2 * Copyright (C) 1994-1995, Index Data I/S
4 * Sebastian Hammer, Adam Dickmeiss
7 * Revision 1.39 1995-12-05 13:20:18 adam
8 * Bug fix: file_read sometimes returned early EOF.
10 * Revision 1.38 1995/12/04 17:59:21 adam
11 * More work on regular expression conversion.
13 * Revision 1.37 1995/12/04 14:22:27 adam
14 * Extra arg to recType_byName.
15 * Started work on new regular expression parsed input to
18 * Revision 1.36 1995/11/30 08:34:29 adam
19 * Started work on commit facility.
20 * Changed a few malloc/free to xmalloc/xfree.
22 * Revision 1.35 1995/11/28 14:26:21 adam
23 * Bug fix: recordId with constant wasn't right.
24 * Bug fix: recordId dictionary entry wasn't deleted when needed.
26 * Revision 1.34 1995/11/28 09:09:38 adam
27 * Zebra config renamed.
28 * Use setting 'recordId' to identify record now.
29 * Bug fix in recindex.c: rec_release_blocks was invokeded even
30 * though the blocks were already released.
31 * File traversal properly deletes records when needed.
33 * Revision 1.33 1995/11/27 09:56:20 adam
34 * Record info elements better enumerated. Internal store of records.
36 * Revision 1.32 1995/11/25 10:24:05 adam
37 * More record fields - they are enumerated now.
38 * New options: flagStoreData flagStoreKey.
40 * Revision 1.31 1995/11/24 11:31:35 adam
41 * Commands add & del read filenames from stdin if source directory is
43 * Match criteria supports 'constant' strings.
45 * Revision 1.30 1995/11/22 17:19:16 adam
46 * Record management uses the bfile system.
48 * Revision 1.29 1995/11/21 15:01:14 adam
49 * New general match criteria implemented.
50 * New feature: document groups.
52 * Revision 1.28 1995/11/21 09:20:30 adam
53 * Yet more work on record match.
55 * Revision 1.27 1995/11/20 16:59:45 adam
56 * New update method: the 'old' keys are saved for each records.
58 * Revision 1.26 1995/11/20 11:56:24 adam
59 * Work on new traversal.
61 * Revision 1.25 1995/11/16 15:34:54 adam
62 * Uses new record management system in both indexer and server.
64 * Revision 1.24 1995/11/15 19:13:08 adam
65 * Work on record management.
67 * Revision 1.23 1995/10/27 14:00:10 adam
68 * Implemented detection of database availability.
70 * Revision 1.22 1995/10/17 18:02:07 adam
71 * New feature: databases. Implemented as prefix to words in dictionary.
73 * Revision 1.21 1995/10/10 12:24:38 adam
74 * Temporary sort files are compressed.
76 * Revision 1.20 1995/10/06 13:52:05 adam
77 * Bug fixes. Handler may abort further scanning.
79 * Revision 1.19 1995/10/04 12:55:16 adam
80 * Bug fix in ranked search. Use=Any keys inserted.
82 * Revision 1.18 1995/10/04 09:37:08 quinn
85 * Revision 1.17 1995/10/03 14:28:57 adam
86 * Buffered read in extract works.
88 * Revision 1.16 1995/10/03 14:28:45 adam
89 * Work on more effecient read handler in extract.
91 * Revision 1.15 1995/10/02 15:42:53 adam
92 * Extract uses file descriptors instead of FILE pointers.
94 * Revision 1.14 1995/10/02 15:29:13 adam
95 * More logging in file_extract.
97 * Revision 1.13 1995/09/29 14:01:39 adam
100 * Revision 1.12 1995/09/28 14:22:56 adam
101 * Sort uses smaller temporary files.
103 * Revision 1.11 1995/09/28 12:10:31 adam
104 * Bug fixes. Field prefix used in queries.
106 * Revision 1.10 1995/09/28 09:19:41 adam
107 * xfree/xmalloc used everywhere.
108 * Extract/retrieve method seems to work for text records.
110 * Revision 1.9 1995/09/27 12:22:28 adam
111 * More work on extract in record control.
112 * Field name is not in isam keys but in prefix in dictionary words.
114 * Revision 1.8 1995/09/14 07:48:22 adam
115 * Record control management.
117 * Revision 1.7 1995/09/11 13:09:32 adam
118 * More work on relevance feedback.
