Commit | Line | Data |
---|---|---|
726d810a | 1 | /* Copyright (C) 1999,2000,2001 Free Software Foundation, Inc. |
54e09076 MD |
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 | ||
1bbd0b84 GB |
41 | |
42 | ||
54e09076 MD |
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! | |
15d9c4e3 | 47 | * For scsh compatibility sort-list and sort-list! are also defined. |
54e09076 | 48 | * In cases where a stable-sort is required use stable-sort or |
15d9c4e3 | 49 | * stable-sort!. An additional feature is |
54e09076 | 50 | * (restricted-vector-sort! vector less? startpos endpos) |
15d9c4e3 | 51 | * which allows you to sort part of a vector. |
54e09076 MD |
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. */ | |
a0599745 | 60 | #include "libguile/scmconfig.h" |
54e09076 MD |
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 | ||
783e7774 | 79 | #include <string.h> |
a0599745 | 80 | #include "libguile/_scm.h" |
54e09076 | 81 | |
a0599745 MD |
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" | |
fdc28395 | 87 | #include "libguile/root.h" |
a0599745 | 88 | #include "libguile/vectors.h" |
c96d76b8 | 89 | #include "libguile/lang.h" |
54e09076 | 90 | |
a0599745 MD |
91 | #include "libguile/validate.h" |
92 | #include "libguile/sort.h" | |
54e09076 MD |
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. | |
15d9c4e3 MD |
102 | |
103 | This sorting routine is not much more efficient than the stable | |
104 | version but doesn't consume extra memory. | |
54e09076 MD |
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 | ||
a34af05e MD |
167 | static const char s_buggy_less[] = "buggy less predicate used when sorting"; |
168 | ||
54e09076 MD |
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)) | |
a34af05e MD |
230 | { |
231 | left_ptr += size; | |
232 | /* The comparison predicate may be buggy */ | |
233 | if (left_ptr > hi) | |
5d2d2ffc | 234 | scm_misc_error (NULL, s_buggy_less, SCM_EOL); |
a34af05e | 235 | } |
54e09076 MD |
236 | |
237 | while ((*cmp) (less, (void *) pivot, (void *) right_ptr)) | |
a34af05e MD |
238 | { |
239 | right_ptr -= size; | |
240 | /* The comparison predicate may be buggy */ | |
241 | if (right_ptr < lo) | |
5d2d2ffc | 242 | scm_misc_error (NULL, s_buggy_less, SCM_EOL); |
a34af05e | 243 | } |
54e09076 MD |
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)) | |
a34af05e MD |
322 | { |
323 | tmp_ptr -= size; | |
324 | /* The comparison predicate may be buggy */ | |
325 | if (tmp_ptr < base_ptr) | |
5d2d2ffc | 326 | scm_misc_error (NULL, s_buggy_less, SCM_EOL); |
a34af05e | 327 | } |
54e09076 MD |
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 | { | |
726d810a | 377 | SCM env = SCM_EXTEND_ENV (SCM_CLOSURE_FORMALS (code), |
4c6fe5fc MD |
378 | scm_cons (*(SCM *) a, |
379 | scm_cons (*(SCM *) b, SCM_EOL)), | |
380 | SCM_ENV (code)); | |
54e09076 | 381 | /* Evaluate the closure body */ |
f9450cdb | 382 | return !SCM_FALSEP (scm_eval_body (SCM_CLOSURE_BODY (code), env)); |
54e09076 MD |
383 | } /* closureless */ |
384 | ||
385 | static int | |
