X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=366c6ecd1199e519339ed665e3d6c901531fea7a;hp=23ba68ed36e7729582b91b64a46a2cb8bb7ed3ac;hb=1ccf2613ceef2359f589cb3dd7e72a899c618b2f;hpb=e80772c51b80b8e5c9c3cfb037c988308a4014b9 diff --git a/index/extract.c b/index/extract.c index 23ba68e..366c6ec 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,10 +1,257 @@ /* - * Copyright (C) 1994-1995, Index Data I/S + * Copyright (C) 1994-2000, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * * $Log: extract.c,v $ - * Revision 1.36 1995-11-30 08:34:29 adam + * Revision 1.105 2000-12-05 10:01:44 adam + * Fixed bug regarding user-defined attribute sets. + * + * Revision 1.104 2000/09/05 14:04:05 adam + * Updates for prefix 'yaz_' for YAZ log functions. + * + * Revision 1.103 2000/05/18 12:01:36 adam + * System call times(2) used again. More 64-bit fixes. + * + * Revision 1.102 2000/05/15 15:32:33 adam + * Added 64 bit file input. + * + * Revision 1.101 2000/05/15 13:02:39 adam + * Minor change. + * + * Revision 1.100 2000/03/20 19:08:36 adam + * Added remote record import using Z39.50 extended services and Segment + * Requests. + * + * Revision 1.99 2000/02/24 10:57:02 adam + * Sequence number incremented after each incomplete-field. + * + * Revision 1.98 1999/09/07 07:19:21 adam + * Work on character mapping. Implemented replace rules. + * + * Revision 1.97 1999/07/06 12:28:04 adam + * Updated record index structure. Format includes version ID. Compression + * algorithm ID is stored for each record block. + * + * Revision 1.96 1999/05/26 07:49:13 adam + * C++ compilation. + * + * Revision 1.95 1999/05/21 12:00:17 adam + * Better diagnostics for extraction process. + * + * Revision 1.94 1999/05/20 12:57:18 adam + * Implemented TCL filter. Updated recctrl system. + * + * Revision 1.93 1999/05/15 14:36:38 adam + * Updated dictionary. Implemented "compression" of dictionary. + * + * Revision 1.92 1999/03/09 16:27:49 adam + * More work on SDRKit integration. + * + * Revision 1.91 1999/02/12 13:29:22 adam + * Implemented position-flag for registers. + * + * Revision 1.90 1999/02/02 14:50:52 adam + * Updated WIN32 code specific sections. Changed header. + * + * Revision 1.89 1998/10/28 10:54:38 adam + * SDRKit integration. + * + * Revision 1.88 1998/10/16 08:14:29 adam + * Updated record control system. + * + * Revision 1.87 1998/10/15 13:10:33 adam + * Fixed bug in Zebra that caused it to stop indexing when empty + * record was read. + * + * Revision 1.86 1998/10/13 20:33:53 adam + * Fixed one log message and change use ordinal to be an unsigned char. + * + * Revision 1.85 1998/09/22 10:03:41 adam + * Changed result sets to be persistent in the sense that they can + * be re-searched if needed. + * Fixed memory leak in rsm_or. + * + * Revision 1.84 1998/06/11 15:42:22 adam + * Changed the way use attributes are specified in the recordId + * specification. + * + * Revision 1.83 1998/06/08 14:43:10 adam + * Added suport for EXPLAIN Proxy servers - added settings databasePath + * and explainDatabase to facilitate this. Increased maximum number + * of databases and attributes in one register. + * + * Revision 1.82 1998/05/20 10:12:15 adam + * Implemented automatic EXPLAIN database maintenance. + * Modified Zebra to work with ASN.1 compiled version of YAZ. + * + * Revision 1.81 1998/03/11 11:19:04 adam + * Changed the way sequence numbers are generated. + * + * Revision 1.80 1998/03/05 08:45:11 adam + * New result set model and modular ranking system. Moved towards + * descent server API. System information stored as "SGML" records. + * + * Revision 1.79 1998/02/17 10:32:52 adam + * Fixed bug: binary files weren't opened with flag b on NT. + * + * Revision 1.78 1998/02/10 12:03:05 adam + * Implemented Sort. + * + * Revision 1.77 1998/01/12 15:04:08 adam + * The test option (-s) only uses read-lock (and not write lock). + * + * Revision 1.76 1997/10/27 14:33:04 adam + * Moved towards generic character mapping depending on "structure" + * field in abstract syntax file. Fixed a few memory leaks. Fixed + * bug with negative integers when doing searches with relational + * operators. + * + * Revision 1.75 1997/09/17 12:19:12 adam + * Zebra version corresponds to YAZ version 1.4. + * Changed Zebra server so that it doesn't depend on global common_resource. + * + * Revision 1.74 1997/09/09 13:38:06 adam + * Partial port to WIN95/NT. + * + * Revision 1.73 1997/09/04 13:57:20 adam + * New file extract/retrieve method tellf (added). + * Added O_BINARY for open calls. + * + * Revision 1.72 1997/07/15 16:32:29 adam + * Bug fix: Match handler didn't terminate the resulting string! + * + * Revision 1.71 1997/07/15 16:28:41 adam + * Bug fix: storeData didn't work with files with