Commit | Line | Data |
---|---|---|
504d99c5 | 1 | /* Copyright (C) 1999,2000,2001,2002 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. */ | |
af7a3945 RB |
60 | #if HAVE_CONFIG_H |
61 | # include <config.h> | |
62 | #endif | |
63 | ||
64 | /* do we still need this here? */ | |
a0599745 | 65 | #include "libguile/scmconfig.h" |
54e09076 MD |
66 | |
67 | /* AIX requires this to be the first thing in the file. The #pragma | |
68 | directive is indented so pre-ANSI compilers will ignore it, rather | |
69 | than choke on it. */ | |
70 | #ifndef __GNUC__ | |
71 | # if HAVE_ALLOCA_H | |
72 | # include <alloca.h> | |
73 | # else | |
74 | # ifdef _AIX | |
75 | #pragma alloca | |
76 | # else | |
77 | # ifndef alloca /* predefined by HP cc +Olibcalls */ | |
78 | char *alloca (); | |
79 | # endif | |
80 | # endif | |
81 | # endif | |
82 | #endif | |
83 | ||
783e7774 | 84 | #include <string.h> |
a0599745 | 85 | #include "libguile/_scm.h" |
54e09076 | 86 | |
a0599745 MD |
87 | #include "libguile/eval.h" |
88 | #include "libguile/unif.h" | |
89 | #include "libguile/ramap.h" | |
90 | #include "libguile/alist.h" | |
91 | #include "libguile/feature.h" | |
fdc28395 | 92 | #include "libguile/root.h" |
a0599745 | 93 | #include "libguile/vectors.h" |
c96d76b8 | 94 | #include "libguile/lang.h" |
54e09076 | 95 | |
a0599745 MD |
96 | #include "libguile/validate.h" |
97 | #include "libguile/sort.h" | |
54e09076 MD |
98 | |
99 | /* The routine quicksort was extracted from the GNU C Library qsort.c | |
100 | written by Douglas C. Schmidt (schmidt@ics.uci.edu) | |
101 | and adapted to guile by adding an extra pointer less | |
102 | to quicksort by Roland Orre <orre@nada.kth.se>. | |
103 | ||
104 | The reason to do this instead of using the library function qsort | |
105 | was to avoid dependency of the ANSI-C extensions for local functions | |
106 | and also to avoid obscure pool based solutions. | |
15d9c4e3 MD |
107 | |
108 | This sorting routine is not much more efficient than the stable | |
109 | version but doesn't consume extra memory. | |
54e09076 MD |
110 | */ |
111 | ||
112 | /* Byte-wise swap two items of size SIZE. */ | |
113 | #define SWAP(a, b, size) \ | |
114 | do \ | |
115 | { \ | |
116 | register size_t __size = (size); \ | |
117 | register char *__a = (a), *__b = (b); \ | |
118 | do \ | |
119 | { \ | |
120 | char __tmp = *__a; \ | |
121 | *__a++ = *__b; \ | |
122 | *__b++ = __tmp; \ | |
123 | } while (--__size > 0); \ | |
124 | } while (0) | |
125 | ||
126 | /* Discontinue quicksort algorithm when partition gets below this size. | |
127 | This particular magic number was chosen to work best on a Sun 4/260. */ | |
128 | #define MAX_THRESH 4 | |
129 | ||
130 | /* Stack node declarations used to store unfulfilled partition obligations. */ | |
131 | typedef struct | |
132 | { | |
133 | char *lo; | |
134 | char *hi; | |
135 | } | |
136 | stack_node; | |
137 | ||
138 | /* The next 4 #defines implement a very fast in-line stack abstraction. */ | |
139 | #define STACK_SIZE (8 * sizeof(unsigned long int)) | |
140 | #define PUSH(low, high) ((void) ((top->lo = (low)), (top->hi = (high)), ++top)) | |
141 | #define POP(low, high) ((void) (--top, (low = top->lo), (high = top->hi))) | |
142 | #define STACK_NOT_EMPTY (stack < top) | |
143 | ||
144 | ||
145 | /* Order size using quicksort. This implementation incorporates | |
146 | four optimizations discussed in Sedgewick: | |
147 | ||
148 | 1. Non-recursive, using an explicit stack of pointer that store the | |
149 | next array partition to sort. To save time, this maximum amount | |
150 | of space required to store an array of MAX_INT is allocated on the | |
151 | stack. Assuming a 32-bit integer, this needs only 32 * | |
152 | sizeof(stack_node) == 136 bits. Pretty cheap, actually. | |
153 | ||
154 | 2. Chose the pivot element using a median-of-three decision tree. | |
155 | This reduces the probability of selecting a bad pivot value and | |
156 | eliminates certain extraneous comparisons. | |
157 | ||
158 | 3. Only quicksorts TOTAL_ELEMS / MAX_THRESH partitions, leaving | |
159 | insertion sort to order the MAX_THRESH items within each partition. | |
160 | This is a big win, since insertion sort is faster for small, mostly | |
161 | sorted array segments. | |
162 | ||
