X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=2f4bf846a6a72f8bebb39c6f3a1d049df9113de6;hp=8b189cae712fd4d7c57de23dbe518cff748467e7;hb=2b1851bd5565e3d21f9cf9a37661a584c063b75f;hpb=4e2fc80e64f3d7895a871ea5b08aa80045a59f2c diff --git a/index/extract.c b/index/extract.c index 8b189ca..2f4bf84 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,450 +1,122 @@ /* - * Copyright (C) 1994-1995, Index Data I/S + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * - * $Log: extract.c,v $ - * Revision 1.42 1995-12-07 17:38:46 adam - * Work locking mechanisms for concurrent updates/commit. - * - * Revision 1.41 1995/12/06 16:06:42 adam - * Better diagnostics. Work on 'real' dictionary deletion. - * - * Revision 1.40 1995/12/05 16:57:40 adam - * More work on regular patterns. - * - * Revision 1.39 1995/12/05 13:20:18 adam - * Bug fix: file_read sometimes returned early EOF. - * - * Revision 1.38 1995/12/04 17:59:21 adam - * More work on regular expression conversion. - * - * Revision 1.37 1995/12/04 14:22:27 adam - * Extra arg to recType_byName. - * Started work on new regular expression parsed input to - * structured records. - * - * Revision 1.36 1995/11/30 08:34:29 adam - * Started work on commit facility. - * Changed a few malloc/free to xmalloc/xfree. - * - * Revision 1.35 1995/11/28 14:26:21 adam - * Bug fix: recordId with constant wasn't right. - * Bug fix: recordId dictionary entry wasn't deleted when needed. - * - * Revision 1.34 1995/11/28 09:09:38 adam - * Zebra config renamed. - * Use setting 'recordId' to identify record now. - * Bug fix in recindex.c: rec_release_blocks was invokeded even - * though the blocks were already released. - * File traversal properly deletes records when needed. - * - * Revision 1.33 1995/11/27 09:56:20 adam - * Record info elements better enumerated. Internal store of records. - * - * Revision 1.32 1995/11/25 10:24:05 adam - * More record fields - they are enumerated now. - * New options: flagStoreData flagStoreKey. - * - * Revision 1.31 1995/11/24 11:31:35 adam - * Commands add & del read filenames from stdin if source directory is - * empty. - * Match criteria supports 'constant' strings. - * - * Revision 1.30 1995/11/22 17:19:16 adam - * Record management uses the bfile system. - * - * Revision 1.29 1995/11/21 15:01:14 adam - * New general match criteria implemented. - * New feature: document groups. - * - * Revision 1.28 1995/11/21 09:20:30 adam - * Yet more work on record match. - * - * Revision 1.27 1995/11/20 16:59:45 adam - * New update method: the 'old' keys are saved for each records. - * - * Revision 1.26 1995/11/20 11:56:24 adam - * Work on new traversal. - * - * Revision 1.25 1995/11/16 15:34:54 adam - * Uses new record management system in both indexer and server. - * - * Revision 1.24 1995/11/15 19:13:08 adam - * Work on record management. - * - * Revision 1.23 1995/10/27 14:00:10 adam - * Implemented detection of database availability. - * - * Revision 1.22 1995/10/17 18:02:07 adam - * New feature: databases. Implemented as prefix to words in dictionary. - * - * Revision 1.21 1995/10/10 12:24:38 adam - * Temporary sort files are compressed. - * - * Revision 1.20 1995/10/06 13:52:05 adam - * Bug fixes. Handler may abort further scanning. - * - * Revision 1.19 1995/10/04 12:55:16 adam - * Bug fix in ranked search. Use=Any keys inserted. - * - * Revision 1.18 1995/10/04 09:37:08 quinn - * Fixed bug. - * - * Revision 1.17 1995/10/03 14:28:57 adam - * Buffered read in extract works. - * - * Revision 1.16 1995/10/03 14:28:45 adam - * Work on more effecient read handler in extract. - * - * Revision 1.15 1995/10/02 15:42:53 adam - * Extract uses file descriptors instead of FILE pointers. - * - * Revision 1.14 1995/10/02 15:29:13 adam - * More logging in file_extract. - * - * Revision 1.13 1995/09/29 14:01:39 adam - * Bug fixes. - * - * Revision 1.12 1995/09/28 14:22:56 adam - * Sort uses smaller temporary files. - * - * Revision 1.11 1995/09/28 12:10:31 adam - * Bug fixes. Field prefix used in queries. - * - * Revision 1.10 1995/09/28 09:19:41 adam - * xfree/xmalloc used everywhere. - * Extract/retrieve method seems to work for text records. - * - * Revision 1.9 1995/09/27 12:22:28 adam - * More work on extract in record control. - * Field name is not in isam keys but in prefix in dictionary words. - * - * Revision 1.8 1995/09/14 07:48:22 adam - * Record control management. - * - * Revision 1.7 1995/09/11 13:09:32 adam - * More work on relevance feedback. - * - * Revision 1.6 1995/09/08 14:52:27 adam - * Minor changes. Dictionary is lower case now. - * - * Revision 1.5 1995/09/06 16:11:16 adam - * Option: only one word key per file. - * - * Revision 1.4 1995/09/05 15:28:39 adam - * More work on search engine. - * - * Revision 1.3 1995/09/04 12:33:41 adam - * Various cleanup. YAZ util used instead. - * - * Revision 1.2 1995/09/04 09:10:34 adam - * More work on index add/del/update. - * Merge sort implemented. - * Initial work on z39 server. - * - * Revision 1.1 1995/09/01 14:06:35 adam - * Split of work into more files. - * + * $Id: extract.c,v 1.114 2002-04-04 14:14:13 adam Exp $ */ #include #include +#ifdef WIN32 +#include +#else #include +#endif #include -#include -#include -#include #include "index.h" +#include -#include "recindex.h" - -static Dict matchDict; - -static Records records = NULL; - -static char **key_buf; -static size_t ptr_top; -static size_t ptr_i; -static size_t key_buf_used; -static int key_file_no; +#if _FILE_OFFSET_BITS == 64 +#define PRINTF_OFF_T "%Ld" +#else +#define PRINTF_OFF_T "%ld" +#endif static int records_inserted = 0; static int records_updated = 0; static int records_deleted = 0; +static int records_processed = 0; -void key_open (int mem) -{ - if (mem < 50000) - mem = 50000; - key_buf = xmalloc (mem); - ptr_top = mem/sizeof(char*); - ptr_i = 0; - - key_buf_used = 0; - key_file_no = 0; - - if (!