1 /* PSPP - computes sample statistics.
2 Copyright (C) 1997-9, 2000 Free Software Foundation, Inc.
3 Written by Ben Pfaff <blp@gnu.org>.
5 This program is free software; you can redistribute it and/or
6 modify it under the terms of the GNU General Public License as
7 published by the Free Software Foundation; either version 2 of the
8 License, or (at your option) any later version.
10 This program is distributed in the hope that it will be useful, but
11 WITHOUT ANY WARRANTY; without even the implied warranty of
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 General Public License for more details.
15 You should have received a copy of the GNU General Public License
16 along with this program; if not, write to the Free Software
17 Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
20 /* Copyright (C) 2001 Free Software Foundation, Inc.
22 This file is part of the GNU ISO C++ Library. This library is free
23 software; you can redistribute it and/or modify it under the
24 terms of the GNU General Public License as published by the
25 Free Software Foundation; either version 2, or (at your option)
28 This library is distributed in the hope that it will be useful,
29 but WITHOUT ANY WARRANTY; without even the implied warranty of
30 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
31 GNU General Public License for more details.
33 You should have received a copy of the GNU General Public License along
34 with this library; see the file COPYING. If not, write to the Free
35 Software Foundation, 59 Temple Place - Suite 330, Boston, MA 02111-1307,
38 As a special exception, you may use this file as part of a free software
39 library without restriction. Specifically, if other files instantiate
40 templates or use macros or inline functions from this file, or you compile
41 this file and link it with other files to produce an executable, this
42 file does not by itself cause the resulting executable to be covered by
43 the GNU General Public License. This exception does not however
44 invalidate any other reasons why the executable file might be covered by
45 the GNU General Public License. */
50 * Hewlett-Packard Company
52 * Permission to use, copy, modify, distribute and sell this software
53 * and its documentation for any purpose is hereby granted without fee,
54 * provided that the above copyright notice appear in all copies and
55 * that both that copyright notice and this permission notice appear
56 * in supporting documentation. Hewlett-Packard Company makes no
57 * representations about the suitability of this software for any
58 * purpose. It is provided "as is" without express or implied warranty.
62 * Silicon Graphics Computer Systems, Inc.
64 * Permission to use, copy, modify, distribute and sell this software
65 * and its documentation for any purpose is hereby granted without fee,
66 * provided that the above copyright notice appear in all copies and
67 * that both that copyright notice and this permission notice appear
68 * in supporting documentation. Silicon Graphics makes no
69 * representations about the suitability of this software for any
70 * purpose. It is provided "as is" without express or implied warranty.
77 #include "algorithm.h"
80 /* Byte-wise swap two items of size SIZE. */
81 #define SWAP(a, b, size) \
84 register size_t __size = (size); \
85 register char *__a = (a), *__b = (b); \
91 } while (--__size > 0); \
94 /* Makes the elements in ARRAY unique, by moving up duplicates,
95 and returns the new number of elements in the array. Sorted
96 arrays only. Arguments same as for sort() above. */
98 unique (void *array, size_t count, size_t size,
99 algo_compare_func *compare, void *aux)
102 char *last = first + size * count;
103 char *result = array;
111 if (compare (result, first, aux))
115 memcpy (result, first, size);
122 /* Helper function that calls sort(), then unique(). */
124 sort_unique (void *array, size_t count, size_t size,
125 algo_compare_func *compare, void *aux)
127 sort (array, count, size, compare, aux);
128 return unique (array, count, size, compare, aux);
135 xmalloc (size_t size)
137 return malloc (size);
141 compare_ints (const void *a_, const void *b_, void *aux)
155 try_unique (const char *title,
156 int *in, size_t in_cnt,
161 in_cnt = unique (in, in_cnt, sizeof *in, compare_ints, NULL);
162 if (in_cnt != out_cnt)
164 fprintf (stderr, "unique_test: %s: in_cnt %d, expected %d\n",
165 title, (int) in_cnt, (int) out_cnt);
169 for (i = 0; i < out_cnt; i++)
172 fprintf (stderr, "unique_test: %s: idx %d = %d, expected %d\n",
173 title, (int) i, in[i], i);
180 int a_in[] = {0, 0, 0, 1, 2, 3, 3, 4, 5, 5};
181 int b_in[] = {0, 1, 2, 2, 2, 3};
185 try_unique ("a", a_in, sizeof a_in / sizeof *a_in, 6);
186 try_unique ("b", b_in, sizeof b_in / sizeof *b_in, 4);
187 try_unique ("c", c_in, sizeof c_in / sizeof *c_in, 1);
188 try_unique ("d", &d_in, 0, 0);
191 #endif /* TEST_UNIQUE */
193 /* Reorders ARRAY, which contains COUNT elements of SIZE bytes
194 each, so that the elements for which PREDICATE returns nonzero
