X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=18cac3d597c58577d9346d69cc32ed0805d2f728;hp=35ec2884a98bc6abf1ea594274f9ef337d58a3cc;hb=7c3a0352f0492609a3b6b26b63a72b0b2d207aab;hpb=3b0bd1aac42b5a96187b4d45af5230a2bf14fd10 diff --git a/index/extract.c b/index/extract.c index 35ec288..18cac3d 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,764 +1,283 @@ -/* - * Copyright (C) 1994-1996, Index Data I/S - * All rights reserved. - * Sebastian Hammer, Adam Dickmeiss - * - * $Log: extract.c,v $ - * 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 1.161 2004-09-14 14:38:07 quinn Exp $ + Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003,2004 + 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 +#include +#ifdef WIN32 +#include +#else #include +#endif #include -#include #include "index.h" +#include +#include -#include "zinfo.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; - -static int records_inserted = 0; -static int records_updated = 0; -static int records_deleted = 0; -static int records_processed = 0; +#if _FILE_OFFSET_BITS == 64 +#define PRINTF_OFF_T "%Ld" +#else +#define PRINTF_OFF_T "%ld" +#endif -static ZebTargetInfo *zti = NULL; +#define USE_SHELLSORT 0 -static void logRecord (int showFlag) +#if USE_SHELLSORT +static void shellsort(void *ar, int r, size_t s, + int (*cmp)(const void *a, const void *b)) { - if (!showFlag) - ++records_processed; - if (showFlag || !(records_processed % 1000)) - { - logf (LOG_LOG, "Records: %7d i/u/d %d/%d/%d", - records_processed, records_inserted, records_updated, - records_deleted); - } + 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 -void key_open (int mem) +static void logRecord (ZebraHandle zh) { - 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; - - key_buf_used = 0; - key_file_no = 0; - - if (!(matchDict = dict_open (GMATCH_DICT, 50, 1))) + ++zh->records_processed; + if (!(zh->records_processed % 1000)) { - logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT); - exit (1); + logf (LOG_LOG, "Records: "ZINT_FORMAT" i/u/d " + ZINT_FORMAT"/"ZINT_FORMAT"/"ZINT_FORMAT, + zh->records_processed, zh->records_inserted, zh->records_updated, + zh->records_deleted); } - assert (!records); - records = rec_open (1); -#if 1 - zti = zebTargetInfo_open (records, 1); -#endif } -struct encode_info { - int sysno; - int seqno; - char buf[768]; -}; - -void encode_key_init (struct encode_info *i) +static void extract_init (struct recExtractCtrl *p, RecWord *w) { - i->sysno = 0; - i->seqno = 0; + w->zebra_maps = p->zebra_maps; + w->seqno = 1; + w->attrSet = VAL_BIB1; + w->attrUse = 1016; + w->reg_type = 'w'; + w->extractCtrl = p; } -char *encode_key_int (int d, char *bp) +static const char **searchRecordKey (ZebraHandle zh, + struct recKeys *reckeys, + int attrSetS, int attrUseS) { - 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; -} +#if IT_KEY_NEW +/* #error searchRecordKey does not work yet in this mode.. */ + static const char *ws[32]; + void *decode_handle = iscz1_start(); + int off = 0; + int startSeq = -1; + int seqno = 0; + int i; -void encode_key_write (char *k, struct encode_info *i, FILE *outf) -{ - struct it_key key; - char *bp = i->buf; + for (i = 0; i<32; i++) + ws[i] = NULL; - 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) + while (off < reckeys->buf_used) { - logf (LOG_FATAL|LOG_ERRNO, "fwrite"); - exit (1); - } -} - -#define SORT_EXTRA 0 + const char *src = reckeys->buf + off; + struct it_key key; + char *dst = (char*) &key; + int attrSet, attrUse; -#if SORT_EXTRA -static int key_y_len; + iscz1_decode(decode_handle, &dst, &src); + assert(key.len < 4 && key.len > 2); -static int key_y_compare (const void *p1, const void *p2) -{ - int r; + attrSet = (int) key.mem[0]; + attrUse = (int) key.mem[1]; + seqno = (int) key.mem[2]; - 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); -} + if (attrUseS == attrUse && attrSetS == attrSet) + { + int woff; -static int key_x_compare (const void *p1, const void *p2) -{ - return strcmp (*(char**) p1, *(char**) p2); -} -#endif + if (startSeq == -1) + startSeq = seqno; + woff = seqno - startSeq; + if (woff >= 0 && woff < 31) + ws[woff] = src; + } -void key_flush (void) -{ - FILE *outf; - char out_fname[200]; - char *prevcp, *cp; - struct encode_info encode_info; -#if SORT_EXTRA + while (*src++) + ; + off = src - reckeys->buf; + } + iscz1_stop(decode_handle); + assert (off == reckeys->buf_used); + return ws; +#else + 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 - - 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); + for (i = 0; i<32; i++) + ws[i] = NULL; - if (!(outf = fopen (out_fname, "w"))) +#if SU_SCHEME + chS = zebraExplain_lookupSU (zh->reg->zei, attrSetS, attrUseS); + if (chS < 0) + return ws; +#endif + while (off < reckeys->buf_used) { - 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]; + + const char *src = reckeys->buf + off; + const char *wstart; + int lead; - 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)) + lead = *src++; +#if SU_SCHEME + if ((lead & 3)<3) + { + memcpy (&ch, src, sizeof(ch)); + src += sizeof(ch); + } +#else + if (!