(matchDict = dict_open (GMATCH_DICT, 50, 1))) - { - logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT); - exit (1); - } - assert (!records); - records = rec_open (1); -} - -struct encode_info { - int sysno; - int seqno; - char buf[512]; -}; - -void encode_key_init (struct encode_info *i) -{ - i->sysno = 0; - i->seqno = 0; -} - -char *encode_key_int (int d, char *bp) -{ - if (d <= 63) - *bp++ = d; - else if (d <= 16383) - { - *bp++ = 64 + (d>>8); - *bp++ = d & 255; - } - else if (d <= 4194303) - { - *bp++ = 128 + (d>>16); - *bp++ = (d>>8) & 255; - *bp++ = d & 255; - } - else - { - *bp++ = 192 + (d>>24); - *bp++ = (d>>16) & 255; - *bp++ = (d>>8) & 255; - *bp++ = d & 255; - } - return bp; -} - -void encode_key_write (char *k, struct encode_info *i, FILE *outf) -{ - struct it_key key; - char *bp = i->buf; - - while ((*bp++ = *k++)) - ; - memcpy (&key, k+1, sizeof(struct it_key)); - bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp); - if (i->sysno != key.sysno) - { - i->sysno = key.sysno; - i->seqno = 0; - } - bp = encode_key_int (key.seqno - i->seqno, bp); - i->seqno = key.seqno; - if (fwrite (i->buf, bp - i->buf, 1, outf) != 1) - { - logf (LOG_FATAL|LOG_ERRNO, "fwrite"); - exit (1); - } -} - -void key_flush (void) -{ - FILE *outf; - char out_fname[200]; - char *prevcp, *cp; - struct encode_info encode_info; - - if (ptr_i <= 0) - return; - - key_file_no++; - logf (LOG_LOG, "sorting section %d", key_file_no); - qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare); - sprintf (out_fname, TEMP_FNAME, key_file_no); - - if (!(outf = fopen (out_fname, "w"))) - { - logf (LOG_FATAL|LOG_ERRNO, "fopen (4) %s", out_fname); - exit (1); - } - logf (LOG_LOG, "writing section %d", key_file_no); - prevcp = cp = key_buf[ptr_top-ptr_i]; - - encode_key_init (&encode_info); - encode_key_write (cp, &encode_info, outf); - while (--ptr_i > 0) - { - cp = key_buf[ptr_top-ptr_i]; - if (strcmp (cp, prevcp)) - { - encode_key_init (&encode_info); - encode_key_write (cp, &encode_info, outf); - prevcp = cp; - } - else - encode_key_write (cp + strlen(cp), &encode_info, outf); - } - if (fclose (outf)) - { - logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname); - exit (1); - } - logf (LOG_LOG, "finished section %d", key_file_no); - ptr_i = 0; - key_buf_used = 0; -} - -int key_close (void) -{ - key_flush (); - xfree (key_buf); - rec_close (&records); - dict_close (matchDict); - - logf (LOG_LOG, "Records inserted %6d", records_inserted); - logf (LOG_LOG, "Records updated %6d", records_updated); - logf (LOG_LOG, "Records deleted %6d", records_deleted); - return key_file_no; -} - -static void wordInit (RecWord *p) -{ - p->attrSet = 1; - p->attrUse = 1016; - p->which = Word_String; -} - -struct recKeys { - int buf_used; - int buf_max; - char *buf; -} reckeys; - -static void addRecordKey (const RecWord *p) +static void logRecord (int showFlag) { - char *dst; - char attrSet; - short attrUse; - size_t i; - - if (reckeys.buf_used+1024 > reckeys.buf_max) - { - char *b; - - b = xmalloc (reckeys.buf_max += 65000); - if (reckeys.buf_used > 0) - memcpy (b, reckeys.buf, reckeys.buf_used); - xfree (reckeys.buf); - reckeys.buf = b; - } - dst = reckeys.buf + reckeys.buf_used; - switch (p->which) + if (!showFlag) + ++records_processed; + if (showFlag || !(records_processed % 1000)) { - case Word_String: - attrSet = p->attrSet; - memcpy (dst, &attrSet, sizeof(attrSet)); - dst += sizeof(attrSet); - - attrUse = p->attrUse; - memcpy (dst, &attrUse, sizeof(attrUse)); - dst += sizeof(attrUse); - - for (i = 0; p->u.string[i]; i++) - *dst++ = p->u.string[i]; - *dst++ = '\0'; - - memcpy (dst, &p->seqno, sizeof(p->seqno)); - dst += sizeof(p->seqno); - - break; - default: - return; + logf (LOG_LOG, "Records: %7d i/u/d %d/%d/%d", + records_processed, records_inserted, records_updated, + records_deleted); } - reckeys.buf_used = dst - reckeys.buf; } -static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, - const char *databaseName) +static void extract_init (struct recExtractCtrl *p, RecWord *w) { - int off = 0; - while (off < reckeys->buf_used) - { - const char *src = reckeys->buf + off; - char attrSet; - short attrUse; - struct it_key key; - - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); - - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); - - if (key_buf_used + 1024 > (ptr_top-ptr_i)*sizeof(char*)) - key_flush (); - ++ptr_i; - key_buf[ptr_top-ptr_i] = (char*)key_buf + key_buf_used; - key_buf_used += index_word_prefix ((char*)key_buf + key_buf_used, - attrSet, attrUse, databaseName); - while (*src) - ((char*)key_buf) [key_buf_used++] = index_char_cvt (*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); - key.sysno = sysno; - memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key)); - key_buf_used += sizeof(key); - off = src - reckeys->buf; - } - assert (off == reckeys->buf_used); + w->zebra_maps = p->zebra_maps; + w->seqnos = p->seqno; + w->attrSet = VAL_BIB1; + w->attrUse = 1016; + w->reg_type = 'w'; + w->extractCtrl = p; } -static const char **searchRecordKey (struct recKeys *reckeys, - int attrSetS, int attrUseS) +static const char **searchRecordKey (ZebraHandle zh, + struct recKeys *reckeys, + int attrSetS, int attrUseS) { static const char *ws[32]; int off = 0; int startSeq = -1; int i; + int seqno = 0; +#if SU_SCHEME + int chS, ch; +#else + short attrUse; + char attrSet; +#endif for (i = 0; i<32; i++) ws[i] = NULL; - + +#if SU_SCHEME + chS = zebraExplain_lookupSU (zh->reg->zei, attrSetS, attrUseS); + if (chS < 0) + return ws; +#endif 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 SU_SCHEME + if ((lead & 3)<3) + { + memcpy (&ch, src, sizeof(ch)); + src += sizeof(ch); + } +#else + if (!