X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=619fdfdf267a3243bc6c404837ee7cecb79eb65a;hp=ae8a6c44d2d555d4f4f1b19962f92daa0012fb46;hb=519fefb91135ad52134b9fc4e82b3874f5525a2b;hpb=247f1ebfc107315f1c0b2b59bb59fe4774f3b0d4 diff --git a/index/extract.c b/index/extract.c index ae8a6c4..619fdfd 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,810 +1,1257 @@ -/* - * Copyright (C) 1994-1998, Index Data I/S - * All rights reserved. - * Sebastian Hammer, Adam Dickmeiss - * - * $Log: extract.c,v $ - * 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 - * Cleanup of various buffer size entities. - * - * Revision 1.58 1996/05/14 06:16:38 adam - * Compact use/set bytes used in search service. - * - * Revision 1.57 1996/05/13 14:23:04 adam - * Work on compaction of set/use bytes in dictionary. - * - * Revision 1.56 1996/05/09 09:54:42 adam - * Server supports maps from one logical attributes to a list of physical - * attributes. - * The extraction process doesn't make space consuming 'any' keys. - * - * Revision 1.55 1996/05/09 07:28:55 quinn - * Work towards phrases and multiple registers - * - * Revision 1.54 1996/05/01 13:46:35 adam - * First work on multiple records in one file. - * New option, -offset, to the "unread" command in the filter module. - * - * Revision 1.53 1996/04/26 12:09:43 adam - * Added a few comments. - * - * Revision 1.52 1996/04/25 13:27:57 adam - * Function recordExtract modified so that files with no keys (possibly empty) - * are ignored. - * - * Revision 1.51 1996/03/19 11:08:42 adam - * Bug fix: Log preamble wasn't always turned off after recordExtract. - * - * Revision 1.50 1996/02/12 18:45:36 adam - * New fileVerboseFlag in record group control. - * - * Revision 1.49 1996/02/05 12:29:57 adam - * Logging reduced a bit. - * The remaining running time is estimated during register merge. - * - * Revision 1.48 1996/02/01 20:53:26 adam - * The temporary per-record keys are compacted a little, and duplication - * of the per-records keys are avoided when they are saved in the record - * information buffer. - * - * Revision 1.47 1996/01/17 14:57:48 adam - * Prototype changed for reader functions in extract/retrieve. File - * is identified by 'void *' instead of 'int. - * - * Revision 1.46 1995/12/15 14:57:16 adam - * Bug fix. - * - * Revision 1.45 1995/12/15 12:37:41 adam - * In addRecordKeyAny: Writes key only when attrSet != -1. - * - * Revision 1.44 1995/12/12 16:00:54 adam - * System call sync(2) used after update/commit. - * Locking (based on fcntl) uses F_EXLCK and F_SHLCK instead of F_WRLCK - * and F_RDLCK. - * - * Revision 1.43 1995/12/11 09:12:46 adam - * The rec_get function returns NULL if record doesn't exist - will - * happen in the server if the result set records have been deleted since - * the creation of the set (i.e. the search). - * The server saves a result temporarily if it is 'volatile', i.e. the - * set is register dependent. - * - * 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 1.126 2002-10-22 12:51:08 adam Exp $ + Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002 + Index Data Aps + +This file is part of the Zebra server. + +Zebra is free software; you can redistribute it and/or modify it under +the terms of the GNU General Public License as published by the Free +Software Foundation; either version 2, or (at your option) any later +version. + +Zebra is distributed in the hope that it will be useful, but WITHOUT ANY +WARRANTY; without even the implied warranty of MERCHANTABILITY or +FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License +for more details. + +You should have received a copy of the GNU General Public License +along with Zebra; see the file LICENSE.zebra. If not, write to the +Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA +02111-1307, USA. +*/ + + #include #include -#ifdef WINDOWS +#ifdef WIN32 #include #else #include #endif #include -#include -#include -#include #include "index.h" +#include +#include -#include "zinfo.h" - -static Dict matchDict; - -static Records records = NULL; -static SortIdx sortIdx = 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; +#define USE_SHELLSORT 0 -static ZebTargetInfo *zti = NULL; +#if USE_SHELLSORT +static void shellsort(void *ar, int r, size_t s, + int (*cmp)(const void *a, const void *b)) +{ + char *a = ar; + char v[100]; + int h, i, j, k; + static const int incs[16] = { 1391376, 463792, 198768, 86961, 33936, + 13776, 4592, 1968, 861, 336, + 112, 48, 21, 7, 3, 1 }; + for ( k = 0; k < 16; k++) + for (h = incs[k], i = h; i < r; i++) + { + memcpy (v, a+s*i, s); + j = i; + while (j > h && (*cmp)(a + s*(j-h), v) > 0) + { + memcpy (a + s*j, a + s*(j-h), s); + j -= h; + } + memcpy (a+s*j, v, s); + } +} +#endif -static void logRecord (int showFlag) +static void logRecord (ZebraHandle zh) { - if (!showFlag) - ++records_processed; - if (showFlag || !(records_processed % 1000)) + ++zh->records_processed; + if (!(zh->records_processed % 1000)) { logf (LOG_LOG, "Records: %7d i/u/d %d/%d/%d", - records_processed, records_inserted, records_updated, - records_deleted); + zh->records_processed, zh->records_inserted, zh->records_updated, + zh->records_deleted); } } -int key_open (BFiles bfs, int mem, int rw) +static void extract_init (struct recExtractCtrl *p, RecWord *w) +{ + w->zebra_maps = p->zebra_maps; + w->seqno = 1; + w->attrSet = VAL_BIB1; + w->attrUse = 1016; + w->reg_type = 'w'; + w->extractCtrl = p; +} + +static const char **searchRecordKey (ZebraHandle zh, + struct recKeys *reckeys, + int attrSetS, int attrUseS) { - if (!mem) - mem = atoi(res_get_def (common_resource, "memMax", "4"))*1024*1024; - if (mem < 50000) - mem = 50000; - key_buf = xmalloc (mem); - ptr_top = mem/sizeof(char*); - ptr_i = 0; + 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 - key_buf_used = 0; - key_file_no = 0; + for (i = 0; i<32; i++) + ws[i] = NULL; - if (!