/*
- * Copyright (C) 1995-1998, Index Data I/S
+ * Copyright (C) 1995-2002, Index Data
* All rights reserved.
- * Sebastian Hammer, Adam Dickmeiss
- *
- * $Log: index.h,v $
- * Revision 1.56 1998-01-12 15:04:08 adam
- * The test option (-s) only uses read-lock (and not write lock).
- *
- * Revision 1.55 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 1.54 1997/09/29 09:08:36 adam
- * Revised locking system to be thread safe for the server.
- *
- * Revision 1.53 1997/09/25 14:54:43 adam
- * WIN32 files lock support.
- *
- * Revision 1.52 1997/09/22 12:39:06 adam
- * Added get_pos method for the ranked result sets.
- *
- * Revision 1.51 1997/09/18 08:59:19 adam
- * Extra generic handle for the character mapping routines.
- *
- * Revision 1.50 1997/09/17 12:19:13 adam
- * Zebra version corresponds to YAZ version 1.4.
- * Changed Zebra server so that it doesn't depend on global common_resource.
- *
- * Revision 1.49 1997/09/05 15:30:08 adam
- * Changed prototype for chr_map_input - added const.
- * Added support for C++, headers uses extern "C" for public definitions.
- *
- * Revision 1.48 1997/02/12 20:39:45 adam
- * Implemented options -f <n> that limits the log to the first <n>
- * records.
- * Changed some log messages also.
- *
- * Revision 1.47 1996/12/23 15:30:44 adam
- * Work on truncation.
- * Bug fix: result sets weren't deleted after server shut down.
- *
- * Revision 1.46 1996/11/08 11:10:19 adam
- * Buffers used during file match got bigger.
- * Compressed ISAM support everywhere.
- * Bug fixes regarding masking characters in queries.
- * Redesigned Regexp-2 queries.
- *
- * Revision 1.45 1996/10/29 14:09:42 adam
- * Use of cisam system - enabled if setting isamc is 1.
- *
- * Revision 1.44 1996/06/06 12:08:40 quinn
- * Added showRecord function
- *
- * Revision 1.43 1996/06/04 10:18:12 adam
- * Search/scan uses character mapping module.
- *
- * Revision 1.42 1996/06/04 08:20:16 quinn
- * Smallish
- *
- * Revision 1.41 1996/06/04 07:54:55 quinn
- * Added output-map.
- *
- * Revision 1.40 1996/05/31 09:06:58 quinn
- * Work on character-set handling
- *
- * Revision 1.39 1996/05/14 14:04:33 adam
- * In zebraidx, the 'stat' command is improved. Statistics about ISAM/DICT
- * is collected.
- *
- * Revision 1.38 1996/04/12 07:02:23 adam
- * File update of single files.
- *
- * Revision 1.37 1996/03/26 16:01:13 adam
- * New setting lockPath: directory of various lock files.
- *
- * Revision 1.36 1996/03/21 14:50:09 adam
- * File update uses modify-time instead of change-time.
- *
- * Revision 1.35 1996/02/12 18:45:36 adam
- * New fileVerboseFlag in record group control.
- *
- * Revision 1.34 1995/12/11 11:43:29 adam
- * Locking based on fcntl instead of flock.
- * Setting commitEnable removed. Command line option -n can be used to
- * prevent commit if commit setting is defined in the configuration file.
- *
- * Revision 1.33 1995/12/08 16:22:53 adam
- * Work on update while servers are running. Three lock files introduced.
- * The servers reload their registers when necessary, but they don't
- * reestablish result sets yet.
- *
- * Revision 1.32 1995/12/07 17:38:46 adam
- * Work locking mechanisms for concurrent updates/commit.
- *
- * Revision 1.31 1995/12/06 12:41:22 adam
- * New command 'stat' for the index program.
- * Filenames can be read from stdin by specifying '-'.
- * Bug fix/enhancement of the transformation from terms to regular
- * expressons in the search engine.
- *
- * Revision 1.30 1995/12/05 11:25:02 adam
- * Include of zebraver.h.
- *
- * Revision 1.29 1995/11/28 09:09:40 adam
- * Zebra config renamed.
- * Use setting 'recordId' to identify record now.
- * Bug fix in recindex.c: rec_release_blocks was invokeded even
- * though the blocks were already released.