(lead & 1)) + { + memcpy (&attrSet, src, sizeof(attrSet)); + src += sizeof(attrSet); + } + if (!(lead & 2)) + { + memcpy (&attrUse, src, sizeof(attrUse)); + src += sizeof(attrUse); + } +#endif 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); + if (lead & 60) + seqno += ((lead>>2) & 15)-1; + else + { + memcpy (&seqno, src, sizeof(seqno)); + src += sizeof(seqno); + } + if ( +#if SU_SCHEME + ch == chS +#else + attrUseS == attrUse && attrSetS == attrSet #endif - if (attrUseS == attrUse && attrSetS == attrSet) + ) { int woff; @@ -462,125 +134,100 @@ static const char **searchRecordKey (struct recKeys *reckeys, return ws; } -static void addRecordKeyAny (const RecWord *p) -{ - if (p->attrSet != 1 || p->attrUse != 1016) - { - RecWord w; +struct file_read_info { + 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; +}; - memcpy (&w, p, sizeof(w)); - w.attrSet = 1; - w.attrUse = 1016; - addRecordKey (&w); - } - addRecordKey (p); +static struct file_read_info *file_read_start (int fd) +{ + struct file_read_info *fi = (struct file_read_info *) + xmalloc (sizeof(*fi)); + + fi->fd = fd; + fi->file_max = 0; + fi->file_moffset = 0; + fi->sdrbuf = 0; + fi->sdrmax = 0; + return fi; } - -#define FILE_READ_BUFSIZE 4096 - -static int file_noread; -#if FILE_READ_BUFSIZE -static char *file_buf; -static int file_offset; -static int file_bufsize; -#endif - -static void file_read_start (int fd) +static void file_read_stop (struct file_read_info *fi) { - file_noread = 0; -#if FILE_READ_BUFSIZE - file_offset = 0; - file_buf = xmalloc (FILE_READ_BUFSIZE); - file_bufsize = read (fd, file_buf, FILE_READ_BUFSIZE); -#endif + xfree (fi); } -static void file_read_stop (int fd) +static off_t file_seek (void *handle, off_t offset) { -#if FILE_READ_BUFSIZE - xfree (file_buf); - file_buf = NULL; -#endif + struct file_read_info *p = (struct file_read_info *) handle; + p->file_offset = offset; + if (p->sdrbuf) + return offset; + return lseek (p->fd, offset, SEEK_SET); } -static int file_read (int fd, char *buf, size_t count) +static off_t file_tell (void *handle) { -#if FILE_READ_BUFSIZE - int l = file_bufsize - file_offset; + struct file_read_info *p = (struct file_read_info *) handle; + return p->file_offset; +} - if (count > l) +static int file_read (void *handle, char *buf, size_t count) +{ + struct file_read_info *p = (struct file_read_info *) handle; + int fd = p->fd; + int r; + if (p->sdrbuf) { - int r; - if (l > 0) - memcpy (buf, file_buf + file_offset, l); - count = count-l; - if (count > FILE_READ_BUFSIZE) - { - if ((r = read (fd, buf + l, count)) == -1) - { - logf (LOG_FATAL|LOG_ERRNO, "read"); - exit (1); - } - file_bufsize = 0; - file_offset = 0; - file_noread += l+r; - return l+r; - } - file_bufsize = r = read (fd, file_buf, FILE_READ_BUFSIZE); - if (r == -1) - { - logf (LOG_FATAL|LOG_ERRNO, "read"); - exit (1); - } - else if (r <= count) - { - file_offset = r; - memcpy (buf + l, file_buf, r); - file_noread += l+r; - return l+r; - } - else - { - file_offset = count; - memcpy (buf + l, file_buf, count - l); - file_noread += count; - return count; - } + 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); } - memcpy (buf, file_buf + file_offset, count); - file_offset += count; - file_noread += count; - return count; -#else - int r; - r = read (fd, buf, count); + else + r = read (fd, buf, count); if (r > 0) - file_noread += r; + { + p->file_offset += r; + if (p->file_offset > p->file_max) + p->file_max = p->file_offset; + } return r; -#endif } -static int atois (const char **s) +static void file_begin (void *handle) { - int val = 0, c; - while ( (c=**s) >= '0' && c <= '9') - { - val = val*10 + c - '0'; - ++(*s); - } - return val; + struct file_read_info *p = (struct file_read_info *) handle; + + p->file_offset = p->file_moffset; + if (!p->sdrbuf && p->file_moffset) + lseek (p->fd, p->file_moffset, SEEK_SET); + p->file_more = 0; +} + +static void file_end (void *handle, off_t offset) +{ + struct file_read_info *p = (struct file_read_info *) handle; + + assert (p->file_more == 0); + p->file_more = 1; + p->file_moffset = offset; } -static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, - const char *fname, - const char *spec) +static char *fileMatchStr (ZebraHandle zh, + struct recKeys *reckeys, struct recordGroup *rGroup, + const char *fname, const char *spec) { - static char dstBuf[2048]; + static char dstBuf[2048]; /* static here ??? */ char *dst = dstBuf; const char *s = spec; static const char **w; - int i; while (1) { @@ -590,21 +237,39 @@ 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 (zh->reg->dh, attset_str))) + { + data1_att *att; + attSet = attset->reference; + att = data1_getattbyname(zh->reg->dh, attset, attname_str); + if (att) + attUse = att->value; + else + attUse = atoi (attname_str); + } + w = searchRecordKey (zh, reckeys, attSet, attUse); assert (w); if (*s == ')') @@ -633,8 +298,8 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, } if (first) { - logf (LOG_WARN, "Record in file %s didn't contain match" - " fields in (%d,%d)", fname, attrSet, attrUse); + logf (LOG_WARN, "Record didn't contain match" + " fields in (%s,%s)", attset_str, attname_str); return NULL; } } @@ -701,36 +366,115 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, fname, rGroup->groupName ? rGroup->groupName : "none"); return NULL; } + *dst = '\0'; return dstBuf; } -static int recordExtract (SYSNO *sysno, const char *fname, - struct recordGroup *rGroup, int deleteFlag, - int fd, RecType recType, char *subType) +struct recordLogInfo { + const char *fname; + int recordOffset; + struct recordGroup *rGroup; +}; + +static void recordLogPreamble (int level, const char *msg, void *info) { - struct recExtractCtrl