2 * Copyright (C) 1994-1997, Index Data I/S
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.25 1997-09-17 12:19:14 adam
8 * Zebra version corresponds to YAZ version 1.4.
9 * Changed Zebra server so that it doesn't depend on global common_resource.
11 * Revision 1.24 1997/09/09 13:38:07 adam
12 * Partial port to WIN95/NT.
14 * Revision 1.23 1997/09/04 13:57:39 adam
15 * Added O_BINARY for open calls.
17 * Revision 1.22 1997/02/12 20:39:45 adam
18 * Implemented options -f <n> that limits the log to the first <n>
20 * Changed some log messages also.
22 * Revision 1.21 1996/11/08 11:10:23 adam
23 * Buffers used during file match got bigger.
24 * Compressed ISAM support everywhere.
25 * Bug fixes regarding masking characters in queries.
26 * Redesigned Regexp-2 queries.
28 * Revision 1.20 1996/11/01 08:58:41 adam
29 * Interface to isamc system now includes update and delete.
31 * Revision 1.19 1996/10/29 14:09:46 adam
32 * Use of cisam system - enabled if setting isamc is 1.
34 * Revision 1.18 1996/06/04 10:18:59 adam
35 * Minor changes - removed include of ctype.h.
37 * Revision 1.17 1996/05/14 15:47:07 adam
38 * Cleanup of various buffer size entities.
40 * Revision 1.16 1996/04/09 10:05:20 adam
41 * Bug fix: prev_name buffer possibly too small; allocated in key_file_init.
43 * Revision 1.15 1996/03/21 14:50:09 adam
44 * File update uses modify-time instead of change-time.
46 * Revision 1.14 1996/02/07 14:06:37 adam
47 * Better progress report during register merge.
48 * New command: clean - removes temporary shadow files.
50 * Revision 1.13 1996/02/05 12:30:00 adam
51 * Logging reduced a bit.
52 * The remaining running time is estimated during register merge.
54 * Revision 1.12 1995/12/06 17:49:19 adam
55 * Uses dict_delete now.
57 * Revision 1.11 1995/12/06 16:06:43 adam
58 * Better diagnostics. Work on 'real' dictionary deletion.
60 * Revision 1.10 1995/12/06 12:41:22 adam
61 * New command 'stat' for the index program.
62 * Filenames can be read from stdin by specifying '-'.
63 * Bug fix/enhancement of the transformation from terms to regular
64 * expressons in the search engine.
66 * Revision 1.9 1995/10/10 12:24:39 adam
67 * Temporary sort files are compressed.
69 * Revision 1.8 1995/10/04 16:57:19 adam
70 * Key input and merge sort in one pass.
72 * Revision 1.7 1995/10/02 15:18:52 adam
73 * New member in recRetrieveCtrl: diagnostic.
75 * Revision 1.6 1995/09/29 15:51:56 adam
76 * First work on multi-way read.
78 * Revision 1.5 1995/09/29 14:01:43 adam
81 * Revision 1.4 1995/09/28 14:22:57 adam
82 * Sort uses smaller temporary files.
84 * Revision 1.3 1995/09/06 16:11:17 adam
85 * Option: only one word key per file.
87 * Revision 1.2 1995/09/04 12:33:42 adam
88 * Various cleanup. YAZ util used instead.
90 * Revision 1.1 1995/09/04 09:10:37 adam
91 * More work on index add/del/update.
92 * Merge sort implemented.
93 * Initial work on z39 server.