386 | applyless (SCM less, const void *a, const void *b) | |
387 | { | |
fdc28395 | 388 | return SCM_NFALSEP (scm_call_2 (less, *(SCM *) a, *(SCM *) b)); |
54e09076 MD |
389 | } /* applyless */ |
390 | ||
391 | static cmp_fun_t | |
392 | scm_cmp_function (SCM p) | |
393 | { | |
394 | switch (SCM_TYP7 (p)) | |
395 | { | |
396 | case scm_tc7_subr_2: | |
397 | case scm_tc7_rpsubr: | |
398 | case scm_tc7_asubr: | |
399 | return subr2less; | |
400 | case scm_tc7_subr_2o: | |
401 | return subr2oless; | |
402 | case scm_tc7_lsubr: | |
403 | return lsubrless; | |
404 | case scm_tcs_closures: | |
405 | return closureless; | |
406 | default: | |
407 | return applyless; | |
408 | } | |
409 | } /* scm_cmp_function */ | |
410 | ||
54e09076 MD |
411 | |
412 | /* Question: Is there any need to make this a more general array sort? | |
413 | It is probably enough to manage the vector type. */ | |
414 | /* endpos equal as for substring, i.e. endpos is not included. */ | |
da4a1dba | 415 | /* More natural with length? */ |
1bbd0b84 | 416 | |
a1ec6916 | 417 | SCM_DEFINE (scm_restricted_vector_sort_x, "restricted-vector-sort!", 4, 0, 0, |
1bbd0b84 | 418 | (SCM vec, SCM less, SCM startpos, SCM endpos), |
e3239868 DH |
419 | "Sort the vector @var{vec}, using @var{less} for comparing\n" |
420 | "the vector elements. @var{startpos} and @var{endpos} delimit\n" | |
421 | "the range of the vector which gets sorted. The return value\n" | |
422 | "is not specified.") | |
1bbd0b84 | 423 | #define FUNC_NAME s_scm_restricted_vector_sort_x |
54e09076 MD |
424 | { |
425 | size_t vlen, spos, len, size = sizeof (SCM); | |
426 | SCM *vp; | |
427 | ||
b5c2579a | 428 | SCM_VALIDATE_VECTOR (1,vec); |
6b5a304f | 429 | SCM_VALIDATE_NIM (2,less); |
b5c2579a | 430 | |
54e09076 | 431 | vp = SCM_VELTS (vec); /* vector pointer */ |
b5c2579a | 432 | vlen = SCM_VECTOR_LENGTH (vec); |
54e09076 | 433 | |
729dbac3 DH |
434 | SCM_VALIDATE_INUM_MIN_COPY (3, startpos, 0, spos); |
435 | SCM_ASSERT_RANGE (3,startpos, spos <= vlen); | |
3b3b36dd | 436 | SCM_VALIDATE_INUM_RANGE (4,endpos,0,vlen+1); |
54e09076 MD |
437 | len = SCM_INUM (endpos) - spos; |
438 | ||
439 | quicksort (&vp[spos], len, size, scm_cmp_function (less), less); | |
440 | return SCM_UNSPECIFIED; | |
441 | /* return vec; */ | |
1bbd0b84 GB |
442 | } |
443 | #undef FUNC_NAME | |
54e09076 MD |
444 | |
445 | /* (sorted? sequence less?) | |
446 | * is true when sequence is a list (x0 x1 ... xm) or a vector #(x0 ... xm) | |
447 | * such that for all 1 <= i <= m, | |
448 | * (not (less? (list-ref list i) (list-ref list (- i 1)))). */ | |
a1ec6916 | 449 | SCM_DEFINE (scm_sorted_p, "sorted?", 2, 0, 0, |
1bbd0b84 | 450 | (SCM items, SCM less), |
e3239868 DH |
451 | "Return @code{#t} iff @var{items} is a list or a vector such that\n" |
452 | "for all 1 <= i <= m, the predicate @var{less} returns true when\n" | |
453 | "applied to all elements i - 1 and i") | |
1bbd0b84 | 454 | #define FUNC_NAME s_scm_sorted_p |
54e09076 | 455 | { |
c014a02e | 456 | long len, j; /* list/vector length, temp j */ |
54e09076 MD |
457 | SCM item, rest; /* rest of items loop variable */ |
