Annotation of src/usr.bin/sort/file.c, Revision 1.7
1.7 ! tobias 1: /* $OpenBSD: file.c,v 1.6 2015/04/01 19:06:18 millert Exp $ */
1.1 millert 2:
3: /*-
4: * Copyright (C) 2009 Gabor Kovesdan <gabor@FreeBSD.org>
5: * Copyright (C) 2012 Oleg Moskalenko <mom040267@gmail.com>
6: * All rights reserved.
7: *
8: * Redistribution and use in source and binary forms, with or without
9: * modification, are permitted provided that the following conditions
10: * are met:
11: * 1. Redistributions of source code must retain the above copyright
12: * notice, this list of conditions and the following disclaimer.
13: * 2. Redistributions in binary form must reproduce the above copyright
14: * notice, this list of conditions and the following disclaimer in the
15: * documentation and/or other materials provided with the distribution.
16: *
17: * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
18: * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19: * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20: * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
21: * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22: * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23: * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24: * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25: * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26: * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27: * SUCH DAMAGE.
28: */
29:
30: #include <sys/mman.h>
31: #include <sys/stat.h>
32: #include <sys/types.h>
33: #include <sys/queue.h>
34:
35: #include <err.h>
36: #include <fcntl.h>
37: #include <stdio.h>
38: #include <stdlib.h>
39: #include <string.h>
40: #include <unistd.h>
41: #include <wchar.h>
42: #include <wctype.h>
43:
44: #include "coll.h"
45: #include "file.h"
46: #include "radixsort.h"
47:
48: unsigned long long free_memory = 1000000;
49: unsigned long long available_free_memory = 1000000;
50:
51: bool use_mmap;
52:
53: const char *tmpdir = "/var/tmp";
54: const char *compress_program;
55:
56: size_t max_open_files = 16;
57:
58: /*
59: * How much space we read from file at once
60: */
61: #define READ_CHUNK 4096
62:
63: /*
64: * File reader structure
65: */
66: struct file_reader {
67: struct reader_buffer rb;
68: FILE *file;
69: char *fname;
70: unsigned char *buffer;
71: unsigned char *mmapaddr;
72: unsigned char *mmapptr;
73: size_t bsz;
74: size_t cbsz;
75: size_t mmapsize;
76: size_t strbeg;
77: int fd;
78: char elsymb;
79: };
80:
81: /*
82: * Structure to be used in file merge process.
83: */
84: struct file_header {
85: struct file_reader *fr;
86: struct sort_list_item *si; /* current top line */
87: size_t file_pos;
88: };
89:
90: /*
91: * List elements of "cleanable" files list.
92: */
93: struct CLEANABLE_FILE {
94: char *fn;
95: LIST_ENTRY(CLEANABLE_FILE) files;
96: };
97:
98: /*
99: * List header of "cleanable" files list.
100: */
101: static LIST_HEAD(CLEANABLE_FILES, CLEANABLE_FILE) tmp_files;
102:
103: /*
104: * Init tmp files list
105: */
106: void
107: init_tmp_files(void)
108: {
109:
110: LIST_INIT(&tmp_files);
111: }
112:
113: /*
114: * Save name of a tmp file for signal cleanup
115: */
116: void
117: tmp_file_atexit(const char *tmp_file)
118: {
119:
120: if (tmp_file) {
121: struct CLEANABLE_FILE *item =
122: sort_malloc(sizeof(struct CLEANABLE_FILE));
123: item->fn = sort_strdup(tmp_file);
124: LIST_INSERT_HEAD(&tmp_files, item, files);
125: }
126: }
127:
128: /*
129: * Clear tmp files
130: */
131: void
132: clear_tmp_files(void)
133: {
134: struct CLEANABLE_FILE *item;
135:
136: LIST_FOREACH(item, &tmp_files, files) {
137: if (item != NULL && item->fn != NULL)
138: unlink(item->fn);
139: }
140: }
141:
142: /*
143: * Check whether a file is a temporary file
144: */
145: static bool
146: file_is_tmp(const char *fn)
147: {
148: struct CLEANABLE_FILE *item;
149: bool ret = false;
150:
151: if (fn) {
152: LIST_FOREACH(item, &tmp_files, files) {
153: if (item != NULL && item->fn != NULL)
154: if (strcmp(item->fn, fn) == 0) {
155: ret = true;
156: break;
157: }
158: }
159: }
160:
161: return ret;
162: }
163:
164: /*
165: * Generate new temporary file name
166: */
167: char *