163 | 4. The larger of the two sub-partitions is always pushed onto the | |
164 | stack first, with the algorithm then concentrating on the | |
165 | smaller partition. This *guarantees* no more than log (n) | |
166 | stack size is needed (actually O(1) in this case)! */ | |
167 | ||
168 | typedef int (*cmp_fun_t) (SCM less, | |
169 | const void*, | |
170 | const void*); | |
171 | ||
a34af05e MD |
172 | static const char s_buggy_less[] = "buggy less predicate used when sorting"; |
173 | ||
54e09076 MD |
174 | static void |
175 | quicksort (void *const pbase, | |
176 | size_t total_elems, | |
177 | size_t size, | |
178 | cmp_fun_t cmp, | |
179 | SCM less) | |
180 | { | |
181 | register char *base_ptr = (char *) pbase; | |
182 | ||
183 | /* Allocating SIZE bytes for a pivot buffer facilitates a better | |
184 | algorithm below since we can do comparisons directly on the pivot. */ | |
185 | char *pivot_buffer = (char *) alloca (size); | |
186 | const size_t max_thresh = MAX_THRESH * size; | |
187 | ||
188 | if (total_elems == 0) | |
189 | /* Avoid lossage with unsigned arithmetic below. */ | |
190 | return; | |
191 | ||
192 | if (total_elems > MAX_THRESH) | |
193 | { | |
194 | char *lo = base_ptr; | |
195 | char *hi = &lo[size * (total_elems - 1)]; | |
196 | /* Largest size needed for 32-bit int!!! */ | |
197 | stack_node stack[STACK_SIZE]; | |
198 | stack_node *top = stack + 1; | |
199 | ||
200 | while (STACK_NOT_EMPTY) | |
201 | { | |
202 | char *left_ptr; | |
203 | char *right_ptr; | |
204 | ||
205 | char *pivot = pivot_buffer; | |
206 | ||
207 | /* Select median value from among LO, MID, and HI. Rearrange | |
208 | LO and HI so the three values are sorted. This lowers the | |
209 | probability of picking a pathological pivot value and | |
210 | skips a comparison for both the LEFT_PTR and RIGHT_PTR. */ | |
211 | ||
212 | char *mid = lo + size * ((hi - lo) / size >> 1); | |
213 | ||
214 | if ((*cmp) (less, (void *) mid, (void *) lo)) | |
215 | SWAP (mid, lo, size); | |
216 | if ((*cmp) (less, (void *) hi, (void *) mid)) | |
217 | SWAP (mid, hi, size); | |
218 | else | |
219 | goto jump_over; | |
220 | if ((*cmp) (less, (void *) mid, (void *) lo)) | |
221 | SWAP (mid, lo, size); | |
222 | jump_over:; | |
223 | memcpy (pivot, mid, size); | |
224 | pivot = pivot_buffer; | |
225 | ||
226 | left_ptr = lo + size; | |
227 | right_ptr = hi - size; | |
228 | ||
229 | /* Here's the famous ``collapse the walls'' section of quicksort. | |
230 | Gotta like those tight inner loops! They are the main reason | |
231 | that this algorithm runs much faster than others. */ | |
232 | do | |
233 | { | |
234 | while ((*cmp) (less, (void *) left_ptr, (void *) pivot)) | |
a34af05e MD |
235 | { |
236 | left_ptr += size; | |
237 | /* The comparison predicate may be buggy */ | |
238 | if (left_ptr > hi) | |
5d2d2ffc | 239 | scm_misc_error (NULL, s_buggy_less, SCM_EOL); |
a34af05e | 240 | } |
54e09076 MD |
241 | |
242 | while ((*cmp) (less, (void *) pivot, (void *) right_ptr)) | |
a34af05e MD |
243 | { |
244 | right_ptr -= size; | |
245 | /* The comparison predicate may be buggy */ | |
246 | if (right_ptr < lo) | |
5d2d2ffc | 247 | scm_misc_error (NULL, s_buggy_less, SCM_EOL); |
a34af05e | 248 | } |
54e09076 MD |
249 | |
250 | if (left_ptr < right_ptr) | |
251 | { | |
252 | SWAP (left_ptr, right_ptr, size); | |
253 | left_ptr += size; | |
254 | right_ptr -= size; | |
255 | } | |
256 | else if (left_ptr == right_ptr) | |
257 | { | |
258 | left_ptr += size; | |
259 | right_ptr -= size; | |
260 | break; | |
261 | } | |
262 | } | |
263 | while (left_ptr <= right_ptr); | |
264 | ||
265 | /* Set up pointers for next iteration. First determine whether | |
266 | left and right partitions are below the threshold size. If so, | |
267 | ignore one or both. Otherwise, push the larger partition's | |
268 | bounds on the stack and continue sorting the smaller one. */ | |
269 | ||
270 | if ((size_t) (right_ptr - lo) <= max_thresh) | |
271 | { | |
272 | if ((size_t) (hi - left_ptr) <= max_thresh) | |
273 | /* Ignore both small partitions. */ | |
274 | POP (lo, hi); | |
275 | else | |
276 | /* Ignore small left partition. */ | |
277 | lo = left_ptr; | |
278 | } | |
279 | else if ((size_t) (hi - left_ptr) <= max_thresh) | |
280 | /* Ignore small right partition. */ | |
