1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1990, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Peter McIlroy and by Dan Bernstein at New York University, 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 */ 34 35 /* 36 * Radixsort routines. 37 * 38 * Program r_sort_a() is unstable but uses O(logN) extra memory for a stack. 39 * Use radixsort(a, n, trace, endchar) for this case. 40 * 41 * For stable sorting (using N extra pointers) use sradixsort(), which calls 42 * r_sort_b(). 43 * 44 * For a description of this code, see D. McIlroy, P. McIlroy, K. Bostic, 45 * "Engineering Radix Sort". 46 */ 47 48 #include <sys/types.h> 49 #include <stdlib.h> 50 #include <stddef.h> 51 #include <errno.h> 52 53 typedef struct { 54 const u_char **sa; 55 int sn, si; 56 } stack; 57 58 static inline void simplesort 59 (const u_char **, int, int, const u_char *, u_int); 60 static void r_sort_a(const u_char **, int, int, const u_char *, u_int); 61 static void r_sort_b(const u_char **, const u_char **, int, int, 62 const u_char *, u_int); 63 64 #define THRESHOLD 20 /* Divert to simplesort(). */ 65 #define SIZE 512 /* Default stack size. */ 66 67 #define SETUP { \ 68 if (tab == NULL) { \ 69 tr = tr0; \ 70 for (c = 0; c < endch; c++) \ 71 tr0[c] = c + 1; \ 72 tr0[c] = 0; \ 73 for (c++; c < 256; c++) \ 74 tr0[c] = c; \ 75 endch = 0; \ 76 } else { \ 77 endch = tab[endch]; \ 78 tr = tab; \ 79 if (endch != 0 && endch != 255) { \ 80 errno = EINVAL; \ 81 return (-1); \ 82 } \ 83 } \ 84 } 85 86 int 87 radixsort(const u_char **a, int n, const u_char *tab, u_int endch) 88 { 89 const u_char *tr; 90 int c; 91 u_char tr0[256]; 92 93 SETUP; 94 r_sort_a(a, n, 0, tr, endch); 95 return (0); 96 } 97 98 int 99 sradixsort(const u_char **a, int n, const u_char *tab, u_int endch) 100 { 101 const u_char *tr, **ta; 102 int c; 103 u_char tr0[256]; 104 105 SETUP; 106 if (n < THRESHOLD) 107 simplesort(a, n, 0, tr, endch); 108 else { 109 if ((ta = malloc(n * sizeof(a))) == NULL) 110 return (-1); 111 r_sort_b(a, ta, n, 0, tr, endch); 112 free(ta); 113 } 114 return (0); 115 } 116 117 #define empty(s) (s >= sp) 118 #define pop(a, n, i) a = (--sp)->sa, n = sp->sn, i = sp->si 119 #define push(a, n, i) sp->sa = a, sp->sn = n, (sp++)->si = i 120 #define swap(a, b, t) t = a, a = b, b = t 121 122 /* Unstable, in-place sort. */ 123 static void 124 r_sort_a(const u_char **a, int n, int i, const u_char *tr, u_int endch) 125 { 126 static int count[256], nc, bmin; 127 int c; 128 const u_char **ak, *r; 129 stack s[SIZE], *sp, *sp0, *sp1, temp; 130 int *cp, bigc; 131 const u_char **an, *t, **aj, **top[256]; 132 133 /* Set up stack. */ 134 sp = s; 135 push(a, n, i); 136 while (!empty(s)) { 137 pop(a, n, i); 138 if (n < THRESHOLD) { 139 simplesort(a, n, i, tr, endch); 140 continue; 141 } 142 an = a + n; 143 144 /* Make character histogram. */ 145 if (nc == 0) { 146 bmin = 255; /* First occupied bin, excluding eos. */ 147 for (ak = a; ak < an;) { 148 c = tr[(*ak++)[i]]; 149 if (++count[c] == 1 && c != endch) { 150 if (c < bmin) 151 bmin = c; 152 nc++; 153 } 154 } 155 if (sp + nc > s + SIZE) { /* Get more stack. */ 156 r_sort_a(a, n, i, tr, endch); 157 continue; 158 } 159 } 160 161 /* 162 * Special case: if all strings have the same 163 * character at position i, move on to the next 164 * character. 