s_scm_stable_sort_x
851 {
852 long len; /* list/vector length */
853
854 if (SCM_NULL_OR_NIL_P (items))
855 return items;
856
857 SCM_VALIDATE_NIM (2, less);
858 if (SCM_CONSP (items))
859 {
860 SCM_VALIDATE_LIST_COPYLEN (1, items, len);
861 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
862 }
863 else if (SCM_VECTORP (items))
864 {
865 SCM *temp;
866 len = SCM_VECTOR_LENGTH (items);
867
868 /*
869 the following array does not contain any new references to
870 SCM objects, so we can get away with allocing it on the heap.
871 */
872 temp = scm_malloc (len * sizeof(SCM));
873
874 scm_merge_vector_step (items,
875 temp,
876 scm_cmp_function (less),
877 less,
878 0,
879 len - 1);
880 free(temp);
881 return items;
882 }
883 else
884 SCM_WRONG_TYPE_ARG (1, items);
885 }
886 #undef FUNC_NAME
887
888 /* stable_sort manages lists and vectors */
889 SCM_DEFINE (scm_stable_sort, "stable-sort", 2, 0, 0,
890 (SCM items, SCM less),
891 "Sort the sequence @var{items}, which may be a list or a\n"
892 "vector. @var{less} is used for comparing the sequence elements.\n"
893 "This is a stable sort.")
894 #define FUNC_NAME s_scm_stable_sort
895 {
896
897 if (SCM_NULL_OR_NIL_P (items))
898 return items;
899
900 SCM_VALIDATE_NIM (2, less);
901 if (SCM_CONSP (items))
902 {
903 long len; /* list/vector length */
904 SCM_VALIDATE_LIST_COPYLEN (1, items, len);
905 items = scm_list_copy (items);
906 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
907 }
908 #ifdef HAVE_ARRAYS
909 /* support ordinary vectors even if arrays not available? */
910 else if (SCM_VECTORP (items))
911 {
912 long len = SCM_VECTOR_LENGTH (items);
913 SCM *temp = scm_malloc (len * sizeof (SCM));
914 SCM retvec = scm_make_uve (len, scm_array_prototype (items));
915 scm_array_copy_x (items, retvec);
916
917 scm_merge_vector_step (retvec,
918 temp,
919 scm_cmp_function (less),
920 less,
921 0,
922 len - 1);
923 free (temp);
924 return retvec;
925 }
926 #endif
927 else
928 SCM_WRONG_TYPE_ARG (1, items);
929 }
930 #undef FUNC_NAME
931
932 /* stable */
933 SCM_DEFINE (scm_sort_list_x, "sort-list!", 2, 0, 0,
934 (SCM items, SCM less),
935 "Sort the list @var{items}, using @var{less} for comparing the\n"
936 "list elements. The sorting is destructive, that means that the\n"
937 "input list is modified to produce the sorted result.\n"
938 "This is a stable sort.")
939 #define FUNC_NAME s_scm_sort_list_x
940 {
941 long len;
942 SCM_VALIDATE_LIST_COPYLEN (1, items, len);
943 SCM_VALIDATE_NIM (2, less);
944 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
945 }
946 #undef FUNC_NAME
947
948 /* stable */
949 SCM_DEFINE (scm_sort_list, "sort-list", 2, 0, 0,
950 (SCM items, SCM less),
951 "Sort the list @var{items}, using @var{less} for comparing the\n"
952 "list elements. This is a stable sort.")
953 #define FUNC_NAME s_scm_sort_list
954 {
955 long len;
956 SCM_VALIDATE_LIST_COPYLEN (1, items, len);
957 SCM_VALIDATE_NIM (2, less);
958 items = scm_list_copy (items);
959 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
960 }
961 #undef FUNC_NAME
962
963 void
964 scm_init_sort ()
965 {
966 #include "libguile/sort.x"
967
968 scm_add_feature ("sort");
969 }
970
971 /*
972 Local Variables:
973 c-file-style: "gnu"
974 End:
975 */