281 | hi = right_ptr; | |
282 | else if ((right_ptr - lo) > (hi - left_ptr)) | |
283 | { | |
284 | /* Push larger left partition indices. */ | |
285 | PUSH (lo, right_ptr); | |
286 | lo = left_ptr; | |
287 | } | |
288 | else | |
289 | { | |
290 | /* Push larger right partition indices. */ | |
291 | PUSH (left_ptr, hi); | |
292 | hi = right_ptr; | |
293 | } | |
294 | } | |
295 | } | |
296 | ||
297 | /* Once the BASE_PTR array is partially sorted by quicksort the rest | |
298 | is completely sorted using insertion sort, since this is efficient | |
299 | for partitions below MAX_THRESH size. BASE_PTR points to the beginning | |
300 | of the array to sort, and END_PTR points at the very last element in | |
301 | the array (*not* one beyond it!). */ | |
302 | ||
303 | { | |
304 | char *const end_ptr = &base_ptr[size * (total_elems - 1)]; | |
305 | char *tmp_ptr = base_ptr; | |
306 | char *thresh = min (end_ptr, base_ptr + max_thresh); | |
307 | register char *run_ptr; | |
308 | ||
309 | /* Find smallest element in first threshold and place it at the | |
310 | array's beginning. This is the smallest array element, | |
311 | and the operation speeds up insertion sort's inner loop. */ | |
312 | ||
313 | for (run_ptr = tmp_ptr + size; run_ptr <= thresh; run_ptr += size) | |
314 | if ((*cmp) (less, (void *) run_ptr, (void *) tmp_ptr)) | |
315 | tmp_ptr = run_ptr; | |
316 | ||
317 | if (tmp_ptr != base_ptr) | |
318 | SWAP (tmp_ptr, base_ptr, size); | |
319 | ||
320 | /* Insertion sort, running from left-hand-side up to right-hand-side. */ | |
321 | ||
322 | run_ptr = base_ptr + size; | |
323 | while ((run_ptr += size) <= end_ptr) | |
324 | { | |
325 | tmp_ptr = run_ptr - size; | |
326 | while ((*cmp) (less, (void *) run_ptr, (void *) tmp_ptr)) | |
a34af05e MD |
327 | { |
328 | tmp_ptr -= size; | |
329 | /* The comparison predicate may be buggy */ | |
330 | if (tmp_ptr < base_ptr) | |
5d2d2ffc | 331 | scm_misc_error (NULL, s_buggy_less, SCM_EOL); |
a34af05e | 332 | } |
54e09076 MD |
333 | |
334 | tmp_ptr += size; | |
335 | if (tmp_ptr != run_ptr) | |
336 | { | |
337 | char *trav; | |
338 | ||
339 | trav = run_ptr + size; | |
340 | while (--trav >= run_ptr) | |
341 | { | |
342 | char c = *trav; | |
343 | char *hi, *lo; | |
344 | ||
345 | for (hi = lo = trav; (lo -= size) >= tmp_ptr; hi = lo) | |
346 | *hi = *lo; | |
347 | *hi = c; | |
348 | } | |
349 | } | |
350 | } | |
351 | } | |
352 | } /* quicksort */ | |
353 | ||
354 | ||
355 | /* comparison routines */ | |
356 | ||
357 | static int | |
358 | subr2less (SCM less, const void *a, const void *b) | |
359 | { | |
360 | return SCM_NFALSEP (SCM_SUBRF (less) (*(SCM *) a, *(SCM *) b)); | |
361 | } /* subr2less */ | |
362 | ||
54e09076 MD |
363 | static int |
364 | lsubrless (SCM less, const void *a, const void *b) | |
365 | { | |
366 | return SCM_NFALSEP (SCM_SUBRF (less) | |
367 | (scm_cons (*(SCM *) a, | |
368 | scm_cons (*(SCM *) b, SCM_EOL)))); | |
369 | } /* lsubrless */ | |
370 | ||
371 | static int | |
372 | closureless (SCM code, const void *a, const void *b) | |
373 | { | |
726d810a | 374 | SCM env = SCM_EXTEND_ENV (SCM_CLOSURE_FORMALS (code), |
4c6fe5fc MD |
375 | scm_cons (*(SCM *) a, |
376 | scm_cons (*(SCM *) b, SCM_EOL)), | |
377 | SCM_ENV (code)); | |
54e09076 | 378 | /* Evaluate the closure body */ |
f9450cdb | 379 | return !SCM_FALSEP (scm_eval_body (SCM_CLOSURE_BODY (code), env)); |
54e09076 MD |
380 | } /* closureless */ |
381 | ||
382 | static int | |
383 | applyless (SCM less, const void *a, const void *b) | |
384 | { | |
fdc28395 | 385 | return SCM_NFALSEP (scm_call_2 (less, *(SCM *) a, *(SCM *) b)); |
54e09076 MD |
386 | } /* applyless */ |
387 | ||
388 | static cmp_fun_t | |
389 | scm_cmp_function (SCM p) | |
390 | { | |
391 | switch (SCM_TYP7 (p)) | |
392 | { | |
393 | case scm_tc7_subr_2: | |
504d99c5 | 394 | case scm_tc7_subr_2o: |
54e09076 MD |
395 | case scm_tc7_rpsubr: |
396 | case scm_tc7_asubr: | |
397 | return subr2less; | |
54e09076 MD |
398 | case scm_tc7_lsubr: |
399 | return lsubrless; | |
400 | case scm_tcs_closures: | |
401 | return closureless; | |
402 | default: | |
403 | return applyless; | |
404 | } | |
405 | } /* scm_cmp_function */ | |
406 | ||
54e09076 MD |
407 | |
408 | /* Question: Is there any need to make this a more general array sort? | |
