X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=0486bb6eb55198e787c5bb2cc8cb9c102b0503ce;hp=c6897dc0ee6f0d4a3e819091d640ce51c832dfe8;hb=4ed5fbcd29d2a98b048d1d94510b262d352b4f7c;hpb=3f05d9800b0d3b0acc2fcf9b90b71e630d09a90d diff --git a/index/extract.c b/index/extract.c index c6897dc..0486bb6 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,10 +1,131 @@ /* - * Copyright (C) 1994-1996, Index Data I/S + * Copyright (C) 1994-1999, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * * $Log: extract.c,v $ - * Revision 1.60 1996-06-04 10:18:12 adam + * Revision 1.93 1999-05-15 14:36:38 adam + * Updated dictionary. Implemented "compression" of dictionary. + * + * Revision 1.92 1999/03/09 16:27:49 adam + * More work on SDRKit integration. + * + * Revision 1.91 1999/02/12 13:29:22 adam + * Implemented position-flag for registers. + * + * Revision 1.90 1999/02/02 14:50:52 adam + * Updated WIN32 code specific sections. Changed header. + * + * Revision 1.89 1998/10/28 10:54:38 adam + * SDRKit integration. + * + * Revision 1.88 1998/10/16 08:14:29 adam + * Updated record control system. + * + * Revision 1.87 1998/10/15 13:10:33 adam + * Fixed bug in Zebra that caused it to stop indexing when empty + * record was read. + * + * Revision 1.86 1998/10/13 20:33:53 adam + * Fixed one log message and change use ordinal to be an unsigned char. + * + * Revision 1.85 1998/09/22 10:03:41 adam + * Changed result sets to be persistent in the sense that they can + * be re-searched if needed. + * Fixed memory leak in rsm_or. + * + * Revision 1.84 1998/06/11 15:42:22 adam + * Changed the way use attributes are specified in the recordId + * specification. + * + * Revision 1.83 1998/06/08 14:43:10 adam + * Added suport for EXPLAIN Proxy servers - added settings databasePath + * and explainDatabase to facilitate this. Increased maximum number + * of databases and attributes in one register. + * + * Revision 1.82 1998/05/20 10:12:15 adam + * Implemented automatic EXPLAIN database maintenance. + * Modified Zebra to work with ASN.1 compiled version of YAZ. + * + * Revision 1.81 1998/03/11 11:19:04 adam + * Changed the way sequence numbers are generated. + * + * Revision 1.80 1998/03/05 08:45:11 adam + * New result set model and modular ranking system. Moved towards + * descent server API. System information stored as "SGML" records. + * + * Revision 1.79 1998/02/17 10:32:52 adam + * Fixed bug: binary files weren't opened with flag b on NT. + * + * Revision 1.78 1998/02/10 12:03:05 adam + * Implemented Sort. + * + * Revision 1.77 1998/01/12 15:04:08 adam + * The test option (-s) only uses read-lock (and not write lock). + * + * Revision 1.76 1997/10/27 14:33:04 adam + * Moved towards generic character mapping depending on "structure" + * field in abstract syntax file. Fixed a few memory leaks. Fixed + * bug with negative integers when doing searches with relational + * operators. + * + * Revision 1.75 1997/09/17 12:19:12 adam + * Zebra version corresponds to YAZ version 1.4. + * Changed Zebra server so that it doesn't depend on global common_resource. + * + * Revision 1.74 1997/09/09 13:38:06 adam + * Partial port to WIN95/NT. + * + * Revision 1.73 1997/09/04 13:57:20 adam + * New file extract/retrieve method tellf (added). + * Added O_BINARY for open calls. + * + * Revision 1.72 1997/07/15 16:32:29 adam + * Bug fix: Match handler didn't terminate the resulting string! + * + * Revision 1.71 1997/07/15 16:28:41 adam + * Bug fix: storeData didn't work with files with multiple records. + * Bug fix: fixed memory management with records; not really well + * thought through. + * + * Revision 1.70 1997/07/01 13:00:42 adam + * Bug fix in routine searchRecordKey: uninitialized variables. + * + * Revision 1.69 1997/04/29 09:26:03 adam + * Bug fix: generic recordId handling didn't work for compressed internal + * keys. + * + * Revision 1.68 1997/02/12 20:39:45 adam + * Implemented options -f that limits the log to the first + * records. + * Changed some log messages also. + * + * Revision 1.67 1996/11/15 15:02:14 adam + * Minor changes regarding logging. + * + * Revision 1.66 1996/11/14 09:52:21 adam + * Strings in record keys bound by IT_MAX_WORD. + * + * Revision 1.65 1996/11/14 