fsort.c revision 1.12 1 /* $NetBSD: fsort.c,v 1.12 2001/02/05 14:25:34 itojun Exp $ */
2
3 /*-
4 * Copyright (c) 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.
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. All advertising materials mentioning features or use of this software
19 * must display the following acknowledgement:
20 * This product includes software developed by the University of
21 * California, Berkeley and its contributors.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 */
38
39 /*
40 * Read in the next bin. If it fits in one segment sort it;
41 * otherwise refine it by segment deeper by one character,
42 * and try again on smaller bins. Sort the final bin at this level
43 * of recursion to keep the head of fstack at 0.
44 * After PANIC passes, abort to merge sort.
45 */
46 #include "sort.h"
47 #include "fsort.h"
48
49 #ifndef lint
50 __RCSID("$NetBSD: fsort.c,v 1.12 2001/02/05 14:25:34 itojun Exp $");
51 __SCCSID("@(#)fsort.c 8.1 (Berkeley) 6/6/93");
52 #endif /* not lint */
53
54 #include <stdlib.h>
55 #include <string.h>
56
57 const u_char **keylist = 0;
58 u_char *buffer = 0, *linebuf = 0;
59 size_t bufsize = DEFLLEN;
60 size_t linebuf_size;
61 struct tempfile fstack[MAXFCT];
62 extern char *toutpath;
63 #define FSORTMAX 4
64 int PANIC = FSORTMAX;
65
66 #define MSTART (MAXFCT - MERGE_FNUM)
67
68 void
69 fsort(binno, depth, top, filelist, nfiles, outfp, ftbl)
70 int binno, depth, top;
71 struct filelist *filelist;
72 int nfiles;
73 FILE *outfp;
74 struct field *ftbl;
75 {
76 const u_char **keypos;
77 u_char *bufend, *tmpbuf;
78 u_char *weights;
79 int ntfiles, mfct = 0, total, i, maxb, lastb, panic = 0;
80 int c, nelem, base;
81 long sizes [NBINS+1];
82 get_func_t get;
83 struct recheader *crec;
84 struct field tfield[2];
85 FILE *prevfp, *tailfp[FSORTMAX+1];
86
87 memset(tailfp, 0, sizeof(tailfp));
88 prevfp = outfp;
89 memset(tfield, 0, sizeof(tfield));
90 if (ftbl[0].flags & R)
91 tfield[0].weights = Rascii;
92 else
93 tfield[0].weights = ascii;
94 tfield[0].icol.num = 1;
95 weights = ftbl[0].weights;
96 if (!buffer) {
97 buffer = malloc(bufsize);
98 keylist = malloc(MAXNUM * sizeof(u_char *));
99 memset(keylist, 0, MAXNUM * sizeof(u_char *));
100 if (!SINGL_FLD) {
101 linebuf_size = DEFLLEN;
102 linebuf = malloc(linebuf_size);
103 }
104 }
105 bufend = buffer + bufsize;
106 if (binno >= 0) {
107 base = top + nfiles;
108 get = getnext;
109 } else {
110 base = 0;
111 if (SINGL_FLD)
112 get = makeline;
113 else
114 get = makekey;
115 }
116 for (;;) {
117 memset(sizes, 0, sizeof(sizes));
118 c = ntfiles = 0;
119 if (binno == weights[REC_D] &&
120 !(SINGL_FLD && ftbl[0].flags & F)) { /* pop */
121 rd_append(weights[REC_D], top,
122 nfiles, prevfp, buffer, bufend);
123 break;
124 } else if (binno == weights[REC_D]) {
125 depth = 0; /* start over on flat weights */
126 ftbl = tfield;
127 weights = ftbl[0].weights;
128 }
129 while (c != EOF) {
130 keypos = keylist;
131 nelem = 0;
132 crec = (RECHEADER *) buffer;
133
134 do_read:
135 while((c = get(binno, top, filelist, nfiles, crec,
136 bufend, ftbl)) == 0) {
137 *keypos++ = crec->data + depth;
138 if (++nelem == MAXNUM) {
139 c = BUFFEND;
140 break;
141 }
142 crec =(RECHEADER *) ((char *) crec +
143 SALIGN(crec->length) + sizeof(TRECHEADER));
144 }
145 if (c == BUFFEND && nelem < min(9, MAXNUM)) {
146 const u_char **keyp;
147 u_char *oldb = buffer;
148
149 /* buffer was too small for data, allocate
150 * bigger buffer */
151 bufsize *= 2;
152 buffer = realloc(buffer, bufsize);
153 if (!buffer) {
154 err(2, "failed to realloc buffer to %ld bytes",
155 (unsigned long) bufsize);
156 }
157 bufend = buffer + bufsize;
158
159 /* patch up keylist[] */
160 for(keyp = &keypos[-1]; keyp >= keylist; keyp--)
161 *keyp = buffer + (*keyp - oldb);
162
163 crec = (RECHEADER *) (buffer + ((u_char *)crec - oldb));
164 goto do_read;
165 }
166 if (c == BUFFEND || ntfiles || mfct) { /* push */
167 if (panic >= PANIC) {
168 fstack[MSTART + mfct].fp = ftmp();
169 if ((stable_sort)
170 ? sradixsort(keylist, nelem,
171 weights, REC_D)
