1 /* $Id: kinput.c,v 1.54 2003-03-05 16:44:02 adam Exp $
2 Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002
5 This file is part of the Zebra server.
7 Zebra is free software; you can redistribute it and/or modify it under
8 the terms of the GNU General Public License as published by the Free
9 Software Foundation; either version 2, or (at your option) any later
12 Zebra is distributed in the hope that it will be useful, but WITHOUT ANY
13 WARRANTY; without even the implied warranty of MERCHANTABILITY or
14 FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
17 You should have received a copy of the GNU General Public License
18 along with Zebra; see the file LICENSE.zebra. If not, write to the
19 Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
38 #define KEY_SIZE (1+sizeof(struct it_key))
39 #define INP_NAME_MAX 768
40 #define INP_BUF_START 60000
41 #define INP_BUF_ADD 400000
46 off_t offset; /* file offset */
47 unsigned char *buf; /* buffer block */
48 size_t buf_size; /* number of read bytes in block */
49 size_t chunk; /* number of bytes allocated */
50 size_t buf_ptr; /* current position in buffer */
51 char *prev_name; /* last word read */
52 int sysno; /* last sysno */
53 int seqno; /* last seqno */
54 off_t length; /* length of file */
55 /* handler invoked in each read */
56 void (*readHandler)(struct key_file *keyp, void *rinfo);
61 void getFnameTmp (Res res, char *fname, int no)
65 pre = res_get_def (res, "keyTmpDir", ".");
66 sprintf (fname, "%s/key%d.tmp", pre, no);
69 void extract_get_fname_tmp (ZebraHandle zh, char *fname, int no)
73 pre = res_get_def (zh->res, "keyTmpDir", ".");
74 sprintf (fname, "%s/key%d.tmp", pre, no);
77 void key_file_chunk_read (struct key_file *f)
79 int nr = 0, r = 0, fd;
81 getFnameTmp (f->res, fname, f->no);
82 fd = open (fname, O_BINARY|O_RDONLY);
88 logf (LOG_WARN|LOG_ERRNO, "cannot open %s", fname);
93 if ((f->length = lseek (fd, 0L, SEEK_END)) == (off_t) -1)
95 logf (LOG_WARN|LOG_ERRNO, "cannot seek %s", fname);
100 if (lseek (fd, f->offset, SEEK_SET) == -1)
102 logf (LOG_WARN|LOG_ERRNO, "cannot seek %s", fname);
106 while (f->chunk - nr > 0)
108 r = read (fd, f->buf + nr, f->chunk - nr);
115 logf (LOG_WARN|LOG_ERRNO, "read of %s", fname);
121 (*f->readHandler)(f, f->readInfo);
125 void key_file_destroy (struct key_file *f)
128 xfree (f->prev_name);
132 struct key_file *key_file_init (int no, int chunk, Res res)
136 f = (struct key_file *) xmalloc (sizeof(*f));
144 f->readHandler = NULL;
145 f->buf = (unsigned char *) xmalloc (f->chunk);
146 f->prev_name = (char *) xmalloc (INP_NAME_MAX);
147 *f->prev_name = '\0';
148 key_file_chunk_read (f);
152 int key_file_getc (struct key_file *f)
154 if (f->buf_ptr < f->buf_size)
155 return f->buf[(f->buf_ptr)++];
156 if (f->buf_size < f->chunk)
158 f->offset += f->buf_size;
159 key_file_chunk_read (f);
160 if (f->buf_ptr < f->buf_size)
161 return f->buf[(f->buf_ptr)++];
166 int key_file_decode (struct key_file *f)
170 c = key_file_getc (f);
177 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
180 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
181 d = (d << 8) + (key_file_getc (f) & 0xff);
184 d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
185 d = (d << 8) + (key_file_getc (f) & 0xff);
186 d = (d << 8) + (key_file_getc (f) & 0xff);
192 int key_file_read (struct key_file *f, char *key)
197 c = key_file_getc (f);
200 strcpy (key, f->prev_name);
209 while ((key[i++] = key_file_getc (f)))
211 strcpy (f->prev_name, key);
214 d = key_file_decode (f);
217 itkey.sysno = d + f->sysno;