168: new_tmp_file_name(void)
169: {
170: char *ret;
1.7 ! tobias 171: int fd;
1.1 millert 172:
1.7 ! tobias 173: sort_asprintf(&ret, "%s/.bsdsort.XXXXXXXXXX", tmpdir);
! 174: if ((fd = mkstemp(ret)) == -1)
! 175: err(2, "%s", ret);
! 176: close(fd);
1.1 millert 177: tmp_file_atexit(ret);
178: return ret;
179: }
180:
181: /*
182: * Initialize file list
183: */
184: void
185: file_list_init(struct file_list *fl, bool tmp)
186: {
187:
188: if (fl) {
189: fl->count = 0;
190: fl->sz = 0;
191: fl->fns = NULL;
192: fl->tmp = tmp;
193: }
194: }
195:
196: /*
197: * Add a file name to the list
198: */
199: void
200: file_list_add(struct file_list *fl, char *fn, bool allocate)
201: {
202:
203: if (fl && fn) {
204: if (fl->count >= fl->sz || fl->fns == NULL) {
205: fl->sz = fl->sz * 2 + 1;
206: fl->fns = sort_reallocarray(fl->fns, fl->sz,
207: sizeof(char *));
208: }
209: fl->fns[fl->count] = allocate ? sort_strdup(fn) : fn;
210: fl->count += 1;
211: }
212: }
213:
214: /*
215: * Populate file list from array of file names
216: */
217: void
218: file_list_populate(struct file_list *fl, int argc, char **argv, bool allocate)
219: {
220:
221: if (fl && argv) {
222: int i;
223:
224: for (i = 0; i < argc; i++)
225: file_list_add(fl, argv[i], allocate);
226: }
227: }
228:
229: /*
230: * Clean file list data and delete the files,
231: * if this is a list of temporary files
232: */
233: void
234: file_list_clean(struct file_list *fl)
235: {
236:
237: if (fl) {
238: if (fl->fns) {
239: size_t i;
240:
241: for (i = 0; i < fl->count; i++) {
242: if (fl->fns[i]) {
243: if (fl->tmp)
244: unlink(fl->fns[i]);
245: sort_free(fl->fns[i]);
246: fl->fns[i] = 0;
247: }
248: }
249: sort_free(fl->fns);
250: fl->fns = NULL;
251: }
252: fl->sz = 0;
253: fl->count = 0;
254: fl->tmp = false;
255: }
256: }
257:
258: /*
259: * Init sort list
260: */
261: void
262: sort_list_init(struct sort_list *l)
263: {
264:
265: if (l) {
266: l->count = 0;
267: l->size = 0;
268: l->memsize = sizeof(struct sort_list);
269: l->list = NULL;
270: }
271: }
272:
273: /*
274: * Add string to sort list
275: */
276: void
277: sort_list_add(struct sort_list *l, struct bwstring *str)
278: {
279:
280: if (l && str) {
281: size_t indx = l->count;
282:
283: if ((l->list == NULL) || (indx >= l->size)) {
284: size_t newsize = (l->size + 1) + 1024;
285:
286: l->list = sort_reallocarray(l->list, newsize,
287: sizeof(struct sort_list_item *));
288: l->memsize += (newsize - l->size) *
289: sizeof(struct sort_list_item *);
290: l->size = newsize;
291: }
292: l->list[indx] = sort_list_item_alloc();
293: sort_list_item_set(l->list[indx], str);
294: l->memsize += sort_list_item_size(l->list[indx]);
295: l->count += 1;
296: }
297: }
298:
299: /*
300: * Clean sort list data
301: */
302: void
303: sort_list_clean(struct sort_list *l)
304: {
305:
306: if (l) {
307: if (l->list) {
308: size_t i;
309:
310: for (i = 0; i < l->count; i++) {
311: struct sort_list_item *item;
312:
313: item = l->list[i];
314:
315: if (item) {
316: sort_list_item_clean(item);
317: sort_free(item);
318: l->list[i] = NULL;
319: }
320: }
321: sort_free(l->list);
322: l->list = NULL;
323: }
324: l->count = 0;
325: l->size = 0;
326: l->memsize = sizeof(struct sort_list);
327: }
328: }
329:
330: /*
331: * Write sort list to file
332: */
333: void
334: sort_list_dump(struct sort_list *l, const char *fn)
335: {
336:
337: if (l && fn) {
338: FILE *f;
339:
340: f = openfile(fn, "w");
341: if (f == NULL)
342: err(2, "%s", fn);
343:
344: if (l->list) {
345: size_t i;
346: if (!sort_opts_vals.uflag) {
347: for (i = 0; i < l->count; ++i)
348: bwsfwrite(l->list[i]->str, f,
349: sort_opts_vals.zflag);
350: } else {
351: struct sort_list_item *last_printed_item = NULL;
352: struct sort_list_item *item;
353: for (i = 0; i < l->count; ++i) {
354: item = l->list[i];
355: if ((last_printed_item == NULL) ||
356: list_coll(&last_printed_item, &item)) {
357: bwsfwrite(item->str, f, sort_opts_vals.zflag);
358: last_printed_item = item;
359: }
360: }
361: }
362: }
363:
364: closefile(f, fn);
365: }
366: }
367:
368: /*
369: * Checks if the given file is sorted. Stops at the first disorder,
370: * prints the disordered line and returns 1.