120 * Revision 1.6 1995/09/08 14:52:27 adam
121 * Minor changes. Dictionary is lower case now.
123 * Revision 1.5 1995/09/06 16:11:16 adam
124 * Option: only one word key per file.
126 * Revision 1.4 1995/09/05 15:28:39 adam
127 * More work on search engine.
129 * Revision 1.3 1995/09/04 12:33:41 adam
130 * Various cleanup. YAZ util used instead.
132 * Revision 1.2 1995/09/04 09:10:34 adam
133 * More work on index add/del/update.
134 * Merge sort implemented.
135 * Initial work on z39 server.
137 * Revision 1.1 1995/09/01 14:06:35 adam
138 * Split of work into more files.
147 #include <alexutil.h>
151 #include "recindex.h"
153 static Dict matchDict;
155 static Records records = NULL;
157 static char **key_buf;
158 static size_t ptr_top;
160 static size_t key_buf_used;
161 static int key_file_no;
163 static int records_inserted = 0;
164 static int records_updated = 0;
165 static int records_deleted = 0;
167 #define MATCH_DICT "match"
169 void key_open (int mem)
173 key_buf = xmalloc (mem);
174 ptr_top = mem/sizeof(char*);
180 if (!(matchDict = dict_open (MATCH_DICT, 20, 1)))
182 logf (LOG_FATAL, "dict_open fail of %s", MATCH_DICT);
186 records = rec_open (1);
195 void encode_key_init (struct encode_info *i)
201 char *encode_key_int (int d, char *bp)
210 else if (d <= 4194303)
212 *bp++ = 128 + (d>>16);
213 *bp++ = (d>>8) & 255;
218 *bp++ = 192 + (d>>24);
219 *bp++ = (d>>16) & 255;
220 *bp++ = (d>>8) & 255;
226 void encode_key_write (char *k, struct encode_info *i, FILE *outf)
231 while ((*bp++ = *k++))
233 memcpy (&key, k+1, sizeof(struct it_key));
234 bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp);
235 if (i->sysno != key.sysno)
237 i->sysno = key.sysno;
240 bp = encode_key_int (key.seqno - i->seqno, bp);
241 i->seqno = key.seqno;
242 if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
244 logf (LOG_FATAL|LOG_ERRNO, "fwrite");
249 void key_flush (void)
254 struct encode_info encode_info;
260 logf (LOG_LOG, "sorting section %d", key_file_no);
261 qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare);
262 sprintf (out_fname, TEMP_FNAME, key_file_no);
264 if (!(outf = fopen (out_fname, "w")))
266 logf (LOG_FATAL|LOG_ERRNO, "fopen (4) %s", out_fname);
269 logf (LOG_LOG, "writing section %d", key_file_no);
270 prevcp = cp = key_buf[ptr_top-ptr_i];
272 encode_key_init (&encode_info);
273 encode_key_write (cp, &encode_info, outf);
276 cp = key_buf[ptr_top-ptr_i];
277 if (strcmp (cp, prevcp))
279 encode_key_init (&encode_info);
280 encode_key_write (cp, &encode_info, outf);
284 encode_key_write (cp + strlen(cp), &encode_info, outf);
288 logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname);
291 logf (LOG_LOG, "finished section %d", key_file_no);
300 rec_close (&records);
301 dict_close (matchDict);
303 logf (LOG_LOG, "Records inserted %6d", records_inserted);
304 logf (LOG_LOG, "Records updated %6d", records_updated);
305 logf (LOG_LOG, "Records deleted %6d", records_deleted);
309 static void wordInit (RecWord *p)
313 p->which = Word_String;
322 static void addRecordKey (const RecWord *p)
329 if (reckeys.buf_used+1024 > reckeys.buf_max)
333 b = xmalloc (reckeys.buf_max += 65000);
334 if (reckeys.buf_used > 0)
335 memcpy (b, reckeys.buf, reckeys.buf_used);
339 dst = reckeys.buf + reckeys.buf_used;
343 attrSet = p->attrSet;
344 memcpy (dst, &attrSet, sizeof(attrSet));
345 dst += sizeof(attrSet);
347 attrUse = p->attrUse;
348 memcpy (dst, &attrUse, sizeof(attrUse));
349 dst += sizeof(attrUse);
351 for (i = 0; p->u.string[i]; i++)