220 f->sysno = itkey.sysno;
223 d = key_file_decode (f);
224 itkey.seqno = d + f->seqno;
225 f->seqno = itkey.seqno;
226 memcpy (key + i, &itkey, sizeof(struct it_key));
227 return i + sizeof (struct it_key);
232 struct key_file **file;
237 int (*cmp)(const void *p1, const void *p2);
238 struct zebra_register *reg;
247 struct heap_info *key_heap_init (int nkeys,
248 int (*cmp)(const void *p1, const void *p2))
250 struct heap_info *hi;
253 hi = (struct heap_info *) xmalloc (sizeof(*hi));
254 hi->info.file = (struct key_file **)
255 xmalloc (sizeof(*hi->info.file) * (1+nkeys));
256 hi->info.buf = (char **) xmalloc (sizeof(*hi->info.buf) * (1+nkeys));
258 hi->ptr = (int *) xmalloc (sizeof(*hi->ptr) * (1+nkeys));
260 for (i = 0; i<= nkeys; i++)
263 hi->info.buf[i] = (char *) xmalloc (INP_NAME_MAX);
268 hi->no_deletions = 0;
269 hi->no_insertions = 0;
270 hi->no_iterations = 0;
274 void key_heap_destroy (struct heap_info *hi, int nkeys)
277 yaz_log (LOG_DEBUG, "key_heap_destroy");
278 for (i = 0; i<=nkeys; i++)
279 xfree (hi->info.buf[i]);
281 xfree (hi->info.buf);
283 xfree (hi->info.file);
287 static void key_heap_swap (struct heap_info *hi, int i1, int i2)
292 hi->ptr[i1] = hi->ptr[i2];
297 static void key_heap_delete (struct heap_info *hi)
299 int cur = 1, child = 2;
301 assert (hi->heapnum > 0);
303 key_heap_swap (hi, 1, hi->heapnum);
305 while (child <= hi->heapnum) {
306 if (child < hi->heapnum &&
307 (*hi->cmp)(&hi->info.buf[hi->ptr[child]],
308 &hi->info.buf[hi->ptr[child+1]]) > 0)
310 if ((*hi->cmp)(&hi->info.buf[hi->ptr[cur]],
311 &hi->info.buf[hi->ptr[child]]) > 0)
313 key_heap_swap (hi, cur, child);
322 static void key_heap_insert (struct heap_info *hi, const char *buf, int nbytes,
327 cur = ++(hi->heapnum);
328 memcpy (hi->info.buf[hi->ptr[cur]], buf, nbytes);
329 hi->info.file[hi->ptr[cur]] = kf;
332 while (parent && (*hi->cmp)(&hi->info.buf[hi->ptr[parent]],
333 &hi->info.buf[hi->ptr[cur]]) > 0)
335 key_heap_swap (hi, cur, parent);
341 static int heap_read_one (struct heap_info *hi, char *name, char *key)
344 char rbuf[INP_NAME_MAX];
350 strcpy (name, hi->info.buf[n]);
351 kf = hi->info.file[n];
353 memcpy (key, hi->info.buf[n] + r+1, KEY_SIZE);
354 key_heap_delete (hi);
355 if ((r = key_file_read (kf, rbuf)))
356 key_heap_insert (hi, rbuf, r, kf);
361 struct heap_cread_info {
362 char prev_name[INP_NAME_MAX];
363 char cur_name[INP_NAME_MAX];
365 struct heap_info *hi;
370 int heap_cread_item (void *vp, char **dst, int *insertMode)
372 struct heap_cread_info *p = (struct heap_cread_info *) vp;
373 struct heap_info *hi = p->hi;
377 *insertMode = p->key[0];
378 memcpy (*dst, p->key+1, sizeof(struct it_key));
379 (*dst) += sizeof(struct it_key);
383 strcpy (p->prev_name, p->cur_name);
384 if (!(p->more = heap_read_one (hi, p->cur_name, p->key)))
386 if (*p->cur_name && strcmp (p->cur_name, p->prev_name))
391 *insertMode = p->key[0];
392 memcpy (*dst, p->key+1, sizeof(struct it_key));
393 (*dst) += sizeof(struct it_key);
397 int heap_inpc (struct heap_info *hi)
399 struct heap_cread_info hci;
400 ISAMC_I isamc_i = (ISAMC_I) xmalloc (sizeof(*isamc_i));
402 hci.key = (char *) xmalloc (KEY_SIZE);
405 hci.more = heap_read_one (hi, hci.cur_name, hci.key);
407 isamc_i->clientData = &hci;
408 isamc_i->read_item = heap_cread_item;
412 char this_name[INP_NAME_MAX];
413 ISAMC_P isamc_p, isamc_p2;
416 strcpy (this_name, hci.cur_name);
417 assert (hci.cur_name[1]);
419 if ((dict_info = dict_lookup (hi->reg->dict, hci.cur_name)))
421 memcpy (&isamc_p, dict_info+1, sizeof(ISAMC_P));