multiple records. + * Bug fix: fixed memory management with records; not really well + * thought through. + * + * Revision 1.70 1997/07/01 13:00:42 adam + * Bug fix in routine searchRecordKey: uninitialized variables. + * + * Revision 1.69 1997/04/29 09:26:03 adam + * Bug fix: generic recordId handling didn't work for compressed internal + * keys. + * + * Revision 1.68 1997/02/12 20:39:45 adam + * Implemented options -f that limits the log to the first + * records. + * Changed some log messages also. + * + * Revision 1.67 1996/11/15 15:02:14 adam + * Minor changes regarding logging. + * + * Revision 1.66 1996/11/14 09:52:21 adam + * Strings in record keys bound by IT_MAX_WORD. + * + * Revision 1.65 1996/11/14 08:57:56 adam + * Reduction of storeKeys area. + * + * Revision 1.64 1996/11/08 11:10:16 adam + * Buffers used during file match got bigger. + * Compressed ISAM support everywhere. + * Bug fixes regarding masking characters in queries. + * Redesigned Regexp-2 queries. + * + * Revision 1.63 1996/10/29 14:09:39 adam + * Use of cisam system - enabled if setting isamc is 1. + * + * Revision 1.62 1996/10/11 10:57:01 adam + * New module recctrl. Used to manage records (extract/retrieval). + * Several files have been moved to the recctrl sub directory. + * + * Revision 1.61 1996/06/06 12:08:37 quinn + * Added showRecord function + * + * Revision 1.60 1996/06/04 10:18:12 adam + * Search/scan uses character mapping module. + * + * Revision 1.59 1996/05/14 15:47:07 adam + * 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. * @@ -129,19 +376,38 @@ */ #include #include +#ifdef WIN32 +#include +#else #include +#endif #include -#include -#include #include +#include +#include #include "index.h" -#include "recindex.h" +#include "zinfo.h" + +#if _FILE_OFFSET_BITS == 64 +#define PRINTF_OFF_T "%Ld" +#else +#define PRINTF_OFF_T "%ld" +#endif + +#ifndef ZEBRASDR +#define ZEBRASDR 0 +#endif + +#if ZEBRASDR +#include "zebrasdr.h" +#endif static Dict matchDict; static Records records = NULL; +static SortIdx sortIdx = NULL; static char **key_buf; static size_t ptr_top; @@ -152,39 +418,85 @@ 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; -#define MATCH_DICT "match" +static ZebraExplainInfo zti = NULL; -void key_open (int mem) + +static void logRecord (int showFlag) { + 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); + } +} + +static int explain_extract (void *handle, Record drec, data1_node *n); + +int key_open (struct recordGroup *rGroup, int mem) +{ + BFiles bfs = rGroup->bfs; + int rw = rGroup->flagRw; + data1_handle dh = rGroup->dh; + char *recordCompression; + int record_compression = REC_COMPRESS_NONE; + if (!mem) + mem = atoi(res_get_def (common_resource, "memMax", "16"))*1024*1024; if (mem < 50000) mem = 50000; - key_buf = xmalloc (mem); + key_buf = (char **) xmalloc (mem); ptr_top = mem/sizeof(char*); ptr_i = 0; key_buf_used = 0; key_file_no = 0; - if (!(matchDict = dict_open (MATCH_DICT, 20, 1))) + if (!(matchDict = dict_open (bfs, GMATCH_DICT, 50, rw, 0))) { - logf (LOG_FATAL, "dict_open fail of %s", MATCH_DICT); - exit (1); + logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT); + return -1; } assert (!records); - records = rec_open (1); + recordCompression = res_get_def (common_resource, + "recordCompression", "none"); + if (!strcmp (recordCompression, "none")) + record_compression = REC_COMPRESS_NONE; + if (!strcmp (recordCompression, "bzip2")) + record_compression = REC_COMPRESS_BZIP2; + records = rec_open (bfs, rw, record_compression); + if (!records) + { + dict_close (matchDict); + return -1; + } + zti = zebraExplain_open (records, dh, common_resource, + rw, rGroup, explain_extract); + if (!zti) + { + rec_close (&records); + dict_close (matchDict); + return -1; + } + sortIdx = sortIdx_open (bfs, 1); + return 0; } struct encode_info { int sysno; int seqno; - char buf[512]; + int cmd; + char buf[768]; }; void encode_key_init (struct encode_info *i) { i->sysno = 0; i->seqno = 0; + i->cmd = -1; } char *encode_key_int (int d, char *bp) @@ -226,8 +538,11 @@ void encode_key_write (char *k, struct encode_info *i, FILE *outf) i->sysno = key.sysno; i->seqno = 0; } + else if (!i->seqno && !key.seqno && i->cmd == *k) + return; bp = encode_key_int (key.seqno - i->seqno, bp); i->seqno = key.seqno; + i->cmd = *k; if (fwrite (i->buf, bp - i->buf, 1, outf) != 1) { logf (LOG_FATAL|LOG_ERRNO, "fwrite"); @@ -235,24 +550,49 @@ void encode_key_write (char *k, struct encode_info *i, FILE *outf) } } +#define SORT_EXTRA 0 + +#if SORT_EXTRA +static int key_y_len; + +static int key_y_compare (const void *p1, const void *p2) +{ + int r; + + if ((r = key_compare (*(char**) p1 + key_y_len + 1, + *(char**) p2 + key_y_len + 1))) + return r; + return *(*(char**) p1 + key_y_len) - *(*(char**) p2 + key_y_len); +} + +static int key_x_compare (const void *p1, const void *p2) +{ + return strcmp (*(char**) p1, *(char**) p2); +} +#endif + void key_flush (void) { FILE *outf; char out_fname[200]; char *prevcp, *cp; struct encode_info encode_info; +#if SORT_EXTRA + int i; +#endif if (ptr_i <= 0) return; key_file_no++; logf (LOG_LOG, "sorting section %d", key_file_no); +#if !SORT_EXTRA qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare); - sprintf (out_fname, TEMP_FNAME, key_file_no); + getFnameTmp (common_resource, out_fname, key_file_no); - if (!