352 *dst++ = p->u.string[i];
355 memcpy (dst, &p->seqno, sizeof(p->seqno));
356 dst += sizeof(p->seqno);
362 reckeys.buf_used = dst - reckeys.buf;
365 static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys,
366 const char *databaseName)
369 while (off < reckeys->buf_used)
371 const char *src = reckeys->buf + off;
376 memcpy (&attrSet, src, sizeof(attrSet));
377 src += sizeof(attrSet);
379 memcpy (&attrUse, src, sizeof(attrUse));
380 src += sizeof(attrUse);
382 if (key_buf_used + 1024 > (ptr_top-ptr_i)*sizeof(char*))
385 key_buf[ptr_top-ptr_i] = (char*)key_buf + key_buf_used;
386 key_buf_used += index_word_prefix ((char*)key_buf + key_buf_used,
387 attrSet, attrUse, databaseName);
389 ((char*)key_buf) [key_buf_used++] = index_char_cvt (*src++);
391 ((char*)key_buf) [key_buf_used++] = '\0';
393 ((char*) key_buf)[key_buf_used++] = cmd;
395 memcpy (&key.seqno, src, sizeof(key.seqno));
396 src += sizeof(key.seqno);
398 memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key));
399 key_buf_used += sizeof(key);
400 off = src - reckeys->buf;
402 assert (off == reckeys->buf_used);
405 static const char **searchRecordKey (struct recKeys *reckeys,
406 int attrSetS, int attrUseS)
408 static const char *ws[32];
413 for (i = 0; i<32; i++)
416 while (off < reckeys->buf_used)
418 const char *src = reckeys->buf + off;
424 memcpy (&attrSet, src, sizeof(attrSet));
425 src += sizeof(attrSet);
427 memcpy (&attrUse, src, sizeof(attrUse));
428 src += sizeof(attrUse);
434 memcpy (&seqno, src, sizeof(seqno));
435 src += sizeof(seqno);
438 logf (LOG_LOG, "(%d,%d) %d %s", attrSet, attrUse, seqno, wstart);
440 if (attrUseS == attrUse && attrSetS == attrSet)
447 woff = seqno - startSeq;
448 if (woff >= 0 && woff < 31)
452 off = src - reckeys->buf;
454 assert (off == reckeys->buf_used);
458 static void addRecordKeyAny (const RecWord *p)
460 if (p->attrSet != 1 || p->attrUse != 1016)
464 memcpy (&w, p, sizeof(w));
473 #define FILE_READ_BUFSIZE 4096
475 static int file_noread;
476 #if FILE_READ_BUFSIZE
477 static char *file_buf;
478 static int file_offset;
479 static int file_bufsize;
482 static void file_read_start (int fd)
485 #if FILE_READ_BUFSIZE
487 file_buf = xmalloc (FILE_READ_BUFSIZE);
488 file_bufsize = read (fd, file_buf, FILE_READ_BUFSIZE);
492 static void file_read_stop (int fd)
494 #if FILE_READ_BUFSIZE
500 static int file_read (int fd, char *buf, size_t count)
502 #if FILE_READ_BUFSIZE
503 int l = file_bufsize - file_offset;
509 memcpy (buf, file_buf + file_offset, l);
511 if (count > FILE_READ_BUFSIZE)
513 if ((r = read (fd, buf + l, count)) == -1)
515 logf (LOG_FATAL|LOG_ERRNO, "read");
523 file_bufsize = r = read (fd, file_buf, FILE_READ_BUFSIZE);
526 logf (LOG_FATAL|LOG_ERRNO, "read");
532 memcpy (buf + l, file_buf, r);
533 file_noread += (l+r);
539 memcpy (buf + l, file_buf, count - l);
540 file_noread += count;
544 memcpy (buf, file_buf + file_offset, count);
545 file_offset += count;
546 file_noread += count;
550 r = read (fd, buf, count);
557 static int atois (const char **s)
560 while ( (c=**s) >= '0' && c <= '9')
562 val = val*10 + c - '0';
568 static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup,
572 static char dstBuf[2048];
574 const char *s = spec;
575 static const char **w;
580 while (*s == ' ' || *s == '\t')
587 int attrSet, attrUse;
591 attrSet = atois (&s);
594 logf (LOG_WARN, "Missing , in match criteria %s in group %s",
595 spec, rGroup->groupName ? rGroup->groupName : "none");
599 attrUse = atois (&s);
