* configure.in: don't append threads.doc to EXTRA_DOT_DOC_FILES,
[bpt/guile.git] / libguile / sort.c
1 /* Copyright (C) 1999, 2000 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 /* Software engineering face-lift by Greg J. Badros, 11-Dec-1999,
42 gjb@cs.washington.edu, http://www.cs.washington.edu/homes/gjb */
43
44
45 /* Written in December 1998 by Roland Orre <orre@nada.kth.se>
46 * This implements the same sort interface as slib/sort.scm
47 * for lists and vectors where slib defines:
48 * sorted?, merge, merge!, sort, sort!
49 * For scsh compatibility sort-list and sort-list! are also defined.
50 * In cases where a stable-sort is required use stable-sort or
51 * stable-sort!. An additional feature is
52 * (restricted-vector-sort! vector less? startpos endpos)
53 * which allows you to sort part of a vector.
54 * Thanks to Aubrey Jaffer for the slib/sort.scm library.
55 * Thanks to Richard A. O'Keefe (based on Prolog code by D.H.D.Warren)
56 * for the merge sort inspiration.
57 * Thanks to Douglas C. Schmidt (schmidt@ics.uci.edu) for the
58 * quicksort code.
59 */
60
61 /* We need this to get the definitions for HAVE_ALLOCA_H, etc. */
62 #include "libguile/scmconfig.h"
63
64 /* AIX requires this to be the first thing in the file. The #pragma
65 directive is indented so pre-ANSI compilers will ignore it, rather
66 than choke on it. */
67 #ifndef __GNUC__
68 # if HAVE_ALLOCA_H
69 # include <alloca.h>
70 # else
71 # ifdef _AIX
72 #pragma alloca
73 # else
74 # ifndef alloca /* predefined by HP cc +Olibcalls */
75 char *alloca ();
76 # endif
77 # endif
78 # endif
79 #endif
80
81 #include <string.h>
82 #include "libguile/_scm.h"
83
84 #include "libguile/eval.h"
85 #include "libguile/unif.h"
86 #include "libguile/ramap.h"
87 #include "libguile/alist.h"
88 #include "libguile/feature.h"
89 #include "libguile/vectors.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 subr2oless (SCM less, const void *a, const void *b)
360 {
361 return SCM_NFALSEP (SCM_SUBRF (less) (*(SCM *) a,
362 *(SCM *) b,
363 SCM_UNDEFINED));
364 } /* subr2oless */
365
366 static int
367 lsubrless (SCM less, const void *a, const void *b)
368 {
369 return SCM_NFALSEP (SCM_SUBRF (less)
370 (scm_cons (*(SCM *) a,
371 scm_cons (*(SCM *) b, SCM_EOL))));
372 } /* lsubrless */
373
374 static int
375 closureless (SCM code, const void *a, const void *b)
376 {
377 SCM env = SCM_EXTEND_ENV (SCM_CAR (SCM_CODE (code)),
378 scm_cons (*(SCM *) a,
379 scm_cons (*(SCM *) b, SCM_EOL)),
380 SCM_ENV (code));
381 /* Evaluate the closure body */
382 return SCM_NFALSEP (scm_eval_body (SCM_CDR (SCM_CODE (code)), env));
383 } /* closureless */
384
385 static int
386 applyless (SCM less, const void *a, const void *b)
387 {
388 return SCM_NFALSEP (scm_apply (less,
389 scm_cons (*(SCM *) a,
390 scm_cons (*(SCM *) b, SCM_EOL)),
391 SCM_EOL));
392 } /* applyless */
393
394 static cmp_fun_t
395 scm_cmp_function (SCM p)
396 {
397 switch (SCM_TYP7 (p))
398 {
399 case scm_tc7_subr_2:
400 case scm_tc7_rpsubr:
401 case scm_tc7_asubr:
402 return subr2less;
403 case scm_tc7_subr_2o:
404 return subr2oless;
405 case scm_tc7_lsubr:
406 return lsubrless;
407 case scm_tcs_closures:
408 return closureless;
409 default:
410 return applyless;
411 }
412 } /* scm_cmp_function */
413
414
415 /* Question: Is there any need to make this a more general array sort?
