From 7e75317bed8eecabcb57e59b16093a32238738e2 Mon Sep 17 00:00:00 2001 From: Adam Dickmeiss Date: Wed, 20 Feb 2002 17:30:01 +0000 Subject: [PATCH] Work on new API. Locking system re-implemented --- CHANGELOG | 6 + LICENSE.zebra | 2 +- README | 4 +- configure.in | 6 +- include/zebraver.h | 8 +- index/Makefile.am | 65 +- index/extract.c | 2980 +++++++++++++++++++++++-------------------------- index/index.h | 308 +---- index/invstat.c | 25 +- index/kinput.c | 252 +---- index/lockidx.c | 25 +- index/lockutil.c | 63 +- index/main.c | 216 +--- index/trav.c | 90 +- index/zebraapi.c | 1394 ++++++----------------- index/zebraapi.h | 59 +- index/zinfo.c | 80 +- index/zinfo.h | 6 +- index/zserver.c | 283 +---- index/zserver.h | 31 +- index/zsets.c | 20 +- recctrl/recgrs.c | 8 +- rset/rstemp.c | 110 +- test/Makefile.am | 2 +- test/api/Makefile.am | 9 + test/api/t1.c | 21 + test/api/t1.cfg | 2 + test/api/t2.c | 35 + test/api/t2.cfg | 4 + test/gils/Makefile.am | 4 + zebra.spec | 2 +- zmbol.spec | 2 +- 32 files changed, 2117 insertions(+), 4005 deletions(-) create mode 100644 test/api/Makefile.am create mode 100644 test/api/t1.c create mode 100644 test/api/t1.cfg create mode 100644 test/api/t2.c create mode 100644 test/api/t2.cfg diff --git a/CHANGELOG b/CHANGELOG index 29990f0..6fa36f5 100644 --- a/CHANGELOG +++ b/CHANGELOG @@ -1,4 +1,10 @@ +--- 1.2 2002/MM/DD + +New Zebra API. Locking system re-implemented. Beware! + +--- 1.1.stable 2002/02/20 + Rank weight can be controlled with attribute type 9. Default value is 34. Recommended values between 1-36. diff --git a/LICENSE.zebra b/LICENSE.zebra index b986a65..e66fe41 100644 --- a/LICENSE.zebra +++ b/LICENSE.zebra @@ -1,5 +1,5 @@ Zebra -Copyright (c) 1995-1999 Index Data ApS. +Copyright (c) 1995-2002 Index Data ApS. All rights reserved. diff --git a/README b/README index 9fe345e..bba6caa 100644 --- a/README +++ b/README @@ -1,8 +1,8 @@ -Copyright (C) 1995-2001, Index Data ApS. +Copyright (C) 1995-2002, Index Data ApS. See the file LICENSE.zebra for details. README for Zebra - $Id: README,v 1.5 2001-02-21 11:23:17 adam Exp $ + $Id: README,v 1.6 2002-02-20 17:30:01 adam Exp $ Documentation about this software can be found in the subdirectory 'doc' and Zebra's home page: http://www.indexdata.dk/zebra/ diff --git a/configure.in b/configure.in index 6f82b86..0f7e458 100644 --- a/configure.in +++ b/configure.in @@ -1,5 +1,5 @@ -dnl Zebra, Index Data Aps, 1994-2000 -dnl $Id: configure.in,v 1.29 2001-10-15 19:53:43 adam Exp $ +dnl Zebra, Index Data Aps, 1994-2002 +dnl $Id: configure.in,v 1.30 2002-02-20 17:30:01 adam Exp $ dnl AC_INIT(include/zebraver.h) AC_MSG_CHECKING(for package) @@ -152,5 +152,5 @@ AC_OUTPUT([ include/Makefile tab/Makefile doc/Makefile - test/Makefile test/gils/Makefile test/usmarc/Makefile + test/Makefile test/gils/Makefile test/usmarc/Makefile test/api/Makefile ]) diff --git a/include/zebraver.h b/include/zebraver.h index 3baa595..7b906f5 100644 --- a/include/zebraver.h +++ b/include/zebraver.h @@ -1,14 +1,14 @@ /* - * Copyright (C) 1994-2001, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * - * $Id: zebraver.h,v 1.18 2001-10-15 19:53:43 adam Exp $ + * $Id: zebraver.h,v 1.19 2002-02-20 17:30:01 adam Exp $ */ #ifndef ZEBRAVER -#define ZEBRAVER "1.1" +#define ZEBRAVER "1.1.1" #endif #ifndef ZEBRADATE -#define ZEBRADATE "$Date: 2001-10-15 19:53:43 $" +#define ZEBRADATE "$Date: 2002-02-20 17:30:01 $" #endif diff --git a/index/Makefile.am b/index/Makefile.am index db03d64..2f5d6ad 100644 --- a/index/Makefile.am +++ b/index/Makefile.am @@ -1,49 +1,44 @@ -## $Id: Makefile.am,v 1.6 2001-10-15 20:57:18 adam Exp $ +## $Id: Makefile.am,v 1.7 2002-02-20 17:30:01 adam Exp $ noinst_PROGRAMS = apitest kdump +noinst_LIBRARIES = libzebra.a + +libzebra_a_SOURCES = dir.c dirs.c trav.c kinput.c kcompare.c \ + attribute.c symtab.c recindex.c recstat.c lockutil.c \ + zebraapi.c zinfo.c invstat.c sortidx.c compact.c zsets.c zrpn.c \ + rank1.c trunc.c retrieve.c extract.c + +if ISZMBOL +zmbol_libs = ../isam/libisam.a ../isamc/libisamc.a ../isamb/libisamb.a +else +zmbol_libs = +endif + +libzebra_a_LIBADD = \ + ../rset/librset.a ../dict/libdict.a ../isams/libisams.a \ + $(zmbol_libs) ../recctrl/librecctrl.a ../bfile/libbfile.a ../dfa/libdfa.a \ + ../util/libutil.a + if ISZMBOL bin_PROGRAMS = zmbolidx zmbolsrv -zmbolidx_SOURCES = main.c dir.c dirs.c trav.c extract.c kinput.c kcompare.c \ - symtab.c recindex.c recstat.c lockutil.c lockidx.c \ - zinfo.c invstat.c sortidx.c compact.c \ - index.h recindex.h recindxp.h zebraapi.h zinfo.h zserver.h - -zmbolsrv_SOURCES = zserver.c kcompare.c zrpn.c zsets.c attribute.c recindex.c \ - lockutil.c locksrv.c zinfo.c trunc.c sortidx.c rank1.c zebraapi.c \ - retrieve.c kinput.c \ - index.h recindex.h recindxp.h zebraapi.h zinfo.h zserver.h else bin_PROGRAMS = zebraidx zebrasrv -zebraidx_SOURCES = main.c dir.c dirs.c trav.c extract.c kinput.c kcompare.c \ - symtab.c recindex.c recstat.c lockutil.c lockidx.c \ - zinfo.c invstat.c sortidx.c compact.c \ - index.h recindex.h recindxp.h zebraapi.h zinfo.h zserver.h - -zebrasrv_SOURCES = zserver.c kcompare.c zrpn.c zsets.c attribute.c recindex.c \ - lockutil.c locksrv.c zinfo.c trunc.c sortidx.c rank1.c zebraapi.c \ - retrieve.c kinput.c \ - index.h recindex.h recindxp.h zebraapi.h zinfo.h zserver.h endif -apitest_SOURCES = apitest.c kcompare.c zrpn.c zsets.c attribute.c recindex.c \ - lockutil.c locksrv.c zinfo.c trunc.c sortidx.c rank1.c zebraapi.c \ - retrieve.c kinput.c \ - index.h recindex.h recindxp.h zebraapi.h zinfo.h zserver.h - +zmbolidx_SOURCES = main.c +zmbolsrv_SOURCES = zserver.c +zebraidx_SOURCES = main.c +zebrasrv_SOURCES = zserver.c +apitest_SOURCES = apitest.c kdump_SOURCES=kdump.c kcompare.c CFLAGS = -I$(srcdir)/../include $(YAZINC) $(TCL_INCLUDE) -if ISZMBOL -zmbol_libs = ../isam/libisam.a ../isamc/libisamc.a ../isamb/libisamb.a -else -zmbol_libs = -endif +LDADD = libzebra.a $(YAZLIB) $(TCL_LIB) -LDADD = \ - ../rset/librset.a ../dict/libdict.a ../isams/libisams.a \ - $(zmbol_libs) \ - ../recctrl/librecctrl.a ../bfile/libbfile.a ../dfa/libdfa.a \ - ../util/libutil.a \ - $(YAZLIB) $(TCL_LIB) +libzebra.a: $(libzebra_a_OBJECTS) $(libzebra_a_DEPENDENCIES) + -test ! -d atmp && mkdir atmp + mkdir atmp; for i in $(libzebra_a_LIBADD); do cd atmp; ar x ../$$i; cd ..; done + ar qc libzebra.a $(libzebra_a_OBJECTS) libzebra.a atmp/*.o + $(RANLIB) libzebra.a diff --git a/index/extract.c b/index/extract.c index 86de3b2..a726916 100644 --- a/index/extract.c +++ b/index/extract.c @@ -1,392 +1,9 @@ /* - * Copyright (C) 1994-2001, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * - * $Log: extract.c,v $ - * Revision 1.109 2001-10-15 19:53:43 adam - * POSIX thread updates. First work on term sets. - * - * Revision 1.108 2001/06/14 11:44:56 adam - * Bug fix: default storeKeys setting wasn't read when group was specified. - * - * Revision 1.107 2001/05/28 13:58:48 adam - * Call flushSortKeys when record is skipped to fix bad re-use of - * sort keys to whatever next record that comes in. - * - * Revision 1.106 2000/12/05 12:22:53 adam - * Termlist source implemented (so that we can index values of XML/SGML - * attributes). - * - * Revision 1.105 2000/12/05 10:01:44 adam - * Fixed bug regarding user-defined attribute sets. - * - * Revision 1.104 2000/09/05 14:04:05 adam - * Updates for prefix 'yaz_' for YAZ log functions. - * - * Revision 1.103 2000/05/18 12:01:36 adam - * System call times(2) used again. More 64-bit fixes. - * - * Revision 1.102 2000/05/15 15:32:33 adam - * Added 64 bit file input. - * - * Revision 1.101 2000/05/15 13:02:39 adam - * Minor change. - * - * Revision 1.100 2000/03/20 19:08:36 adam - * Added remote record import using Z39.50 extended services and Segment - * Requests. - * - * Revision 1.99 2000/02/24 10:57:02 adam - * Sequence number incremented after each incomplete-field. - * - * Revision 1.98 1999/09/07 07:19:21 adam - * Work on character mapping. Implemented replace rules. - * - * Revision 1.97 1999/07/06 12:28:04 adam - * Updated record index structure. Format includes version ID. Compression - * algorithm ID is stored for each record block. - * - * Revision 1.96 1999/05/26 07:49:13 adam - * C++ compilation. - * - * Revision 1.95 1999/05/21 12:00:17 adam - * Better diagnostics for extraction process. - * - * Revision 1.94 1999/05/20 12:57:18 adam - * Implemented TCL filter. Updated recctrl system. - * - * Revision 1.93 1999/05/15 14:36:38 adam - * Updated dictionary. Implemented "compression" of dictionary. - * - * Revision 1.92 1999/03/09 16:27:49 adam - * More work on SDRKit integration. - * - * Revision 1.91 1999/02/12 13:29:22 adam - * Implemented position-flag for registers. - * - * Revision 1.90 1999/02/02 14:50:52 adam - * Updated WIN32 code specific sections. Changed header. - * - * Revision 1.89 1998/10/28 10:54:38 adam - * SDRKit integration. - * - * Revision 1.88 1998/10/16 08:14:29 adam - * Updated record control system. - * - * Revision 1.87 1998/10/15 13:10:33 adam - * Fixed bug in Zebra that caused it to stop indexing when empty - * record was read. - * - * Revision 1.86 1998/10/13 20:33:53 adam - * Fixed one log message and change use ordinal to be an unsigned char. - * - * Revision 1.85 1998/09/22 10:03:41 adam - * Changed result sets to be persistent in the sense that they can - * be re-searched if needed. - * Fixed memory leak in rsm_or. - * - * Revision 1.84 1998/06/11 15:42:22 adam - * Changed the way use attributes are specified in the recordId - * specification. - * - * Revision 1.83 1998/06/08 14:43:10 adam - * Added suport for EXPLAIN Proxy servers - added settings databasePath - * and explainDatabase to facilitate this. Increased maximum number - * of databases and attributes in one register. - * - * Revision 1.82 1998/05/20 10:12:15 adam - * Implemented automatic EXPLAIN database maintenance. - * Modified Zebra to work with ASN.1 compiled version of YAZ. - * - * Revision 1.81 1998/03/11 11:19:04 adam - * Changed the way sequence numbers are generated. - * - * Revision 1.80 1998/03/05 08:45:11 adam - * New result set model and modular ranking system. Moved towards - * descent server API. System information stored as "SGML" records. - * - * Revision 1.79 1998/02/17 10:32:52 adam - * Fixed bug: binary files weren't opened with flag b on NT. - * - * Revision 1.78 1998/02/10 12:03:05 adam - * Implemented Sort. - * - * Revision 1.77 1998/01/12 15:04:08 adam - * The test option (-s) only uses read-lock (and not write lock). - * - * Revision 1.76 1997/10/27 14:33:04 adam - * Moved towards generic character mapping depending on "structure" - * field in abstract syntax file. Fixed a few memory leaks. Fixed - * bug with negative integers when doing searches with relational - * operators. - * - * Revision 1.75 1997/09/17 12:19:12 adam - * Zebra version corresponds to YAZ version 1.4. - * Changed Zebra server so that it doesn't depend on global common_resource. - * - * Revision 1.74 1997/09/09 13:38:06 adam - * Partial port to WIN95/NT. - * - * Revision 1.73 1997/09/04 13:57:20 adam - * New file extract/retrieve method tellf (added). - * Added O_BINARY for open calls. - * - * Revision 1.72 1997/07/15 16:32:29 adam - * Bug fix: Match handler didn't terminate the resulting string! - * - * Revision 1.71 1997/07/15 16:28:41 adam - * Bug fix: storeData didn't work with files with multiple records. - * Bug fix: fixed memory management with records; not really well - * thought through. - * - * Revision 1.70 1997/07/01 13:00:42 adam - * Bug fix in routine searchRecordKey: uninitialized variables. - * - * Revision 1.69 1997/04/29 09:26:03 adam - * Bug fix: generic recordId handling didn't work for compressed internal - * keys. - * - * Revision 1.68 1997/02/12 20:39:45 adam - * Implemented options -f that limits the log to the first - * records. - * Changed some log messages also. - * - * Revision 1.67 1996/11/15 15:02:14 adam - * Minor changes regarding logging. - * - * Revision 1.66 1996/11/14 09:52:21 adam - * Strings in record keys bound by IT_MAX_WORD. - * - * Revision 1.65 1996/11/14 08:57:56 adam - * Reduction of storeKeys area. - * - * Revision 1.64 1996/11/08 11:10:16 adam - * Buffers used during file match got bigger. - * Compressed ISAM support everywhere. - * Bug fixes regarding masking characters in queries. - * Redesigned Regexp-2 queries. - * - * Revision 1.63 1996/10/29 14:09:39 adam - * Use of cisam system - enabled if setting isamc is 1. - * - * Revision 1.62 1996/10/11 10:57:01 adam - * New module recctrl. Used to manage records (extract/retrieval). - * Several files have been moved to the recctrl sub directory. - * - * Revision 1.61 1996/06/06 12:08:37 quinn - * Added showRecord function - * - * Revision 1.60 1996/06/04 10:18:12 adam - * Search/scan uses character mapping module. - * - * Revision 1.59 1996/05/14 15:47:07 adam - * Cleanup of various buffer size entities. - * - * Revision 1.58 1996/05/14 06:16:38 adam - * Compact use/set bytes used in search service. - * - * Revision 1.57 1996/05/13 14:23:04 adam - * Work on compaction of set/use bytes in dictionary. - * - * Revision 1.56 1996/05/09 09:54:42 adam - * Server supports maps from one logical attributes to a list of physical - * attributes. - * The extraction process doesn't make space consuming 'any' keys. - * - * Revision 1.55 1996/05/09 07:28:55 quinn - * Work towards phrases and multiple registers - * - * Revision 1.54 1996/05/01 13:46:35 adam - * First work on multiple records in one file. - * New option, -offset, to the "unread" command in the filter module. - * - * Revision 1.53 1996/04/26 12:09:43 adam - * Added a few comments. - * - * Revision 1.52 1996/04/25 13:27:57 adam - * Function recordExtract modified so that files with no keys (possibly empty) - * are ignored. - * - * Revision 1.51 1996/03/19 11:08:42 adam - * Bug fix: Log preamble wasn't always turned off after recordExtract. - * - * Revision 1.50 1996/02/12 18:45:36 adam - * New fileVerboseFlag in record group control. - * - * Revision 1.49 1996/02/05 12:29:57 adam - * Logging reduced a bit. - * The remaining running time is estimated during register merge. - * - * Revision 1.48 1996/02/01 20:53:26 adam - * The temporary per-record keys are compacted a little, and duplication - * of the per-records keys are avoided when they are saved in the record - * information buffer. - * - * Revision 1.47 1996/01/17 14:57:48 adam - * Prototype changed for reader functions in extract/retrieve. File - * is identified by 'void *' instead of 'int. - * - * Revision 1.46 1995/12/15 14:57:16 adam - * Bug fix. - * - * Revision 1.45 1995/12/15 12:37:41 adam - * In addRecordKeyAny: Writes key only when attrSet != -1. - * - * Revision 1.44 1995/12/12 16:00:54 adam - * System call sync(2) used after update/commit. - * Locking (based on fcntl) uses F_EXLCK and F_SHLCK instead of F_WRLCK - * and F_RDLCK. - * - * Revision 1.43 1995/12/11 09:12:46 adam - * The rec_get function returns NULL if record doesn't exist - will - * happen in the server if the result set records have been deleted since - * the creation of the set (i.e. the search). - * The server saves a result temporarily if it is 'volatile', i.e. the - * set is register dependent. - * - * Revision 1.42 1995/12/07 17:38:46 adam - * Work locking mechanisms for concurrent updates/commit. - * - * Revision 1.41 1995/12/06 16:06:42 adam - * Better diagnostics. Work on 'real' dictionary deletion. - * - * Revision 1.40 1995/12/05 16:57:40 adam - * More work on regular patterns. - * - * Revision 1.39 1995/12/05 13:20:18 adam - * Bug fix: file_read sometimes returned early EOF. - * - * Revision 1.38 1995/12/04 17:59:21 adam - * More work on regular expression conversion. - * - * Revision 1.37 1995/12/04 14:22:27 adam - * Extra arg to recType_byName. - * Started work on new regular expression parsed input to - * structured records. - * - * Revision 1.36 1995/11/30 08:34:29 adam - * Started work on commit facility. - * Changed a few malloc/free