458 | SCM *vp; | |
459 | cmp_fun_t cmp = scm_cmp_function (less); | |
460 | ||
c96d76b8 | 461 | if (SCM_NULL_OR_NIL_P (items)) |
54e09076 | 462 | return SCM_BOOL_T; |
1bbd0b84 | 463 | |
6b5a304f | 464 | SCM_VALIDATE_NIM (2,less); |
54e09076 MD |
465 | |
466 | if (SCM_CONSP (items)) | |
467 | { | |
468 | len = scm_ilength (items); /* also checks that it's a pure list */ | |
1bbd0b84 | 469 | SCM_ASSERT_RANGE (1,items,len >= 0); |
54e09076 MD |
470 | if (len <= 1) |
471 | return SCM_BOOL_T; | |
472 | ||
473 | item = SCM_CAR (items); | |
474 | rest = SCM_CDR (items); | |
475 | j = len - 1; | |
476 | while (j > 0) | |
477 | { | |
4b479d98 | 478 | if ((*cmp) (less, SCM_CARLOC(rest), &item)) |
54e09076 MD |
479 | return SCM_BOOL_F; |
480 | else | |
481 | { | |
482 | item = SCM_CAR (rest); | |
483 | rest = SCM_CDR (rest); | |
484 | j--; | |
485 | } | |
486 | } | |
487 | return SCM_BOOL_T; | |
488 | } | |
489 | else | |
490 | { | |
b5c2579a DH |
491 | SCM_VALIDATE_VECTOR (1, items); |
492 | ||
493 | vp = SCM_VELTS (items); /* vector pointer */ | |
494 | len = SCM_VECTOR_LENGTH (items); | |
495 | j = len - 1; | |
496 | while (j > 0) | |
54e09076 | 497 | { |
b5c2579a DH |
498 | if ((*cmp) (less, &vp[1], vp)) |
499 | return SCM_BOOL_F; | |
500 | else | |
501 | { | |
502 | vp++; | |
503 | j--; | |
504 | } | |
54e09076 | 505 | } |
b5c2579a | 506 | return SCM_BOOL_T; |
54e09076 | 507 | } |
b5c2579a | 508 | |
54e09076 | 509 | return SCM_BOOL_F; |
1bbd0b84 GB |
510 | } |
511 | #undef FUNC_NAME | |
54e09076 MD |
512 | |
513 | /* (merge a b less?) | |
514 | takes two lists a and b such that (sorted? a less?) and (sorted? b less?) | |
515 | and returns a new list in which the elements of a and b have been stably | |
516 | interleaved so that (sorted? (merge a b less?) less?). | |
517 | Note: this does _not_ accept vectors. */ | |
a1ec6916 | 518 | SCM_DEFINE (scm_merge, "merge", 3, 0, 0, |
1bbd0b84 | 519 | (SCM alist, SCM blist, SCM less), |
8f85c0c6 NJ |
520 | "Merge two already sorted lists into one.\n" |
521 | "Given two lists @var{alist} and @var{blist}, such that\n" | |
522 | "@code{(sorted? alist less?)} and @code{(sorted? blist less?)},\n" | |
523 | "return a new list in which the elements of @var{alist} and\n" | |
e3239868 DH |
524 | "@var{blist} have been stably interleaved so that\n" |
525 | "@code{(sorted? (merge alist blist less?) less?)}.\n" | |
526 | "Note: this does _not_ accept vectors.") | |
1bbd0b84 | 527 | #define FUNC_NAME s_scm_merge |
54e09076 | 528 | { |
c014a02e | 529 | long alen, blen; /* list lengths */ |
54e09076 MD |
530 | SCM build, last; |
531 | cmp_fun_t cmp = scm_cmp_function (less); | |
6b5a304f | 532 | SCM_VALIDATE_NIM (3,less); |
54e09076 | 533 | |
c96d76b8 | 534 | if (SCM_NULL_OR_NIL_P (alist)) |
54e09076 | 535 | return blist; |
c96d76b8 | 536 | else if (SCM_NULL_OR_NIL_P (blist)) |
54e09076 MD |
537 | return alist; |
538 | else | |
539 | { | |
3b3b36dd GB |
540 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1,alist,alen); |
541 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2,blist,blen); | |