08:57:56 adam + * Reduction of storeKeys area. + * + * Revision 1.64 1996/11/08 11:10:16 adam + * Buffers used during file match got bigger. + * Compressed ISAM support everywhere. + * Bug fixes regarding masking characters in queries. + * Redesigned Regexp-2 queries. + * + * Revision 1.63 1996/10/29 14:09:39 adam + * Use of cisam system - enabled if setting isamc is 1. + * + * Revision 1.62 1996/10/11 10:57:01 adam + * New module recctrl. Used to manage records (extract/retrieval). + * Several files have been moved to the recctrl sub directory. + * + * Revision 1.61 1996/06/06 12:08:37 quinn + * Added showRecord function + * + * Revision 1.60 1996/06/04 10:18:12 adam * Search/scan uses character mapping module. * * Revision 1.59 1996/05/14 15:47:07 adam @@ -217,18 +338,32 @@ */ #include #include +#ifdef WIN32 +#include +#else #include +#endif #include -#include #include +#include +#include #include "index.h" #include "zinfo.h" +#ifndef ZEBRASDR +#define ZEBRASDR 0 +#endif + +#if ZEBRASDR +#include "zebrasdr.h" +#endif + static Dict matchDict; static Records records = NULL; +static SortIdx sortIdx = NULL; static char **key_buf; static size_t ptr_top; @@ -241,7 +376,7 @@ static int records_updated = 0; static int records_deleted = 0; static int records_processed = 0; -static ZebTargetInfo *zti = NULL; +static ZebraExplainInfo zti = NULL; static void logRecord (int showFlag) { @@ -255,8 +390,15 @@ static void logRecord (int showFlag) } } -void key_open (int mem) +static int explain_extract (void *handle, Record drec, data1_node *n); + +int key_open (struct recordGroup *rGroup, int mem) { + BFiles bfs = rGroup->bfs; + int rw = rGroup->flagRw; + data1_handle dh = rGroup->dh; + if (!mem) + mem = atoi(res_get_def (common_resource, "memMax", "4"))*1024*1024; if (mem < 50000) mem = 50000; key_buf = xmalloc (mem); @@ -266,21 +408,34 @@ void key_open (int mem) key_buf_used = 0; key_file_no = 0; - if (!(matchDict = dict_open (GMATCH_DICT, 50, 1))) + if (!(matchDict = dict_open (bfs, GMATCH_DICT, 50, rw, 0))) { logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT); - exit (1); + return -1; } assert (!records); - records = rec_open (1); -#if 1 - zti = zebTargetInfo_open (records, 1); -#endif + records = rec_open (bfs, rw); + if (!records) + { + dict_close (matchDict); + return -1; + } + zti = zebraExplain_open (records, dh, common_resource, + rw, rGroup, explain_extract); + if (!zti) + { + rec_close (&records); + dict_close (matchDict); + return -1; + } + sortIdx = sortIdx_open (bfs, 1); + return 0; } struct encode_info { int sysno; int seqno; + int cmd; char buf[768]; }; @@ -288,6 +443,7 @@ void encode_key_init (struct encode_info *i) { i->sysno = 0; i->seqno = 0; + i->cmd = -1; } char *encode_key_int (int d, char *bp) @@ -329,8 +485,11 @@ void encode_key_write (char *k, struct encode_info *i, FILE *outf) i->sysno = key.sysno; i->seqno = 0; } + else if (!i->seqno && !key.seqno && i->cmd == *k) + return; bp = encode_key_int (key.seqno - i->seqno, bp); i->seqno = key.seqno; + i->cmd = *k; if (fwrite (i->buf, bp - i->buf, 1, outf) != 1) { logf (LOG_FATAL|LOG_ERRNO, "fwrite"); @@ -338,24 +497,49 @@ void encode_key_write (char *k, struct encode_info *i, FILE *outf) } } +#define SORT_EXTRA 0 + +#if SORT_EXTRA +static int key_y_len; + +static int key_y_compare (const void *p1, const void *p2) +{ + int r; + + if ((r = key_compare (*(char**) p1 + key_y_len + 1, + *(char**) p2 + key_y_len + 1))) + return r; + return *(*(char**) p1 + key_y_len) - *(*(char**) p2 + key_y_len); +} + +static int key_x_compare (const void *p1, const void *p2) +{ + return strcmp (*(char**) p1, *(char**) p2); +} +#endif + void key_flush (void) { FILE *outf; char out_fname[200]; char *prevcp, *cp; struct encode_info encode_info; +#if SORT_EXTRA + int i; +#endif if (ptr_i <= 0) return; key_file_no++; logf (LOG_LOG, "sorting section %d", key_file_no); +#if !SORT_EXTRA qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare); - sprintf (out_fname, TEMP_FNAME, key_file_no); + getFnameTmp (out_fname, key_file_no); - if (!(outf = fopen (out_fname, "w"))) + if (!