to xmalloc/xfree. - * - * Revision 1.35 1995/11/28 14:26:21 adam - * Bug fix: recordId with constant wasn't right. - * Bug fix: recordId dictionary entry wasn't deleted when needed. - * - * Revision 1.34 1995/11/28 09:09:38 adam - * Zebra config renamed. - * Use setting 'recordId' to identify record now. - * Bug fix in recindex.c: rec_release_blocks was invokeded even - * though the blocks were already released. - * File traversal properly deletes records when needed. - * - * Revision 1.33 1995/11/27 09:56:20 adam - * Record info elements better enumerated. Internal store of records. - * - * Revision 1.32 1995/11/25 10:24:05 adam - * More record fields - they are enumerated now. - * New options: flagStoreData flagStoreKey. - * - * Revision 1.31 1995/11/24 11:31:35 adam - * Commands add & del read filenames from stdin if source directory is - * empty. - * Match criteria supports 'constant' strings. - * - * Revision 1.30 1995/11/22 17:19:16 adam - * Record management uses the bfile system. - * - * Revision 1.29 1995/11/21 15:01:14 adam - * New general match criteria implemented. - * New feature: document groups. - * - * Revision 1.28 1995/11/21 09:20:30 adam - * Yet more work on record match. - * - * Revision 1.27 1995/11/20 16:59:45 adam - * New update method: the 'old' keys are saved for each records. - * - * Revision 1.26 1995/11/20 11:56:24 adam - * Work on new traversal. - * - * Revision 1.25 1995/11/16 15:34:54 adam - * Uses new record management system in both indexer and server. - * - * Revision 1.24 1995/11/15 19:13:08 adam - * Work on record management. - * - * Revision 1.23 1995/10/27 14:00:10 adam - * Implemented detection of database availability. - * - * Revision 1.22 1995/10/17 18:02:07 adam - * New feature: databases. Implemented as prefix to words in dictionary. - * - * Revision 1.21 1995/10/10 12:24:38 adam - * Temporary sort files are compressed. - * - * Revision 1.20 1995/10/06 13:52:05 adam - * Bug fixes. Handler may abort further scanning. - * - * Revision 1.19 1995/10/04 12:55:16 adam - * Bug fix in ranked search. Use=Any keys inserted. - * - * Revision 1.18 1995/10/04 09:37:08 quinn - * Fixed bug. - * - * Revision 1.17 1995/10/03 14:28:57 adam - * Buffered read in extract works. - * - * Revision 1.16 1995/10/03 14:28:45 adam - * Work on more effecient read handler in extract. - * - * Revision 1.15 1995/10/02 15:42:53 adam - * Extract uses file descriptors instead of FILE pointers. - * - * Revision 1.14 1995/10/02 15:29:13 adam - * More logging in file_extract. - * - * Revision 1.13 1995/09/29 14:01:39 adam - * Bug fixes. - * - * Revision 1.12 1995/09/28 14:22:56 adam - * Sort uses smaller temporary files. - * - * Revision 1.11 1995/09/28 12:10:31 adam - * Bug fixes. Field prefix used in queries. - * - * Revision 1.10 1995/09/28 09:19:41 adam - * xfree/xmalloc used everywhere. - * Extract/retrieve method seems to work for text records. - * - * Revision 1.9 1995/09/27 12:22:28 adam - * More work on extract in record control. - * Field name is not in isam keys but in prefix in dictionary words. - * - * Revision 1.8 1995/09/14 07:48:22 adam - * Record control management. - * - * Revision 1.7 1995/09/11 13:09:32 adam - * More work on relevance feedback. - * - * Revision 1.6 1995/09/08 14:52:27 adam - * Minor changes. Dictionary is lower case now. - * - * Revision 1.5 1995/09/06 16:11:16 adam - * Option: only one word key per file. - * - * Revision 1.4 1995/09/05 15:28:39 adam - * More work on search engine. - * - * Revision 1.3 1995/09/04 12:33:41 adam - * Various cleanup. YAZ util used instead. - * - * Revision 1.2 1995/09/04 09:10:34 adam - * More work on index add/del/update. - * Merge sort implemented. - * Initial work on z39 server. - * - * Revision 1.1 1995/09/01 14:06:35 adam - * Split of work into more files. - * + * $Id: extract.c,v 1.110 2002-02-20 17:30:01 adam Exp $ */ #include #include @@ -401,7 +18,7 @@ #include #include #include "index.h" - +#include "zserver.h" #include "zinfo.h" #if _FILE_OFFSET_BITS == 64 @@ -418,25 +35,11 @@ #include "zebrasdr.h" #endif -static Dict matchDict; - -static Records records = NULL; -static SortIdx sortIdx = NULL; - -static char **key_buf; -static size_t ptr_top; -static size_t ptr_i; -static size_t key_buf_used; -static int key_file_no; - static int records_inserted = 0; static int records_updated = 0; static int records_deleted = 0; static int records_processed = 0; -static ZebraExplainInfo zti = NULL; - - static void logRecord (int showFlag) { if (!showFlag) @@ -449,1412 +52,1609 @@ static void logRecord (int showFlag) } } -static int explain_extract (void *handle, Record drec, data1_node *n); +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; +} -int key_open (struct recordGroup *rGroup, int mem) +static const char **searchRecordKey (ZebraHandle zh, + struct recKeys *reckeys, + int attrSetS, int attrUseS) { - BFiles bfs = rGroup->bfs; - int rw = rGroup->flagRw; - data1_handle dh = rGroup->dh; - char *recordCompression; - int record_compression = REC_COMPRESS_NONE; - if (!mem) - mem = atoi(res_get_def (common_resource, "memMax", "16"))*1024*1024; - if (mem < 50000) - mem = 50000; - key_buf = (char **) xmalloc (mem); - ptr_top = mem/sizeof(char*); - ptr_i = 0; - - key_buf_used = 0; - key_file_no = 0; - - if (!(matchDict = dict_open (bfs, GMATCH_DICT, 50, rw, 0))) - { - logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT); - return -1; - } - assert (!records); - recordCompression = res_get_def (common_resource, - "recordCompression", "none"); - if (!strcmp (recordCompression, "none")) - record_compression = REC_COMPRESS_NONE; - if (!strcmp (recordCompression, "bzip2")) - record_compression = REC_COMPRESS_BZIP2; - records = rec_open (bfs, rw, record_compression); - if (!records) - { - dict_close (matchDict); - return -1; - } - zti = zebraExplain_open (records, dh, common_resource, - rw, rGroup, explain_extract); - if (!zti) + static const char *ws[32]; + int off = 0; + int startSeq = -1; + int i; + int seqno = 0; +#if SU_SCHEME + int chS, ch; +#else + short attrUse; + char attrSet; +#endif + + for (i = 0; i<32; i++) + ws[i] = NULL; + +#if SU_SCHEME + chS = zebraExplain_lookupSU (zh->service->zei, attrSetS, attrUseS); + if (chS < 0) + return ws; +#endif + while (off < reckeys->buf_used) { - rec_close (&records); - dict_close (matchDict); - return -1; + + 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 + attrUseS == attrUse && attrSetS == attrSet +#endif + ) + { + int woff; + + + if (startSeq == -1) + startSeq = seqno; + woff = seqno - startSeq; + if (woff >= 0 && woff < 31) + ws[woff] = wstart; + } + + off = src - reckeys->buf; } - sortIdx = sortIdx_open (bfs, 1); - return 0; + assert (off == reckeys->buf_used); + return ws; } -struct encode_info { - int sysno; - int seqno; - int cmd; - char buf[768]; +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; }; -void encode_key_init (struct encode_info *i) +static struct file_read_info *file_read_start (int fd) { - i->sysno = 0; - i->seqno = 0; - i->cmd = -1; + 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; } -char *encode_key_int (int d, char *bp) +static void file_read_stop (struct file_read_info *fi) { - if (d <= 63) - *bp++ = d; - else if (d <= 16383) - { - *bp++ = 64 + (d>>8); - *bp++ = d & 255; - } - else if (d <= 4194303) - { - *bp++ = 128 + (d>>16); - *bp++ = (d>>8) & 255; - *bp++ = d & 255; - } - else - { - *bp++ = 192 + (d>>24); - *bp++ = (d>>16) & 255; - *bp++ = (d>>8) & 255; - *bp++ = d & 255; - } - return bp; + xfree (fi); } -void encode_key_write (char *k, struct encode_info *i, FILE *outf) +static off_t file_seek (void *handle, off_t offset) { - struct it_key key; - char *bp = i->buf; - - 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) + 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) { - i->sysno = key.sysno; - i->seqno = 0; + 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 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) + else + r = read (fd, buf, count); + if (r > 0) { - logf (LOG_FATAL|LOG_ERRNO, "fwrite"); - exit (1); + p->file_offset += r; + if (p->file_offset > p->file_max) + p->file_max = p->file_offset; } + return r; } -#define SORT_EXTRA 0 - -#if SORT_EXTRA -static int key_y_len; - -static int key_y_compare (const void *p1, const void *p2) +static void file_begin (void *handle) { - int r; + struct file_read_info *p = (struct file_read_info *) handle; - if ((r = key_compare (*(char**) p1 + key_y_len + 1, - *(char**) p2 + key_y_len + 1))) - return r; - return *(*(char**) p1 + key_y_len) - *(*(char**) p2 + key_y_len); + 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 int key_x_compare (const void *p1, const void *p2) +static void file_end (void *handle, off_t offset) { - return strcmp (*(char**) p1, *(char**) p2); + struct file_read_info *p = (struct file_read_info *) handle; + + assert (p->file_more == 0); + p->file_more = 1; + p->file_moffset = offset; } -#endif -void key_flush (void) +static char *fileMatchStr (ZebraHandle zh, + struct recKeys *reckeys, struct recordGroup *rGroup, + const char *fname, const char *spec) { - FILE *outf; - char out_fname[200]; - char *prevcp, *cp; - struct encode_info encode_info; -#if SORT_EXTRA - int i; -#endif - - if (ptr_i <= 0) - return; - - key_file_no++; - logf (LOG_LOG, "sorting section %d", key_file_no); -#if !SORT_EXTRA - qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare); - getFnameTmp (common_resource, out_fname, key_file_no); + static char dstBuf[2048]; /* static here ??? */ + char *dst = dstBuf; + const char *s = spec; + static const char **w; - 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); - prevcp = cp = key_buf[ptr_top-ptr_i]; - - encode_key_init (&encode_info); - encode_key_write (cp, &encode_info, outf); - while (--ptr_i > 0) + while (1) { - cp = key_buf[ptr_top-ptr_i]; - if (strcmp (cp, prevcp)) + while (*s == ' ' || *s == '\t') + s++; + if (!*s) + break; + if (*s == '(') { - encode_key_init (&encode_info); - encode_key_write (cp, &encode_info, outf); - prevcp = cp; + 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->service->dh, attset_str))) + { + data1_att *att; + attSet = attset->reference; + att = data1_getattbyname(zh->service->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 - encode_key_write (cp + strlen(cp), &encode_info, outf); - } -#else - qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_x_compare); - getFnameTmp (out_fname, key_file_no); + else if (*s == '$') + { + int spec_len; + char special[64]; + const char *spec_src = NULL; + const char *s1 = ++s; + while (*s1 && *s1 != ' ' && *s1 != '\t') + s1++; - 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])) + 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 == '\'') { - 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) + int stopMarker = *s++; + char tmpString[64]; + int i = 0; + + while (*s && *s != stopMarker) { - cp = key_buf[ptr_top-ptr_i]; - encode_key_write (cp+key_y_len, &encode_info, outf); + if (i < 63) + tmpString[i++] = *s++; } - if (!i) - break; - prevcp = key_buf[ptr_top-ptr_i]; + if (*s) + s++; + tmpString[i] = '\0'; + strcpy (dst, tmpString); + dst += strlen (tmpString); } -#endif - if (fclose (outf)) + 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_FATAL|LOG_ERRNO, "fclose %s", out_fname); - exit (1); + logf (LOG_WARN, "No match criteria for record %s in group %s", + fname, rGroup->groupName ? rGroup->groupName : "none"); + return NULL; } - logf (LOG_LOG, "finished section %d", key_file_no); - ptr_i = 0; - key_buf_used = 0; -} - -int key_close (struct recordGroup *rGroup) -{ - int rw = rGroup->flagRw; - if (rw) - zebraExplain_runNumberIncrement (zti, 1); - zebraExplain_close (zti, rw); - key_flush (); - xfree (key_buf); - rec_close (&records); - dict_close (matchDict); - sortIdx_close (sortIdx); - - logRecord (1); - return key_file_no; + *dst = '\0'; + return dstBuf; } -static void wordInit (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 struct sortKey { - char *string; - int length; - int attrSet; - int attrUse; - struct sortKey *next; -} *sortKeys = NULL; - -static struct recKeys { - int buf_used; - int buf_max; - char *buf; - char prevAttrSet; - short prevAttrUse; - int prevSeqNo; -} reckeys; - -static void addIndexString (RecWord *p, const char *string, int length) -{ - char *dst; - unsigned char attrSet; - unsigned short attrUse; - int lead = 0; - int diff = 0; - int *pseqno = &p->seqnos[p->reg_type]; - - if (reckeys.buf_used+1024 > reckeys.buf_max) - { - char *b; - - b = (char *) xmalloc (reckeys.buf_max += 128000); - if (reckeys.buf_used > 0) - memcpy (b, reckeys.buf, reckeys.buf_used); - xfree (reckeys.buf); - reckeys.buf = b; - } - dst = reckeys.buf + reckeys.buf_used; - - attrSet = p->attrSet; - if (reckeys.buf_used > 0 && reckeys.prevAttrSet == attrSet) - lead |= 1; - else - reckeys.prevAttrSet = attrSet; - attrUse = p->attrUse; - if (reckeys.buf_used > 0 && reckeys.prevAttrUse == attrUse) - lead |= 2; - else - reckeys.prevAttrUse = attrUse; -#if 1 - diff = 1 + *pseqno - reckeys.prevSeqNo; - if (diff >= 1 && diff <= 15) - lead |= (diff << 2); - else - diff = 0; -#endif - reckeys.prevSeqNo = *pseqno; - - *dst++ = lead; - -#if SU_SCHEME - if ((lead & 3) < 3) - { - int ch = zebraExplain_lookupSU (zti, attrSet, attrUse); - if (ch < 0) - { - ch = zebraExplain_addSU (zti, attrSet, attrUse); - } - 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'; - - if (!diff) - { - memcpy (dst, pseqno, sizeof(*pseqno)); - dst += sizeof(*pseqno); - } - reckeys.buf_used = dst - reckeys.buf; - if (*pseqno) - (*pseqno)++; -} - -static void addSortString (RecWord *p, const char *string, int length) +struct recordLogInfo { + const char *fname; + int recordOffset; + struct recordGroup *rGroup; +}; + +static void recordLogPreamble (int level, const char *msg, void *info) { - struct sortKey *sk; - - for (sk = sortKeys; sk; sk = sk->next) - if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse) - return; - - sk = (struct sortKey *) xmalloc (sizeof(*sk)); - sk->next = sortKeys; - sortKeys = sk; - - sk->string = (char *) xmalloc (length); - sk->length = length; - memcpy (sk->string, string, length); + struct recordLogInfo *p = (struct recordLogInfo *) info; + FILE *outf = yaz_log_file (); - sk->attrSet = p->attrSet; - sk->attrUse = p->attrUse; + 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 void addString (RecWord *p, const char *string, int length) -{ - assert (length > 0); - if (zebra_maps_is_sort (p->zebra_maps, p->reg_type)) - addSortString (p, string, length); - else - addIndexString (p, string, length); -} -static void addIncompleteField (RecWord *p) +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) { - const char *b = p->string; - int remain = p->length; - const char **map = 0; - - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + RecordAttr *recordAttr; + int r; + char *matchStr; + SYSNO sysnotmp; + Record rec; + struct recordLogInfo logInfo; + off_t recordOffset = 0; - while (map) + if (fi->fd != -1) { - char buf[IT_MAX_WORD+1]; - int i, remain; + struct recExtractCtrl extractCtrl; - /* Skip spaces */ - while (map && *map && **map == *CHR_SPACE) + /* we are going to read from a file, so prepare the extraction */ + int i; + + zh->keys.buf_used = 0; + zh->keys.prevAttrUse = -1; + zh->keys.prevAttrSet = -1; + zh->keys.prevSeqNo = 0; + + recordOffset = fi->file_moffset; + extractCtrl.offset = fi->file_moffset; + extractCtrl.readf = file_read; + extractCtrl.seekf = file_seek; + extractCtrl.tellf = file_tell; + extractCtrl.endf = file_end; + extractCtrl.fh = fi; + extractCtrl.subType = subType; + extractCtrl.init = extract_init; + extractCtrl.tokenAdd = extract_token_add; + extractCtrl.schemaAdd = extract_schema_add; + extractCtrl.dh = zh->service->dh; + extractCtrl.handle = zh; + for (i = 0; i<256; i++) { - remain = p->length - (b - p->string); - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + if (zebra_maps_is_positioned(zh->service->zebra_maps, i)) + extractCtrl.seqno[i] = 