409 | It is probably enough to manage the vector type. */ | |
410 | /* endpos equal as for substring, i.e. endpos is not included. */ | |
da4a1dba | 411 | /* More natural with length? */ |
1bbd0b84 | 412 | |
a1ec6916 | 413 | SCM_DEFINE (scm_restricted_vector_sort_x, "restricted-vector-sort!", 4, 0, 0, |
1bbd0b84 | 414 | (SCM vec, SCM less, SCM startpos, SCM endpos), |
e3239868 DH |
415 | "Sort the vector @var{vec}, using @var{less} for comparing\n" |
416 | "the vector elements. @var{startpos} and @var{endpos} delimit\n" | |
417 | "the range of the vector which gets sorted. The return value\n" | |
418 | "is not specified.") | |
1bbd0b84 | 419 | #define FUNC_NAME s_scm_restricted_vector_sort_x |
54e09076 MD |
420 | { |
421 | size_t vlen, spos, len, size = sizeof (SCM); | |
422 | SCM *vp; | |
423 | ||
34d19ef6 HWN |
424 | SCM_VALIDATE_VECTOR (1, vec); |
425 | SCM_VALIDATE_NIM (2, less); | |
b5c2579a | 426 | |
34d19ef6 | 427 | vp = SCM_WRITABLE_VELTS (vec); /* vector pointer */ |
b5c2579a | 428 | vlen = SCM_VECTOR_LENGTH (vec); |
54e09076 | 429 | |
729dbac3 | 430 | SCM_VALIDATE_INUM_MIN_COPY (3, startpos, 0, spos); |
34d19ef6 HWN |
431 | SCM_ASSERT_RANGE (3, startpos, spos <= vlen); |
432 | SCM_VALIDATE_INUM_RANGE (4, endpos,0, vlen+1); | |
54e09076 MD |
433 | len = SCM_INUM (endpos) - spos; |
434 | ||
435 | quicksort (&vp[spos], len, size, scm_cmp_function (less), less); | |
34d19ef6 | 436 | |
504d99c5 | 437 | return scm_return_first (SCM_UNSPECIFIED, vec); |
54e09076 | 438 | /* return vec; */ |
1bbd0b84 GB |
439 | } |
440 | #undef FUNC_NAME | |
54e09076 MD |
441 | |
442 | /* (sorted? sequence less?) | |
443 | * is true when sequence is a list (x0 x1 ... xm) or a vector #(x0 ... xm) | |
444 | * such that for all 1 <= i <= m, | |
445 | * (not (less? (list-ref list i) (list-ref list (- i 1)))). */ | |
a1ec6916 | 446 | SCM_DEFINE (scm_sorted_p, "sorted?", 2, 0, 0, |
1bbd0b84 | 447 | (SCM items, SCM less), |
e3239868 DH |
448 | "Return @code{#t} iff @var{items} is a list or a vector such that\n" |
449 | "for all 1 <= i <= m, the predicate @var{less} returns true when\n" | |
450 | "applied to all elements i - 1 and i") | |
1bbd0b84 | 451 | #define FUNC_NAME s_scm_sorted_p |
54e09076 | 452 | { |
c014a02e | 453 | long len, j; /* list/vector length, temp j */ |
54e09076 | 454 | SCM item, rest; /* rest of items loop variable */ |
34d19ef6 | 455 | SCM const *vp; |
54e09076 MD |
456 | cmp_fun_t cmp = scm_cmp_function (less); |
457 | ||
c96d76b8 | 458 | if (SCM_NULL_OR_NIL_P (items)) |
54e09076 | 459 | return SCM_BOOL_T; |
1bbd0b84 | 460 | |
34d19ef6 | 461 | SCM_VALIDATE_NIM (2, less); |
54e09076 MD |
462 | |
463 | if (SCM_CONSP (items)) | |
464 | { | |
465 | len = scm_ilength (items); /* also checks that it's a pure list */ | |
34d19ef6 | 466 | SCM_ASSERT_RANGE (1, items, len >= 0); |
54e09076 MD |
467 | if (len <= 1) |
468 | return SCM_BOOL_T; | |
469 | ||
470 | item = SCM_CAR (items); | |
471 | rest = SCM_CDR (items); | |
472 | j = len - 1; | |
473 | while (j > 0) | |
474 | { | |
4b479d98 | 475 | if ((*cmp) (less, SCM_CARLOC(rest), &item)) |
54e09076 MD |
476 | return SCM_BOOL_F; |
477 | else | |
478 | { | |
479 | item = SCM_CAR (rest); | |
480 | rest = SCM_CDR (rest); | |
481 | j--; | |
482 | } | |
483 | } | |
484 | return SCM_BOOL_T; | |
485 | } | |
486 | else | |
487 | { | |
b5c2579a DH |
488 | SCM_VALIDATE_VECTOR (1, items); |
489 | ||
490 | vp = SCM_VELTS (items); /* vector pointer */ | |
491 | len = SCM_VECTOR_LENGTH (items); | |
492 | j = len - 1; | |
493 | while (j > 0) | |
54e09076 | 494 | { |
b5c2579a DH |
495 | if ((*cmp) (less, &vp[1], vp)) |
496 | return SCM_BOOL_F; | |
497 | else | |
498 | { | |
499 | vp++; | |
500 | j--; | |
501 | } | |
54e09076 | 502 | } |
b5c2579a | 503 | return SCM_BOOL_T; |
54e09076 | 504 | } |
b5c2579a | 505 | |
54e09076 | 506 | return SCM_BOOL_F; |
1bbd0b84 GB |
507 | } |
508 | #undef FUNC_NAME | |
54e09076 MD |
509 | |
510 | /* (merge a b less?) | |
511 | takes two lists a and b such that (sorted? a less?) and (sorted? b less?) | |
512 | and returns a new list in which the elements of a and b have been stably | |
513 | interleaved so that (sorted? (merge a b less?) less?). | |
514 | Note: this does _not_ accept vectors. */ | |
a1ec6916 | 515 | SCM_DEFINE (scm_merge, "merge", 3, 0, 0, |