(outf = fopen (out_fname, "wb"))) { - logf (LOG_FATAL|LOG_ERRNO, "fopen (4) %s", out_fname); + logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); exit (1); } logf (LOG_LOG, "writing section %d", key_file_no); @@ -375,6 +559,42 @@ void key_flush (void) else encode_key_write (cp + strlen(cp), &encode_info, outf); } +#else + qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_x_compare); + getFnameTmp (out_fname, key_file_no); + + if (!(outf = fopen (out_fname, "wb"))) + { + logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); + exit (1); + } + logf (LOG_LOG, "writing section %d", key_file_no); + i = ptr_i; + prevcp = key_buf[ptr_top-i]; + while (1) + if (!--i || strcmp (prevcp, key_buf[ptr_top-i])) + { + key_y_len = strlen(prevcp)+1; +#if 0 + logf (LOG_LOG, "key_y_len: %2d %02x %02x %s", + key_y_len, prevcp[0], prevcp[1], 2+prevcp); +#endif + qsort (key_buf + ptr_top-ptr_i, ptr_i - i, + sizeof(char*), key_y_compare); + cp = key_buf[ptr_top-ptr_i]; + --key_y_len; + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + while (--ptr_i > i) + { + cp = key_buf[ptr_top-ptr_i]; + encode_key_write (cp+key_y_len, &encode_info, outf); + } + if (!i) + break; + prevcp = key_buf[ptr_top-ptr_i]; + } +#endif if (fclose (outf)) { logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname); @@ -385,42 +605,56 @@ void key_flush (void) key_buf_used = 0; } -int key_close (void) +int key_close (struct recordGroup *rGroup) { + int rw = rGroup->flagRw; + if (rw) + zebraExplain_runNumberIncrement (zti, 1); + zebraExplain_close (zti, rw, 0); key_flush (); xfree (key_buf); -#if 1 - zebTargetInfo_close (zti, 1); -#endif rec_close (&records); dict_close (matchDict); + sortIdx_close (sortIdx); logRecord (1); return key_file_no; } -static void wordInit (RecWord *p) +static void wordInit (struct recExtractCtrl *p, RecWord *w) { - p->attrSet = 1; - p->attrUse = 1016; - p->which = Word_String; + w->zebra_maps = p->zebra_maps; + w->seqnos = p->seqno; + w->attrSet = VAL_BIB1; + w->attrUse = 1016; + w->reg_type = 'w'; } -struct recKeys { +static struct sortKey { + char *string; + int length; + int attrSet; + int attrUse; + struct sortKey *next; +} *sortKeys = NULL; + +static struct recKeys { int buf_used; int buf_max; char *buf; char prevAttrSet; short prevAttrUse; + int prevSeqNo; } reckeys; -static void addRecordKey (const RecWord *p) +static void addIndexString (RecWord *p, const char *string, int length) { char *dst; - char attrSet; - short attrUse; - size_t i; + unsigned char attrSet; + unsigned short attrUse; int lead = 0; + int diff = 0; + int *pseqno = &p->seqnos[p->reg_type]; if (reckeys.buf_used+1024 > reckeys.buf_max) { @@ -444,7 +678,15 @@ static void addRecordKey (const RecWord *p) lead |= 2; else reckeys.prevAttrUse = attrUse; - +#if 1 + diff = 1 + *pseqno - reckeys.prevSeqNo; + if (diff >= 1 && diff <= 15) + lead |= (diff << 2); + else + diff = 0; +#endif + reckeys.prevSeqNo = *pseqno; + *dst++ = lead; if (!(lead & 1)) @@ -457,43 +699,178 @@ static void addRecordKey (const RecWord *p) memcpy (dst, &attrUse, sizeof(attrUse)); dst += sizeof(attrUse); } - switch (p->which) - { - case Word_String: - *dst++ = 'w'; - break; - case Word_Phrase: - *dst++ = 'p'; - break; - case Word_Numeric: - *dst++ = 'n'; - } - for (i = 0; p->u.string[i]; i++) - *dst++ = p->u.string[i]; + *dst++ = p->reg_type; + memcpy (dst, string, length); + dst += length; *dst++ = '\0'; - memcpy (dst, &p->seqno, sizeof(p->seqno)); - dst += sizeof(p->seqno); + if (!diff) + { + memcpy (dst, pseqno, sizeof(*pseqno)); + dst += sizeof(*pseqno); + } reckeys.buf_used = dst - reckeys.buf; + if (*pseqno) + (*pseqno)++; } -static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, - const char *databaseName) +static void addSortString (RecWord *p, const char *string, int length) { - char attrSet = -1; - short attrUse = -1; - int off = 0; + struct sortKey *sk; + + for (sk = sortKeys; sk; sk = sk->next) + if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse) + return; + + sk = xmalloc (sizeof(*sk)); + sk->next = sortKeys; + sortKeys = sk; + + sk->string = xmalloc (length); + sk->length = length; + memcpy (sk->string, string, length); + + sk->attrSet = p->attrSet; + sk->attrUse = p->attrUse; +} + +static void addString (RecWord *p, const char *string, int length) +{ + assert (length > 0); + if (zebra_maps_is_sort (p->zebra_maps, p->reg_type)) + addSortString (p, string, length); + else + addIndexString (p, string, length); +} + +static void addIncompleteField (RecWord *p) +{ + const char *b = p->string; + int remain = p->length; + const char **map = 0; - if (zebTargetInfo_curDatabase (zti, databaseName)) + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + + while (map) { - if (zebTargetInfo_newDatabase (zti, databaseName)) - abort (); + char buf[IT_MAX_WORD+1]; + int i, remain; + + /* Skip spaces */ + while (map && *map && **map == *CHR_SPACE) + { + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + else + map = 0; + } + if (!map) + break; + i = 0; + while (map && *map && **map != *CHR_SPACE) + { + const char *cp = *map; + + while (i < IT_MAX_WORD && *cp) + buf[i++] = *(cp++); + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + else + map = 0; + } + if (!i) + return; + addString (p, buf, i); + } +} + +static void addCompleteField (RecWord *p) +{ + const char *b = p->string; + char buf[IT_MAX_WORD+1]; + const char **map = 0; + int i = 0, remain = p->length; + + if (remain > 0) + map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain); + + while (remain > 0 && i < IT_MAX_WORD) + { + while (map && *map && **map == *CHR_SPACE) + { + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + else + map = 0; + } + if (!map) + break; + + if (i && i < IT_MAX_WORD) + buf[i++] = *CHR_SPACE; + while (map && *map && **map != *CHR_SPACE) + { + const char *cp = *map; + + if (i >= IT_MAX_WORD) + break; + while (i < IT_MAX_WORD && *cp) + buf[i++] = *(cp++); + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, + remain); + else + map = 0; + } } + if (!i) + return; + addString (p, buf, i); +} + +static void addRecordKey (RecWord *p) +{ + if (zebra_maps_is_complete (p->zebra_maps, p->reg_type)) + addCompleteField (p); + else + addIncompleteField(p); +} + +static void flushSortKeys (SYSNO sysno, int cmd) +{ + struct sortKey *sk = sortKeys; + + sortIdx_sysno (sortIdx, sysno); + while (sk) + { + struct sortKey *sk_next = sk->next; + sortIdx_type (sortIdx, sk->attrUse); + sortIdx_add (sortIdx, sk->string, sk->length); + xfree (sk->string); + xfree (sk); + sk = sk_next; + } + sortKeys = NULL; +} + +static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys) +{ + unsigned char attrSet = (unsigned char) -1; + unsigned short attrUse = (unsigned short) -1; + int seqno = 0; + int off = 0; + + zebraExplain_recordCountIncrement (zti, cmd ? 1 : -1); while (off < reckeys->buf_used) { const char *src = reckeys->buf + off; struct it_key key; - int lead; + int lead, ch; lead = *src++; @@ -512,20 +889,26 @@ static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, ++ptr_i; key_buf[ptr_top-ptr_i] = (char*)key_buf + key_buf_used; - lead = zebTargetInfo_lookupSU (zti, attrSet, attrUse); - if (lead < 0) - lead = zebTargetInfo_addSU (zti, attrSet, attrUse); - assert (lead > 0); - ((char*) key_buf) [key_buf_used++] = lead; + ch = zebraExplain_lookupSU (zti, attrSet, attrUse); + if (ch < 0) + ch = zebraExplain_addSU (zti, attrSet, attrUse); + assert (ch > 0); + key_buf_used += key_SU_code (ch, ((char*)key_buf) + key_buf_used); + while (*src) ((char*)key_buf) [key_buf_used++] = *src++; src++; ((char*)key_buf) [key_buf_used++] = '\0'; - ((char*) key_buf)[key_buf_used++] = cmd; - memcpy (&key.seqno, src, sizeof(key.seqno)); - src += sizeof(key.seqno); + if (lead & 60) + seqno += ((lead>>2) & 15)-1; + else + { + memcpy (&seqno, src, sizeof(seqno)); + src += sizeof(seqno); + } + key.seqno = seqno; key.sysno = sysno; memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key)); key_buf_used += sizeof(key); @@ -535,40 +918,48 @@ static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, } static const char **searchRecordKey (struct recKeys *reckeys, - int attrSetS, int attrUseS) + int attrSetS, int attrUseS) { static const char *ws[32]; int off = 0; int startSeq = -1; int i; + short attrUse; + char attrSet; + int seqno = 0; for (i = 0; i<32; i++) ws[i] = NULL; while (off < reckeys->buf_used) { - const char *src = reckeys->buf + off; - char attrSet; - short attrUse; - int seqno; - const char *wstart; - - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); + const char *src = reckeys->buf + off; + const char *wstart; + int lead; + + lead = *src++; + if (!