2 * Copyright (C) 1994-1998, Index Data I/S
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.27 1998-02-17 10:32:52 adam
8 * Fixed bug: binary files weren't opened with flag b on NT.
10 * Revision 1.26 1998/01/29 13:39:13 adam
11 * Compress ISAM is default.
13 * Revision 1.25 1997/09/17 12:19:14 adam
14 * Zebra version corresponds to YAZ version 1.4.
15 * Changed Zebra server so that it doesn't depend on global common_resource.
17 * Revision 1.24 1997/09/09 13:38:07 adam
18 * Partial port to WIN95/NT.
20 * Revision 1.23 1997/09/04 13:57:39 adam
21 * Added O_BINARY for open calls.
23 * Revision 1.22 1997/02/12 20:39:45 adam
24 * Implemented options -f <n> that limits the log to the first <n>
26 * Changed some log messages also.
28 * Revision 1.21 1996/11/08 11:10:23 adam
29 * Buffers used during file match got bigger.
30 * Compressed ISAM support everywhere.
31 * Bug fixes regarding masking characters in queries.
32 * Redesigned Regexp-2 queries.
34 * Revision 1.20 1996/11/01 08:58:41 adam
35 * Interface to isamc system now includes update and delete.
37 * Revision 1.19 1996/10/29 14:09:46 adam
38 * Use of cisam system - enabled if setting isamc is 1.
40 * Revision 1.18 1996/06/04 10:18:59 adam
41 * Minor changes - removed include of ctype.h.
43 * Revision 1.17 1996/05/14 15:47:07 adam
44 * Cleanup of various buffer size entities.
46 * Revision 1.16 1996/04/09 10:05:20 adam
47 * Bug fix: prev_name buffer possibly too small; allocated in key_file_init.
49 * Revision 1.15 1996/03/21 14:50:09 adam
50 * File update uses modify-time instead of change-time.
52 * Revision 1.14 1996/02/07 14:06:37 adam
53 * Better progress report during register merge.
54 * New command: clean - removes temporary shadow files.
56 * Revision 1.13 1996/02/05 12:30:00 adam
57 * Logging reduced a bit.
58 * The remaining running time is estimated during register merge.
60 * Revision 1.12 1995/12/06 17:49:19 adam
61 * Uses dict_delete now.
63 * Revision 1.11 1995/12/06 16:06:43 adam
64 * Better diagnostics. Work on 'real' dictionary deletion.
66 * Revision 1.10 1995/12/06 12:41:22 adam
67 * New command 'stat' for the index program.
68 * Filenames can be read from stdin by specifying '-'.
69 * Bug fix/enhancement of the transformation from terms to regular
70 * expressons in the search engine.
72 * Revision 1.9 1995/10/10 12:24:39 adam
73 * Temporary sort files are compressed.
75 * Revision 1.8 1995/10/04 16:57:19 adam
76 * Key input and merge sort in one pass.
78 * Revision 1.7 1995/10/02 15:18:52 adam
79 * New member in recRetrieveCtrl: diagnostic.
81 * Revision 1.6 1995/09/29 15:51:56 adam
82 * First work on multi-way read.
84 * Revision 1.5 1995/09/29 14:01:43 adam
87 * Revision 1.4 1995/09/28 14:22:57 adam
88 * Sort uses smaller temporary files.
90 * Revision 1.3 1995/09/06 16:11:17 adam
91 * Option: only one word key per file.
93 * Revision 1.2 1995/09/04 12:33:42 adam
94 * Various cleanup. YAZ util used instead.
96 * Revision 1.1 1995/09/04 09:10:37 adam
97 * More work on index add/del/update.
98 * Merge sort implemented.
99 * Initial work on z39 server.