1bbd0b84 | 516 | (SCM alist, SCM blist, SCM less), |
8f85c0c6 NJ |
517 | "Merge two already sorted lists into one.\n" |
518 | "Given two lists @var{alist} and @var{blist}, such that\n" | |
519 | "@code{(sorted? alist less?)} and @code{(sorted? blist less?)},\n" | |
520 | "return a new list in which the elements of @var{alist} and\n" | |
e3239868 DH |
521 | "@var{blist} have been stably interleaved so that\n" |
522 | "@code{(sorted? (merge alist blist less?) less?)}.\n" | |
523 | "Note: this does _not_ accept vectors.") | |
1bbd0b84 | 524 | #define FUNC_NAME s_scm_merge |
54e09076 | 525 | { |
c014a02e | 526 | long alen, blen; /* list lengths */ |
54e09076 MD |
527 | SCM build, last; |
528 | cmp_fun_t cmp = scm_cmp_function (less); | |
34d19ef6 | 529 | SCM_VALIDATE_NIM (3, less); |
54e09076 | 530 | |
c96d76b8 | 531 | if (SCM_NULL_OR_NIL_P (alist)) |
54e09076 | 532 | return blist; |
c96d76b8 | 533 | else if (SCM_NULL_OR_NIL_P (blist)) |
54e09076 MD |
534 | return alist; |
535 | else | |
536 | { | |
34d19ef6 HWN |
537 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1, alist, alen); |
538 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2, blist, blen); | |
4b479d98 | 539 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
540 | { |
541 | build = scm_cons (SCM_CAR (blist), SCM_EOL); | |
542 | blist = SCM_CDR (blist); | |
543 | blen--; | |
544 | } | |
c56cc3c8 MD |
545 | else |
546 | { | |
547 | build = scm_cons (SCM_CAR (alist), SCM_EOL); | |
548 | alist = SCM_CDR (alist); | |
549 | alen--; | |
550 | } | |
54e09076 MD |
551 | last = build; |
552 | while ((alen > 0) && (blen > 0)) | |
553 | { | |
4b479d98 | 554 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
555 | { |
556 | SCM_SETCDR (last, scm_cons (SCM_CAR (blist), SCM_EOL)); | |
557 | blist = SCM_CDR (blist); | |
558 | blen--; | |
559 | } | |
c56cc3c8 MD |
560 | else |
561 | { | |
562 | SCM_SETCDR (last, scm_cons (SCM_CAR (alist), SCM_EOL)); | |
563 | alist = SCM_CDR (alist); | |
564 | alen--; | |
565 | } | |
54e09076 MD |
566 | last = SCM_CDR (last); |
567 | } | |
568 | if ((alen > 0) && (blen == 0)) | |
569 | SCM_SETCDR (last, alist); | |
570 | else if ((alen == 0) && (blen > 0)) | |
571 | SCM_SETCDR (last, blist); | |
572 | } | |
573 | return build; | |
1bbd0b84 GB |
574 | } |
575 | #undef FUNC_NAME | |
576 | ||
54e09076 MD |
577 | |
578 | static SCM | |
579 | scm_merge_list_x (SCM alist, SCM blist, | |
580 | long alen, long blen, | |
581 | cmp_fun_t cmp, SCM less) | |
582 | { | |
583 | SCM build, last; | |
584 | ||
c96d76b8 | 585 | if (SCM_NULL_OR_NIL_P (alist)) |
54e09076 | 586 | return blist; |
c96d76b8 | 587 | else if (SCM_NULL_OR_NIL_P (blist)) |
54e09076 MD |
588 | return alist; |
589 | else | |
590 | { | |
4b479d98 | 591 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
592 | { |
593 | build = blist; | |
594 | blist = SCM_CDR (blist); | |
595 | blen--; | |
596 | } | |
c56cc3c8 MD |
597 | else |
598 | { | |
599 | build = alist; | |
600 | alist = SCM_CDR (alist); | |
601 | alen--; | |
602 | } | |
54e09076 MD |
603 | last = build; |
604 | while ((alen > 0) && (blen > 0)) | |
605 | { | |
4b479d98 | 606 | if ((*cmp) (less, SCM_CARLOC (blist), SCM_CARLOC (alist))) |
54e09076 MD |
607 | { |
608 | SCM_SETCDR (last, blist); | |
609 | blist = SCM_CDR (blist); | |
610 | blen--; | |
611 | } | |
c56cc3c8 MD |
612 | else |
613 | { | |
614 | SCM_SETCDR (last, alist); | |
615 | alist = SCM_CDR (alist); | |
616 | alen--; | |
617 | } | |
54e09076 MD |
618 | last = SCM_CDR (last); |
619 | } | |
620 | if ((alen > 0) && (blen == 0)) | |
621 | SCM_SETCDR (last, alist); | |
622 | else if ((alen == 0) && (blen > 0)) | |
623 | SCM_SETCDR (last, blist); | |
624 | } | |
625 | return build; | |
626 | } /* scm_merge_list_x */ | |
627 | ||
a1ec6916 | 628 | SCM_DEFINE (scm_merge_x, "merge!", 3, 0, 0, |
1bbd0b84 | 629 | (SCM alist, SCM blist, SCM less), |
e3239868 DH |
630 | "Takes two lists @var{alist} and @var{blist} such that\n" |
631 | "@code{(sorted? alist less?)} and @code{(sorted? blist less?)} and\n" | |
632 | "returns a new list in which the elements of @var{alist} and\n" | |
633 | "@var{blist} have been stably interleaved so that\n" | |
634 | " @code{(sorted? (merge alist blist less?) less?)}.\n" | |
635 | "This is the destructive variant of @code{merge}\n" | |
