X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fextract.c;h=557c2f27451baad2ee99845bfa016997c2b40e73;hp=2f4bf846a6a72f8bebb39c6f3a1d049df9113de6;hb=b3890b03023cb9904ac1b10e2eee86669246d1dd;hpb=2b1851bd5565e3d21f9cf9a37661a584c063b75f diff --git a/index/extract.c b/index/extract.c index 2f4bf84..557c2f2 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,10 +1,26 @@ -/* - * Copyright (C) 1994-2002, Index Data - * All rights reserved. - * Sebastian Hammer, Adam Dickmeiss - * - * $Id: extract.c,v 1.114 2002-04-04 14:14:13 adam Exp $ - */ +/* $Id: extract.c,v 1.146 2003-10-07 09:18:43 adam Exp $ + Copyright (C) 1995,1996,1997,1998,1999,2000,2001,2002,2003 + Index Data Aps + +This file is part of the Zebra server. + +Zebra is free software; you can redistribute it and/or modify it under +the terms of the GNU General Public License as published by the Free +Software Foundation; either version 2, or (at your option) any later +version. + +Zebra is distributed in the hope that it will be useful, but WITHOUT ANY +WARRANTY; without even the implied warranty of MERCHANTABILITY or +FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License +for more details. + +You should have received a copy of the GNU General Public License +along with Zebra; see the file LICENSE.zebra. If not, write to the +Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA +02111-1307, USA. +*/ + + #include #include #ifdef WIN32 @@ -15,6 +31,7 @@ #include #include "index.h" +#include #include #if _FILE_OFFSET_BITS == 64 @@ -23,27 +40,48 @@ #define PRINTF_OFF_T "%ld" #endif -static int records_inserted = 0; -static int records_updated = 0; -static int records_deleted = 0; -static int records_processed = 0; +#define USE_SHELLSORT 0 -static 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)) + char *a = ar; + char v[100]; + int h, i, j, k; + static const int incs[16] = { 1391376, 463792, 198768, 86961, 33936, + 13776, 4592, 1968, 861, 336, + 112, 48, 21, 7, 3, 1 }; + for ( k = 0; k < 16; k++) + for (h = incs[k], i = h; i < r; i++) + { + memcpy (v, a+s*i, s); + j = i; + while (j > h && (*cmp)(a + s*(j-h), v) > 0) + { + memcpy (a + s*j, a + s*(j-h), s); + j -= h; + } + memcpy (a+s*j, v, s); + } +} +#endif + +static void logRecord (ZebraHandle zh) +{ + ++zh->records_processed; + if (!(zh->records_processed % 1000)) { logf (LOG_LOG, "Records: %7d i/u/d %d/%d/%d", - records_processed, records_inserted, records_updated, - records_deleted); + zh->records_processed, zh->records_inserted, zh->records_updated, + zh->records_deleted); } } static void extract_init (struct recExtractCtrl *p, RecWord *w) { w->zebra_maps = p->zebra_maps; - w->seqnos = p->seqno; + w->seqno = 1; w->attrSet = VAL_BIB1; w->attrUse = 1016; w->reg_type = 'w'; @@ -323,8 +361,9 @@ static char *fileMatchStr (ZebraHandle zh, spec_src = rGroup->groupName; else if (!strcmp (special, "database")) spec_src = rGroup->databaseName; - else if (!strcmp (special, "filename")) + else if (!strcmp (special, "filename")) { spec_src = fname; + } else if (!strcmp (special, "type")) spec_src = rGroup->recordType; else @@ -376,31 +415,18 @@ struct recordLogInfo { 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); -} - - static int recordExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, struct recordGroup *rGroup, int deleteFlag, struct file_read_info *fi, - RecType recType, char *subType, void *clientData) + RecType recType, char *subType, void *clientData, + int force_update) { RecordAttr *recordAttr; int r; char *matchStr; SYSNO sysnotmp; Record rec; - struct recordLogInfo logInfo; off_t recordOffset = 0; if (fi->fd != -1) @@ -414,7 +440,7 @@ static int recordExtract (ZebraHandle zh, zh->reg->keys.prevAttrUse = -1; zh->reg->keys.prevAttrSet = -1; zh->reg->keys.prevSeqNo = 0; - zh->reg->sortKeys = 0; + zh->reg->sortKeys.buf_used = 0; recordOffset = fi->file_moffset; extractCtrl.offset = fi->file_moffset; @@ -441,29 +467,41 @@ static int recordExtract (ZebraHandle zh, 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); + if (rGroup->flagRw) + { + char msg[512]; + sprintf (msg, "%s:" PRINTF_OFF_T , fname, recordOffset); + yaz_log_init_prefix2 (msg); + } r = (*recType->extract)(clientData, &extractCtrl); - log_event_start (NULL, NULL); - + yaz_log_init_prefix2 (0); if (r == RECCTRL_EXTRACT_EOF) return 0; - else if (r == RECCTRL_EXTRACT_ERROR) + else if (r == RECCTRL_EXTRACT_ERROR_GENERIC) { /* error occured during extraction ... */ if (rGroup->flagRw && - records_processed < rGroup->fileVerboseLimit) + zh->records_processed < rGroup->fileVerboseLimit) { logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, rGroup->recordType, fname, recordOffset); } return 0; } + else if (r == RECCTRL_EXTRACT_ERROR_NO_SUCH_FILTER) + { + /* error occured during extraction ... */ + if (rGroup->flagRw && + zh->records_processed < rGroup->fileVerboseLimit) + { + logf (LOG_WARN, "no filter for %s %s " + PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + } + return 0; + } if (zh->reg->keys.buf_used == 0) { /* the extraction process returned no information - the record @@ -514,7 +552,7 @@ static int recordExtract (ZebraHandle zh, logf (LOG_WARN, "cannot delete record above (seems new)"); return 1; } - if (records_processed < rGroup->fileVerboseLimit) + if (zh->records_processed < rGroup->fileVerboseLimit) logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, rGroup->recordType, fname, recordOffset); rec = rec_new (zh->reg->records); @@ -527,34 +565,40 @@ static int recordExtract (ZebraHandle zh, { dict_insert (zh->reg->matchDict, matchStr, sizeof(*sysno), sysno); } - extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); + 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 (zh->reg->records, *sysno); assert (rec); recordAttr = rec_init_attr (zh->reg->zei, rec); - if (recordAttr->runNumber == + if (!force_update && recordAttr->runNumber == zebraExplain_runNumberIncrement (zh->reg->zei, 0)) { - logf (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, rGroup->recordType, - fname, recordOffset); + yaz_log (LOG_LOG, "run number = %d", recordAttr->runNumber); + yaz_log (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, + rGroup->recordType, fname, recordOffset); extract_flushSortKeys (zh, *sysno, -1, &zh->reg->sortKeys); rec_rm (&rec); - logRecord (0); + logRecord (zh); return 1; } delkeys.buf_used = rec->size[recInfo_delKeys]; delkeys.buf = rec->info[recInfo_delKeys]; - extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys); + + 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) { @@ -567,16 +611,16 @@ static int recordExtract (ZebraHandle zh, } else { - if (records_processed < rGroup->fileVerboseLimit) + if (zh->records_processed < rGroup->fileVerboseLimit) logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, rGroup->recordType, fname, recordOffset); - records_deleted++; + zh->records_deleted++; if (matchStr) dict_delete (zh->reg->matchDict, matchStr); rec_del (zh->reg->records, &rec); } rec_rm (&rec); - logRecord (0); + logRecord (zh); return 1; } else @@ -590,11 +634,12 @@ static int recordExtract (ZebraHandle zh, } else { - if (records_processed < rGroup->fileVerboseLimit) + if (zh->records_processed < rGroup->fileVerboseLimit) logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, rGroup->recordType, fname, recordOffset); + extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); - records_updated++; + zh->records_updated++; } } } @@ -612,17 +657,10 @@ static int recordExtract (ZebraHandle zh, xfree (rec->info[recInfo_delKeys]); if (zh->reg->keys.buf_used > 0 && rGroup->flagStoreKeys == 1) { -#if 1 rec->size[recInfo_delKeys] = zh->reg->keys.buf_used; rec->info[recInfo_delKeys] = zh->reg->keys.buf; zh->reg->keys.buf = NULL; zh->reg->keys.buf_max = 0; -#else - rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used); - rec->size[recInfo_delKeys] = reckeys.buf_used; - memcpy (rec->info[recInfo_delKeys], reckeys.buf, - rec->size[recInfo_delKeys]); -#endif } else { @@ -630,6 +668,14 @@ static int recordExtract (ZebraHandle zh, 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); @@ -679,7 +725,7 @@ static int recordExtract (ZebraHandle zh, /* commit this record */ rec_put (zh->reg->records, &rec); - logRecord (0); + logRecord (zh); return 1; } @@ -728,7 +774,7 @@ int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, } if (!rGroup->recordType) { - if (records_processed < rGroup->fileVerboseLimit) + if (zh->records_processed < rGroup->fileVerboseLimit) logf (LOG_LOG, "? %s", fname); return 