(lead & 1)) { - encode_key_init (&encode_info); - encode_key_write (cp, &encode_info, outf); - prevcp = cp; + memcpy (&attrSet, src, sizeof(attrSet)); + src += sizeof(attrSet); } - 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, "w"))) - { - logf (LOG_FATAL|LOG_ERRNO, "fopen (4) %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])) + if (!(lead & 2)) { - 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); + memcpy (&attrUse, src, sizeof(attrUse)); + src += sizeof(attrUse); + } #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]; + 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 - if (fclose (outf)) - { - logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname); - exit (1); + ) + { + int woff; + + + if (startSeq == -1) + startSeq = seqno; + woff = seqno - startSeq; + if (woff >= 0 && woff < 31) + ws[woff] = wstart; + } + + off = src - reckeys->buf; } - logf (LOG_LOG, "finished section %d", key_file_no); - ptr_i = 0; - key_buf_used = 0; + assert (off == reckeys->buf_used); + return ws; +#endif } -int key_close (void) +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; +}; + +static struct file_read_info *file_read_start (int fd) { - key_flush (); - xfree (key_buf); -#if 1 - zebTargetInfo_close (zti, 1); -#endif - rec_close (&records); - dict_close (matchDict); + struct file_read_info *fi = (struct file_read_info *) + xmalloc (sizeof(*fi)); - logRecord (1); - return key_file_no; + fi->fd = fd; + fi->file_max = 0; + fi->file_moffset = 0; + fi->sdrbuf = 0; + fi->sdrmax = 0; + return fi; } -static void wordInit (RecWord *p) +static void file_read_stop (struct file_read_info *fi) { - p->attrSet = 1; - p->attrUse = 1016; - p->which = Word_String; + xfree (fi); } -struct recKeys { - int buf_used; - int buf_max; - char *buf; - char prevAttrSet; - short prevAttrUse; - int prevSeqNo; -} reckeys; +static off_t file_seek (void *handle, off_t offset) +{ + 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 void addRecordKey (const RecWord *p) +static off_t file_tell (void *handle) { - char *dst; - char attrSet; - short attrUse; - size_t i; - int lead = 0; - int diff = 0; + struct file_read_info *p = (struct file_read_info *) handle; + return p->file_offset; +} - if (reckeys.buf_used+1024 > reckeys.buf_max) +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) { - char *b; - - b = xmalloc (reckeys.buf_max += 128000); - if (reckeys.buf_used > 0) - memcpy (b, reckeys.buf, reckeys.buf_used); - xfree (reckeys.buf); - reckeys.buf = b; + 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); } - dst = reckeys.buf + reckeys.buf_used; - - 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); - else - diff = 0; -#endif - reckeys.prevSeqNo = p->seqno; - - *dst++ = lead; - - if (!(lead & 1)) - { - memcpy (dst, &attrSet, sizeof(attrSet)); - dst += sizeof(attrSet); - } - if (!(lead & 2)) - { - memcpy (dst, &attrUse, sizeof(attrUse)); - dst += sizeof(attrUse); - } - switch (p->which) - { - case Word_String: - *dst++ = 'w'; - break; - case Word_Phrase: - *dst++ = 'p'; - break; - case Word_Numeric: - *dst++ = 'n'; - } - for (i = 0; p->u.string[i] && i < IT_MAX_WORD-3; i++) - *dst++ = p->u.string[i]; - *dst++ = '\0'; - - if (!diff) - { - memcpy (dst, &p->seqno, sizeof(p->seqno)); - dst += sizeof(p->seqno); - } - reckeys.buf_used = dst - reckeys.buf; -} - -static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, - const char *databaseName) -{ - char attrSet = -1; - short attrUse = -1; - int seqno = 0; - int off = 0; - - if (zebTargetInfo_curDatabase (zti, databaseName)) - { - if (zebTargetInfo_newDatabase (zti, databaseName)) - abort (); - } - while (off < reckeys->buf_used) - { - const char *src = reckeys->buf + off; - struct it_key key; - int lead, ch; - - lead = *src++; - - if (!(lead & 1)) - { - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); - } - if (!(lead & 2)) - { - 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); - if (ch < 0) - ch = zebTargetInfo_addSU (zti, attrSet, attrUse); - assert (ch > 0); - ((char*) key_buf) [key_buf_used++] = ch; - while (*src) - ((char*)key_buf) [key_buf_used++] = *src++; - src++; - ((char*)key_buf) [key_buf_used++] = '\0'; - ((char*) key_buf)[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*)key_buf + key_buf_used, &key, sizeof(key)); - 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) -{ - static const char *ws[32]; - int off = 0; - int startSeq = -1; - int i; - short attrUse; - char attrSet; - int seqno = 0; - - for (i = 0; i<32; i++) - ws[i] = NULL; - - while (off < reckeys->buf_used) - { - - const char *src = reckeys->buf + off; - const char *wstart; - int lead; - - lead = *src++; - - if (!(lead & 1)) - { - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); - } - if (!