416 It is probably enough to manage the vector type. */
417 /* endpos equal as for substring, i.e. endpos is not included. */
418 /* More natural with length? */
419
420 SCM_DEFINE (scm_restricted_vector_sort_x, "restricted-vector-sort!", 4, 0, 0,
421 (SCM vec, SCM less, SCM startpos, SCM endpos),
422 "Sort the vector @var{vec}, using @var{less} for comparing\n"
423 "the vector elements. @var{startpos} and @var{endpos} delimit\n"
424 "the range of the vector which gets sorted. The return value\n"
425 "is not specified.")
426 #define FUNC_NAME s_scm_restricted_vector_sort_x
427 {
428 size_t vlen, spos, len, size = sizeof (SCM);
429 SCM *vp;
430
431 SCM_VALIDATE_VECTOR (1,vec);
432 SCM_VALIDATE_NIM (2,less);
433
434 vp = SCM_VELTS (vec); /* vector pointer */
435 vlen = SCM_VECTOR_LENGTH (vec);
436
437 SCM_VALIDATE_INUM_COPY (3,startpos,spos);
438 SCM_ASSERT_RANGE (3,startpos,(spos >= 0) && (spos <= vlen));
439 SCM_VALIDATE_INUM_RANGE (4,endpos,0,vlen+1);
440 len = SCM_INUM (endpos) - spos;
441
442 quicksort (&vp[spos], len, size, scm_cmp_function (less), less);
443 return SCM_UNSPECIFIED;
444 /* return vec; */
445 }
446 #undef FUNC_NAME
447
448 /* (sorted? sequence less?)
449 * is true when sequence is a list (x0 x1 ... xm) or a vector #(x0 ... xm)
450 * such that for all 1 <= i <= m,
451 * (not (less? (list-ref list i) (list-ref list (- i 1)))). */
452 SCM_DEFINE (scm_sorted_p, "sorted?", 2, 0, 0,
453 (SCM items, SCM less),
454 "Return @code{#t} iff @var{items} is a list or a vector such that\n"
455 "for all 1 <= i <= m, the predicate @var{less} returns true when\n"
456 "applied to all elements i - 1 and i")
457 #define FUNC_NAME s_scm_sorted_p
458 {
459 long len, j; /* list/vector length, temp j */
460 SCM item, rest; /* rest of items loop variable */
461 SCM *vp;
462 cmp_fun_t cmp = scm_cmp_function (less);
463
464 if (SCM_NULLP (items))
465 return SCM_BOOL_T;
466
467 SCM_VALIDATE_NIM (2,less);
468
469 if (SCM_CONSP (items))
470 {
471 len = scm_ilength (items); /* also checks that it's a pure list */
472 SCM_ASSERT_RANGE (1,items,len >= 0);
473 if (len <= 1)
474 return SCM_BOOL_T;
475
476 item = SCM_CAR (items);
477 rest = SCM_CDR (items);
478 j = len - 1;
479 while (j > 0)
480 {
481 if ((*cmp) (less, SCM_CARLOC(rest), &item))
482 return SCM_BOOL_F;
483 else
484 {
485 item = SCM_CAR (rest);
486 rest = SCM_CDR (rest);
487 j--;
488 }
489 }
490 return SCM_BOOL_T;
491 }
492 else
493 {
494 SCM_VALIDATE_VECTOR (1, items);
495
496 vp = SCM_VELTS (items); /* vector pointer */
497 len = SCM_VECTOR_LENGTH (items);
498 j = len - 1;
499 while (j > 0)
500 {
501 if ((*cmp) (less, &vp[1], vp))
502 return SCM_BOOL_F;
503 else
504 {
505 vp++;
506 j--;
507 }
508 }
509 return SCM_BOOL_T;
510 }
511
512 return SCM_BOOL_F;
513 }
514 #undef FUNC_NAME
515
516 /* (merge a b less?)
