/*
- * Copyright (C) 1994-1995, Index Data I/S
+ * Copyright (C) 1994-2002, Index Data
* All rights reserved.
* Sebastian Hammer, Adam Dickmeiss
*
- * $Log: extract.c,v $
- * Revision 1.27 1995-11-20 16:59:45 adam
- * New update method: the 'old' keys are saved for each records.
- *
- * Revision 1.26 1995/11/20 11:56:24 adam
- * Work on new traversal.
- *
- * Revision 1.25 1995/11/16 15:34:54 adam
- * Uses new record management system in both indexer and server.
- *
- * Revision 1.24 1995/11/15 19:13:08 adam
- * Work on record management.
- *
- * Revision 1.23 1995/10/27 14:00:10 adam
- * Implemented detection of database availability.
- *
- * Revision 1.22 1995/10/17 18:02:07 adam
- * New feature: databases. Implemented as prefix to words in dictionary.
- *
- * Revision 1.21 1995/10/10 12:24:38 adam
- * Temporary sort files are compressed.
- *
- * Revision 1.20 1995/10/06 13:52:05 adam
- * Bug fixes. Handler may abort further scanning.
- *
- * Revision 1.19 1995/10/04 12:55:16 adam
- * Bug fix in ranked search. Use=Any keys inserted.
- *
- * Revision 1.18 1995/10/04 09:37:08 quinn
- * Fixed bug.
- *
- * Revision 1.17 1995/10/03 14:28:57 adam
- * Buffered read in extract works.
- *
- * Revision 1.16 1995/10/03 14:28:45 adam
- * Work on more effecient read handler in extract.
- *
- * Revision 1.15 1995/10/02 15:42:53 adam
- * Extract uses file descriptors instead of FILE pointers.
- *
- * Revision 1.14 1995/10/02 15:29:13 adam
- * More logging in file_extract.
- *
- * Revision 1.13 1995/09/29 14:01:39 adam
- * Bug fixes.
- *
- * Revision 1.12 1995/09/28 14:22:56 adam
- * Sort uses smaller temporary files.
- *
- * Revision 1.11 1995/09/28 12:10:31 adam
- * Bug fixes. Field prefix used in queries.
- *
- * Revision 1.10 1995/09/28 09:19:41 adam
- * xfree/xmalloc used everywhere.
- * Extract/retrieve method seems to work for text records.
- *
- * Revision 1.9 1995/09/27 12:22:28 adam
- * More work on extract in record control.
- * Field name is not in isam keys but in prefix in dictionary words.
- *
- * Revision 1.8 1995/09/14 07:48:22 adam
- * Record control management.
- *
- * Revision 1.7 1995/09/11 13:09:32 adam
- * More work on relevance feedback.
- *
- * Revision 1.6 1995/09/08 14:52:27 adam
- * Minor changes. Dictionary is lower case now.
- *
- * Revision 1.5 1995/09/06 16:11:16 adam
- * Option: only one word key per file.
- *
- * Revision 1.4 1995/09/05 15:28:39 adam
- * More work on search engine.
- *
- * Revision 1.3 1995/09/04 12:33:41 adam
- * Various cleanup. YAZ util used instead.
- *
- * Revision 1.2 1995/09/04 09:10:34 adam
- * More work on index add/del/update.
- * Merge sort implemented.
- * Initial work on z39 server.
- *
- * Revision 1.1 1995/09/01 14:06:35 adam
- * Split of work into more files.
- *
+ * $Id: extract.c,v 1.114 2002-04-04 14:14:13 adam Exp $
*/
#include <stdio.h>
#include <assert.h>
+#ifdef WIN32
+#include <io.h>
+#else
#include <unistd.h>
+#endif
#include <fcntl.h>
-#include <ctype.h>
-#include <alexutil.h>
-#include <recctrl.h>
#include "index.h"
+#include <charmap.h>
+
+#if _FILE_OFFSET_BITS == 64
+#define PRINTF_OFF_T "%Ld"
+#else
+#define PRINTF_OFF_T "%ld"
+#endif
+
+static int records_inserted = 0;
+static int records_updated = 0;
+static int records_deleted = 0;
+static int records_processed = 0;
-#if RECORD_BASE
-#include "recindex.h"
+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 void extract_init (struct recExtractCtrl *p, RecWord *w)
+{
+ w->zebra_maps = p->zebra_maps;
+ w->seqnos = p->seqno;
+ w->attrSet = VAL_BIB1;
+ w->attrUse = 1016;
+ w->reg_type = 'w';
+ w->extractCtrl = p;
+}
+
+static const char **searchRecordKey (ZebraHandle zh,
+ struct recKeys *reckeys,
+ int attrSetS, int attrUseS)
+{
+ static const char *ws[32];
+ int off = 0;
+ int startSeq = -1;
+ int i;
+ int seqno = 0;
+#if SU_SCHEME
+ int chS, ch;
+#else
+ short attrUse;
+ char attrSet;
#endif
-static Dict file_idx;
+ for (i = 0; i<32; i++)
+ ws[i] = NULL;
+#if SU_SCHEME
+ chS = zebraExplain_lookupSU (zh->reg->zei, attrSetS, attrUseS);
+ if (chS < 0)
+ return ws;
+#endif
+ while (off < reckeys->buf_used)
+ {
-#if RECORD_BASE
-static Records records = NULL;
+ const char *src = reckeys->buf + off;
+ const char *wstart;
+ int lead;
+
+ lead = *src++;
+#if SU_SCHEME
+ if ((lead & 3)<3)
+ {
+ memcpy (&ch, src, sizeof(ch));
+ src += sizeof(ch);
+ }
+#else
+ if (!(lead & 1))
+ {
+ memcpy (&attrSet, src, sizeof(attrSet));
+ src += sizeof(attrSet);
+ }
+ if (!(lead & 2))
+ {
+ memcpy (&attrUse, src, sizeof(attrUse));
+ src += sizeof(attrUse);
+ }
+#endif
+ wstart = src;
+ while (*src++)
+ ;
+ if (lead & 60)
+ seqno += ((lead>>2) & 15)-1;
+ else
+ {
+ memcpy (&seqno, src, sizeof(seqno));
+ src += sizeof(seqno);
+ }
+ if (
+#if SU_SCHEME
+ ch == chS
#else
-static int sys_idx_fd = -1;
-static SYSNO sysno_next;
+ attrUseS == attrUse && attrSetS == attrSet
#endif
+ )
+ {
+ int woff;
+
-static int key_cmd;
-static int key_sysno;
-static const char *key_databaseName;
-static char **key_buf;
-static size_t ptr_top;
-static size_t ptr_i;
-static size_t key_buf_used;
-static int key_file_no;
+ if (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; /* 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)
+{
+ 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;
+}
+
+static void file_read_stop (struct file_read_info *fi)
+{
+ xfree (fi);
+}
+
+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 off_t file_tell (void *handle)
+{
+ struct file_read_info *p = (struct file_read_info *) handle;
+ return p->file_offset;
+}
+
+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)
+ {
+ 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);
+ }
+ 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 void file_begin (void *handle)