(lead & 2)) - { - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); - } - wstart = src; - while (*src++) - ; - 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 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) + r = read (fd, buf, count); + if (r > 0) { p->file_offset += r; if (p->file_offset > p->file_max) @@ -769,43 +288,31 @@ static int file_read (void *handle, char *buf, size_t count) static void file_begin (void *handle) { - struct file_read_info *p = handle; + struct file_read_info *p = (struct file_read_info *) handle; p->file_offset = p->file_moffset; - if (p->file_moffset) + if (!p->sdrbuf && p->file_moffset) lseek (p->fd, p->file_moffset, SEEK_SET); p->file_more = 0; } static void file_end (void *handle, off_t offset) { - struct file_read_info *p = handle; + struct file_read_info *p = (struct file_read_info *) handle; assert (p->file_more == 0); p->file_more = 1; p->file_moffset = offset; } -static int atois (const char **s) -{ - int val = 0, c; - while ( (c=**s) >= '0' && c <= '9') - { - val = val*10 + c - '0'; - ++(*s); - } - return val; -} - -static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, - const char *fname, - const char *spec) +static char *fileMatchStr (ZebraHandle zh, + struct recKeys *reckeys, + 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) { @@ -815,21 +322,39 @@ 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 == ')') @@ -840,7 +365,7 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, else { logf (LOG_WARN, "Missing ) in match criteria %s in group %s", - spec, rGroup->groupName ? rGroup->groupName : "none"); + spec, zh->m_group ? zh->m_group : "none"); return NULL; } s++; @@ -859,7 +384,7 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, if (first) { logf (LOG_WARN, "Record didn't contain match" - " fields in (%d,%d)", attrSet, attrUse); + " fields in (%s,%s)", attset_str, attname_str); return NULL; } } @@ -880,13 +405,14 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, s = s1; if (!strcmp (special, "group")) - spec_src = rGroup->groupName; + spec_src = zh->m_group; else if (!strcmp (special, "database")) - spec_src = rGroup->databaseName; - else if (!strcmp (special, "filename")) + spec_src = zh->basenames[0]; + else if (!strcmp (special, "filename")) { spec_src = fname; + } else if (!strcmp (special, "type")) - spec_src = rGroup->recordType; + spec_src = zh->m_record_type; else spec_src = NULL; if (spec_src) @@ -915,7 +441,7 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, else { logf (LOG_WARN, "Syntax error in match criteria %s in group %s", - spec, rGroup->groupName ? rGroup->groupName : "none"); + spec, zh->m_group ? zh->m_group : "none"); return NULL; } *dst++ = 1; @@ -923,7 +449,7 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, if (dst == dstBuf) { logf (LOG_WARN, "No match criteria for record %s in group %s", - fname, rGroup->groupName ? rGroup->groupName : "none"); + fname, zh->m_group ? zh->m_group : "none"); return NULL; } *dst = '\0'; @@ -935,84 +461,132 @@ struct recordLogInfo { int recordOffset; struct recordGroup *rGroup; }; - -static void recordLogPreamble (int level, const char *msg, void *info) + +void create_rec_keys_codec(struct recKeys *keys) { - 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); + keys->buf_used = 0; +#if IT_KEY_NEW + iscz1_reset(keys->codec_handle); +#else + keys->prevAttrUse = -1; + keys->prevAttrSet = -1; + keys->prevSeqNo = 0; +#endif } - -static int recordExtract (SYSNO *sysno, const char *fname, - struct recordGroup *rGroup, int deleteFlag, - struct file_read_info *fi, RecType recType, - char *subType) + +static int file_extract_record(ZebraHandle zh, + SYSNO *sysno, const char *fname, + int deleteFlag, + struct file_read_info *fi, + int force_update) { - struct recExtractCtrl extractCtrl; + RecordAttr *recordAttr; int r; - char *matchStr; + const char *matchStr; SYSNO sysnotmp; - off_t recordOffset = 0; Record rec; - struct recordLogInfo logInfo; + off_t recordOffset = 0; + RecType recType; + void *clientData; + char subType[128]; + + if (!(recType = + recType_byName (zh->reg->recTypes, zh->m_record_type, subType, + &clientData))) + { + logf (LOG_WARN, "No such record type: %s", zh->m_record_type); + return 0; + } + + /* announce database */ + if (zebraExplain_curDatabase (zh->reg->zei, zh->basenames[0])) + { + if (zebraExplain_newDatabase (zh->reg->zei, zh->basenames[0], + zh->m_explain_database)) + return 0; + } if (fi->fd != -1) { + struct recExtractCtrl extractCtrl; + /* we are going to read from a file, so prepare the extraction */ - extractCtrl.fh = fi; - extractCtrl.subType = subType; - extractCtrl.init = wordInit; - extractCtrl.add = addRecordKey; - - reckeys.buf_used = 0; - reckeys.prevAttrUse = -1; - reckeys.prevAttrSet = -1; - reckeys.prevSeqNo = 0; - - recordOffset = fi->file_moffset; - extractCtrl.offset = recordOffset; - extractCtrl.readf = file_read; - extractCtrl.seekf = file_seek; - extractCtrl.endf = file_end; - extractCtrl.map_chrs_input = map_chrs_input; - extractCtrl.flagShowRecords = rGroup->flagShowRecords; - if (rGroup->flagShowRecords) - printf ("File: %s %ld\n", fname, (long) recordOffset); - - logInfo.fname = fname; - logInfo.recordOffset = recordOffset; - logInfo.rGroup = rGroup; - log_event_start (recordLogPreamble, &logInfo); - - r = (*recType->extract)(&extractCtrl); - - log_event_start (NULL, NULL); - - if (r) + int i; + + create_rec_keys_codec(&zh->reg->keys); + + zh->reg->sortKeys.buf_used = 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 = !zh->m_flag_rw; + + if (!zh->m_flag_rw) + printf ("File: %s " PRINTF_OFF_T "\n", fname, recordOffset); + if (zh->m_flag_rw) { + 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 (zh->m_flag_rw && + zh->records_processed < zh->m_file_verbose_limit) + { + logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, zh->m_record_type, + fname, recordOffset); + } + return 0; + } + else if (r == RECCTRL_EXTRACT_ERROR_NO_SUCH_FILTER) + { /* error occured during extraction ... */ - if (!rGroup->flagShowRecords && - records_processed < rGroup->fileVerboseLimit) + if (zh->m_flag_rw && + zh->records_processed < zh->m_file_verbose_limit) { - logf (LOG_WARN, "fail %s %s %ld code = %d", rGroup->recordType, - fname, (long) recordOffset, r); + logf (LOG_WARN, "no filter for %s %s " + PRINTF_OFF_T, zh->m_record_type, + fname, recordOffset); } return 0; } - if (reckeys.buf_used == 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->flagShowRecords) + if (!zh->m_flag_rw) return 1; - logf (LOG_WARN, "No keys generated for file %s", fname); - logf (LOG_WARN, " The file is probably empty"); - return 0; + + logf (LOG_WARN, "empty %s %s " PRINTF_OFF_T, zh->m_record_type, + fname, recordOffset); + return 1; } } @@ -1023,17 +597,20 @@ static int recordExtract (SYSNO *sysno, const char *fname, { sysnotmp = 0; sysno = &sysnotmp; - if (rGroup->recordId && *rGroup->recordId) + if (zh->m_record_id && *zh->m_record_id) { char *rinfo; - matchStr = fileMatchStr (&reckeys, rGroup, fname, - rGroup->recordId); + matchStr = fileMatchStr (zh, &zh->reg->keys, fname, + zh->m_record_id); if (matchStr) { - rinfo = dict_lookup (matchDict, matchStr); + rinfo = dict_lookup (zh->reg->matchDict, matchStr); if (rinfo) + { + assert(*rinfo == sizeof(*sysno)); memcpy (sysno, rinfo+1, sizeof(*sysno)); + } } else { @@ -1048,53 +625,80 @@ static int recordExtract (SYSNO *sysno, const char *fname, /* new record */ if (deleteFlag) { - logf (LOG_LOG, "Cannot delete new record"); + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, zh->m_record_type, + fname, 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); + if (zh->records_processed < zh->m_file_verbose_limit) + logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, zh->m_record_type, + 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_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); + extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); - records_inserted++; + zh->records_inserted++; } else { /* record already exists */ struct recKeys delkeys; + struct sortKeys sortKeys; - rec = rec_get (records, *sysno); + rec = rec_get (zh->reg->records, *sysno); assert (rec); + + recordAttr = rec_init_attr (zh->reg->zei, rec); + + if (!force_update && 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, + zh->m_record_type, 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]; - flushRecordKeys (*sysno, 0, &delkeys, rec->info[recInfo_databaseName]); + + sortKeys.buf_used = rec->size[recInfo_sortKeys]; + sortKeys.buf = rec->info[recInfo_sortKeys]; + + extract_flushSortKeys (zh, *sysno, 0, &sortKeys); + extract_flushRecordKeys (zh, *sysno, 0, &delkeys); if (deleteFlag) { /* record going to be deleted */ if (!delkeys.buf_used) { - logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + zh->m_record_type, fname, 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 (zh->records_processed < zh->m_file_verbose_limit) + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + zh->m_record_type, fname, recordOffset); + zh->records_deleted++; if (matchStr) - dict_delete (matchDict, matchStr); - rec_del (records, &rec); + dict_delete (zh->reg->matchDict, matchStr); + rec_del (zh->reg->records, &rec); } - logRecord (0); + rec_rm (&rec); + logRecord (zh); return 1; } else @@ -1102,24 +706,25 @@ static int recordExtract (SYSNO *sysno, const char *fname, /* record going to be updated */ if (!delkeys.buf_used) { - logf (LOG_LOG, "update %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); + logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, + zh->m_record_type, fname, 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++; + if (zh->records_processed < zh->m_file_verbose_limit) + logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, + zh->m_record_type, fname, recordOffset); + extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); + 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]); + rec_strdup (zh->m_record_type, &rec->size[recInfo_fileType]); /* update filename */ xfree (rec->info[recInfo_filename]); @@ -1128,19 +733,12 @@ static int recordExtract (SYSNO *sysno, const char *fname, /* update delete keys */ xfree (rec->info[recInfo_delKeys]); - if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1) + if (zh->reg->keys.buf_used > 0 && zh->m_store_keys == 1) { -#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 + 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 { @@ -1148,25 +746,45 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec->size[recInfo_delKeys] = 0; } + /* update sort keys */ + xfree (rec->info[recInfo_sortKeys]); + + rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used; + rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf; + zh->reg->sortKeys.buf = NULL; + zh->reg->sortKeys.buf_max = 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) + if (zh->m_store_data) { - int size = fi->file_moffset - recordOffset; - if (!size) - size = fi->file_max - recordOffset; - rec->size[recInfo_storeData] = size; - rec->info[recInfo_storeData] = xmalloc (size); + 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 %ld in %s", fname, - (long) recordOffset); + logf (LOG_ERRNO|LOG_FATAL, "seek to " PRINTF_OFF_T " in %s", + recordOffset, fname); exit (1); } - if (read (fi->fd, rec->info[recInfo_storeData], size) < size) + if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize) + < recordAttr->recordSize) { logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", - fi->file_max, fname); + recordAttr->recordSize, fname); exit (1); } } @@ -1178,140 +796,84 @@ static int recordExtract (SYSNO *sysno, const char *fname, /* update database name */ xfree (rec->info[recInfo_databaseName]); rec->info[recInfo_databaseName] = - rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]); + rec_strdup (zh->basenames[0], &rec->size[recInfo_databaseName]); /* update offset */ - xfree (rec->info[recInfo_offset]); - - rec->size[recInfo_offset] = sizeof(recordOffset); - rec->info[recInfo_offset] = xmalloc (sizeof(recordOffset)); - memcpy (rec->info[recInfo_offset], &recordOffset, sizeof(recordOffset)); + recordAttr->recordOffset = recordOffset; /* commit this record */ - rec_put (records, &rec); - logRecord (0); + rec_put (zh->reg->records, &rec); + logRecord (zh); return 1; } -int fileExtract (SYSNO *sysno, const char *fname, - const struct recordGroup *rGroupP, int deleteFlag) +int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, + 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; + const char *original_record_type = 0; - memcpy (rGroup, rGroupP, sizeof(*rGroupP)); - - if (!rGroup->groupName || !*rGroup->groupName) + if (!zh->m_group || !*zh->m_group) *gprefix = '\0'; else - sprintf (gprefix, "%s.", rGroup->groupName); - + sprintf (gprefix, "%s.", zh->m_group); + 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); break; } /* determine file type - depending on extension */ - if (!rGroup->recordType) + original_record_type = zh->m_record_type; + if (!zh->m_record_type) { sprintf (ext_res, "%srecordType.%s", gprefix, ext); - if (!(rGroup->recordType = res_get (common_resource, ext_res))) - { - 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; - } - } - } - if (!rGroup->recordType) - { - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "? record %s", fname); - return 0; + zh->m_record_type = res_get (zh->res, ext_res); } - if (!(recType = recType_byName (rGroup->recordType, subType))) + if (!zh->m_record_type) { - logf (LOG_WARN, "No such record type: %s", rGroup->recordType); + if (zh->records_processed < zh->m_file_verbose_limit) + logf (LOG_LOG, "? %s", fname); return 0; } - /* determine match criteria */ - if (!rGroup->recordId) + if (!zh->m_record_id) { sprintf (ext_res, "%srecordId.%s", gprefix, ext); - rGroup->recordId = res_get (common_resource, ext_res); - } - - /* determine database name */ - if (!rGroup->databaseName) - { - 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); - } - } - if (!rGroup->databaseName) - rGroup->databaseName = "Default"; - - if (rGroup->flagStoreData == -1) - { - 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); + zh->m_record_id = res_get (zh->res, ext_res); } - if (rGroup->flagStoreData == -1) - rGroup->flagStoreData = 0; - if (rGroup->flagStoreKeys == -1) + if (sysno && deleteFlag) + fd = -1; + else { - const char *sval; + char full_rep[1024]; - sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext); - if (!(sval = res_get (common_resource, ext_res))) + if (zh->path_reg && !yaz_is_abspath (fname)) { - sprintf (ext_res, "%sstoreKeys", gprefix); - sval = res_get (common_resource, ext_res); + strcpy (full_rep, zh->path_reg); + strcat (full_rep, "/"); + strcat (full_rep, fname); } - if (sval) - rGroup->flagStoreKeys = atoi (sval); - } - if (rGroup->flagStoreKeys == -1) - rGroup->flagStoreKeys = 0; + else + strcpy (full_rep, fname); + - if (sysno && deleteFlag) - fd = -1; - else - { - if ((fd = open (fname, O_RDONLY)) == -1) + if ((fd = open (full_rep, O_BINARY|O_RDONLY)) == -1) { - logf (LOG_WARN|LOG_ERRNO, "open %s", fname); + logf (LOG_WARN|LOG_ERRNO, "open %s", full_rep); + zh->m_record_type = original_record_type; return 0; } } @@ -1319,12 +881,1205 @@ int fileExtract (SYSNO *sysno, const char *fname, do { file_begin (fi); - r = recordExtract (sysno, fname, rGroup, deleteFlag, fi, - recType, subType); + r = file_extract_record (zh, sysno, fname, deleteFlag, fi, 1); } while (r && !sysno && fi->file_more); file_read_stop (fi); if (fd != -1) close (fd); + zh->m_record_type = original_record_type; return r; } +/* + If sysno is provided, then it's used to identify the reocord. + If not, and match_criteria is provided, then sysno is guessed + If not, and a record is provided, then sysno is got from there + + */ +int buffer_extract_record (ZebraHandle zh, + const char *buf, size_t buf_size, + int delete_flag, + int test_mode, + const char *recordType, + SYSNO *sysno, + const char *match_criteria, + const char *fname, + int force_update, + int allow_update) +{ + RecordAttr *recordAttr; + struct recExtractCtrl extractCtrl; + int i, r; + const char *matchStr = 0; + RecType recType = NULL; + char subType[1024]; + void *clientData; + Record rec; + long recordOffset = 0; + struct zebra_fetch_control fc; + const char *pr_fname = fname; /* filename to print .. */ + + if (!pr_fname) + pr_fname = ""; /* make it printable if file is omitted */ + + 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; + + create_rec_keys_codec(&zh->reg->keys); + + zh->reg->sortKeys.buf_used = 0; + + if (zebraExplain_curDatabase (zh->reg->zei, zh->basenames[0])) + { + if (zebraExplain_newDatabase (zh->reg->zei, zh->basenames[0], + zh->m_explain_database)) + return 0; + } + + if (recordType && *recordType) { + logf (LOG_DEBUG, "Record type explicitly specified: %s", recordType); + recType = recType_byName (zh->reg->recTypes, recordType, subType, + &clientData); + } else { + if (!