371: */
372: int
373: check(const char *fn)
374: {
375: struct bwstring *s1, *s2, *s1disorder, *s2disorder;
376: struct file_reader *fr;
377: struct keys_array *ka1, *ka2;
378: int res;
379: size_t pos, posdisorder;
380:
381: s1 = s2 = s1disorder = s2disorder = NULL;
382: ka1 = ka2 = NULL;
383:
384: fr = file_reader_init(fn);
385:
386: res = 0;
387: pos = 1;
388: posdisorder = 1;
389:
390: if (fr == NULL) {
391: err(2, "%s", fn);
392: goto end;
393: }
394:
395: s1 = file_reader_readline(fr);
396: if (s1 == NULL)
397: goto end;
398:
399: ka1 = keys_array_alloc();
400: preproc(s1, ka1);
401:
402: s2 = file_reader_readline(fr);
403: if (s2 == NULL)
404: goto end;
405:
406: ka2 = keys_array_alloc();
407: preproc(s2, ka2);
408:
409: for (;;) {
410:
411: if (debug_sort) {
412: bwsprintf(stdout, s2, "s1=<", ">");
413: bwsprintf(stdout, s1, "s2=<", ">");
414: }
415: int cmp = key_coll(ka2, ka1, 0);
416: if (debug_sort)
417: printf("; cmp1=%d", cmp);
418:
419: if (!cmp && sort_opts_vals.complex_sort &&
420: !(sort_opts_vals.uflag) && !(sort_opts_vals.sflag)) {
421: cmp = top_level_str_coll(s2, s1);
422: if (debug_sort)
423: printf("; cmp2=%d", cmp);
424: }
425: if (debug_sort)
426: printf("\n");
427:
428: if ((sort_opts_vals.uflag && (cmp <= 0)) || (cmp < 0)) {
429: if (!(sort_opts_vals.csilentflag)) {
430: s2disorder = bwsdup(s2);
431: posdisorder = pos;
432: if (debug_sort)
433: s1disorder = bwsdup(s1);
434: }
435: res = 1;
436: goto end;
437: }
438:
439: pos++;
440:
441: clean_keys_array(s1, ka1);
442: sort_free(ka1);
443: ka1 = ka2;
444: ka2 = NULL;
445:
446: bwsfree(s1);
447: s1 = s2;
448:
449: s2 = file_reader_readline(fr);
450: if (s2 == NULL)
451: goto end;
452:
453: ka2 = keys_array_alloc();
454: preproc(s2, ka2);
455: }
456:
457: end:
458: if (ka1) {
459: clean_keys_array(s1, ka1);
460: sort_free(ka1);
461: }
462:
463: if (s1)
464: bwsfree(s1);
465:
466: if (ka2) {
467: clean_keys_array(s2, ka2);
468: sort_free(ka2);
469: }
470:
471: if (s2)
472: bwsfree(s2);
473:
474: if (fn == NULL || *fn == 0 || strcmp(fn, "-") == 0) {
475: for (;;) {
476: s2 = file_reader_readline(fr);
477: if (s2 == NULL)
478: break;
479: bwsfree(s2);
480: }
481: }
482:
483: file_reader_free(fr);
484:
485: if (s2disorder) {
486: bws_disorder_warnx(s2disorder, fn, posdisorder);
487: if (s1disorder) {
488: bws_disorder_warnx(s1disorder, fn, posdisorder);
489: if (s1disorder != s2disorder)
490: bwsfree(s1disorder);
491: }
492: bwsfree(s2disorder);
493: s1disorder = NULL;
494: s2disorder = NULL;
495: }
496:
497: if (res)
498: exit(res);
499:
500: return 0;
501: }
502:
503: /*
504: * Opens a file. If the given filename is "-", stdout will be
505: * opened.