600 w = searchRecordKey (reckeys, attrSet, attrUse);
605 for (i = 0; i<32; i++)
610 logf (LOG_WARN, "Missing ) in match criteria %s in group %s",
611 spec, rGroup->groupName ? rGroup->groupName : "none");
616 for (i = 0; i<32; i++)
617 if (matchFlag[i] && w[i])
629 logf (LOG_WARN, "Record in file %s didn't contain match"
630 " fields in (%d,%d)", fname, attrSet, attrUse);
638 const char *spec_src = NULL;
639 const char *s1 = ++s;
640 while (*s1 && *s1 != ' ' && *s1 != '\t')
646 memcpy (special, s, spec_len);
647 special[spec_len] = '\0';
650 if (!strcmp (special, "group"))
651 spec_src = rGroup->groupName;
652 else if (!strcmp (special, "database"))
653 spec_src = rGroup->databaseName;
654 else if (!strcmp (special, "filename"))
656 else if (!strcmp (special, "type"))
657 spec_src = rGroup->recordType;
662 strcpy (dst, spec_src);
663 dst += strlen (spec_src);
666 else if (*s == '\"' || *s == '\'')
668 int stopMarker = *s++;
672 while (*s && *s != stopMarker)
675 tmpString[i++] = *s++;
680 strcpy (dst, tmpString);
681 dst += strlen (tmpString);
685 logf (LOG_WARN, "Syntax error in match criteria %s in group %s",
686 spec, rGroup->groupName ? rGroup->groupName : "none");
693 logf (LOG_WARN, "No match criteria for record %s in group %s",
694 fname, rGroup->groupName ? rGroup->groupName : "none");
700 static int recordExtract (SYSNO *sysno, const char *fname,
701 struct recordGroup *rGroup, int deleteFlag,
702 int fd, RecType recType, char *subType)
704 struct recExtractCtrl extractCtrl;
714 extractCtrl.subType = subType;
715 extractCtrl.init = wordInit;
716 extractCtrl.add = addRecordKeyAny;
718 reckeys.buf_used = 0;
719 extractCtrl.readf = file_read;
720 r = (*recType->extract)(&extractCtrl);
724 logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);
729 /* perform match if sysno not known and if match criteria is specified */
736 if (rGroup->recordId && *rGroup->recordId)
740 matchStr = fileMatchStr (&reckeys, rGroup, fname,
744 rinfo = dict_lookup (matchDict, matchStr);
746 memcpy (sysno, rinfo+1, sizeof(*sysno));
750 logf (LOG_WARN, "Record not inserted");
761 logf (LOG_LOG, "? %s", fname);
764 logf (LOG_LOG, "add %s %s", rGroup->recordType, fname);
765 rec = rec_new (records);
770 dict_insert (matchDict, matchStr, sizeof(*sysno), sysno);
772 flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName);
778 struct recKeys delkeys;
780 rec = rec_get (records, *sysno);
782 delkeys.buf_used = rec->size[recInfo_delKeys];
783 delkeys.buf = rec->info[recInfo_delKeys];
784 flushRecordKeys (*sysno, 0, &delkeys, rec->info[recInfo_databaseName]);
787 if (!delkeys.buf_used)
789 logf (LOG_WARN, "cannot delete %s - no delete keys", fname);
794 logf (LOG_LOG, "delete %s %s", rGroup->recordType, fname);
797 dict_insert (matchDict, matchStr, sizeof(sysnoz), &sysnoz);
798 rec_del (records, &rec);
804 if (!delkeys.buf_used)
806 logf (LOG_WARN, "cannot update %s - no delete keys",
811 logf (LOG_LOG, "update %s %s", rGroup->recordType,
813 flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName);
818 xfree (rec->info[recInfo_fileType]);
819 rec->info[recInfo_fileType] =
820 rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]);
822 xfree (rec->info[recInfo_filename]);
823 rec->info[recInfo_filename] =
824 rec_strdup (fname, &rec->size[recInfo_filename]);
826 xfree (rec->info[recInfo_delKeys]);
827 if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1)
829 rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used);
830 rec->size[recInfo_delKeys] = reckeys.buf_used;