517 takes two lists a and b such that (sorted? a less?) and (sorted? b less?)
518 and returns a new list in which the elements of a and b have been stably
519 interleaved so that (sorted? (merge a b less?) less?).
520 Note: this does _not_ accept vectors. */
521 SCM_DEFINE (scm_merge, "merge", 3, 0, 0,
522 (SCM alist, SCM blist, SCM less),
523 "Takes two lists @var{alist} and @var{blist} such that\n"
524 "@code{(sorted? alist less?)} and @code{(sorted? blist less?)} and\n"
525 "returns a new list in which the elements of @var{alist} and\n"
526 "@var{blist} have been stably interleaved so that\n"
527 "@code{(sorted? (merge alist blist less?) less?)}.\n"
528 "Note: this does _not_ accept vectors.")
529 #define FUNC_NAME s_scm_merge
530 {
531 long alen, blen; /* list lengths */
532 SCM build, last;
533 cmp_fun_t cmp = scm_cmp_function (less);
534 SCM_VALIDATE_NIM (3,less);
535
536 if (SCM_NULLP (alist))
537 return blist;
538 else if (SCM_NULLP (blist))
539 return alist;
540 else
541 {
542 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1,alist,alen);
543 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2,blist,blen);
544 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
545 {
546 build = scm_cons (SCM_CAR (blist), SCM_EOL);
547 blist = SCM_CDR (blist);
548 blen--;
549 }
550 else
551 {
552 build = scm_cons (SCM_CAR (alist), SCM_EOL);
553 alist = SCM_CDR (alist);
554 alen--;
555 }
556 last = build;
557 while ((alen > 0) && (blen > 0))
558 {
559 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
560 {
561 SCM_SETCDR (last, scm_cons (SCM_CAR (blist), SCM_EOL));
562 blist = SCM_CDR (blist);
563 blen--;
564 }
565 else
566 {
567 SCM_SETCDR (last, scm_cons (SCM_CAR (alist), SCM_EOL));
568 alist = SCM_CDR (alist);
569 alen--;
570 }
571 last = SCM_CDR (last);
572 }
573 if ((alen > 0) && (blen == 0))
574 SCM_SETCDR (last, alist);
575 else if ((alen == 0) && (blen > 0))
576 SCM_SETCDR (last, blist);
577 }
578 return build;
579 }
580 #undef FUNC_NAME
581
582
583 static SCM
584 scm_merge_list_x (SCM alist, SCM blist,
585 long alen, long blen,
586 cmp_fun_t cmp, SCM less)
587 {
588 SCM build, last;
589
590 if (SCM_NULLP (alist))
591 return blist;
592 else if (SCM_NULLP (blist))
593 return alist;
594 else
595 {
596 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
597 {
598 build = blist;
599 blist = SCM_CDR (blist);
600 blen--;
601 }
602 else
603 {
604 build = alist;
605 alist = SCM_CDR (alist);
606 alen--;
607 }
608 last = build;
609 while ((alen > 0) && (blen > 0))
610 {
611 if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist)))
612 {
613 SCM_SETCDR (last, blist);
614 blist = SCM_CDR (blist);
615 blen--;
616 }
617 else
618 {
619 SCM_SETCDR (last, alist);
620 alist = SCM_CDR (alist);
621 alen--;
622 }
623 last = SCM_CDR (last);
624 }
625 if ((alen > 0) && (blen == 0))
626 SCM_SETCDR (last, alist);
627 else if ((alen == 0) && (blen > 0))
628 SCM_SETCDR (last, blist);
629 }
630 return build;
631 } /* scm_merge_list_x */
632
633 SCM_DEFINE (scm_merge_x, "merge!", 3, 0, 0,
634 (SCM alist, SCM blist, SCM less),
635 "Takes two lists @var{alist} and @var{blist} such that\n"
636 "@code{(sorted? alist less?)} and @code{(sorted? blist less?)} and\n"
637 "returns a new list in which the elements of @var{alist} and\n"
638 "@var{blist} have been stably interleaved so that\n"
639 " @code{(sorted? (merge alist blist less?) less?)}.\n"
640 "This is the destructive variant of @code{merge}\n"
641 "Note: this does _not_ accept vectors.")
