Annotation of src/usr.bin/sort/fsort.c, Revision 1.3
1.3 ! millert 1: /* $OpenBSD: fsort.c,v 1.2 1997/01/22 05:49:19 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.
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: #ifndef lint
40: #if 0
41: static char sccsid[] = "@(#)fsort.c 8.1 (Berkeley) 6/6/93";
42: #else
1.3 ! millert 43: static char rcsid[] = "$OpenBSD: fsort.c,v 1.2 1997/01/22 05:49:19 millert Exp $";
1.1 millert 44: #endif
45: #endif /* not lint */
46:
47: /*
48: * Read in the next bin. If it fits in one segment sort it;
49: * otherwise refine it by segment deeper by one character,
50: * and try again on smaller bins. Sort the final bin at this level
51: * of recursion to keep the head of fstack at 0.
52: * After PANIC passes, abort to merge sort.
53: */
54: #include "sort.h"
55: #include "fsort.h"
56:
57: #include <stdlib.h>
58: #include <string.h>
59:
60: u_char **keylist = 0, *buffer = 0, *linebuf = 0;
61: struct tempfile fstack[MAXFCT];
62: extern char *toutpath;
63: #define FSORTMAX 4
64: int PANIC = FSORTMAX;
65:
66: void
67: fsort(binno, depth, infiles, nfiles, outfp, ftbl)
68: register int binno, depth, nfiles;
69: register union f_handle infiles;
70: FILE *outfp;
71: register struct field *ftbl;
72: {
73: register u_char *bufend, **keypos, *tmpbuf;
74: u_char *weights;
75: int ntfiles, mfct = 0, total, i, maxb, lastb, panic = 0;
76: register int c, nelem;
77: long sizes [NBINS+1];
78: union f_handle tfiles, mstart = {MAXFCT-16};
79: register int (*get)(int, union f_handle, int, RECHEADER *,
80: u_char *, struct field *);
81: register struct recheader *crec;
82: struct field tfield[2];
83: FILE *prevfp, *tailfp[FSORTMAX+1];
84:
85: memset(tailfp, 0, sizeof(tailfp));
86: prevfp = outfp;
87: memset(tfield, 0, sizeof(tfield));
88: if (ftbl[0].flags & R)
89: tfield[0].weights = Rascii;
90: else
91: tfield[0].weights = ascii;
92: tfield[0].icol.num = 1;
93: weights = ftbl[0].weights;
94: if (!buffer) {
1.3 ! millert 95: buffer = malloc(BUFSIZE + 1);
1.1 millert 96: keylist = malloc(MAXNUM * sizeof(u_char *));
97: if (!SINGL_FLD)
98: linebuf = malloc(MAXLLEN);
99: }
1.3 ! millert 100: bufend = buffer + BUFSIZE;
1.1 millert 101: if (binno >= 0) {
102: tfiles.top = infiles.top + nfiles;
103: get = getnext;
104: } else {
105: tfiles.top = 0;
106: if (SINGL_FLD)
107: get = makeline;
108: else
109: get = makekey;
110: }
111: for (;;) {
112: memset(sizes, 0, sizeof(sizes));
113: c = ntfiles = 0;
114: if (binno == weights[REC_D] &&
115: !(SINGL_FLD && ftbl[0].flags & F)) { /* pop */
116: rd_append(weights[REC_D],
117: infiles, nfiles, prevfp, buffer, bufend);
118: break;
119: } else if (binno == weights[REC_D]) {
120: depth = 0; /* start over on flat weights */
121: ftbl = tfield;
122: weights = ftbl[0].weights;
123: }
124: while (c != EOF) {
125: keypos = keylist;
126: nelem = 0;
127: crec = (RECHEADER *) buffer;
128: while((c = get(binno, infiles, nfiles, crec, bufend,
129: ftbl)) == 0) {
130: *keypos++ = crec->data + depth;
131: if (++nelem == MAXNUM) {
132: c = BUFFEND;
133: break;
134: }
135: crec =(RECHEADER *) ((char *) crec +
136: SALIGN(crec->length) + sizeof(TRECHEADER));
137: }
138: if (c == BUFFEND || ntfiles || mfct) { /* push */
139: if (panic >= PANIC) {
140: fstack[MAXFCT-16+mfct].fp = ftmp();
141: if (radixsort((const u_char **)keylist,
142: nelem, weights, REC_D))
143: err(2, NULL);
144: append(keylist, nelem, depth, fstack[
145: MAXFCT-16+mfct].fp, putrec, ftbl);
146: mfct++;
147: /* reduce number of open files */
148: if (mfct == 16 ||(c == EOF && ntfiles)) {
149: tmpbuf = malloc(bufend -
150: crec->data);
151: memmove(tmpbuf, crec->data,
152: bufend - crec->data);
153: fstack[tfiles.top + ntfiles].fp