636 | "Note: this does _not_ accept vectors.") | |
1bbd0b84 | 637 | #define FUNC_NAME s_scm_merge_x |
54e09076 | 638 | { |
c014a02e | 639 | long alen, blen; /* list lengths */ |
54e09076 | 640 | |
34d19ef6 | 641 | SCM_VALIDATE_NIM (3, less); |
c96d76b8 | 642 | if (SCM_NULL_OR_NIL_P (alist)) |
54e09076 | 643 | return blist; |
c96d76b8 | 644 | else if (SCM_NULL_OR_NIL_P (blist)) |
54e09076 MD |
645 | return alist; |
646 | else | |
647 | { | |
34d19ef6 HWN |
648 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (1, alist, alen); |
649 | SCM_VALIDATE_NONEMPTYLIST_COPYLEN (2, blist, blen); | |
54e09076 MD |
650 | return scm_merge_list_x (alist, blist, |
651 | alen, blen, | |
652 | scm_cmp_function (less), | |
653 | less); | |
654 | } | |
1bbd0b84 GB |
655 | } |
656 | #undef FUNC_NAME | |
54e09076 MD |
657 | |
658 | /* This merge sort algorithm is same as slib's by Richard A. O'Keefe. | |
659 | The algorithm is stable. We also tried to use the algorithm used by | |
660 | scsh's merge-sort but that algorithm showed to not be stable, even | |
661 | though it claimed to be. | |
662 | */ | |
663 | static SCM | |
664 | scm_merge_list_step (SCM * seq, | |
665 | cmp_fun_t cmp, | |
666 | SCM less, | |
c014a02e | 667 | long n) |
54e09076 | 668 | { |
c56cc3c8 MD |
669 | SCM a, b; |
670 | ||
54e09076 MD |
671 | if (n > 2) |
672 | { | |
c014a02e | 673 | long mid = n / 2; |
c56cc3c8 MD |
674 | a = scm_merge_list_step (seq, cmp, less, mid); |
675 | b = scm_merge_list_step (seq, cmp, less, n - mid); | |
676 | return scm_merge_list_x (a, b, mid, n - mid, cmp, less); | |
54e09076 MD |
677 | } |
678 | else if (n == 2) | |
679 | { | |
680 | SCM p = *seq; | |
681 | SCM rest = SCM_CDR (*seq); | |
682 | SCM x = SCM_CAR (*seq); | |
683 | SCM y = SCM_CAR (SCM_CDR (*seq)); | |
684 | *seq = SCM_CDR (rest); | |
685 | SCM_SETCDR (rest, SCM_EOL); | |
686 | if ((*cmp) (less, &y, &x)) | |
687 | { | |
4b479d98 DH |
688 | SCM_SETCAR (p, y); |
689 | SCM_SETCAR (rest, x); | |
54e09076 MD |
690 | } |
691 | return p; | |
692 | } | |
693 | else if (n == 1) | |
694 | { | |
695 | SCM p = *seq; | |
696 | *seq = SCM_CDR (p); | |
697 | SCM_SETCDR (p, SCM_EOL); | |
698 | return p; | |
699 | } | |
700 | else | |
701 | return SCM_EOL; | |
702 | } /* scm_merge_list_step */ | |
703 | ||
704 | ||
54e09076 | 705 | /* scm_sort_x manages lists and vectors, not stable sort */ |
a1ec6916 | 706 | SCM_DEFINE (scm_sort_x, "sort!", 2, 0, 0, |
1bbd0b84 | 707 | (SCM items, SCM less), |
e3239868 DH |
708 | "Sort the sequence @var{items}, which may be a list or a\n" |
709 | "vector. @var{less} is used for comparing the sequence\n" | |
710 | "elements. The sorting is destructive, that means that the\n" | |
711 | "input sequence is modified to produce the sorted result.\n" | |
712 | "This is not a stable sort.") | |
1bbd0b84 | 713 | #define FUNC_NAME s_scm_sort_x |
54e09076 | 714 | { |
c014a02e | 715 | long len; /* list/vector length */ |
c96d76b8 NJ |
716 | if (SCM_NULL_OR_NIL_P (items)) |
717 | return items; | |
b5c2579a | 718 | |
34d19ef6 | 719 | SCM_VALIDATE_NIM (2, less); |
54e09076 MD |
720 | |
721 | if (SCM_CONSP (items)) | |
722 | { | |
34d19ef6 | 723 | SCM_VALIDATE_LIST_COPYLEN (1, items, len); |
54e09076 MD |
724 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); |
725 | } | |
726 | else if (SCM_VECTORP (items)) | |
727 | { | |
b5c2579a | 728 | len = SCM_VECTOR_LENGTH (items); |
54e09076 MD |
729 | scm_restricted_vector_sort_x (items, |
730 | less, | |
731 | SCM_MAKINUM (0L), | |
732 | SCM_MAKINUM (len)); | |
733 | return items; | |
734 | } | |
735 | else | |
276dd677 | 736 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 737 | } |
0f981281 | 738 | #undef FUNC_NAME |
54e09076 MD |
739 | |
740 | /* scm_sort manages lists and vectors, not stable sort */ | |
1bbd0b84 | 741 | |
a1ec6916 | 742 | SCM_DEFINE (scm_sort, "sort", 2, 0, 0, |
1bbd0b84 | 743 | (SCM items, SCM less), |
e3239868 DH |
744 | "Sort the sequence @var{items}, which may be a list or a\n" |
745 | "vector. @var{less} is used for comparing the sequence\n" | |
746 | "elements. This is not a stable sort.") | |
1bbd0b84 | 747 | #define FUNC_NAME s_scm_sort |
54e09076 | 748 | { |
c96d76b8 NJ |
749 | if (SCM_NULL_OR_NIL_P (items)) |
750 | return items; | |
b5c2579a | 751 | |