(matchDict = dict_open (bfs, GMATCH_DICT, 50, rw))) - { - logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT); - return -1; - } - assert (!records); - records = rec_open (bfs, rw); - if (!records) - { - dict_close (matchDict); - return -1; - } - zti = zebTargetInfo_open (records, rw); - if (!zti) +#if SU_SCHEME + chS = zebraExplain_lookupSU (zh->reg->zei, attrSetS, attrUseS); + if (chS < 0) + return ws; +#endif + while (off < reckeys->buf_used) { - rec_close (&records); - dict_close (matchDict); - return -1; + + 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++) + ; + 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 + ) + { + int woff; + + + if (startSeq == -1) + startSeq = seqno; + woff = seqno - startSeq; + if (woff >= 0 && woff < 31) + ws[woff] = wstart; + } + + off = src - reckeys->buf; } - sortIdx = sortIdx_open (bfs, 1); - return 0; + assert (off == reckeys->buf_used); + return ws; } -struct encode_info { - int sysno; - int seqno; - char buf[768]; +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; }; -void encode_key_init (struct encode_info *i) +static struct file_read_info *file_read_start (int fd) { - i->sysno = 0; - i->seqno = 0; -} + struct file_read_info *fi = (struct file_read_info *) + xmalloc (sizeof(*fi)); -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; + fi->fd = fd; + fi->file_max = 0; + fi->file_moffset = 0; + fi->sdrbuf = 0; + fi->sdrmax = 0; + return fi; } -void encode_key_write (char *k, struct encode_info *i, FILE *outf) +static void file_read_stop (struct file_read_info *fi) { - 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); - } + xfree (fi); } -#define SORT_EXTRA 0 - -#if SORT_EXTRA -static int key_y_len; - -static int key_y_compare (const void *p1, const void *p2) +static off_t file_seek (void *handle, off_t offset) { - 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); + 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 key_x_compare (const void *p1, const void *p2) +static off_t file_tell (void *handle) { - return strcmp (*(char**) p1, *(char**) p2); + struct file_read_info *p = (struct file_read_info *) handle; + return p->file_offset; } -#endif -void key_flush (void) +static int file_read (void *handle, char *buf, size_t count) { - 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); - 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); - 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); - } -#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"))) + struct file_read_info *p = (struct file_read_info *) handle; + int fd = p->fd; + int r; + if (p->sdrbuf) { - logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); - exit (1); + 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); } - 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)) + else + r = read (fd, buf, count); + if (r > 0) { - logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname); - exit (1); + p->file_offset += r; + if (p->file_offset > p->file_max) + p->file_max = p->file_offset; } - logf (LOG_LOG, "finished section %d", key_file_no); - ptr_i = 0; - key_buf_used = 0; + return r; } -int key_close () +static void file_begin (void *handle) { - key_flush (); - xfree (key_buf); -#if 1 - zebTargetInfo_close (zti, 1); -#endif - rec_close (&records); - dict_close (matchDict); - sortIdx_close (sortIdx); + struct file_read_info *p = (struct file_read_info *) handle; - logRecord (1); - return key_file_no; + 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 wordInit (struct recExtractCtrl *p, RecWord *w) +static void file_end (void *handle, off_t offset) { - w->zebra_maps = p->zebra_maps; - w->attrSet = 1; - w->attrUse = 1016; - w->reg_type = 'w'; + struct file_read_info *p = (struct file_read_info *) handle; + + assert (p->file_more == 0); + p->file_more = 1; + p->file_moffset = offset; } -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 addIndexString (RecWord *p, const char *string, int length) +static char *fileMatchStr (ZebraHandle zh, + struct recKeys *reckeys, struct recordGroup *rGroup, + const char *fname, const char *spec) { - char *dst; - char attrSet; - short attrUse; - int lead = 0; - int diff = 0; + static char dstBuf[2048]; /* static here ??? */ + char *dst = dstBuf; + const char *s = spec; + static const char **w; - if (reckeys.buf_used+1024 > reckeys.buf_max) + while (1) + { + while (*s == ' ' || *s == '\t') + s++; + if (!*s) + break; + if (*s == '(') + { + char attset_str[64], attname_str[64]; + data1_attset *attset; + int i; + char matchFlag[32]; + int attSet = 1, attUse = 1; + int first = 1; + + s++; + 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 == ')') + { + for (i = 0; i<32; i++) + matchFlag[i] = 1; + } + else + { + logf (LOG_WARN, "Missing ) in match criteria %s in group %s", + spec, rGroup->groupName ? rGroup->groupName : "none"); + return NULL; + } + s++; + + for (i = 0; i<32; i++) + if (matchFlag[i] && w[i]) + { + if (first) + { + *dst++ = ' '; + first = 0; + } + strcpy (dst, w[i]); + dst += strlen(w[i]); + } + if (first) + { + logf (LOG_WARN, "Record didn't contain match" + " fields in (%s,%s)", attset_str, attname_str); + return NULL; + } + } + else if (*s == '$') + { + int spec_len; + char special[64]; + const char *spec_src = NULL; + const char *s1 = ++s; + while (*s1 && *s1 != ' ' && *s1 != '\t') + s1++; + + spec_len = s1 - s; + if (spec_len > 63) + spec_len = 63; + memcpy (special, s, spec_len); + special[spec_len] = '\0'; + s = s1; + + if (!strcmp (special, "group")) + spec_src = rGroup->groupName; + else if (!strcmp (special, "database")) + spec_src = rGroup->databaseName; + else if (!strcmp (special, "filename")) + spec_src = fname; + else if (!strcmp (special, "type")) + spec_src = rGroup->recordType; + else + spec_src = NULL; + if (spec_src) + { + strcpy (dst, spec_src); + dst += strlen (spec_src); + } + } + else if (*s == '\"' || *s == '\'') + { + int stopMarker = *s++; + char tmpString[64]; + int i = 0; + + while (*s && *s != stopMarker) + { + if (i < 63) + tmpString[i++] = *s++; + } + if (*s) + s++; + tmpString[i] = '\0'; + strcpy (dst, tmpString); + dst += strlen (tmpString); + } + else + { + logf (LOG_WARN, "Syntax error in match criteria %s in group %s", + spec, rGroup->groupName ? rGroup->groupName : "none"); + return NULL; + } + *dst++ = 1; + } + if (dst == dstBuf) + { + logf (LOG_WARN, "No match criteria for record %s in group %s", + fname, rGroup->groupName ? rGroup->groupName : "none"); + return NULL; + } + *dst = '\0'; + return dstBuf; +} + +struct recordLogInfo { + const char *fname; + int recordOffset; + struct recordGroup *rGroup; +}; + +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; + off_t recordOffset = 0; + + if (fi->fd != -1) + { + 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); + if (rGroup->flagRw) + { + char msg[512]; + sprintf (msg, "%s:" PRINTF_OFF_T , fname, recordOffset); + yaz_log_init_prefix2 (msg); + } + + r = (*recType->extract)(clientData, &extractCtrl); + + yaz_log_init_prefix2 (0); + if (r == RECCTRL_EXTRACT_EOF) + return 0; + else if (r == RECCTRL_EXTRACT_ERROR_GENERIC) + { + /* error occured during extraction ... */ + if (rGroup->flagRw && + zh->records_processed < rGroup->fileVerboseLimit) + { + logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + } + return 0; + } + else if (r == RECCTRL_EXTRACT_ERROR_NO_SUCH_FILTER) + { + /* error occured during extraction ... */ + if (rGroup->flagRw && + zh->records_processed < rGroup->fileVerboseLimit) + { + logf (LOG_WARN, "no filter for %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 */ + + matchStr = NULL; + if (!sysno) + { + sysnotmp = 0; + sysno = &sysnotmp; + if (rGroup->recordId && *rGroup->recordId) + { + char *rinfo; + + matchStr = fileMatchStr (zh, &zh->reg->keys, rGroup, fname, + rGroup->recordId); + if (matchStr) + { + rinfo = dict_lookup (zh->reg->matchDict, matchStr); + if (rinfo) + memcpy (sysno, rinfo+1, sizeof(*sysno)); + } + else + { + logf (LOG_WARN, "Bad match criteria"); + return 0; + } + } + } + + if (! *sysno) + { + /* new record */ + if (deleteFlag) + { + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + logf (LOG_WARN, "cannot delete record above (seems new)"); + return 1; + } + if (zh->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 (zh->reg->matchDict, matchStr, sizeof(*sysno), sysno); + } + extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); + extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); + + zh->records_inserted++; + } + 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)) + { + yaz_log (LOG_LOG, "run number = %d", recordAttr->runNumber); + yaz_log (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + extract_flushSortKeys (zh, *sysno, -1, &zh->reg->sortKeys); + rec_rm (&rec); + logRecord (zh); + 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 (deleteFlag) + { + /* record going to be deleted */ + if (!delkeys.buf_used) + { + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + logf (LOG_WARN, "cannot delete file above, storeKeys false"); + } + else + { + if (zh->records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + zh->records_deleted++; + if (matchStr) + dict_delete (zh->reg->matchDict, matchStr); + rec_del (zh->reg->records, &rec); + } + rec_rm (&rec); + logRecord (zh); + return 1; + } + else + { + /* record going to be updated */ + if (!delkeys.buf_used) + { + logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + logf (LOG_WARN, "cannot update file above, storeKeys false"); + } + else + { + if (zh->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); + zh->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 (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 + { + rec->info[recInfo_delKeys] = NULL; + 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] = recordAttr->recordSize; + rec->info[recInfo_storeData] = (char *) + xmalloc (recordAttr->recordSize); + if (lseek (fi->fd, recordOffset, SEEK_SET) < 0) + { + 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 + { + 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]); + + /* update offset */ + recordAttr->recordOffset = recordOffset; + + /* commit this record */ + rec_put (zh->reg->records, &rec); + logRecord (zh); + return 1; +} + +int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, + const struct recordGroup *rGroupP, int deleteFlag) +{ + int r, i, fd; + char gprefix[128]; + char ext[128]; + char ext_res[128]; + char subType[128]; + RecType recType; + struct recordGroup rGroupM; + struct recordGroup *rGroup = &rGroupM; + struct file_read_info *fi; + void *clientData; + + memcpy (rGroup, rGroupP, sizeof(*rGroupP)); + + if (!rGroup->groupName || !*rGroup->groupName) + *gprefix = '\0'; + else + sprintf (gprefix, "%s.", rGroup->groupName); + + logf (LOG_DEBUG, "fileExtract %s", fname); + + /* determine file extension */ + *ext = '\0'; + for (i = strlen(fname); --i >= 0; ) + if (fname[i] == '/') + break; + else if (fname[i] == '.') + { + strcpy (ext, fname+i+1); + break; + } + /* determine file type - depending on extension */ + if (!rGroup->recordType) + { + sprintf (ext_res, "%srecordType.%s", gprefix, ext); + if (!