+{
+ struct file_read_info *p = (struct file_read_info *) handle;
-static int key_del_max;
-static int key_del_used;
-static char *key_del_buf;
+ 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 (ZebraHandle zh,
+ struct recKeys *reckeys, struct recordGroup *rGroup,
+ const char *fname, const char *spec)
+{
+ static char dstBuf[2048]; /* static here ??? */
+ char *dst = dstBuf;
+ const char *s = spec;
+ static const char **w;
+
+ while (1)
+ {
+ while (*s == ' ' || *s == '\t')
+ s++;
+ if (!*s)
+ break;
+ if (*s == '(')
+ {
+ char attset_str[64], attname_str[64];
+ data1_attset *attset;
+ int i;
+ char matchFlag[32];
+ int attSet = 1, attUse = 1;
+ int first = 1;
+
+ s++;
+ for (i = 0; *s && *s != ',' && *s != ')'; s++)
+ if (i < 63)
+ attset_str[i++] = *s;
+ attset_str[i] = '\0';
+
+ if (*s == ',')
+ {
+ s++;
+ for (i = 0; *s && *s != ')'; s++)
+ if (i < 63)
+ attname_str[i++] = *s;
+ attname_str[i] = '\0';
+ }
+
+ if ((attset = data1_get_attset (zh->reg->dh, attset_str)))
+ {
+ data1_att *att;
+ attSet = attset->reference;
+ att = data1_getattbyname(zh->reg->dh, attset, attname_str);
+ if (att)
+ attUse = att->value;
+ else
+ attUse = atoi (attname_str);
+ }
+ w = searchRecordKey (zh, reckeys, attSet, attUse);
+ assert (w);
+
+ if (*s == ')')
+ {
+ for (i = 0; i<32; i++)
+ matchFlag[i] = 1;
+ }
+ else
+ {
+ logf (LOG_WARN, "Missing ) in match criteria %s in group %s",
+ spec, rGroup->groupName ? rGroup->groupName : "none");
+ return NULL;
+ }
+ s++;
+
+ for (i = 0; i<32; i++)
+ if (matchFlag[i] && w[i])
+ {
+ if (first)
+ {
+ *dst++ = ' ';
+ first = 0;
+ }
+ strcpy (dst, w[i]);
+ dst += strlen(w[i]);
+ }
+ if (first)
+ {
+ logf (LOG_WARN, "Record didn't contain match"
+ " fields in (%s,%s)", attset_str, attname_str);
+ return NULL;
+ }
+ }
+ else if (*s == '$')
+ {
+ int spec_len;
+ char special[64];
+ const char *spec_src = NULL;
+ const char *s1 = ++s;
+ while (*s1 && *s1 != ' ' && *s1 != '\t')
+ s1++;
+
+ spec_len = s1 - s;
+ if (spec_len > 63)
+ spec_len = 63;
+ memcpy (special, s, spec_len);
+ special[spec_len] = '\0';
+ s = s1;
+
+ if (!strcmp (special, "group"))
+ spec_src = rGroup->groupName;
+ else if (!strcmp (special, "database"))
+ spec_src = rGroup->databaseName;
+ else if (!strcmp (special, "filename"))
+ spec_src = fname;
+ else if (!strcmp (special, "type"))
+ spec_src = rGroup->recordType;
+ else
+ spec_src = NULL;
+ if (spec_src)
+ {
+ strcpy (dst, spec_src);
+ dst += strlen (spec_src);
+ }
+ }
+ else if (*s == '\"' || *s == '\'')
+ {
+ int stopMarker = *s++;
+ char tmpString[64];
+ int i = 0;
+
+ while (*s && *s != stopMarker)
+ {
+ if (i < 63)
+ tmpString[i++] = *s++;
+ }
+ if (*s)
+ s++;
+ tmpString[i] = '\0';
+ strcpy (dst, tmpString);
+ dst += strlen (tmpString);
+ }
+ else
+ {
+ logf (LOG_WARN, "Syntax error in match criteria %s in group %s",
+ spec, rGroup->groupName ? rGroup->groupName : "none");
+ return NULL;
+ }
+ *dst++ = 1;
+ }
+ if (dst == dstBuf)
+ {
+ logf (LOG_WARN, "No match criteria for record %s in group %s",
+ fname, rGroup->groupName ? rGroup->groupName : "none");
+ return NULL;
+ }
+ *dst = '\0';
+ return dstBuf;
+}
+
+struct recordLogInfo {
+ const char *fname;
+ int recordOffset;
+ struct recordGroup *rGroup;
+};
+
+static 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)
+{
+ RecordAttr *recordAttr;
+ int r;
+ char *matchStr;
+ SYSNO sysnotmp;
+ Record rec;
+ struct recordLogInfo logInfo;
+ off_t recordOffset = 0;
+
+ if (fi->fd != -1)
+ {
+ struct recExtractCtrl extractCtrl;
+
+ /* we are going to read from a file, so prepare the extraction */
+ int i;
+
+ zh->reg->keys.buf_used = 0;
+ zh->reg->keys.prevAttrUse = -1;
+ zh->reg->keys.prevAttrSet = -1;
+ zh->reg->keys.prevSeqNo = 0;
+ zh->reg->sortKeys = 0;
+
+ recordOffset = fi->file_moffset;
+ extractCtrl.offset = fi->file_moffset;
+ extractCtrl.readf = file_read;
+ extractCtrl.seekf = file_seek;
+ extractCtrl.tellf = file_tell;
+ extractCtrl.endf = file_end;
+ extractCtrl.fh = fi;
+ extractCtrl.subType = subType;
+ extractCtrl.init = extract_init;
+ extractCtrl.tokenAdd = extract_token_add;
+ extractCtrl.schemaAdd = extract_schema_add;
+ extractCtrl.dh = zh->reg->dh;
+ extractCtrl.handle = zh;
+ for (i = 0; i<256; i++)
+ {
+ if (zebra_maps_is_positioned(zh->reg->zebra_maps, i))
+ extractCtrl.seqno[i] = 1;
+ else
+ extractCtrl.seqno[i] = 0;
+ }
+ extractCtrl.zebra_maps = zh->reg->zebra_maps;
+ extractCtrl.flagShowRecords = !rGroup->flagRw;
+
+ if (!rGroup->flagRw)
+ printf ("File: %s " PRINTF_OFF_T "\n", fname, recordOffset);
+
+ logInfo.fname = fname;
+ logInfo.recordOffset = recordOffset;
+ logInfo.rGroup = rGroup;
+ log_event_start (recordLogPreamble, &logInfo);
+
+ r = (*recType->extract)(clientData, &extractCtrl);
+
+ log_event_start (NULL, NULL);
+
+ if (r == RECCTRL_EXTRACT_EOF)
+ return 0;
+ else if (r == RECCTRL_EXTRACT_ERROR)
+ {
+ /* error occured during extraction ... */
+ if (rGroup->flagRw &&
+ records_processed < rGroup->fileVerboseLimit)
+ {
+ logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, rGroup->recordType,
+ fname, recordOffset);
+ }
+ return 0;
+ }
+ if (zh->reg->keys.buf_used == 0)
+ {
+ /* the extraction process returned no information - the record
+ is probably empty - unless flagShowRecords is in use */
+ if (!rGroup->flagRw)
+ return 1;
+
+ logf (LOG_WARN, "empty %s %s " PRINTF_OFF_T, rGroup->recordType,
+ fname, recordOffset);
+ return 1;
+ }
+ }
+
+ /* perform match if sysno not known and if match criteria is specified */
+
+ matchStr = NULL;
+ if (!sysno)
+ {
+ sysnotmp = 0;
+ sysno = &sysnotmp;
+ if (rGroup->recordId && *rGroup->recordId)
+ {
+ char *rinfo;
+
+ matchStr = fileMatchStr (zh, &zh->reg->keys, rGroup, fname,
+ rGroup->recordId);
+ if (matchStr)
+ {