4b479d98 | 542 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
543 | { |
544 | build = scm_cons (SCM_CAR (blist), SCM_EOL); | |
545 | blist = SCM_CDR (blist); | |
546 | blen--; | |
547 | } | |
c56cc3c8 MD |
548 | else |
549 | { | |
550 | build = scm_cons (SCM_CAR (alist), SCM_EOL); | |
551 | alist = SCM_CDR (alist); | |
552 | alen--; | |
553 | } | |
54e09076 MD |
554 | last = build; |
555 | while ((alen > 0) && (blen > 0)) | |
556 | { | |
4b479d98 | 557 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
558 | { |
559 | SCM_SETCDR (last, scm_cons (SCM_CAR (blist), SCM_EOL)); | |
560 | blist = SCM_CDR (blist); | |
561 | blen--; | |
562 | } | |
c56cc3c8 MD |
563 | else |
564 | { | |
565 | SCM_SETCDR (last, scm_cons (SCM_CAR (alist), SCM_EOL)); | |
566 | alist = SCM_CDR (alist); | |
567 | alen--; | |
568 | } | |
54e09076 MD |
569 | last = SCM_CDR (last); |
570 | } | |
571 | if ((alen > 0) && (blen == 0)) | |
572 | SCM_SETCDR (last, alist); | |
573 | else if ((alen == 0) && (blen > 0)) | |
574 | SCM_SETCDR (last, blist); | |
575 | } | |
576 | return build; | |
1bbd0b84 GB |
577 | } |
578 | #undef FUNC_NAME | |
579 | ||
54e09076 MD |
580 | |
581 | static SCM | |
582 | scm_merge_list_x (SCM alist, SCM blist, | |
583 | long alen, long blen, | |
584 | cmp_fun_t cmp, SCM less) | |
585 | { | |
586 | SCM build, last; | |
587 | ||
c96d76b8 | 588 | if (SCM_NULL_OR_NIL_P (alist)) |
54e09076 | 589 | return blist; |
c96d76b8 | 590 | else if (SCM_NULL_OR_NIL_P (blist)) |
54e09076 MD |
591 | return alist; |
592 | else | |
593 | { | |
4b479d98 | 594 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
595 | { |
596 | build = blist; | |
597 | blist = SCM_CDR (blist); | |
598 | blen--; | |
599 | } | |
c56cc3c8 MD |
600 | else |
601 | { | |
602 | build = alist; | |
603 | alist = SCM_CDR (alist); | |
604 | alen--; | |
605 | } | |
54e09076 MD |
606 | last = build; |
607 | while ((alen > 0) && (blen > 0)) | |
608 | { | |
4b479d98 | 609 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
610 | { |
611 | SCM_SETCDR (last, blist); | |
612 | blist = SCM_CDR (blist); | |
613 | blen--; | |
614 | } | |
c56cc3c8 MD |
615 | else |
616 | { | |
617 | SCM_SETCDR (last, alist); | |
618 | alist = SCM_CDR (alist); | |
619 | alen--; | |
620 | } | |
54e09076 MD |
621 | last = SCM_CDR (last); |
622 | } | |
623 | if ((alen > 0) && (blen == 0)) | |
624 | SCM_SETCDR (last, alist); | |
625 | else if ((alen == 0) && (blen > 0)) | |
626 | SCM_SETCDR (last, blist); | |
627 | } | |
628 | return build; | |
629 | } /* scm_merge_list_x */ | |
630 | ||
a1ec6916 | 631 | SCM_DEFINE (scm_merge_x, "merge!", 3, 0, 0, |
1bbd0b84 | 632 | (SCM alist, SCM blist, SCM less), |
e3239868 DH |
633 | "Takes two lists @var{alist} and @var{blist} such that\n" |
634 | "@code{(sorted? alist less?)} and @code{(sorted? blist less?)} and\n" | |
635 | "returns a new list in which the elements of @var{alist} and\n" | |
636 | "@var{blist} have been stably interleaved so that\n" | |
637 | " @code{(sorted? (merge alist blist less?) less?)}.\n" | |
