Annotation of src/usr.bin/sort/fsort.c, Revision 1.15
1.14 naddy 1: /* $OpenBSD: fsort.c,v 1.13 2006/10/18 23:30:43 millert Exp $ */
1.1 millert 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.
1.9 millert 18: * 3. Neither the name of the University nor the names of its contributors
1.1 millert 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: #ifndef lint
36: #if 0
37: static char sccsid[] = "@(#)fsort.c 8.1 (Berkeley) 6/6/93";
38: #else
1.14 naddy 39: static char rcsid[] = "$OpenBSD: fsort.c,v 1.13 2006/10/18 23:30:43 millert Exp $";
1.1 millert 40: #endif
41: #endif /* not lint */
42:
43: /*
44: * Read in the next bin. If it fits in one segment sort it;
45: * otherwise refine it by segment deeper by one character,
46: * and try again on smaller bins. Sort the final bin at this level
47: * of recursion to keep the head of fstack at 0.
48: * After PANIC passes, abort to merge sort.
1.4 millert 49: */
1.1 millert 50: #include "sort.h"
51: #include "fsort.h"
52:
53: #include <stdlib.h>
54: #include <string.h>
55:
1.15 ! millert 56: u_char *buffer = NULL, *bufend = NULL, *linebuf = NULL;
! 57: size_t bufsize = BUFSIZE, linebuf_size = MAXLLEN;
1.1 millert 58: struct tempfile fstack[MAXFCT];
1.12 deraadt 59: extern char toutpath[];
1.1 millert 60: #define FSORTMAX 4
61: int PANIC = FSORTMAX;
62:
63: void
1.11 deraadt 64: fsort(int binno, int depth, union f_handle infiles, int nfiles, FILE *outfp,
65: struct field *ftbl)
1.1 millert 66: {
1.15 ! millert 67: u_char *weights, **keypos, *tmpbuf;
! 68: static u_char **keylist;
1.1 millert 69: int ntfiles, mfct = 0, total, i, maxb, lastb, panic = 0;
1.5 millert 70: int c, nelem;
1.8 ericj 71: long sizes[NBINS+1];
1.1 millert 72: union f_handle tfiles, mstart = {MAXFCT-16};
1.8 ericj 73: int (*get)(int, union f_handle, int, RECHEADER *,
1.1 millert 74: u_char *, struct field *);
1.8 ericj 75: RECHEADER *crec;
1.1 millert 76: struct field tfield[2];
77: FILE *prevfp, *tailfp[FSORTMAX+1];
78:
79: memset(tailfp, 0, sizeof(tailfp));
80: prevfp = outfp;
81: memset(tfield, 0, sizeof(tfield));
82: if (ftbl[0].flags & R)
83: tfield[0].weights = Rascii;
84: else
85: tfield[0].weights = ascii;
86: tfield[0].icol.num = 1;
87: weights = ftbl[0].weights;
1.15 ! millert 88: if (keylist == NULL) {
! 89: if ((keylist = calloc(MAXNUM, sizeof(u_char *))) == NULL)
! 90: err(2, NULL);
1.14 naddy 91: }
1.1 millert 92: if (binno >= 0) {
93: tfiles.top = infiles.top + nfiles;
94: get = getnext;
95: } else {
96: tfiles.top = 0;
97: if (SINGL_FLD)
98: get = makeline;
99: else
100: get = makekey;
101: }
102: for (;;) {
103: memset(sizes, 0, sizeof(sizes));
104: c = ntfiles = 0;
105: if (binno == weights[REC_D] &&
106: !(SINGL_FLD && ftbl[0].flags & F)) { /* pop */
107: rd_append(weights[REC_D],
108: infiles, nfiles, prevfp, buffer, bufend);
109: break;
110: } else if (binno == weights[REC_D]) {
111: depth = 0; /* start over on flat weights */
112: ftbl = tfield;
113: weights = ftbl[0].weights;
114: }
115: while (c != EOF) {
116: keypos = keylist;
117: nelem = 0;
118: crec = (RECHEADER *) buffer;
119: while((c = get(binno, infiles, nfiles, crec, bufend,
120: ftbl)) == 0) {
121: *keypos++ = crec->data + depth;
122: if (++nelem == MAXNUM) {
123: c = BUFFEND;
124: break;
125: }
126: crec =(RECHEADER *) ((char *) crec +
127: SALIGN(crec->length) + sizeof(TRECHEADER));
128: }
1.8 ericj 129: /*
130: * buffer was too small for data, allocate
131: * a bigger buffer.