+ rinfo = dict_lookup (zh->reg->matchDict, matchStr);
+ if (rinfo)
+ memcpy (sysno, rinfo+1, sizeof(*sysno));
+ }
+ else
+ {
+ logf (LOG_WARN, "Bad match criteria");
+ return 0;
+ }
+ }
+ }
+
+ if (! *sysno)
+ {
+ /* new record */
+ if (deleteFlag)
+ {
+ logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, rGroup->recordType,
+ fname, recordOffset);
+ logf (LOG_WARN, "cannot delete record above (seems new)");
+ return 1;
+ }
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, rGroup->recordType,
+ fname, recordOffset);
+ rec = rec_new (zh->reg->records);
+
+ *sysno = rec->sysno;
+
+ recordAttr = rec_init_attr (zh->reg->zei, rec);
+
+ if (matchStr)
+ {
+ dict_insert (zh->reg->matchDict, matchStr, sizeof(*sysno), sysno);
+ }
+ extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
+ extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys);
+
+ records_inserted++;
+ }
+ else
+ {
+ /* record already exists */
+ struct recKeys delkeys;
+
+ rec = rec_get (zh->reg->records, *sysno);
+ assert (rec);
+
+ recordAttr = rec_init_attr (zh->reg->zei, rec);
+
+ if (recordAttr->runNumber ==
+ zebraExplain_runNumberIncrement (zh->reg->zei, 0))
+ {
+ logf (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, rGroup->recordType,
+ fname, recordOffset);
+ extract_flushSortKeys (zh, *sysno, -1, &zh->reg->sortKeys);
+ rec_rm (&rec);
+ logRecord (0);
+ return 1;
+ }
+ delkeys.buf_used = rec->size[recInfo_delKeys];
+ delkeys.buf = rec->info[recInfo_delKeys];
+ extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys);
+ extract_flushRecordKeys (zh, *sysno, 0, &delkeys);
+ if (deleteFlag)
+ {
+ /* record going to be deleted */
+ if (!delkeys.buf_used)
+ {
+ logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T,
+ rGroup->recordType, fname, recordOffset);
+ logf (LOG_WARN, "cannot delete file above, storeKeys false");
+ }
+ else
+ {
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T,
+ rGroup->recordType, fname, recordOffset);
+ records_deleted++;
+ if (matchStr)
+ dict_delete (zh->reg->matchDict, matchStr);
+ rec_del (zh->reg->records, &rec);
+ }
+ rec_rm (&rec);
+ logRecord (0);
+ return 1;
+ }
+ else
+ {
+ /* record going to be updated */
+ if (!delkeys.buf_used)
+ {
+ logf (LOG_LOG, "update %s %s " PRINTF_OFF_T,
+ rGroup->recordType, fname, recordOffset);
+ logf (LOG_WARN, "cannot update file above, storeKeys false");
+ }
+ else
+ {
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "update %s %s " PRINTF_OFF_T,
+ rGroup->recordType, fname, recordOffset);
+ extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
+ records_updated++;
+ }
+ }
+ }
+ /* update file type */
+ xfree (rec->info[recInfo_fileType]);
+ rec->info[recInfo_fileType] =
+ rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]);
+
+ /* update filename */
+ xfree (rec->info[recInfo_filename]);
+ rec->info[recInfo_filename] =
+ rec_strdup (fname, &rec->size[recInfo_filename]);
+
+ /* update delete keys */
+ xfree (rec->info[recInfo_delKeys]);
+ if (zh->reg->keys.buf_used > 0 && rGroup->flagStoreKeys == 1)
+ {
+#if 1
+ rec->size[recInfo_delKeys] = zh->reg->keys.buf_used;
+ rec->info[recInfo_delKeys] = zh->reg->keys.buf;
+ zh->reg->keys.buf = NULL;
+ zh->reg->keys.buf_max = 0;
+#else
+ rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used);
+ rec->size[recInfo_delKeys] = reckeys.buf_used;
+ memcpy (rec->info[recInfo_delKeys], reckeys.buf,
+ rec->size[recInfo_delKeys]);
+#endif
+ }
+ else
+ {
+ rec->info[recInfo_delKeys] = NULL;
+ rec->size[recInfo_delKeys] = 0;
+ }
+
+ /* save file size of original record */
+ zebraExplain_recordBytesIncrement (zh->reg->zei,
+ - recordAttr->recordSize);
+ recordAttr->recordSize = fi->file_moffset - recordOffset;
+ if (!recordAttr->recordSize)
+ recordAttr->recordSize = fi->file_max - recordOffset;
+ zebraExplain_recordBytesIncrement (zh->reg->zei,
+ recordAttr->recordSize);
+
+ /* set run-number for this record */
+ recordAttr->runNumber = zebraExplain_runNumberIncrement (zh->reg->zei,
+ 0);
+
+ /* update store data */
+ xfree (rec->info[recInfo_storeData]);
+ if (rGroup->flagStoreData == 1)
+ {
+ rec->size[recInfo_storeData] = recordAttr->recordSize;
+ rec->info[recInfo_storeData] = (char *)
+ xmalloc (recordAttr->recordSize);
+ if (lseek (fi->fd, recordOffset, SEEK_SET) < 0)
+ {
+ logf (LOG_ERRNO|LOG_FATAL, "seek to " PRINTF_OFF_T " in %s",
+ recordOffset, fname);
+ exit (1);
+ }
+ if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize)
+ < recordAttr->recordSize)
+ {
+ logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s",
+ recordAttr->recordSize, fname);
+ exit (1);
+ }
+ }
+ else
+ {
+ rec->info[recInfo_storeData] = NULL;
+ rec->size[recInfo_storeData] = 0;
+ }
+ /* update database name */
+ xfree (rec->info[recInfo_databaseName]);
+ rec->info[recInfo_databaseName] =
+ rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]);
+
+ /* update offset */
+ recordAttr->recordOffset = recordOffset;
+
+ /* commit this record */
+ rec_put (zh->reg->records, &rec);
+ logRecord (0);
+ return 1;
+}
+
+int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname,
+ const struct recordGroup *rGroupP, int deleteFlag)
+{
+ int r, i, fd;
+ char gprefix[128];
+ char ext[128];
+ char ext_res[128];
+ char subType[128];
+ RecType recType;
+ struct recordGroup rGroupM;
+ struct recordGroup *rGroup = &rGroupM;
+ struct file_read_info *fi;
+ void *clientData;
+
+ memcpy (rGroup, rGroupP, sizeof(*rGroupP));
+
+ if (!rGroup->groupName || !*rGroup->groupName)
+ *gprefix = '\0';
+ else
+ sprintf (gprefix, "%s.", rGroup->groupName);
+
+ logf (LOG_DEBUG, "fileExtract %s", fname);
+
+ /* determine file extension */
+ *ext = '\0';
+ for (i = strlen(fname); --i >= 0; )
+ if (fname[i] == '/')
+ break;
+ else if (fname[i] == '.')