116 #define KEY_SIZE (1+sizeof(struct it_key))
117 #define INP_NAME_MAX 768
118 #define INP_BUF_START 60000
119 #define INP_BUF_ADD 400000
121 static int no_diffs = 0;
122 static int no_updates = 0;
123 static int no_deletions = 0;
124 static int no_insertions = 0;
125 static int no_iterations = 0;
128 int no; /* file no */
129 off_t offset; /* file offset */
130 unsigned char *buf; /* buffer block */
131 size_t buf_size; /* number of read bytes in block */
132 size_t chunk; /* number of bytes allocated */
133 size_t buf_ptr; /* current position in buffer */
134 char *prev_name; /* last word read */
135 int sysno; /* last sysno */
136 int seqno; /* last seqno */
137 off_t length; /* length of file */
138 /* handler invoked in each read */
139 void (*readHandler)(struct key_file *keyp, void *rinfo);
143 void getFnameTmp (char *fname, int no)
147 pre = res_get_def (common_resource, "keyTmpDir", ".");
148 sprintf (fname, "%s/key%d.tmp", pre, no);
151 void key_file_chunk_read (struct key_file *f)
155 getFnameTmp (fname, f->no);
156 fd = open (fname, O_BINARY|O_RDONLY);
159 logf (LOG_FATAL|LOG_ERRNO, "cannot open %s", fname);
164 if ((f->length = lseek (fd, 0L, SEEK_END)) == (off_t) -1)
166 logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
170 if (lseek (fd, f->offset, SEEK_SET) == -1)
172 logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
175 while (f->chunk - nr > 0)
177 r = read (fd, f->buf + nr, f->chunk - nr);
184 logf (LOG_FATAL|LOG_ERRNO, "read of %s", fname);
190 (*f->readHandler)(f, f->readInfo);
194 struct key_file *key_file_init (int no, int chunk)
198 f = xmalloc (sizeof(*f));
205 f->readHandler = NULL;
206 f->buf = xmalloc (f->chunk);
207 f->prev_name = xmalloc (INP_NAME_MAX);
208 *f->prev_name = '\0';
209 key_file_chunk_read (f);
213 int key_file_getc (struct key_file *f)
215 if (f->buf_ptr < f->buf_size)
216 return f->buf[(f->buf_ptr)++];
217 if (f->buf_size < f->chunk)
219 f->offset += f->buf_size;
220 key_file_chunk_read (f);
221 if (f->buf_ptr < f->buf_size)
222 return f->buf[(f->buf_ptr)++];
227 int key_file_decode (struct key_file *f)
231 c = key_file_getc (f);
238 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
241 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
242 d = (d << 8) + (key_file_getc (f) & 0xff);
245 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
246 d = (d << 8) + (key_file_getc (f) & 0xff);
247 d = (d << 8) + (key_file_getc (f) & 0xff);
253 int key_file_read (struct key_file *f, char *key)
258 c = key_file_getc (f);
261 strcpy (key, f->prev_name);
270 while ((key[i++] = key_file_getc (f)))
272 strcpy (f->prev_name, key);
275 d = key_file_decode (f);
278 itkey.sysno = d + f->sysno;
281 f->sysno = itkey.sysno;
284 d = key_file_decode (f);
285 itkey.seqno = d + f->seqno;
286 f->seqno = itkey.seqno;
287 memcpy (key + i, &itkey, sizeof(struct it_key));
288 return i + sizeof (struct it_key);
293 struct key_file **file;
298 int (*cmp)(const void *p1, const void *p2);
304 struct heap_info *key_heap_init (int nkeys,
305 int (*cmp)(const void *p1, const void *p2))
307 struct heap_info *hi;
310 hi = xmalloc (sizeof(*hi));
311 hi->info.file = xmalloc (sizeof(*hi->info.file) * (1+nkeys));
312 hi->info.buf = xmalloc (sizeof(*hi->info.buf) * (1+nkeys));