extractCtrl; + struct recordLogInfo *p = (struct recordLogInfo *) info; + FILE *outf = yaz_log_file (); + + if (level & LOG_LOG) + return ; + fprintf (outf, "File %s, offset %d, type %s\n", + p->rGroup->recordType, p->recordOffset, p->fname); + log_event_start (NULL, NULL); +} + + +static int recordExtract (ZebraHandle zh, + SYSNO *sysno, const char *fname, + struct recordGroup *rGroup, int deleteFlag, + struct file_read_info *fi, + RecType recType, char *subType, void *clientData) +{ + RecordAttr *recordAttr; int r; char *matchStr; SYSNO sysnotmp; Record rec; + struct recordLogInfo logInfo; + off_t recordOffset = 0; - if (fd != -1) + if (fi->fd != -1) { - extractCtrl.fd = fd; - /* extract keys */ - extractCtrl.subType = subType; - extractCtrl.init = wordInit; - extractCtrl.add = addRecordKeyAny; - - reckeys.buf_used = 0; - extractCtrl.readf = file_read; - r = (*recType->extract)(&extractCtrl); - - if (r) - { - logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r); + struct recExtractCtrl extractCtrl; + + /* we are going to read from a file, so prepare the extraction */ + int i; + + zh->reg->keys.buf_used = 0; + zh->reg->keys.prevAttrUse = -1; + zh->reg->keys.prevAttrSet = -1; + zh->reg->keys.prevSeqNo = 0; + zh->reg->sortKeys = 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 = extract_init; + extractCtrl.tokenAdd = extract_token_add; + extractCtrl.schemaAdd = extract_schema_add; + extractCtrl.dh = zh->reg->dh; + extractCtrl.handle = zh; + for (i = 0; i<256; i++) + { + if (zebra_maps_is_positioned(zh->reg->zebra_maps, i)) + extractCtrl.seqno[i] = 1; + else + extractCtrl.seqno[i] = 0; + } + extractCtrl.zebra_maps = zh->reg->zebra_maps; + extractCtrl.flagShowRecords = !rGroup->flagRw; + + if (!rGroup->flagRw) + printf ("File: %s " PRINTF_OFF_T "\n", fname, recordOffset); + + logInfo.fname = fname; + logInfo.recordOffset = recordOffset; + logInfo.rGroup = rGroup; + log_event_start (recordLogPreamble, &logInfo); + + r = (*recType->extract)(clientData, &extractCtrl); + + log_event_start (NULL, NULL); + + if (r == RECCTRL_EXTRACT_EOF) + return 0; + else if (r == RECCTRL_EXTRACT_ERROR) + { + /* error occured during extraction ... */ + if (rGroup->flagRw && + records_processed < rGroup->fileVerboseLimit) + { + logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + } return 0; } + if (zh->reg->keys.buf_used == 0) + { + /* the extraction process returned no information - the record + is probably empty - unless flagShowRecords is in use */ + if (!rGroup->flagRw) + return 1; + + logf (LOG_WARN, "empty %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + return 1; + } } /* perform match if sysno not known and if match criteria is specified */ @@ -744,99 +488,141 @@ static int recordExtract (SYSNO *sysno, const char *fname, { char *rinfo; - matchStr = fileMatchStr (&reckeys, rGroup, fname, + matchStr = fileMatchStr (zh, &zh->reg->keys, rGroup, fname, rGroup->recordId); if (matchStr) { - rinfo = dict_lookup (matchDict, matchStr); + rinfo = dict_lookup (zh->reg->matchDict, matchStr); if (rinfo) memcpy (sysno, rinfo+1, sizeof(*sysno)); } else { - logf (LOG_WARN, "Record not inserted"); + logf (LOG_WARN, "Bad match criteria"); return 0; } } } - /* new record ? */ if (! *sysno) { + /* new record */ if (deleteFlag) { - logf (LOG_LOG, "? %s", fname); + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + logf (LOG_WARN, "cannot delete record above (seems new)"); return 1; } - logf (LOG_LOG, "add %s %s", rGroup->recordType, fname); - rec = rec_new (records); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + rec = rec_new (zh->reg->records); + *sysno = rec->sysno; + recordAttr = rec_init_attr (zh->reg->zei, rec); + if (matchStr) { - dict_insert (matchDict, matchStr, sizeof(*sysno), sysno); + dict_insert (zh->reg->matchDict, matchStr, sizeof(*sysno), sysno); } - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); + extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); + extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); records_inserted++; } else { + /* record already exists */ struct recKeys delkeys; - rec = rec_get (records, *sysno); - + rec = rec_get (zh->reg->records, *sysno); + assert (rec); + + recordAttr = rec_init_attr (zh->reg->zei, rec); + + if (recordAttr->runNumber == + zebraExplain_runNumberIncrement (zh->reg->zei, 0)) + { + logf (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + extract_flushSortKeys (zh, *sysno, -1, &zh->reg->sortKeys); + 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]); + extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys); + extract_flushRecordKeys (zh, *sysno, 0, &delkeys); if (deleteFlag) { + /* record going to be deleted */ if (!delkeys.buf_used) { - logf (LOG_WARN, "cannot delete %s: storeKeys false", - fname); + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + logf (LOG_WARN, "cannot delete file above, storeKeys false"); } else { - logf (LOG_LOG, "delete %s %s", rGroup->recordType, fname); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); records_deleted++; if (matchStr) - dict_delete (matchDict, matchStr); - rec_del (records, &rec); + dict_delete (zh->reg->matchDict, matchStr); + rec_del (zh->reg->records, &rec); } + rec_rm (&rec); + logRecord (0); return 1; } else { + /* record going to be updated */ if (!delkeys.buf_used) { - logf (LOG_WARN, "cannot update %s: storeKeys false", - fname); + logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + logf (LOG_WARN, "cannot update file above, storeKeys false"); } else { - logf (LOG_LOG, "update %s %s", rGroup->recordType, - fname); - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); records_updated++; } } } + /* update file type */ xfree (rec->info[recInfo_fileType]); rec->info[recInfo_fileType] = rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]); + /* update