(outf = fopen (out_fname, "w"))) + if (!(outf = fopen (out_fname, "wb"))) { - logf (LOG_FATAL|LOG_ERRNO, "fopen (4) %s", out_fname); + logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); exit (1); } logf (LOG_LOG, "writing section %d", key_file_no); @@ -272,6 +612,42 @@ void key_flush (void) else encode_key_write (cp + strlen(cp), &encode_info, outf); } +#else + qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_x_compare); + getFnameTmp (out_fname, key_file_no); + + if (!(outf = fopen (out_fname, "wb"))) + { + logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); + exit (1); + } + logf (LOG_LOG, "writing section %d", key_file_no); + i = ptr_i; + prevcp = key_buf[ptr_top-i]; + while (1) + if (!--i || strcmp (prevcp, key_buf[ptr_top-i])) + { + key_y_len = strlen(prevcp)+1; +#if 0 + logf (LOG_LOG, "key_y_len: %2d %02x %02x %s", + key_y_len, prevcp[0], prevcp[1], 2+prevcp); +#endif + qsort (key_buf + ptr_top-ptr_i, ptr_i - i, + sizeof(char*), key_y_compare); + cp = key_buf[ptr_top-ptr_i]; + --key_y_len; + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + while (--ptr_i > i) + { + cp = key_buf[ptr_top-ptr_i]; + encode_key_write (cp+key_y_len, &encode_info, outf); + } + if (!i) + break; + prevcp = key_buf[ptr_top-ptr_i]; + } +#endif if (fclose (outf)) { logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname); @@ -282,107 +658,355 @@ void key_flush (void) key_buf_used = 0; } -int key_close (void) +int key_close (struct recordGroup *rGroup) { + int rw = rGroup->flagRw; + if (rw) + zebraExplain_runNumberIncrement (zti, 1); + zebraExplain_close (zti, rw); key_flush (); xfree (key_buf); rec_close (&records); dict_close (matchDict); + sortIdx_close (sortIdx); - logf (LOG_LOG, "Records inserted %6d", records_inserted); - logf (LOG_LOG, "Records updated %6d", records_updated); - logf (LOG_LOG, "Records deleted %6d", records_deleted); + logRecord (1); return key_file_no; } -static void wordInit (RecWord *p) +static void wordInit (struct recExtractCtrl *p, RecWord *w) { - p->attrSet = 1; - p->attrUse = 1016; - p->which = Word_String; + w->zebra_maps = p->zebra_maps; + w->seqnos = p->seqno; + w->attrSet = VAL_BIB1; + w->attrUse = 1016; + w->reg_type = 'w'; + w->extractCtrl = p; } -struct recKeys { +static struct sortKey { + char *string; + int length; + int attrSet; + int attrUse; + struct sortKey *next; +} *sortKeys = NULL; + +static struct recKeys { int buf_used; int buf_max; char *buf; + char prevAttrSet; + short prevAttrUse; + int prevSeqNo; } reckeys; -static void addRecordKey (const RecWord *p) +static void addIndexString (RecWord *p, const char *string, int length) { char *dst; - char attrSet; - short attrUse; - size_t i; + unsigned char attrSet; + unsigned short attrUse; + int lead = 0; + int diff = 0; + int *pseqno = &p->seqnos[p->reg_type]; if (reckeys.buf_used+1024 > reckeys.buf_max) { char *b; - b = xmalloc (reckeys.buf_max += 65000); + b = (char *) xmalloc (reckeys.buf_max += 128000); if (reckeys.buf_used > 0) memcpy (b, reckeys.buf, reckeys.buf_used); xfree (reckeys.buf); reckeys.buf = b; } dst = reckeys.buf + reckeys.buf_used; - switch (p->which) + + 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 + *pseqno - reckeys.prevSeqNo; + if (diff >= 1 && diff <= 15) + lead |= (diff << 2); + else + diff = 0; +#endif + reckeys.prevSeqNo = *pseqno; + + *dst++ = lead; + +#if SU_SCHEME + if ((lead & 3) < 3) + { + int ch = zebraExplain_lookupSU (zti, attrSet, attrUse); + if (ch < 0) + { + ch = zebraExplain_addSU (zti, attrSet, attrUse); + yaz_log (LOG_LOG, "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)) { - case Word_String: - attrSet = p->attrSet; memcpy (dst, &attrSet, sizeof(attrSet)); dst += sizeof(attrSet); - - attrUse = p->attrUse; + } + if (!