638 | "This is the destructive variant of @code{merge}\n" | |
639 | "Note: this does _not_ accept vectors.") | |
1bbd0b84 | 640 | #define FUNC_NAME s_scm_merge_x |
54e09076 | 641 | { |
c014a02e | 642 | long alen, blen; /* list lengths */ |
54e09076 | 643 | |
6b5a304f | 644 | SCM_VALIDATE_NIM (3,less); |
c96d76b8 | 645 | if (SCM_NULL_OR_NIL_P (alist)) |
54e09076 | 646 | return blist; |
c96d76b8 | 647 | else if (SCM_NULL_OR_NIL_P (blist)) |
54e09076 MD |
648 | return alist; |
649 | else | |
650 | { | |
3b3b36dd GB |
651 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1,alist,alen); |
652 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2,blist,blen); | |
54e09076 MD |
653 | return scm_merge_list_x (alist, blist, |
654 | alen, blen, | |
655 | scm_cmp_function (less), | |
656 | less); | |
657 | } | |
1bbd0b84 GB |
658 | } |
659 | #undef FUNC_NAME | |
54e09076 MD |
660 | |
661 | /* This merge sort algorithm is same as slib's by Richard A. O'Keefe. | |
662 | The algorithm is stable. We also tried to use the algorithm used by | |
663 | scsh's merge-sort but that algorithm showed to not be stable, even | |
664 | though it claimed to be. | |
665 | */ | |
666 | static SCM | |
667 | scm_merge_list_step (SCM * seq, | |
668 | cmp_fun_t cmp, | |
669 | SCM less, | |
c014a02e | 670 | long n) |
54e09076 | 671 | { |
c56cc3c8 MD |
672 | SCM a, b; |
673 | ||
54e09076 MD |
674 | if (n > 2) |
675 | { | |
c014a02e | 676 | long mid = n / 2; |
c56cc3c8 MD |
677 | a = scm_merge_list_step (seq, cmp, less, mid); |
678 | b = scm_merge_list_step (seq, cmp, less, n - mid); | |
679 | return scm_merge_list_x (a, b, mid, n - mid, cmp, less); | |
54e09076 MD |
680 | } |
681 | else if (n == 2) | |
682 | { | |
683 | SCM p = *seq; | |
684 | SCM rest = SCM_CDR (*seq); | |
685 | SCM x = SCM_CAR (*seq); | |
686 | SCM y = SCM_CAR (SCM_CDR (*seq)); | |
687 | *seq = SCM_CDR (rest); | |
688 | SCM_SETCDR (rest, SCM_EOL); | |
689 | if ((*cmp) (less, &y, &x)) | |
690 | { | |
4b479d98 DH |
691 | SCM_SETCAR (p, y); |
692 | SCM_SETCAR (rest, x); | |
54e09076 MD |
693 | } |
694 | return p; | |
695 | } | |
696 | else if (n == 1) | |
697 | { | |
698 | SCM p = *seq; | |
699 | *seq = SCM_CDR (p); | |
700 | SCM_SETCDR (p, SCM_EOL); | |
701 | return p; | |
702 | } | |
703 | else | |
704 | return SCM_EOL; | |
705 | } /* scm_merge_list_step */ | |
706 | ||
707 | ||
54e09076 | 708 | /* scm_sort_x manages lists and vectors, not stable sort */ |
a1ec6916 | 709 | SCM_DEFINE (scm_sort_x, "sort!", 2, 0, 0, |
1bbd0b84 | 710 | (SCM items, SCM less), |
e3239868 DH |
711 | "Sort the sequence @var{items}, which may be a list or a\n" |
712 | "vector. @var{less} is used for comparing the sequence\n" | |
713 | "elements. The sorting is destructive, that means that the\n" | |
714 | "input sequence is modified to produce the sorted result.\n" | |
715 | "This is not a stable sort.") | |
1bbd0b84 | 716 | #define FUNC_NAME s_scm_sort_x |
54e09076 | 717 | { |
c014a02e | 718 | long len; /* list/vector length */ |
c96d76b8 NJ |
719 | if (SCM_NULL_OR_NIL_P (items)) |
720 | return items; | |
b5c2579a | 721 | |