0; } @@ -814,9 +860,21 @@ int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, fd = -1; else { - if ((fd = open (fname, O_BINARY|O_RDONLY)) == -1) + char full_rep[1024]; + + if (zh->path_reg && !yaz_is_abspath (fname)) + { + strcpy (full_rep, zh->path_reg); + strcat (full_rep, "/"); + strcat (full_rep, fname); + } + else + strcpy (full_rep, fname); + + + if ((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); return 0; } } @@ -825,15 +883,13 @@ int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, { file_begin (fi); r = recordExtract (zh, sysno, fname, rGroup, deleteFlag, fi, - recType, subType, clientData); + recType, subType, clientData, 1); } while (r && !sysno && fi->file_more); file_read_stop (fi); if (fd != -1) close (fd); return r; } - - int extract_rec_in_mem (ZebraHandle zh, const char *recordType, const char *buf, size_t buf_size, const char *databaseName, int delete_flag, @@ -841,14 +897,56 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, int store_keys, int store_data, const char *match_criteria) { + struct recordGroup rGroup; + rGroup.groupName = NULL; + rGroup.databaseName = (char *)databaseName; + rGroup.path = NULL; + rGroup.recordId = NULL; + rGroup.recordType = (char *)recordType; + rGroup.flagStoreData = store_data; + rGroup.flagStoreKeys = store_keys; + rGroup.flagRw = 1; + rGroup.databaseNamePath = 0; + rGroup.explainDatabase = 0; + rGroup.fileVerboseLimit = 100000; + rGroup.followLinks = -1; + return (bufferExtractRecord (zh, + buf, buf_size, + &rGroup, + delete_flag, + test_mode, + recordType, + sysno, + match_criteria, + "", + 0,1)); +} +/* + 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 bufferExtractRecord (ZebraHandle zh, + const char *buf, size_t buf_size, + struct recordGroup *rGroup, + int delete_flag, + int test_mode, + const char *recordType, + int *sysno, + const char *match_criteria, + const char *fname, + int force_update, + int allow_update) + +{ RecordAttr *recordAttr; struct recExtractCtrl extractCtrl; int i, r; char *matchStr = 0; - RecType recType; + RecType recType = NULL; char subType[1024]; void *clientData; - const char *fname = ""; Record rec; long recordOffset = 0; struct zebra_fetch_control fc; @@ -867,26 +965,45 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, extractCtrl.endf = zebra_record_int_end; extractCtrl.fh = &fc; - /* announce database */ - if (zebraExplain_curDatabase (zh->reg->zei, databaseName)) - { - if (zebraExplain_newDatabase (zh->reg->zei, databaseName, 0)) - return 0; - } - if (!(recType = - recType_byName (zh->reg->recTypes, recordType, subType, - &clientData))) - { - logf (LOG_WARN, "No such record type: %s", recordType); - return 0; - } - zh->reg->keys.buf_used = 0; zh->reg->keys.prevAttrUse = -1; zh->reg->keys.prevAttrSet = -1; zh->reg->keys.prevSeqNo = 0; - zh->reg->sortKeys = 0; + zh->reg->sortKeys.buf_used = 0; + + /* announce database */ + if (!(rGroup->databaseName)) { + logf (LOG_WARN, "Invalid record group, no database name given"); + return 0; + } + + if (zebraExplain_curDatabase (zh->reg->zei, rGroup->databaseName)) + { + if (zebraExplain_newDatabase (zh->reg->zei, rGroup->databaseName, 0)) + return 0; + } + + if (*recordType) { + logf (LOG_DEBUG, "Record type explicitly specified: %s", recordType); + recType = recType_byName (zh->reg->recTypes, recordType, subType, + &clientData); + } else { + if (!(rGroup->recordType)) { + logf (LOG_WARN, "No such record type defined"); + return 0; + } + logf (LOG_DEBUG, "Get record type from rgroup: %s",rGroup->recordType); + recType = recType_byName (zh->reg->recTypes, rGroup->recordType, subType, + &clientData); + recordType = rGroup->recordType; + } + + if (!recType) { + logf (LOG_WARN, "No such record type: %s", rGroup->recordType); + return 0; + } + extractCtrl.subType = subType; extractCtrl.init = extract_init; extractCtrl.tokenAdd = extract_token_add; @@ -907,19 +1024,16 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, if (r == RECCTRL_EXTRACT_EOF) return 0; - else if (r == RECCTRL_EXTRACT_ERROR) + else if (r == RECCTRL_EXTRACT_ERROR_GENERIC) { /* error occured during extraction ... */ -#if 1 - yaz_log (LOG_WARN, "extract error"); -#else - if (rGroup->flagRw && - records_processed < rGroup->fileVerboseLimit) - { - logf (LOG_WARN, "fail %s %s %ld", rGroup->recordType, - fname, (long) recordOffset); - } -#endif + 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) @@ -933,6 +1047,27 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, return 1; } /* match criteria */ + matchStr = NULL; + + if (! *sysno && match_criteria) { + char *rinfo; + if (*match_criteria) { + matchStr = (char *)match_criteria; + } else { + if (rGroup->recordId && *rGroup->recordId) { + matchStr = fileMatchStr (zh, &zh->reg->keys, rGroup, fname, + rGroup->recordId); + } + } + if (matchStr) { + rinfo = dict_lookup (zh->reg->matchDict, matchStr); + if (rinfo) + memcpy (sysno, rinfo+1, sizeof(*sysno)); + } else { + logf (LOG_WARN, "Bad match criteria (recordID)"); + return 0; + } + } if (! *sysno) { @@ -957,30 +1092,49 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, dict_insert (zh->reg->matchDict, matchStr, sizeof(*sysno), sysno); } - extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); - } + 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, 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 (recordAttr->runNumber == - zebraExplain_runNumberIncrement (zh->reg->zei, 0)) - { - logf (LOG_LOG, "skipped %s %s %ld", recordType, - fname, (long) recordOffset); - rec_rm (&rec); - return 1; + if (!force_update) { + if (recordAttr->runNumber == + zebraExplain_runNumberIncrement (zh->reg->zei, 0)) + { + logf (LOG_LOG, "skipped %s %s %ld", recordType, + 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]; - extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys); + + 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) { @@ -995,13 +1149,13 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, { logf (LOG_LOG, "delete %s %s %ld", recordType, fname, (long) recordOffset); -#if 0 + zh->records_deleted++; if (matchStr) - dict_delete (matchDict, matchStr); -#endif + dict_delete (zh->reg->matchDict, matchStr); rec_del (zh->reg->records, &rec); } rec_rm (&rec); + logRecord(zh); return 1; } else @@ -1017,7 +1171,9 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, { logf (LOG_LOG, "update %s %s %ld", recordType, fname, (long) recordOffset); + extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys); extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys); + zh->records_updated++; } } } @@ -1033,7 +1189,7 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, /* update delete keys */ xfree (rec->info[recInfo_delKeys]); - if (zh->reg->keys.buf_used > 0 && store_keys == 1) + if (zh->reg->keys.buf_used > 0 && rGroup->flagStoreKeys == 1) { rec->size[recInfo_delKeys] = zh->reg->keys.buf_used; rec->info[recInfo_delKeys] = zh->reg->keys.buf; @@ -1046,6 +1202,14 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, 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); @@ -1065,7 +1229,7 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, /* update store data */ xfree (rec->info[recInfo_storeData]); - if (store_data == 1) + if (rGroup->flagStoreData == 1) { rec->size[recInfo_storeData] = recordAttr->recordSize; rec->info[recInfo_storeData] = (char *) @@ -1096,17 +1260,19 @@ int extract_rec_in_mem (ZebraHandle zh, const char *recordType, /* update database name */ xfree (rec->info[recInfo_databaseName]); rec->info[recInfo_databaseName] = - rec_strdup (databaseName, &rec->size[recInfo_databaseName]); + rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]); /* update offset */ recordAttr->recordOffset = recordOffset; /* commit this record */ rec_put (zh->reg->records, &rec); - + logRecord(zh); return 0; } + + int explain_extract (void *handle, Record rec, data1_node *n) { ZebraHandle zh = (ZebraHandle) handle; @@ -1126,7 +1292,7 @@ int explain_extract (void *handle, Record rec, data1_node *n) zh->reg->keys.prevAttrUse = -1; zh->reg->keys.prevAttrSet = -1; zh->reg->keys.prevSeqNo = 0; - zh->reg->sortKeys = 0; + zh->reg->sortKeys.buf_used = 0; extractCtrl.init = extract_init; extractCtrl.tokenAdd = extract_token_add; @@ -1137,17 +1303,22 @@ int explain_extract (void *handle, Record rec, data1_node *n) extractCtrl.zebra_maps = zh->reg->zebra_maps; extractCtrl.flagShowRecords = 0; extractCtrl.handle = handle; - - grs_extract_tree(&extractCtrl, n); + + if (n) + grs_extract_tree(&extractCtrl, n); if (rec->size[recInfo_delKeys]) { struct recKeys delkeys; - struct sortKey *sortKeys = 0; + struct sortKeys sortkeys; delkeys.buf_used = rec->size[recInfo_delKeys]; delkeys.buf = rec->info[recInfo_delKeys]; - extract_flushSortKeys (zh, rec->sysno, 0, &sortKeys); + + 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); @@ -1158,6 +1329,13 @@ int explain_extract (void *handle, Record rec, data1_node *n) 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; } @@ -1176,7 +1354,14 @@ void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno, if (!zh->reg->key_buf) { - int mem = 8*1024*1024; + 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; @@ -1265,8 +1450,8 @@ void extract_flushWriteKeys (ZebraHandle zh) (zh->reg->key_file_no)++; logf (LOG_LOG, "sorting section %d", (zh->reg->key_file_no)); #if !SORT_EXTRA - qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i, sizeof(char*), - key_qsort_compare); + qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i, + sizeof(char*), key_qsort_compare); extract_get_fname_tmp (zh, out_fname, zh->reg->key_file_no); if (!(outf = fopen (out_fname, "wb"))) @@ -1285,6 +1470,7 @@ void extract_flushWriteKeys (ZebraHandle zh) 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; @@ -1292,6 +1478,7 @@ void extract_flushWriteKeys (ZebraHandle zh) 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); @@ -1323,6 +1510,7 @@ void extract_flushWriteKeys (ZebraHandle zh) 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]; @@ -1346,11 +1534,11 @@ void extract_add_index_string (RecWord *p, const char *string, unsigned short attrUse; int lead = 0; int diff = 0; - int *pseqno = &p->seqnos[p->reg_type]; + int *pseqno = &p->seqno; ZebraHandle zh = p->extractCtrl->handle; ZebraExplainInfo zei = zh->reg->zei; struct recKeys *keys = &zh->reg->keys; - + if (keys->buf_used+1024 > keys->buf_max) { char *b; @@ -1421,30 +1609,43 @@ void extract_add_index_string (RecWord *p, const char *string, dst += sizeof(*pseqno); } keys->buf_used = dst - keys->buf; - if (*pseqno) - (*pseqno)++; } static void extract_add_sort_string (RecWord *p, const char *string, int length) { - struct sortKey *sk; ZebraHandle zh = p->extractCtrl->handle; + struct sortKeys *sk = &zh->reg->sortKeys; + int off = 0; - for (sk = zh->reg->sortKeys; sk; sk = sk->next) - if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse) - return; - - sk = (struct sortKey *) xmalloc (sizeof(*sk)); - sk->next = zh->reg->sortKeys; - zh->reg->sortKeys = sk; - - sk->string = (char *) xmalloc (length); - sk->length = length; - memcpy (sk->string, string, length); - - sk->attrSet = p->attrSet; - sk->attrUse = p->attrUse; + 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, string, length); + sk->buf_used = off + length; } void extract_add_string (RecWord *p, const char *string, int length) @@ -1497,8 +1698,8 @@ static void extract_add_incomplete_field (RecWord *p) if (!i) return; extract_add_string (p, buf, i); + p->seqno++; } - (p->seqnos[p->reg_type])++; /* to separate this from next one */ } static void extract_add_complete_field (RecWord *p) @@ -1550,6 +1751,12 @@ static void extract_add_complete_field (RecWord *p) 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))) { @@ -1569,22 +1776,28 @@ void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid) } void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno, - int cmd, struct sortKey **skp) + int cmd, struct sortKeys *sk) { - struct sortKey *sk = *skp; SortIdx sortIdx = zh->reg->sortIdx; + int off = 0; sortIdx_sysno (sortIdx, sysno); - while (sk) + + while (off < sk->buf_used) { - 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; + 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; } - *skp = 0; } void encode_key_init (struct encode_info *i) @@ -1592,6 +1805,10 @@ 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; } char *encode_key_int (int d, char *bp) @@ -1618,7 +1835,13 @@ char *encode_key_int (int d, char *bp) } return bp; } +#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; @@ -1645,3 +1868,108 @@ void encode_key_write (char *k, struct encode_info *i, FILE *outf) } } +void encode_key_flush (struct encode_info *i, FILE *outf) +{ /* dummy routine */ +} + +#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