110 #define KEY_SIZE (1+sizeof(struct it_key))
111 #define INP_NAME_MAX 768
112 #define INP_BUF_START 60000
113 #define INP_BUF_ADD 400000
115 static int no_diffs = 0;
116 static int no_updates = 0;
117 static int no_deletions = 0;
118 static int no_insertions = 0;
119 static int no_iterations = 0;
122 int no; /* file no */
123 off_t offset; /* file offset */
124 unsigned char *buf; /* buffer block */
125 size_t buf_size; /* number of read bytes in block */
126 size_t chunk; /* number of bytes allocated */
127 size_t buf_ptr; /* current position in buffer */
128 char *prev_name; /* last word read */
129 int sysno; /* last sysno */
130 int seqno; /* last seqno */
131 off_t length; /* length of file */
132 /* handler invoked in each read */
133 void (*readHandler)(struct key_file *keyp, void *rinfo);
137 void getFnameTmp (char *fname, int no)
141 pre = res_get_def (common_resource, "keyTmpDir", ".");
142 sprintf (fname, "%s/key%d.tmp", pre, no);
145 void key_file_chunk_read (struct key_file *f)
149 getFnameTmp (fname, f->no);
150 fd = open (fname, O_BINARY|O_RDONLY);
153 logf (LOG_FATAL|LOG_ERRNO, "cannot open %s", fname);
158 if ((f->length = lseek (fd, 0L, SEEK_END)) == (off_t) -1)
160 logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
164 if (lseek (fd, f->offset, SEEK_SET) == -1)
166 logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname);
169 while (f->chunk - nr > 0)
171 r = read (fd, f->buf + nr, f->chunk - nr);
178 logf (LOG_FATAL|LOG_ERRNO, "read of %s", fname);
184 (*f->readHandler)(f, f->readInfo);
188 struct key_file *key_file_init (int no, int chunk)
192 f = xmalloc (sizeof(*f));
199 f->readHandler = NULL;
200 f->buf = xmalloc (f->chunk);
201 f->prev_name = xmalloc (INP_NAME_MAX);
202 *f->prev_name = '\0';
203 key_file_chunk_read (f);
207 int key_file_getc (struct key_file *f)
209 if (f->buf_ptr < f->buf_size)
210 return f->buf[(f->buf_ptr)++];
211 if (f->buf_size < f->chunk)
213 f->offset += f->buf_size;
214 key_file_chunk_read (f);
215 if (f->buf_ptr < f->buf_size)
216 return f->buf[(f->buf_ptr)++];
221 int key_file_decode (struct key_file *f)
225 c = key_file_getc (f);
232 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
235 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
236 d = (d << 8) + (key_file_getc (f) & 0xff);
239 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
240 d = (d << 8) + (key_file_getc (f) & 0xff);
241 d = (d << 8) + (key_file_getc (f) & 0xff);
247 int key_file_read (struct key_file *f, char *key)
252 c = key_file_getc (f);
255 strcpy (key, f->prev_name);
264 while ((key[i++] = key_file_getc (f)))
266 strcpy (f->prev_name, key);
269 d = key_file_decode (f);
272 itkey.sysno = d + f->sysno;
275 f->sysno = itkey.sysno;
278 d = key_file_decode (f);
279 itkey.seqno = d + f->seqno;
280 f->seqno = itkey.seqno;
281 memcpy (key + i, &itkey, sizeof(struct it_key));
282 return i + sizeof (struct it_key);
287 struct key_file **file;
292 int (*cmp)(const void *p1, const void *p2);
298 struct heap_info *key_heap_init (int nkeys,
299 int (*cmp)(const void *p1, const void *p2))
301 struct heap_info *hi;
304 hi = xmalloc (sizeof(*hi));
305 hi->info.file = xmalloc (sizeof(*hi->info.file) * (1+nkeys));
306 hi->info.buf = xmalloc (sizeof(*hi->info.buf) * (1+nkeys));