6b5a304f | 722 | SCM_VALIDATE_NIM (2,less); |
54e09076 MD |
723 | |
724 | if (SCM_CONSP (items)) | |
725 | { | |
3b3b36dd | 726 | SCM_VALIDATE_LIST_COPYLEN (1,items,len); |
54e09076 MD |
727 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); |
728 | } | |
729 | else if (SCM_VECTORP (items)) | |
730 | { | |
b5c2579a | 731 | len = SCM_VECTOR_LENGTH (items); |
54e09076 MD |
732 | scm_restricted_vector_sort_x (items, |
733 | less, | |
734 | SCM_MAKINUM (0L), | |
735 | SCM_MAKINUM (len)); | |
736 | return items; | |
737 | } | |
738 | else | |
276dd677 | 739 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 740 | } |
0f981281 | 741 | #undef FUNC_NAME |
54e09076 MD |
742 | |
743 | /* scm_sort manages lists and vectors, not stable sort */ | |
1bbd0b84 | 744 | |
a1ec6916 | 745 | SCM_DEFINE (scm_sort, "sort", 2, 0, 0, |
1bbd0b84 | 746 | (SCM items, SCM less), |
e3239868 DH |
747 | "Sort the sequence @var{items}, which may be a list or a\n" |
748 | "vector. @var{less} is used for comparing the sequence\n" | |
749 | "elements. This is not a stable sort.") | |
1bbd0b84 | 750 | #define FUNC_NAME s_scm_sort |
54e09076 | 751 | { |
c96d76b8 NJ |
752 | if (SCM_NULL_OR_NIL_P (items)) |
753 | return items; | |
b5c2579a | 754 | |
6b5a304f | 755 | SCM_VALIDATE_NIM (2,less); |
54e09076 MD |
756 | if (SCM_CONSP (items)) |
757 | { | |
c014a02e | 758 | long len; |
e9e225e5 | 759 | |
3b3b36dd | 760 | SCM_VALIDATE_LIST_COPYLEN (1,items,len); |
54e09076 MD |
761 | items = scm_list_copy (items); |
762 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); | |
763 | } | |
afe5177e GH |
764 | #ifdef HAVE_ARRAYS |
765 | /* support ordinary vectors even if arrays not available? */ | |
54e09076 MD |
766 | else if (SCM_VECTORP (items)) |
767 | { | |
c014a02e | 768 | long len = SCM_VECTOR_LENGTH (items); |
e9e225e5 GH |
769 | SCM sortvec = scm_make_uve (len, scm_array_prototype (items)); |
770 | ||
54e09076 MD |
771 | scm_array_copy_x (items, sortvec); |
772 | scm_restricted_vector_sort_x (sortvec, | |
773 | less, | |
774 | SCM_MAKINUM (0L), | |
775 | SCM_MAKINUM (len)); | |
776 | return sortvec; | |
777 | } | |
afe5177e | 778 | #endif |
54e09076 | 779 | else |
276dd677 | 780 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 781 | } |
0f981281 | 782 | #undef FUNC_NAME |
54e09076 MD |
783 | |
784 | static void | |
785 | scm_merge_vector_x (void *const vecbase, | |
786 | void *const tempbase, | |
787 | cmp_fun_t cmp, | |
788 | SCM less, | |
c014a02e ML |
789 | long low, |
790 | long mid, | |
791 | long high) | |
54e09076 MD |
792 | { |
793 | register SCM *vp = (SCM *) vecbase; | |
794 | register SCM *temp = (SCM *) tempbase; | |
c014a02e ML |
795 | long it; /* Index for temp vector */ |
796 | long i1 = low; /* Index for lower vector segment */ | |
797 | long i2 = mid + 1; /* Index for upper vector segment */ | |
54e09076 MD |
798 | |
799 | /* Copy while both segments contain more characters */ | |
800 | for (it = low; (i1 <= mid) && (i2 <= high); ++it) | |
801 | if ((*cmp) (less, &vp[i2], &vp[i1])) | |
802 | temp[it] = vp[i2++]; | |
803 | else | |