+ {
+ strcpy (ext, fname+i+1);
+ break;
+ }
+ /* determine file type - depending on extension */
+ if (!rGroup->recordType)
+ {
+ sprintf (ext_res, "%srecordType.%s", gprefix, ext);
+ if (!(rGroup->recordType = res_get (zh->res, ext_res)))
+ {
+ sprintf (ext_res, "%srecordType", gprefix);
+ rGroup->recordType = res_get (zh->res, ext_res);
+ }
+ }
+ if (!rGroup->recordType)
+ {
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "? %s", fname);
+ return 0;
+ }
+ if (!*rGroup->recordType)
+ return 0;
+ if (!(recType =
+ recType_byName (zh->reg->recTypes, rGroup->recordType, subType,
+ &clientData)))
+ {
+ logf (LOG_WARN, "No such record type: %s", rGroup->recordType);
+ return 0;
+ }
+
+ /* determine match criteria */
+ if (!rGroup->recordId)
+ {
+ sprintf (ext_res, "%srecordId.%s", gprefix, ext);
+ rGroup->recordId = res_get (zh->res, ext_res);
+ }
+
+ /* determine database name */
+ if (!rGroup->databaseName)
+ {
+ sprintf (ext_res, "%sdatabase.%s", gprefix, ext);
+ if (!(rGroup->databaseName = res_get (zh->res, ext_res)))
+ {
+ sprintf (ext_res, "%sdatabase", gprefix);
+ rGroup->databaseName = res_get (zh->res, ext_res);
+ }
+ }
+ if (!rGroup->databaseName)
+ rGroup->databaseName = "Default";
+
+ /* determine if explain database */
+
+ sprintf (ext_res, "%sexplainDatabase", gprefix);
+ rGroup->explainDatabase =
+ atoi (res_get_def (zh->res, ext_res, "0"));
+
+ /* announce database */
+ if (zebraExplain_curDatabase (zh->reg->zei, rGroup->databaseName))
+ {
+ if (zebraExplain_newDatabase (zh->reg->zei, rGroup->databaseName,
+ rGroup->explainDatabase))
+ return 0;
+ }
+
+ if (rGroup->flagStoreData == -1)
+ {
+ const char *sval;
+ sprintf (ext_res, "%sstoreData.%s", gprefix, ext);
+ if (!(sval = res_get (zh->res, ext_res)))
+ {
+ sprintf (ext_res, "%sstoreData", gprefix);
+ sval = res_get (zh->res, ext_res);
+ }
+ if (sval)
+ rGroup->flagStoreData = atoi (sval);
+ }
+ if (rGroup->flagStoreData == -1)
+ rGroup->flagStoreData = 0;
+
+ if (rGroup->flagStoreKeys == -1)
+ {
+ const char *sval;
+
+ sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext);
+ sval = res_get (zh->res, ext_res);
+ if (!sval)
+ {
+ sprintf (ext_res, "%sstoreKeys", gprefix);
+ sval = res_get (zh->res, ext_res);
+ }
+ if (!sval)
+ sval = res_get (zh->res, "storeKeys");
+ if (sval)
+ rGroup->flagStoreKeys = atoi (sval);
+ }
+ if (rGroup->flagStoreKeys == -1)
+ rGroup->flagStoreKeys = 0;
+
+ if (sysno && deleteFlag)
+ fd = -1;
+ else
+ {
+ if ((fd = open (fname, O_BINARY|O_RDONLY)) == -1)
+ {
+ logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
+ return 0;
+ }
+ }
+ fi = file_read_start (fd);
+ do
+ {
+ file_begin (fi);
+ r = recordExtract (zh, sysno, fname, rGroup, deleteFlag, fi,
+ recType, subType, clientData);
+ } while (r && !sysno && fi->file_more);
+ file_read_stop (fi);
+ if (fd != -1)
+ close (fd);
+ return r;
+}
+
+
+int extract_rec_in_mem (ZebraHandle zh, const char *recordType,
+ const char *buf, size_t buf_size,
+ const char *databaseName, int delete_flag,
+ int test_mode, int *sysno,
+ int store_keys, int store_data,
+ const char *match_criteria)
+{
+ RecordAttr *recordAttr;
+ struct recExtractCtrl extractCtrl;
+ int i, r;
+ char *matchStr = 0;
+ RecType recType;
+ char subType[1024];
+ void *clientData;
+ const char *fname = "<no file>";
+ Record rec;
+ long recordOffset = 0;
+ struct zebra_fetch_control fc;
+
+ fc.fd = -1;
+ fc.record_int_buf = buf;
+ fc.record_int_len = buf_size;
+ fc.record_int_pos = 0;
+ fc.offset_end = 0;
+ fc.record_offset = 0;
+
+ extractCtrl.offset = 0;
+ extractCtrl.readf = zebra_record_int_read;
+ extractCtrl.seekf = zebra_record_int_seek;
+ extractCtrl.tellf = zebra_record_int_tell;
+ extractCtrl.endf = zebra_record_int_end;
+ extractCtrl.fh = &fc;
+
+ /* announce database */
+ if (zebraExplain_curDatabase (zh->reg->zei, databaseName))
+ {
+ if (zebraExplain_newDatabase (zh->reg->zei, databaseName, 0))
+ return 0;
+ }
+ if (!(recType =
+ recType_byName (zh->reg->recTypes, recordType, subType,
+ &clientData)))
+ {
+ logf (LOG_WARN, "No such record type: %s", recordType);
+ return 0;
+ }
+
+ zh->reg->keys.buf_used = 0;
+ zh->reg->keys.prevAttrUse = -1;
+ zh->reg->keys.prevAttrSet = -1;
+ zh->reg->keys.prevSeqNo = 0;
+ zh->reg->sortKeys = 0;
+
+ extractCtrl.subType = subType;
+ extractCtrl.init = extract_init;
+ extractCtrl.tokenAdd = extract_token_add;
+ extractCtrl.schemaAdd = extract_schema_add;
+ extractCtrl.dh = zh->reg->dh;
+ extractCtrl.handle = zh;
+ extractCtrl.zebra_maps = zh->reg->zebra_maps;
+ extractCtrl.flagShowRecords = 0;
+ for (i = 0; i<256; i++)
+ {
+ if (zebra_maps_is_positioned(zh->reg->zebra_maps, i))
+ extractCtrl.seqno[i] = 1;
+ else
+ extractCtrl.seqno[i] = 0;
+ }
+
+ r = (*recType->extract)(clientData, &extractCtrl);
+
+ if (r == RECCTRL_EXTRACT_EOF)
+ return 0;
+ else if (r == RECCTRL_EXTRACT_ERROR)
+ {
+ /* error occured during extraction ... */
+#if 1
+ yaz_log (LOG_WARN, "extract error");
+#else
+ if (rGroup->flagRw &&
+ records_processed < rGroup->fileVerboseLimit)
+ {
+ logf (LOG_WARN, "fail %s %s %ld", rGroup->recordType,
+ fname, (long) recordOffset);
+ }
+#endif
+ return 0;
+ }
+ if (zh->reg->keys.buf_used == 0)
+ {
+ /* the extraction process returned no information - the record
+ is probably empty - unless flagShowRecords is in use */
+ if (test_mode)
+ return 1;
+ logf (LOG_WARN, "No keys generated for record");
+ logf (LOG_WARN, " The file is probably empty");
+ return 1;
+ }
+ /* match criteria */
+
+ if (! *sysno)
+ {
+ /* new record */
+ if (delete_flag)
+ {
+ logf (LOG_LOG, "delete %s %s %ld", recordType,
+ fname, (long) recordOffset);
+ logf (LOG_WARN, "cannot delete record above (seems new)");
+ return 1;
+ }
+ logf (LOG_LOG, "add %s %s %ld", recordType, fname,
+ (long) recordOffset);
+ rec = rec_new (zh->reg->records);
+
+ *sysno = rec->sysno;
+
+ recordAttr = rec_init_attr (zh->reg->zei, rec);
+
+ if (matchStr)
+ {
+ dict_insert (zh->reg->matchDict, matchStr,
+ sizeof(*sysno), sysno);
+ }
+ extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
+ extract_flushSortKeys (zh, *sysno, 1, &zh->reg->sortKeys);
+ }
+ else
+ {
+ /* record already exists */
+ struct recKeys delkeys;
+
+ rec = rec_get (zh->reg->records, *sysno);
+ assert (rec);
+
+ recordAttr = rec_init_attr (zh->reg->zei, rec);
+
+ if (recordAttr->runNumber ==
+ zebraExplain_runNumberIncrement (zh->reg->zei, 0))
+ {
+ logf (LOG_LOG, "skipped %s %s %ld", recordType,
+ fname, (long) recordOffset);
+ rec_rm (&rec);
+ return 1;
+ }
+ delkeys.buf_used = rec->size[recInfo_delKeys];
+ delkeys.buf = rec->info[recInfo_delKeys];
+ extract_flushSortKeys (zh, *sysno, 0, &zh->reg->sortKeys);
+ extract_flushRecordKeys (zh, *sysno, 0, &delkeys);
+ if (delete_flag)
+ {
+ /* record going to be deleted */
+ if (!delkeys.buf_used)
+ {
+ logf (LOG_LOG, "delete %s %s %ld", recordType,
+ fname, (long) recordOffset);
+ logf (LOG_WARN, "cannot delete file above, storeKeys false");
+ }
+ else
+ {
+ logf (LOG_LOG, "delete %s %s %ld", recordType,
+ fname, (long) recordOffset);
+#if 0
+ if (matchStr)
+ dict_delete (matchDict, matchStr);
+#endif
+ rec_del (zh->reg->records, &rec);
+ }
+ rec_rm (&rec);
+ return 1;
+ }
+ else
+ {
+ /* record going to be updated */
+ if (!delkeys.buf_used)
+ {
+ logf (LOG_LOG, "update %s %s %ld", recordType,
+ fname, (long) recordOffset);
+ logf (LOG_WARN, "cannot update file above, storeKeys false");
+ }
+ else
+ {
+ logf (LOG_LOG, "update %s %s %ld", recordType,
+ fname, (long) recordOffset);
+ extract_flushRecordKeys (zh, *sysno, 1, &zh->reg->keys);
+ }
+ }
+ }
+ /* update file type */
+ xfree (rec->info[recInfo_fileType]);
+ rec->info[recInfo_fileType] =
+ rec_strdup (recordType, &rec->size[recInfo_fileType]);
+
+ /* update filename */
+ xfree (rec->info[recInfo_filename]);
+ rec->info[recInfo_filename] =
+ rec_strdup (fname, &rec->size[recInfo_filename]);
+
+ /* update delete keys */
+ xfree (rec->info[recInfo_delKeys]);
+ if (zh->reg->keys.buf_used > 0 && store_keys == 1)
+ {
+ rec->size[recInfo_delKeys] = zh->reg->keys.buf_used;
+ rec->info[recInfo_delKeys] = zh->reg->keys.buf;
+ zh->reg->keys.buf = NULL;
+ zh->reg->keys.buf_max = 0;
+ }
+ else
+ {
+ rec->info[recInfo_delKeys] = NULL;
+ rec->size[recInfo_delKeys] = 0;
+ }
-void key_open (int mem)
-{
-#if !RECORD_BASE
- void *file_key;
+ /* 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
- 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;
+ zebraExplain_recordBytesIncrement (zh->reg->zei,
+ recordAttr->recordSize);
- key_del_buf = NULL;
- key_del_max = 0;
+ /* set run-number for this record */
+ recordAttr->runNumber =
+ zebraExplain_runNumberIncrement (zh->reg->zei, 0);
- if (!(file_idx = dict_open (FNAME_FILE_DICT, 40, 1)))
+ /* update store data */
+ xfree (rec->info[recInfo_storeData]);
+ if (store_data == 1)
{
- logf (LOG_FATAL, "dict_open fail of %s", "fileidx");
- exit (1);
- }
-#if RECORD_BASE
- assert (!records);
- records = rec_open (1);
+ rec->size[recInfo_storeData] = recordAttr->recordSize;
+ rec->info[recInfo_storeData] = (char *)
+ xmalloc (recordAttr->recordSize);
+#if 1
+ memcpy (rec->info[recInfo_storeData], buf, recordAttr->recordSize);
#else
- file_key = dict_lookup (file_idx, ".");
- if (file_key)
- memcpy (&sysno_next, (char*)file_key+1, sizeof(sysno_next));
+ if (lseek (fi->fd, recordOffset, SEEK_SET) < 0)
+ {
+ logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s",
+ (long) recordOffset, fname);
+ exit (1);
+ }
+ if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize)
+ < recordAttr->recordSize)
+ {
+ logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s",
+ recordAttr->recordSize, fname);
+ exit (1);
+ }
+#endif
+ }
else
- sysno_next = 1;
- if ((sys_idx_fd = open (FNAME_SYS_IDX, O_RDWR|O_CREAT, 0666)) == -1)
{
- logf (LOG_FATAL|LOG_ERRNO, "open %s", FNAME_SYS_IDX);
- exit (1);
+ rec->info[recInfo_storeData] = NULL;
+ rec->size[recInfo_storeData] = 0;
}
-#endif
-}
+ /* update database name */
+ xfree (rec->info[recInfo_databaseName]);
+ rec->info[recInfo_databaseName] =
+ rec_strdup (databaseName, &rec->size[recInfo_databaseName]);
-struct encode_info {
- int sysno;
- int seqno;
- char buf[512];
-};
+ /* update offset */
+ recordAttr->recordOffset = recordOffset;
+
+ /* commit this record */
+ rec_put (zh->reg->records, &rec);
-void encode_key_init (struct encode_info *i)
-{
- i->sysno = 0;
- i->seqno = 0;
+ return 0;
}
-char *encode_key_int (int d, char *bp)
+int explain_extract (void *handle, Record rec, data1_node *n)
{
- if (d <= 63)
- *bp++ = d;
- else if (d <= 16383)
- {
- *bp++ = 64 + (d>>8);
- *bp++ = d & 255;
- }
- else if (d <= 4194303)
+ ZebraHandle zh = (ZebraHandle) handle;
+ struct recExtractCtrl extractCtrl;
+ int i;
+
+ if (zebraExplain_curDatabase (zh->reg->zei,
+ rec->info[recInfo_databaseName]))
{
- *bp++ = 128 + (d>>16);
- *bp++ = (d>>8) & 255;
- *bp++ = d & 255;
+ abort();
+ if (zebraExplain_newDatabase (zh->reg->zei,
+ rec->info[recInfo_databaseName], 0))
+ abort ();
}
- else
+
+ zh->reg->keys.buf_used = 0;
+ zh->reg->keys.prevAttrUse = -1;
+ zh->reg->keys.prevAttrSet = -1;
+ zh->reg->keys.prevSeqNo = 0;
+ zh->reg->sortKeys = 0;
+
+ extractCtrl.init = extract_init;
+ extractCtrl.tokenAdd = extract_token_add;
+ extractCtrl.schemaAdd = extract_schema_add;
+ extractCtrl.dh = zh->reg->dh;
+ for (i = 0; i<256; i++)
+ extractCtrl.seqno[i] = 0;
+ extractCtrl.zebra_maps = zh->reg->zebra_maps;
+ extractCtrl.flagShowRecords = 0;
+ extractCtrl.handle = handle;
+
+ grs_extract_tree(&extractCtrl, n);
+
+ if (rec->size[recInfo_delKeys])
{
- *bp++ = 192 + (d>>24);
- *bp++ = (d>>16) & 255;
- *bp++ = (d>>8) & 255;
- *bp++ = d & 255;
+ struct recKeys delkeys;
+ struct sortKey *sortKeys = 0;
+
+ delkeys.buf_used = rec->size[recInfo_delKeys];
+ delkeys.buf = rec->info[recInfo_delKeys];
+ extract_flushSortKeys (zh, rec->sysno, 0, &sortKeys);
+ extract_flushRecordKeys (zh, rec->sysno, 0, &delkeys);
}
- return bp;
+ extract_flushRecordKeys (zh, rec->sysno, 1, &zh->reg->keys);