195 precede those for which PREDICATE returns zero. AUX is
196 passed to each predicate as auxiliary data. Returns the
197 number of elements for which PREDICATE returns nonzero. Not
200 partition (void *array, size_t count, size_t size,
201 algo_predicate_func *predicate, void *aux)
204 char *last = first + count * size;
208 /* Move FIRST forward to point to first element that fails
214 else if (!predicate (first, aux))
221 /* Move LAST backward to point to last element that passes
229 else if (predicate (last, aux))
235 /* By swapping FIRST and LAST we extend the starting and
236 ending sequences that pass and fail, respectively,
238 SWAP (first, last, size);
243 /* A algo_random_func that uses random.h. */
245 algo_default_random (unsigned max, void *aux unused)
247 return rng_get_unsigned (pspp_rng ()) % max;
250 /* Randomly reorders ARRAY, which contains COUNT elements of SIZE
251 bytes each. Uses RANDOM as a source of random data, passing
252 AUX as the auxiliary data. RANDOM may be null to use a
253 default random source. */
255 random_shuffle (void *array_, size_t count, size_t size,
256 algo_random_func *random, void *aux)
258 unsigned char *array = array_;
262 random = algo_default_random;
264 for (i = 1; i < count; i++)
265 SWAP (array + i * size, array + random (i + 1, aux) * size, size);
268 /* Copies the COUNT elements of SIZE bytes each from ARRAY to
269 RESULT, except that elements for which PREDICATE is false are
270 not copied. Returns the number of elements copied. AUX is
271 passed to PREDICATE as auxiliary data. */
273 copy_if (const void *array, size_t count, size_t size,
275 algo_predicate_func *predicate, void *aux)
277 const unsigned char *input = array;
278 const unsigned char *last = input + size * count;
279 unsigned char *output = result;
281 while (input <= last)
283 if (predicate (input, aux))
285 memcpy (output, input, size);
297 /* A predicate and its auxiliary data. */
300 algo_predicate_func *predicate;
305 not (const void *data, void *pred_aux_)
307 const struct pred_aux *pred_aux = pred_aux_;
309 return !pred_aux->predicate (data, pred_aux->aux);
312 /* Copies the COUNT elements of SIZE bytes each from ARRAY to
313 RESULT, except that elements for which PREDICATE is true are
314 not copied. Returns the number of elements copied. AUX is
315 passed to PREDICATE as auxiliary data. */
317 remove_copy_if (const void *array, size_t count, size_t size,
319 algo_predicate_func *predicate, void *aux)
321 struct pred_aux pred_aux;
322 pred_aux.predicate = predicate;
324 return copy_if (array, count, size, result, not, &pred_aux);
327 /* Copyright (C) 1991, 1992, 1996, 1997, 1999 Free Software Foundation, Inc.
328 This file is part of the GNU C Library.
329 Written by Douglas C. Schmidt (schmidt@ics.uci.edu).
331 The GNU C Library is free software; you can redistribute it and/or
332 modify it under the terms of the GNU Lesser General Public
333 License as published by the Free Software Foundation; either
334 version 2.1 of the License, or (at your option) any later version.
336 The GNU C Library is distributed in the hope that it will be useful,
337 but WITHOUT ANY WARRANTY; without even the implied warranty of
338 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
339 Lesser General Public License for more details.
341 You should have received a copy of the GNU Lesser General Public
342 License along with the GNU C Library; if not, write to the Free
343 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
346 /* If you consider tuning this algorithm, you should consult first:
347 Engineering a sort function; Jon Bentley and M. Douglas McIlroy;
348 Software - Practice and Experience; Vol. 23 (11), 1249-1265, 1993. */
355 /* Discontinue quicksort algorithm when partition gets below this size.
356 This particular magic number was chosen to work best on a Sun 4/260. */
359 /* Stack node declarations used to store unfulfilled partition obligations. */
366 /* The next 4 #defines implement a very fast in-line stack abstraction. */
367 /* The stack needs log (total_elements) entries (we could even subtract
368 log(MAX_THRESH)). Since total_elements has type size_t, we get as
369 upper bound for log (total_elements):
370 bits per byte (CHAR_BIT) * sizeof(size_t). */
371 #define STACK_SIZE (CHAR_BIT * sizeof(size_t))
372 #define PUSH(low, high) ((void) ((top->lo = (low)), (top->hi = (high)), ++top))
373 #define POP(low, high) ((void) (--top, (low = top->lo), (high = top->hi)))
374 #define STACK_NOT_EMPTY (stack < top)
377 /* Order size using quicksort. This implementation incorporates
378 four optimizations discussed in Sedgewick:
380 1. Non-recursive, using an explicit stack of pointer that store the
381 next array partition to sort. To save time, this maximum amount
382 of space required to store an array of SIZE_MAX is allocated on the
383 stack. Assuming a 32-bit (64 bit) integer for size_t, this needs
384 only 32 * sizeof(stack_node) == 256 bytes (for 64 bit: 1024 bytes).