- * File traversal properly deletes records when needed.
- *
- * Revision 1.28 1995/11/27 13:58:53 adam
- * New option -t. storeStore data implemented in server.
- *
- * Revision 1.27 1995/11/25 10:24:06 adam
- * More record fields - they are enumerated now.
- * New options: flagStoreData flagStoreKey.
- *
- * Revision 1.26 1995/11/22 17:19:17 adam
- * Record management uses the bfile system.
- *
- * Revision 1.25 1995/11/21 15:29:12 adam
- * Config file 'base' read by default by both indexer and server.
- *
- * Revision 1.24 1995/11/21 15:01:15 adam
- * New general match criteria implemented.
- * New feature: document groups.
- *
- * Revision 1.23 1995/11/20 16:59:45 adam
- * New update method: the 'old' keys are saved for each records.
- *
- * Revision 1.22 1995/11/20 11:56:26 adam
- * Work on new traversal.
- *
- * Revision 1.21 1995/11/16 15:34:55 adam
- * Uses new record management system in both indexer and server.
- *
- * Revision 1.20 1995/11/15 14:46:18 adam
- * Started work on better record management system.
- *
- * Revision 1.19 1995/10/27 14:00:11 adam
- * Implemented detection of database availability.
- *
- * Revision 1.18 1995/10/17 18:02:08 adam
- * New feature: databases. Implemented as prefix to words in dictionary.
- *
- * Revision 1.17 1995/10/13 16:01:49 adam
- * Work on relations.
- *
- * Revision 1.16 1995/10/10 12:24:38 adam
- * Temporary sort files are compressed.
- *
- * Revision 1.15 1995/10/04 16:57:19 adam
- * Key input and merge sort in one pass.
- *
- * Revision 1.14 1995/09/29 14:01:40 adam
- * Bug fixes.
- *
- * Revision 1.13 1995/09/28 14:22:56 adam
- * Sort uses smaller temporary files.
- *
- * Revision 1.12 1995/09/28 12:10:32 adam
- * Bug fixes. Field prefix used in queries.
- *
- * Revision 1.11 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 1.10 1995/09/14 07:48:23 adam
- * Record control management.
- *
- * Revision 1.9 1995/09/11 13:09:33 adam
- * More work on relevance feedback.
- *
- * Revision 1.8 1995/09/08 14:52:27 adam
- * Minor changes. Dictionary is lower case now.
- *
- * Revision 1.7 1995/09/06 16:11:16 adam
- * Option: only one word key per file.
- *
- * Revision 1.6 1995/09/05 15:28:39 adam
- * More work on search engine.
- *
- * Revision 1.5 1995/09/04 12:33:42 adam
- * Various cleanup. YAZ util used instead.
- *
- * Revision 1.4 1995/09/04 09:10:35 adam
- * More work on index add/del/update.
- * Merge sort implemented.
- * Initial work on z39 server.
- *
- * Revision 1.3 1995/09/01 14:06:35 adam
- * Split of work into more files.
- *
- * Revision 1.2 1995/09/01 10:30:24 adam
- * More work on indexing. Not working yet.
- *
- * Revision 1.1 1995/08/31 14:50:24 adam
- * New simple file index tool.