(lead & 1)) + { + memcpy (&attrSet, src, sizeof(attrSet)); + src += sizeof(attrSet); + } + if (!(lead & 2)) + { + memcpy (&attrUse, src, sizeof(attrUse)); + src += sizeof(attrUse); + } wstart = src; while (*src++) ; - - memcpy (&seqno, src, sizeof(seqno)); - src += sizeof(seqno); - -#if 0 - logf (LOG_LOG, "(%d,%d) %d %s", attrSet, attrUse, seqno, wstart); -#endif + if (lead & 60) + seqno += ((lead>>2) & 15)-1; + else + { + memcpy (&seqno, src, sizeof(seqno)); + src += sizeof(seqno); + } if (attrUseS == attrUse && attrSetS == attrSet) { int woff; @@ -588,11 +979,13 @@ static const char **searchRecordKey (struct recKeys *reckeys, } struct file_read_info { - off_t file_max; - off_t file_offset; - off_t file_moffset; + off_t file_max; /* maximum offset so far */ + off_t file_offset; /* current offset */ + off_t file_moffset; /* offset of rec/rec boundary */ int file_more; int fd; + char *sdrbuf; + int sdrmax; }; static struct file_read_info *file_read_start (int fd) @@ -602,12 +995,13 @@ static struct file_read_info *file_read_start (int fd) fi->fd = fd; fi->file_max = 0; fi->file_moffset = 0; + fi->sdrbuf = 0; + fi->sdrmax = 0; return fi; } static void file_read_stop (struct file_read_info *fi) { - assert (fi); xfree (fi); } @@ -615,15 +1009,32 @@ static off_t file_seek (void *handle, off_t offset) { struct file_read_info *p = handle; p->file_offset = offset; + if (p->sdrbuf) + return offset; return lseek (p->fd, offset, SEEK_SET); } +static off_t file_tell (void *handle) +{ + struct file_read_info *p = handle; + return p->file_offset; +} + static int file_read (void *handle, char *buf, size_t count) { struct file_read_info *p = handle; int fd = p->fd; int r; - r = read (fd, buf, count); + if (p->sdrbuf) + { + r = count; + if (r > p->sdrmax - p->file_offset) + r = p->sdrmax - p->file_offset; + if (r) + memcpy (buf, p->sdrbuf + p->file_offset, r); + } + else + r = read (fd, buf, count); if (r > 0) { p->file_offset += r; @@ -638,7 +1049,7 @@ static void file_begin (void *handle) struct file_read_info *p = handle; p->file_offset = p->file_moffset; - if (p->file_moffset) + if (!p->sdrbuf && p->file_moffset) lseek (p->fd, p->file_moffset, SEEK_SET); p->file_more = 0; } @@ -652,26 +1063,13 @@ static void file_end (void *handle, off_t offset) p->file_moffset = offset; } -static int atois (const char **s) -{ - int val = 0, c; - while ( (c=**s) >= '0' && c <= '9') - { - val = val*10 + c - '0'; - ++(*s); - } - return val; -} - static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, - const char *fname, - const char *spec) + const char *fname, const char *spec) { static char dstBuf[2048]; char *dst = dstBuf; const char *s = spec; static const char **w; - int i; while (1) { @@ -681,21 +1079,39 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, break; if (*s == '(') { + char attset_str[64], attname_str[64]; + data1_attset *attset; + int i; char matchFlag[32]; - int attrSet, attrUse; + int attSet = 1, attUse = 1; int first = 1; s++; - attrSet = atois (&s); - if (*s != ',') - { - logf (LOG_WARN, "Missing , in match criteria %s in group %s", - spec, rGroup->groupName ? rGroup->groupName : "none"); - return NULL; - } - s++; - attrUse = atois (&s); - w = searchRecordKey (reckeys, attrSet, attrUse); + for (i = 0; *s && *s != ',' && *s != ')'; s++) + if (i < 63) + attset_str[i++] = *s; + attset_str[i] = '\0'; + + if (*s == ',') + { + s++; + for (i = 0; *s && *s != ')'; s++) + if (i < 63) + attname_str[i++] = *s; + attname_str[i] = '\0'; + } + + if ((attset = data1_get_attset (rGroup->dh, attset_str))) + { + data1_att *att; + attSet = attset->reference; + att = data1_getattbyname(rGroup->dh, attset, attname_str); + if (att) + attUse = att->value; + else + attUse = atoi (attname_str); + } + w = searchRecordKey (reckeys, attSet, attUse); assert (w); if (*s == ')') @@ -725,7 +1141,7 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, if (first) { logf (LOG_WARN, "Record didn't contain match" - " fields in (%d,%d)", attrSet, attrUse); + " fields in (%s,%s)", attset_str, attname_str); return NULL; } } @@ -792,12 +1208,13 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, fname, rGroup->groupName ? rGroup->groupName : "none"); return NULL; } + *dst = '\0'; return dstBuf; } struct recordLogInfo { const char *fname; - char *op; + int recordOffset; struct recordGroup *rGroup; }; @@ -808,61 +1225,95 @@ static void recordLogPreamble (int level, const char *msg, void *info) if (level & LOG_LOG) return ; - if (p->op) - fprintf (outf, "%s of ", p->op); - fprintf (outf, "%s type %s\n", p->rGroup->recordType, p->fname); + fprintf (outf, "File %s, offset %d, type %s\n", + p->rGroup->recordType, p->recordOffset, p->fname); log_event_start (NULL, NULL); } +void addSchema (struct recExtractCtrl *p, Odr_oid *oid) +{ + zebraExplain_addSchema (zti, oid); +} + static int recordExtract (SYSNO *sysno, const char *fname, struct recordGroup *rGroup, int deleteFlag, - struct file_read_info *fi, RecType recType, - char *subType) + struct file_read_info *fi, + RecType recType, char *subType) { - struct recExtractCtrl extractCtrl; + RecordAttr *recordAttr; int r; char *matchStr; SYSNO sysnotmp; - off_t recordOffset = 0; Record rec; struct recordLogInfo logInfo; + off_t recordOffset = 0; - logInfo.fname = fname; - logInfo.op = NULL; - logInfo.rGroup = rGroup; - log_event_start (recordLogPreamble, &logInfo); - if (fi->fd != -1) { + struct recExtractCtrl extractCtrl; + /* we are going to read from a file, so prepare the extraction */ - extractCtrl.fh = fi; - extractCtrl.subType = subType; - extractCtrl.init = wordInit; - extractCtrl.add = addRecordKey; - - reckeys.buf_used = 0; - reckeys.prevAttrUse = -1; - reckeys.prevAttrSet = -1; - - recordOffset = fi->file_moffset; - extractCtrl.offset = recordOffset; - extractCtrl.readf = file_read; - extractCtrl.seekf = file_seek; - extractCtrl.endf = file_end; + int i; + + reckeys.buf_used = 0; + reckeys.prevAttrUse = -1; + reckeys.prevAttrSet = -1; + reckeys.prevSeqNo = 0; + + recordOffset = fi->file_moffset; + extractCtrl.offset = fi->file_moffset; + extractCtrl.readf = file_read; + extractCtrl.seekf = file_seek; + extractCtrl.tellf = file_tell; + extractCtrl.endf = file_end; + extractCtrl.fh = fi; + extractCtrl.subType = subType; + extractCtrl.init = wordInit; + extractCtrl.addWord = addRecordKey; + extractCtrl.addSchema = addSchema; + extractCtrl.dh = rGroup->dh; + for (i = 0; i<256; i++) + { + if (zebra_maps_is_positioned(rGroup->zebra_maps, i)) + extractCtrl.seqno[i] = 1; + else + extractCtrl.seqno[i] = 0; + } + extractCtrl.zebra_maps = rGroup->zebra_maps; + extractCtrl.flagShowRecords = !rGroup->flagRw; + + if (!rGroup->flagRw) + printf ("File: %s %ld\n", fname, (long) recordOffset); + + logInfo.fname = fname; + logInfo.recordOffset = recordOffset; + logInfo.rGroup = rGroup; + log_event_start (recordLogPreamble, &logInfo); + r = (*recType->extract)(&extractCtrl); + log_event_start (NULL, NULL); + if (r) { /* error occured during extraction ... */ - logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r); + if (rGroup->flagRw && + records_processed < rGroup->fileVerboseLimit) + { + logf (LOG_WARN, "fail %s %s %ld code = %d", rGroup->recordType, + fname, (long) recordOffset, r); + } return 0; } if (reckeys.buf_used == 0) { /* the extraction process returned no information - the record - is probably empty */ - logf (LOG_WARN, "Empty file %s", fname); - return 0; + is probably empty - unless flagShowRecords is in use */ + if (!rGroup->flagRw) + return 1; + logf (LOG_WARN, "No keys generated for file %s", fname); + logf (LOG_WARN, " The file is probably empty"); + return 1; } } @@ -898,21 +1349,26 @@ static int recordExtract (SYSNO *sysno, const char *fname, /* new record */ if (deleteFlag) { - logf (LOG_LOG, "Cannot delete new record"); + logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType, + fname, (long) recordOffset); + logf (LOG_WARN, "cannot delete record above (seems new)"); return 1; } - logInfo.op = "add"; - if (rGroup->fileVerboseFlag) - logf (LOG_LOG, "add %s %s+%ld", rGroup->recordType, + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "add %s %s %ld", rGroup->recordType, fname, (long) recordOffset); rec = rec_new (records); + *sysno = rec->sysno; + recordAttr = rec_init_attr (zti, rec); + if (matchStr) { dict_insert (matchDict, matchStr, sizeof(*sysno), sysno); } - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); + flushRecordKeys (*sysno, 1, &reckeys); + flushSortKeys (*sysno, 1); records_inserted++; } @@ -923,20 +1379,33 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec = rec_get (records, *sysno); assert (rec); + + recordAttr = rec_init_attr (zti, rec); + + if (recordAttr->runNumber == zebraExplain_runNumberIncrement (zti, 0)) + { + logf (LOG_LOG, "skipped %s %s %ld", rGroup->recordType, + fname, (long) recordOffset); + rec_rm (&rec); + logRecord (0); + return 1; + } delkeys.buf_used = rec->size[recInfo_delKeys]; delkeys.buf = rec->info[recInfo_delKeys]; - flushRecordKeys (*sysno, 0, &delkeys, rec->info[recInfo_databaseName]); + flushSortKeys (*sysno, 0); + flushRecordKeys (*sysno, 0, &delkeys); if (deleteFlag) { /* record going to be deleted */ - logInfo.op = "delete"; if (!delkeys.buf_used) { - logf (LOG_WARN, "cannot delete; storeKeys false"); + logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType, + fname, (long) recordOffset); + logf (LOG_WARN, "cannot delete file above, storeKeys false"); } else { - if (rGroup->fileVerboseFlag) + if (records_processed < rGroup->fileVerboseLimit) logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType, fname, (long) recordOffset); records_deleted++; @@ -944,23 +1413,25 @@ static int recordExtract (SYSNO *sysno, const char *fname, dict_delete (matchDict, matchStr); rec_del (records, &rec); } + rec_rm (&rec); logRecord (0); return 1; } else { /* record going to be updated */ - logInfo.op = "update"; if (!delkeys.buf_used) { - logf (LOG_WARN, "cannot update; storeKeys false"); + logf (LOG_LOG, "update %s %s %ld", rGroup->recordType, + fname, (long) recordOffset); + logf (LOG_WARN, "cannot update file above, storeKeys false"); } else { - if (rGroup->fileVerboseFlag) + if (records_processed < rGroup->fileVerboseLimit) logf (LOG_LOG, "update %s %s %ld", rGroup->recordType, fname, (long) recordOffset); - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); + flushRecordKeys (*sysno, 1, &reckeys); records_updated++; } } @@ -997,23 +1468,33 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec->size[recInfo_delKeys] = 0; } + /* save file size of original record */ + zebraExplain_recordBytesIncrement (zti, - recordAttr->recordSize); + recordAttr->recordSize = fi->file_moffset - recordOffset; + if (!recordAttr->recordSize) + recordAttr->recordSize = fi->file_max - recordOffset; + zebraExplain_recordBytesIncrement (zti, recordAttr->recordSize); + + /* set run-number for this record */ + recordAttr->runNumber = zebraExplain_runNumberIncrement (zti, 0); + /* update store data */ xfree (rec->info[recInfo_storeData]); if (rGroup->flagStoreData == 1) { - rec->size[recInfo_storeData] = fi->file_max; - rec->info[recInfo_storeData] = xmalloc (fi->file_max); + rec->size[recInfo_storeData] = recordAttr->recordSize; + rec->info[recInfo_storeData] = xmalloc (recordAttr->recordSize); if (lseek (fi->fd, recordOffset, SEEK_SET) < 0) { - logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s", fname, - (long) recordOffset); + logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s", + (long) recordOffset, fname); exit (1); } - if (read (fi->fd, rec->info[recInfo_storeData], fi->file_max) - < fi->file_max) + if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize) + < recordAttr->recordSize) { logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", - fi->file_max, fname); + recordAttr->recordSize, fname); exit (1); } } @@ -1028,11 +1509,7 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]); /* update offset */ - xfree (rec->info[recInfo_offset]); - - rec->size[recInfo_offset] = sizeof(recordOffset); - rec->info[recInfo_offset] = xmalloc (sizeof(recordOffset)); - memcpy (rec->info[recInfo_offset], &recordOffset, sizeof(recordOffset)); + recordAttr->recordOffset = recordOffset; /* commit this record */ rec_put (records, &rec); @@ -1063,12 +1540,10 @@ int fileExtract (SYSNO *sysno, const char *fname, logf (LOG_DEBUG, "fileExtract %s", fname); /* determine file extension */ + *ext = '\0'; for (i = strlen(fname); --i >= 0; ) if (fname[i] == '/') - { - strcpy (ext, ""); break; - } else if (fname[i] == '.') { strcpy (ext, fname+i+1); @@ -1081,21 +1556,19 @@ int fileExtract (SYSNO *sysno, const char *fname, if (!