804 | temp[it] = vp[i1++]; | |
805 | ||
806 | /* Copy while first segment contains more characters */ | |
807 | while (i1 <= mid) | |
808 | temp[it++] = vp[i1++]; | |
809 | ||
810 | /* Copy while second segment contains more characters */ | |
811 | while (i2 <= high) | |
812 | temp[it++] = vp[i2++]; | |
813 | ||
814 | /* Copy back from temp to vp */ | |
815 | for (it = low; it <= high; ++it) | |
816 | vp[it] = temp[it]; | |
817 | } /* scm_merge_vector_x */ | |
818 | ||
819 | static void | |
820 | scm_merge_vector_step (void *const vp, | |
821 | void *const temp, | |
822 | cmp_fun_t cmp, | |
823 | SCM less, | |
c014a02e ML |
824 | long low, |
825 | long high) | |
54e09076 MD |
826 | { |
827 | if (high > low) | |
828 | { | |
c014a02e | 829 | long mid = (low + high) / 2; |
54e09076 MD |
830 | scm_merge_vector_step (vp, temp, cmp, less, low, mid); |
831 | scm_merge_vector_step (vp, temp, cmp, less, mid+1, high); | |
832 | scm_merge_vector_x (vp, temp, cmp, less, low, mid, high); | |
833 | } | |
834 | } /* scm_merge_vector_step */ | |
835 | ||
836 | ||
54e09076 MD |
837 | /* stable-sort! manages lists and vectors */ |
838 | ||
a1ec6916 | 839 | SCM_DEFINE (scm_stable_sort_x, "stable-sort!", 2, 0, 0, |
1bbd0b84 | 840 | (SCM items, SCM less), |
e3239868 DH |
841 | "Sort the sequence @var{items}, which may be a list or a\n" |
842 | "vector. @var{less} is used for comparing the sequence elements.\n" | |
843 | "The sorting is destructive, that means that the input sequence\n" | |
844 | "is modified to produce the sorted result.\n" | |
845 | "This is a stable sort.") | |
1bbd0b84 | 846 | #define FUNC_NAME s_scm_stable_sort_x |
54e09076 | 847 | { |
c014a02e | 848 | long len; /* list/vector length */ |
54e09076 | 849 | |
c96d76b8 NJ |
850 | if (SCM_NULL_OR_NIL_P (items)) |
851 | return items; | |
b5c2579a | 852 | |
6b5a304f | 853 | SCM_VALIDATE_NIM (2,less); |
54e09076 MD |
854 | if (SCM_CONSP (items)) |
855 | { | |
3b3b36dd | 856 | SCM_VALIDATE_LIST_COPYLEN (1,items,len); |
54e09076 MD |
857 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); |
858 | } | |
859 | else if (SCM_VECTORP (items)) | |
860 | { | |
861 | SCM *temp, *vp; | |
b5c2579a | 862 | len = SCM_VECTOR_LENGTH (items); |
54e09076 MD |
863 | temp = malloc (len * sizeof(SCM)); |
864 | vp = SCM_VELTS (items); | |
865 | scm_merge_vector_step (vp, | |
866 | temp, | |
867 | scm_cmp_function (less), | |
868 | less, | |
869 | 0, | |
870 | len - 1); | |
871 | free(temp); | |
872 | return items; | |
873 | } | |
874 | else | |
276dd677 | 875 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 876 | } |
0f981281 | 877 | #undef FUNC_NAME |
54e09076 MD |
878 | |
879 | /* stable_sort manages lists and vectors */ | |
1bbd0b84 | 880 | |
a1ec6916 | 881 | SCM_DEFINE (scm_stable_sort, "stable-sort", 2, 0, 0, |
1bbd0b84 | 882 | (SCM items, SCM less), |
e3239868 DH |
883 | "Sort the sequence @var{items}, which may be a list or a\n" |
884 | "vector. @var{less} is used for comparing the sequence elements.\n" | |
885 | "This is a stable sort.") | |
1bbd0b84 | 886 | #define FUNC_NAME s_scm_stable_sort |