308 hi->ptr = xmalloc (sizeof(*hi->ptr) * (1+nkeys));
310 for (i = 0; i<= nkeys; i++)
313 hi->info.buf[i] = xmalloc (INP_NAME_MAX);
318 static void key_heap_swap (struct heap_info *hi, int i1, int i2)
323 hi->ptr[i1] = hi->ptr[i2];
328 static void key_heap_delete (struct heap_info *hi)
330 int cur = 1, child = 2;
332 assert (hi->heapnum > 0);
334 key_heap_swap (hi, 1, hi->heapnum);
336 while (child <= hi->heapnum) {
337 if (child < hi->heapnum &&
338 (*hi->cmp)(&hi->info.buf[hi->ptr[child]],
339 &hi->info.buf[hi->ptr[child+1]]) > 0)
341 if ((*hi->cmp)(&hi->info.buf[hi->ptr[cur]],
342 &hi->info.buf[hi->ptr[child]]) > 0)
344 key_heap_swap (hi, cur, child);
353 static void key_heap_insert (struct heap_info *hi, const char *buf, int nbytes,
358 cur = ++(hi->heapnum);
359 memcpy (hi->info.buf[hi->ptr[cur]], buf, nbytes);
360 hi->info.file[hi->ptr[cur]] = kf;
363 while (parent && (*hi->cmp)(&hi->info.buf[hi->ptr[parent]],
364 &hi->info.buf[hi->ptr[cur]]) > 0)
366 key_heap_swap (hi, cur, parent);
372 static int heap_read_one (struct heap_info *hi, char *name, char *key)
375 char rbuf[INP_NAME_MAX];
381 strcpy (name, hi->info.buf[n]);
382 kf = hi->info.file[n];
384 memcpy (key, hi->info.buf[n] + r+1, KEY_SIZE);
385 key_heap_delete (hi);
386 if ((r = key_file_read (kf, rbuf)))
387 key_heap_insert (hi, rbuf, r, kf);
392 struct heap_cread_info {
393 char prev_name[INP_NAME_MAX];
394 char cur_name[INP_NAME_MAX];
396 struct heap_info *hi;
401 int heap_cread_item (void *vp, char **dst, int *insertMode)
403 struct heap_cread_info *p = vp;
404 struct heap_info *hi = p->hi;
408 *insertMode = p->key[0];
409 memcpy (*dst, p->key+1, sizeof(struct it_key));
410 (*dst) += sizeof(struct it_key);
414 strcpy (p->prev_name, p->cur_name);
415 if (!(p->more = heap_read_one (hi, p->cur_name, p->key)))
417 if (*p->cur_name && strcmp (p->cur_name, p->prev_name))
422 *insertMode = p->key[0];
423 memcpy (*dst, p->key+1, sizeof(struct it_key));
424 (*dst) += sizeof(struct it_key);
428 int heap_inpc (struct heap_info *hi)
430 struct heap_cread_info hci;
431 ISAMC_I isamc_i = xmalloc (sizeof(*isamc_i));
433 hci.key = xmalloc (KEY_SIZE);
436 hci.more = heap_read_one (hi, hci.cur_name, hci.key);
438 isamc_i->clientData = &hci;
439 isamc_i->read_item = heap_cread_item;
443 char this_name[INP_NAME_MAX];
444 ISAMC_P isamc_p, isamc_p2;
447 strcpy (this_name, hci.cur_name);
448 logf (LOG_DEBUG, "inserting %s", 1+hci.cur_name);
450 if ((dict_info = dict_lookup (hi->dict, hci.cur_name)))
452 memcpy (&isamc_p, dict_info+1, sizeof(ISAMC_P));
453 isamc_p2 = isc_merge (hi->isamc, isamc_p, isamc_i);
457 if (!dict_delete (hi->dict, this_name))
463 if (isamc_p2 != isamc_p)
464 dict_insert (hi->dict, this_name,
465 sizeof(ISAMC_P), &isamc_p2);
470 isamc_p = isc_merge (hi->isamc, 0, isamc_i);
472 dict_insert (hi->dict, this_name, sizeof(ISAMC_P), &isamc_p);
479 int heap_inp (struct heap_info *hi)
482 char next_name[INP_NAME_MAX];
483 char cur_name[INP_NAME_MAX];
484 int key_buf_size = INP_BUF_START;
490 next_key = xmalloc (KEY_SIZE);
491 key_buf = xmalloc (key_buf_size);
492 more = heap_read_one (hi, cur_name, key_buf);
493 while (more) /* EOF ? */
496 key_buf_ptr = KEY_SIZE;
499 if (!(more = heap_read_one (hi, next_name, next_key)))