506: */
507: FILE *
508: openfile(const char *fn, const char *mode)
509: {
510: FILE *file;
511:
512: if (strcmp(fn, "-") == 0) {
513: return (mode && mode[0] == 'r') ? stdin : stdout;
514: } else {
515: mode_t orig_file_mask = 0;
516: int is_tmp = file_is_tmp(fn);
517:
518: if (is_tmp && (mode[0] == 'w'))
519: orig_file_mask = umask(S_IWGRP | S_IWOTH |
520: S_IRGRP | S_IROTH);
521:
522: if (is_tmp && (compress_program != NULL)) {
523: char *cmd;
524:
525: fflush(stdout);
526:
527: if (mode[0] == 'r')
1.6 millert 528: sort_asprintf(&cmd, "%s -d < %s",
1.5 tobias 529: compress_program, fn);
1.1 millert 530: else if (mode[0] == 'w')
1.6 millert 531: sort_asprintf(&cmd, "%s > %s",
1.1 millert 532: compress_program, fn);
533: else
534: err(2, "Wrong file mode");
535:
536: if ((file = popen(cmd, mode)) == NULL)
537: err(2, NULL);
538:
539: sort_free(cmd);
540:
541: } else if ((file = fopen(fn, mode)) == NULL)
542: err(2, "%s", fn);
543:
544: if (is_tmp && (mode[0] == 'w'))
545: umask(orig_file_mask);
546: }
547:
548: return file;
549: }
550:
551: /*
552: * Close file
553: */
554: void
555: closefile(FILE *f, const char *fn)
556: {
557: if (f == NULL) {
558: ;
559: } else if (f == stdin) {
560: ;
561: } else if (f == stdout) {
562: fflush(f);
563: } else {
564: if (file_is_tmp(fn) && compress_program != NULL) {
565: if (pclose(f) < 0)
566: err(2, NULL);
567: } else
568: fclose(f);
569: }
570: }
571:
572: /*
573: * Reads a file into the internal buffer.
574: */
575: struct file_reader *
576: file_reader_init(const char *fsrc)
577: {
578: struct file_reader *ret;
579:
580: if (fsrc == NULL)
581: fsrc = "-";
582:
1.2 millert 583: ret = sort_calloc(1, sizeof(struct file_reader));
1.1 millert 584:
585: ret->elsymb = '\n';
586: if (sort_opts_vals.zflag)
587: ret->elsymb = 0;
588:
589: ret->fname = sort_strdup(fsrc);
590:
591: if (strcmp(fsrc, "-") && (compress_program == NULL) && use_mmap) {
592: struct stat stat_buf;
593: void *addr;
594: size_t sz = 0;
595: int fd;
596:
597: fd = open(fsrc, O_RDONLY);
598: if (fd < 0)
599: err(2, "%s", fsrc);
600:
601: if (fstat(fd, &stat_buf) < 0)
602: err(2, "%s", fsrc);
603: sz = stat_buf.st_size;
604:
605: addr = mmap(NULL, sz, PROT_READ, 0, fd, 0);
606: if (addr == MAP_FAILED) {
607: close(fd);
608: } else {
609: ret->fd = fd;
610: ret->mmapaddr = addr;
611: ret->mmapsize = sz;
612: ret->mmapptr = ret->mmapaddr;
613: posix_madvise(addr, sz, POSIX_MADV_SEQUENTIAL);
614: }
615: }
616:
617: if (ret->mmapaddr == NULL) {
618: ret->file = openfile(fsrc, "r");
619: if (ret->file == NULL)
620: err(2, "%s", fsrc);
621:
622: if (strcmp(fsrc, "-")) {
623: ret->cbsz = READ_CHUNK;
624: ret->buffer = sort_malloc(ret->cbsz);
625: ret->bsz = 0;
626: ret->strbeg = 0;
627:
628: ret->bsz = fread(ret->buffer, 1, ret->cbsz, ret->file);
629: if (ret->bsz == 0) {
630: if (ferror(ret->file))
631: err(2, NULL);
632: }
633: }
634: }
635:
636: return ret;
637: }
638:
639: struct bwstring *
640: file_reader_readline(struct file_reader *fr)
641: {
642: struct bwstring *ret = NULL;
643:
644: if (fr->mmapaddr) {
645: unsigned char *mmapend;
646:
647: mmapend = fr->mmapaddr + fr->mmapsize;
648: if (fr->mmapptr >= mmapend)
649: return NULL;
650: else {
651: unsigned char *strend;
652: size_t sz;
653:
654: sz = mmapend - fr->mmapptr;
655: strend = memchr(fr->mmapptr, fr->elsymb, sz);
656:
657: if (strend == NULL) {
658: ret = bwscsbdup(fr->mmapptr, sz);