132: */
133: if (c == BUFFEND && nelem == 0) {
134: bufsize *= 2;
1.15 ! millert 135: tmpbuf = realloc(buffer, bufsize);
! 136: if (!tmpbuf)
1.8 ericj 137: err(2, "failed to realloc buffer");
1.15 ! millert 138: crec = (RECHEADER *)
! 139: (tmpbuf + ((u_char *)crec - buffer));
! 140: buffer = tmpbuf;
! 141: bufend = buffer + bufsize - 1;
1.8 ericj 142: continue;
143: }
1.1 millert 144: if (c == BUFFEND || ntfiles || mfct) { /* push */
145: if (panic >= PANIC) {
146: fstack[MAXFCT-16+mfct].fp = ftmp();
147: if (radixsort((const u_char **)keylist,
148: nelem, weights, REC_D))
149: err(2, NULL);
150: append(keylist, nelem, depth, fstack[
151: MAXFCT-16+mfct].fp, putrec, ftbl);
152: mfct++;
153: /* reduce number of open files */
154: if (mfct == 16 ||(c == EOF && ntfiles)) {
155: fstack[tfiles.top + ntfiles].fp
156: = ftmp();
157: fmerge(0, mstart, mfct, geteasy,
158: fstack[tfiles.top+ntfiles].fp,
159: putrec, ftbl);
1.4 millert 160: ntfiles++;
1.1 millert 161: mfct = 0;
162: }
163: } else {
164: fstack[tfiles.top + ntfiles].fp= ftmp();
165: onepass(keylist, depth, nelem, sizes,
166: weights, fstack[tfiles.top+ntfiles].fp);
1.4 millert 167: ntfiles++;
1.1 millert 168: }
169: }
170: }
171: get = getnext;
172: if (!ntfiles && !mfct) { /* everything in memory--pop */
173: if (nelem > 1 && radixsort((const u_char **)keylist,
174: nelem, weights, REC_D))
175: err(2, NULL);
176: append(keylist, nelem, depth, outfp, putline, ftbl);
177: break; /* pop */
178: }
179: if (panic >= PANIC) {
180: if (!ntfiles)
181: fmerge(0, mstart, mfct, geteasy,
182: outfp, putline, ftbl);
183: else
184: fmerge(0, tfiles, ntfiles, geteasy,
185: outfp, putline, ftbl);
186: break;
187: }
188: total = maxb = lastb = 0; /* find if one bin dominates */
189: for (i = 0; i < NBINS; i++)
190: if (sizes[i]) {
191: if (sizes[i] > sizes[maxb])
192: maxb = i;
193: lastb = i;
194: total += sizes[i];
195: }
196: if (sizes[maxb] < max((total / 2) , BUFSIZE))
197: maxb = lastb; /* otherwise pop after last bin */
198: fstack[tfiles.top].lastb = lastb;
199: fstack[tfiles.top].maxb = maxb;
200:
201: /* start refining next level. */
202: get(-1, tfiles, ntfiles, crec, bufend, 0); /* rewind */
203: for (i = 0; i < maxb; i++) {
204: if (!sizes[i]) /* bin empty; step ahead file offset */
205: get(i, tfiles, ntfiles, crec, bufend, 0);
206: else
207: fsort(i, depth+1, tfiles, ntfiles, outfp, ftbl);
208: }
209: if (lastb != maxb) {
210: if (prevfp != outfp)
211: tailfp[panic] = prevfp;
212: prevfp = ftmp();
213: for (i = maxb+1; i <= lastb; i++)
214: if (!sizes[i])
215: get(i, tfiles, ntfiles, crec, bufend,0);
216: else
217: fsort(i, depth+1, tfiles, ntfiles,
218: prevfp, ftbl);
219: }
220:
221: /* sort biggest (or last) bin at this level */
222: depth++;
223: panic++;
224: binno = maxb;
225: infiles.top = tfiles.top; /* getnext will free tfiles, */
226: nfiles = ntfiles; /* so overwrite them */
227: }
228: if (prevfp != outfp) {
229: concat(outfp, prevfp);
230: fclose(prevfp);
231: }
232: for (i = panic; i >= 0; --i)
233: if (tailfp[i]) {
234: concat(outfp, tailfp[i]);
235: fclose(tailfp[i]);
236: }
237: }
238:
239: /*
1.4 millert 240: * This is one pass of radix exchange, dumping the bins to disk.
1.1 millert 241: */
242: #define swap(a, b, t) t = a, a = b, b = t
243: void
1.11 deraadt 244: onepass(u_char **a, int depth, long n, long sizes[], u_char *tr, FILE *fp)
1.1 millert 245: {
1.8 ericj 246: size_t tsizes[NBINS+1];
1.1 millert 247: u_char **bin[257], **top[256], ***bp, ***bpmax, ***tp;
1.7 millert 248: static int histo[256];
1.1 millert 249: int *hp;
1.8 ericj 250: int c;
1.1 millert 251: u_char **an, *t, **aj;
1.8 ericj 252: u_char **ak, *r;
1.1 millert 253:
254: memset(tsizes, 0, sizeof(tsizes));
255: depth += sizeof(TRECHEADER);
1.8 ericj 256: an = &a[n];
1.1 millert 257: for (ak = a; ak < an; ak++) {
258: histo[c = tr[**ak]]++;
259: tsizes[c] += ((RECHEADER *) (*ak -= depth))->length;
260: }
261:
262: bin[0] = a;
263: bpmax = bin + 256;
264: tp = top, hp = histo;
265: for (bp = bin; bp < bpmax; bp++) {
266: *tp++ = *(bp+1) = *bp + (c = *hp);
267: *hp++ = 0;
268: if (c <= 1)
269: continue;
270: }
1.4 millert 271: for (aj = a; aj < an; *aj = r, aj = bin[c+1])
1.1 millert 272: for(r = *aj; aj < (ak = --top[c = tr[r[depth]]]) ;)
273: swap(*ak, r, t);
274:
275: for (ak = a, c = 0; c < 256; c++) {
276: an = bin[c+1];
277: n = an - ak;
278: tsizes[c] += n * sizeof(TRECHEADER);
279: /* tell getnext how many elements in this bin, this segment. */
1.8 ericj 280: EWRITE(&tsizes[c], sizeof(size_t), 1, fp);
1.1 millert 281: sizes[c] += tsizes[c];
282: for (; ak < an; ++ak)
283: putrec((RECHEADER *) *ak, fp);
284: }
285: }