642 #define FUNC_NAME s_scm_merge_x
643 {
644 long alen, blen; /* list lengths */
645
646 SCM_VALIDATE_NIM (3,less);
647 if (SCM_NULLP (alist))
648 return blist;
649 else if (SCM_NULLP (blist))
650 return alist;
651 else
652 {
653 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1,alist,alen);
654 SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2,blist,blen);
655 return scm_merge_list_x (alist, blist,
656 alen, blen,
657 scm_cmp_function (less),
658 less);
659 }
660 }
661 #undef FUNC_NAME
662
663 /* This merge sort algorithm is same as slib's by Richard A. O'Keefe.
664 The algorithm is stable. We also tried to use the algorithm used by
665 scsh's merge-sort but that algorithm showed to not be stable, even
666 though it claimed to be.
667 */
668 static SCM
669 scm_merge_list_step (SCM * seq,
670 cmp_fun_t cmp,
671 SCM less,
672 int n)
673 {
674 SCM a, b;
675
676 if (n > 2)
677 {
678 long mid = n / 2;
679 a = scm_merge_list_step (seq, cmp, less, mid);
680 b = scm_merge_list_step (seq, cmp, less, n - mid);
681 return scm_merge_list_x (a, b, mid, n - mid, cmp, less);
682 }
683 else if (n == 2)
684 {
685 SCM p = *seq;
686 SCM rest = SCM_CDR (*seq);
687 SCM x = SCM_CAR (*seq);
688 SCM y = SCM_CAR (SCM_CDR (*seq));
689 *seq = SCM_CDR (rest);
690 SCM_SETCDR (rest, SCM_EOL);
691 if ((*cmp) (less, &y, &x))
692 {
693 SCM_SETCAR (p, y);
694 SCM_SETCAR (rest, x);
695 }
696 return p;
697 }
698 else if (n == 1)
699 {
700 SCM p = *seq;
701 *seq = SCM_CDR (p);
702 SCM_SETCDR (p, SCM_EOL);
703 return p;
704 }
705 else
706 return SCM_EOL;
707 } /* scm_merge_list_step */
708
709
710 /* scm_sort_x manages lists and vectors, not stable sort */
711 SCM_DEFINE (scm_sort_x, "sort!", 2, 0, 0,
712 (SCM items, SCM less),
713 "Sort the sequence @var{items}, which may be a list or a\n"
714 "vector. @var{less} is used for comparing the sequence\n"
715 "elements. The sorting is destructive, that means that the\n"
716 "input sequence is modified to produce the sorted result.\n"
717 "This is not a stable sort.")
718 #define FUNC_NAME s_scm_sort_x
719 {
720 long len; /* list/vector length */
721 if (SCM_NULLP(items))
722 return SCM_EOL;
723
724 SCM_VALIDATE_NIM (2,less);
725
726 if (SCM_CONSP (items))
727 {
728 SCM_VALIDATE_LIST_COPYLEN (1,items,len);
729 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
730 }
731 else if (SCM_VECTORP (items))
732 {
733 len = SCM_VECTOR_LENGTH (items);
734 scm_restricted_vector_sort_x (items,
735 less,
736 SCM_MAKINUM (0L),
737 SCM_MAKINUM (len));
738 return items;
739 }
740 else
741 SCM_WRONG_TYPE_ARG (1, items);
742 }
743 #undef FUNC_NAME
744
745 /* scm_sort manages lists and vectors, not stable sort */
746
747 SCM_DEFINE (scm_sort, "sort", 2, 0, 0,
748 (SCM items, SCM less),
749 "Sort the sequence @var{items}, which may be a list or a\n"
750 "vector. @var{less} is used for comparing the sequence\n"
751 "elements. This is not a stable sort.")