659: fr->mmapptr = mmapend;
660: } else {
661: ret = bwscsbdup(fr->mmapptr, strend -
662: fr->mmapptr);
663: fr->mmapptr = strend + 1;
664: }
665: }
666:
667: } else if (fr->file != stdin) {
668: unsigned char *strend;
669: size_t bsz1, remsz, search_start;
670:
671: search_start = 0;
672: remsz = 0;
673: strend = NULL;
674:
675: if (fr->bsz > fr->strbeg)
676: remsz = fr->bsz - fr->strbeg;
677:
678: /* line read cycle */
679: for (;;) {
680: if (remsz > search_start)
681: strend = memchr(fr->buffer + fr->strbeg +
682: search_start, fr->elsymb, remsz -
683: search_start);
684: else
685: strend = NULL;
686:
687: if (strend)
688: break;
689: if (feof(fr->file))
690: break;
691:
692: if (fr->bsz != fr->cbsz)
693: /* NOTREACHED */
694: err(2, "File read software error 1");
695:
696: if (remsz > (READ_CHUNK >> 1)) {
697: search_start = fr->cbsz - fr->strbeg;
698: fr->cbsz += READ_CHUNK;
699: fr->buffer = sort_realloc(fr->buffer,
700: fr->cbsz);
701: bsz1 = fread(fr->buffer + fr->bsz, 1,
702: READ_CHUNK, fr->file);
703: if (bsz1 == 0) {
704: if (ferror(fr->file))
705: err(2, NULL);
706: break;
707: }
708: fr->bsz += bsz1;
709: remsz += bsz1;
710: } else {
711: if (remsz > 0 && fr->strbeg>0)
712: bcopy(fr->buffer + fr->strbeg,
713: fr->buffer, remsz);
714:
715: fr->strbeg = 0;
716: search_start = remsz;
717: bsz1 = fread(fr->buffer + remsz, 1,
718: fr->cbsz - remsz, fr->file);
719: if (bsz1 == 0) {
720: if (ferror(fr->file))
721: err(2, NULL);
722: break;
723: }
724: fr->bsz = remsz + bsz1;
725: remsz = fr->bsz;
726: }
727: }
728:
729: if (strend == NULL)
730: strend = fr->buffer + fr->bsz;
731:
732: if ((fr->buffer + fr->strbeg <= strend) &&
733: (fr->strbeg < fr->bsz) && (remsz>0))
734: ret = bwscsbdup(fr->buffer + fr->strbeg, strend -
735: fr->buffer - fr->strbeg);
736:
737: fr->strbeg = (strend - fr->buffer) + 1;
738:
739: } else {
740: size_t len = 0;
741:
742: ret = bwsfgetln(fr->file, &len, sort_opts_vals.zflag,
743: &(fr->rb));
744: }
745:
746: return ret;
747: }
748:
749: static void
750: file_reader_clean(struct file_reader *fr)
751: {
752:
753: if (fr) {
754: if (fr->mmapaddr)
755: munmap(fr->mmapaddr, fr->mmapsize);
756:
757: if (fr->fd)
758: close(fr->fd);
759:
760: if (fr->buffer)
761: sort_free(fr->buffer);
762:
763: if (fr->file)
764: if (fr->file != stdin)
765: closefile(fr->file, fr->fname);
766:
767: if (fr->fname)
768: sort_free(fr->fname);
769:
770: memset(fr, 0, sizeof(struct file_reader));
771: }
772: }
773:
774: void
775: file_reader_free(struct file_reader *fr)
776: {
777:
778: if (fr) {
779: file_reader_clean(fr);
780: sort_free(fr);
781: }
782: }
783:
784: int
785: procfile(const char *fsrc, struct sort_list *list, struct file_list *fl)
786: {
787: struct file_reader *fr;
788:
789: fr = file_reader_init(fsrc);
790: if (fr == NULL)
791: err(2, "%s", fsrc);
792:
793: /* file browse cycle */
794: for (;;) {
795: struct bwstring *bws;
796:
797: bws = file_reader_readline(fr);
798:
799: if (bws == NULL)
800: break;
801:
802: sort_list_add(list, bws);
803:
804: if (list->memsize >= available_free_memory) {
805: char *fn;
806:
807: fn = new_tmp_file_name();
808: sort_list_to_file(list, fn);
809: file_list_add(fl, fn, false);
810: sort_list_clean(list);
811: }
812: }
813:
814: file_reader_free(fr);
815:
816: return 0;
817: }
818:
819: /*
820: * Compare file headers. Files with EOF always go to the end of the list.