314 hi->ptr = xmalloc (sizeof(*hi->ptr) * (1+nkeys));
316 for (i = 0; i<= nkeys; i++)
319 hi->info.buf[i] = xmalloc (INP_NAME_MAX);
324 static void key_heap_swap (struct heap_info *hi, int i1, int i2)
329 hi->ptr[i1] = hi->ptr[i2];
334 static void key_heap_delete (struct heap_info *hi)
336 int cur = 1, child = 2;
338 assert (hi->heapnum > 0);
340 key_heap_swap (hi, 1, hi->heapnum);
342 while (child <= hi->heapnum) {
343 if (child < hi->heapnum &&
344 (*hi->cmp)(&hi->info.buf[hi->ptr[child]],
345 &hi->info.buf[hi->ptr[child+1]]) > 0)
347 if ((*hi->cmp)(&hi->info.buf[hi->ptr[cur]],
348 &hi->info.buf[hi->ptr[child]]) > 0)
350 key_heap_swap (hi, cur, child);
359 static void key_heap_insert (struct heap_info *hi, const char *buf, int nbytes,
364 cur = ++(hi->heapnum);
365 memcpy (hi->info.buf[hi->ptr[cur]], buf, nbytes);
366 hi->info.file[hi->ptr[cur]] = kf;
369 while (parent && (*hi->cmp)(&hi->info.buf[hi->ptr[parent]],
370 &hi->info.buf[hi->ptr[cur]]) > 0)
372 key_heap_swap (hi, cur, parent);
378 static int heap_read_one (struct heap_info *hi, char *name, char *key)
381 char rbuf[INP_NAME_MAX];
387 strcpy (name, hi->info.buf[n]);
388 kf = hi->info.file[n];
390 memcpy (key, hi->info.buf[n] + r+1, KEY_SIZE);
391 key_heap_delete (hi);
392 if ((r = key_file_read (kf, rbuf)))
393 key_heap_insert (hi, rbuf, r, kf);
398 struct heap_cread_info {
399 char prev_name[INP_NAME_MAX];
400 char cur_name[INP_NAME_MAX];
402 struct heap_info *hi;
407 int heap_cread_item (void *vp, char **dst, int *insertMode)
409 struct heap_cread_info *p = vp;
410 struct heap_info *hi = p->hi;
414 *insertMode = p->key[0];
415 memcpy (*dst, p->key+1, sizeof(struct it_key));
416 (*dst) += sizeof(struct it_key);
420 strcpy (p->prev_name, p->cur_name);
421 if (!(p->more = heap_read_one (hi, p->cur_name, p->key)))
423 if (*p->cur_name && strcmp (p->cur_name, p->prev_name))
428 *insertMode = p->key[0];
429 memcpy (*dst, p->key+1, sizeof(struct it_key));
430 (*dst) += sizeof(struct it_key);
434 int heap_inpc (struct heap_info *hi)
436 struct heap_cread_info hci;
437 ISAMC_I isamc_i = xmalloc (sizeof(*isamc_i));
439 hci.key = xmalloc (KEY_SIZE);
442 hci.more = heap_read_one (hi, hci.cur_name, hci.key);
444 isamc_i->clientData = &hci;
445 isamc_i->read_item = heap_cread_item;
449 char this_name[INP_NAME_MAX];
450 ISAMC_P isamc_p, isamc_p2;
453 strcpy (this_name, hci.cur_name);
454 logf (LOG_DEBUG, "inserting %s", 1+hci.cur_name);
455 assert (hci.cur_name[1]);
457 if ((dict_info = dict_lookup (hi->dict, hci.cur_name)))
459 memcpy (&isamc_p, dict_info+1, sizeof(ISAMC_P));
460 isamc_p2 = isc_merge (hi->isamc, isamc_p, isamc_i);
464 if (!dict_delete (hi->dict, this_name))
470 if (isamc_p2 != isamc_p)
471 dict_insert (hi->dict, this_name,
472 sizeof(ISAMC_P), &isamc_p2);
477 isamc_p = isc_merge (hi->isamc, 0, isamc_i);
479 dict_insert (hi->dict, this_name, sizeof(ISAMC_P), &isamc_p);
486 int heap_inp (struct heap_info *hi)
489 char next_name[INP_NAME_MAX];
490 char cur_name[INP_NAME_MAX];
491 int key_buf_size = INP_BUF_START;
497 next_key = xmalloc (KEY_SIZE);
498 key_buf = xmalloc (key_buf_size);
499 more = heap_read_one (hi, cur_name, key_buf);
500 while (more) /* EOF ? */
503 key_buf_ptr = KEY_SIZE;
506 if (!(more = heap_read_one (hi, next_name, next_key)))