831 memcpy (rec->info[recInfo_delKeys], reckeys.buf,
832 rec->size[recInfo_delKeys]);
836 rec->info[recInfo_delKeys] = NULL;
837 rec->size[recInfo_delKeys] = 0;
840 xfree (rec->info[recInfo_storeData]);
841 if (rGroup->flagStoreData == 1)
843 rec->size[recInfo_storeData] = file_noread;
844 rec->info[recInfo_storeData] = xmalloc (file_noread);
845 #if FILE_READ_BUFSIZE
846 if (file_noread < FILE_READ_BUFSIZE)
847 memcpy (rec->info[recInfo_storeData], file_buf, file_noread);
851 if (lseek (fd, 0L, SEEK_SET) < 0)
853 logf (LOG_ERRNO|LOG_FATAL, "seek to 0 in %s", fname);
856 if (read (fd, rec->info[recInfo_storeData], file_noread)
859 logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s",
867 rec->info[recInfo_storeData] = NULL;
868 rec->size[recInfo_storeData] = 0;
870 xfree (rec->info[recInfo_databaseName]);
871 rec->info[recInfo_databaseName] =
872 rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]);
874 rec_put (records, &rec);
878 int fileExtract (SYSNO *sysno, const char *fname,
879 const struct recordGroup *rGroupP, int deleteFlag)
887 struct recordGroup rGroupM;
888 struct recordGroup *rGroup = &rGroupM;
890 memcpy (rGroup, rGroupP, sizeof(*rGroupP));
892 if (!rGroup->groupName || !*rGroup->groupName)
895 sprintf (gprefix, "%s.", rGroup->groupName);
897 logf (LOG_DEBUG, "fileExtract %s", fname);
899 /* determine file extension */
900 for (i = strlen(fname); --i >= 0; )
906 else if (fname[i] == '.')
908 strcpy (ext, fname+i+1);
911 /* determine file type - depending on extension */
912 if (!rGroup->recordType)
914 sprintf (ext_res, "%srecordType.%s", gprefix, ext);
915 if (!(rGroup->recordType = res_get (common_resource, ext_res)))
917 sprintf (ext_res, "%srecordType", gprefix);
918 if (!(rGroup->recordType = res_get (common_resource, ext_res)))
920 logf (LOG_LOG, "? %s", fname);
925 if (!rGroup->recordType)
927 logf (LOG_LOG, "? record %s", fname);
930 if (!(recType = recType_byName (rGroup->recordType, subType)))
932 logf (LOG_WARN, "No such record type: %s", rGroup->recordType);
936 /* determine match criteria */
937 if (!rGroup->recordId)
939 sprintf (ext_res, "%srecordId.%s", gprefix, ext);
940 rGroup->recordId = res_get (common_resource, ext_res);
943 /* determine database name */
944 if (!rGroup->databaseName)
946 sprintf (ext_res, "%sdatabase.%s", gprefix, ext);
947 if (!(rGroup->databaseName = res_get (common_resource, ext_res)))
949 sprintf (ext_res, "%sdatabase", gprefix);
950 rGroup->databaseName = res_get (common_resource, ext_res);
953 if (!rGroup->databaseName)
954 rGroup->databaseName = "Default";
956 if (rGroup->flagStoreData == -1)
959 sprintf (ext_res, "%sstoreData.%s", gprefix, ext);
960 if (!(sval = res_get (common_resource, ext_res)))
962 sprintf (ext_res, "%sstoreData", gprefix);
963 sval = res_get (common_resource, ext_res);
966 rGroup->flagStoreData = atoi (sval);
968 if (rGroup->flagStoreData == -1)
969 rGroup->flagStoreData = 0;
971 if (rGroup->flagStoreKeys == -1)
975 sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext);
976 if (!(sval = res_get (common_resource, ext_res)))
978 sprintf (ext_res, "%sstoreKeys", gprefix);
979 sval = res_get (common_resource, ext_res);
982 rGroup->flagStoreKeys = atoi (sval);
984 if (rGroup->flagStoreKeys == -1)
985 rGroup->flagStoreKeys = 0;
987 if (sysno && deleteFlag)
991 if ((fd = open (fname, O_RDONLY)) == -1)
993 logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
997 file_read_start (fd);
998 recordExtract (sysno, fname, rGroup, deleteFlag, fd, recType, subType);