422 isamc_p2 = isc_merge (hi->reg->isamc, isamc_p, isamc_i);
426 if (!dict_delete (hi->reg->dict, this_name))
432 if (isamc_p2 != isamc_p)
433 dict_insert (hi->reg->dict, this_name,
434 sizeof(ISAMC_P), &isamc_p2);
439 isamc_p = isc_merge (hi->reg->isamc, 0, isamc_i);
441 dict_insert (hi->reg->dict, this_name, sizeof(ISAMC_P), &isamc_p);
450 /* for debugging only */
451 static void print_dict_item (ZebraMaps zm, const char *s)
454 char keybuf[IT_MAX_WORD+1];
456 const char *from = s + 2;
460 const char *res = zebra_maps_output (zm, reg_type, &from);
468 yaz_log (LOG_LOG, "%s", keybuf);
472 int heap_inpb (struct heap_info *hi)
474 struct heap_cread_info hci;
475 ISAMC_I isamc_i = (ISAMC_I) xmalloc (sizeof(*isamc_i));
477 hci.key = (char *) xmalloc (KEY_SIZE);
480 hci.more = heap_read_one (hi, hci.cur_name, hci.key);
482 isamc_i->clientData = &hci;
483 isamc_i->read_item = heap_cread_item;
487 char this_name[INP_NAME_MAX];
488 ISAMC_P isamc_p, isamc_p2;
491 strcpy (this_name, hci.cur_name);
492 assert (hci.cur_name[1]);
496 print_dict_item (hi->reg->zebra_maps, hci.cur_name);
498 if ((dict_info = dict_lookup (hi->reg->dict, hci.cur_name)))
500 memcpy (&isamc_p, dict_info+1, sizeof(ISAMC_P));
501 isamc_p2 = isamb_merge (hi->reg->isamb, isamc_p, isamc_i);
505 if (!dict_delete (hi->reg->dict, this_name))
511 if (isamc_p2 != isamc_p)
512 dict_insert (hi->reg->dict, this_name,
513 sizeof(ISAMC_P), &isamc_p2);
518 isamc_p = isamb_merge (hi->reg->isamb, 0, isamc_i);
520 dict_insert (hi->reg->dict, this_name, sizeof(ISAMC_P), &isamc_p);
528 int heap_inpd (struct heap_info *hi)
530 struct heap_cread_info hci;
531 ISAMD_I isamd_i = (ISAMD_I) xmalloc (sizeof(*isamd_i));
533 hci.key = (char *) xmalloc (KEY_SIZE);
536 hci.more = heap_read_one (hi, hci.cur_name, hci.key);
538 isamd_i->clientData = &hci;
539 isamd_i->read_item = heap_cread_item;
543 char this_name[INP_NAME_MAX];
545 char dictentry[ISAMD_MAX_DICT_LEN+1];
548 strcpy (this_name, hci.cur_name);
550 /* print_dict_item (hi->reg->zebra_maps, hci.cur_name); */
551 /*!*/ /* FIXME: depend on isamd-debug */
553 assert (hci.cur_name[1]);
555 if ((dict_info = dict_lookup (hi->reg->dict, hci.cur_name)))
557 dictlen=dict_info[0];
558 memcpy (dictentry, dict_info+1, dictlen );
560 logf(LOG_LOG,"dictentry before. len=%d: %d %d %d %d %d %d %d %d %d",
561 dictlen,dictentry[0], dictentry[1], dictentry[2],
562 dictentry[3], dictentry[4], dictentry[5],
563 dictentry[6], dictentry[7], dictentry[8]); /*!*/
565 dictlen= isamd_append(hi->reg->isamd, dictentry, dictlen, isamd_i);
566 /* logf dictentry after */
570 if ( (dictlen!=dict_info[0]) ||
571 (0!=memcmp(dictentry, dict_info+1, dictlen)) )
573 dict_insert(hi->reg->dict, this_name,
580 if (!dict_delete (hi->reg->dict, this_name))
582 logf (LOG_FATAL, "dict_delete failed");
590 memset (dictentry, '\0', ISAMD_MAX_DICT_LEN);
591 dictlen= isamd_append(hi->reg->isamd, dictentry, dictlen, isamd_i);
592 /* logf dictentry first */
595 dict_insert(hi->reg->dict, this_name,
603 int heap_inp (struct heap_info *hi)
606 char next_name[INP_NAME_MAX];
607 char cur_name[INP_NAME_MAX];
608 int key_buf_size = INP_BUF_START;
614 next_key = (char *) xmalloc (KEY_SIZE);
615 key_buf = (char *) xmalloc (key_buf_size);
616 more = heap_read_one (hi, cur_name, key_buf);
617 while (more) /* EOF ? */
620 key_buf_ptr = KEY_SIZE;
623 if (!(more = heap_read_one (hi, next_name, next_key)))
625 if (*next_name && strcmp (next_name, cur_name))
627 memcpy (key_buf + key_buf_ptr, next_key, KEY_SIZE);
628 key_buf_ptr += KEY_SIZE;
629 if (key_buf_ptr+(int) KEY_SIZE >= key_buf_size)