+ extract_flushSortKeys (zh, rec->sysno, 1, &zh->reg->sortKeys);
+
+ xfree (rec->info[recInfo_delKeys]);
+ rec->size[recInfo_delKeys] = zh->reg->keys.buf_used;
+ rec->info[recInfo_delKeys] = zh->reg->keys.buf;
+ zh->reg->keys.buf = NULL;
+ zh->reg->keys.buf_max = 0;
+ return 0;
}
-void encode_key_write (char *k, struct encode_info *i, FILE *outf)
+void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno,
+ int cmd, struct recKeys *reckeys)
{
- struct it_key key;
- char *bp = i->buf;
+#if SU_SCHEME
+#else
+ unsigned char attrSet = (unsigned char) -1;
+ unsigned short attrUse = (unsigned short) -1;
+#endif
+ int seqno = 0;
+ int off = 0;
+ int ch = 0;
+ ZebraExplainInfo zei = zh->reg->zei;
- 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)
+ if (!zh->reg->key_buf)
{
- i->sysno = key.sysno;
- i->seqno = 0;
+ int mem = 8*1024*1024;
+ zh->reg->key_buf = (char**) xmalloc (mem);
+ zh->reg->ptr_top = mem/sizeof(char*);
+ zh->reg->ptr_i = 0;
+ zh->reg->key_buf_used = 0;
+ zh->reg->key_file_no = 0;
}
- bp = encode_key_int (key.seqno - i->seqno, bp);
- i->seqno = key.seqno;
- if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
+ zebraExplain_recordCountIncrement (zei, cmd ? 1 : -1);
+ while (off < reckeys->buf_used)
{
- logf (LOG_FATAL|LOG_ERRNO, "fwrite");
- exit (1);
+ const char *src = reckeys->buf + off;
+ struct it_key key;
+ int lead;
+
+ lead = *src++;
+
+#if SU_SCHEME
+ if ((lead & 3) < 3)
+ {
+ memcpy (&ch, src, sizeof(ch));
+ src += sizeof(ch);
+ }
+#else
+ if (!(lead & 1))
+ {
+ memcpy (&attrSet, src, sizeof(attrSet));
+ src += sizeof(attrSet);
+ }
+ if (!(lead & 2))
+ {
+ memcpy (&attrUse, src, sizeof(attrUse));
+ src += sizeof(attrUse);
+ }
+#endif
+ if (zh->reg->key_buf_used + 1024 >
+ (zh->reg->ptr_top -zh->reg->ptr_i)*sizeof(char*))
+ extract_flushWriteKeys (zh);
+ ++(zh->reg->ptr_i);
+ (zh->reg->key_buf)[zh->reg->ptr_top - zh->reg->ptr_i] =
+ (char*)zh->reg->key_buf + zh->reg->key_buf_used;
+#if SU_SCHEME
+#else
+ ch = zebraExplain_lookupSU (zei, attrSet, attrUse);
+ if (ch < 0)
+ ch = zebraExplain_addSU (zei, attrSet, attrUse);
+#endif
+ assert (ch > 0);
+ zh->reg->key_buf_used +=
+ key_SU_encode (ch,((char*)zh->reg->key_buf) +
+ zh->reg->key_buf_used);
+
+ while (*src)
+ ((char*)zh->reg->key_buf) [(zh->reg->key_buf_used)++] = *src++;
+ src++;
+ ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = '\0';
+ ((char*)(zh->reg->key_buf))[(zh->reg->key_buf_used)++] = cmd;
+
+ if (lead & 60)
+ seqno += ((lead>>2) & 15)-1;
+ else
+ {
+ memcpy (&seqno, src, sizeof(seqno));
+ src += sizeof(seqno);
+ }
+ key.seqno = seqno;
+ key.sysno = sysno;
+ memcpy ((char*)zh->reg->key_buf + zh->reg->key_buf_used, &key, sizeof(key));
+ (zh->reg->key_buf_used) += sizeof(key);
+ off = src - reckeys->buf;
}
+ assert (off == reckeys->buf_used);
}
-void key_flush (void)
+void extract_flushWriteKeys (ZebraHandle zh)
{
FILE *outf;
char out_fname[200];
char *prevcp, *cp;
struct encode_info encode_info;
-
- if (ptr_i <= 0)
+ int ptr_i = zh->reg->ptr_i;
+#if SORT_EXTRA
+ int i;
+#endif
+ if (!zh->reg->key_buf || ptr_i <= 0)
return;
- key_file_no++;
- logf (LOG_LOG, "sorting section %d", key_file_no);
- qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare);
- sprintf (out_fname, TEMP_FNAME, key_file_no);
+ (zh->reg->key_file_no)++;
+ logf (LOG_LOG, "sorting section %d", (zh->reg->key_file_no));
+#if !SORT_EXTRA
+ qsort (zh->reg->key_buf + zh->reg->ptr_top - ptr_i, ptr_i, sizeof(char*),
+ key_qsort_compare);
+ extract_get_fname_tmp (zh, out_fname, zh->reg->key_file_no);
- if (!(outf = fopen (out_fname, "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);
- prevcp = cp = key_buf[ptr_top-ptr_i];
+ 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 = key_buf[ptr_top-ptr_i];
+ cp = (zh->reg->key_buf)[zh->reg->ptr_top - ptr_i];
if (strcmp (cp, prevcp))
{
encode_key_init (&encode_info);
else
encode_key_write (cp + strlen(cp), &encode_info, outf);
}
+#else
+ qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_x_compare);
+ extract_get_fname_tmp (out_fname, key_file_no);
+
+ if (!(outf = fopen (out_fname, "wb")))
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname);
+ exit (1);
+ }
+ logf (LOG_LOG, "writing section %d", key_file_no);
+ i = ptr_i;
+ prevcp = key_buf[ptr_top-i];
+ while (1)
+ if (!--i || strcmp (prevcp, key_buf[ptr_top-i]))
+ {
+ key_y_len = strlen(prevcp)+1;
+#if 0
+ logf (LOG_LOG, "key_y_len: %2d %02x %02x %s",
+ key_y_len, prevcp[0], prevcp[1], 2+prevcp);
+#endif
+ qsort (key_buf + ptr_top-ptr_i, ptr_i - i,
+ sizeof(char*), key_y_compare);
+ cp = key_buf[ptr_top-ptr_i];
+ --key_y_len;
+ encode_key_init (&encode_info);
+ encode_key_write (cp, &encode_info, outf);
+ while (--ptr_i > i)
+ {
+ cp = key_buf[ptr_top-ptr_i];
+ encode_key_write (cp+key_y_len, &encode_info, outf);
+ }
+ if (!i)
+ break;
+ prevcp = key_buf[ptr_top-ptr_i];
+ }
+#endif
if (fclose (outf))
{
logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname);
exit (1);
}
- logf (LOG_LOG, "finished section %d", key_file_no);
- ptr_i = 0;
- key_buf_used = 0;
+ logf (LOG_LOG, "finished section %d", zh->reg->key_file_no);
+ zh->reg->ptr_i = 0;
+ zh->reg->key_buf_used = 0;
}
-int key_close (void)
+void extract_add_index_string (RecWord *p, const char *string,
+ int length)
{
- key_flush ();
- xfree (key_buf);
-#if RECORD_BASE
- rec_close (&records);
-#else
- close (sys_idx_fd);
- dict_insert (file_idx, ".", sizeof(sysno_next), &sysno_next);
-#endif
- dict_close (file_idx);
+ char *dst;
+ unsigned char attrSet;
+ unsigned short attrUse;
+ int lead = 0;
+ int diff = 0;
+ int *pseqno = &p->seqnos[p->reg_type];
+ ZebraHandle zh = p->extractCtrl->handle;
+ ZebraExplainInfo zei = zh->reg->zei;
+ struct recKeys *keys = &zh->reg->keys;
- xfree (key_del_buf);
- key_del_buf = NULL;
- key_del_max = 0;
- return key_file_no;
-}
-
-static void wordInit (RecWord *p)
-{
- p->attrSet = 1;
- p->attrUse = 1016;
- p->which = Word_String;
-}
+ if (keys->buf_used+1024 > keys->buf_max)
+ {
+ char *b;
-static void wordAdd (const RecWord *p)