172 : radixsort(keylist, nelem,
173 weights, REC_D) )
174 err(2, NULL);
175 append(keylist, nelem, depth,
176 fstack[MSTART + mfct].fp, putrec,
177 ftbl);
178 mfct++;
179 /* reduce number of open files */
180 if (mfct == MERGE_FNUM ||(c == EOF && ntfiles)) {
181 tmpbuf = malloc(bufend -
182 crec->data);
183 memmove(tmpbuf, crec->data,
184 bufend - crec->data);
185 fstack[base + ntfiles].fp
186 = ftmp();
187 fmerge(0, MSTART, filelist,
188 mfct, geteasy,
189 fstack[base].fp,
190 putrec, ftbl);
191 ++ntfiles;
192 mfct = 0;
193 memmove(crec->data, tmpbuf,
194 bufend - crec->data);
195 free(tmpbuf);
196 }
197 } else {
198 fstack[base + ntfiles].fp= ftmp();
199 onepass(keylist, depth, nelem, sizes,
200 weights, fstack[base + ntfiles].fp);
201 ++ntfiles;
202 }
203 }
204 }
205 if (!ntfiles && !mfct) { /* everything in memory--pop */
206 if (nelem > 1) {
207 if ((stable_sort)
208 ? sradixsort(keylist, nelem, weights, REC_D)
209 : radixsort(keylist, nelem, weights, REC_D) )
210 err(2, NULL);
211 }
212 append(keylist, nelem, depth, outfp, putline, ftbl);
213 break; /* pop */
214 }
215 if (panic >= PANIC) {
216 if (!ntfiles)
217 fmerge(0, MSTART, filelist, mfct, geteasy,
218 outfp, putline, ftbl);
219 else
220 fmerge(0, base, filelist, ntfiles, geteasy,
221 outfp, putline, ftbl);
222 break;
223
224 }
225 total = maxb = lastb = 0; /* find if one bin dominates */
226 for (i = 0; i < NBINS; i++)
227 if (sizes[i]) {
228 if (sizes[i] > sizes[maxb])
229 maxb = i;
230 lastb = i;
231 total += sizes[i];
232 }
233 if (sizes[maxb] < max((total / 2) , BUFSIZE))
234 maxb = lastb; /* otherwise pop after last bin */
235 fstack[base].lastb = lastb;
236 fstack[base].maxb = maxb;
237
238 /* start refining next level. */
239 getnext(-1, base, NULL, ntfiles, crec, bufend, 0); /* rewind */
240 for (i = 0; i < maxb; i++) {
241 if (!sizes[i]) /* bin empty; step ahead file offset */
242 getnext(i, base, NULL,ntfiles, crec, bufend, 0);
243 else {
244 fsort(i, depth+1, base, filelist, ntfiles,
245 outfp, ftbl);
246 }
247 }
248
249 get = getnext;
250
251 if (lastb != maxb) {
252 if (prevfp != outfp)
253 tailfp[panic] = prevfp;
254 prevfp = ftmp();
255 for (i = maxb+1; i <= lastb; i++)
256 if (!sizes[i]) {
257 getnext(i, base, NULL, ntfiles, crec,
258 bufend,0);
259 } else {
260 fsort(i, depth+1, base, filelist,
261 ntfiles, prevfp, ftbl);
262 }
263 }
264
265 /* sort biggest (or last) bin at this level */
266 depth++;
267 panic++;
268 binno = maxb;
269 top = base;
270 nfiles = ntfiles; /* so overwrite them */
271 }
272 if (prevfp != outfp) {
273 concat(outfp, prevfp);
274 fclose(prevfp);
275 }
276 for (i = panic; i >= 0; --i)
277 if (tailfp[i]) {
278 concat(outfp, tailfp[i]);
279 fclose(tailfp[i]);
280 }
281 }
282
283 /*
284 This is one pass of radix exchange, dumping the bins to disk.
285 */
286 #define swap(a, b, t) t = a, a = b, b = t
287 void
288 onepass(a, depth, n, sizes, tr, fp)
289 const u_char **a;
290 int depth;
291 long n, sizes[];
292 u_char *tr;
293 FILE *fp;
294 {
295 size_t tsizes[NBINS+1];
296 const u_char **bin[257], ***bp, ***bpmax, **top[256], ***tp;
297 static int histo[256];
298 int *hp;
299 int c;
300 const u_char **an, *t, **aj;
301 const u_char **ak, *r;
302
303 memset(tsizes, 0, sizeof(tsizes));
304 depth += sizeof(TRECHEADER);
305 an = &a[n];
306 for (ak = a; ak < an; ak++) {
307 histo[c = tr[**ak]]++;
308 tsizes[c] += ((const RECHEADER *) (*ak -= depth))->length;
309 }
310
311 bin[0] = a;
312 bpmax = bin + 256;
313 tp = top, hp = histo;
314 for (bp = bin; bp < bpmax; bp++) {
315 *tp++ = *(bp+1) = *bp + (c = *hp);
316 *hp++ = 0;
317 if (c <= 1)
318 continue;
319 }
320 for (aj = a; aj < an; *aj = r, aj = bin[c+1])
321 for (r = *aj; aj < (ak = --top[c = tr[r[depth]]]) ;)
322 swap(*ak, r, t);
323
324 for (ak = a, c = 0; c < 256; c++) {
325 an = bin[c+1];
326 n = an - ak;
327 tsizes[c] += n * sizeof(TRECHEADER);
328 /* tell getnext how many elements in this bin, this segment. */
329 EWRITE(&tsizes[c], sizeof(size_t), 1, fp);
330 sizes[c] += tsizes[c];
331 for (; ak < an; ++ak)
332 putrec((const RECHEADER *) *ak, fp);
333 }
334 }
335