(zh->m_record_type)) { + logf (LOG_WARN, "No such record type defined"); + return 0; + } + logf (LOG_DEBUG, "Get record type from rgroup: %s",zh->m_record_type); + recType = recType_byName (zh->reg->recTypes, zh->m_record_type, subType, + &clientData); + recordType = zh->m_record_type; + } + + if (!recType) { + logf (LOG_WARN, "No such record type: %s", zh->m_record_type); + return 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_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 */ + matchStr = NULL; + + if (! *sysno) { + char *rinfo; + if (match_criteria && *match_criteria) { + matchStr = match_criteria; + } else { + if (zh->m_record_id && *zh->m_record_id) { + matchStr = fileMatchStr (zh, &zh->reg->keys, pr_fname, + zh->m_record_id); + if (!matchStr) + { + logf (LOG_WARN, "Bad match criteria (recordID)"); + return 1; + } + } + } + if (matchStr) { + rinfo = dict_lookup (zh->reg->matchDict, matchStr); + if (rinfo) + { + assert(*rinfo == sizeof(*sysno)); + memcpy (sysno, rinfo+1, sizeof(*sysno)); + } + } + } + + if (! *sysno) + { + /* new record */ + if (delete_flag) + { + logf (LOG_LOG, "delete %s %s %ld", recordType, + pr_fname, (long) recordOffset); + logf (LOG_WARN, "cannot delete record above (seems new)"); + return 1; + } + logf (LOG_LOG, "add %s %s %ld", recordType, pr_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_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); + extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); + + zh->records_inserted++; + } + else + { + /* record already exists */ + struct recKeys delkeys; + struct sortKeys sortKeys; + + if (!allow_update) { + logf (LOG_LOG, "skipped %s %s %ld", + recordType, pr_fname, (long) recordOffset); + logRecord(zh); + return -1; + } + + rec = rec_get (zh->reg->records, *sysno); + assert (rec); + + recordAttr = rec_init_attr (zh->reg->zei, rec); + + if (!force_update) { + if (recordAttr->runNumber == + zebraExplain_runNumberIncrement (zh->reg->zei, 0)) + { + logf (LOG_LOG, "skipped %s %s %ld", recordType, + pr_fname, (long) 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]; + + sortKeys.buf_used = rec->size[recInfo_sortKeys]; + sortKeys.buf = rec->info[recInfo_sortKeys]; + + extract_flushSortKeys (zh, *sysno, 0, &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, + pr_fname, (long) recordOffset); + logf (LOG_WARN, "cannot delete file above, storeKeys false"); + } + else + { + logf (LOG_LOG, "delete %s %s %ld", recordType, + pr_fname, (long) recordOffset); + zh->records_deleted++; + if (matchStr) + dict_delete (zh->reg->matchDict, matchStr); + rec_del (zh->reg->records, &rec); + } + rec_rm (&rec); + logRecord(zh); + return 0; + } + else + { + /* record going to be updated */ + if (!delkeys.buf_used) + { + logf (LOG_LOG, "update %s %s %ld", recordType, + pr_fname, (long) recordOffset); + logf (LOG_WARN, "cannot update file above, storeKeys false"); + } + else + { + logf (LOG_LOG, "update %s %s %ld", recordType, + pr_fname, (long) recordOffset); + extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); + 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 (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 && zh->m_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; + } + + /* update sort keys */ + xfree (rec->info[recInfo_sortKeys]); + + rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used; + rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf; + zh->reg->sortKeys.buf = NULL; + zh->reg->sortKeys.buf_max = 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 (zh->m_store_data) + { + rec->size[recInfo_storeData] = recordAttr->recordSize; + rec->info[recInfo_storeData] = (char *) + xmalloc (recordAttr->recordSize); + memcpy (rec->info[recInfo_storeData], buf, recordAttr->recordSize); + } + 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 (zh->basenames[0], &rec->size[recInfo_databaseName]); + + /* update offset */ + recordAttr->recordOffset = recordOffset; + + /* commit this record */ + rec_put (zh->reg->records, &rec); + logRecord(zh); + 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 (); + } + + create_rec_keys_codec(&zh->reg->keys); + + zh->reg->sortKeys.buf_used = 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; + + if (n) + grs_extract_tree(&extractCtrl, n); + + if (rec->size[recInfo_delKeys]) + { + struct recKeys delkeys; + struct sortKeys sortkeys; + + delkeys.buf_used = rec->size[recInfo_delKeys]; + delkeys.buf = rec->info[recInfo_delKeys]; + + sortkeys.buf_used = rec->size[recInfo_sortKeys]; + sortkeys.buf = rec->info[recInfo_sortKeys]; + + 