-{
- struct it_key key;
- size_t i;
-
- 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,
- p->attrSet, p->attrUse,
- key_databaseName);
- switch (p->which)
- {
- case Word_String:
- for (i = 0; p->u.string[i]; i++)
- ((char*)key_buf) [key_buf_used++] =
- index_char_cvt (p->u.string[i]);
- ((char*)key_buf) [key_buf_used++] = '\0';
- break;
- default:
- return ;
+ 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;
}
- ((char*) key_buf)[key_buf_used++] = ((key_cmd == 'a') ? 1 : 0);
- key.sysno = key_sysno;
- key.seqno = p->seqno;
- memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key));
- key_buf_used += sizeof(key);
+ dst = keys->buf + keys->buf_used;
- if (key_cmd == 'a' && key_del_used >= 0)
+ attrSet = p->attrSet;
+ if (keys->buf_used > 0 && keys->prevAttrSet == attrSet)
+ lead |= 1;
+ else
+ keys->prevAttrSet = attrSet;
+ attrUse = p->attrUse;
+ if (keys->buf_used > 0 && keys->prevAttrUse == attrUse)
+ lead |= 2;
+ else
+ keys->prevAttrUse = attrUse;
+#if 1
+ diff = 1 + *pseqno - keys->prevSeqNo;
+ if (diff >= 1 && diff <= 15)
+ lead |= (diff << 2);
+ else
+ diff = 0;
+#endif
+ keys->prevSeqNo = *pseqno;
+
+ *dst++ = lead;
+
+#if SU_SCHEME
+ if ((lead & 3) < 3)
{
- char attrSet;
- short attrUse;
- if (key_del_used + 1024 > key_del_max)
- {
- char *kbn;
-
- if (!(kbn = malloc (key_del_max += 64000)))
- {
- logf (LOG_FATAL, "malloc");
- exit (1);
- }
- if (key_del_buf)
- memcpy (kbn, key_del_buf, key_del_used);
- free (key_del_buf);
- key_del_buf = kbn;
- }
- switch (p->which)
+ int ch = zebraExplain_lookupSU (zei, attrSet, attrUse);
+ if (ch < 0)
{
- case Word_String:
- for (i = 0; p->u.string[i]; i++)
- ((char*)key_del_buf) [key_del_used++] = p->u.string[i];
- ((char*)key_del_buf) [key_del_used++] = '\0';
- break;
- default:
- return ;
+ ch = zebraExplain_addSU (zei, attrSet, attrUse);
+ yaz_log (LOG_DEBUG, "addSU set=%d use=%d SU=%d",
+ attrSet, attrUse, ch);
}
- attrSet = p->attrSet;
- memcpy (key_del_buf + key_del_used, &attrSet, sizeof(attrSet));
- key_del_used += sizeof(attrSet);
-
- attrUse = p->attrUse;
- memcpy (key_del_buf + key_del_used, &attrUse, sizeof(attrUse));
- key_del_used += sizeof(attrUse);
+ assert (ch > 0);
+ memcpy (dst, &ch, sizeof(ch));
+ dst += sizeof(ch);
+ }
+#else
+ if (!(lead & 1))
+ {
+ memcpy (dst, &attrSet, sizeof(attrSet));
+ dst += sizeof(attrSet);
+ }
+ if (!(lead & 2))
+ {
+ memcpy (dst, &attrUse, sizeof(attrUse));
+ dst += sizeof(attrUse);
+ }
+#endif
+ *dst++ = p->reg_type;
+ memcpy (dst, string, length);
+ dst += length;
+ *dst++ = '\0';
- memcpy (key_del_buf + key_del_used, &p->seqno, sizeof(p->seqno));
- key_del_used += sizeof(p->seqno);
+ if (!diff)
+ {
+ memcpy (dst, pseqno, sizeof(*pseqno));
+ dst += sizeof(*pseqno);
}
+ keys->buf_used = dst - keys->buf;
+ if (*pseqno)
+ (*pseqno)++;
}
-static void wordAddAny (const RecWord *p)
+static void extract_add_sort_string (RecWord *p, const char *string,
+ int length)
{
- if (p->attrSet != 1 || p->attrUse != 1016)
- {
- RecWord w;
+ struct sortKey *sk;
+ ZebraHandle zh = p->extractCtrl->handle;
- memcpy (&w, p, sizeof(w));
- w.attrSet = 1;
- w.attrUse = 1016;
- wordAdd (&w);
- }
- wordAdd (p);
-}
+ for (sk = zh->reg->sortKeys; sk; sk = sk->next)
+ if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse)
+ return;
-static char *file_buf;
-static int file_offset;
-static int file_bufsize;
+ sk = (struct sortKey *) xmalloc (sizeof(*sk));
+ sk->next = zh->reg->sortKeys;
+ zh->reg->sortKeys = sk;
-static void file_read_start (int fd)
-{
- file_offset = 0;
- file_buf = xmalloc (4096);
- file_bufsize = read (fd, file_buf, 4096);
+ sk->string = (char *) xmalloc (length);
+ sk->length = length;
+ memcpy (sk->string, string, length);
+
+ sk->attrSet = p->attrSet;
+ sk->attrUse = p->attrUse;
}
-static void file_read_stop (int fd)
+void extract_add_string (RecWord *p, const char *string, int length)
{
- xfree (file_buf);
+ 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 int file_read (int fd, char *buf, size_t count)
+static void extract_add_incomplete_field (RecWord *p)
{
- int l = file_bufsize - file_offset;
+ const char *b = p->string;
+ int remain = p->length;
+ const char **map = 0;
- if (count > l)
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+
+ while (map)
{
- 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;
- return r;
- }
- file_bufsize = r = read (fd, file_buf, 4096);
- if (r == -1)
- {
- logf (LOG_FATAL|LOG_ERRNO, "read");
- exit (1);
- }
- else if (r <= count)
- {
- file_offset = r;
- memcpy (buf + l, file_buf, r);
- return l + r;
- }
- else
- {
- file_offset = count;
- memcpy (buf + l, file_buf, count - l);
- return count;
- }
+ char buf[IT_MAX_WORD+1];
+ int i, remain;
+
+ /* Skip spaces */
+ while (map && *map && **map == *CHR_SPACE)
+ {
+ remain = p->length - (b - p->string);
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+ else
+ map = 0;
+ }
+ if (!map)
+ break;
+ i = 0;
+ while (map && *map && **map != *CHR_SPACE)
+ {
+ const char *cp = *map;
+
+ while (i < IT_MAX_WORD && *cp)
+ buf[i++] = *(cp++);
+ remain = p->length - (b - p->string);
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+ else
+ map = 0;
+ }
+ if (!i)
+ return;
+ extract_add_string (p, buf, i);
}
- memcpy (buf, file_buf + file_offset, count);
- file_offset += count;
- return count;
+ (p->seqnos[p->reg_type])++; /* to separate this from next one */
}
-SYSNO file_extract (int cmd, const char *fname, const char *kname,
- char *databaseName)
+static void extract_add_complete_field (RecWord *p)
{
- int i, r;
- char ext[128];
- SYSNO sysno;
- char ext_res[128];
- const char *file_type;
- void *file_info;
- struct recExtractCtrl extractCtrl;
- RecType rt;
+ const char *b = p->string;
+ char buf[IT_MAX_WORD+1];
+ const char **map = 0;
+ int i = 0, remain = p->length;
- key_del_used = -1;
- key_databaseName = databaseName;
- for (i = strlen(fname); --i >= 0; )
- if (fname[i] == '/')
- {
- strcpy (ext, "");
- break;
- }
- else if (fname[i] == '.')