34d19ef6 | 752 | SCM_VALIDATE_NIM (2, less); |
54e09076 MD |
753 | if (SCM_CONSP (items)) |
754 | { | |
c014a02e | 755 | long len; |
e9e225e5 | 756 | |
34d19ef6 | 757 | SCM_VALIDATE_LIST_COPYLEN (1, items, len); |
54e09076 MD |
758 | items = scm_list_copy (items); |
759 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); | |
760 | } | |
3cfe6eab | 761 | #if SCM_HAVE_ARRAYS |
afe5177e | 762 | /* support ordinary vectors even if arrays not available? */ |
54e09076 MD |
763 | else if (SCM_VECTORP (items)) |
764 | { | |
c014a02e | 765 | long len = SCM_VECTOR_LENGTH (items); |
e9e225e5 GH |
766 | SCM sortvec = scm_make_uve (len, scm_array_prototype (items)); |
767 | ||
54e09076 MD |
768 | scm_array_copy_x (items, sortvec); |
769 | scm_restricted_vector_sort_x (sortvec, | |
770 | less, | |
771 | SCM_MAKINUM (0L), | |
772 | SCM_MAKINUM (len)); | |
773 | return sortvec; | |
774 | } | |
afe5177e | 775 | #endif |
54e09076 | 776 | else |
276dd677 | 777 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 778 | } |
0f981281 | 779 | #undef FUNC_NAME |
54e09076 MD |
780 | |
781 | static void | |
1d1559ce HWN |
782 | scm_merge_vector_x (SCM vec, |
783 | SCM * temp, | |
54e09076 MD |
784 | cmp_fun_t cmp, |
785 | SCM less, | |
c014a02e ML |
786 | long low, |
787 | long mid, | |
788 | long high) | |
54e09076 | 789 | { |
c014a02e ML |
790 | long it; /* Index for temp vector */ |
791 | long i1 = low; /* Index for lower vector segment */ | |
792 | long i2 = mid + 1; /* Index for upper vector segment */ | |
54e09076 MD |
793 | |
794 | /* Copy while both segments contain more characters */ | |
795 | for (it = low; (i1 <= mid) && (i2 <= high); ++it) | |
1d1559ce HWN |
796 | { |
797 | /* | |
798 | Every call of LESS might invoke GC. For full correctness, we | |
799 | should reset the generation of vecbase and tempbase between | |
800 | every call of less. | |
54e09076 | 801 | |
1d1559ce HWN |
802 | */ |
803 | register SCM *vp = SCM_WRITABLE_VELTS(vec); | |
804 | ||
805 | if ((*cmp) (less, &vp[i2], &vp[i1])) | |
806 | temp[it] = vp[i2++]; | |
807 | else | |
808 | temp[it] = vp[i1++]; | |
809 | } | |
54e09076 | 810 | |
1d1559ce HWN |
811 | { |
812 | register SCM *vp = SCM_WRITABLE_VELTS(vec); | |
813 | ||
814 | /* Copy while first segment contains more characters */ | |
815 | while (i1 <= mid) | |
816 | temp[it++] = vp[i1++]; | |
817 | ||
818 | /* Copy while second segment contains more characters */ | |
819 | while (i2 <= high) | |
820 | temp[it++] = vp[i2++]; | |
821 | ||
822 | /* Copy back from temp to vp */ | |
823 | for (it = low; it <= high; ++it) | |
824 | vp[it] = temp[it]; | |
825 | } | |
826 | } /* scm_merge_vector_x */ | |
54e09076 MD |
827 | |
828 | static void | |
1d1559ce HWN |
829 | scm_merge_vector_step (SCM vp, |
830 | SCM * temp, | |
54e09076 MD |
831 | cmp_fun_t cmp, |
832 | SCM less, | |
c014a02e ML |
833 | long low, |
834 | long high) | |
54e09076 MD |
835 | { |
836 | if (high > low) | |
837 | { | |
c014a02e | 838 | long mid = (low + high) / 2; |
54e09076 MD |
839 | scm_merge_vector_step (vp, temp, cmp, less, low, mid); |
840 | scm_merge_vector_step (vp, temp, cmp, less, mid+1, high); | |
841 | scm_merge_vector_x (vp, temp, cmp, less, low, mid, high); | |
842 | } | |
843 | } /* scm_merge_vector_step */ | |
844 | ||
845 | ||
54e09076 MD |
846 | /* stable-sort! manages lists and vectors */ |
847 | ||
a1ec6916 | 848 | SCM_DEFINE (scm_stable_sort_x, "stable-sort!", 2, 0, 0, |
1bbd0b84 | 849 | (SCM items, SCM less), |
e3239868 DH |
850 | "Sort the sequence @var{items}, which may be a list or a\n" |
851 | "vector. @var{less} is used for comparing the sequence elements.\n" | |
852 | "The sorting is destructive, that means that the input sequence\n" | |
853 | "is modified to produce the sorted result.\n" | |
854 | "This is a stable sort.") | |
1bbd0b84 | 855 | #define FUNC_NAME s_scm_stable_sort_x |
54e09076 | 856 | { |
c014a02e | 857 | long len; /* list/vector length */ |
54e09076 | 858 | |
c96d76b8 NJ |
859 | if (SCM_NULL_OR_NIL_P (items)) |
860 | return items; | |
b5c2579a | 861 | |
34d19ef6 | 862 | SCM_VALIDATE_NIM (2, less); |
54e09076 MD |
863 | if (SCM_CONSP (items)) |
864 | { | |
34d19ef6 | 865 | SCM_VALIDATE_LIST_COPYLEN (1, items, len); |
54e09076 MD |
866 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); |
867 | } | |