- *
+ * Sebastian Hammer, Adam Dickmeiss, Heikki Levanto
+ * $Id: index.h,v 1.77 2002-04-05 08:46:26 adam Exp $
*/
+
+#ifndef INDEX_H
+#define INDEX_H
+
#include <time.h>
#include <zebraver.h>
#include <zebrautl.h>
#include <zebramap.h>
+#include <sortidx.h>
+
+#if HAVE_SYS_TIMES_H
+#include <sys/times.h>
+#endif
#include <dict.h>
+#include <isams.h>
#include <isam.h>
#include <isamc.h>
-#include <data1.h>
+#include <isamd.h>
+#define ISAM_DEFAULT "c"
+#include <yaz/data1.h>
+#include <recctrl.h>
+#include "recindex.h"
+#include "zebraapi.h"
+#include "zinfo.h"
+#include <passwddb.h>
+#include <rset.h>
+
+YAZ_BEGIN_CDECL
+
+#define SU_SCHEME 1
#define IT_MAX_WORD 256
#define IT_KEY_HAVE_SEQNO 1
time_t mtime;
};
-
-struct recordGroup {
- char *groupName;
- char *databaseName;
- char *path;
- char *recordId;
- char *recordType;
- int flagStoreData;
- int flagStoreKeys;
- int flagRw;
- int fileVerboseLimit;
- data1_handle dh;
- BFiles bfs;
- ZebraMaps zebra_maps;
-};
-
-void getFnameTmp (char *fname, int no);
+void getFnameTmp (Res res, char *fname, int no);
struct dirs_info *dirs_open (Dict dict, const char *rep, int rw);
struct dirs_info *dirs_fopen (Dict dict, const char *path);
void dirs_del (struct dirs_info *p, const char *src);
void dirs_free (struct dirs_info **pp);
-struct dir_entry *dir_open (const char *rep);
+struct dir_entry *dir_open (const char *rep, const char *base);
void dir_sort (struct dir_entry *e);
void dir_free (struct dir_entry **e_p);
-void repositoryUpdate (struct recordGroup *rGroup);
-void repositoryAdd (struct recordGroup *rGroup);
-void repositoryDelete (struct recordGroup *rGroup);
-void repositoryShow (struct recordGroup *rGroup);
+void repositoryUpdate (ZebraHandle zh);
+void repositoryAdd (ZebraHandle zh);
+void repositoryDelete (ZebraHandle zh);
+void repositoryShow (ZebraHandle zh);
-int key_open (BFiles bfs, int mem, int rw);
-int key_close (void);
+int key_open (ZebraHandle zh, int mem);
+int key_close (ZebraHandle zh);
int key_compare (const void *p1, const void *p2);
int key_get_pos (const void *p);
int key_compare_it (const void *p1, const void *p2);
int key_qsort_compare (const void *p1, const void *p2);
void key_logdump (int mask, const void *p);
-void inv_prstat (BFiles bfs);
-void key_input (BFiles bfs, int nkeys, int cache);
-ISAMC_M key_isamc_m (Res res);
+void inv_prstat (ZebraHandle zh);
+void inv_compact (BFiles bfs);
+void key_input (ZebraHandle zh, int nkeys, int cache, Res res);
+ISAMS_M key_isams_m (Res res, ISAMS_M me);
+ISAMC_M key_isamc_m (Res res, ISAMC_M me);
+ISAMD_M key_isamd_m (Res res, ISAMD_M me);
int merge_sort (char **buf, int from, int to);
+int key_SU_code (int ch, char *out);
#define FNAME_DICT "dict"
#define FNAME_ISAM "isam"
#define FNAME_ISAMC "isamc"
+#define FNAME_ISAMS "isams"
+#define FNAME_ISAMH "isamh"
+#define FNAME_ISAMD "isamd"
#define FNAME_CONFIG "zebra.cfg"
#define GMATCH_DICT "gmatch"
int index_word_prefix (char *string, int attset_ordinal,
int local_attribute, const char *databaseName);
-int fileExtract (SYSNO *sysno, const char *fname,
- const struct recordGroup *rGroup, int deleteFlag);
-void zebraIndexLockMsg (const char *str);
-void zebraIndexUnlock (void);
-void zebraIndexLock (BFiles bfs, int commitNow, const char *rval);
-int zebraIndexWait (int commitPhase);
+void zebraIndexLockMsg (ZebraHandle zh, const char *str);
+void zebraIndexUnlock (ZebraHandle zh);
+int zebraIndexLock (BFiles bfs, ZebraHandle zh, int commitNow, const char *rval);
+int zebraIndexWait (ZebraHandle zh, int commitPhase);
#define FNAME_MAIN_LOCK "zebraidx.LCK"
#define FNAME_COMMIT_LOCK "zebracmt.LCK"
#define FNAME_TOUCH_TIME "zebraidx.time"