165 */ 166 if (nc == 1 && count[bmin] == n) { 167 push(a, n, i+1); 168 nc = count[bmin] = 0; 169 continue; 170 } 171 172 /* 173 * Set top[]; push incompletely sorted bins onto stack. 174 * top[] = pointers to last out-of-place element in bins. 175 * count[] = counts of elements in bins. 176 * Before permuting: top[c-1] + count[c] = top[c]; 177 * during deal: top[c] counts down to top[c-1]. 178 */ 179 sp0 = sp1 = sp; /* Stack position of biggest bin. */ 180 bigc = 2; /* Size of biggest bin. */ 181 if (endch == 0) /* Special case: set top[eos]. */ 182 top[0] = ak = a + count[0]; 183 else { 184 ak = a; 185 top[255] = an; 186 } 187 for (cp = count + bmin; nc > 0; cp++) { 188 while (*cp == 0) /* Find next non-empty pile. */ 189 cp++; 190 if (*cp > 1) { 191 if (*cp > bigc) { 192 bigc = *cp; 193 sp1 = sp; 194 } 195 push(ak, *cp, i+1); 196 } 197 top[cp-count] = ak += *cp; 198 nc--; 199 } 200 swap(*sp0, *sp1, temp); /* Play it safe -- biggest bin last. */ 201 202 /* 203 * Permute misplacements home. Already home: everything 204 * before aj, and in bin[c], items from top[c] on. 205 * Inner loop: 206 * r = next element to put in place; 207 * ak = top[r[i]] = location to put the next element. 208 * aj = bottom of 1st disordered bin. 209 * Outer loop: 210 * Once the 1st disordered bin is done, ie. aj >= ak, 211 * aj<-aj + count[c] connects the bins in a linked list; 212 * reset count[c]. 213 */ 214 for (aj = a; aj < an; *aj = r, aj += count[c], count[c] = 0) 215 for (r = *aj; aj < (ak = --top[c = tr[r[i]]]);) 216 swap(*ak, r, t); 217 } 218 } 219 220 /* Stable sort, requiring additional memory. */ 221 static void 222 r_sort_b(const u_char **a, const u_char **ta, int n, int i, const u_char *tr, 223 u_int endch) 224 { 225 static int count[256], nc, bmin; 226 int c; 227 const u_char **ak, **ai; 228 stack s[512], *sp, *sp0, *sp1, temp; 229 const u_char **top[256]; 230 int *cp, bigc; 231 232 sp = s; 233 push(a, n, i); 234 while (!empty(s)) { 235 pop(a, n, i); 236 if (n < THRESHOLD) { 237 simplesort(a, n, i, tr, endch); 238 continue; 239 } 240 241 if (nc == 0) { 242 bmin = 255; 243 for (ak = a + n; --ak >= a;) { 244 c = tr[(*ak)[i]]; 245 if (++count[c] == 1 && c != endch) { 246 if (c < bmin) 247 bmin = c; 248 nc++; 249 } 250 } 251 if (sp + nc > s + SIZE) { 252 r_sort_b(a, ta, n, i, tr, endch); 253 continue; 254 } 255 } 256 257 sp0 = sp1 = sp; 258 bigc = 2; 259 if (endch == 0) { 260 top[0] = ak = a + count[0]; 261 count[0] = 0; 262 } else { 263 ak = a; 264 top[255] = a + n; 265 count[255] = 0; 266 } 267 for (cp = count + bmin; nc > 0; cp++) { 268 while (*cp == 0) 269 cp++; 270 if ((c = *cp) > 1) { 271 if (c > bigc) { 272 bigc = c; 273 sp1 = sp; 274 } 275 push(ak, c, i+1); 276 } 277 top[cp-count] = ak += c; 278 *cp = 0; /* Reset count[]. */ 279 nc--; 280 } 281 swap(*sp0, *sp1, temp); 282 283 for (ak = ta + n, ai = a+n; ak > ta;) /* Copy to temp. */ 284 *--ak = *--ai; 285 for (ak = ta+n; --ak >= ta;) /* Deal to piles. */ 286 *--top[tr[(*ak)[i]]] = *ak; 287 } 288 } 289 290 /* insertion sort */ 291 static inline void 292 simplesort(const u_char **a, int n, int b, const u_char *tr, u_int endch) 293 { 294 u_char ch; 295 const u_char **ak, **ai, *s, *t; 296 297 for (ak = a+1; --n >= 1; ak++) 298 for (ai = ak; ai > a; ai--) { 299 for (s = ai[0] + b, t = ai[-1] + b; 300 (ch = tr[*s]) != endch; s++, t++) 301 if (ch != tr[*t]) 302 break; 303 if (ch >= tr[*t]) 304 break; 305 swap(ai[0], ai[-1], s); 306 } 307 } 308