385 Pretty cheap, actually.
387 2. Chose the pivot element using a median-of-three decision tree.
388 This reduces the probability of selecting a bad pivot value and
389 eliminates certain extraneous comparisons.
391 3. Only quicksorts TOTAL_ELEMS / MAX_THRESH partitions, leaving
392 insertion sort to order the MAX_THRESH items within each partition.
393 This is a big win, since insertion sort is faster for small, mostly
394 sorted array segments.
396 4. The larger of the two sub-partitions is always pushed onto the
397 stack first, with the algorithm then concentrating on the
398 smaller partition. This *guarantees* no more than log (total_elems)
399 stack size is needed (actually O(1) in this case)! */
402 sort (void *const pbase, size_t total_elems, size_t size,
403 algo_compare_func *cmp, void *aux)
405 register char *base_ptr = (char *) pbase;
407 const size_t max_thresh = MAX_THRESH * size;
409 if (total_elems == 0)
410 /* Avoid lossage with unsigned arithmetic below. */
413 if (total_elems > MAX_THRESH)
416 char *hi = &lo[size * (total_elems - 1)];
417 stack_node stack[STACK_SIZE];
418 stack_node *top = stack + 1;
420 while (STACK_NOT_EMPTY)
425 /* Select median value from among LO, MID, and HI. Rearrange
426 LO and HI so the three values are sorted. This lowers the
427 probability of picking a pathological pivot value and
428 skips a comparison for both the LEFT_PTR and RIGHT_PTR in
431 char *mid = lo + size * ((hi - lo) / size >> 1);
433 if ((*cmp) ((void *) mid, (void *) lo, aux) < 0)
434 SWAP (mid, lo, size);
435 if ((*cmp) ((void *) hi, (void *) mid, aux) < 0)
436 SWAP (mid, hi, size);
439 if ((*cmp) ((void *) mid, (void *) lo, aux) < 0)
440 SWAP (mid, lo, size);
443 left_ptr = lo + size;
444 right_ptr = hi - size;
446 /* Here's the famous ``collapse the walls'' section of quicksort.
447 Gotta like those tight inner loops! They are the main reason
448 that this algorithm runs much faster than others. */
451 while ((*cmp) ((void *) left_ptr, (void *) mid, aux) < 0)
454 while ((*cmp) ((void *) mid, (void *) right_ptr, aux) < 0)
457 if (left_ptr < right_ptr)
459 SWAP (left_ptr, right_ptr, size);
462 else if (mid == right_ptr)
467 else if (left_ptr == right_ptr)
474 while (left_ptr <= right_ptr);
476 /* Set up pointers for next iteration. First determine whether
477 left and right partitions are below the threshold size. If so,
478 ignore one or both. Otherwise, push the larger partition's
479 bounds on the stack and continue sorting the smaller one. */
481 if ((size_t) (right_ptr - lo) <= max_thresh)
483 if ((size_t) (hi - left_ptr) <= max_thresh)
484 /* Ignore both small partitions. */
487 /* Ignore small left partition. */
490 else if ((size_t) (hi - left_ptr) <= max_thresh)
491 /* Ignore small right partition. */
493 else if ((right_ptr - lo) > (hi - left_ptr))
495 /* Push larger left partition indices. */
496 PUSH (lo, right_ptr);
501 /* Push larger right partition indices. */
508 /* Once the BASE_PTR array is partially sorted by quicksort the rest
509 is completely sorted using insertion sort, since this is efficient
510 for partitions below MAX_THRESH size. BASE_PTR points to the beginning
511 of the array to sort, and END_PTR points at the very last element in
512 the array (*not* one beyond it!). */
514 #define min(x, y) ((x) < (y) ? (x) : (y))
517 char *const end_ptr = &base_ptr[size * (total_elems - 1)];
518 char *tmp_ptr = base_ptr;
519 char *thresh = min(end_ptr, base_ptr + max_thresh);
520 register char *run_ptr;
522 /* Find smallest element in first threshold and place it at the
523 array's beginning. This is the smallest array element,
524 and the operation speeds up insertion sort's inner loop. */
526 for (run_ptr = tmp_ptr + size; run_ptr <= thresh; run_ptr += size)
527 if ((*cmp) ((void *) run_ptr, (void *) tmp_ptr, aux) < 0)
530 if (tmp_ptr != base_ptr)
531 SWAP (tmp_ptr, base_ptr, size);
533 /* Insertion sort, running from left-hand-side up to right-hand-side. */
535 run_ptr = base_ptr + size;
536 while ((run_ptr += size) <= end_ptr)
538 tmp_ptr = run_ptr - size;
539 while ((*cmp) ((void *) run_ptr, (void *) tmp_ptr, aux) < 0)
543 if (tmp_ptr != run_ptr)
547 trav = run_ptr + size;
548 while (--trav >= run_ptr)
553 for (hi = lo = trav; (lo -= size) >= tmp_ptr; hi = lo)