54e09076 | 887 | { |
c014a02e | 888 | long len; /* list/vector length */ |
c96d76b8 NJ |
889 | if (SCM_NULL_OR_NIL_P (items)) |
890 | return items; | |
b5c2579a | 891 | |
6b5a304f | 892 | SCM_VALIDATE_NIM (2,less); |
54e09076 MD |
893 | if (SCM_CONSP (items)) |
894 | { | |
3b3b36dd | 895 | SCM_VALIDATE_LIST_COPYLEN (1,items,len); |
54e09076 MD |
896 | items = scm_list_copy (items); |
897 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); | |
898 | } | |
afe5177e GH |
899 | #ifdef HAVE_ARRAYS |
900 | /* support ordinary vectors even if arrays not available? */ | |
54e09076 MD |
901 | else if (SCM_VECTORP (items)) |
902 | { | |
903 | SCM retvec; | |
904 | SCM *temp, *vp; | |
b5c2579a | 905 | len = SCM_VECTOR_LENGTH (items); |
54e09076 MD |
906 | retvec = scm_make_uve (len, scm_array_prototype (items)); |
907 | scm_array_copy_x (items, retvec); | |
908 | temp = malloc (len * sizeof (SCM)); | |
909 | vp = SCM_VELTS (retvec); | |
910 | scm_merge_vector_step (vp, | |
911 | temp, | |
912 | scm_cmp_function (less), | |
913 | less, | |
914 | 0, | |
915 | len - 1); | |
916 | free (temp); | |
917 | return retvec; | |
918 | } | |
afe5177e | 919 | #endif |
54e09076 | 920 | else |
276dd677 | 921 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 922 | } |
0f981281 | 923 | #undef FUNC_NAME |
54e09076 | 924 | |
1bbd0b84 | 925 | /* stable */ |
a1ec6916 | 926 | SCM_DEFINE (scm_sort_list_x, "sort-list!", 2, 0, 0, |
1bbd0b84 | 927 | (SCM items, SCM less), |
e3239868 DH |
928 | "Sort the list @var{items}, using @var{less} for comparing the\n" |
929 | "list elements. The sorting is destructive, that means that the\n" | |
930 | "input list is modified to produce the sorted result.\n" | |
931 | "This is a stable sort.") | |
1bbd0b84 | 932 | #define FUNC_NAME s_scm_sort_list_x |
54e09076 | 933 | { |
c014a02e | 934 | long len; |
3b3b36dd | 935 | SCM_VALIDATE_LIST_COPYLEN (1,items,len); |
6b5a304f | 936 | SCM_VALIDATE_NIM (2,less); |
54e09076 | 937 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); |
1bbd0b84 | 938 | } |
0f981281 | 939 | #undef FUNC_NAME |
54e09076 | 940 | |
1bbd0b84 | 941 | /* stable */ |
a1ec6916 | 942 | SCM_DEFINE (scm_sort_list, "sort-list", 2, 0, 0, |
e3239868 DH |
943 | (SCM items, SCM less), |
944 | "Sort the list @var{items}, using @var{less} for comparing the\n" | |
945 | "list elements. This is a stable sort.") | |
1bbd0b84 | 946 | #define FUNC_NAME s_scm_sort_list |
54e09076 | 947 | { |
c014a02e | 948 | long len; |
3b3b36dd | 949 | SCM_VALIDATE_LIST_COPYLEN (1,items,len); |
6b5a304f | 950 | SCM_VALIDATE_NIM (2,less); |
54e09076 MD |
951 | items = scm_list_copy (items); |
952 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); | |
1bbd0b84 | 953 | } |
0f981281 | 954 | #undef FUNC_NAME |
54e09076 MD |
955 | |
956 | void | |
957 | scm_init_sort () | |
958 | { | |
a0599745 | 959 | #include "libguile/sort.x" |
54e09076 MD |
960 | |
961 | scm_add_feature ("sort"); | |
962 | } | |
89e00824 ML |
963 | |
964 | /* | |
965 | Local Variables: | |
966 | c-file-style: "gnu" | |
967 | End: | |
968 | */ |