filename */ xfree (rec->info[recInfo_filename]); rec->info[recInfo_filename] = rec_strdup (fname, &rec->size[recInfo_filename]); + /* update delete keys */ xfree (rec->info[recInfo_delKeys]); - if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1) + if (zh->reg->keys.buf_used > 0 && rGroup->flagStoreKeys == 1) { +#if 1 + rec->size[recInfo_delKeys] = zh->reg->keys.buf_used; + rec->info[recInfo_delKeys] = zh->reg->keys.buf; + zh->reg->keys.buf = NULL; + zh->reg->keys.buf_max = 0; +#else rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used); rec->size[recInfo_delKeys] = reckeys.buf_used; memcpy (rec->info[recInfo_delKeys], reckeys.buf, rec->size[recInfo_delKeys]); +#endif } else { @@ -844,29 +630,38 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec->size[recInfo_delKeys] = 0; } + /* save file size of original record */ + zebraExplain_recordBytesIncrement (zh->reg->zei, + - recordAttr->recordSize); + recordAttr->recordSize = fi->file_moffset - recordOffset; + if (!recordAttr->recordSize) + recordAttr->recordSize = fi->file_max - recordOffset; + zebraExplain_recordBytesIncrement (zh->reg->zei, + recordAttr->recordSize); + + /* set run-number for this record */ + recordAttr->runNumber = zebraExplain_runNumberIncrement (zh->reg->zei, + 0); + + /* update store data */ xfree (rec->info[recInfo_storeData]); if (rGroup->flagStoreData == 1) { - rec->size[recInfo_storeData] = file_noread; - rec->info[recInfo_storeData] = xmalloc (file_noread); -#if FILE_READ_BUFSIZE - if (file_noread < FILE_READ_BUFSIZE) - memcpy (rec->info[recInfo_storeData], file_buf, file_noread); - else -#endif + rec->size[recInfo_storeData] = recordAttr->recordSize; + rec->info[recInfo_storeData] = (char *) + xmalloc (recordAttr->recordSize); + if (lseek (fi->fd, recordOffset, SEEK_SET) < 0) { - if (lseek (fd, 0L, SEEK_SET) < 0) - { - logf (LOG_ERRNO|LOG_FATAL, "seek to 0 in %s", fname); - exit (1); - } - if (read (fd, rec->info[recInfo_storeData], file_noread) - < file_noread) - { - logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", - file_noread, fname); - exit (1); - } + logf (LOG_ERRNO|LOG_FATAL, "seek to " PRINTF_OFF_T " in %s", + recordOffset, fname); + exit (1); + } + if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize) + < recordAttr->recordSize) + { + logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", + recordAttr->recordSize, fname); + exit (1); } } else @@ -874,18 +669,24 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec->info[recInfo_storeData] = NULL; rec->size[recInfo_storeData] = 0; } + /* update database name */ xfree (rec->info[recInfo_databaseName]); rec->info[recInfo_databaseName] = rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]); - rec_put (records, &rec); + /* update offset */ + recordAttr->recordOffset = recordOffset; + + /* commit this record */ + rec_put (zh->reg->records, &rec); + logRecord (0); return 1; } -int fileExtract (SYSNO *sysno, const char *fname, +int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, const struct recordGroup *rGroupP, int deleteFlag) { - int i, fd; + int r, i, fd; char gprefix[128]; char ext[128]; char ext_res[128]; @@ -893,6 +694,8 @@ int fileExtract (SYSNO *sysno, const char *fname, RecType recType; struct recordGroup rGroupM; struct recordGroup *rGroup = &rGroupM; + struct file_read_info *fi; + void *clientData; memcpy (rGroup, rGroupP, sizeof(*rGroupP)); @@ -904,12 +707,10 @@ 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); @@ -919,22 +720,23 @@ int fileExtract (SYSNO *sysno, const char *fname, if (!rGroup->recordType) { sprintf (ext_res, "%srecordType.%s", gprefix, ext); - if (!(rGroup->recordType = res_get (common_resource, ext_res))) + if (!(rGroup->recordType = res_get (zh->res, ext_res))) { sprintf (ext_res, "%srecordType", gprefix); - if (!(rGroup->recordType = res_get (common_resource, ext_res))) - { - logf (LOG_LOG, "? %s", fname); - return 0; - } + rGroup->recordType = res_get (zh->res, ext_res); } } if (!rGroup->recordType) { - 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 (zh->reg->recTypes, rGroup->recordType, subType, + &clientData))) { logf (LOG_WARN, "No such record type: %s", rGroup->recordType); return 0; @@ -944,30 +746,44 @@ int fileExtract (SYSNO *sysno, const char *fname, if (!rGroup->recordId) { sprintf (ext_res, "%srecordId.%s", gprefix, ext); - rGroup->recordId = res_get (common_resource, ext_res); + rGroup->recordId = res_get (zh->res, ext_res); } /* determine database name */ if (!rGroup->databaseName) { sprintf (ext_res, "%sdatabase.%s", gprefix, ext); - if (!(rGroup->databaseName = res_get (common_resource, ext_res))) + if (!(rGroup->databaseName = res_get (zh->res, ext_res))) { sprintf (ext_res, "%sdatabase", gprefix); - rGroup->databaseName = res_get (common_resource, ext_res); + rGroup->databaseName = res_get (zh->res, ext_res); } } if (!rGroup->databaseName) rGroup->databaseName = "Default"; + /* determine if explain database */ + + sprintf (ext_res, "%sexplainDatabase", gprefix); + rGroup->explainDatabase = + atoi (res_get_def (zh->res, ext_res, "0")); + + /* announce database */ + if (zebraExplain_curDatabase (zh->reg->zei, rGroup->databaseName)) + { + if (zebraExplain_newDatabase (zh->reg->zei, rGroup->databaseName, + rGroup->explainDatabase)) + return 0; + } + if (rGroup->flagStoreData == -1) { const char *sval; sprintf (ext_res, "%sstoreData.%s", gprefix, ext); - if (!(sval = res_get (common_resource, ext_res))) + if (!(sval = res_get (zh->res, ext_res))) { sprintf (ext_res, "%sstoreData", gprefix); - sval = res_get (common_resource, ext_res); + sval = res_get (zh->res, ext_res); } if (sval) rGroup->flagStoreData = atoi (sval); @@ -980,11 +796,14 @@ int fileExtract (SYSNO *sysno, const char *fname, const char *sval; sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext); - if (!