(rGroup->recordType = res_get (zh->res, ext_res))) + { + sprintf (ext_res, "%srecordType", gprefix); + rGroup->recordType = res_get (zh->res, ext_res); + } + } + if (!rGroup->recordType) + { + if (zh->records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "? %s", fname); + return 0; + } + 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; + } + + /* determine match criteria */ + if (!rGroup->recordId) + { + sprintf (ext_res, "%srecordId.%s", gprefix, ext); + 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 (zh->res, ext_res))) + { + sprintf (ext_res, "%sdatabase", gprefix); + 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 (zh->res, ext_res))) + { + sprintf (ext_res, "%sstoreData", gprefix); + sval = res_get (zh->res, ext_res); + } + if (sval) + rGroup->flagStoreData = atoi (sval); + } + if (rGroup->flagStoreData == -1) + rGroup->flagStoreData = 0; + + if (rGroup->flagStoreKeys == -1) { - char *b; + const char *sval; - b = xmalloc (reckeys.buf_max += 128000); - if (reckeys.buf_used > 0) - memcpy (b, reckeys.buf, reckeys.buf_used); - xfree (reckeys.buf); - reckeys.buf = b; + sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext); + sval = res_get (zh->res, ext_res); + if (!sval) + { + sprintf (ext_res, "%sstoreKeys", gprefix); + sval = res_get (zh->res, ext_res); + } + if (!sval) + sval = res_get (zh->res, "storeKeys"); + if (sval) + rGroup->flagStoreKeys = atoi (sval); } - dst = reckeys.buf + reckeys.buf_used; + if (rGroup->flagStoreKeys == -1) + rGroup->flagStoreKeys = 0; - attrSet = p->attrSet; - if (reckeys.buf_used > 0 && reckeys.prevAttrSet == attrSet) - lead |= 1; - else - reckeys.prevAttrSet = attrSet; - attrUse = p->attrUse; - if (reckeys.buf_used > 0 && reckeys.prevAttrUse == attrUse) - lead |= 2; - else - reckeys.prevAttrUse = attrUse; -#if 1 - diff = 1 + p->seqno - reckeys.prevSeqNo; - if (diff >= 1 && diff <= 15) - lead |= (diff << 2); + if (sysno && deleteFlag) + fd = -1; else - diff = 0; -#endif - reckeys.prevSeqNo = p->seqno; + { + char full_rep[1024]; - *dst++ = lead; + if (zh->path_reg && !yaz_is_abspath (fname)) + { + strcpy (full_rep, zh->path_reg); + strcat (full_rep, "/"); + strcat (full_rep, fname); + } + else + strcpy (full_rep, fname); + - if (!(lead & 1)) + if ((fd = open (full_rep, O_BINARY|O_RDONLY)) == -1) + { + logf (LOG_WARN|LOG_ERRNO, "open %s", full_rep); + return 0; + } + } + fi = file_read_start (fd); + do { - memcpy (dst, &attrSet, sizeof(attrSet)); - dst += sizeof(attrSet); + 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 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 (!(lead & 2)) + if (!(recType = + recType_byName (zh->reg->recTypes, recordType, subType, + &clientData))) { - memcpy (dst, &attrUse, sizeof(attrUse)); - dst += sizeof(attrUse); + logf (LOG_WARN, "No such record type: %s", recordType); + return 0; } - *dst++ = p->reg_type; - memcpy (dst, string, length); - dst += length; - *dst++ = '\0'; - if (!diff) + 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++) { - memcpy (dst, &p->seqno, sizeof(p->seqno)); - dst += sizeof(p->seqno); + if (zebra_maps_is_positioned(zh->reg->zebra_maps, i)) + extractCtrl.seqno[i] = 1; + else + extractCtrl.seqno[i] = 0; } - reckeys.buf_used = dst - reckeys.buf; - (p->seqno)++; -} - -static void addSortString (RecWord *p, const char *string, int length) -{ - 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; + r = (*recType->extract)(clientData, &extractCtrl); - sk->string = xmalloc (p->length); - sk->length = p->length; - memcpy (sk->string, p->string, p->length); - sk->attrSet = p->attrSet; - sk->attrUse = p->attrUse; -} + if (r == RECCTRL_EXTRACT_EOF) + return 0; + else if (r == RECCTRL_EXTRACT_ERROR_GENERIC) + { + /* error occured during extraction ... */ + yaz_log (LOG_WARN, "extract error: generic"); + return 0; + } + else if (r == RECCTRL_EXTRACT_ERROR_NO_SUCH_FILTER) + { + /* error occured during extraction ... */ + yaz_log (LOG_WARN, "extract error: no such filter"); + 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 */ -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); -} + 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); -static void addIncompleteField (RecWord *p) -{ - const char *b = p->string; - int remain = p->length; - const char **map = 0; + *sysno = rec->sysno; - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + recordAttr = rec_init_attr (zh->reg->zei, rec); - while (map) + 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 { - char buf[IT_MAX_WORD+1]; - int i, remain; + /* record already exists */ + struct recKeys delkeys; - /* Skip spaces */ - while (map && *map && **map == *CHR_SPACE) + 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)) { - remain = p->length - (b - p->string); - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); - else - map = 0; + logf (LOG_LOG, "skipped %s %s %ld", recordType, + fname, (long) recordOffset); + rec_rm (&rec); + return 1; } - if (!map) - break; - i = 0; - while (map && *map && **map != *CHR_SPACE) - { - const char *cp = *map; + 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]); - 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); + /* 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; } -} -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; + /* 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); - if (remain > 0) - map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain); + /* set run-number for this record */ + recordAttr->runNumber = + zebraExplain_runNumberIncrement (zh->reg->zei, 0); - while (remain > 0 && i < IT_MAX_WORD) + /* update store data */ + xfree (rec->info[recInfo_storeData]); + if (store_data == 1) { - 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; + 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]); - if (i && i < IT_MAX_WORD) - buf[i++] = *CHR_SPACE; - while (map && *map && **map != *CHR_SPACE) - { - const char *cp = *map; + /* update offset */ + recordAttr->recordOffset = recordOffset; + + /* commit this record */ + rec_put (zh->reg->records, &rec); - 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); + return 0; } -static void addRecordKey (RecWord *p) +int explain_extract (void *handle, Record rec, data1_node *n) { - if (zebra_maps_is_complete (p->zebra_maps, p->reg_type)) - addCompleteField (p); - else - addIncompleteField(p); -} + 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 (); + } -static void flushSortKeys (SYSNO sysno, int cmd) -{ - struct sortKey *sk = sortKeys; + 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); - sortIdx_sysno (sortIdx, sysno); - while (sk) + if (rec->size[recInfo_delKeys]) { - 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; + 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); } - sortKeys = NULL; + 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; } -static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, - const char *databaseName) +void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno, + int cmd, struct recKeys *reckeys) { - char attrSet = -1; - short attrUse = -1; +#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 (zebTargetInfo_curDatabase (zti, databaseName)) + if (!zh->reg->key_buf) { - if (zebTargetInfo_newDatabase (zti, databaseName)) - abort (); + int mem= 1024*1024* atoi( res_get_def( zh->res, "memmax", "8")); + if (mem <= 0) + { + logf(LOG_WARN, "Invalid memory setting, using default 8 MB"); + mem= 1024*1024*8; + } + /* FIXME: That "8" should be in a default settings include */ + /* not hard-coded here! -H */ + 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, ch; + 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)); @@ -815,21 +1262,29 @@ static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, 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; - - ch = zebTargetInfo_lookupSU (zti, attrSet, 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 = zebTargetInfo_addSU (zti, attrSet, attrUse); + ch = zebraExplain_addSU (zei, attrSet, attrUse); +#endif assert (ch > 0); - ((char*) key_buf) [key_buf_used++] = ch; + zh->reg->key_buf_used += + key_SU_encode (ch,((char*)zh->reg->key_buf) + + zh->reg->key_buf_used); + while (*src) - ((char*)key_buf) [key_buf_used++] = *src++; + ((char*)zh->reg->key_buf) [(zh->reg->key_buf_used)++] = *src++; src++; - ((char*)key_buf) [key_buf_used++] = '\0'; - ((char*) key_buf)[key_buf_used++] = cmd; + ((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; @@ -840,690 +1295,548 @@ static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, } key.seqno = seqno; key.sysno = sysno; - memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key)); - key_buf_used += sizeof(key); + 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); } -static const char **searchRecordKey (struct recKeys *reckeys, - int attrSetS, int attrUseS) +void extract_flushWriteKeys (ZebraHandle zh) { - static const char *ws[32]; - int off = 0; - int startSeq = -1; + 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; - short attrUse; - char attrSet; - int seqno = 0; +#endif + if (!zh->reg->key_buf || ptr_i <= 0) + return; - for (i = 0; i<32; i++) - ws[i] = NULL; - - while (off < reckeys->buf_used) - { + (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); - const char *src = reckeys->buf + off; - const char *wstart; - int lead; + 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]; - lead = *src++; - - if (!(lead & 1)) - { - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); - } - if (!(lead & 2)) + 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)) { - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); + encode_key_flush ( &encode_info, outf); + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + prevcp = cp; } - wstart = src; - while (*src++) - ; - if (lead & 60) - seqno += ((lead>>2) & 15)-1; else - { - memcpy (&seqno, src, sizeof(seqno)); - src += sizeof(seqno); - } - if (attrUseS == attrUse && attrSetS == attrSet) - { - int woff; - - - if (startSeq == -1) - startSeq = seqno; - woff = seqno - startSeq; - if (woff >= 0 && woff < 31) - ws[woff] = wstart; - } - - off = src - reckeys->buf; - } - assert (off == reckeys->buf_used); - return ws; -} - -struct file_read_info { - off_t file_max; - off_t file_offset; - off_t file_moffset; - int file_more; - int fd; -}; - -static struct file_read_info *file_read_start (int fd) -{ - struct file_read_info *fi = xmalloc (sizeof(*fi)); - - fi->fd = fd; - fi->file_max = 0; - fi->file_moffset = 0; - return fi; -} - -static void file_read_stop (struct file_read_info *fi) -{ - assert (fi); - xfree (fi); -} - -static off_t file_seek (void *handle, off_t offset) -{ - struct file_read_info *p = handle; - p->file_offset = 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 (r > 0) - { - p->file_offset += r; - if (p->file_offset > p->file_max) - p->file_max = p->file_offset; + encode_key_write (cp + strlen(cp), &encode_info, outf); } - return r; -} - -static void file_begin (void *handle) -{ - struct file_read_info *p = handle; - - p->file_offset = p->file_moffset; - if (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 = handle; - - assert (p->file_more == 0); - p->file_more = 1; - p->file_moffset = offset; -} + encode_key_flush ( &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); -static int atois (const char **s) -{ - int val = 0, c; - while ( (c=**s) >= '0' && c <= '9') + if (!