1; else - map = 0; + extractCtrl.seqno[i] = 0; } - if (!map) - break; - i = 0; - while (map && *map && **map != *CHR_SPACE) - { - const char *cp = *map; + extractCtrl.zebra_maps = zh->service->zebra_maps; + extractCtrl.flagShowRecords = !rGroup->flagRw; - while (i < IT_MAX_WORD && *cp) - buf[i++] = *(cp++); - remain = p->length - (b - p->string); - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); - else - map = 0; - } - if (!i) - return; - addString (p, buf, i); - } - (p->seqnos[p->reg_type])++; /* to separate this from next one */ -} + if (!rGroup->flagRw) + printf ("File: %s " PRINTF_OFF_T "\n", fname, recordOffset); -static void addCompleteField (RecWord *p) -{ - const char *b = p->string; - char buf[IT_MAX_WORD+1]; - const char **map = 0; - int i = 0, remain = p->length; + logInfo.fname = fname; + logInfo.recordOffset = recordOffset; + logInfo.rGroup = rGroup; + log_event_start (recordLogPreamble, &logInfo); - if (remain > 0) - map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain); + r = (*recType->extract)(clientData, &extractCtrl); - while (remain > 0 && i < IT_MAX_WORD) - { - while (map && *map && **map == *CHR_SPACE) - { - remain = p->length - (b - p->string); - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); - else - map = 0; - } - if (!map) - break; + log_event_start (NULL, NULL); - if (i && i < IT_MAX_WORD) - buf[i++] = *CHR_SPACE; - while (map && *map && **map != *CHR_SPACE) + if (r == RECCTRL_EXTRACT_EOF) + return 0; + else if (r == RECCTRL_EXTRACT_ERROR) { - const char *cp = *map; - - if (i >= IT_MAX_WORD) - break; - while (i < IT_MAX_WORD && *cp) - buf[i++] = *(cp++); - remain = p->length - (b - p->string); - if (remain > 0) - map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, - remain); - else - map = 0; - } + /* 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->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; + } } - if (!i) - return; - addString (p, buf, i); -} -static void addRecordKey (RecWord *p) -{ - WRBUF wrbuf; - if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0, - p->string, p->length))) + /* perform match if sysno not known and if match criteria is specified */ + + matchStr = NULL; + if (!sysno) { - p->string = wrbuf_buf(wrbuf); - p->length = wrbuf_len(wrbuf); + sysnotmp = 0; + sysno = &sysnotmp; + if (rGroup->recordId && *rGroup->recordId) + { + char *rinfo; + + matchStr = fileMatchStr (zh, &zh->keys, rGroup, fname, + rGroup->recordId); + if (matchStr) + { + rinfo = dict_lookup (zh->service->matchDict, matchStr); + if (rinfo) + memcpy (sysno, rinfo+1, sizeof(*sysno)); + } + else + { + logf (LOG_WARN, "Bad match criteria"); + return 0; + } + } } - if (zebra_maps_is_complete (p->zebra_maps, p->reg_type)) - addCompleteField (p); - else - addIncompleteField(p); -} -static void flushSortKeys (SYSNO sysno, int cmd) -{ - struct sortKey *sk = sortKeys; - - sortIdx_sysno (sortIdx, sysno); - while (sk) + if (! *sysno) { - struct sortKey *sk_next = sk->next; - if (cmd >= 0) - { - sortIdx_type (sortIdx, sk->attrUse); - sortIdx_add (sortIdx, sk->string, sk->length); - } - xfree (sk->string); - xfree (sk); - sk = sk_next; - } - sortKeys = NULL; -} + /* 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->service->records); -static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys) -{ -#if SU_SCHEME -#else - unsigned char attrSet = (unsigned char) -1; - unsigned short attrUse = (unsigned short) -1; -#endif - int seqno = 0; - int off = 0; - int ch = 0; + *sysno = rec->sysno; - zebraExplain_recordCountIncrement (zti, cmd ? 1 : -1); - while (off < reckeys->buf_used) + recordAttr = rec_init_attr (zh->service->zei, rec); + + if (matchStr) + { + dict_insert (zh->service->matchDict, matchStr, sizeof(*sysno), sysno); + } + extract_flushRecordKeys (zh, *sysno, 1, &zh->keys); + extract_flushSortKeys (zh, *sysno, 1, &zh->sortKeys); + + records_inserted++; + } + else { - const char *src = reckeys->buf + off; - struct it_key key; - int lead; - - lead = *src++; + /* record already exists */ + struct recKeys delkeys; -#if SU_SCHEME - if ((lead & 3) < 3) + rec = rec_get (zh->service->records, *sysno); + assert (rec); + + recordAttr = rec_init_attr (zh->service->zei, rec); + + if (recordAttr->runNumber == + zebraExplain_runNumberIncrement (zh->service->zei, 0)) { - memcpy (&ch, src, sizeof(ch)); - src += sizeof(ch); + logf (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, rGroup->recordType, + fname, recordOffset); + extract_flushSortKeys (zh, *sysno, -1, &zh->sortKeys); + rec_rm (&rec); + logRecord (0); + return 1; } -#else - if (!(lead & 1)) + delkeys.buf_used = rec->size[recInfo_delKeys]; + delkeys.buf = rec->info[recInfo_delKeys]; + extract_flushSortKeys (zh, *sysno, 0, &zh->sortKeys); + extract_flushRecordKeys (zh, *sysno, 0, &delkeys); + if (deleteFlag) { - memcpy (&attrSet, src, sizeof(attrSet)); - src += sizeof(attrSet); + /* 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->service->matchDict, matchStr); + rec_del (zh->service->records, &rec); + } + rec_rm (&rec); + logRecord (0); + return 1; } - if (!(lead & 2)) + else { - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); + /* 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->keys); + records_updated++; + } } -#endif - if (key_buf_used + 1024 > (ptr_top-ptr_i)*sizeof(char*)) - key_flush (); - ++ptr_i; + } + /* update file type */ + xfree (rec->info[recInfo_fileType]); + rec->info[recInfo_fileType] = + rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]); - key_buf[ptr_top-ptr_i] = (char*)key_buf + key_buf_used; + /* update filename */ + xfree (rec->info[recInfo_filename]); + rec->info[recInfo_filename] = + rec_strdup (fname, &rec->size[recInfo_filename]); -#if SU_SCHEME + /* update delete keys */ + xfree (rec->info[recInfo_delKeys]); + if (zh->keys.buf_used > 0 && rGroup->flagStoreKeys == 1) + { +#if 1 + rec->size[recInfo_delKeys] = zh->keys.buf_used; + rec->info[recInfo_delKeys] = zh->keys.buf; + zh->keys.buf = NULL; + zh->keys.buf_max = 0; #else - ch = zebraExplain_lookupSU (zti, attrSet, attrUse); - if (ch < 0) - { - ch = zebraExplain_addSU (zti, attrSet, attrUse); - yaz_log (LOG_LOG, "addSU cmd=%d set=%d use=%d SU=%d", - cmd, attrSet, attrUse, ch); - } + 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 - assert (ch > 0); - key_buf_used += key_SU_encode (ch, ((char*)key_buf) + key_buf_used); + } + else + { + rec->info[recInfo_delKeys] = NULL; + rec->size[recInfo_delKeys] = 0; + } - while (*src) - ((char*)key_buf) [key_buf_used++] = *src++; - src++; - ((char*)key_buf) [key_buf_used++] = '\0'; - ((char*) key_buf)[key_buf_used++] = cmd; + /* save file size of original record */ + zebraExplain_recordBytesIncrement (zh->service->zei, + - recordAttr->recordSize); + recordAttr->recordSize = fi->file_moffset - recordOffset; + if (!recordAttr->recordSize) + recordAttr->recordSize = fi->file_max - recordOffset; + zebraExplain_recordBytesIncrement (zh->service->zei, + recordAttr->recordSize); - if (lead & 60) - seqno += ((lead>>2) & 15)-1; - else + /* set run-number for this record */ + recordAttr->runNumber = zebraExplain_runNumberIncrement (zh->service->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) { - memcpy (&seqno, src, sizeof(seqno)); - src += sizeof(seqno); + 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); } - key.seqno = seqno; - key.sysno = sysno; - memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key)); - key_buf_used += sizeof(key); - off = src - reckeys->buf; } - assert (off == reckeys->buf_used); + 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->service->records, &rec); + logRecord (0); + return 1; } -static const char **searchRecordKey (struct recKeys *reckeys, - int attrSetS, int attrUseS) +int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, + const struct recordGroup *rGroupP, int deleteFlag) { - 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 + 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; - for (i = 0; i<32; i++) - ws[i] = NULL; + memcpy (rGroup, rGroupP, sizeof(*rGroupP)); + + if (!rGroup->groupName || !*rGroup->groupName) + *gprefix = '\0'; + else + sprintf (gprefix, "%s.", rGroup->groupName); -#if SU_SCHEME - chS = zebraExplain_lookupSU (zti, attrSetS, attrUseS); - if (chS < 0) - return ws; -#endif - while (off < reckeys->buf_used) - { + logf (LOG_DEBUG, "fileExtract %s", fname); - 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)) + /* determine file extension */ + *ext = '\0'; + for (i = strlen(fname); --i >= 0; ) + if (fname[i] == '/') + break; + else if (fname[i] == '.') { - memcpy (&attrUse, src, sizeof(attrUse)); - src += sizeof(attrUse); + strcpy (ext, fname+i+1); + break; } -#endif - wstart = src; - while (*src++) - ; - if (lead & 60) - seqno += ((lead>>2) & 15)-1; - else + /* determine file type - depending on extension */ + if (!rGroup->recordType) + { + sprintf (ext_res, "%srecordType.%s", gprefix, ext); + if (!(rGroup->recordType = res_get (zh->service->res, ext_res))) { - memcpy (&seqno, src, sizeof(seqno)); - src += sizeof(seqno); + sprintf (ext_res, "%srecordType", gprefix); + rGroup->recordType = res_get (zh->service->res, ext_res); } - if ( -#if SU_SCHEME - ch == chS -#else - attrUseS == attrUse && attrSetS == attrSet -#endif - ) - { - int woff; + } + 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->service->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->service->res, ext_res); + } - if (startSeq == -1) - startSeq = seqno; - woff = seqno - startSeq; - if (woff >= 0 && woff < 31) - ws[woff] = wstart; + /* determine database name */ + if (!rGroup->databaseName) + { + sprintf (ext_res, "%sdatabase.%s", gprefix, ext); + if (!(rGroup->databaseName = res_get (zh->service->res, ext_res))) + { + sprintf (ext_res, "%sdatabase", gprefix); + rGroup->databaseName = res_get (zh->service->res, ext_res); } - - 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); -} + if (!rGroup->databaseName) + rGroup->databaseName = "Default"; -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); -} + /* determine if explain database */ + + sprintf (ext_res, "%sexplainDatabase", gprefix); + rGroup->explainDatabase = + atoi (res_get_def (zh->service->res, ext_res, "0")); -static off_t file_tell (void *handle) -{ - struct file_read_info *p = (struct file_read_info *) handle; - return p->file_offset; -} + /* announce database */ + if (zebraExplain_curDatabase (zh->service->zei, rGroup->databaseName)) + { + if (zebraExplain_newDatabase (zh->service->zei, rGroup->databaseName, + rGroup->explainDatabase)) + return 0; + } -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) + if (rGroup->flagStoreData == -1) { - 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); + const char *sval; + sprintf (ext_res, "%sstoreData.%s", gprefix, ext); + if (!(sval = res_get (zh->service->res, ext_res))) + { + sprintf (ext_res, "%sstoreData", gprefix); + sval = res_get (zh->service->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->service->res, ext_res); + if (!sval) + { + sprintf (ext_res, "%sstoreKeys", gprefix); + sval = res_get (zh->service->res, ext_res); + } + if (!sval) + sval = res_get (zh->service->res, "storeKeys"); + if (sval) + rGroup->flagStoreKeys = atoi (sval); } + if (rGroup->flagStoreKeys == -1) + rGroup->flagStoreKeys = 0; + + if (sysno && deleteFlag) + fd = -1; 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; + 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; } -static void file_begin (void *handle) + +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) { - struct file_read_info *p = (struct file_read_info *) handle; + RecordAttr *recordAttr; + struct recExtractCtrl extractCtrl; + int i, r; + char *matchStr; + RecType recType; + char subType[1024]; + void *clientData; + const char *fname = ""; + 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; - p->file_offset = p->file_moffset; - if (!p->sdrbuf && p->file_moffset) - lseek (p->fd, p->file_moffset, SEEK_SET); - p->file_more = 0; -} + /* announce database */ + if (zebraExplain_curDatabase (zh->service->zei, databaseName)) + { + if (zebraExplain_newDatabase (zh->service->zei, databaseName, 0)) + return 0; + } + if (!(recType = + recType_byName (zh->service->recTypes, recordType, subType, + &clientData))) + { + logf (LOG_WARN, "No such record type: %s", recordType); + return 0; + } -static void file_end (void *handle, off_t offset) -{ - struct file_read_info *p = (struct file_read_info *) handle; + zh->keys.buf_used = 0; + zh->keys.prevAttrUse = -1; + zh->keys.prevAttrSet = -1; + zh->keys.prevSeqNo = 0; + zh->sortKeys = 0; + + extractCtrl.subType = subType; + extractCtrl.init = extract_init; + extractCtrl.tokenAdd = extract_token_add; + extractCtrl.schemaAdd = extract_schema_add; + extractCtrl.dh = zh->service->dh; + extractCtrl.handle = zh; + extractCtrl.zebra_maps = zh->service->zebra_maps; + extractCtrl.flagShowRecords = 0; + for (i = 0; i<256; i++) + { + if (zebra_maps_is_positioned(zh->service->zebra_maps, i)) + extractCtrl.seqno[i] = 1; + else + extractCtrl.seqno[i] = 0; + } - assert (p->file_more == 0); - p->file_more = 1; - p->file_moffset = offset; -} + r = (*recType->extract)(clientData, &extractCtrl); -static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup, - const char *fname, const char *spec) -{ - static char dstBuf[2048]; - char *dst = dstBuf; - const char *s = spec; - static const char **w; + 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->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 */ - while (1) + + if (! *sysno) { - while (*s == ' ' || *s == '\t') - s++; - if (!*s) - break; - if (*s == '(') + /* new record */ + if (delete_flag) { - char attset_str[64], attname_str[64]; - data1_attset *attset; - int i; - char matchFlag[32]; - int attSet = 1, attUse = 1; - int first = 1; + 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->service->records); - s++; - for (i = 0; *s && *s != ',' && *s != ')'; s++) - if (i < 63) - attset_str[i++] = *s; - attset_str[i] = '\0'; + *sysno = rec->sysno; - if (*s == ',') - { - s++; - for (i = 0; *s && *s != ')'; s++) - if (i < 63) - attname_str[i++] = *s; - attname_str[i] = '\0'; - } - - if ((attset = data1_get_attset (rGroup->dh, attset_str))) - { - data1_att *att; - attSet = attset->reference; - att = data1_getattbyname(rGroup->dh, attset, attname_str); - if (att) - attUse = att->value; - else - attUse = atoi (attname_str); - } - w = searchRecordKey (reckeys, attSet, attUse); - assert (w); + recordAttr = rec_init_attr (zh->service->zei, rec); - if (*s == ')') + if (matchStr) + { + dict_insert (zh->service->matchDict, matchStr, + sizeof(*sysno), sysno); + } + extract_flushRecordKeys (zh, *sysno, 1, &zh->keys); + extract_flushSortKeys (zh, *sysno, 1, &zh->sortKeys); + } + else + { + /* record already exists */ + struct recKeys delkeys; + + rec = rec_get (zh->service->records, *sysno); + assert (rec); + + recordAttr = rec_init_attr (zh->service->zei, rec); + + if (recordAttr->runNumber == + zebraExplain_runNumberIncrement (zh->service->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->sortKeys); + extract_flushRecordKeys (zh, *sysno, 0, &delkeys); + if (delete_flag) + { + /* record going to be deleted */ + if (!delkeys.buf_used) { - for (i = 0; i<32; i++) - matchFlag[i] = 1; + logf (LOG_LOG, "delete %s %s %ld", recordType, + fname, (long) recordOffset); + logf (LOG_WARN, "cannot delete file above, storeKeys false"); } 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; + logf (LOG_LOG, "delete %s %s %ld", recordType, + fname, (long) recordOffset); +#if 0 + if (matchStr) + dict_delete (matchDict, matchStr); +#endif + rec_del (zh->service->records, &rec); } + rec_rm (&rec); + return 1; } - else if (*s == '$') + else { - 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) + /* record going to be updated */ + if (!delkeys.buf_used) { - strcpy (dst, spec_src); - dst += strlen (spec_src); + 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->keys); } } - else if (*s == '\"' || *s == '\'') - { - int stopMarker = *s++; - char tmpString[64]; - int i = 0; + } + /* 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->keys.buf_used > 0 && store_keys == 1) + { + rec->size[recInfo_delKeys] = zh->keys.buf_used; + rec->info[recInfo_delKeys] = zh->keys.buf; + zh->keys.buf = NULL; + zh->keys.buf_max = 0; + } + else + { + rec->info[recInfo_delKeys] = NULL; + rec->size[recInfo_delKeys] = 0; + } - while (*s && *s != stopMarker) - { - if (i < 63) - tmpString[i++] = *s++; - } - if (*s) - s++; - tmpString[i] = '\0'; - strcpy (dst, tmpString); - dst += strlen (tmpString); + /* save file size of original record */ + zebraExplain_recordBytesIncrement (zh->service->zei, + - recordAttr->recordSize); +#if 0 + recordAttr->recordSize = fi->file_moffset - recordOffset; + if (!recordAttr->recordSize) + recordAttr->recordSize = fi->file_max - recordOffset; +#else + recordAttr->recordSize = buf_size; +#endif + zebraExplain_recordBytesIncrement (zh->service->zei, + recordAttr->recordSize); + + /* set run-number for this record */ + recordAttr->runNumber = + zebraExplain_runNumberIncrement (zh->service->zei, 0); + + /* update store data */ + xfree (rec->info[recInfo_storeData]); + if (store_data == 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 + if (lseek (fi->fd, recordOffset, SEEK_SET) < 0) + { + logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s", + (long) recordOffset, fname); + exit (1); } - else + if (read (fi->fd, rec->info[recInfo_storeData], recordAttr->recordSize) + < recordAttr->recordSize) { - logf (LOG_WARN, "Syntax error in match criteria %s in group %s", - spec, rGroup->groupName ? rGroup->groupName : "none"); - return NULL; + logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s", + recordAttr->recordSize, fname); + exit (1); } - *dst++ = 1; +#endif } - if (dst == dstBuf) + else { - logf (LOG_WARN, "No match criteria for record %s in group %s", - fname, rGroup->groupName ? rGroup->groupName : "none"); - return NULL; + rec->info[recInfo_storeData] = NULL; + rec->size[recInfo_storeData] = 0; } - *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 (); + /* update database name */ + xfree (rec->info[recInfo_databaseName]); + rec->info[recInfo_databaseName] = + rec_strdup (databaseName, &rec->size[recInfo_databaseName]); - 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); -} + /* update offset */ + recordAttr->recordOffset = recordOffset; + + /* commit this record */ + rec_put (zh->service->records, &rec); -void addSchema (struct recExtractCtrl *p, Odr_oid *oid) -{ - zebraExplain_addSchema (zti, oid); + return 0; } -static int recordExtract (SYSNO *sysno, const char *fname, - struct recordGroup *rGroup, int deleteFlag, - struct file_read_info *fi, - RecType recType, char *subType, void *clientData) +int explain_extract (void *handle, Record rec, data1_node *n) { - RecordAttr *recordAttr; - int r; - char *matchStr; - SYSNO sysnotmp; - Record rec; - struct recordLogInfo logInfo; - off_t recordOffset = 0; + ZebraHandle zh = (ZebraHandle) handle; + struct recExtractCtrl extractCtrl; + int i; - if (fi->fd != -1) + if (zebraExplain_curDatabase (zh->service->zei, + rec->info[recInfo_databaseName])) { - struct recExtractCtrl extractCtrl; + abort(); + if (zebraExplain_newDatabase (zh->service->zei, + rec->info[recInfo_databaseName], 0)) + abort (); + } - /* we are going to read from a file, so prepare the extraction */ - int i; + zh->keys.buf_used = 0; + zh->keys.prevAttrUse = -1; + zh->keys.prevAttrSet = -1; + zh->keys.prevSeqNo = 0; + zh->sortKeys = 0; + + extractCtrl.init = extract_init; + extractCtrl.tokenAdd = extract_token_add; + extractCtrl.schemaAdd = extract_schema_add; + extractCtrl.dh = zh->service->dh; + for (i = 0; i<256; i++) + extractCtrl.seqno[i] = 0; + extractCtrl.zebra_maps = zh->service->zebra_maps; + extractCtrl.flagShowRecords = 0; + extractCtrl.handle = handle; + + grs_extract_tree(&extractCtrl, n); - reckeys.buf_used = 0; - reckeys.prevAttrUse = -1; - reckeys.prevAttrSet = -1; - reckeys.prevSeqNo = 0; - - recordOffset = fi->file_moffset; - extractCtrl.offset = fi->file_moffset; - extractCtrl.readf = file_read; - extractCtrl.seekf = file_seek; - extractCtrl.tellf = file_tell; - extractCtrl.endf = file_end; - extractCtrl.fh = fi; - extractCtrl.subType = subType; - extractCtrl.init = wordInit; - extractCtrl.tokenAdd = addRecordKey; - extractCtrl.schemaAdd = addSchema; - extractCtrl.dh = rGroup->dh; - for (i = 0; i<256; i++) - { - if (zebra_maps_is_positioned(rGroup->zebra_maps, i)) - extractCtrl.seqno[i] = 1; - else - extractCtrl.seqno[i] = 0; - } - extractCtrl.zebra_maps = rGroup->zebra_maps; - extractCtrl.flagShowRecords = !rGroup->flagRw; + logf (LOG_LOG, "flush explain record, sysno=%d", rec->sysno); - if (!rGroup->flagRw) - printf ("File: %s " PRINTF_OFF_T "\n", fname, recordOffset); + if (rec->size[recInfo_delKeys]) + { + struct recKeys delkeys; + struct sortKey *sortKeys = 0; - logInfo.fname = fname; - logInfo.recordOffset = recordOffset; - logInfo.rGroup = rGroup; - log_event_start (recordLogPreamble, &logInfo); + 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); + } + extract_flushRecordKeys (zh, rec->sysno, 1, &zh->keys); + extract_flushSortKeys (zh, rec->sysno, 1, &zh->sortKeys); - r = (*recType->extract)(clientData, &extractCtrl); + xfree (rec->info[recInfo_delKeys]); + rec->size[recInfo_delKeys] = zh->keys.buf_used; + rec->info[recInfo_delKeys] = zh->keys.buf; + zh->keys.buf = NULL; + zh->keys.buf_max = 0; + return 0; +} - log_event_start (NULL, NULL); +void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno, + int cmd, struct recKeys *reckeys) +{ +#if SU_SCHEME +#else + unsigned char attrSet = (unsigned char) -1; + unsigned short attrUse = (unsigned short) -1; +#endif + int seqno = 0; + int off = 0; + int ch = 0; + ZebraExplainInfo zei = zh->service->zei; - if (r == RECCTRL_EXTRACT_EOF) - return 0; - else if (r == RECCTRL_EXTRACT_ERROR) + if (!zh->key_buf) + { + int mem = 8*1024*1024; + zh->key_buf = (char**) xmalloc (mem); + zh->ptr_top = mem/sizeof(char*); + zh->ptr_i = 0; + zh->key_buf_used = 0; + zh->key_file_no = 0; + } + zebraExplain_recordCountIncrement (zei, cmd ? 1 : -1); + while (off < reckeys->buf_used) + { + const char *src = reckeys->buf + off; + struct it_key key; + int lead; + + lead = *src++; + +#if SU_SCHEME + if ((lead & 3) < 3) { - /* error occured during extraction ... */ - if (rGroup->flagRw && - records_processed < rGroup->fileVerboseLimit) - { - logf (LOG_WARN, "fail %s %s " PRINTF_OFF_T, rGroup->recordType, - fname, recordOffset); - } - return 0; - } - if (reckeys.buf_used == 0) + memcpy (&ch, src, sizeof(ch)); + src += sizeof(ch); + } +#else + if (!(lead & 1)) { - /* 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; + memcpy (&attrSet, src, sizeof(attrSet)); + src += sizeof(attrSet); } - } - - /* 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) + if (!(lead & 2)) { - char *rinfo; - - matchStr = fileMatchStr (&reckeys, rGroup, fname, - rGroup->recordId); - if (matchStr) - { - rinfo = dict_lookup (matchDict, matchStr); - if (rinfo) - memcpy (sysno, rinfo+1, sizeof(*sysno)); - } - else - { - logf (LOG_WARN, "Bad match criteria"); - return 0; - } + memcpy (&attrUse, src, sizeof(attrUse)); + src += sizeof(attrUse); } - } +#endif + if (zh->key_buf_used + 1024 > (zh->ptr_top-zh->ptr_i)*sizeof(char*)) + extract_flushWriteKeys (zh); + ++(zh->ptr_i); + (zh->key_buf)[zh->ptr_top - zh->ptr_i] = + (char*)zh->key_buf + zh->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->key_buf_used += + key_SU_encode (ch,((char*)zh->key_buf) + zh->key_buf_used); - if (! *sysno) - { - /* new record */ - if (deleteFlag) + while (*src) + ((char*)zh->key_buf) [(zh->key_buf_used)++] = *src++; + src++; + ((char*)(zh->key_buf))[(zh->key_buf_used)++] = '\0'; + ((char*)(zh->key_buf))[(zh->key_buf_used)++] = cmd; + + if (lead & 60) + seqno += ((lead>>2) & 15)-1; + else { - logf (LOG_LOG, "delete %s %s " PRINTF_OFF_T, rGroup->recordType, - fname, recordOffset); - logf (LOG_WARN, "cannot delete record above (seems new)"); - return 1; + memcpy (&seqno, src, sizeof(seqno)); + src += sizeof(seqno); } - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "add %s %s " PRINTF_OFF_T, rGroup->recordType, - fname, recordOffset); - rec = rec_new (records); - - *sysno = rec->sysno; + key.seqno = seqno; + key.sysno = sysno; + memcpy ((char*)zh->key_buf + zh->key_buf_used, &key, sizeof(key)); + (zh->key_buf_used) += sizeof(key); + off = src - reckeys->buf; + } + assert (off == reckeys->buf_used); +} - recordAttr = rec_init_attr (zti, rec); +void extract_flushWriteKeys (ZebraHandle zh) +{ + FILE *outf; + char out_fname[200]; + char *prevcp, *cp; + struct encode_info encode_info; + int ptr_i = zh->ptr_i; +#if SORT_EXTRA + int i; +#endif + if (!zh->key_buf || ptr_i <= 0) + return; - if (matchStr) - { - dict_insert (matchDict, matchStr, sizeof(*sysno), sysno); - } - flushRecordKeys (*sysno, 1, &reckeys); - flushSortKeys (*sysno, 1); + (zh->key_file_no)++; + logf (LOG_LOG, "sorting section %d", (zh->key_file_no)); +#if !SORT_EXTRA + qsort (zh->key_buf + zh->ptr_top - ptr_i, ptr_i, sizeof(char*), + key_qsort_compare); + extract_get_fname_tmp (zh, out_fname, zh->key_file_no); - records_inserted++; + if (!(outf = fopen (out_fname, "wb"))) + { + logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); + exit (1); } - else + logf (LOG_LOG, "writing section %d", zh->key_file_no); + prevcp = cp = (zh->key_buf)[zh->ptr_top - ptr_i]; + + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + + while (--ptr_i > 0) { - /* record already exists */ - struct recKeys delkeys; - - rec = rec_get (records, *sysno); - assert (rec); - - recordAttr = rec_init_attr (zti, rec); - - if (recordAttr->runNumber == zebraExplain_runNumberIncrement (zti, 0)) - { - logf (LOG_LOG, "skipped %s %s " PRINTF_OFF_T, rGroup->recordType, - fname, recordOffset); - flushSortKeys (*sysno, -1); - rec_rm (&rec); - logRecord (0); - return 1; - } - delkeys.buf_used = rec->size[recInfo_delKeys]; - delkeys.buf = rec->info[recInfo_delKeys]; - flushSortKeys (*sysno, 0); - flushRecordKeys (*sysno, 0, &delkeys); - if (deleteFlag) + cp = (zh->key_buf)[zh->ptr_top - ptr_i]; + if (strcmp (cp, prevcp)) { - /* 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 (matchDict, matchStr); - rec_del (records, &rec); - } - rec_rm (&rec); - logRecord (0); - return 1; + encode_key_init (&encode_info); + encode_key_write (cp, &encode_info, outf); + prevcp = cp; } 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])) { - /* 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 + 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) { - if (records_processed < rGroup->fileVerboseLimit) - logf (LOG_LOG, "update %s %s " PRINTF_OFF_T, - rGroup->recordType, fname, recordOffset); - flushRecordKeys (*sysno, 1, &reckeys); - records_updated++; + 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); } - /* update file type */ - xfree (rec->info[recInfo_fileType]); - rec->info[recInfo_fileType] = - rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]); + logf (LOG_LOG, "finished section %d", zh->key_file_no); + zh->ptr_i = 0; + zh->key_buf_used = 0; +} - /* update filename */ - xfree (rec->info[recInfo_filename]); - rec->info[recInfo_filename] = - rec_strdup (fname, &rec->size[recInfo_filename]); +void extract_add_index_string (RecWord *p, const char *string, + int length) +{ + 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->service->zei; + struct recKeys *keys = &zh->keys; - /* update delete keys */ - xfree (rec->info[recInfo_delKeys]); - if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1) + if (keys->buf_used+1024 > keys->buf_max) { + char *b; + + b = (char *) xmalloc (keys->buf_max += 128000); + if (keys->buf_used > 0) + memcpy (b, keys->buf, keys->buf_used); + xfree (keys->buf); + keys->buf = b; + } + dst = keys->buf + keys->buf_used; + + 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 - rec->size[recInfo_delKeys] = reckeys.buf_used; - rec->info[recInfo_delKeys] = reckeys.buf; - reckeys.buf = NULL; - reckeys.buf_max = 0; -#else - rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used); - rec->size[recInfo_delKeys] = reckeys.buf_used; - memcpy (rec->info[recInfo_delKeys], reckeys.buf, - rec->size[recInfo_delKeys]); + 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) + { + int ch = zebraExplain_lookupSU (zei, attrSet, attrUse); + if (ch < 0) + { + ch = zebraExplain_addSU (zei, attrSet, attrUse); + yaz_log (LOG_LOG, "addSU set=%d use=%d SU=%d", + attrSet, attrUse, ch); + } + assert (ch > 0); + memcpy (dst, &ch, sizeof(ch)); + dst += sizeof(ch); } - else +#else + if (!(lead & 1)) { - rec->info[recInfo_delKeys] = NULL; - rec->size[recInfo_delKeys] = 0; + 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'; - /* save file size of original record */ - zebraExplain_recordBytesIncrement (zti, - recordAttr->recordSize); - recordAttr->recordSize = fi->file_moffset - recordOffset; - if (!recordAttr->recordSize) - recordAttr->recordSize = fi->file_max - recordOffset; - zebraExplain_recordBytesIncrement (zti, recordAttr->recordSize); - - /* set run-number for this record */ - recordAttr->runNumber = zebraExplain_runNumberIncrement (zti, 0); - - /* update store data */ - xfree (rec->info[recInfo_storeData]); - if (rGroup->flagStoreData == 1) + if (!diff) { - 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); - } + memcpy (dst, pseqno, sizeof(*pseqno)); + 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 sortKey *sortKeys = zh->sortKeys; + + for (sk = sortKeys; sk; sk = sk->next) + if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse) + return; + + sk = (struct sortKey *) xmalloc (sizeof(*sk)); + sk->next = sortKeys; + sortKeys = sk; + + sk->string = (char *) xmalloc (length); + sk->length = length; + memcpy (sk->string, string, length); + + sk->attrSet = p->attrSet; + sk->attrUse = p->attrUse; +} + +void extract_add_string (RecWord *p, const char *string, int length) +{ + assert (length > 0); + if (zebra_maps_is_sort (p->zebra_maps, p->reg_type)) + extract_add_sort_string (p, string, length); else + extract_add_index_string (p, string, length); +} + +static void extract_add_incomplete_field (RecWord *p) +{ + const char *b = p->string; + int remain = p->length; + const char **map = 0; + + if (remain > 0) + map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + + while (map) { - 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]); + 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; - /* update offset */ - recordAttr->recordOffset = recordOffset; - - /* commit this record */ - rec_put (records, &rec); - logRecord (0); - return 1; + 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); + } + (p->seqnos[p->reg_type])++; /* to separate this from next one */ } -int fileExtract (SYSNO *sysno, const char *fname, - const struct recordGroup *rGroupP, int deleteFlag) +static void extract_add_complete_field (RecWord *p) { - 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); + const char *b = p->string; + char buf[IT_MAX_WORD+1]; + const char **map = 0; + int i = 0, remain = p->length; - logf (LOG_DEBUG, "fileExtract %s", fname); + if (remain > 0) + map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain); - /* 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 (common_resource, ext_res))) - { - sprintf (ext_res, "%srecordType", gprefix); - rGroup->recordType = res_get (common_resource, 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 (rGroup->recTypes, rGroup->recordType, subType, - &clientData))) + while (remain > 0 && i < IT_MAX_WORD) { - logf (LOG_WARN, "No such record type: %s", rGroup->recordType); - return 0; - } + 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; - /* determine match criteria */ - if (!rGroup->recordId) - { - sprintf (ext_res, "%srecordId.