501 if (*next_name && strcmp (next_name, cur_name))
503 memcpy (key_buf + key_buf_ptr, next_key, KEY_SIZE);
504 key_buf_ptr += KEY_SIZE;
505 if (key_buf_ptr+KEY_SIZE >= key_buf_size)
508 new_key_buf = xmalloc (key_buf_size + INP_BUF_ADD);
509 memcpy (new_key_buf, key_buf, key_buf_size);
510 key_buf_size += INP_BUF_ADD;
512 key_buf = new_key_buf;
516 nmemb = key_buf_ptr / KEY_SIZE;
517 assert (nmemb*KEY_SIZE == key_buf_ptr);
518 if ((info = dict_lookup (hi->dict, cur_name)))
520 ISAM_P isam_p, isam_p2;
521 logf (LOG_DEBUG, "updating %s", 1+cur_name);
522 memcpy (&isam_p, info+1, sizeof(ISAM_P));
523 isam_p2 = is_merge (hi->isam, isam_p, nmemb, key_buf);
527 if (!dict_delete (hi->dict, cur_name))
533 if (isam_p2 != isam_p)
534 dict_insert (hi->dict, cur_name, sizeof(ISAM_P), &isam_p2);
540 logf (LOG_DEBUG, "inserting %s", 1+cur_name);
542 isam_p = is_merge (hi->isam, 0, nmemb, key_buf);
543 dict_insert (hi->dict, cur_name, sizeof(ISAM_P), &isam_p);
545 memcpy (key_buf, next_key, KEY_SIZE);
546 strcpy (cur_name, next_name);
551 struct progressInfo {
558 void progressFunc (struct key_file *keyp, void *info)
560 struct progressInfo *p = info;
561 time_t now, remaining;
563 if (keyp->buf_size <= 0 || p->totalBytes <= 0)
567 if (now >= p->lastTime+10)
570 remaining = (now - p->startTime)*
571 ((double) p->totalBytes/p->totalOffset - 1.0);
572 if (remaining <= 130)
573 logf (LOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining",
574 (100.0*p->totalOffset) / p->totalBytes, (long) remaining);
576 logf (LOG_LOG, "Merge %2.1f%% completed; %ld minutes remaining",
577 (100.0*p->totalOffset) / p->totalBytes, (long) remaining/60);
579 p->totalOffset += keyp->buf_size;
586 void key_input (BFiles bfs, int nkeys, int cache)
592 struct key_file **kf;
595 struct heap_info *hi;
596 struct progressInfo progressInfo;
604 getFnameTmp (fname, nkeys+1);
605 if (access (fname, R_OK) == -1)
612 dict = dict_open (bfs, FNAME_DICT, cache, 1);
615 logf (LOG_FATAL, "dict_open fail");
618 if (res_get_match (common_resource, "isam", "c", NULL))
620 isamc = isc_open (bfs,
621 FNAME_ISAMC, 1, key_isamc_m (common_resource));
624 logf (LOG_FATAL, "isc_open fail");
630 isam = is_open (bfs, FNAME_ISAM, key_compare, 1,
631 sizeof(struct it_key), common_resource);
634 logf (LOG_FATAL, "is_open fail");
638 kf = xmalloc ((1+nkeys) * sizeof(*kf));
639 progressInfo.totalBytes = 0;
640 progressInfo.totalOffset = 0;
641 time (&progressInfo.startTime);
642 time (&progressInfo.lastTime);
643 for (i = 1; i<=nkeys; i++)
645 kf[i] = key_file_init (i, 32768);
646 kf[i]->readHandler = progressFunc;
647 kf[i]->readInfo = &progressInfo;
648 progressInfo.totalBytes += kf[i]->length;
649 progressInfo.totalOffset += kf[i]->buf_size;
651 hi = key_heap_init (nkeys, key_qsort_compare);
656 for (i = 1; i<=nkeys; i++)
657 if ((r = key_file_read (kf[i], rbuf)))
658 key_heap_insert (hi, rbuf, r, kf[i]);
669 for (i = 1; i<=nkeys; i++)
671 getFnameTmp (rbuf, i);
674 logf (LOG_LOG, "Iterations . . .%7d", no_iterations);
675 logf (LOG_LOG, "Distinct words .%7d", no_diffs);
676 logf (LOG_LOG, "Updates. . . . .%7d", no_updates);
677 logf (LOG_LOG, "Deletions. . . .%7d", no_deletions);
678 logf (LOG_LOG, "Insertions . . .%7d", no_insertions);