(lead & 2)) + { memcpy (dst, &attrUse, sizeof(attrUse)); dst += sizeof(attrUse); - - for (i = 0; p->u.string[i]; i++) - *dst++ = p->u.string[i]; - *dst++ = '\0'; - - memcpy (dst, &p->seqno, sizeof(p->seqno)); - dst += sizeof(p->seqno); + } +#endif + *dst++ = p->reg_type; + memcpy (dst, string, length); + dst += length; + *dst++ = '\0'; - break; - default: - return; + if (!diff) + { + memcpy (dst, pseqno, sizeof(*pseqno)); + dst += sizeof(*pseqno); } reckeys.buf_used = dst - reckeys.buf; + if (*pseqno) + (*pseqno)++; +} + +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 = (struct sortKey *) xmalloc (sizeof(*sk)); + sk->next = sortKeys; + sortKeys = sk; + + sk->string = (char *) xmalloc (length); + sk->length = length; + memcpy (sk->string, string, length); + + sk->attrSet = p->attrSet; + sk->attrUse = p->attrUse; +} + +static void addString (RecWord *p, const char *string, int length) +{ + assert (length > 0); + if (zebra_maps_is_sort (p->zebra_maps, p->reg_type)) + addSortString (p, string, length); + else + addIndexString (p, string, length); +} + +static void addIncompleteField (RecWord *p) +{ + const char *b = p->string; + int remain = p->length; + const char **map = 0; + + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + + while (map) + { + char buf[IT_MAX_WORD+1]; + int i, remain; + + /* Skip spaces */ + while (map && *map && **map == *CHR_SPACE) + { + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + else + map = 0; + } + if (!map) + break; + i = 0; + while (map && *map && **map != *CHR_SPACE) + { + const char *cp = *map; + + while (i < IT_MAX_WORD && *cp) + buf[i++] = *(cp++); + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + else + map = 0; + } + if (!i) + return; + addString (p, buf, i); + } + (p->seqnos[p->reg_type])++; /* to separate this from next one */ +} + +static void addCompleteField (RecWord *p) +{ + const char *b = p->string; + char buf[IT_MAX_WORD+1]; + const char **map = 0; + int i = 0, remain = p->length; + + if (remain > 0) + map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain); + + while (remain > 0 && i < IT_MAX_WORD) + { + while (map && *map && **map == *CHR_SPACE) + { + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + else + map = 0; + } + if (!map) + break; + + if (i && i < IT_MAX_WORD) + buf[i++] = *CHR_SPACE; + while (map && *map && **map != *CHR_SPACE) + { + const char *cp = *map; + + if (i >= IT_MAX_WORD) + break; + while (i < IT_MAX_WORD && *cp) + buf[i++] = *(cp++); + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, + remain); + else + map = 0; + } + } + if (!i) + return; + addString (p, buf, i); } -static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, - const char *databaseName) +static void addRecordKey (RecWord *p) { + WRBUF wrbuf; + if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0, + p->string, p->length))) + { + p->string = wrbuf_buf(wrbuf); + p->length = wrbuf_len(wrbuf); + } + if (zebra_maps_is_complete (p->zebra_maps, p->reg_type)) + addCompleteField (p); + else + addIncompleteField(p); +} + +static void flushSortKeys (SYSNO sysno, int cmd) +{ + struct sortKey *sk = sortKeys; + + sortIdx_sysno (sortIdx, sysno); + while (sk) + { + struct sortKey *sk_next = sk->next; + sortIdx_type (sortIdx, sk->attrUse); + sortIdx_add (sortIdx, sk->string, sk->length); + xfree (sk->string); + xfree (sk); + sk = sk_next; + } + sortKeys = NULL; +} + +static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys) +{ +#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; + + zebraExplain_recordCountIncrement (zti, cmd ? 1 : -1); while (off < reckeys->buf_used) { const char *src = reckeys->buf + off; - char attrSet; - short attrUse; struct it_key key; - - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); - - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); - + 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 (key_buf_used + 1024 > (ptr_top-ptr_i)*sizeof(char*)) key_flush (); ++ptr_i; + key_buf[ptr_top-ptr_i] = (char*)key_buf + key_buf_used; - key_buf_used += index_word_prefix ((char*)key_buf + key_buf_used, - attrSet, attrUse, databaseName); + +#if SU_SCHEME +#else + ch = zebraExplain_lookupSU (zti, attrSet, attrUse); + if (ch < 0) + { + ch = zebraExplain_addSU (zti, attrSet, attrUse); + yaz_log (LOG_LOG, "addSU cmd=%d set=%d use=%d SU=%d", + cmd, attrSet, attrUse, ch); + } +#endif + assert (ch > 0); + key_buf_used += key_SU_code (ch, ((char*)key_buf) + key_buf_used); + while (*src) - ((char*)key_buf) [key_buf_used++] = index_char_cvt (*src++); + ((char*)key_buf) [key_buf_used++] = *src++; src++; ((char*)key_buf) [key_buf_used++] = '\0'; - ((char*) key_buf)[key_buf_used++] = cmd; - memcpy (&key.seqno, src, sizeof(key.seqno)); - src += sizeof(key.seqno); + if (lead & 60) + seqno += ((lead>>2) & 15)-1; + else + { + memcpy (&seqno, src, sizeof(seqno)); + src += sizeof(seqno); + } + key.seqno = seqno; key.sysno = sysno; memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key)); key_buf_used += sizeof(key); @@ -392,41 +1016,71 @@ static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys, } static const char **searchRecordKey (struct recKeys *reckeys, - int attrSetS, int attrUseS) + int attrSetS, int attrUseS) { static const char *ws[32]; int off = 0; int startSeq = -1; int i; + int seqno = 0; +#if SU_SCHEME + int chS, ch; +#else + short attrUse; + char attrSet; +#endif for (i = 0; i<32; i++) ws[i] = NULL; - + +#if SU_SCHEME + chS = zebraExplain_lookupSU (zti, attrSetS, attrUseS); + if (chS < 0) + return ws; +#endif while (off < reckeys->buf_used) { - const char *src = reckeys->buf + off; - char attrSet; - short attrUse; - int seqno; - const char *wstart; - - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); - - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); + const char *src = reckeys->buf + off; + const char *wstart; + int lead; + + lead = *src++; +#if SU_SCHEME + if ((lead & 3)<3) + { + memcpy (&ch, src, sizeof(ch)); + src += sizeof(ch); + } +#else + if (!(lead & 1)) + { + memcpy (&attrSet, src, sizeof(attrSet)); + src += sizeof(attrSet); + } + if (!(lead & 2)) + { + memcpy (&attrUse, src, sizeof(attrUse)); + src += sizeof(attrUse); + } +#endif wstart = src; while (*src++) ; - - memcpy (&seqno, src, sizeof(seqno)); - src += sizeof(seqno); - -#if 0 - logf (LOG_LOG, "(%d,%d) %d %s", attrSet, attrUse, seqno, wstart); + if (lead & 60) + seqno += ((lead>>2) & 15)-1; + else + { + memcpy (&seqno, src, sizeof(seqno)); + src += sizeof(seqno); + } + if ( +#if SU_SCHEME + ch == chS +#else + attrUseS == attrUse && attrSetS == attrSet #endif - if (attrUseS == attrUse && attrSetS == attrSet) + ) { int woff; @@ -444,110 +1098,99 @@ static const char **searchRecordKey (struct recKeys *reckeys, return ws; } -static void addRecordKeyAny (const RecWord *p) -{ - if (p->attrSet != 1 || p->attrUse != 1016) - { - RecWord w; +struct file_read_info { + off_t file_max; /* maximum offset so far */ + off_t file_offset; /* current offset */ + off_t file_moffset; /* offset of rec/rec boundary */ + int file_more; + int fd; + char *sdrbuf; + int sdrmax; +}; - memcpy (&w, p, sizeof(w)); - w.attrSet = 1; - w.attrUse = 1016; - addRecordKey (&w); - } - addRecordKey (p); +static struct file_read_info *file_read_start (int fd) +{ + struct file_read_info *fi = (struct file_read_info *) + xmalloc (sizeof(*fi)); + + fi->fd = fd; + fi->file_max = 0; + fi->file_moffset = 0; + fi->sdrbuf = 0; + fi->sdrmax = 0; + return fi; } -#define FILE_READ_BUFSIZE 4096 - -static char *file_buf; -static int file_offset; -static int file_bufsize; -static int file_noread; - -static void file_read_start (int fd) +static void file_read_stop (struct file_read_info *fi) { - file_offset = 0; - file_buf = xmalloc (FILE_READ_BUFSIZE); - file_bufsize = read (fd, file_buf, FILE_READ_BUFSIZE); - file_noread = 0; + xfree (fi); } -static void file_read_stop (int fd) +static off_t file_seek (void *handle, off_t offset) { - xfree (file_buf); - file_buf = NULL; + struct file_read_info *p = (struct file_read_info *) handle; + p->file_offset = offset; + if (p->sdrbuf) + return offset; + return lseek (p->fd, offset, SEEK_SET); } -static int file_read (int fd, char *buf, size_t count) +static off_t file_tell (void *handle) { - int l = file_bufsize - file_offset; + struct file_read_info *p = (struct file_read_info *) handle; + return p->file_offset; +} - if (count > l) +static int file_read (void *handle, char *buf, size_t count) +{ + struct file_read_info *p = (struct file_read_info *) handle; + int fd = p->fd; + int r; + if (p->sdrbuf) { - int r; - if (l > 0) - memcpy (buf, file_buf + file_offset, l); - count = count-l; - if (count > file_bufsize) - { - if ((r = read (fd, buf + l, count)) == -1) - { - logf (LOG_FATAL|LOG_ERRNO, "read"); - exit (1); - } - file_bufsize = 0; - file_offset = 0; - file_noread += r; - return r; - } - file_bufsize = r = read (fd, file_buf, FILE_READ_BUFSIZE); - if (r == -1) - { - logf (LOG_FATAL|LOG_ERRNO, "read"); - exit (1); - } - else if (r <= count) - { - file_offset = r; - memcpy (buf + l, file_buf, r); - file_noread += (l+r); - return l + r; - } - else - { - file_offset = count; - memcpy (buf + l, file_buf, count - l); - file_noread += count; - return count; - } + r = count; + if (r > p->sdrmax - p->file_offset) + r = p->sdrmax - p->file_offset; + if (r) + memcpy (buf, p->sdrbuf + p->file_offset, r); } - memcpy (buf, file_buf + file_offset, count); - file_offset += count; - file_noread += count; - return count; + else + 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; + } + return r; } -static int atois (const char **s) +static void file_begin (void *handle) { - int val = 0, c; - while ( (c=**s) >= '0' && c <= '9') - { - val = val*10 + c - '0'; - ++(*s); - } - return val; + struct file_read_info *p = (struct file_read_info *) handle; + + p->file_offset = p->file_moffset; + if (!p->sdrbuf && p->file_moffset) + lseek (p->fd, p->file_moffset, SEEK_SET); + p->file_more = 0; +} + +static void file_end (void *handle, off_t offset) +{ + struct file_read_info *p = (struct file_read_info *) handle; + + assert (p->file_more == 0); + p->file_more = 1; + p->file_moffset = offset; } static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, - const char *fname, - const char *spec) + const char *fname, const char *spec) { static char dstBuf[2048]; char *dst = dstBuf; const char *s = spec; static const char **w; - int i; while (1) { @@ -557,21 +1200,39 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, break; if (*s == '(') { + char attset_str[64], attname_str[64]; + data1_attset *attset; + int i; char matchFlag[32]; - int attrSet, attrUse; + int attSet = 1, attUse = 1; int first = 1; s++; - attrSet = atois (&s); - if (*s != ',') - { - logf (LOG_WARN, "Missing , in match criteria %s in group %s", - spec, rGroup->groupName ? rGroup->groupName : "none"); - return NULL; - } - s++; - attrUse = atois (&s); - w = searchRecordKey (reckeys, attrSet, attrUse); + for (i = 0; *s && *s != ',' && *s != ')'; s++) + if (i < 63) + attset_str[i++] = *s; + attset_str[i] = '\0'; + + if (*s == ',') + { + s++; + for (i = 0; *s && *s != ')'; s++) + if (i < 63) + attname_str[i++] = *s; + attname_str[i] = '\0'; + } + + if ((attset = data1_get_attset (rGroup->dh, attset_str))) + { + data1_att *att; + attSet = attset->reference; + att = data1_getattbyname(rGroup->dh, attset, attname_str); + if (att) + attUse = att->value; + else + attUse = atoi (attname_str); + } + w = searchRecordKey (reckeys, attSet, attUse); assert (w); if (*s == ')') @@ -600,8 +1261,8 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, } if (first) { - logf (LOG_WARN, "Record in file %s didn't contain match" - " fields in (%d,%d)", fname, attrSet, attrUse); + logf (LOG_WARN, "Record didn't contain match" + " fields in (%s,%s)", attset_str, attname_str); return NULL; } } @@ -668,37 +1329,116 @@ static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, fname, rGroup->groupName ? rGroup->groupName : "none"); return NULL; } + *dst = '\0'; return dstBuf; } +struct recordLogInfo { + const char *fname; + int recordOffset; + struct recordGroup *rGroup; +}; + +static void recordLogPreamble (int level, const char *msg, void *info) +{ + struct recordLogInfo *p = (struct recordLogInfo *) info; + FILE *outf = yaz_log_file (); + + if (level & LOG_LOG) + return ; + fprintf (outf, "File %s, offset %d, type %s\n", + p->rGroup->recordType, p->recordOffset, p->fname); + log_event_start (NULL, NULL); +} + +void addSchema (struct recExtractCtrl *p, Odr_oid *oid) +{ + zebraExplain_addSchema (zti, oid); +} + static int recordExtract (SYSNO *sysno, const char *fname, struct recordGroup *rGroup, int deleteFlag, - int fd, - RecType recType) + struct file_read_info *fi, + RecType recType, char *subType, void *clientData) { - struct recExtractCtrl extractCtrl; + RecordAttr *recordAttr; int r; char *matchStr; SYSNO sysnotmp; Record rec; + struct recordLogInfo logInfo; + off_t recordOffset = 0; - if (fd != -1) + if (fi->fd != -1) { - extractCtrl.fd = fd; - /* extract keys */ - extractCtrl.subType = ""; - extractCtrl.init = wordInit; - extractCtrl.add = addRecordKeyAny; - - reckeys.buf_used = 0; - extractCtrl.readf = file_read; - r = (*recType->extract)(&extractCtrl); - - if (r) - { - logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r); + struct recExtractCtrl extractCtrl; + + /* we are going to read from a file, so prepare the extraction */ + int i; + + reckeys.buf_used = 0; + reckeys.prevAttrUse = -1; + reckeys.prevAttrSet = -1; + reckeys.prevSeqNo = 0; + + recordOffset = fi->file_moffset; + extractCtrl.offset = fi->file_moffset; + extractCtrl.readf = file_read; + extractCtrl.seekf = file_seek; + extractCtrl.tellf = file_tell; + extractCtrl.endf = file_end; + extractCtrl.fh = fi; + extractCtrl.subType = subType; + extractCtrl.init = wordInit; + extractCtrl.tokenAdd = addRecordKey; + extractCtrl.schemaAdd = addSchema; + extractCtrl.dh = rGroup->dh; + for (i = 0; i<256; i++) + { + if (zebra_maps_is_positioned(rGroup->zebra_maps, i)) + extractCtrl.seqno[i] = 1; + else + extractCtrl.seqno[i] = 0; + } + extractCtrl.zebra_maps = rGroup->zebra_maps; + extractCtrl.flagShowRecords = !rGroup->flagRw; + + if (!rGroup->flagRw) + printf ("File: %s " PRINTF_OFF_T "\n", fname, recordOffset); + + logInfo.fname = fname; + logInfo.recordOffset = recordOffset; + logInfo.rGroup = rGroup; + log_event_start (recordLogPreamble, &logInfo); + + r = (*recType->extract)(clientData, &extractCtrl); + + log_event_start (NULL, NULL); + + if (r == RECCTRL_EXTRACT_EOF) + return 0; + else if (r == RECCTRL_EXTRACT_ERROR) + { + /* error occured during extraction ... */ + if (rGroup->flagRw && + records_processed < rGroup->fileVerboseLimit) + { + logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + } return 0; } + if (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, "empty %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + return 1; + } } /* perform match if sysno not known and if match criteria is specified */ @@ -722,89 +1462,129 @@ static int recordExtract (SYSNO *sysno, const char *fname, } else { - logf (LOG_WARN, "Record not inserted"); + logf (LOG_WARN, "Bad match criteria"); return 0; } } } - /* new record ? */ if (! *sysno) { + /* new record */ if (deleteFlag) { - logf (LOG_LOG, "? %s", fname); + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + logf (LOG_WARN, "cannot delete record above (seems new)"); return 1; } - logf (LOG_LOG, "add %s %s", rGroup->recordType, fname); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); rec = rec_new (records); + *sysno = rec->sysno; + recordAttr = rec_init_attr (zti, rec); + if (matchStr) { dict_insert (matchDict, matchStr, sizeof(*sysno), sysno); } - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); + flushRecordKeys (*sysno, 1, &reckeys); + flushSortKeys (*sysno, 1); records_inserted++; } else { + /* record already exists */ struct recKeys delkeys; rec = rec_get (records, *sysno); - + assert (rec); + + recordAttr = rec_init_attr (zti, rec); + + if (recordAttr->runNumber == zebraExplain_runNumberIncrement (zti, 0)) + { + logf (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + rec_rm (&rec); + logRecord (0); + return 1; + } delkeys.buf_used = rec->size[recInfo_delKeys]; delkeys.buf = rec->info[recInfo_delKeys]; - flushRecordKeys (*sysno, 0, &delkeys, rec->info[recInfo_databaseName]); + flushSortKeys (*sysno, 0); + flushRecordKeys (*sysno, 0, &delkeys); if (deleteFlag) { + /* record going to be deleted */ if (!delkeys.buf_used) { - logf (LOG_WARN, "cannot delete %s - no delete keys", fname); + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + logf (LOG_WARN, "cannot delete file above, storeKeys false"); } else { - SYSNO sysnoz = 0; - logf (LOG_LOG, "delete %s %s", rGroup->recordType, fname); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); records_deleted++; if (matchStr) - dict_insert (matchDict, matchStr, sizeof(sysnoz), &sysnoz); + dict_delete (matchDict, matchStr); rec_del (records, &rec); } + rec_rm (&rec); + logRecord (0); return 1; } else { + /* record going to be updated */ if (!delkeys.buf_used) { - logf (LOG_WARN, "cannot update %s - no delete keys", - fname); + logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + logf (LOG_WARN, "cannot update file above, storeKeys false"); } else { - logf (LOG_LOG, "update %s %s", rGroup->recordType, - fname); - flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); + flushRecordKeys (*sysno, 1, &reckeys); records_updated++; } } } + /* update file type */ xfree (rec->info[recInfo_fileType]); rec->info[recInfo_fileType] = rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]); + /* update filename */ xfree (rec->info[recInfo_filename]); rec->info[recInfo_filename] = rec_strdup (fname, &rec->size[recInfo_filename]); + /* update delete keys */ xfree (rec->info[recInfo_delKeys]); if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1) { +#if 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 } else { @@ -812,27 +1592,35 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec->size[recInfo_delKeys] = 0; } + /* save file size of original record */ + zebraExplain_recordBytesIncrement (zti, - recordAttr->recordSize); + recordAttr->recordSize = fi->file_moffset - recordOffset; + if (!recordAttr->recordSize) + recordAttr->recordSize = fi->file_max - recordOffset; + zebraExplain_recordBytesIncrement (zti, recordAttr->recordSize); + + /* set run-number for this record */ + recordAttr->runNumber = zebraExplain_runNumberIncrement (zti, 0); + + /* update store data */ xfree (rec->info[recInfo_storeData]); if (rGroup->flagStoreData == 1) { - rec->size[recInfo_storeData] = file_noread; - rec->info[recInfo_storeData] = xmalloc (file_noread); - if (file_noread < FILE_READ_BUFSIZE) - memcpy (rec->info[recInfo_storeData], file_buf, file_noread); - else + rec->size[recInfo_storeData] = recordAttr->recordSize; + rec->info[recInfo_storeData] = (char *) + xmalloc (recordAttr->recordSize); + if (lseek (fi->fd, recordOffset, SEEK_SET) < 0) { - if (lseek (fd, 0L, SEEK_SET) < 0) - { - logf (LOG_ERRNO|LOG_FATAL, "seek to 0 in %s", fname); - exit (1); - } - if (read (fd, rec->info[recInfo_storeData], file_noread) - < file_noread) - { - logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", - file_noread, fname); - exit (1); - } + logf (LOG_ERRNO|LOG_FATAL, "seek to " PRINTF_OFF_T " in %s", + recordOffset, fname); + exit (1); + } + if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize) + < recordAttr->recordSize) + { + logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", + recordAttr->recordSize, fname); + exit (1); } } else @@ -840,24 +1628,33 @@ static int recordExtract (SYSNO *sysno, const char *fname, rec->info[recInfo_storeData] = NULL; rec->size[recInfo_storeData] = 0; } + /* update database name */ xfree (rec->info[recInfo_databaseName]); rec->info[recInfo_databaseName] = rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]); + /* update offset */ + recordAttr->recordOffset = recordOffset; + + /* commit this record */ rec_put (records, &rec); + logRecord (0); return 1; } int fileExtract (SYSNO *sysno, const char *fname, const struct recordGroup *rGroupP, int deleteFlag) { - int i, fd; + int r, i, fd; char gprefix[128]; char ext[128]; char ext_res[128]; + char subType[128]; RecType recType; struct recordGroup rGroupM; struct recordGroup *rGroup = &rGroupM; + struct file_read_info *fi; + void *clientData; memcpy (rGroup, rGroupP, sizeof(*rGroupP)); @@ -869,12 +1666,10 @@ int fileExtract (SYSNO *sysno, const char *fname, logf (LOG_DEBUG, "fileExtract %s", fname); /* determine file extension */ + *ext = '\0'; for (i = strlen(fname); --i >= 0; ) if (fname[i] == '/') - { - strcpy (ext, ""); break; - } else if (fname[i] == '.') { strcpy (ext, fname+i+1); @@ -887,19 +1682,20 @@ int fileExtract (SYSNO *sysno, const char *fname, if (!(rGroup->recordType = res_get (common_resource, ext_res))) { sprintf (ext_res, "%srecordType", gprefix); - if (!(rGroup->recordType = res_get (common_resource, ext_res))) - { - logf (LOG_LOG, "? %s", fname); - return 0; - } + rGroup->recordType = res_get (common_resource, ext_res); } } if (!rGroup->recordType) { - logf (LOG_LOG, "? record %s", fname); + if (records_processed < rGroup->fileVerboseLimit) + logf (LOG_LOG, "? %s", fname); return 0; } - if (!(recType = recType_byName (rGroup->recordType))) + if (!*rGroup->recordType) + return 0; + if (!(recType = + recType_byName (rGroup->recTypes, rGroup->recordType, subType, + &clientData))) { logf (LOG_WARN, "No such record type: %s", rGroup->recordType); return 0; @@ -925,6 +1721,20 @@ int fileExtract (SYSNO *sysno, const char *fname, if (!rGroup->databaseName) rGroup->databaseName = "Default"; + /* determine if explain database */ + + sprintf (ext_res, "%sexplainDatabase", gprefix); + rGroup->explainDatabase = + atoi (res_get_def (common_resource, ext_res, "0")); + + /* announce database */ + if (zebraExplain_curDatabase (zti, rGroup->databaseName)) + { + if (zebraExplain_newDatabase (zti, rGroup->databaseName, + rGroup->explainDatabase)) + return 0; + } + if (rGroup->flagStoreData == -1) { const char *sval; @@ -960,17 +1770,72 @@ int fileExtract (SYSNO *sysno, const char *fname, fd = -1; else { - if ((fd = open (fname, O_RDONLY)) == -1) + if ((fd = open (fname, O_BINARY|O_RDONLY)) == -1) { logf (LOG_WARN|LOG_ERRNO, "open %s", fname); return 0; } } - file_read_start (fd); - recordExtract (sysno, fname, rGroup, deleteFlag, fd, recType); - file_read_stop (fd); + fi = file_read_start (fd); + do + { + file_begin (fi); + r = recordExtract (sysno, fname, rGroup, deleteFlag, fi, + recType, subType, clientData); + } while (r && !sysno && fi->file_more); + file_read_stop (fi); if (fd != -1) close (fd); - return 1; + return r; } +static int explain_extract (void *handle, Record rec, data1_node *n) +{ + struct recordGroup *rGroup = (struct recordGroup*) handle; + struct recExtractCtrl extractCtrl; + int i; + + if (zebraExplain_curDatabase (zti, rec->info[recInfo_databaseName])) + { + if (zebraExplain_newDatabase (zti, rec->info[recInfo_databaseName], 0)) + abort (); + } + + reckeys.buf_used = 0; + reckeys.prevAttrUse = -1; + reckeys.prevAttrSet = -1; + reckeys.prevSeqNo = 0; + + extractCtrl.init = wordInit; + extractCtrl.tokenAdd = addRecordKey; + extractCtrl.schemaAdd = addSchema; + extractCtrl.dh = rGroup->dh; + for (i = 0; i<256; i++) + extractCtrl.seqno[i] = 0; + extractCtrl.zebra_maps = rGroup->zebra_maps; + extractCtrl.flagShowRecords = !rGroup->flagRw; + + grs_extract_tree(&extractCtrl, n); + + logf (LOG_DEBUG, "flush explain record, sysno=%d", rec->sysno); + + if (rec->size[recInfo_delKeys]) + { + struct recKeys delkeys; + + delkeys.buf_used = rec->size[recInfo_delKeys]; + delkeys.buf = rec->info[recInfo_delKeys]; + flushSortKeys (rec->sysno, 0); + flushRecordKeys (rec->sysno, 0, &delkeys); + } + flushRecordKeys (rec->sysno, 1, &reckeys); + flushSortKeys (rec->sysno, 1); + + xfree (rec->info[recInfo_delKeys]); + rec->size[recInfo_delKeys] = reckeys.buf_used; + rec->info[recInfo_delKeys] = reckeys.buf; + reckeys.buf = NULL; + reckeys.buf_max = 0; + + return 0; +}