%s", gprefix, ext); - rGroup->recordId = res_get (common_resource, ext_res); + if (i && i < IT_MAX_WORD) + buf[i++] = *CHR_SPACE; + while (map && *map && **map != *CHR_SPACE) + { + const char *cp = *map; + + if (i >= IT_MAX_WORD) + break; + while (i < IT_MAX_WORD && *cp) + buf[i++] = *(cp++); + remain = p->length - (b - p->string); + if (remain > 0) + map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, + remain); + else + map = 0; + } } + if (!i) + return; + extract_add_string (p, buf, i); +} - /* determine database name */ - if (!rGroup->databaseName) +void extract_token_add (RecWord *p) +{ + WRBUF wrbuf; + if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0, + p->string, p->length))) { - sprintf (ext_res, "%sdatabase.%s", gprefix, ext); - if (!(rGroup->databaseName = res_get (common_resource, ext_res))) - { - sprintf (ext_res, "%sdatabase", gprefix); - rGroup->databaseName = res_get (common_resource, ext_res); - } + p->string = wrbuf_buf(wrbuf); + p->length = wrbuf_len(wrbuf); } - if (!rGroup->databaseName) - rGroup->databaseName = "Default"; + if (zebra_maps_is_complete (p->zebra_maps, p->reg_type)) + extract_add_complete_field (p); + else + extract_add_incomplete_field(p); +} - /* determine if explain database */ - - sprintf (ext_res, "%sexplainDatabase", gprefix); - rGroup->explainDatabase = - atoi (res_get_def (common_resource, ext_res, "0")); +void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid) +{ + ZebraHandle zh = (ZebraHandle) (p->handle); + zebraExplain_addSchema (zh->service->zei, oid); +} - /* announce database */ - if (zebraExplain_curDatabase (zti, rGroup->databaseName)) +void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno, + int cmd, struct sortKey **skp) +{ + struct sortKey *sk = *skp; + SortIdx sortIdx = zh->service->sortIdx; + + sortIdx_sysno (sortIdx, sysno); + while (sk) { - if (zebraExplain_newDatabase (zti, rGroup->databaseName, - rGroup->explainDatabase)) - 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; +} - if (rGroup->flagStoreData == -1) +void encode_key_init (struct encode_info *i) +{ + i->sysno = 0; + i->seqno = 0; + i->cmd = -1; +} + +char *encode_key_int (int d, char *bp) +{ + if (d <= 63) + *bp++ = d; + else if (d <= 16383) { - const char *sval; - sprintf (ext_res, "%sstoreData.%s", gprefix, ext); - if (!(sval = res_get (common_resource, ext_res))) - { - sprintf (ext_res, "%sstoreData", gprefix); - sval = res_get (common_resource, ext_res); - } - if (sval) - rGroup->flagStoreData = atoi (sval); + *bp++ = 64 + (d>>8); + *bp++ = d & 255; } - if (rGroup->flagStoreData == -1) - rGroup->flagStoreData = 0; - - if (rGroup->flagStoreKeys == -1) + else if (d <= 4194303) { - const char *sval; - - sprintf (ext_res, "%sstoreKeys.%s", gprefix, ext); - sval = res_get (common_resource, ext_res); - if (!sval) - { - sprintf (ext_res, "%sstoreKeys", gprefix); - sval = res_get (common_resource, ext_res); - } - if (!sval) - sval = res_get (common_resource, "storeKeys"); - if (sval) - rGroup->flagStoreKeys = atoi (sval); + *bp++ = 128 + (d>>16); + *bp++ = (d>>8) & 255; + *bp++ = d & 255; } - 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; - } + *bp++ = 192 + (d>>24); + *bp++ = (d>>16) & 255; + *bp++ = (d>>8) & 255; + *bp++ = d & 255; } - fi = file_read_start (fd); - do - { - file_begin (fi); - r = recordExtract (sysno, fname, rGroup, deleteFlag, fi, - recType, subType, clientData); - } while (r && !sysno && fi->file_more); - file_read_stop (fi); - if (fd != -1) - close (fd); - return r; + return bp; } -static int explain_extract (void *handle, Record rec, data1_node *n) +void encode_key_write (char *k, struct encode_info *i, FILE *outf) { - struct recordGroup *rGroup = (struct recordGroup*) handle; - struct recExtractCtrl extractCtrl; - int i; + struct it_key key; + char *bp = i->buf; - if (zebraExplain_curDatabase (zti, rec->info[recInfo_databaseName])) + 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 (zebraExplain_newDatabase (zti, rec->info[recInfo_databaseName], 0)) - abort (); + i->sysno = key.sysno; + i->seqno = 0; } - - reckeys.buf_used = 0; - reckeys.prevAttrUse = -1; - reckeys.prevAttrSet = -1; - reckeys.prevSeqNo = 0; - - extractCtrl.init = wordInit; - extractCtrl.tokenAdd = addRecordKey; - extractCtrl.schemaAdd = addSchema; - extractCtrl.dh = rGroup->dh; - for (i = 0; i<256; i++) - extractCtrl.seqno[i] = 0; - extractCtrl.zebra_maps = rGroup->zebra_maps; - extractCtrl.flagShowRecords = !rGroup->flagRw; - - grs_extract_tree(&extractCtrl, n); - - logf (LOG_DEBUG, "flush explain record, sysno=%d", rec->sysno); - - if (rec->size[recInfo_delKeys]) + 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) { - struct recKeys delkeys; - - delkeys.buf_used = rec->size[recInfo_delKeys]; - delkeys.buf = rec->info[recInfo_delKeys]; - flushSortKeys (rec->sysno, 0); - flushRecordKeys (rec->sysno, 0, &delkeys); + logf (LOG_FATAL|LOG_ERRNO, "fwrite"); + exit (1); } - flushRecordKeys (rec->sysno, 1, &reckeys); - flushSortKeys (rec->sysno, 1); - - xfree (rec->info[recInfo_delKeys]); - rec->size[recInfo_delKeys] = reckeys.buf_used; - rec->info[recInfo_delKeys] = reckeys.buf; - reckeys.buf = NULL; - reckeys.buf_max = 0; - - return 0; } + diff --git a/index/index.h b/index/index.h index aba2cbc..6064462 100644 --- a/index/index.h +++ b/index/index.h @@ -1,8 +1,8 @@ /* - * Copyright (C) 1995-0000, Index Data + * Copyright (C) 1995-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss, Heikki Levanto - * (log at the end) + * $Id: index.h,v 1.73 2002-02-20 17:30:01 adam Exp $ */ #ifndef INDEX_H @@ -25,6 +25,7 @@ #endif #include #include +#include "zebraapi.h" YAZ_BEGIN_CDECL @@ -68,13 +69,6 @@ struct recordGroup { int fileVerboseLimit; int databaseNamePath; int explainDatabase; -#if ZEBRASDR - int useSDR; -#endif - data1_handle dh; - BFiles bfs; - ZebraMaps zebra_maps; - RecTypes recTypes; }; void getFnameTmp (Res res, char *fname, int no); @@ -93,21 +87,21 @@ struct dir_entry *dir_open (const char *rep); 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 (struct recordGroup *rGroup, int mem); -int key_close (struct recordGroup *group); +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 inv_prstat (ZebraHandle zh); void inv_compact (BFiles bfs); -void key_input (BFiles bfs, int nkeys, int cache, Res res); +void key_input (ZebraHandle zh, int nkeys, int cache, Res res); ISAMS_M key_isams_m (Res res, ISAMS_M me); #if ZMBOL ISAMC_M key_isamc_m (Res res, ISAMC_M me); @@ -136,13 +130,13 @@ int index_char_cvt (int c); int index_word_prefix (char *string, int attset_ordinal, int local_attribute, const char *databaseName); -int fileExtract (SYSNO *sysno, const char *fname, +int fileExtract (ZebraHandle zh, SYSNO *sysno, const char *fname, const struct recordGroup *rGroup, int deleteFlag); -void zebraIndexLockMsg (const char *str); -void zebraIndexUnlock (void); -int 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" @@ -158,266 +152,28 @@ int zebra_unlock (ZebraLockHandle h); int zebra_lock_fd (ZebraLockHandle h); void zebra_lock_prefix (Res res, char *dst); +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; +// 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); + YAZ_END_CDECL #endif -/* - * $Log: index.h,v $ - * Revision 1.72 2001-11-19 23:08:29 adam - * Added const qualifier for name parameter of key_SU_decode. - * - * Revision 1.71 2001/11/19 23:05:22 adam - * Added a few prototypes. - * - * Revision 1.70 2000/12/05 10:01:44 adam - * Fixed bug regarding user-defined attribute sets. - * - * Revision 1.69 2000/03/20 19:08:36 adam - * Added remote record import using Z39.50 extended services and Segment - * Requests. - * - * Revision 1.68 2000/02/24 11:00:07 adam - * Fixed bug: indexer would run forever when lock dir was non-existant. - * - * Revision 1.67 1999/11/30 13:48:03 adam - * Improved installation. Updated for inclusion of YAZ header files. - * - * Revision 1.66 1999/07/14 13:21:34 heikki - * Added isam-d files. Compiles (almost) clean. Doesn't work at all - * - * Revision 1.65 1999/07/14 10:59:26 adam - * Changed functions isc_getmethod, isams_getmethod. - * Improved fatal error handling (such as missing EXPLAIN schema). - * - * Revision 1.64 1999/06/30 15:07:23 heikki - * Adding isamh stuff - * - * Revision 1.63 1999/05/26 07:49:13 adam - * C++ compilation. - * - * Revision 1.62 1999/05/12 13:08:06 adam - * First version of ISAMS. - * - * Revision 1.61 1999/03/09 16:27:49 adam - * More work on SDRKit integration. - * - * Revision 1.60 1998/10/16 08:14:31 adam - * Updated record control system. - * - * Revision 1.59 1998/06/08 14:43:11 adam - * Added suport for EXPLAIN Proxy servers - added settings databasePath - * and explainDatabase to facilitate this. Increased maximum number - * of databases and attributes in one register. - * - * Revision 1.58 1998/05/20 10:12:16 adam - * Implemented automatic EXPLAIN database maintenance. - * Modified Zebra to work with ASN.1 compiled version of YAZ. - * - * Revision 1.57 1998/03/05 08:45:12 adam - * New result set model and modular ranking system. Moved towards - * descent server API. System information stored as "SGML" records. - * - * 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 that limits the log to the first - * 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. - * - */ diff --git a/index/invstat.c b/index/invstat.c index ce82cd6..5a23007 100644 --- a/index/invstat.c +++ b/index/invstat.c @@ -14,6 +14,7 @@ #if ZMBOL #include "../isamc/isamd-p.h" #endif +#include "zserver.h" struct inv_stat_info { ISAMS isams; @@ -128,8 +129,9 @@ static int inv_stat_handle (char *name, const char *info, int pos, return 0; } -void inv_prstat (BFiles bfs) +void inv_prstat (ZebraHandle zh) { + BFiles bfs = zh->service->bfs; Dict dict; ISAMS isams = NULL; #if ZMBOL @@ -156,11 +158,11 @@ void inv_prstat (BFiles bfs) logf (LOG_FATAL, "dict_open fail"); exit (1); } - if (res_get_match (common_resource, "isam", "s", ISAM_DEFAULT)) + if (res_get_match (zh->service->res, "isam", "s", ISAM_DEFAULT)) { struct ISAMS_M_s isams_m; isams = isams_open (bfs, FNAME_ISAMS, 0, - key_isams_m(common_resource, &isams_m)); + key_isams_m(zh->service->res, &isams_m)); if (!isams) { logf (LOG_FATAL, "isams_open fail"); @@ -168,32 +170,32 @@ void inv_prstat (BFiles bfs) } } #if ZMBOL - else if (res_get_match (common_resource, "isam", "i", ISAM_DEFAULT)) + else if (res_get_match (zh->service->res, "isam", "i", ISAM_DEFAULT)) { isam = is_open (bfs, FNAME_ISAM, key_compare, 0, - sizeof(struct it_key), common_resource); + sizeof(struct it_key), zh->service->res); if (!isam) { logf (LOG_FATAL, "is_open fail"); exit (1); } } - else if (res_get_match (common_resource, "isam", "d", ISAM_DEFAULT)) + else if (res_get_match (zh->service->res, "isam", "d", ISAM_DEFAULT)) { struct ISAMD_M_s isamd_m; isamd = isamd_open (bfs, FNAME_ISAMD, 0, - key_isamd_m(common_resource,&isamd_m)); + key_isamd_m(zh->service->res,&isamd_m)); if (!isamd) { logf (LOG_FATAL, "isamd_open fail"); exit (1); } } - else if (res_get_match (common_resource, "isam", "c", ISAM_DEFAULT)) + else if (res_get_match (zh->service->res, "isam", "c", ISAM_DEFAULT)) { struct ISAMC_M_s isamc_m; isamc = isc_open (bfs, FNAME_ISAMC, 0, - key_isamc_m (common_resource, &isamc_m)); + key_isamc_m (zh->service->res, &isamc_m)); if (!isamc) { logf (LOG_FATAL, "isc_open fail"); @@ -323,7 +325,10 @@ void inv_prstat (BFiles bfs) /* * * $Log: invstat.c,v $ - * Revision 1.21 2000-07-13 10:14:20 heikki + * Revision 1.22 2002-02-20 17:30:01 adam + * Work on new API. Locking system re-implemented + * + * Revision 1.21 2000/07/13 10:14:20 heikki * Removed compiler warnings when making zebra * * Revision 1.20 1999/12/01 13:30:30 adam diff --git a/index/kinput.c b/index/kinput.c index 92e64ec..b9d7f06 100644 --- a/index/kinput.c +++ b/index/kinput.c @@ -1,9 +1,9 @@ /* - * Copyright (C) 1994-2000, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss, Heikki Levanto * - * (log at the end) + * $Id: kinput.c,v 1.45 2002-02-20 17:30:01 adam Exp $ * * Bugs * - Allocates a lot of memory for the merge process, but never releases it. @@ -412,6 +412,7 @@ int heap_inpc (struct heap_info *hi) } } xfree (isamc_i); + xfree (hci.key); return 0; } @@ -672,7 +673,7 @@ void zebra_index_merge (ZebraHandle zh) for (i = 1; i<=nkeys; i++) { - extract_get_fname_tmp (zh, rbuf, i); + extract_get_fname_tmp (zh, rbuf, i); unlink (rbuf); } logf (LOG_LOG, "Iterations . . .%7d", no_iterations); @@ -683,16 +684,9 @@ void zebra_index_merge (ZebraHandle zh) zh->key_file_no = 0; } -void key_input (BFiles bfs, int nkeys, int cache, Res res) +void key_input (ZebraHandle zh, int nkeys, int cache, Res res) { - Dict dict; - ISAMS isams = NULL; -#if ZMBOL - ISAM isam = NULL; - ISAMC isamc = NULL; - ISAMD isamd = NULL; -#endif struct key_file **kf; char rbuf[1024]; int i, r; @@ -713,58 +707,6 @@ void key_input (BFiles bfs, int nkeys, int cache, Res res) if (!nkeys) return ; } - dict = dict_open (bfs, FNAME_DICT, cache, 1, 0); - if (!dict) - { - logf (LOG_FATAL, "dict_open fail"); - exit (1); - } - if (res_get_match (res, "isam", "s", ISAM_DEFAULT)) - { - struct ISAMS_M_s isams_m; - isams = isams_open (bfs, FNAME_ISAMS, 1, - key_isams_m (res, &isams_m)); - if (!isams) - { - logf (LOG_FATAL, "isams_open fail"); - exit (1); - } - logf (LOG_LOG, "isams opened"); - } -#if ZMBOL - else if (res_get_match (res, "isam", "i", ISAM_DEFAULT)) - { - isam = is_open (bfs, FNAME_ISAM, key_compare, 1, - sizeof(struct it_key), res); - if (!isam) - { - logf (LOG_FATAL, "is_open fail"); - exit (1); - } - } - else if (res_get_match (res, "isam", "d", ISAM_DEFAULT)) - { - struct ISAMD_M_s isamd_m; - isamd = isamd_open (bfs, FNAME_ISAMD, 1, - key_isamd_m (res,&isamd_m)); - if (!isamd) - { - logf (LOG_FATAL, "isamd_open fail"); - exit (1); - } - } - else if (res_get_match (res, "isam", "c", ISAM_DEFAULT)) - { - struct ISAMC_M_s isamc_m; - isamc = isc_open (bfs, FNAME_ISAMC, 1, - key_isamc_m (res, &isamc_m)); - if (!isamc) - { - logf (LOG_FATAL, "isc_open fail"); - exit (1); - } - } -#endif kf = (struct key_file **) xmalloc ((1+nkeys) * sizeof(*kf)); progressInfo.totalBytes = 0; progressInfo.totalOffset = 0; @@ -779,40 +721,28 @@ void key_input (BFiles bfs, int nkeys, int cache, Res res) progressInfo.totalOffset += kf[i]->buf_size; } hi = key_heap_init (nkeys, key_qsort_compare); - hi->dict = dict; - hi->isams = isams; + hi->dict = zh->service->dict; + hi->isams = zh->service->isams; #if ZMBOL - hi->isam = isam; - hi->isamc = isamc; - hi->isamd = isamd; + hi->isam = zh->service->isam; + hi->isamc = zh->service->isamc; + hi->isamd = zh->service->isamd; #endif for (i = 1; i<=nkeys; i++) if ((r = key_file_read (kf[i], rbuf))) key_heap_insert (hi, rbuf, r, kf[i]); - if (isams) + if (hi->isams) heap_inps (hi); #if ZMBOL - else if (isamc) + else if (hi->isamc) heap_inpc (hi); - else if (isam) + else if (hi->isam) heap_inp (hi); - else if (isamd) + else if (hi->isamd) heap_inpd (hi); #endif - dict_close (dict); - if (isams) - isams_close (isams); -#if ZMBOL - if (isam) - is_close (isam); - if (isamc) - isc_close (isamc); - if (isamd) - isamd_close (isamd); -#endif - for (i = 1; i<=nkeys; i++) { getFnameTmp (res, rbuf, i); @@ -827,157 +757,3 @@ void key_input (BFiles bfs, int nkeys, int cache, Res res) /* xmalloc_trav("unfreed"); while hunting leaks */ } - - -/* - * $Log: kinput.c,v $ - * Revision 1.44 2000-05-18 12:01:36 adam - * System call times(2) used again. More 64-bit fixes. - * - * Revision 1.43 2000/03/20 19:08:36 adam - * Added remote record import using Z39.50 extended services and Segment - * Requests. - * - * Revision 1.42 1999/12/01 21:58:48 adam - * Proper handle of illegal use of isams. - * - * Revision 1.41 1999/11/30 13:48:03 adam - * Improved installation. Updated for inclusion of YAZ header files. - * - * Revision 1.40 1999/09/08 12:12:39 adam - * Removed log message. - * - * Revision 1.39 1999/08/18 10:39:20 heikki - * Added a comment on memory leaks - * - * Revision 1.38 1999/08/18 08:38:04 heikki - * Memory leak hunting - * - * Revision 1.37 1999/07/14 13:21:34 heikki - * Added isam-d files. Compiles (almost) clean. Doesn't work at all - * - * Revision 1.36 1999/07/14 10:59:26 adam - * Changed functions isc_getmethod, isams_getmethod. - * Improved fatal error handling (such as missing EXPLAIN schema). - * - * Revision 1.35 1999/06/30 15:07:23 heikki - * Adding isamh stuff - * - * Revision 1.34 1999/05/26 07:49:13 adam - * C++ compilation. - * - * Revision 1.33 1999/05/15 14:36:38 adam - * Updated dictionary. Implemented "compression" of dictionary. - * - * Revision 1.32 1999/05/12 13:08:06 adam - * First version of ISAMS. - * - * Revision 1.31 1999/02/02 14:50:56 adam - * Updated WIN32 code specific sections. Changed header. - * - * Revision 1.30 1998/10/28 10:53:57 adam - * Added type cast to prevent warning. - * - * Revision 1.29 1998/06/11 15:41:39 adam - * Minor changes. - * - * Revision 1.28 1998/03/05 08:45:12 adam - * New result set model and modular ranking system. Moved towards - * descent server API. System information stored as "SGML" records. - * - * Revision 1.27 1998/02/17 10:32:52 adam - * Fixed bug: binary files weren't opened with flag b on NT. - * - * Revision 1.26 1998/01/29 13:39:13 adam - * Compress ISAM is default. - * - * Revision 1.25 1997/09/17 12:19:14 adam - * Zebra version corresponds to YAZ version 1.4. - * Changed Zebra server so that it doesn't depend on global common_resource. - * - * Revision 1.24 1997/09/09 13:38:07 adam - * Partial port to WIN95/NT. - * - * Revision 1.23 1997/09/04 13:57:39 adam - * Added O_BINARY for open calls. - * - * Revision 1.22 1997/02/12 20:39:45 adam - * Implemented options -f that limits the log to the first - * records. - * Changed some log messages also. - * - * Revision 1.21 1996/11/08 11:10:23 adam - * Buffers used during file match got bigger. - * Compressed ISAM support everywhere. - * Bug fixes regarding masking characters in queries. - * Redesigned Regexp-2 queries. - * - * Revision 1.20 1996/11/01 08:58:41 adam - * Interface to isamc system now includes update and delete. - * - * Revision 1.19 1996/10/29 14:09:46 adam - * Use of cisam system - enabled if setting isamc is 1. - * - * Revision 1.18 1996/06/04 10:18:59 adam - * Minor changes - removed include of ctype.h. - * - * Revision 1.17 1996/05/14 15:47:07 adam - * Cleanup of various buffer size entities. - * - * Revision 1.16 1996/04/09 10:05:20 adam - * Bug fix: prev_name buffer possibly too small; allocated in key_file_init. - * - * Revision 1.15 1996/03/21 14:50:09 adam - * File update uses modify-time instead of change-time. - * - * Revision 1.14 1996/02/07 14:06:37 adam - * Better progress report during register merge. - * New command: clean - removes temporary shadow files. - * - * Revision 1.13 1996/02/05 12:30:00 adam - * Logging reduced a bit. - * The remaining running time is estimated during register merge. - * - * Revision 1.12 1995/12/06 17:49:19 adam - * Uses dict_delete now. - * - * Revision 1.11 1995/12/06 16:06:43 adam - * Better diagnostics. Work on 'real' dictionary deletion. - * - * Revision 1.10 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 1.9 1995/10/10 12:24:39 adam - * Temporary sort files are compressed. - * - * Revision 1.8 1995/10/04 16:57:19 adam - * Key input and merge sort in one pass. - * - * Revision 1.7 1995/10/02 15:18:52 adam - * New member in recRetrieveCtrl: diagnostic. - * - * Revision 1.6 1995/09/29 15:51:56 adam - * First work on multi-way read. - * - * Revision 1.5 1995/09/29 14:01:43 adam - * Bug fixes. - * - * Revision 1.4 1995/09/28 14:22:57 adam - * Sort uses smaller temporary files. - * - * Revision 1.3 1995/09/06 16:11:17 adam - * Option: only one word key per file. - * - * Revision 1.2 1995/09/04 12:33:42 adam - * Various cleanup. YAZ util used instead. - * - * Revision 1.1 1995/09/04 09:10:37 adam - * More work on index add/del/update. - * Merge sort implemented. - * Initial work on z39 server. - * - */ - diff --git a/index/lockidx.c b/index/lockidx.c index 26af684..4904de4 100644 --- a/index/lockidx.c +++ b/index/lockidx.c @@ -4,7 +4,10 @@ * Sebastian Hammer, Adam Dickmeiss * * $Log: lockidx.c,v $ - * Revision 1.20 2000-10-16 20:16:00 adam + * Revision 1.21 2002-02-20 17:30:01 adam + * Work on new API. Locking system re-implemented + * + * Revision 1.20 2000/10/16 20:16:00 adam * Fixed problem with close of lock file for WIN32. * * Revision 1.19 2000/09/05 14:04:05 adam @@ -86,12 +89,13 @@ #include #include "index.h" +#include "zserver.h" static ZebraLockHandle server_lock_main = NULL; static ZebraLockHandle server_lock_cmt = NULL; static ZebraLockHandle server_lock_org = NULL; -int zebraIndexWait (int commitPhase) +int zebraIndexWait (ZebraHandle zh, int commitPhase) { ZebraLockHandle h; @@ -101,7 +105,7 @@ int zebraIndexWait (int commitPhase) { char path[1024]; - zebra_lock_prefix (common_resource, path); + zebra_lock_prefix (zh->service->res, path); strcat (path, FNAME_COMMIT_LOCK); server_lock_cmt = zebra_lock_create (path, 1); if (!server_lock_cmt) @@ -116,7 +120,7 @@ int zebraIndexWait (int commitPhase) { char path[1024]; - zebra_lock_prefix (common_resource, path); + zebra_lock_prefix (zh->service->res, path); strcat (path, FNAME_ORG_LOCK); server_lock_org = zebra_lock_create (path, 1); if (!server_lock_org) @@ -153,7 +157,7 @@ int zebraIndexWait (int commitPhase) } -void zebraIndexLockMsg (const char *str) +void zebraIndexLockMsg (ZebraHandle zh, const char *str) { char path[1024]; int l, r, fd; @@ -168,17 +172,17 @@ void zebraIndexLockMsg (const char *str) logf (LOG_FATAL|LOG_ERRNO, "write lock file"); exit (1); } - zebra_lock_prefix (common_resource, path); + zebra_lock_prefix (zh->service->res, path); strcat (path, FNAME_TOUCH_TIME); fd = creat (path, 0666); close (fd); } -void zebraIndexUnlock (void) +void zebraIndexUnlock (ZebraHandle zh) { char path[1024]; - zebra_lock_prefix (common_resource, path); + zebra_lock_prefix (zh->service->res, path); strcat (path, FNAME_MAIN_LOCK); #ifdef WIN32 zebra_lock_destroy (server_lock_main); @@ -192,7 +196,8 @@ void zebraIndexUnlock (void) server_lock_main = 0; } -int zebraIndexLock (BFiles bfs, int commitNow, const char *rval) +int zebraIndexLock (BFiles bfs, ZebraHandle zh, int commitNow, + const char *rval) { char path[1024]; char buf[256]; @@ -201,7 +206,7 @@ int zebraIndexLock (BFiles bfs, int commitNow, const char *rval) if (server_lock_main) return 0; - zebra_lock_prefix (common_resource, path); + zebra_lock_prefix (zh->service->res, path); strcat (path, FNAME_MAIN_LOCK); while (1) { diff --git a/index/lockutil.c b/index/lockutil.c index 123f7e6..325ac4a 100644 --- a/index/lockutil.c +++ b/index/lockutil.c @@ -1,50 +1,9 @@ /* - * Copyright (C) 1994-1999, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * - * $Log: lockutil.c,v $ - * Revision 1.12 1999-05-26 07:49:13 adam - * C++ compilation. - * - * Revision 1.11 1999/02/02 14:50:59 adam - * Updated WIN32 code specific sections. Changed header. - * - * Revision 1.10 1997/09/29 09:08:36 adam - * Revised locking system to be thread safe for the server. - * - * Revision 1.9 1997/09/25 14:54:43 adam - * WIN32 files lock support. - * - * Revision 1.8 1997/09/17 12:19:15 adam - * Zebra version corresponds to YAZ version 1.4. - * Changed Zebra server so that it doesn't depend on global common_resource. - * - * Revision 1.7 1997/09/09 13:38:08 adam - * Partial port to WIN95/NT. - * - * Revision 1.6 1996/10/29 14:08:14 adam - * Uses resource lockDir instead of lockPath. - * - * Revision 1.5 1996/03/26 16:01:13 adam - * New setting lockPath: directory of various lock files. - * - * Revision 1.4 1995/12/13 08:46:10 adam - * Locking uses F_WRLCK and F_RDLCK again! - * - * Revision 1.3 1995/12/12 16:00:57 adam - * System call sync(2) used after update/commit. - * Locking (based on fcntl) uses F_EXLCK and F_SHLCK instead of F_WRLCK - * and F_RDLCK. - * - * Revision 1.2 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 1.1 1995/12/07 17:38:47 adam - * Work locking mechanisms for concurrent updates/commit. - * + * $Id: lockutil.c,v 1.13 2002-02-20 17:30:01 adam Exp $ */ #include #include @@ -120,6 +79,24 @@ static int unixLock (int fd, int type, int cmd) } #endif +int zebra_lock_w (ZebraLockHandle h) +{ +#ifdef WIN32 + return _locking (h->fd, _LK_LOCK, 1); +#else + return unixLock (h->fd, F_WRLCK, F_SETLKW); +#endif +} + +int zebra_lock_r (ZebraLockHandle h) +{ +#ifdef WIN32 + return _locking (h->fd, _LK_LOCK, 1); +#else + return unixLock (h->fd, F_RDLCK, F_SETLKW); +#endif +} + int zebra_lock (ZebraLockHandle h) { #ifdef WIN32 diff --git a/index/main.c b/index/main.c index 60310e4..3508f2a 100644 --- a/index/main.c +++ b/index/main.c @@ -1,8 +1,8 @@ /* - * Copyright (C) 1994-2001, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * - * $Id: main.c,v 1.81 2001-11-19 23:29:09 adam Exp $ + * $Id: main.c,v 1.82 2002-02-20 17:30:01 adam Exp $ */ #include #include @@ -14,28 +14,19 @@ #endif #include +#include "zebraapi.h" +#include "zserver.h" #include "index.h" #include "recindex.h" -#ifndef ZEBRASDR -#define ZEBRASDR 0 -#endif - -#if ZEBRASDR -#include "zebrasdr.h" -#endif - char *prog; -Res common_resource = 0; - - int main (int argc, char **argv) { int ret; int cmd = 0; char *arg; - char *configName = FNAME_CONFIG; + char *configName = 0; int nsections = 0; int disableCommit = 0; size_t mem_max = 0; @@ -43,6 +34,8 @@ int main (int argc, char **argv) char nbuf[100]; #endif struct recordGroup rGroupDef; + ZebraService zs = 0; + ZebraHandle zh = 0; nmem_init (); @@ -52,10 +45,6 @@ int main (int argc, char **argv) yaz_log_init_prefix (nbuf); #endif -#if ZEBRASDR - zebraSdr_std (); - rGroupDef.useSDR = 0; -#endif rGroupDef.groupName = NULL; rGroupDef.databaseName = NULL; rGroupDef.path = NULL; @@ -67,10 +56,6 @@ int main (int argc, char **argv) rGroupDef.databaseNamePath = 0; rGroupDef.explainDatabase = 0; rGroupDef.fileVerboseLimit = 100000; - rGroupDef.zebra_maps = NULL; - rGroupDef.dh = data1_create (); - rGroupDef.recTypes = recTypes_init (rGroupDef.dh); - recTypes_default_handlers (rGroupDef.recTypes); prog = *argv; if (argc < 2) @@ -93,25 +78,18 @@ int main (int argc, char **argv) " -v Set logging to .\n" " -l Write log to .\n" " -f Display information for the first records.\n" -#if ZEBRASDR - " -S Use SDRKit\n" -#endif " -V Show version.\n", *argv ); exit (1); } while ((ret = options ("sVt:c:g:d:m:v:nf:l:" -#if ZEBRASDR - "S" -#endif , argv, argc, &arg)) != -2) { if (ret == 0) { - const char *rval; if(cmd == 0) /* command */ { - if (!common_resource) + if (!zs) { #if ZMBOL logf (LOG_LOG, "zmbol version %s %s", @@ -120,27 +98,9 @@ int main (int argc, char **argv) logf (LOG_LOG, "zebra version %s %s", ZEBRAVER, ZEBRADATE); #endif - common_resource = res_open (configName ? - configName : FNAME_CONFIG); - if (!common_resource) - { - logf (LOG_FATAL, "cannot read file `%s'", configName); - exit (1); - } - data1_set_tabpath (rGroupDef.dh, res_get (common_resource, - "profilePath")); - - rGroupDef.bfs = - bfs_create (res_get (common_resource, "register")); - if (!rGroupDef.bfs) - { - logf (LOG_FATAL, "Cannot access register"); - exit(1); - } + zs = zebra_start (configName ? configName : FNAME_CONFIG); - bf_lockDir (rGroupDef.bfs, - res_get (common_resource, "lockDir")); - rGroupDef.zebra_maps = zebra_maps_open (common_resource); + zh = zebra_open (zs); } if (!strcmp (arg, "update")) cmd = 'u'; @@ -154,83 +114,23 @@ int main (int argc, char **argv) cmd = 'd'; else if (!strcmp (arg, "init")) { - zebraIndexUnlock(); - rval = res_get (common_resource, "shadow"); - zebraIndexLock (rGroupDef.bfs, 0, rval); - if (rval && *rval) - bf_cache (rGroupDef.bfs, rval); - zebraIndexLockMsg ("w"); - bf_reset (rGroupDef.bfs); + zebra_init (zh); } else if (!strcmp (arg, "commit")) { - rval = res_get (common_resource, "shadow"); - zebraIndexLock (rGroupDef.bfs, 1, rval); - if (rval && *rval) - bf_cache (rGroupDef.bfs, rval); - else - { - logf (LOG_FATAL, "Cannot perform commit"); - logf (LOG_FATAL, "No shadow area defined"); - exit (1); - } - if (bf_commitExists (rGroupDef.bfs)) - { - logf (LOG_LOG, "commit start"); - zebraIndexLockMsg ("c"); - zebraIndexWait (1); - logf (LOG_LOG, "commit execute"); - bf_commitExec (rGroupDef.bfs); -#ifndef WIN32 - sync (); -#endif - zebraIndexLockMsg ("d"); - zebraIndexWait (0); - logf (LOG_LOG, "commit clean"); - bf_commitClean (rGroupDef.bfs, rval); - } - else - logf (LOG_LOG, "nothing to commit"); + zebra_commit (zh); } else if (!strcmp (arg, "clean")) { - rval = res_get (common_resource, "shadow"); - zebraIndexLock (rGroupDef.bfs, 1, rval); - if (bf_commitExists (rGroupDef.bfs)) - { - zebraIndexLockMsg ("d"); - zebraIndexWait (0); - logf (LOG_LOG, "commit clean"); - bf_commitClean (rGroupDef.bfs, rval); - } - else - logf (LOG_LOG, "nothing to clean"); + assert (!"todo"); } else if (!strcmp (arg, "stat") || !strcmp (arg, "status")) { - Records records; - rval = res_get (common_resource, "shadow"); - zebraIndexLock (rGroupDef.bfs, 0, rval); - if (rval && *rval) - { - bf_cache (rGroupDef.bfs, rval); - zebraIndexLockMsg ("r"); - } - records = rec_open (rGroupDef.bfs, 0, 0); - rec_prstat (records); - rec_close (&records); - inv_prstat (rGroupDef.bfs); + assert (!"todo"); } else if (!strcmp (arg, "compact")) { - rval = res_get (common_resource, "shadow"); - zebraIndexLock (rGroupDef.bfs, 0, rval); - if (rval && *rval) - { - bf_cache (rGroupDef.bfs, rval); - zebraIndexLockMsg ("r"); - } - inv_compact(rGroupDef.bfs); + zebra_compact (zh); } else { @@ -240,81 +140,28 @@ int main (int argc, char **argv) } else { - struct recordGroup rGroup; -#if ZMBOL -#else - /* For zebra, delete lock file and reset register */ - if (rGroupDef.flagRw) - { - zebraIndexUnlock(); - bf_reset (rGroupDef.bfs); - } -#endif - rval = res_get (common_resource, "shadow"); - zebraIndexLock (rGroupDef.bfs, 0, rval); - if (rGroupDef.flagRw) - { - if (rval && *rval && !disableCommit) - { - bf_cache (rGroupDef.bfs, rval); - zebraIndexLockMsg ("r"); - } - else - { - bf_cache (rGroupDef.bfs, 0); - zebraIndexLockMsg ("w"); - } - zebraIndexWait (0); - } - memcpy (&rGroup, &rGroupDef, sizeof(rGroup)); - rGroup.path = arg; + memcpy (&zh->rGroup, &rGroupDef, sizeof(rGroupDef)); + zebra_begin_trans (zh); + + zh->rGroup.path = arg; switch (cmd) { case 'u': - if (!key_open (&rGroup, mem_max)) - { - logf (LOG_LOG, "updating %s", rGroup.path); - repositoryUpdate (&rGroup); - nsections = key_close (&rGroup); - } - break; - case 'U': - if (!key_open (&rGroup, mem_max)) - { - logf (LOG_LOG, "updating (pass 1) %s", rGroup.path); - repositoryUpdate (&rGroup); - key_close (&rGroup); - } - nsections = 0; + zebra_repository_update (zh); break; case 'd': - if (!key_open (&rGroup,mem_max)) - { - logf (LOG_LOG, "deleting %s", rGroup.path); - repositoryDelete (&rGroup); - nsections = key_close (&rGroup); - } + zebra_repository_delete (zh); break; case 's': - logf (LOG_LOG, "dumping %s", rGroup.path); - repositoryShow (&rGroup); + logf (LOG_LOG, "dumping %s", zh->rGroup.path); + zebra_repository_show (zh); nsections = 0; break; - case 'm': - nsections = -1; - break; default: nsections = 0; } cmd = 0; - if (nsections) - { - logf (LOG_LOG, "merging with index"); - key_input (rGroup.bfs, nsections, 60, common_resource); -#ifndef WIN32 - sync (); -#endif - } + zebra_end_trans (zh); log_event_end (NULL, NULL); } } @@ -325,7 +172,7 @@ int main (int argc, char **argv) #else fprintf (stderr, "Zebra %s %s\n", ZEBRAVER, ZEBRADATE); #endif - fprintf (stderr, " (C) 1994-2001, Index Data ApS\n"); + fprintf (stderr, " (C) 1994-2002, Index Data ApS\n"); #ifdef WIN32 #ifdef _DEBUG fprintf (stderr, " WIN32 Debug\n"); @@ -358,20 +205,11 @@ int main (int argc, char **argv) rGroupDef.recordType = arg; else if (ret == 'n') disableCommit = 1; -#if ZEBRASDR - else if (ret == 'S') - rGroupDef.useSDR = 1; -#endif else logf (LOG_WARN, "unknown option '-%s'", arg); } - recTypes_destroy (rGroupDef.recTypes); - if (common_resource) - { - zebraIndexUnlock (); - bfs_destroy (rGroupDef.bfs); - } - data1_destroy (rGroupDef.dh); + zebra_close (zh); + zebra_stop (zs); exit (0); return 0; } diff --git