752 #define FUNC_NAME s_scm_sort
753 {
754 if (SCM_NULLP(items))
755 return SCM_EOL;
756
757 SCM_VALIDATE_NIM (2,less);
758 if (SCM_CONSP (items))
759 {
760 long len;
761
762 SCM_VALIDATE_LIST_COPYLEN (1,items,len);
763 items = scm_list_copy (items);
764 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
765 }
766 #ifdef HAVE_ARRAYS
767 /* support ordinary vectors even if arrays not available? */
768 else if (SCM_VECTORP (items))
769 {
770 long len = SCM_VECTOR_LENGTH (items);
771 SCM sortvec = scm_make_uve (len, scm_array_prototype (items));
772
773 scm_array_copy_x (items, sortvec);
774 scm_restricted_vector_sort_x (sortvec,
775 less,
776 SCM_MAKINUM (0L),
777 SCM_MAKINUM (len));
778 return sortvec;
779 }
780 #endif
781 else
782 SCM_WRONG_TYPE_ARG (1, items);
783 }
784 #undef FUNC_NAME
785
786 static void
787 scm_merge_vector_x (void *const vecbase,
788 void *const tempbase,
789 cmp_fun_t cmp,
790 SCM less,
791 long low,
792 long mid,
793 long high)
794 {
795 register SCM *vp = (SCM *) vecbase;
796 register SCM *temp = (SCM *) tempbase;
797 long it; /* Index for temp vector */
798 long i1 = low; /* Index for lower vector segment */
799 long i2 = mid + 1; /* Index for upper vector segment */
800
801 /* Copy while both segments contain more characters */
802 for (it = low; (i1 <= mid) && (i2 <= high); ++it)
803 if ((*cmp) (less, &vp[i2], &vp[i1]))
804 temp[it] = vp[i2++];
805 else
806 temp[it] = vp[i1++];
807
808 /* Copy while first segment contains more characters */
809 while (i1 <= mid)
810 temp[it++] = vp[i1++];
811
812 /* Copy while second segment contains more characters */
813 while (i2 <= high)
814 temp[it++] = vp[i2++];
815
816 /* Copy back from temp to vp */
817 for (it = low; it <= high; ++it)
818 vp[it] = temp[it];
819 } /* scm_merge_vector_x */
820
821 static void
822 scm_merge_vector_step (void *const vp,
823 void *const temp,
824 cmp_fun_t cmp,
825 SCM less,
826 long low,
827 long high)
828 {
829 if (high > low)
830 {
831 long mid = (low + high) / 2;
832 scm_merge_vector_step (vp, temp, cmp, less, low, mid);
833 scm_merge_vector_step (vp, temp, cmp, less, mid+1, high);
834 scm_merge_vector_x (vp, temp, cmp, less, low, mid, high);
835 }
836 } /* scm_merge_vector_step */
837
838
839 /* stable-sort! manages lists and vectors */
840
841 SCM_DEFINE (scm_stable_sort_x, "stable-sort!", 2, 0, 0,
842 (SCM items, SCM less),
843 "Sort the sequence @var{items}, which may be a list or a\n"
844 "vector. @var{less} is used for comparing the sequence elements.\n"
845 "The sorting is destructive, that means that the input sequence\n"
846 "is modified to produce the sorted result.\n"
847 "This is a stable sort.")