632 new_key_buf = (char *) xmalloc (key_buf_size + INP_BUF_ADD);
633 memcpy (new_key_buf, key_buf, key_buf_size);
634 key_buf_size += INP_BUF_ADD;
636 key_buf = new_key_buf;
640 nmemb = key_buf_ptr / KEY_SIZE;
641 assert (nmemb * (int) KEY_SIZE == key_buf_ptr);
642 if ((info = dict_lookup (hi->reg->dict, cur_name)))
644 ISAM_P isam_p, isam_p2;
645 memcpy (&isam_p, info+1, sizeof(ISAM_P));
646 isam_p2 = is_merge (hi->reg->isam, isam_p, nmemb, key_buf);
650 if (!dict_delete (hi->reg->dict, cur_name))
656 if (isam_p2 != isam_p)
657 dict_insert (hi->reg->dict, cur_name,
658 sizeof(ISAM_P), &isam_p2);
665 isam_p = is_merge (hi->reg->isam, 0, nmemb, key_buf);
666 dict_insert (hi->reg->dict, cur_name, sizeof(ISAM_P), &isam_p);
668 memcpy (key_buf, next_key, KEY_SIZE);
669 strcpy (cur_name, next_name);
674 int heap_inps (struct heap_info *hi)
676 struct heap_cread_info hci;
677 ISAMS_I isams_i = (ISAMS_I) xmalloc (sizeof(*isams_i));
679 hci.key = (char *) xmalloc (KEY_SIZE);
682 hci.more = heap_read_one (hi, hci.cur_name, hci.key);
684 isams_i->clientData = &hci;
685 isams_i->read_item = heap_cread_item;
689 char this_name[INP_NAME_MAX];
693 strcpy (this_name, hci.cur_name);
694 assert (hci.cur_name[1]);
696 if (!(dict_info = dict_lookup (hi->reg->dict, hci.cur_name)))
698 isams_p = isams_merge (hi->reg->isams, isams_i);
700 dict_insert (hi->reg->dict, this_name, sizeof(ISAMS_P), &isams_p);
704 logf (LOG_FATAL, "isams doesn't support this kind of update");
712 struct progressInfo {
719 void progressFunc (struct key_file *keyp, void *info)
721 struct progressInfo *p = (struct progressInfo *) info;
722 time_t now, remaining;
724 if (keyp->buf_size <= 0 || p->totalBytes <= 0)
728 if (now >= p->lastTime+10)
731 remaining = (time_t) ((now - p->startTime)*
732 ((double) p->totalBytes/p->totalOffset - 1.0));
733 if (remaining <= 130)
734 logf (LOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining",
735 (100.0*p->totalOffset) / p->totalBytes, (long) remaining);
737 logf (LOG_LOG, "Merge %2.1f%% completed; %ld minutes remaining",
738 (100.0*p->totalOffset) / p->totalBytes, (long) remaining/60);
740 p->totalOffset += keyp->buf_size;
747 void zebra_index_merge (ZebraHandle zh)
749 struct key_file **kf;
752 struct heap_info *hi;
753 struct progressInfo progressInfo;
754 int nkeys = zh->reg->key_file_no;
762 extract_get_fname_tmp (zh, fname, nkeys+1);
763 if (access (fname, R_OK) == -1)
770 kf = (struct key_file **) xmalloc ((1+nkeys) * sizeof(*kf));
771 progressInfo.totalBytes = 0;
772 progressInfo.totalOffset = 0;
773 time (&progressInfo.startTime);
774 time (&progressInfo.lastTime);
775 for (i = 1; i<=nkeys; i++)
777 kf[i] = key_file_init (i, 8192, zh->res);
778 kf[i]->readHandler = progressFunc;
779 kf[i]->readInfo = &progressInfo;
780 progressInfo.totalBytes += kf[i]->length;
781 progressInfo.totalOffset += kf[i]->buf_size;
783 hi = key_heap_init (nkeys, key_qsort_compare);
786 for (i = 1; i<=nkeys; i++)
787 if ((r = key_file_read (kf[i], rbuf)))
788 key_heap_insert (hi, rbuf, r, kf[i]);
800 for (i = 1; i<=nkeys; i++)
802 extract_get_fname_tmp (zh, rbuf, i);
805 logf (LOG_LOG, "Iterations . . .%7d", hi->no_iterations);
806 logf (LOG_LOG, "Distinct words .%7d", hi->no_diffs);
807 logf (LOG_LOG, "Updates. . . . .%7d", hi->no_updates);
808 logf (LOG_LOG, "Deletions. . . .%7d", hi->no_deletions);
809 logf (LOG_LOG, "Insertions . . .%7d", hi->no_insertions);
810 zh->reg->key_file_no = 0;
812 key_heap_destroy (hi, nkeys);
813 for (i = 1; i<=nkeys; i++)
814 key_file_destroy (kf[i]);