typedef struct zebra_lock_info *ZebraLockHandle;
-ZebraLockHandle zebra_lock_create(const char *file, int excl_flag);
+ZebraLockHandle zebra_lock_create(const char *dir,
+ const char *file, int excl_flag);
void zebra_lock_destroy (ZebraLockHandle h);
int zebra_lock (ZebraLockHandle h);
int zebra_lock_nb (ZebraLockHandle h);
int zebra_unlock (ZebraLockHandle h);
int zebra_lock_fd (ZebraLockHandle h);
void zebra_lock_prefix (Res res, char *dst);
+char *zebra_mk_fname (const char *dir, const char *name);
+
+int zebra_lock_w (ZebraLockHandle h);
+int zebra_lock_r (ZebraLockHandle h);
+
+void zebra_load_atts (data1_handle dh, Res res);
+
+int key_SU_decode (int *ch, const unsigned char *out);
+int key_SU_encode (int ch, char *out);
+
+// extern Res common_resource;
+
+struct encode_info {
+ int sysno;
+ int seqno;
+ int cmd;
+ char buf[768];
+};
+
+void encode_key_init (struct encode_info *i);
+char *encode_key_int (int d, char *bp);
+void encode_key_write (char *k, struct encode_info *i, FILE *outf);
+
+typedef struct {
+ char *term;
+ char *db;
+ int sysno;
+ int score;
+} *ZebraPosSet;
+
+typedef struct zebra_set *ZebraSet;
+
+typedef struct zebra_rank_class {
+ struct rank_control *control;
+ int init_flag;
+ void *class_handle;
+ struct zebra_rank_class *next;
+} *ZebraRankClass;
+
+struct recKeys {
+ int buf_used;
+ int buf_max;
+ char *buf;
+ char prevAttrSet;
+ short prevAttrUse;
+ int prevSeqNo;
+};
+
+struct sortKey {
+ char *string;
+ int length;
+ int attrSet;
+ int attrUse;
+ struct sortKey *next;
+};
+
+struct zebra_register {
+ char *name;
+
+ ISAMS isams;
+ ISAM isam;
+ ISAMC isamc;
+ ISAMD isamd;
+ Dict dict;
+ Dict matchDict;
+ SortIdx sortIdx;
+ int registerState; /* 0 (no commit pages), 1 (use commit pages) */
+ time_t registerChange;
+ BFiles bfs;
+ Records records;
+ ZebraExplainInfo zei;
+
+ char *server_path_prefix;
+ data1_handle dh;
+ ZebraMaps zebra_maps;
+ ZebraRankClass rank_classes;
+ RecTypes recTypes;
+ int seqno;
+ int last_val;
+ int stop_flag;
+ int active; /* 0=shutdown, 1=enabled and inactive, 2=activated */
+
+
+
+ struct recKeys keys;
+ struct sortKey *sortKeys;
+
+ char **key_buf;
+ size_t ptr_top;
+ size_t ptr_i;
+ size_t key_buf_used;
+ int key_file_no;
+};
+
+struct zebra_service {
+ int stop_flag;
+ Res global_res;
+ char *configName;
+ struct zebra_session *sessions;
+ struct zebra_register *regs;
+ Zebra_mutex_cond session_lock;
+ Passwd_db passwd_db;
+ char *path_root;
+};
+
+
+struct zebra_session {
+ struct zebra_session *next;
+ struct zebra_service *service;
+ struct zebra_register *reg;
+
+ char *admin_databaseName;
+
+ char **basenames;
+ int num_basenames;
+ char *reg_name;
+ char *path_reg;
+
+ ZebraLockHandle lock_normal;
+ ZebraLockHandle lock_shadow;
+
+ int trans_no;
+ int destroyed;
+ ZebraSet sets;
+ Res res;
+ int errCode;
+ int hits;
+ char *errString;
+#if HAVE_SYS_TIMES_H
+ struct tms tms1;
+ struct tms tms2;
+#endif
+ struct recordGroup rGroup;
+};
+
+struct rank_control {
+ char *name;
+ void *(*create)(struct zebra_register *reg);
+ void (*destroy)(struct zebra_register *reg, void *class_handle);
+ void *(*begin)(struct zebra_register *reg, void *class_handle, RSET rset);
+ void (*end)(struct zebra_register *reg, void *set_handle);
+ int (*calc)(void *set_handle, int sysno);
+ void (*add)(void *set_handle, int seqno, int term_index);
+};
+
+struct term_set_entry {
+ char *term;
+ struct term_set_entry *next;
+};
+
+struct term_set_list {
+ struct term_set_entry *first;
+ struct term_set_entry *last;
+};
+
+RSET rpn_search (ZebraHandle zh, NMEM mem,
+ Z_RPNQuery *rpn, int num_bases, char **basenames,
+ const char *setname, ZebraSet sset);
+
+