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; + + xfree (rec->info[recInfo_sortKeys]); + rec->size[recInfo_sortKeys] = zh->reg->sortKeys.buf_used; + rec->info[recInfo_sortKeys] = zh->reg->sortKeys.buf; + zh->reg->sortKeys.buf = NULL; + zh->reg->sortKeys.buf_max = 0; + + return 0; +} + +void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno, + int cmd, struct recKeys *reckeys) +{ +#if IT_KEY_NEW + void *decode_handle = iscz1_start(); +#else + int seqno = 0; +#if SU_SCHEME +#else + unsigned char attrSet = (unsigned char) -1; + unsigned short attrUse = (unsigned short) -1; +#endif +#endif + int off = 0; + int ch = 0; + ZebraExplainInfo zei = zh->reg->zei; + + if (!zh->reg->key_buf) + { + 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); +#if IT_KEY_NEW + while (off < reckeys->buf_used) + { + const char *src = reckeys->buf + off; + struct it_key key; + char *dst = (char*) &key; + int attrSet, attrUse; + + iscz1_decode(decode_handle, &dst, &src); + assert(key.len < 4 && key.len > 2); + + attrSet = (int) key.mem[0]; + attrUse = (int) key.mem[1]; /* sequence in mem[2] */ + + if (zh->reg->key_buf_used + 1024 > + (zh->reg->ptr_top -zh->reg->ptr_i)*sizeof(char*)) + extract_flushWriteKeys (zh,0); + ++(zh->reg->ptr_i); + assert(zh->reg->ptr_i > 0); + (zh->reg->key_buf)[zh->reg->ptr_top - zh->reg->ptr_i] = + (char*)zh->reg->key_buf + zh->reg->key_buf_used; + + ch = zebraExplain_lookupSU (zei, attrSet, attrUse); + if (ch < 0) + ch = zebraExplain_addSU (zei, attrSet, attrUse); + + 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; + + key.len = 2; + key.mem[0] = sysno; + key.mem[1] = key.mem[2]; /* sequence .. */ + + 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; + } +#else + 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,0); + assert(zh->reg->ptr_i >= 0); + ++(zh->reg->ptr_i); + assert(zh->reg->ptr_i > 0); + (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; + } +#endif + assert (off == reckeys->buf_used); +#if IT_KEY_NEW + iscz1_stop(decode_handle); +#endif +} + +void extract_flushWriteKeys (ZebraHandle zh, int final) + /* optimizing: if final=1, and no files written yet */ + /* push the keys directly to merge, sidestepping the */ + /* temp file altogether. Speeds small updates */ +{ + FILE *outf; + char out_fname[200]; + char *prevcp, *cp; + struct encode_info encode_info; + int ptr_i = zh->reg->ptr_i; + int temp_policy; +#if SORT_EXTRA + int i; +#endif + if (!zh->reg->key_buf || ptr_i <= 0) + { + logf (LOG_DEBUG, " nothing to flush section=%d buf=%p i=%d", + zh->reg->key_file_no, zh->reg->key_buf, ptr_i); + logf (LOG_DEBUG, " buf=%p ", + zh->reg->key_buf); + logf (LOG_DEBUG, " ptr=%d ",zh->reg->ptr_i); + logf (LOG_DEBUG, " reg=%p ",zh->reg); + + return; + } + + (zh->reg->key_file_no)++; + logf (LOG_LOG, "sorting section %d", (zh->reg->key_file_no)); + logf (LOG_DEBUG, " sort_buff at %p n=%d", + zh->reg->key_buf + zh->reg->ptr_top - ptr_i,ptr_i); +#if !SORT_EXTRA + qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i, + sizeof(char*), key_qsort_compare); + + /* zebra.cfg: tempfiles: + Y: always use temp files (old way) + A: use temp files, if more than one (auto) + = if this is both the last and the first + N: never bother with temp files (new) */ + + temp_policy=toupper(res_get_def(zh->res,"tempfiles","auto")[0]); + if (temp_policy != 'Y' && temp_policy != 'N' && temp_policy != 'A') { + logf (LOG_WARN, "Illegal tempfiles setting '%c'. using 'Auto' ", + temp_policy); + temp_policy='A'; + } + + if ( ( temp_policy =='N' ) || /* always from memory */ + ( ( temp_policy =='A' ) && /* automatic */ + (zh->reg->key_file_no == 1) && /* this is first time */ + (final) ) ) /* and last (=only) time */ + { /* go directly from memory */ + zh->reg->key_file_no =0; /* signal not to read files */ + zebra_index_merge(zh); + zh->reg->ptr_i = 0; + zh->reg->key_buf_used = 0; + return; + } + + /* Not doing directly from memory, write into a temp file */ + 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_flush ( &encode_info, outf); + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + prevcp = cp; + } + else + encode_key_write (cp + strlen(cp), &encode_info, outf); + } + 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); + + 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); + } + encode_key_flush ( &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 *str, int length) +{ + char *dst; + ZebraHandle zh = p->extractCtrl->handle; + struct recKeys *keys = &zh->reg->keys; +#if IT_KEY_NEW + struct it_key key; + const char *src = (char*) &key; +#else + unsigned char attrSet; + unsigned short attrUse; + int lead = 0; + int diff = 0; + int *pseqno = &p->seqno; + ZebraExplainInfo zei = zh->reg->zei; +#endif + + 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; + +#if IT_KEY_NEW + key.len = 3; + key.mem[0] = p->attrSet; + key.mem[1] = p->attrUse; + key.mem[2] = p->seqno; + +#if 0 + /* just for debugging .. */ + yaz_log(LOG_LOG, "set=%d use=%d seqno=%d", p->attrSet, p->attrUse, + p->seqno); +#endif + + iscz1_encode(keys->codec_handle, &dst, &src); + + *dst++ = p->reg_type; + memcpy (dst, str, length); + dst += length; + *dst++ = '\0'; +#else + /* leader byte is encoded as follows: + bit 0 : 1 if