508 if (*next_name && strcmp (next_name, cur_name))
510 memcpy (key_buf + key_buf_ptr, next_key, KEY_SIZE);
511 key_buf_ptr += KEY_SIZE;
512 if (key_buf_ptr+KEY_SIZE >= key_buf_size)
515 new_key_buf = xmalloc (key_buf_size + INP_BUF_ADD);
516 memcpy (new_key_buf, key_buf, key_buf_size);
517 key_buf_size += INP_BUF_ADD;
519 key_buf = new_key_buf;
523 nmemb = key_buf_ptr / KEY_SIZE;
524 assert (nmemb*KEY_SIZE == key_buf_ptr);
525 if ((info = dict_lookup (hi->dict, cur_name)))
527 ISAM_P isam_p, isam_p2;
528 logf (LOG_DEBUG, "updating %s", 1+cur_name);
529 memcpy (&isam_p, info+1, sizeof(ISAM_P));
530 isam_p2 = is_merge (hi->isam, isam_p, nmemb, key_buf);
534 if (!dict_delete (hi->dict, cur_name))
540 if (isam_p2 != isam_p)
541 dict_insert (hi->dict, cur_name, sizeof(ISAM_P), &isam_p2);
547 logf (LOG_DEBUG, "inserting %s", 1+cur_name);
549 isam_p = is_merge (hi->isam, 0, nmemb, key_buf);
550 dict_insert (hi->dict, cur_name, sizeof(ISAM_P), &isam_p);
552 memcpy (key_buf, next_key, KEY_SIZE);
553 strcpy (cur_name, next_name);
558 struct progressInfo {
565 void progressFunc (struct key_file *keyp, void *info)
567 struct progressInfo *p = info;
568 time_t now, remaining;
570 if (keyp->buf_size <= 0 || p->totalBytes <= 0)
574 if (now >= p->lastTime+10)
577 remaining = (now - p->startTime)*
578 ((double) p->totalBytes/p->totalOffset - 1.0);
579 if (remaining <= 130)
580 logf (LOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining",
581 (100.0*p->totalOffset) / p->totalBytes, (long) remaining);
583 logf (LOG_LOG, "Merge %2.1f%% completed; %ld minutes remaining",
584 (100.0*p->totalOffset) / p->totalBytes, (long) remaining/60);
586 p->totalOffset += keyp->buf_size;
593 void key_input (BFiles bfs, int nkeys, int cache)
599 struct key_file **kf;
602 struct heap_info *hi;
603 struct progressInfo progressInfo;
611 getFnameTmp (fname, nkeys+1);
612 if (access (fname, R_OK) == -1)
619 dict = dict_open (bfs, FNAME_DICT, cache, 1);
622 logf (LOG_FATAL, "dict_open fail");
625 if (!res_get_match (common_resource, "isam", "i", NULL))
627 isamc = isc_open (bfs,
628 FNAME_ISAMC, 1, key_isamc_m (common_resource));
631 logf (LOG_FATAL, "isc_open fail");
637 isam = is_open (bfs, FNAME_ISAM, key_compare, 1,
638 sizeof(struct it_key), common_resource);
641 logf (LOG_FATAL, "is_open fail");
645 kf = xmalloc ((1+nkeys) * sizeof(*kf));
646 progressInfo.totalBytes = 0;
647 progressInfo.totalOffset = 0;
648 time (&progressInfo.startTime);
649 time (&progressInfo.lastTime);
650 for (i = 1; i<=nkeys; i++)
652 kf[i] = key_file_init (i, 32768);
653 kf[i]->readHandler = progressFunc;
654 kf[i]->readInfo = &progressInfo;
655 progressInfo.totalBytes += kf[i]->length;
656 progressInfo.totalOffset += kf[i]->buf_size;
658 hi = key_heap_init (nkeys, key_qsort_compare);
663 for (i = 1; i<=nkeys; i++)
664 if ((r = key_file_read (kf[i], rbuf)))
665 key_heap_insert (hi, rbuf, r, kf[i]);
676 for (i = 1; i<=nkeys; i++)
678 getFnameTmp (rbuf, i);
681 logf (LOG_LOG, "Iterations . . .%7d", no_iterations);
682 logf (LOG_LOG, "Distinct words .%7d", no_diffs);
683 logf (LOG_LOG, "Updates. . . . .%7d", no_updates);
684 logf (LOG_LOG, "Deletions. . . .%7d", no_deletions);
685 logf (LOG_LOG, "Insertions . . .%7d", no_insertions);