+void rpn_scan (ZebraHandle zh, ODR stream, Z_AttributesPlusTerm *zapt,
+ oid_value attributeset,
+ int num_bases, char **basenames,
+ int *position, int *num_entries, ZebraScanEntry **list,
+ int *is_partial);
+
+RSET rset_trunc (ZebraHandle zh, ISAMS_P *isam_p, int no,
+ const char *term, int length_term, const char *flags);
+
+void resultSetAddTerm (ZebraHandle zh, ZebraSet s, int reg_type,
+ const char *db, int set,
+ int use, const char *term);
+ZebraSet resultSetAdd (ZebraHandle zh, const char *name, int ov);
+ZebraSet resultSetGet (ZebraHandle zh, const char *name);
+ZebraSet resultSetAddRPN (ZebraHandle zh, ODR stream, ODR decode,
+ Z_RPNQuery *rpn, int num_bases,
+ char **basenames, const char *setname);
+RSET resultSetRef (ZebraHandle zh, Z_ResultSetId *resultSetId);
+void resultSetDestroy (ZebraHandle zh, int num_names, char **names,
+ int *statuses);
+
+
+ZebraPosSet zebraPosSetCreate (ZebraHandle zh, const char *name,
+ int num, int *positions);
+void zebraPosSetDestroy (ZebraHandle zh, ZebraPosSet records, int num);
+
+void resultSetSort (ZebraHandle zh, NMEM nmem,
+ int num_input_setnames, const char **input_setnames,
+ const char *output_setname,
+ Z_SortKeySpecList *sort_sequence, int *sort_status);
+void resultSetSortSingle (ZebraHandle zh, NMEM nmem,
+ ZebraSet sset, RSET rset,
+ Z_SortKeySpecList *sort_sequence, int *sort_status);
+void resultSetRank (ZebraHandle zh, ZebraSet zebraSet, RSET rset);
+void resultSetInvalidate (ZebraHandle zh);
+
+int zebra_server_lock_init (ZebraService zh);
+int zebra_server_lock_destroy (ZebraService zh);
+int zebra_server_lock (ZebraService zh, int lockCommit);
+void zebra_server_unlock (ZebraService zh, int commitPhase);
+int zebra_server_lock_get_state (ZebraService zh, time_t *timep);
+
+typedef struct attent
+{
+ int attset_ordinal;
+ data1_local_attribute *local_attributes;
+} attent;
+
+void zebraRankInstall (struct zebra_register *reg, struct rank_control *ctrl);
+ZebraRankClass zebraRankLookup (ZebraHandle zh, const char *name);
+void zebraRankDestroy (struct zebra_register *reg);
+
+int att_getentbyatt(ZebraHandle zh, attent *res, oid_value set, int att);
+
+extern struct rank_control *rank1_class;
+
+int zebra_record_fetch (ZebraHandle zh, int sysno, int score, ODR stream,
+ oid_value input_format, Z_RecordComposition *comp,
+ oid_value *output_format, char **rec_bufp,
+ int *rec_lenp, char **basenamep);
+
+void extract_get_fname_tmp (ZebraHandle zh, char *fname, int no);
+void zebra_index_merge (ZebraHandle zh);
+
+
+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);
+
+void extract_flushWriteKeys (ZebraHandle zh);
+
+struct zebra_fetch_control {
+ off_t offset_end;
+ off_t record_offset;
+ off_t record_int_pos;
+ const char *record_int_buf;
+ int record_int_len;
+ int fd;
+};
+
+int zebra_record_ext_read (void *fh, char *buf, size_t count);
+off_t zebra_record_ext_seek (void *fh, off_t offset);
+off_t zebra_record_ext_tell (void *fh);
+off_t zebra_record_int_seek (void *fh, off_t offset);
+off_t zebra_record_int_tell (void *fh);
+int zebra_record_int_read (void *fh, char *buf, size_t count);
+void zebra_record_int_end (void *fh, off_t offset);
+
+void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno,
+ int cmd, struct recKeys *reckeys);
+void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno,
+ int cmd, struct sortKey **skp);
+void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid);
+void extract_token_add (RecWord *p);
+int explain_extract (void *handle, Record rec, data1_node *n);
+
+int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname,
+ const struct recordGroup *rGroup, int deleteFlag);
+YAZ_END_CDECL
-extern Res common_resource;
+#endif