(sval = res_get (common_resource, ext_res))) + sval = res_get (zh->res, ext_res); + if (!sval) { sprintf (ext_res, "%sstoreKeys", gprefix); - sval = res_get (common_resource, ext_res); + sval = res_get (zh->res, ext_res); } + if (!sval) + sval = res_get (zh->res, "storeKeys"); if (sval) rGroup->flagStoreKeys = atoi (sval); } @@ -995,17 +814,834 @@ int fileExtract (SYSNO *sysno, const char *fname, 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; } } - file_read_start (fd); - recordExtract (sysno, fname, rGroup, deleteFlag, fd, recType, subType); - file_read_stop (fd); + fi = file_read_start (fd); + do + { + file_begin (fi); + r = recordExtract (zh, sysno, fname, rGroup, deleteFlag, fi, + recType, subType, clientData); + } while (r && !sysno && fi->file_more); + file_read_stop (fi); if (fd != -1) close (fd); - return 1; + return r; +} + + +int extract_rec_in_mem (ZebraHandle zh, const char *recordType, + const char *buf, size_t buf_size, + const char *databaseName, int delete_flag, + int test_mode, int *sysno, + int store_keys, int store_data, + const char *match_criteria) +{ + RecordAttr *recordAttr; + struct recExtractCtrl extractCtrl; + int i, r; + char *matchStr = 0; + RecType recType; + char subType[1024]; + void *clientData; + const char *fname = ""; + Record rec; + long recordOffset = 0; + struct zebra_fetch_control fc; + + fc.fd = -1; + fc.record_int_buf = buf; + fc.record_int_len = buf_size; + fc.record_int_pos = 0; + fc.offset_end = 0; + fc.record_offset = 0; + + extractCtrl.offset = 0; + extractCtrl.readf = zebra_record_int_read; + extractCtrl.seekf = zebra_record_int_seek; + extractCtrl.tellf = zebra_record_int_tell; + extractCtrl.endf = zebra_record_int_end; + extractCtrl.fh = &fc; + + /* announce database */ + if (zebraExplain_curDatabase (zh->reg->zei, databaseName)) + { + if (zebraExplain_newDatabase (zh->reg->zei, databaseName, 0)) + return 0; + } + if (!(recType = + recType_byName (zh->reg->recTypes, recordType, subType, + &clientData))) + { + logf (LOG_WARN, "No such record type: %s", recordType); + return 0; + } + + zh->reg->keys.buf_used = 0; + zh->reg->keys.prevAttrUse = -1; + zh->reg->keys.prevAttrSet = -1; + zh->reg->keys.prevSeqNo = 0; + zh->reg->sortKeys = 0; + + extractCtrl.subType = subType; + extractCtrl.init = extract_init; + extractCtrl.tokenAdd = extract_token_add; + extractCtrl.schemaAdd = extract_schema_add; + extractCtrl.dh = zh->reg->dh; + extractCtrl.handle = zh; + extractCtrl.zebra_maps = zh->reg->zebra_maps; + extractCtrl.flagShowRecords = 0; + for (i = 0; i<256; i++) + { + if (zebra_maps_is_positioned(zh->reg->zebra_maps, i)) + extractCtrl.seqno[i] = 1; + else + extractCtrl.seqno[i] = 0; + } + + r = (*recType->extract)(clientData, &extractCtrl); + + if (r == RECCTRL_EXTRACT_EOF) + return 0; + else if (r == RECCTRL_EXTRACT_ERROR) + { + /* error occured during extraction ... */ +#if 1 + yaz_log (LOG_WARN, "extract error"); +#else + if (rGroup->flagRw && + records_processed < rGroup->fileVerboseLimit) + { + logf (LOG_WARN, "fail %s %s %ld", rGroup->recordType, + fname, (long) recordOffset); + } +#endif + return 0; + } + if (zh->reg->keys.buf_used == 0) + { + /* the extraction process returned no information - the record + is probably empty - unless flagShowRecords is in use */ + if (test_mode) + return 1; + logf (LOG_WARN, "No keys generated for record"); + logf (LOG_WARN, " The file is probably empty"); + return 1; + } + /* match criteria */ + + if (! *sysno) + { + /* new record */ + if (delete_flag) + { + logf (LOG_LOG, "delete %s %s %ld", recordType, + fname, (long) recordOffset); + logf (LOG_WARN, "cannot delete record above (seems new)"); + return 1; + } + logf (LOG_LOG, "add %s %s %ld", recordType, fname, + (long) recordOffset); + rec = rec_new (zh->reg->records); + + *sysno = rec->sysno; + + recordAttr = rec_init_attr (zh->reg->zei, rec); + + if (matchStr) + { + dict_insert (zh->reg->matchDict, matchStr, + sizeof(*sysno), sysno); + } + extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); + extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); + } + else + { + /* record already exists */ + struct recKeys delkeys; + + rec = rec_get (zh->reg->records, *sysno); + assert (rec); + + recordAttr = rec_init_attr (zh->reg->zei, rec); + + if (recordAttr->runNumber == + zebraExplain_runNumberIncrement (zh->reg->zei, 0)) + { + logf (LOG_LOG, "skipped %s %s %ld", recordType, + fname, (long) recordOffset); + rec_rm (&rec); + return 1; + } + delkeys.buf_used = rec->size[recInfo_delKeys]; + delkeys.buf = rec->info[recInfo_delKeys]; + extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys); + extract_flushRecordKeys (zh, *sysno, 0, &delkeys); + if (delete_flag) + { + /* record going to be deleted */ + if (!delkeys.buf_used) + { + logf (LOG_LOG, "delete %s %s %ld", recordType, + fname, (long) recordOffset); + logf (LOG_WARN, "cannot delete file above, storeKeys false"); + } + else + { + logf (LOG_LOG, "delete %s %s %ld", recordType, + fname, (long) recordOffset); +#if 0 + if (matchStr) + dict_delete (matchDict, matchStr); +#endif + rec_del (zh->reg->records, &rec); + } + rec_rm (&rec); + return 1; + } + else + { + /* record going to be updated */ + if (!delkeys.buf_used) + { + logf (LOG_LOG, "update %s %s %ld", recordType, + fname, (long) recordOffset); + logf (LOG_WARN, "cannot update file above, storeKeys false"); + } + else + { + logf (LOG_LOG, "update %s %s %ld", recordType, + fname, (long) recordOffset); + extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); + } + } + } + /* update file type */ + xfree (rec->info[recInfo_fileType]); + rec->info[recInfo_fileType] = + rec_strdup (recordType, &rec->size[recInfo_fileType]); + + /* update