(rGroup->recordType = res_get (common_resource, ext_res))) { sprintf (ext_res, "%srecordType", gprefix); - if (!(rGroup->recordType = res_get (common_resource, ext_res))) - { - if (rGroup->fileVerboseFlag) - logf (LOG_LOG, "? %s", fname); - return 0; - } + rGroup->recordType = res_get (common_resource, ext_res); } } if (!rGroup->recordType) { - if (rGroup->fileVerboseFlag) - logf (LOG_LOG, "? record %s", fname); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "? %s", fname); return 0; } - if (!(recType = recType_byName (rGroup->recordType, subType))) + if (!*rGroup->recordType) + return 0; + if (!(recType = + recType_byName (rGroup->recTypes, rGroup->recordType, subType))) { logf (LOG_WARN, "No such record type: %s", rGroup->recordType); return 0; @@ -1121,6 +1594,20 @@ int fileExtract (SYSNO *sysno, const char *fname, if (!rGroup->databaseName) rGroup->databaseName = "Default"; + /* determine if explain database */ + + sprintf (ext_res, "%sexplainDatabase", gprefix); + rGroup->explainDatabase = + atoi (res_get_def (common_resource, ext_res, "0")); + + /* announce database */ + if (zebraExplain_curDatabase (zti, rGroup->databaseName)) + { + if (zebraExplain_newDatabase (zti, rGroup->databaseName, + rGroup->explainDatabase)) + abort (); + } + if (rGroup->flagStoreData == -1) { const char *sval; @@ -1152,11 +1639,64 @@ int fileExtract (SYSNO *sysno, const char *fname, if (rGroup->flagStoreKeys == -1) rGroup->flagStoreKeys = 0; +#if ZEBRASDR + if (rGroup->useSDR) + { + ZebraSdrHandle h; + char xname[128], *xp; + + strncpy (xname, fname, 127); + if (!(xp = strchr (xname, '.'))) + return 0; + *xp = '\0'; + if (strcmp (xp+1, "sdr.bits")) + return 0; + + h = zebraSdr_open (xname); + if (!h) + { + logf (LOG_WARN, "sdr open %s", xname); + return 0; + } + for (;;) + { + unsigned char *buf; + char sdr_name[128]; + int r, segmentno; + + segmentno = zebraSdr_segment (h, 0); + sprintf (sdr_name, "%%%s.%d", xname, segmentno); + +#if 0 + if (segmentno > 20) + break; +#endif + r = zebraSdr_read (h, &buf); + + if (!r) + break; + + fi = file_read_start (0); + fi->sdrbuf = buf; + fi->sdrmax = r; + do + { + file_begin (fi); + r = recordExtract (sysno, sdr_name, rGroup, deleteFlag, fi, + recType, subType); + } while (r && !sysno && fi->file_more); + file_read_stop (fi); + free (buf); + } + zebraSdr_close (h); + return 1; + } +#endif if (sysno && deleteFlag) fd = -1; else { - if ((fd = open (fname, O_RDONLY)) == -1) + if ((fd = open (fname, O_BINARY|O_RDONLY)) == -1) { logf (LOG_WARN|LOG_ERRNO, "open %s", fname); return 0; @@ -1169,10 +1709,59 @@ int fileExtract (SYSNO *sysno, const char *fname, r = recordExtract (sysno, fname, rGroup, deleteFlag, fi, recType, subType); } while (r && !sysno && fi->file_more); - log_event_start (NULL, NULL); file_read_stop (fi); if (fd != -1) close (fd); return r; } +static int explain_extract (void *handle, Record rec, data1_node *n) +{ + struct recordGroup *rGroup = (struct recordGroup*) handle; + struct recExtractCtrl extractCtrl; + int i; + + if (zebraExplain_curDatabase (zti, rec->info[recInfo_databaseName])) + { + if (zebraExplain_newDatabase (zti, rec->info[recInfo_databaseName], 0)) + abort (); + } + + reckeys.buf_used = 0; + reckeys.prevAttrUse = -1; + reckeys.prevAttrSet = -1; + reckeys.prevSeqNo = 0; + + extractCtrl.init = wordInit; + extractCtrl.addWord = addRecordKey; + extractCtrl.addSchema = addSchema; + extractCtrl.dh = rGroup->dh; + for (i = 0; i<256; i++) + extractCtrl.seqno[i] = 0; + extractCtrl.zebra_maps = rGroup->zebra_maps; + extractCtrl.flagShowRecords = !rGroup->flagRw; + + grs_extract_tree(&extractCtrl, n); + + logf (LOG_DEBUG, "flush explain record, sysno=%d", rec->sysno); + + if (rec->size[recInfo_delKeys]) + { + struct recKeys delkeys; + + delkeys.buf_used = rec->size[recInfo_delKeys]; + delkeys.buf = rec->info[recInfo_delKeys]; + flushSortKeys (rec->sysno, 0); + flushRecordKeys (rec->sysno, 0, &delkeys); + } + flushRecordKeys (rec->sysno, 1, &reckeys); + flushSortKeys (rec->sysno, 1); + + xfree (rec->info[recInfo_delKeys]); + rec->size[recInfo_delKeys] = reckeys.buf_used; + rec->info[recInfo_delKeys] = reckeys.buf; + reckeys.buf = NULL; + reckeys.buf_max = 0; + + return 0; +}