821: */
822: static int
823: file_header_cmp(struct file_header *f1, struct file_header *f2)
824: {
825:
826: if (f1 == f2)
827: return 0;
828: else {
829: if (f1->fr == NULL) {
830: return (f2->fr == NULL) ? 0 : 1;
831: } else if (f2->fr == NULL)
832: return -1;
833: else {
834: int ret;
835:
836: ret = list_coll(&(f1->si), &(f2->si));
837: if (!ret)
838: return (f1->file_pos < f2->file_pos) ? -1 : 1;
839: return ret;
840: }
841: }
842: }
843:
844: /*
845: * Allocate and init file header structure
846: */
847: static void
848: file_header_init(struct file_header **fh, const char *fn, size_t file_pos)
849: {
850:
851: if (fh && fn) {
852: struct bwstring *line;
853:
854: *fh = sort_malloc(sizeof(struct file_header));
855: (*fh)->file_pos = file_pos;
856: (*fh)->fr = file_reader_init(fn);
857: if ((*fh)->fr == NULL) {
858: err(2, "Cannot open %s for reading",
859: strcmp(fn, "-") == 0 ? "stdin" : fn);
860: }
861: line = file_reader_readline((*fh)->fr);
862: if (line == NULL) {
863: file_reader_free((*fh)->fr);
864: (*fh)->fr = NULL;
865: (*fh)->si = NULL;
866: } else {
867: (*fh)->si = sort_list_item_alloc();
868: sort_list_item_set((*fh)->si, line);
869: }
870: }
871: }
872:
873: /*
874: * Close file
875: */
876: static void
877: file_header_close(struct file_header **fh)
878: {
879:
880: if (fh && *fh) {
881: if ((*fh)->fr) {
882: file_reader_free((*fh)->fr);
883: (*fh)->fr = NULL;
884: }
885: if ((*fh)->si) {
886: sort_list_item_clean((*fh)->si);
887: sort_free((*fh)->si);
888: (*fh)->si = NULL;
889: }
890: sort_free(*fh);
891: *fh = NULL;
892: }
893: }
894:
895: /*
896: * Swap two array elements
897: */
898: static void
899: file_header_swap(struct file_header **fh, size_t i1, size_t i2)
900: {
901: struct file_header *tmp;
902:
903: tmp = fh[i1];
904: fh[i1] = fh[i2];
905: fh[i2] = tmp;
906: }
907:
908: /* heap algorithm ==>> */
909:
910: /*
911: * See heap sort algorithm
912: * "Raises" last element to its right place
913: */
914: static void
915: file_header_heap_swim(struct file_header **fh, size_t indx)
916: {
917:
918: if (indx > 0) {
919: size_t parent_index;
920:
921: parent_index = (indx - 1) >> 1;
922:
923: if (file_header_cmp(fh[indx], fh[parent_index]) < 0) {
924: /* swap child and parent and continue */
925: file_header_swap(fh, indx, parent_index);
926: file_header_heap_swim(fh, parent_index);
927: }
928: }
929: }
930:
931: /*
932: * Sink the top element to its correct position
933: */
934: static void
935: file_header_heap_sink(struct file_header **fh, size_t indx, size_t size)
936: {
937: size_t left_child_index;
938: size_t right_child_index;
939:
940: left_child_index = indx + indx + 1;
941: right_child_index = left_child_index + 1;
942:
943: if (left_child_index < size) {
944: size_t min_child_index;
945:
946: min_child_index = left_child_index;
947:
948: if ((right_child_index < size) &&