- {
- strcpy (ext, fname+i+1);
- break;
- }
- sprintf (ext_res, "fileExtension.%s", ext);
- if (!(file_type = res_get (common_resource, ext_res)))
- return 0;
- if (!(rt = recType_byName (file_type)))
- return 0;
- logf (LOG_DEBUG, "%c %s k=%s", cmd, fname, kname);
- file_info = dict_lookup (file_idx, kname);
- if (!file_info)
- {
-#if RECORD_BASE
- Record rec = rec_new (records);
-
- sysno = rec->sysno;
- dict_insert (file_idx, kname, sizeof(sysno), &sysno);
- rec->info[0] = rec_strdup (file_type, &rec->size[0]);
- rec->info[1] = rec_strdup (kname, &rec->size[1]);
- rec_put (records, &rec);
-#else
- sysno = sysno_next++;
- dict_insert (file_idx, kname, sizeof(sysno), &sysno);
- lseek (sys_idx_fd, sysno * SYS_IDX_ENTRY_LEN, SEEK_SET);
- write (sys_idx_fd, file_type, strlen (file_type)+1);
- write (sys_idx_fd, kname, strlen(kname)+1);
-#endif
- }
- else
- memcpy (&sysno, (char*) file_info+1, sizeof(sysno));
+ if (remain > 0)
+ map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain);
- if ((extractCtrl.fd = open (fname, O_RDONLY)) == -1)
+ while (remain > 0 && i < IT_MAX_WORD)
{
- logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
- return 0;
- }
- extractCtrl.subType = "";
- extractCtrl.init = wordInit;
- extractCtrl.add = wordAddAny;
-
- file_read_start (extractCtrl.fd);
+ 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;
- extractCtrl.readf = file_read;
- key_sysno = sysno;
- key_cmd = cmd;
- r = (*rt->extract)(&extractCtrl);
+ if (i && i < IT_MAX_WORD)
+ buf[i++] = *CHR_SPACE;
+ while (map && *map && **map != *CHR_SPACE)
+ {
+ const char *cp = *map;
- file_read_stop (extractCtrl.fd);
-
- close (extractCtrl.fd);
- if (r)
- logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);
- return sysno;
+ if (i >= IT_MAX_WORD)
+ break;
+ while (i < IT_MAX_WORD && *cp)
+ buf[i++] = *(cp++);
+ remain = p->length - (b - p->string);
+ if (remain > 0)
+ map = zebra_maps_input (p->zebra_maps, p->reg_type, &b,
+ remain);
+ else
+ map = 0;
+ }
+ }
+ if (!i)
+ return;
+ extract_add_string (p, buf, i);
}
-int fileExtract (SYSNO *sysno, const char *fname, const char *databaseName,
- int deleteFlag)
+void extract_token_add (RecWord *p)
{
- int i, r;
- char ext[128];
- char ext_res[128];
- const char *file_type;
- struct recExtractCtrl extractCtrl;
- RecType rt;
- Record rec;
+ WRBUF wrbuf;
+ if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0,
+ p->string, p->length)))
+ {
+ p->string = wrbuf_buf(wrbuf);
+ p->length = wrbuf_len(wrbuf);
+ }
+ if (zebra_maps_is_complete (p->zebra_maps, p->reg_type))
+ extract_add_complete_field (p);
+ else
+ extract_add_incomplete_field(p);
+}
- logf (LOG_DEBUG, "fileExtractAdd %s", fname);
+void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid)
+{
+ ZebraHandle zh = (ZebraHandle) (p->handle);
+ zebraExplain_addSchema (zh->reg->zei, oid);
+}
- key_del_used = 0;
- for (i = strlen(fname); --i >= 0; )
- if (fname[i] == '/')
- {
- strcpy (ext, "");
- break;
- }
- else if (fname[i] == '.')
- {
- strcpy (ext, fname+i+1);
- break;
- }
- sprintf (ext_res, "fileExtension.%s", ext);
- if (!(file_type = res_get (common_resource, ext_res)))
- return 0;
- if (!(rt = recType_byName (file_type)))
- return 0;
+void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno,
+ int cmd, struct sortKey **skp)
+{
+ struct sortKey *sk = *skp;
+ SortIdx sortIdx = zh->reg->sortIdx;
- if ((extractCtrl.fd = open (fname, O_RDONLY)) == -1)
+ sortIdx_sysno (sortIdx, sysno);
+ while (sk)
{
- logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
- return 0;
+ struct sortKey *sk_next = sk->next;
+ sortIdx_type (sortIdx, sk->attrUse);
+ sortIdx_add (sortIdx, sk->string, sk->length);
+ xfree (sk->string);
+ xfree (sk);
+ sk = sk_next;
}
+ *skp = 0;
+}
- extractCtrl.subType = "";
- extractCtrl.init = wordInit;
- extractCtrl.add = wordAddAny;
+void encode_key_init (struct encode_info *i)
+{
+ i->sysno = 0;
+ i->seqno = 0;
+ i->cmd = -1;
+}
- if (! *sysno)
+char *encode_key_int (int d, char *bp)
+{
+ if (d <= 63)
+ *bp++ = d;
+ else if (d <= 16383)
{
- logf (LOG_LOG, "add record %s", fname);
- rec = rec_new (records);
- *sysno = rec->sysno;
- rec->info[0] = rec_strdup (file_type, &rec->size[0]);
- rec->info[1] = rec_strdup (fname, &rec->size[1]);
- rec->info[3] = rec_strdup (databaseName, &rec->size[3]);
+ *bp++ = 64 + (d>>8);
+ *bp++ = d & 255;
}
- else
+ else if (d <= 4194303)
{
- size_t off;
- char *kb;
-
- if (deleteFlag)
- logf (LOG_LOG, "delete record %s", fname);
- else
- logf (LOG_LOG, "update record %s", fname);
- rec = rec_get (records, *sysno);
-
- key_cmd = 'd';
- key_sysno = *sysno;
- key_databaseName = rec->info[3];
- kb = rec->info[2];
- for (off = 0; off < rec->size[2]; )
- {
- RecWord rw;
- char attrSet;
- short attrUse;
-
- rw.which = Word_String;
- rw.u.string = kb + off;
- while (kb[off])
- off++;
- off++;
- memcpy (&attrSet, kb + off, sizeof(attrSet));
- off += sizeof(attrSet);
- memcpy (&attrUse, kb + off, sizeof(attrUse));
- off += sizeof(attrUse);
- memcpy (&rw.seqno, kb + off, sizeof(rw.seqno));
- off += sizeof(rw.seqno);
- rw.attrSet = attrSet;
- rw.attrUse = attrUse;
-
- (*extractCtrl.add) (&rw);
- }
- assert (off == rec->size[2]);
- free (rec->info[2]);
- rec->info[2] = NULL;
- rec->size[2] = 0;
-
- free (rec->info[3]);
- rec->info[3] = rec_strdup (databaseName, &rec->size[3]);
- }
-
- if (deleteFlag)
+ *bp++ = 128 + (d>>16);
+ *bp++ = (d>>8) & 255;
+ *bp++ = d & 255;
+ }
+ else
{
-#if 0
- rec_del (records, *sysno);
- rec_rm (&rec);
-#endif
- return 1;
+ *bp++ = 192 + (d>>24);
+ *bp++ = (d>>16) & 255;
+ *bp++ = (d>>8) & 255;
+ *bp++ = d & 255;
}
-
- key_databaseName = databaseName;
- key_sysno = *sysno;
- key_cmd = 'a';
+ return bp;
+}
- file_read_start (extractCtrl.fd);
- extractCtrl.readf = file_read;
- r = (*rt->extract)(&extractCtrl);
- file_read_stop (extractCtrl.fd);
+void encode_key_write (char *k, struct encode_info *i, FILE *outf)
+{
+ struct it_key key;
+ char *bp = i->buf;
- close (extractCtrl.fd);
- if (r)
+ 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)
{
- rec_rm (&rec);
- logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);
- return 0;
+ i->sysno = key.sysno;
+ i->seqno = 0;
}
- if (key_del_used > 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)
{
- rec->size[2] = key_del_used;
- rec->info[2] = malloc (rec->size[2]);
- memcpy (rec->info[2], key_del_buf, rec->size[2]);
+ logf (LOG_FATAL|LOG_ERRNO, "fwrite");
+ exit (1);
}
- rec_put (records, &rec);
- return 1;
}
+
projects / idzebra-moved-to-github.git / blobdiff