154: = ftmp();
155: fmerge(0, mstart, mfct, geteasy,
156: fstack[tfiles.top+ntfiles].fp,
157: putrec, ftbl);
158: ++ntfiles;
159: mfct = 0;
160: memmove(crec->data, tmpbuf,
161: bufend - crec->data);
162: free(tmpbuf);
163: }
164: } else {
165: fstack[tfiles.top + ntfiles].fp= ftmp();
166: onepass(keylist, depth, nelem, sizes,
167: weights, fstack[tfiles.top+ntfiles].fp);
168: ++ntfiles;
169: }
170: }
171: }
172: get = getnext;
173: if (!ntfiles && !mfct) { /* everything in memory--pop */
174: if (nelem > 1 && radixsort((const u_char **)keylist,
175: nelem, weights, REC_D))
176: err(2, NULL);
177: append(keylist, nelem, depth, outfp, putline, ftbl);
178: break; /* pop */
179: }
180: if (panic >= PANIC) {
181: if (!ntfiles)
182: fmerge(0, mstart, mfct, geteasy,
183: outfp, putline, ftbl);
184: else
185: fmerge(0, tfiles, ntfiles, geteasy,
186: outfp, putline, ftbl);
187: break;
188:
189: }
190: total = maxb = lastb = 0; /* find if one bin dominates */
191: for (i = 0; i < NBINS; i++)
192: if (sizes[i]) {
193: if (sizes[i] > sizes[maxb])
194: maxb = i;
195: lastb = i;
196: total += sizes[i];
197: }
198: if (sizes[maxb] < max((total / 2) , BUFSIZE))
199: maxb = lastb; /* otherwise pop after last bin */
200: fstack[tfiles.top].lastb = lastb;
201: fstack[tfiles.top].maxb = maxb;
202:
203: /* start refining next level. */
204: get(-1, tfiles, ntfiles, crec, bufend, 0); /* rewind */
205: for (i = 0; i < maxb; i++) {
206: if (!sizes[i]) /* bin empty; step ahead file offset */
207: get(i, tfiles, ntfiles, crec, bufend, 0);
208: else
209: fsort(i, depth+1, tfiles, ntfiles, outfp, ftbl);
210: }
211: if (lastb != maxb) {
212: if (prevfp != outfp)
213: tailfp[panic] = prevfp;
214: prevfp = ftmp();
215: for (i = maxb+1; i <= lastb; i++)
216: if (!sizes[i])
217: get(i, tfiles, ntfiles, crec, bufend,0);
218: else
219: fsort(i, depth+1, tfiles, ntfiles,
220: prevfp, ftbl);
221: }
222:
223: /* sort biggest (or last) bin at this level */
224: depth++;
225: panic++;
226: binno = maxb;
227: infiles.top = tfiles.top; /* getnext will free tfiles, */
228: nfiles = ntfiles; /* so overwrite them */
229: }
230: if (prevfp != outfp) {
231: concat(outfp, prevfp);
232: fclose(prevfp);
233: }
234: for (i = panic; i >= 0; --i)
235: if (tailfp[i]) {
236: concat(outfp, tailfp[i]);
237: fclose(tailfp[i]);
238: }
239: }
240:
241: /*
242: This is one pass of radix exchange, dumping the bins to disk.
243: */
244: #define swap(a, b, t) t = a, a = b, b = t
245: void
246: onepass(a, depth, n, sizes, tr, fp)
247: u_char **a;
248: int depth;
249: long n, sizes[];
250: u_char *tr;
251: FILE *fp;
252: {
253: long tsizes[NBINS+1];
254: u_char **bin[257], **top[256], ***bp, ***bpmax, ***tp;
255: static histo[256];
256: int *hp;
257: register int c;
258: u_char **an, *t, **aj;
259: register u_char **ak, *r;
260:
261: memset(tsizes, 0, sizeof(tsizes));
262: depth += sizeof(TRECHEADER);
263: an = a + n;
264: for (ak = a; ak < an; ak++) {
265: histo[c = tr[**ak]]++;
266: tsizes[c] += ((RECHEADER *) (*ak -= depth))->length;
267: }
268:
269: bin[0] = a;
270: bpmax = bin + 256;
271: tp = top, hp = histo;
272: for (bp = bin; bp < bpmax; bp++) {
273: *tp++ = *(bp+1) = *bp + (c = *hp);
274: *hp++ = 0;
275: if (c <= 1)
276: continue;
277: }
278: for(aj = a; aj < an; *aj = r, aj = bin[c+1])
279: for(r = *aj; aj < (ak = --top[c = tr[r[depth]]]) ;)
280: swap(*ak, r, t);
281:
282: for (ak = a, c = 0; c < 256; c++) {
283: an = bin[c+1];
284: n = an - ak;
285: tsizes[c] += n * sizeof(TRECHEADER);
286: /* tell getnext how many elements in this bin, this segment. */
287: EWRITE(tsizes+c, sizeof(long), 1, fp);
288: sizes[c] += tsizes[c];
289: for (; ak < an; ++ak)
290: putrec((RECHEADER *) *ak, fp);
291: }
292: }