848 #define FUNC_NAME s_scm_stable_sort_x
849 {
850 long len; /* list/vector length */
851
852 if (SCM_NULLP (items))
853 return SCM_EOL;
854
855 SCM_VALIDATE_NIM (2,less);
856 if (SCM_CONSP (items))
857 {
858 SCM_VALIDATE_LIST_COPYLEN (1,items,len);
859 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
860 }
861 else if (SCM_VECTORP (items))
862 {
863 SCM *temp, *vp;
864 len = SCM_VECTOR_LENGTH (items);
865 temp = malloc (len * sizeof(SCM));
866 vp = SCM_VELTS (items);
867 scm_merge_vector_step (vp,
868 temp,
869 scm_cmp_function (less),
870 less,
871 0,
872 len - 1);
873 free(temp);
874 return items;
875 }
876 else
877 SCM_WRONG_TYPE_ARG (1, items);
878 }
879 #undef FUNC_NAME
880
881 /* stable_sort manages lists and vectors */
882
883 SCM_DEFINE (scm_stable_sort, "stable-sort", 2, 0, 0,
884 (SCM items, SCM less),
885 "Sort the sequence @var{items}, which may be a list or a\n"
886 "vector. @var{less} is used for comparing the sequence elements.\n"
887 "This is a stable sort.")
888 #define FUNC_NAME s_scm_stable_sort
889 {
890 long len; /* list/vector length */
891 if (SCM_NULLP (items))
892 return SCM_EOL;
893
894 SCM_VALIDATE_NIM (2,less);
895 if (SCM_CONSP (items))
896 {
897 SCM_VALIDATE_LIST_COPYLEN (1,items,len);
898 items = scm_list_copy (items);
899 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
900 }
901 #ifdef HAVE_ARRAYS
902 /* support ordinary vectors even if arrays not available? */
903 else if (SCM_VECTORP (items))
904 {
905 SCM retvec;
906 SCM *temp, *vp;
907 len = SCM_VECTOR_LENGTH (items);
908 retvec = scm_make_uve (len, scm_array_prototype (items));
909 scm_array_copy_x (items, retvec);
910 temp = malloc (len * sizeof (SCM));
911 vp = SCM_VELTS (retvec);
912 scm_merge_vector_step (vp,
913 temp,
914 scm_cmp_function (less),
915 less,
916 0,
917 len - 1);
918 free (temp);
919 return retvec;
920 }
921 #endif
922 else
923 SCM_WRONG_TYPE_ARG (1, items);
924 }
925 #undef FUNC_NAME
926
927 /* stable */
928 SCM_DEFINE (scm_sort_list_x, "sort-list!", 2, 0, 0,
929 (SCM items, SCM less),
930 "Sort the list @var{items}, using @var{less} for comparing the\n"
931 "list elements. The sorting is destructive, that means that the\n"
932 "input list is modified to produce the sorted result.\n"
933 "This is a stable sort.")
934 #define FUNC_NAME s_scm_sort_list_x
935 {
936 long len;
937 SCM_VALIDATE_LIST_COPYLEN (1,items,len);
938 SCM_VALIDATE_NIM (2,less);
939 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
940 }
941 #undef FUNC_NAME
942
943 /* stable */
944 SCM_DEFINE (scm_sort_list, "sort-list", 2, 0, 0,
945 (SCM items, SCM less),
946 "Sort the list @var{items}, using @var{less} for comparing the\n"
947 "list elements. This is a stable sort.")
948 #define FUNC_NAME s_scm_sort_list
949 {
950 long len;
951 SCM_VALIDATE_LIST_COPYLEN (1,items,len);
952 SCM_VALIDATE_NIM (2,less);
953 items = scm_list_copy (items);
954 return scm_merge_list_step (&items, scm_cmp_function (less), less, len);
955 }
956 #undef FUNC_NAME
957
958 void
959 scm_init_sort ()
960 {
961 #ifndef SCM_MAGIC_SNARFER
962 #include "libguile/sort.x"
963 #endif
964
965 scm_add_feature ("sort");
966 }
967
968 /*
969 Local Variables:
970 c-file-style: "gnu"
971 End:
972 */