filename */ + xfree (rec->info[recInfo_filename]); + rec->info[recInfo_filename] = + rec_strdup (fname, &rec->size[recInfo_filename]); + + /* update delete keys */ + xfree (rec->info[recInfo_delKeys]); + if (zh->reg->keys.buf_used > 0 && store_keys == 1) + { + rec->size[recInfo_delKeys] = zh->reg->keys.buf_used; + rec->info[recInfo_delKeys] = zh->reg->keys.buf; + zh->reg->keys.buf = NULL; + zh->reg->keys.buf_max = 0; + } + else + { + rec->info[recInfo_delKeys] = NULL; + rec->size[recInfo_delKeys] = 0; + } + + /* save file size of original record */ + zebraExplain_recordBytesIncrement (zh->reg->zei, + - recordAttr->recordSize); +#if 0 + recordAttr->recordSize = fi->file_moffset - recordOffset; + if (!recordAttr->recordSize) + recordAttr->recordSize = fi->file_max - recordOffset; +#else + recordAttr->recordSize = buf_size; +#endif + zebraExplain_recordBytesIncrement (zh->reg->zei, + recordAttr->recordSize); + + /* set run-number for this record */ + recordAttr->runNumber = + zebraExplain_runNumberIncrement (zh->reg->zei, 0); + + /* update store data */ + xfree (rec->info[recInfo_storeData]); + if (store_data == 1) + { + rec->size[recInfo_storeData] = recordAttr->recordSize; + rec->info[recInfo_storeData] = (char *) + xmalloc (recordAttr->recordSize); +#if 1 + memcpy (rec->info[recInfo_storeData], buf, recordAttr->recordSize); +#else + if (lseek (fi->fd, recordOffset, SEEK_SET) < 0) + { + logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s", + (long) recordOffset, fname); + exit (1); + } + if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize) + < recordAttr->recordSize) + { + logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", + recordAttr->recordSize, fname); + exit (1); + } +#endif + } + else + { + rec->info[recInfo_storeData] = NULL; + rec->size[recInfo_storeData] = 0; + } + /* update database name */ + xfree (rec->info[recInfo_databaseName]); + rec->info[recInfo_databaseName] = + rec_strdup (databaseName, &rec->size[recInfo_databaseName]); + + /* update offset */ + recordAttr->recordOffset = recordOffset; + + /* commit this record */ + rec_put (zh->reg->records, &rec); + + return 0; +} + +int explain_extract (void *handle, Record rec, data1_node *n) +{ + ZebraHandle zh = (ZebraHandle) handle; + struct recExtractCtrl extractCtrl; + int i; + + if (zebraExplain_curDatabase (zh->reg->zei, + rec->info[recInfo_databaseName])) + { + abort(); + if (zebraExplain_newDatabase (zh->reg->zei, + rec->info[recInfo_databaseName], 0)) + abort (); + } + + zh->reg->keys.buf_used = 0; + zh->reg->keys.prevAttrUse = -1; + zh->reg->keys.prevAttrSet = -1; + zh->reg->keys.prevSeqNo = 0; + zh->reg->sortKeys = 0; + + extractCtrl.init = extract_init; + extractCtrl.tokenAdd = extract_token_add; + extractCtrl.schemaAdd = extract_schema_add; + extractCtrl.dh = zh->reg->dh; + for (i = 0; i<256; i++) + extractCtrl.seqno[i] = 0; + extractCtrl.zebra_maps = zh->reg->zebra_maps; + extractCtrl.flagShowRecords = 0; + extractCtrl.handle = handle; + + grs_extract_tree(&extractCtrl, n); + + if (rec->size[recInfo_delKeys]) + { + struct recKeys delkeys; + struct sortKey *sortKeys = 0; + + delkeys.buf_used = rec->size[recInfo_delKeys]; + delkeys.buf = rec->info[recInfo_delKeys]; + extract_flushSortKeys (zh, rec->sysno, 0, &sortKeys); + extract_flushRecordKeys (zh, rec->sysno, 0, &delkeys); + } + extract_flushRecordKeys (zh, rec->sysno, 1, &zh->reg->keys); + extract_flushSortKeys (zh, rec->sysno, 1, &zh->reg->sortKeys); + + xfree (rec->info[recInfo_delKeys]); + rec->size[recInfo_delKeys] = zh->reg->keys.buf_used; + rec->info[recInfo_delKeys] = zh->reg->keys.buf; + zh->reg->keys.buf = NULL; + zh->reg->keys.buf_max = 0; + return 0; +} + +void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno, + int cmd, struct recKeys *reckeys) +{ +#if SU_SCHEME +#else + unsigned char attrSet = (unsigned char) -1; + unsigned short attrUse = (unsigned short) -1; +#endif + int seqno = 0; + int off = 0; + int ch = 0; + ZebraExplainInfo zei = zh->reg->zei; + + if (!zh->reg->key_buf) + { + int mem = 8*1024*1024; + zh->reg->key_buf = (char**) xmalloc (mem); + zh->reg->ptr_top = mem/sizeof(char*); + zh->reg->ptr_i = 0; + zh->reg->key_buf_used = 0; + zh->reg->key_file_no = 0; + } + zebraExplain_recordCountIncrement (zei, cmd ? 1 : -1); + while (off < reckeys->buf_used) + { + const char *src = reckeys->buf + off; + struct it_key key; + int lead; + + lead = *src++; + +#if SU_SCHEME + if ((lead & 3) < 3) + { + memcpy (&ch, src, sizeof(ch)); + src += sizeof(ch); + } +#else + if (!(lead & 1)) + { + memcpy (&attrSet, src, sizeof(attrSet)); + src += sizeof(attrSet); + } + if (!(lead & 2)) + { + memcpy (&attrUse, src, sizeof(attrUse)); + src += sizeof(attrUse); + } +#endif + if (zh->reg->key_buf_used + 1024 > + (zh->reg->ptr_top -zh->reg->ptr_i)*sizeof(char*)) + extract_flushWriteKeys (zh); + ++(zh->reg->ptr_i); + (zh->reg->key_buf)[zh->reg->ptr_top - zh->reg->ptr_i] = + (char*)zh->reg->key_buf + zh->reg->key_buf_used; +#if SU_SCHEME +#else + ch = zebraExplain_lookupSU (zei, attrSet, attrUse); + if (ch < 0) + ch = zebraExplain_addSU (zei, attrSet, attrUse); +#endif + assert (ch > 0); + zh->reg->key_buf_used += + key_SU_encode (ch,((char*)zh->reg->key_buf) + + zh->reg->key_buf_used); + + while (*src) + ((char*)zh->reg->key_buf) [(zh->reg->key_buf_used)++] = *src++; + src++; + ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = '\0'; + ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = cmd; + + 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*)zh->reg->key_buf + zh->reg->key_buf_used, &key, sizeof(key)); + (zh->reg->key_buf_used) += sizeof(key); + off = src - reckeys->buf; + } + assert (off == reckeys->buf_used); +} + +void extract_flushWriteKeys (ZebraHandle zh) +{ + FILE *outf; + char out_fname[200]; + char *prevcp, *cp; + struct encode_info encode_info; + int ptr_i = zh->reg->ptr_i; +#if SORT_EXTRA + int i; +#endif + if (!zh->reg->key_buf || ptr_i <= 0) + return; + + (zh->reg->key_file_no)++; + logf (LOG_LOG, "sorting section %d", (zh->reg->key_file_no)); +#if !SORT_EXTRA + qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i, sizeof(char*), + key_qsort_compare); + extract_get_fname_tmp (zh, out_fname, zh->reg->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", zh->reg->key_file_no); + prevcp = cp = (zh->reg->key_buf)[zh->reg->ptr_top - ptr_i]; + + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + + while (--ptr_i > 0) + { + cp = (zh->reg->key_buf)[zh->reg->ptr_top - ptr_i]; + if (strcmp (cp, prevcp)) + { + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + prevcp = cp; + } + 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); + extract_get_fname_tmp (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); + exit (1); + } + logf (LOG_LOG, "finished section %d", zh->reg->key_file_no); + zh->reg->ptr_i = 0; + zh->reg->key_buf_used = 0; +} + +void extract_add_index_string (RecWord *p, const char *string, + int length) +{ + char *dst; + unsigned char attrSet; + unsigned short attrUse; + int lead = 0; + int diff = 0; + int *pseqno = &p->seqnos[p->reg_type]; + ZebraHandle zh = p->extractCtrl->handle; + ZebraExplainInfo zei = zh->reg->zei; + struct recKeys *keys = &zh->reg->keys; + + if (keys->buf_used+1024 > keys->buf_max) + { + char *b; + + b = (char *) xmalloc (keys->buf_max += 128000); + if (keys->buf_used > 0) + memcpy (b, keys->buf, keys->buf_used); + xfree (keys->buf); + keys->buf = b; + } + dst = keys->buf + keys->buf_used; + + attrSet = p->attrSet; + if (keys->buf_used > 0 && keys->prevAttrSet == attrSet) + lead |= 1; + else + keys->prevAttrSet = attrSet; + attrUse = p->attrUse; + if (keys->buf_used > 0 && keys->prevAttrUse == attrUse) + lead |= 2; + else + keys->prevAttrUse = attrUse; +#if 1 + diff = 1 + *pseqno - keys->prevSeqNo; + if (diff >= 1 && diff <= 15) + lead |= (diff << 2); + else + diff = 0; +#endif + keys->prevSeqNo = *pseqno; + + *dst++ = lead; + +#if SU_SCHEME + if ((lead & 3) < 3) + { + int ch = zebraExplain_lookupSU (zei, attrSet, attrUse); + if (ch < 0) + { + ch = zebraExplain_addSU (zei, attrSet, attrUse); + yaz_log (LOG_DEBUG, "addSU set=%d use=%d SU=%d", + attrSet, attrUse, ch); + } + assert (ch > 0); + memcpy (dst, &ch, sizeof(ch)); + dst += sizeof(ch); + } +#else + if (!(lead & 1)) + { + memcpy (dst, &attrSet, sizeof(attrSet)); + dst += sizeof(attrSet); + } + if (!(lead & 2)) + { + memcpy (dst, &attrUse, sizeof(attrUse)); + dst += sizeof(attrUse); + } +#endif + *dst++ = p->reg_type; + memcpy (dst, string, length); + dst += length; + *dst++ = '\0'; + + if (!diff) + { + memcpy (dst, pseqno, sizeof(*pseqno)); + dst += sizeof(*pseqno); + } + keys->buf_used = dst - keys->buf; + if (*pseqno) + (*pseqno)++; +} + +static void extract_add_sort_string (RecWord *p, const char *string, + int length) +{ + struct sortKey *sk; + ZebraHandle zh = p->extractCtrl->handle; + + for (sk = zh->reg->sortKeys; sk; sk = sk->next) + if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse) + return; + + sk = (struct sortKey *) xmalloc (sizeof(*sk)); + sk->next = zh->reg->sortKeys; + zh->reg->sortKeys = sk; + + sk->string = (char *) xmalloc (length); + sk->length = length; + memcpy (sk->string, string, length); + + sk->attrSet = p->attrSet; + sk->attrUse = p->attrUse; +} + +void extract_add_string (RecWord *p, const char *string, int length) +{ + assert (length > 0); + if (zebra_maps_is_sort (p->zebra_maps, p->reg_type)) + extract_add_sort_string (p, string, length); + else + extract_add_index_string (p, string, length); +} + +static void extract_add_incomplete_field (RecWord *p) +{ + const char *b = p->string; + int remain = p->length; + const char **map = 0; + + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + + while (map) + { + 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; + extract_add_string (p, buf, i); + } + (p->seqnos[p->reg_type])++; /* to separate this from next one */ +} + +static void extract_add_complete_field (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; + extract_add_string (p, buf, i); +} + +void extract_token_add (RecWord *p) +{ + WRBUF wrbuf; + if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0, + p->string, p->length))) + { + p->string = wrbuf_buf(wrbuf); + p->length = wrbuf_len(wrbuf); + } + if (zebra_maps_is_complete (p->zebra_maps, p->reg_type)) + extract_add_complete_field (p); + else + extract_add_incomplete_field(p); +} + +void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid) +{ + ZebraHandle zh = (ZebraHandle) (p->handle); + zebraExplain_addSchema (zh->reg->zei, oid); +} + +void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno, + int cmd, struct sortKey **skp) +{ + struct sortKey *sk = *skp; + SortIdx sortIdx = zh->reg->sortIdx; + + 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; + } + *skp = 0; +} + +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) +{ + if (d <= 63) + *bp++ = d; + else if (d <= 16383) + { + *bp++ = 64 + (d>>8); + *bp++ = d & 255; + } + else if (d <= 4194303) + { + *bp++ = 128 + (d>>16); + *bp++ = (d>>8) & 255; + *bp++ = d & 255; + } + else + { + *bp++ = 192 + (d>>24); + *bp++ = (d>>16) & 255; + *bp++ = (d>>8) & 255; + *bp++ = d & 255; + } + return bp; +} + +void encode_key_write (char *k, struct encode_info *i, FILE *outf) +{ + struct it_key key; + char *bp = i->buf; + + while ((*bp++ = *k++)) + ; + memcpy (&key, k+1, sizeof(struct it_key)); + bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp); + if (i->sysno != key.sysno) + { + 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"); + exit (1); + } }