949: (file_header_cmp(fh[left_child_index],
950: fh[right_child_index]) > 0))
951: min_child_index = right_child_index;
952: if (file_header_cmp(fh[indx], fh[min_child_index]) > 0) {
953: file_header_swap(fh, indx, min_child_index);
954: file_header_heap_sink(fh, min_child_index, size);
955: }
956: }
957: }
958:
959: /* <<== heap algorithm */
960:
961: /*
962: * Adds element to the "left" end
963: */
964: static void
965: file_header_list_rearrange_from_header(struct file_header **fh, size_t size)
966: {
967:
968: file_header_heap_sink(fh, 0, size);
969: }
970:
971: /*
972: * Adds element to the "right" end
973: */
974: static void
975: file_header_list_push(struct file_header *f, struct file_header **fh, size_t size)
976: {
977:
978: fh[size++] = f;
979: file_header_heap_swim(fh, size - 1);
980: }
981:
982: struct last_printed
983: {
984: struct bwstring *str;
985: };
986:
987: /*
988: * Prints the current line of the file
989: */
990: static void
991: file_header_print(struct file_header *fh, FILE *f_out, struct last_printed *lp)
992: {
993:
994: if (fh && fh->fr && f_out && fh->si && fh->si->str) {
995: if (sort_opts_vals.uflag) {
996: if ((lp->str == NULL) || (str_list_coll(lp->str, &(fh->si)))) {
997: bwsfwrite(fh->si->str, f_out, sort_opts_vals.zflag);
998: if (lp->str)
999: bwsfree(lp->str);
1000: lp->str = bwsdup(fh->si->str);
1001: }
1002: } else
1003: bwsfwrite(fh->si->str, f_out, sort_opts_vals.zflag);
1004: }
1005: }
1006:
1007: /*
1008: * Read next line
1009: */
1010: static void
1011: file_header_read_next(struct file_header *fh)
1012: {
1013:
1014: if (fh && fh->fr) {
1015: struct bwstring *tmp;
1016:
1017: tmp = file_reader_readline(fh->fr);
1018: if (tmp == NULL) {
1019: file_reader_free(fh->fr);
1020: fh->fr = NULL;
1021: if (fh->si) {
1022: sort_list_item_clean(fh->si);
1023: sort_free(fh->si);
1024: fh->si = NULL;
1025: }
1026: } else {
1027: if (fh->si == NULL)
1028: fh->si = sort_list_item_alloc();
1029: sort_list_item_set(fh->si, tmp);
1030: }
1031: }
1032: }
1033:
1034: /*
1035: * Merge array of "files headers"
1036: */
1037: static void
1038: file_headers_merge(size_t fnum, struct file_header **fh, FILE *f_out)
1039: {
1040: struct last_printed lp;
1041: size_t i;
1042:
1043: memset(&lp, 0, sizeof(lp));
1044:
1045: /*
1046: * construct the initial sort structure
1047: */
1048: for (i = 0; i < fnum; i++)
1049: file_header_list_push(fh[i], fh, i);
1050:
1051: while (fh[0]->fr) { /* unfinished files are always in front */
1052: /* output the smallest line: */
1053: file_header_print(fh[0], f_out, &lp);
1054: /* read a new line, if possible: */
1055: file_header_read_next(fh[0]);
1056: /* re-arrange the list: */
1057: file_header_list_rearrange_from_header(fh, fnum);
1058: }
1059:
1060: if (lp.str)
1061: bwsfree(lp.str);
1062: }
1063:
1064: /*
1065: * Merges the given files into the output file, which can be
1066: * stdout.