attrset is unchanged; 0 if attrset is changed + bit 1 : 1 if attruse is unchanged; 0 if attruse is changed + */ + 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; + + diff = 1 + *pseqno - keys->prevSeqNo; + if (diff >= 1 && diff <= 15) + lead |= (diff << 2); + else + diff = 0; + + 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, str, length); + dst += length; + *dst++ = '\0'; + + if (!diff) + { + memcpy (dst, pseqno, sizeof(*pseqno)); + dst += sizeof(*pseqno); + } +#endif + keys->buf_used = dst - keys->buf; +} + +static void extract_add_sort_string (RecWord *p, const char *str, + int length) +{ + ZebraHandle zh = p->extractCtrl->handle; + struct sortKeys *sk = &zh->reg->sortKeys; + int off = 0; + + while (off < sk->buf_used) + { + int set, use, slen; + + off += key_SU_decode(&set, sk->buf + off); + off += key_SU_decode(&use, sk->buf + off); + off += key_SU_decode(&slen, sk->buf + off); + off += slen; + if (p->attrSet == set && p->attrUse == use) + return; + } + assert (off == sk->buf_used); + + if (sk->buf_used + IT_MAX_WORD > sk->buf_max) + { + char *b; + + b = (char *) xmalloc (sk->buf_max += 128000); + if (sk->buf_used > 0) + memcpy (b, sk->buf, sk->buf_used); + xfree (sk->buf); + sk->buf = b; + } + off += key_SU_encode(p->attrSet, sk->buf + off); + off += key_SU_encode(p->attrUse, sk->buf + off); + off += key_SU_encode(length, sk->buf + off); + memcpy (sk->buf + off, str, length); + sk->buf_used = off + length; +} + +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, 0); + + 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, 0); + 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, 0); + else + map = 0; + } + if (!i) + return; + extract_add_string (p, buf, i); + p->seqno++; + } +} + +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; + int first; /* first position */ + +yaz_log(LOG_DEBUG, "Complete field, w='%s'", p->string); + + if (remain > 0) + map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain, 1); + + while (remain > 0 && i < IT_MAX_WORD) + { + while (map && *map && **map == *CHR_SPACE) + { + remain = p->length - (b - p->string); + + if (remain > 0) + { + first = i ? 0 : 1; + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain, first); + } + 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 (**map == *CHR_CUT) + { + i = 0; + } + else + { + if (i >= IT_MAX_WORD) + break; +yaz_log(LOG_DEBUG, "Adding string to index '%d'", *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, 0); + } + else + map = 0; + } + } + if (!i) + return; + extract_add_string (p, buf, i); +} + +void extract_token_add (RecWord *p) +{ + WRBUF wrbuf; +#if 0 + yaz_log (LOG_LOG, "token_add " + "reg_type=%c attrSet=%d attrUse=%d seqno=%d s=%.*s", + p->reg_type, p->attrSet, p->attrUse, p->seqno, p->length, + p->string); +#endif + 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 sortKeys *sk) +{ + SortIdx sortIdx = zh->reg->sortIdx; + int off = 0; + + sortIdx_sysno (sortIdx, sysno); + + while (off < sk->buf_used) + { + int set, use, slen; + + off += key_SU_decode(&set, sk->buf + off); + off += key_SU_decode(&use, sk->buf + off); + off += key_SU_decode(&slen, sk->buf + off); + + sortIdx_type(sortIdx, use); + if (cmd == 1) + sortIdx_add(sortIdx, sk->buf + off, slen); + else + sortIdx_add(sortIdx, "", 1); + off += slen; + } +} + +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; +#if IT_KEY_NEW + i->encode_handle = iscz1_start(); +#endif +} + +#if IT_KEY_NEW +#else +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; +} +#endif + +#define OLDENCODE 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, *bp0; + const char *src = (char *) &key; + + /* copy term to output buf */ + while ((*bp++ = *k++)) + ; + /* and copy & align key so we can mangle */ + memcpy (&key, k+1, sizeof(struct it_key)); /* *k is insert/delete */ +#if IT_KEY_NEW + bp0 = bp++; + iscz1_encode(i->encode_handle, &bp, &src); + *bp0 = (*k * 128) + bp - bp0 - 1; /* length and insert/delete combined */ +#else + 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; +#endif + if (fwrite (i->buf, bp - i->buf, 1, outf) != 1) + { + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); + } +} + +void encode_key_flush (struct encode_info *i, FILE *outf) +{ /* dummy routine */ +#if IT_KEY_NEW + iscz1_stop(i->encode_handle); +#endif +} + +#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) +{ + struct it_key key; + char *bp; + + if (*k) /* first time for new key */ + { + bp = i->buf; + while ((*bp++ = *k++)) + ; + i->keylen= bp - i->buf -1; + assert(i->keylen+1+sizeof(struct it_key) < ENCODE_BUFLEN); + } + else + { + bp=i->buf + i->keylen; + *bp++=0; + k++; + } + + memcpy (&key, k+1, sizeof(struct it_key)); + if (0==i->prevsys) /* no previous filter, fill up */ + { + 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) + { + 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; + } +} + +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) + { + return; /* nothing to flush */ + } + *bp++=0; + 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) + { + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); + } + i->keylen=0; /* ok, it's written, forget it */ + i->prevsys=0; /* forget the values too */ + i->prevseq=0; +} +#endif