868 | else if (SCM_VECTORP (items)) | |
869 | { | |
1d1559ce | 870 | SCM *temp; |
b5c2579a | 871 | len = SCM_VECTOR_LENGTH (items); |
34d19ef6 | 872 | |
34d19ef6 | 873 | /* |
1d1559ce HWN |
874 | the following array does not contain any new references to |
875 | SCM objects, so we can get away with allocing it on the heap. | |
876 | */ | |
67329a9e | 877 | temp = scm_malloc (len * sizeof(SCM)); |
34d19ef6 | 878 | |
1d1559ce | 879 | scm_merge_vector_step (items, |
54e09076 MD |
880 | temp, |
881 | scm_cmp_function (less), | |
882 | less, | |
883 | 0, | |
884 | len - 1); | |
885 | free(temp); | |
886 | return items; | |
887 | } | |
888 | else | |
276dd677 | 889 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 890 | } |
0f981281 | 891 | #undef FUNC_NAME |
54e09076 MD |
892 | |
893 | /* stable_sort manages lists and vectors */ | |
a1ec6916 | 894 | SCM_DEFINE (scm_stable_sort, "stable-sort", 2, 0, 0, |
1bbd0b84 | 895 | (SCM items, SCM less), |
e3239868 DH |
896 | "Sort the sequence @var{items}, which may be a list or a\n" |
897 | "vector. @var{less} is used for comparing the sequence elements.\n" | |
898 | "This is a stable sort.") | |
1bbd0b84 | 899 | #define FUNC_NAME s_scm_stable_sort |
54e09076 | 900 | { |
1d1559ce | 901 | |
c96d76b8 NJ |
902 | if (SCM_NULL_OR_NIL_P (items)) |
903 | return items; | |
b5c2579a | 904 | |
34d19ef6 | 905 | SCM_VALIDATE_NIM (2, less); |
54e09076 MD |
906 | if (SCM_CONSP (items)) |
907 | { | |
1d1559ce | 908 | long len; /* list/vector length */ |
34d19ef6 | 909 | SCM_VALIDATE_LIST_COPYLEN (1, items, len); |
54e09076 MD |
910 | items = scm_list_copy (items); |
911 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); | |
912 | } | |
3cfe6eab | 913 | #if SCM_HAVE_ARRAYS |
afe5177e | 914 | /* support ordinary vectors even if arrays not available? */ |
54e09076 MD |
915 | else if (SCM_VECTORP (items)) |
916 | { | |
1d1559ce | 917 | long len = SCM_VECTOR_LENGTH (items); |
67329a9e | 918 | SCM *temp = scm_malloc (len * sizeof (SCM)); |
1d1559ce | 919 | SCM retvec = scm_make_uve (len, scm_array_prototype (items)); |
54e09076 | 920 | scm_array_copy_x (items, retvec); |
34d19ef6 | 921 | |
1d1559ce | 922 | scm_merge_vector_step (retvec, |
54e09076 MD |
923 | temp, |
924 | scm_cmp_function (less), | |
925 | less, | |
926 | 0, | |
927 | len - 1); | |
928 | free (temp); | |
929 | return retvec; | |
930 | } | |
afe5177e | 931 | #endif |
54e09076 | 932 | else |
276dd677 | 933 | SCM_WRONG_TYPE_ARG (1, items); |
1bbd0b84 | 934 | } |
0f981281 | 935 | #undef FUNC_NAME |
54e09076 | 936 | |
1bbd0b84 | 937 | /* stable */ |
a1ec6916 | 938 | SCM_DEFINE (scm_sort_list_x, "sort-list!", 2, 0, 0, |
1bbd0b84 | 939 | (SCM items, SCM less), |
e3239868 DH |
940 | "Sort the list @var{items}, using @var{less} for comparing the\n" |
941 | "list elements. The sorting is destructive, that means that the\n" | |
942 | "input list is modified to produce the sorted result.\n" | |
943 | "This is a stable sort.") | |
1bbd0b84 | 944 | #define FUNC_NAME s_scm_sort_list_x |
54e09076 | 945 | { |
c014a02e | 946 | long len; |
34d19ef6 HWN |
947 | SCM_VALIDATE_LIST_COPYLEN (1, items, len); |
948 | SCM_VALIDATE_NIM (2, less); | |
54e09076 | 949 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); |
1bbd0b84 | 950 | } |
0f981281 | 951 | #undef FUNC_NAME |
54e09076 | 952 | |
1bbd0b84 | 953 | /* stable */ |
a1ec6916 | 954 | SCM_DEFINE (scm_sort_list, "sort-list", 2, 0, 0, |
e3239868 DH |
955 | (SCM items, SCM less), |
956 | "Sort the list @var{items}, using @var{less} for comparing the\n" | |
957 | "list elements. This is a stable sort.") | |
1bbd0b84 | 958 | #define FUNC_NAME s_scm_sort_list |
54e09076 | 959 | { |
c014a02e | 960 | long len; |
34d19ef6 HWN |
961 | SCM_VALIDATE_LIST_COPYLEN (1, items, len); |
962 | SCM_VALIDATE_NIM (2, less); | |
54e09076 MD |
963 | items = scm_list_copy (items); |
964 | return scm_merge_list_step (&items, scm_cmp_function (less), less, len); | |
1bbd0b84 | 965 | } |
0f981281 | 966 | #undef FUNC_NAME |
54e09076 MD |
967 | |
968 | void | |
969 | scm_init_sort () | |
970 | { | |
a0599745 | 971 | #include "libguile/sort.x" |
54e09076 MD |
972 | |
973 | scm_add_feature ("sort"); | |
974 | } | |
89e00824 ML |
975 | |
976 | /* | |
977 | Local Variables: | |
978 | c-file-style: "gnu" | |
979 | End: | |
980 | */ |