1067: */
1068: static void
1069: merge_files_array(size_t argc, char **argv, const char *fn_out)
1070: {
1071:
1072: if (argv && fn_out) {
1073: struct file_header **fh;
1074: FILE *f_out;
1075: size_t i;
1076:
1077: f_out = openfile(fn_out, "w");
1078:
1079: if (f_out == NULL)
1080: err(2, "%s", fn_out);
1081:
1082: fh = sort_malloc((argc + 1) * sizeof(struct file_header *));
1083:
1084: for (i = 0; i < argc; i++)
1085: file_header_init(fh + i, argv[i], (size_t) i);
1086:
1087: file_headers_merge(argc, fh, f_out);
1088:
1089: for (i = 0; i < argc; i++)
1090: file_header_close(fh + i);
1091:
1092: sort_free(fh);
1093:
1094: closefile(f_out, fn_out);
1095: }
1096: }
1097:
1098: /*
1099: * Shrinks the file list until its size smaller than max number of opened files
1100: */
1101: static int
1102: shrink_file_list(struct file_list *fl)
1103: {
1104:
1105: if (fl == NULL || (size_t)fl->count < max_open_files)
1106: return 0;
1107: else {
1108: struct file_list new_fl;
1109: size_t indx = 0;
1110:
1111: file_list_init(&new_fl, true);
1112: while (indx < fl->count) {
1113: char *fnew;
1114: size_t num;
1115:
1116: num = fl->count - indx;
1117: fnew = new_tmp_file_name();
1118:
1119: if ((size_t) num >= max_open_files)
1120: num = max_open_files - 1;
1121: merge_files_array(num, fl->fns + indx, fnew);
1122: if (fl->tmp) {
1123: size_t i;
1124:
1125: for (i = 0; i < num; i++)
1126: unlink(fl->fns[indx + i]);
1127: }
1128: file_list_add(&new_fl, fnew, false);
1129: indx += num;
1130: }
1131: fl->tmp = false; /* already taken care of */
1132: file_list_clean(fl);
1133:
1134: fl->count = new_fl.count;
1135: fl->fns = new_fl.fns;
1136: fl->sz = new_fl.sz;
1137: fl->tmp = new_fl.tmp;
1138:
1139: return 1;
1140: }
1141: }
1142:
1143: /*
1144: * Merge list of files
1145: */
1146: void
1147: merge_files(struct file_list *fl, const char *fn_out)
1148: {
1149:
1150: if (fl && fn_out) {
1151: while (shrink_file_list(fl));
1152:
1153: merge_files_array(fl->count, fl->fns, fn_out);
1154: }
1155: }
1156:
1157: static const char *
1158: get_sort_method_name(int sm)
1159: {
1160:
1161: if (sm == SORT_MERGESORT)
1162: return "mergesort";
1163: else if (sort_opts_vals.sort_method == SORT_RADIXSORT)
1164: return "radixsort";
1165: else if (sort_opts_vals.sort_method == SORT_HEAPSORT)
1166: return "heapsort";
1167: else
1168: return "quicksort";
1169: }
1170:
1171: /*
1172: * Sort list of lines and writes it to the file
1173: */
1174: void
1175: sort_list_to_file(struct sort_list *list, const char *outfile)
1176: {
1177: struct sort_mods *sm = &(keys[0].sm);
1178:
1179: if (!sm->Mflag && !sm->Rflag && !sm->Vflag &&
1180: !sm->gflag && !sm->hflag && !sm->nflag) {
1181: if ((sort_opts_vals.sort_method == SORT_DEFAULT) && byte_sort)
1182: sort_opts_vals.sort_method = SORT_RADIXSORT;
1183:
1184: } else if (sort_opts_vals.sort_method == SORT_RADIXSORT)
1185: err(2, "Radix sort cannot be used with these sort options");
1186:
1187: /*
1188: * to handle stable sort and the unique cases in the
1189: * right order, we need stable basic algorithm
1190: */
1191: if (sort_opts_vals.sflag) {
1192: switch (sort_opts_vals.sort_method){
1193: case SORT_MERGESORT:
1194: break;
1195: case SORT_RADIXSORT:
1196: break;
1197: case SORT_DEFAULT:
1198: sort_opts_vals.sort_method = SORT_MERGESORT;
1199: break;
1200: default:
1201: errx(2, "The chosen sort method cannot be used with "
1202: "stable and/or unique sort");
1203: };
1204: }
1205:
1206: if (sort_opts_vals.sort_method == SORT_DEFAULT)
1207: sort_opts_vals.sort_method = DEFAULT_SORT_ALGORITHM;
1208:
1209: if (debug_sort)
1210: printf("sort_method=%s\n",
1211: get_sort_method_name(sort_opts_vals.sort_method));
1212:
1213: switch (sort_opts_vals.sort_method){
1214: case SORT_RADIXSORT:
1215: rxsort(list->list, list->count);
1216: break;
1217: case SORT_MERGESORT:
1218: mergesort(list->list, list->count,
1219: sizeof(struct sort_list_item *), list_coll);
1220: break;
1221: case SORT_HEAPSORT:
1222: heapsort(list->list, list->count,
1223: sizeof(struct sort_list_item *), list_coll);
1224: break;
1225: case SORT_QSORT:
1226: qsort(list->list, list->count,
1227: sizeof(struct sort_list_item *), list_coll);
1228: break;
1229: default:
1230: DEFAULT_SORT_FUNC(list->list, list->count,
1231: sizeof(struct sort_list_item *), list_coll);
1232: break;
1233: }
1234: sort_list_dump(list, outfile);
1235: }