(outf = fopen (out_fname, "wb"))) { - val = val*10 + c - '0'; - ++(*s); + logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); + exit (1); } - return val; -} - -static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, - const char *fname, - const char *spec) -{ - static char dstBuf[2048]; - char *dst = dstBuf; - const char *s = spec; - static const char **w; - int i; - + logf (LOG_LOG, "writing section %d", key_file_no); + i = ptr_i; + prevcp = key_buf[ptr_top-i]; while (1) - { - while (*s == ' ' || *s == '\t') - s++; - if (!*s) - break; - if (*s == '(') - { - char matchFlag[32]; - int attrSet, attrUse; - 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); - assert (w); - - if (*s == ')') - { - for (i = 0; i<32; i++) - matchFlag[i] = 1; - } - else - { - logf (LOG_WARN, "Missing ) in match criteria %s in group %s", - spec, rGroup->groupName ? rGroup->groupName : "none"); - return NULL; - } - s++; - - for (i = 0; i<32; i++) - if (matchFlag[i] && w[i]) - { - if (first) - { - *dst++ = ' '; - first = 0; - } - strcpy (dst, w[i]); - dst += strlen(w[i]); - } - if (first) - { - logf (LOG_WARN, "Record didn't contain match" - " fields in (%d,%d)", attrSet, attrUse); - return NULL; - } - } - else if (*s == '$') - { - int spec_len; - char special[64]; - const char *spec_src = NULL; - const char *s1 = ++s; - while (*s1 && *s1 != ' ' && *s1 != '\t') - s1++; - - spec_len = s1 - s; - if (spec_len > 63) - spec_len = 63; - memcpy (special, s, spec_len); - special[spec_len] = '\0'; - s = s1; - - if (!strcmp (special, "group")) - spec_src = rGroup->groupName; - else if (!strcmp (special, "database")) - spec_src = rGroup->databaseName; - else if (!strcmp (special, "filename")) - spec_src = fname; - else if (!strcmp (special, "type")) - spec_src = rGroup->recordType; - else - spec_src = NULL; - if (spec_src) - { - strcpy (dst, spec_src); - dst += strlen (spec_src); - } - } - else if (*s == '\"' || *s == '\'') + if (!--i || strcmp (prevcp, key_buf[ptr_top-i])) { - int stopMarker = *s++; - char tmpString[64]; - int i = 0; - - while (*s && *s != stopMarker) + 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) { - if (i < 63) - tmpString[i++] = *s++; + cp = key_buf[ptr_top-ptr_i]; + encode_key_write (cp+key_y_len, &encode_info, outf); } - if (*s) - s++; - tmpString[i] = '\0'; - strcpy (dst, tmpString); - dst += strlen (tmpString); + encode_key_flush ( &encode_info, outf); + if (!i) + break; + prevcp = key_buf[ptr_top-ptr_i]; } - else +#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->seqno; + 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) { - logf (LOG_WARN, "Syntax error in match criteria %s in group %s", - spec, rGroup->groupName ? rGroup->groupName : "none"); - return NULL; + ch = zebraExplain_addSU (zei, attrSet, attrUse); + yaz_log (LOG_DEBUG, "addSU set=%d use=%d SU=%d", + attrSet, attrUse, ch); } - *dst++ = 1; + assert (ch > 0); + memcpy (dst, &ch, sizeof(ch)); + dst += sizeof(ch); } - if (dst == dstBuf) +#else + if (!(lead & 1)) { - logf (LOG_WARN, "No match criteria for record %s in group %s", - fname, rGroup->groupName ? rGroup->groupName : "none"); - return NULL; + memcpy (dst, &attrSet, sizeof(attrSet)); + dst += sizeof(attrSet); } - *dst = '\0'; - return dstBuf; + 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; } -struct recordLogInfo { - const char *fname; - int recordOffset; - struct recordGroup *rGroup; -}; - -static void recordLogPreamble (int level, const char *msg, void *info) +static void extract_add_sort_string (RecWord *p, const char *string, + int length) { - struct recordLogInfo *p = info; - FILE *outf = 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); + 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; } -static int recordExtract (SYSNO *sysno, const char *fname, - struct recordGroup *rGroup, int deleteFlag, - struct file_read_info *fi, RecType recType, - char *subType) +void extract_add_string (RecWord *p, const char *string, int length) { - struct recExtractCtrl extractCtrl; - int r; - char *matchStr; - SYSNO sysnotmp; - off_t recordOffset = 0; - Record rec; - struct recordLogInfo logInfo; + 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); +} - if (fi->fd != -1) - { - /* we are going to read from a file, so prepare the extraction */ - extractCtrl.fh = fi; - extractCtrl.subType = subType; - extractCtrl.init = wordInit; - extractCtrl.add = addRecordKey; - extractCtrl.dh = rGroup->dh; - - reckeys.buf_used = 0; - reckeys.prevAttrUse = -1; - reckeys.prevAttrSet = -1; - reckeys.prevSeqNo = 0; - - recordOffset = fi->file_moffset; - extractCtrl.offset = recordOffset; - extractCtrl.readf = file_read; - extractCtrl.seekf = file_seek; - extractCtrl.tellf = file_tell; - extractCtrl.endf = file_end; - extractCtrl.zebra_maps = rGroup->zebra_maps; - extractCtrl.flagShowRecords = !rGroup->flagRw; - if (!rGroup->flagRw) - printf ("File: %s %ld\n", fname, (long) recordOffset); +static void extract_add_incomplete_field (RecWord *p) +{ + const char *b = p->string; + int remain = p->length; + const char **map = 0; - logInfo.fname = fname; - logInfo.recordOffset = recordOffset; - logInfo.rGroup = rGroup; - log_event_start (recordLogPreamble, &logInfo); + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); - r = (*recType->extract)(&extractCtrl); + while (map) + { + char buf[IT_MAX_WORD+1]; + int i, remain; - log_event_start (NULL, NULL); + /* 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; - if (r) - { - /* error occured during extraction ... */ - 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 - 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 0; - } + 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->seqno++; } +} - /* perform match if sysno not known and if match criteria is specified */ - - matchStr = NULL; - if (!sysno) +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) { - sysnotmp = 0; - sysno = &sysnotmp; - if (rGroup->recordId && *rGroup->recordId) - { - char *rinfo; - - matchStr = fileMatchStr (&reckeys, rGroup, fname, - rGroup->recordId); - if (matchStr) - { - rinfo = dict_lookup (matchDict, matchStr); - if (rinfo) - memcpy (sysno, rinfo+1, sizeof(*sysno)); - } - else - { - logf (LOG_WARN, "Bad match criteria"); - return 0; - } - } + 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); +} - if (! *sysno) +void extract_token_add (RecWord *p) +{ + WRBUF wrbuf; + if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0, + p->string, p->length))) { - /* new record */ - if (deleteFlag) - { - logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); - logf (LOG_WARN, "cannot delete record above (seems new)"); - return 1; - } - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "add %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); - rec = rec_new (records); - *sysno = rec->sysno; + 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); +} - if (matchStr) - { - dict_insert (matchDict, matchStr, sizeof(*sysno), sysno); - } - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); - flushSortKeys (*sysno, 1); +void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid) +{ + ZebraHandle zh = (ZebraHandle) (p->handle); + zebraExplain_addSchema (zh->reg->zei, oid); +} - records_inserted++; - } - else - { - /* record already exists */ - struct recKeys delkeys; +void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno, + int cmd, struct sortKey **skp) +{ + struct sortKey *sk = *skp; + SortIdx sortIdx = zh->reg->sortIdx; - rec = rec_get (records, *sysno); - assert (rec); - delkeys.buf_used = rec->size[recInfo_delKeys]; - delkeys.buf = rec->info[recInfo_delKeys]; - flushSortKeys (*sysno, 0); - flushRecordKeys (*sysno, 0, &delkeys, rec->info[recInfo_databaseName]); - if (deleteFlag) + sortIdx_sysno (sortIdx, sysno); + while (sk) + { + struct sortKey *sk_next = sk->next; + if (cmd == 1) { - /* record going to be deleted */ - if (!delkeys.buf_used) - { - logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); - logf (LOG_WARN, "cannot delete file above, storeKeys false"); - } - else - { - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); - records_deleted++; - if (matchStr) - dict_delete (matchDict, matchStr); - rec_del (records, &rec); - } - logRecord (0); - return 1; + /* insert/update: set it */ + sortIdx_type (sortIdx, sk->attrUse); + sortIdx_add (sortIdx, sk->string, sk->length); } - else - { - /* record going to be updated */ - if (!delkeys.buf_used) - { - logf (LOG_LOG, "update %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); - logf (LOG_WARN, "cannot update file above, storeKeys false"); - } - else - { - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "update %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); - records_updated++; - } + else if (cmd == 0) + { /* delete : zero it */ + sortIdx_type (sortIdx, sk->attrUse); + sortIdx_add (sortIdx, "", 1); } + xfree (sk->string); + xfree (sk); + sk = sk_next; } - /* update file type */ - xfree (rec->info[recInfo_fileType]); - rec->info[recInfo_fileType] = - rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]); + *skp = 0; +} - /* update filename */ - xfree (rec->info[recInfo_filename]); - rec->info[recInfo_filename] = - rec_strdup (fname, &rec->size[recInfo_filename]); +void encode_key_init (struct encode_info *i) +{ + i->sysno = 0; + i->seqno = 0; + i->cmd = -1; + i->prevsys=0; + i->prevseq=0; + i->prevcmd=-1; + i->keylen=0; +} - /* update delete keys */ - xfree (rec->info[recInfo_delKeys]); - if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1) +char *encode_key_int (int d, char *bp) +{ + if (d <= 63) + *bp++ = d; + else if (d <= 16383) { -#if 1 - rec->size[recInfo_delKeys] = reckeys.buf_used; - rec->info[recInfo_delKeys] = reckeys.buf; - reckeys.buf = NULL; - reckeys.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 + *bp++ = 64 + (d>>8); + *bp++ = d & 255; + } + else if (d <= 4194303) + { + *bp++ = 128 + (d>>16); + *bp++ = (d>>8) & 255; + *bp++ = d & 255; } else { - rec->info[recInfo_delKeys] = NULL; - rec->size[recInfo_delKeys] = 0; + *bp++ = 192 + (d>>24); + *bp++ = (d>>16) & 255; + *bp++ = (d>>8) & 255; + *bp++ = d & 255; } + return bp; +} - /* update store data */ - xfree (rec->info[recInfo_storeData]); - if (rGroup->flagStoreData == 1) +#ifdef OLDENCODE +/* this is the old encode_key_write + * may be deleted once we are confident that the new works + * HL 15-oct-2002 + */ +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) { - int size = fi->file_moffset - recordOffset; - if (!size) - size = fi->file_max - recordOffset; - rec->size[recInfo_storeData] = size; - rec->info[recInfo_storeData] = xmalloc (size); - if (lseek (fi->fd, recordOffset, SEEK_SET) < 0) - { - logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s", fname, - (long) recordOffset); - exit (1); - } - if (read (fi->fd, rec->info[recInfo_storeData], size) < size) - { - logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", - fi->file_max, fname); - exit (1); - } + i->sysno = key.sysno; + i->seqno = 0; } - else + 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) { - rec->info[recInfo_storeData] = NULL; - rec->size[recInfo_storeData] = 0; + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); } - /* update database name */ - xfree (rec->info[recInfo_databaseName]); - rec->info[recInfo_databaseName] = - 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)); - - /* commit this record */ - rec_put (records, &rec); - logRecord (0); - return 1; +void encode_key_flush (struct encode_info *i, FILE *outf) +{ /* dummy routine */ } -int fileExtract (SYSNO *sysno, const char *fname, - const struct recordGroup *rGroupP, int deleteFlag) +#else + +/* new encode_key_write + * The idea is to buffer one more key, and compare them + * If we are going to delete and insert the same key, + * we may as well not bother. Should make a difference in + * updates with small modifications (appending to a mbox) + */ +void encode_key_write (char *k, struct encode_info *i, FILE *outf) { - int r, i, fd; - char gprefix[128]; - char ext[128]; - char ext_res[128]; - char subType[128]; - RecType recType; - struct recordGroup rGroupM; - struct recordGroup *rGroup = &rGroupM; - struct file_read_info *fi; + struct it_key key; + char *bp; - memcpy (rGroup, rGroupP, sizeof(*rGroupP)); - - if (!rGroup->groupName || !*rGroup->groupName) - *gprefix = '\0'; + if (*k) /* first time for new key */ + { + bp = i->buf; + while ((*bp++ = *k++)) + ; + i->keylen= bp - i->buf -1; + } else - sprintf (gprefix, "%s.", rGroup->groupName); - - logf (LOG_DEBUG, "fileExtract %s", fname); + { + bp=i->buf + i->keylen; + *bp++=0; + k++; + } - /* determine file extension */ - for (i = strlen(fname); --i >= 0; ) - if (fname[i] == '/') - { - strcpy (ext, ""); - break; - } - else if (fname[i] == '.') - { - strcpy (ext, fname+i+1); - break; - } - /* determine file type - depending on extension */ - if (!rGroup->recordType) + memcpy (&key, k+1, sizeof(struct it_key)); + if (0==i->prevsys) /* no previous filter, fill up */ { - sprintf (ext_res, "%srecordType.%s", gprefix, ext); - if (!(rGroup->recordType = res_get (common_resource, ext_res))) + i->prevsys=key.sysno; + i->prevseq=key.seqno; + i->prevcmd=*k; + } + else if ( (i->prevsys==key.sysno) && + (i->prevseq==key.seqno) && + (i->prevcmd!=*k) ) + { /* same numbers, diff cmd, they cancel out */ + i->prevsys=0; + } + else + { /* different stuff, write previous, move buf */ + bp = encode_key_int ( (i->prevsys - i->sysno) * 2 + i->prevcmd, bp); + if (i->sysno != i->prevsys) + { + i->sysno = i->prevsys; + i->seqno = 0; + } + else if (!i->seqno && !i->prevseq && i->cmd == i->prevcmd) + { + return; /* ??? Filters some sort of duplicates away */ + /* ??? Can this ever happen -H 15oct02 */ + } + bp = encode_key_int (i->prevseq - i->seqno, bp); + i->seqno = i->prevseq; + i->cmd = i->prevcmd; + if (fwrite (i->buf, bp - i->buf, 1, outf) != 1) { - sprintf (ext_res, "%srecordType", gprefix); - if (!(rGroup->recordType = res_get (common_resource, ext_res))) - { - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "? %s", fname); - return 0; - } + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); } + i->keylen=0; /* ok, it's written, forget it */ + i->prevsys=key.sysno; + i->prevseq=key.seqno; + i->prevcmd=*k; } - if (!rGroup->recordType) - { - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "? record %s", fname); - return 0; - } - if (!(recType = recType_byName (rGroup->recordType, subType))) +#ifdef SKIPTHIS_OLDCODE + bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp); + if (i->sysno != key.sysno) { - logf (LOG_WARN, "No such record type: %s", rGroup->recordType); - return 0; + i->sysno = key.sysno; + i->seqno = 0; } - - /* determine match criteria */ - if (!rGroup->recordId) + 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) { - sprintf (ext_res, "%srecordId.%s", gprefix, ext); - rGroup->recordId = res_get (common_resource, ext_res); + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); } + i->keylen=0; /* ok, it's written, forget it */ +#endif +} - /* determine database name */ - if (!rGroup->databaseName) +void encode_key_flush (struct encode_info *i, FILE *outf) +{ /* flush the last key from i */ + char *bp =i->buf + i->keylen; + if (0==i->prevsys) { - sprintf (ext_res, "%sdatabase.%s", gprefix, ext); - if (!(rGroup->databaseName = res_get (common_resource, ext_res))) - { - sprintf (ext_res, "%sdatabase", gprefix); - rGroup->databaseName = res_get (common_resource, ext_res); - } + return; /* nothing to flush */ } - if (!rGroup->databaseName) - rGroup->databaseName = "Default"; - - if (rGroup->flagStoreData == -1) + *bp++=0; + bp = encode_key_int ( (i->prevsys - i->sysno) * 2 + i->prevcmd, bp); + if (i->sysno != i->prevsys) { - const char *sval; - sprintf (ext_res, "%sstoreData.%s", gprefix, ext); - if (!(sval = res_get (common_resource, ext_res))) - { - sprintf (ext_res, "%sstoreData", gprefix); - sval = res_get (common_resource, ext_res); - } - if (sval) - rGroup->flagStoreData = atoi (sval); + i->sysno = i->prevsys; + i->seqno = 0; } - if (rGroup->flagStoreData == -1) - rGroup->flagStoreData = 0; - - if (rGroup->flagStoreKeys == -1) + else if (!i->seqno && !i->prevseq && i->cmd == i->prevcmd) { - const char *sval; - - sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext); - if (!(sval = res_get (common_resource, ext_res))) - { - sprintf (ext_res, "%sstoreKeys", gprefix); - sval = res_get (common_resource, ext_res); - } - if (sval) - rGroup->flagStoreKeys = atoi (sval); + return; /* ??? Filters some sort of duplicates away */ + /* ??? Can this ever happen -H 15oct02 */ } - if (rGroup->flagStoreKeys == -1) - rGroup->flagStoreKeys = 0; - - if (sysno && deleteFlag) - fd = -1; - else + bp = encode_key_int (i->prevseq - i->seqno, bp); + i->seqno = i->prevseq; + i->cmd = i->prevcmd; + if (fwrite (i->buf, bp - i->buf, 1, outf) != 1) { - if ((fd = open (fname, O_BINARY|O_RDONLY)) == -1) - { - logf (LOG_WARN|LOG_ERRNO, "open %s", fname); - return 0; - } + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); } - fi = file_read_start (fd); - do - { - file_begin (fi); - r = recordExtract (sysno, fname, rGroup, deleteFlag, fi, - recType, subType); - } while (r && !sysno && fi->file_more); - file_read_stop (fi); - if (fd != -1) - close (fd); - return r; + i->keylen=0; /* ok, it's written, forget it */ + i->prevsys=0; /* forget the values too */ + i->prevseq=0; } - +#endif