a/index/trav.c b/index/trav.c index 150fc27..b855ead 100644 --- a/index/trav.c +++ b/index/trav.c @@ -4,7 +4,10 @@ * Sebastian Hammer, Adam Dickmeiss * * $Log: trav.c,v $ - * Revision 1.36 1999-05-15 14:36:38 adam + * Revision 1.37 2002-02-20 17:30:01 adam + * Work on new API. Locking system re-implemented + * + * Revision 1.36 1999/05/15 14:36:38 adam * Updated dictionary. Implemented "compression" of dictionary. * * Revision 1.35 1999/02/02 14:51:09 adam @@ -151,6 +154,7 @@ #include #include "index.h" +#include "zserver.h" static int repComp (const char *a, const char *b, size_t len) { @@ -159,7 +163,7 @@ static int repComp (const char *a, const char *b, size_t len) return memcmp (a, b, len); } -static void repositoryExtractR (int deleteFlag, char *rep, +static void repositoryExtractR (ZebraHandle zh, int deleteFlag, char *rep, struct recordGroup *rGroup, int level) { @@ -188,10 +192,10 @@ static void repositoryExtractR (int deleteFlag, char *rep, switch (e[i].kind) { case dirs_file: - fileExtract (NULL, rep, rGroup, deleteFlag); + fileExtract (zh, NULL, rep, rGroup, deleteFlag); break; case dirs_dir: - repositoryExtractR (deleteFlag, rep, rGroup, level+1); + repositoryExtractR (zh, deleteFlag, rep, rGroup, level+1); break; } } @@ -199,9 +203,10 @@ static void repositoryExtractR (int deleteFlag, char *rep, } -static void fileDeleteR (struct dirs_info *di, struct dirs_entry *dst, - const char *base, char *src, - struct recordGroup *rGroup) +static void fileDeleteR (ZebraHandle zh, + struct dirs_info *di, struct dirs_entry *dst, + const char *base, char *src, + struct recordGroup *rGroup) { char tmppath[1024]; size_t src_len = strlen (src); @@ -212,7 +217,7 @@ static void fileDeleteR (struct dirs_info *di, struct dirs_entry *dst, { case dirs_file: sprintf (tmppath, "%s%s", base, dst->path); - fileExtract (&dst->sysno, tmppath, rGroup, 1); + fileExtract (zh, &dst->sysno, tmppath, rGroup, 1); strcpy (tmppath, dst->path); dst = dirs_read (di); @@ -229,7 +234,8 @@ static void fileDeleteR (struct dirs_info *di, struct dirs_entry *dst, } } -static void fileUpdateR (struct dirs_info *di, struct dirs_entry *dst, +static void fileUpdateR (ZebraHandle zh, + struct dirs_info *di, struct dirs_entry *dst, const char *base, char *src, struct recordGroup *rGroup, int level) @@ -264,7 +270,7 @@ static void fileUpdateR (struct dirs_info *di, struct dirs_entry *dst, else if (!e_src) { strcpy (src, dst->path); - fileDeleteR (di, dst, base, src, rGroup); + fileDeleteR (zh, di, dst, base, src, rGroup); return; } else @@ -311,7 +317,7 @@ static void fileUpdateR (struct dirs_info *di, struct dirs_entry *dst, case dirs_file: if (e_src[i_src].mtime > dst->mtime) { - if (fileExtract (&dst->sysno, tmppath, rGroup, 0)) + if (fileExtract (zh, &dst->sysno, tmppath, rGroup, 0)) { dirs_add (di, src, dst->sysno, e_src[i_src].mtime); } @@ -321,7 +327,7 @@ static void fileUpdateR (struct dirs_info *di, struct dirs_entry *dst, dst = dirs_read (di); break; case dirs_dir: - fileUpdateR (di, dst, base, src, rGroup, level+1); + fileUpdateR (zh, di, dst, base, src, rGroup, level+1); dst = dirs_last (di); logf (LOG_DEBUG, "last is %s", dst ? dst->path : "null"); break; @@ -339,11 +345,11 @@ static void fileUpdateR (struct dirs_info *di, struct dirs_entry *dst, switch (e_src[i_src].kind) { case dirs_file: - if (fileExtract (&sysno, tmppath, rGroup, 0)) + if (fileExtract (zh, &sysno, tmppath, rGroup, 0)) dirs_add (di, src, sysno, e_src[i_src].mtime); break; case dirs_dir: - fileUpdateR (di, dst, base, src, rGroup, level+1); + fileUpdateR (zh, di, dst, base, src, rGroup, level+1); if (dst) dst = dirs_last (di); break; @@ -358,12 +364,12 @@ static void fileUpdateR (struct dirs_info *di, struct dirs_entry *dst, switch (dst->kind) { case dirs_file: - fileExtract (&dst->sysno, tmppath, rGroup, 1); + fileExtract (zh, &dst->sysno, tmppath, rGroup, 1); dirs_del (di, dst->path); dst = dirs_read (di); break; case dirs_dir: - fileDeleteR (di, dst, base, src, rGroup); + fileDeleteR (zh, di, dst, base, src, rGroup); dst = dirs_last (di); } } @@ -371,7 +377,7 @@ static void fileUpdateR (struct dirs_info *di, struct dirs_entry *dst, dir_free (&e_src); } -static void groupRes (struct recordGroup *rGroup) +static void groupRes (ZebraService zs, struct recordGroup *rGroup) { char resStr[256]; char gPrefix[256]; @@ -382,21 +388,23 @@ static void groupRes (struct recordGroup *rGroup) sprintf (gPrefix, "%s.", rGroup->groupName); sprintf (resStr, "%srecordId", gPrefix); - rGroup->recordId = res_get (common_resource, resStr); + rGroup->recordId = res_get (zs->res, resStr); sprintf (resStr, "%sdatabasePath", gPrefix); rGroup->databaseNamePath = - atoi (res_get_def (common_resource, resStr, "0")); + atoi (res_get_def (zs->res, resStr, "0")); } -void repositoryShow (struct recordGroup *rGroup) +void repositoryShow (ZebraHandle zh) + { + struct recordGroup *rGroup = &zh->rGroup; char src[1024]; int src_len; struct dirs_entry *dst; Dict dict; struct dirs_info *di; - if (!(dict = dict_open (rGroup->bfs, FMATCH_DICT, 50, 0, 0))) + if (!(dict = dict_open (zh->service->bfs, FMATCH_DICT, 50, 0, 0))) { logf (LOG_FATAL, "dict_open fail of %s", FMATCH_DICT); return; @@ -420,7 +428,8 @@ void repositoryShow (struct recordGroup *rGroup) dict_close (dict); } -static void fileUpdate (Dict dict, struct recordGroup *rGroup, +static void fileUpdate (ZebraHandle zh, + Dict dict, struct recordGroup *rGroup, const char *path) { struct dirs_info *di; @@ -443,13 +452,13 @@ static void fileUpdate (Dict dict, struct recordGroup *rGroup, if (e_dst) { if (sbuf.st_mtime > e_dst->mtime) - if (fileExtract (&e_dst->sysno, src, rGroup, 0)) + if (fileExtract (zh, &e_dst->sysno, src, rGroup, 0)) dirs_add (di, src, e_dst->sysno, sbuf.st_mtime); } else { SYSNO sysno = 0; - if (fileExtract (&sysno, src, rGroup, 0)) + if (fileExtract (zh, &sysno, src, rGroup, 0)) dirs_add (di, src, sysno, sbuf.st_mtime); } dirs_free (&di); @@ -463,7 +472,7 @@ static void fileUpdate (Dict dict, struct recordGroup *rGroup, } di = dirs_open (dict, src, rGroup->flagRw); *dst = '\0'; - fileUpdateR (di, dirs_read (di), src, dst, rGroup, 0); + fileUpdateR (zh, di, dirs_read (di), src, dst, rGroup, 0); dirs_free (&di); } else @@ -473,7 +482,8 @@ static void fileUpdate (Dict dict, struct recordGroup *rGroup, } -static void repositoryExtract (int deleteFlag, struct recordGroup *rGroup, +static void repositoryExtract (ZebraHandle zh, + int deleteFlag, struct recordGroup *rGroup, const char *path) { struct stat sbuf; @@ -484,34 +494,36 @@ static void repositoryExtract (int deleteFlag, struct recordGroup *rGroup, stat (src, &sbuf); if (S_ISREG(sbuf.st_mode)) - fileExtract (NULL, src, rGroup, deleteFlag); + fileExtract (zh, NULL, src, rGroup, deleteFlag); else if (S_ISDIR(sbuf.st_mode)) - repositoryExtractR (deleteFlag, src, rGroup, 0); + repositoryExtractR (zh, deleteFlag, src, rGroup, 0); else logf (LOG_WARN, "Ignoring path %s", src); } -static void repositoryExtractG (int deleteFlag, struct recordGroup *rGroup) +static void repositoryExtractG (ZebraHandle zh, + int deleteFlag, struct recordGroup *rGroup) { if (*rGroup->path == '\0' || !strcmp(rGroup->path, "-")) { char src[1024]; while (scanf ("%s", src) == 1) - repositoryExtract (deleteFlag, rGroup, src); + repositoryExtract (zh, deleteFlag, rGroup, src); } else - repositoryExtract (deleteFlag, rGroup, rGroup->path); + repositoryExtract (zh, deleteFlag, rGroup, rGroup->path); } -void repositoryUpdate (struct recordGroup *rGroup) +void repositoryUpdate (ZebraHandle zh) { - groupRes (rGroup); + struct recordGroup *rGroup = &zh->rGroup; + groupRes (zh->service, rGroup); assert (rGroup->path); if (rGroup->recordId && !strcmp (rGroup->recordId, "file")) { Dict dict; - if (!(dict = dict_open (rGroup->bfs, FMATCH_DICT, 50, + if (!(dict = dict_open (zh->service->bfs, FMATCH_DICT, 50, rGroup->flagRw, 0))) { logf (LOG_FATAL, "dict_open fail of %s", FMATCH_DICT); @@ -521,18 +533,18 @@ void repositoryUpdate (struct recordGroup *rGroup) { char src[1024]; while (scanf ("%s", src) == 1) - fileUpdate (dict, rGroup, src); + fileUpdate (zh, dict, rGroup, src); } else - fileUpdate (dict, rGroup, rGroup->path); + fileUpdate (zh, dict, rGroup, rGroup->path); dict_close (dict); } else - repositoryExtractG (0, rGroup); + repositoryExtractG (zh, 0, rGroup); } -void repositoryDelete (struct recordGroup *rGroup) +void repositoryDelete (ZebraHandle zh) { - repositoryExtractG (1, rGroup); + repositoryExtractG (zh, 1, &zh->rGroup); } diff --git a/index/zebraapi.c b/index/zebraapi.c index 3cd3a10..bb0c06f 100644 --- a/index/zebraapi.c +++ b/index/zebraapi.c @@ -1,155 +1,8 @@ /* - * Copyright (C) 1995-2000, Index Data + * Copyright (C) 1995-2002, Index Data * All rights reserved. * - * $Log: zebraapi.c,v $ - * Revision 1.44 2001-10-15 19:53:43 adam - * POSIX thread updates. First work on term sets. - * - * Revision 1.43 2000/12/05 12:22:53 adam - * Termlist source implemented (so that we can index values of XML/SGML - * attributes). - * - * Revision 1.42 2000/12/05 10:01:44 adam - * Fixed bug regarding user-defined attribute sets. - * - * Revision 1.41 2000/12/01 17:59:08 adam - * Fixed bug regarding online updates on WIN32. - * When zebra.cfg is not available the server will not abort. - * - * Revision 1.40 2000/11/29 15:21:31 adam - * Fixed problem with passwd db. - * - * Revision 1.39 2000/11/29 14:24:01 adam - * Script configure uses yaz pthreads options. Added locking for - * zebra_register_{lock,unlock}. - * - * Revision 1.38 2000/11/08 13:46:58 adam - * Fixed scan: server could break if bad attribute/database was selected. - * Work on remote update. - * - * Revision 1.37 2000/10/17 12:37:09 adam - * Fixed notification of live-updates. Fixed minor problem with mf_init - * where it didn't handle shadow area file names correctly. - * - * Revision 1.36 2000/09/06 08:59:36 adam - * Using read-only (for now) for server. - * - * Revision 1.35 2000/07/07 12:49:20 adam - * Optimized resultSetInsert{Rank,Sort}. - * - * Revision 1.34 2000/06/09 13:56:38 ian - * Added some logging on Authentication and searches. - * - * Revision 1.33 2000/05/18 12:01:36 adam - * System call times(2) used again. More 64-bit fixes. - * - * Revision 1.32 2000/04/19 14:35:59 adam - * WIN32 update (this version is known not to work on Windows). - * - * Revision 1.31 2000/04/05 10:07:02 adam - * Minor zebra compile fix. - * - * Revision 1.30 2000/04/05 09:49:35 adam - * On Unix, zebra/z'mbol uses automake. - * - * Revision 1.29 2000/03/20 19:08:36 adam - * Added remote record import using Z39.50 extended services and Segment - * Requests. - * - * Revision 1.28 2000/03/15 15:00:30 adam - * First work on threaded version. - * - * Revision 1.27 2000/02/24 12:31:17 adam - * Added zebra_string_norm. - * - * Revision 1.26 1999/11/30 13:48:03 adam - * Improved installation. Updated for inclusion of YAZ header files. - * - * Revision 1.25 1999/11/04 15:00:45 adam - * Implemented delete result set(s). - * - * Revision 1.24 1999/10/14 14:33:50 adam - * Added truncation 5=106. - * - * Revision 1.23 1999/09/07 11:36:32 adam - * Minor changes. - * - * Revision 1.22 1999/08/02 10:13:47 adam - * Fixed bug regarding zebra_hits. - * - * Revision 1.21 1999/07/14 10:59:26 adam - * Changed functions isc_getmethod, isams_getmethod. - * Improved fatal error handling (such as missing EXPLAIN schema). - * - * Revision 1.20 1999/07/06 12:28:04 adam - * Updated record index structure. Format includes version ID. Compression - * algorithm ID is stored for each record block. - * - * Revision 1.19 1999/05/26 07:49:13 adam - * C++ compilation. - * - * Revision 1.18 1999/05/15 14:36:38 adam - * Updated dictionary. Implemented "compression" of dictionary. - * - * Revision 1.17 1999/05/12 13:08:06 adam - * First version of ISAMS. - * - * Revision 1.16 1999/02/19 10:38:30 adam - * Implemented chdir-setting. - * - * Revision 1.15 1999/02/17 12:18:12 adam - * Fixed zebra_close so that a NULL pointer is ignored. - * - * Revision 1.14 1999/02/02 14:51:11 adam - * Updated WIN32 code specific sections. Changed header. - * - * Revision 1.13 1998/12/16 12:23:30 adam - * Added facility for database name mapping using resource mapdb. - * - * Revision 1.12 1998/11/16 10:18:10 adam - * Better error reporting for result sets. - * - * Revision 1.11 1998/10/16 08:14:34 adam - * Updated record control system. - * - * Revision 1.10 1998/09/22 10:03:42 adam - * Changed result sets to be persistent in the sense that they can - * be re-searched if needed. - * Fixed memory leak in rsm_or. - * - * Revision 1.9 1998/09/02 13:53:17 adam - * Extra parameter decode added to search routines to implement - * persistent queries. - * - * Revision 1.8 1998/08/24 17:29:23 adam - * Minor changes. - * - * Revision 1.7 1998/06/24 12:16:13 adam - * Support for relations on text operands. Open range support in - * DFA module (i.e. [-j], [g-]). - * - * Revision 1.6 1998/06/22 11:36:47 adam - * Added authentication check facility to zebra. - * - * Revision 1.5 1998/06/13 00:14:08 adam - * Minor changes. - * - * Revision 1.4 1998/06/12 12:22:12 adam - * Work on Zebra API. - * - * Revision 1.3 1998/05/27 16:57:44 adam - * Zebra returns surrogate diagnostic for single records when - * appropriate. - * - * Revision 1.2 1998/05/20 10:12:19 adam - * Implemented automatic EXPLAIN database maintenance. - * Modified Zebra to work with ASN.1 compiled version of YAZ. - * - * Revision 1.1 1998/03/05 08:45:13 adam - * New result set model and modular ranking system. Moved towards - * descent server API. System information stored as "SGML" records. - * + * $Id: zebraapi.c,v 1.45 2002-02-20 17:30:01 adam Exp $ */ #include @@ -179,93 +32,23 @@ static void zebra_chdir (ZebraService zh) #endif } -static 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); - -static int explain_extract (void *handle, Record rec, data1_node *n); -static void extract_index (ZebraHandle zh); - -static void zebra_register_unlock (ZebraHandle zh); - -static int zebra_register_activate (ZebraService zh, int rw); -static int zebra_register_deactivate (ZebraService zh); - -static int zebra_register_lock (ZebraHandle zh, int rw) +static void zebra_flush_reg (ZebraHandle zh) { - time_t lastChange; - int state; - zh->errCode = 0; - zh->errString = 0; - if (!zh->service->active) - { - zh->errCode = 1019; - return 1; - } + if (zh->service->matchDict) + dict_close (zh->service->matchDict); + zh->service->matchDict = 0; + zebraExplain_flush (zh->service->zei, 1, zh); -#if HAVE_SYS_TIMES_H - times (&zh->tms1); -#endif - - zebra_mutex_cond_lock (&zh->service->session_lock); - - state = zebra_server_lock_get_state(zh->service, &lastChange); - - zebra_server_lock (zh->service, state); - - switch (state) - { - case 'c': - state = 1; - break; - default: - state = 0; - } - if (rw) - logf (LOG_LOG, "Register in read/write mode"); - else if (zh->service->registerState == state) - { - logf (LOG_DEBUG, "registerChange = %ld lastChange = %ld", - (long) zh->service->registerChange, (long)lastChange); - if (zh->service->registerChange >= lastChange) - { - return 0; - } - logf (LOG_LOG, "Register completely updated since last access"); - } - else if (zh->service->registerState == -1) - logf (LOG_LOG, "Reading register using state %d pid=%ld", state, - (long) getpid()); - else - logf (LOG_LOG, "Register has changed state from %d to %d", - zh->service->registerState, state); - zh->service->registerChange = lastChange; - - zebra_register_deactivate (zh->service); - - zh->service->registerState = state; - - zebra_register_activate (zh->service, rw); - return 0; + extract_flushWriteKeys (zh); + zebra_index_merge (zh); } -static void zebra_register_unlock (ZebraHandle zh) -{ - if (zh->service->registerState != -1) - zebra_server_unlock (zh->service, zh->service->registerState); - zebra_mutex_cond_unlock (&zh->service->session_lock); -#if HAVE_SYS_TIMES_H - times (&zh->tms2); - logf (LOG_LOG, "user/system: %ld/%ld", - (long) (zh->tms2.tms_utime - zh->tms1.tms_utime), - (long) (zh->tms2.tms_stime - zh->tms1.tms_stime)); +static int zebra_register_activate (ZebraHandle zh, int rw, int useshadow); +static int zebra_register_deactivate (ZebraHandle zh); -#endif -} +static int zebra_begin_read (ZebraHandle zh); +static void zebra_end_read (ZebraHandle zh); ZebraHandle zebra_open (ZebraService zs) { @@ -284,9 +67,19 @@ ZebraHandle zebra_open (ZebraService zs) zh->errCode = 0; zh->errString = 0; + zh->trans_no = 0; + zh->seqno = 0; + zh->last_val = 0; + + zh->lock_normal = zebra_lock_create ("norm.LCK", 0); + zh->lock_shadow = zebra_lock_create ("shadow.LCK", 0); + zh->key_buf = 0; zh->admin_databaseName = 0; - + + zh->keys.buf_max = 0; + zh->keys.buf = 0; + zebra_mutex_cond_lock (&zs->session_lock); zh->next = zs->sessions; @@ -318,7 +111,7 @@ ZebraService zebra_start (const char *configName) // return zh; } zebra_chdir (zh); - zebra_server_lock_init (zh); + zebra_mutex_cond_init (&zh->session_lock); if (!res_get (zh->res, "passwd")) zh->passwd_db = NULL; @@ -334,115 +127,121 @@ ZebraService zebra_start (const char *configName) return zh; } -static int zebra_register_activate (ZebraService zh, int rw) +static int zebra_register_activate (ZebraHandle zh, int rw, int useshadow) { - if (zh->active > 1) - { - yaz_log (LOG_LOG, "zebra_register_activate (ignored since active=%d)", - zh->active); - return 0; - } - yaz_log (LOG_LOG, "zebra_register_activate shadow=%s", - zh->registerState ? "yes" : "no"); + ZebraService zs = zh->service; + int record_compression = REC_COMPRESS_NONE; + char *recordCompression = 0; + + yaz_log (LOG_LOG, "zebra_open_register_activate rw = %d useshadow=%d", + rw, useshadow); - zh->dh = data1_create (); - if (!zh->dh) + zs->dh = data1_create (); + if (!zs->dh) return -1; - zh->bfs = bfs_create (res_get (zh->res, "register")); - if (!zh->bfs) + zs->bfs = bfs_create (res_get (zs->res, "register")); + if (!zs->bfs) { - data1_destroy(zh->dh); + data1_destroy(zs->dh); return -1; } - bf_lockDir (zh->bfs, res_get (zh->res, "lockDir")); - bf_cache (zh->bfs, zh->registerState ? res_get (zh->res, "shadow") : NULL); - data1_set_tabpath (zh->dh, res_get(zh->res, "profilePath")); - zh->recTypes = recTypes_init (zh->dh); - recTypes_default_handlers (zh->recTypes); - - zh->records = NULL; - zh->zebra_maps = zebra_maps_open (zh->res); - zh->rank_classes = NULL; - - zh->records = 0; - zh->dict = 0; - zh->sortIdx = 0; - zh->isams = 0; + bf_lockDir (zs->bfs, res_get (zs->res, "lockDir")); + if (useshadow) + bf_cache (zs->bfs, res_get (zs->res, "shadow")); + data1_set_tabpath (zs->dh, res_get(zs->res, "profilePath")); + zs->recTypes = recTypes_init (zs->dh); + recTypes_default_handlers (zs->recTypes); + + zs->records = NULL; + zs->zebra_maps = zebra_maps_open (zs->res); + zs->rank_classes = NULL; + + zs->records = 0; + zs->dict = 0; + zs->sortIdx = 0; + zs->isams = 0; #if ZMBOL - zh->isam = 0; - zh->isamc = 0; - zh->isamd = 0; + zs->isam = 0; + zs->isamc = 0; + zs->isamd = 0; #endif - zh->zei = 0; + zs->zei = 0; - zebraRankInstall (zh, rank1_class); + zebraRankInstall (zs, rank1_class); + + recordCompression = res_get_def (zh->service->res, + "recordCompression", "none"); + if (!strcmp (recordCompression, "none")) + record_compression = REC_COMPRESS_NONE; + if (!strcmp (recordCompression, "bzip2")) + record_compression = REC_COMPRESS_BZIP2; - if (!(zh->records = rec_open (zh->bfs, rw, 0))) + if (!(zs->records = rec_open (zs->bfs, rw, record_compression))) { logf (LOG_WARN, "rec_open"); return -1; } - if (!(zh->dict = dict_open (zh->bfs, FNAME_DICT, 80, rw, 0))) + if (!(zs->dict = dict_open (zs->bfs, FNAME_DICT, 80, rw, 0))) { logf (LOG_WARN, "dict_open"); return -1; } - if (!(zh->sortIdx = sortIdx_open (zh->bfs, rw))) + if (!(zs->sortIdx = sortIdx_open (zs->bfs, rw))) { logf (LOG_WARN, "sortIdx_open"); return -1; } - if (res_get_match (zh->res, "isam", "s", ISAM_DEFAULT)) + if (res_get_match (zs->res, "isam", "s", ISAM_DEFAULT)) { struct ISAMS_M_s isams_m; - if (!(zh->isams = isams_open (zh->bfs, FNAME_ISAMS, rw, - key_isams_m(zh->res, &isams_m)))) + if (!(zs->isams = isams_open (zs->bfs, FNAME_ISAMS, rw, + key_isams_m(zs->res, &isams_m)))) { logf (LOG_WARN, "isams_open"); return -1; } } #if ZMBOL - else if (res_get_match (zh->res, "isam", "i", ISAM_DEFAULT)) + else if (res_get_match (zs->res, "isam", "i", ISAM_DEFAULT)) { - if (!(zh->isam = is_open (zh->bfs, FNAME_ISAM, key_compare, rw, - sizeof (struct it_key), zh->res))) + if (!(zs->isam = is_open (zs->bfs, FNAME_ISAM, key_compare, rw, + sizeof (struct it_key), zs->res))) { logf (LOG_WARN, "is_open"); return -1; } } - else if (res_get_match (zh->res, "isam", "c", ISAM_DEFAULT)) + else if (res_get_match (zs->res, "isam", "c", ISAM_DEFAULT)) { struct ISAMC_M_s isamc_m; - if (!(zh->isamc = isc_open (zh->bfs, FNAME_ISAMC, - rw, key_isamc_m(zh->res, &isamc_m)))) + if (!(zs->isamc = isc_open (zs->bfs, FNAME_ISAMC, + rw, key_isamc_m(zs->res, &isamc_m)))) { logf (LOG_WARN, "isc_open"); return -1; } } - else if (res_get_match (zh->res, "isam", "d", ISAM_DEFAULT)) + else if (res_get_match (zs->res, "isam", "d", ISAM_DEFAULT)) { struct ISAMD_M_s isamd_m; - if (!(zh->isamd = isamd_open (zh->bfs, FNAME_ISAMD, - rw, key_isamd_m(zh->res, &isamd_m)))) + if (!(zs->isamd = isamd_open (zs->bfs, FNAME_ISAMD, + rw, key_isamd_m(zs->res, &isamd_m)))) { logf (LOG_WARN, "isamd_open"); return -1; } } #endif - zh->zei = zebraExplain_open (zh->records, zh->dh, - zh->res, rw, 0 /* rGroup */, + zs->zei = zebraExplain_open (zs->records, zs->dh, + zs->res, rw, zh, explain_extract); - if (!zh->zei) + if (!zs->zei) { logf (LOG_WARN, "Cannot obtain EXPLAIN information"); return -1; } - zh->active = 2; + zs->active = 2; yaz_log (LOG_LOG, "zebra_register_activate ok"); return 0; } @@ -452,7 +251,7 @@ void zebra_admin_shutdown (ZebraHandle zh) zebra_mutex_cond_lock (&zh->service->session_lock); zh->service->stop_flag = 1; if (!zh->service->sessions) - zebra_register_deactivate(zh->service); + zebra_register_deactivate(zh); zh->service->active = 0; zebra_mutex_cond_unlock (&zh->service->session_lock); } @@ -467,8 +266,9 @@ void zebra_admin_start (ZebraHandle zh) zebra_mutex_cond_unlock (&zs->session_lock); } -static int zebra_register_deactivate (ZebraService zs) +static int zebra_register_deactivate (ZebraHandle zh) { + ZebraService zs = zh->service; zs->stop_flag = 0; if (zs->active <= 1) { @@ -495,6 +295,8 @@ static int zebra_register_deactivate (ZebraService zs) #endif rec_close (&zs->records); } + resultSetInvalidate (zh); + recTypes_destroy (zs->recTypes); zebra_maps_close (zs->zebra_maps); zebraRankDestroy (zs); @@ -515,13 +317,15 @@ void zebra_stop(ZebraService zs) zebra_mutex_cond_lock (&zs->session_lock); while (zs->sessions) + { + zebra_register_deactivate(zs->sessions); zebra_close (zs->sessions); - + } + zebra_mutex_cond_unlock (&zs->session_lock); zebra_mutex_cond_destroy (&zs->session_lock); - zebra_register_deactivate(zs); res_close (zs->res); xfree (zs->configName); xfree (zs); @@ -529,9 +333,13 @@ void zebra_stop(ZebraService zs) void zebra_close (ZebraHandle zh) { - ZebraService zs = zh->service; + ZebraService zs; struct zebra_session **sp; + if (!zh) + return; + + zs = zh->service; yaz_log (LOG_LOG, "zebra_close zh=%p", zh); if (!zh) return ; @@ -542,8 +350,11 @@ void zebra_close (ZebraHandle zh) xfree (zh->key_buf); zh->key_buf = 0; } + xfree (zh->admin_databaseName); zebra_mutex_cond_lock (&zs->session_lock); + zebra_lock_destroy (zh->lock_normal); + zebra_lock_destroy (zh->lock_shadow); sp = &zs->sessions; while (1) { @@ -555,8 +366,8 @@ void zebra_close (ZebraHandle zh) } sp = &(*sp)->next; } - if (!zs->sessions && zs->stop_flag) - zebra_register_deactivate(zs); +// if (!zs->sessions && zs->stop_flag) +// zebra_register_deactivate(zs); zebra_mutex_cond_unlock (&zs->session_lock); xfree (zh); } @@ -634,12 +445,12 @@ void zebra_search_rpn (ZebraHandle zh, ODR stream, ODR decode, const char *setname) { zh->hits = 0; - if (zebra_register_lock (zh, 0)) + if (zebra_begin_read (zh)) return; map_basenames (zh, stream, &num_bases, &basenames); resultSetAddRPN (zh, stream, decode, query, num_bases, basenames, setname); - zebra_register_unlock (zh); + zebra_end_read (zh); logf(LOG_APP,"SEARCH:%d:",zh->hits); } @@ -652,8 +463,9 @@ void zebra_records_retrieve (ZebraHandle zh, ODR stream, ZebraPosSet poset; int i, *pos_array; - if (zebra_register_lock (zh, 0)) + if (zebra_begin_read (zh)) return; + pos_array = (int *) xmalloc (num_recs * sizeof(*pos_array)); for (i = 0; imem, num_input_setnames, input_setnames, output_setname, sort_sequence, sort_status); - zebra_register_unlock (zh); + zebra_end_read(zh); } int zebra_deleleResultSet(ZebraHandle zh, int function, @@ -738,7 +550,7 @@ int zebra_deleleResultSet(ZebraHandle zh, int function, int *statuses) { int i, status; - if (zebra_register_lock (zh, 0)) + if (zebra_begin_read(zh)) return Z_DeleteStatus_systemProblemAtTarget; switch (function) { @@ -749,7 +561,7 @@ int zebra_deleleResultSet(ZebraHandle zh, int function, resultSetDestroy (zh, -1, 0, statuses); break; } - zebra_register_unlock (zh); + zebra_end_read (zh); status = Z_DeleteStatus_success; for (i = 0; iadmin_databaseName); zh->admin_databaseName = xstrdup(database); } void zebra_admin_import_end (ZebraHandle zh) { - zebraExplain_flush (zh->service->zei, 1, zh); - extract_index (zh); - zebra_register_unlock (zh); + zebra_end_trans (zh); } void zebra_admin_import_segment (ZebraHandle zh, Z_Segment *segment) @@ -844,21 +653,19 @@ void zebra_admin_import_segment (ZebraHandle zh, Z_Segment *segment) void zebra_admin_create (ZebraHandle zh, const char *database) { - ZebraService zs = zh->service; - if (zebra_register_lock(zh, 1)) - { - zh->errCode = 1019; - return; - } + ZebraService zs; + + zebra_begin_trans (zh); + + zs = zh->service; /* announce database */ - if (zebraExplain_newDatabase (zs->zei, database, 0 /* explainDatabase */)) + if (zebraExplain_newDatabase (zh->service->zei, database, 0 + /* explainDatabase */)) { zh->errCode = 224; zh->errString = "Database already exist"; } - zebraExplain_flush (zh->service->zei, 1, zh); - extract_index (zh); - zebra_register_unlock(zh); + zebra_end_trans (zh); } int zebra_string_norm (ZebraHandle zh, unsigned reg_id, @@ -880,837 +687,296 @@ int zebra_string_norm (ZebraHandle zh, unsigned reg_id, return wrbuf_len(wrbuf); } -static void extract_init (struct recExtractCtrl *p, RecWord *w) + +void zebra_set_state (int val, int seqno) { - w->zebra_maps = p->zebra_maps; - w->seqnos = p->seqno; - w->attrSet = VAL_BIB1; - w->attrUse = 1016; - w->reg_type = 'w'; - w->extractCtrl = p; + long p = getpid(); + FILE *f = fopen ("state.LCK", "w"); + fprintf (f, "%c %d %ld\n", val, seqno, p); + fclose (f); } -static void extract_add_index_string (RecWord *p, const char *string, - int length) +void zebra_get_state (char *val, int *seqno) { - 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->service->zei; - struct recKeys *keys = &zh->keys; - - if (keys->buf_used+1024 > keys->buf_max) - { - char *b; - - b = (char *) xmalloc (keys->buf_max += 128000); - if (keys->buf_used > 0) - memcpy (b, keys->buf, keys->buf_used); - xfree (keys->buf); - keys->buf = b; - } - dst = keys->buf + keys->buf_used; + FILE *f = fopen ("state.LCK", "r"); - 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) - { - int ch = zebraExplain_lookupSU (zei, attrSet, attrUse); - if (ch < 0) - { - ch = zebraExplain_addSU (zei, attrSet, attrUse); - yaz_log (LOG_LOG, "addSU set=%d use=%d SU=%d", - attrSet, attrUse, ch); - } - assert (ch > 0); - memcpy (dst, &ch, sizeof(ch)); - dst += sizeof(ch); - } -#else - if (!(lead & 1)) - { - 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'; + *val = 'o'; + *seqno = 0; - if (!diff) + if (f) { - memcpy (dst, pseqno, sizeof(*pseqno)); - dst += sizeof(*pseqno); + fscanf (f, "%c %d", val, seqno); + fclose (f); } - keys->buf_used = dst - keys->buf; - if (*pseqno) - (*pseqno)++; } -static void extract_add_sort_string (RecWord *p, const char *string, - int length) +static int zebra_begin_read (ZebraHandle zh) { - struct sortKey *sk; - ZebraHandle zh = p->extractCtrl->handle; - struct sortKey *sortKeys = zh->sortKeys; - - for (sk = sortKeys; sk; sk = sk->next) - if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse) - return; - - sk = (struct sortKey *) xmalloc (sizeof(*sk)); - sk->next = sortKeys; - sortKeys = sk; - - sk->string = (char *) xmalloc (length); - sk->length = length; - memcpy (sk->string, string, length); - - sk->attrSet = p->attrSet; - sk->attrUse = p->attrUse; -} + int dirty = 0; + char val; + int seqno; -static void extract_add_string (RecWord *p, const char *string, int length) -{ - assert (length > 0); - if (zebra_maps_is_sort (p->zebra_maps, p->reg_type)) - extract_add_sort_string (p, string, length); - else - extract_add_index_string (p, string, length); -} + zebra_flush_reg (zh); -static void extract_add_incomplete_field (RecWord *p) -{ - const char *b = p->string; - int remain = p->length; - const char **map = 0; + (zh->trans_no)++; - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); + if (zh->trans_no != 1) + return 0; - while (map) + zebra_get_state (&val, &seqno); + if (val == 'd') + val = 'o'; + if (seqno != zh->seqno) { - 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); + yaz_log (LOG_LOG, "reopen seqno cur/old %d/%d", seqno, zh->seqno); + dirty = 1; } - (p->seqnos[p->reg_type])++; /* to separate this from next one */ -} - -static void extract_add_complete_field (RecWord *p) -{ - const char *b = p->string; - char buf[IT_MAX_WORD+1]; - const char **map = 0; - int i = 0, remain = p->length; - - if (remain > 0) - map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain); - - while (remain > 0 && i < IT_MAX_WORD) + else if (zh->last_val != val) { - while (map && *map && **map == *CHR_SPACE) - { - remain = p->length - (b - p->string); - if (remain > 0) - map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain); - else - map = 0; - } - if (!map) - break; - - if (i && i < IT_MAX_WORD) - buf[i++] = *CHR_SPACE; - while (map && *map && **map != *CHR_SPACE) - { - const char *cp = *map; - - if (i >= IT_MAX_WORD) - break; - while (i < IT_MAX_WORD && *cp) - buf[i++] = *(cp++); - remain = p->length - (b - p->string); - if (remain > 0) - map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, - remain); - else - map = 0; - } + yaz_log (LOG_LOG, "reopen last cur/old %d/%d", val, zh->last_val); + dirty = 1; } - if (!i) - return; - extract_add_string (p, buf, i); -} + if (!dirty) + return 0; -static void extract_token_add (RecWord *p) -{ - WRBUF wrbuf; - if ((wrbuf = zebra_replace(p->zebra_maps, p->reg_type, 0, - p->string, p->length))) - { - p->string = wrbuf_buf(wrbuf); - p->length = wrbuf_len(wrbuf); - } - if (zebra_maps_is_complete (p->zebra_maps, p->reg_type)) - extract_add_complete_field (p); + if (val == 'c') + zebra_lock_r (zh->lock_shadow); else - extract_add_incomplete_field(p); -} + zebra_lock_r (zh->lock_normal); + + zh->last_val = val; + zh->seqno = seqno; -static void extract_schema_add (struct recExtractCtrl *p, Odr_oid *oid) -{ - ZebraHandle zh = (ZebraHandle) (p->handle); - zebraExplain_addSchema (zh->service->zei, oid); + zebra_register_deactivate (zh); + + zebra_register_activate (zh, 0, val == 'c' ? 1 : 0); + return 0; } -static void extract_flushSortKeys (ZebraHandle zh, SYSNO sysno, - int cmd, struct sortKey **skp) +static void zebra_end_read (ZebraHandle zh) { - struct sortKey *sk = *skp; - SortIdx sortIdx = zh->service->sortIdx; - - sortIdx_sysno (sortIdx, sysno); - while (sk) - { - struct sortKey *sk_next = sk->next; - sortIdx_type (sortIdx, sk->attrUse); - sortIdx_add (sortIdx, sk->string, sk->length); - xfree (sk->string); - xfree (sk); - sk = sk_next; - } - *skp = 0; -} + (zh->trans_no)--; -struct encode_info { - int sysno; - int seqno; - int cmd; - char buf[768]; -}; + if (zh->trans_no != 0) + return; -void encode_key_init (struct encode_info *i) -{ - i->sysno = 0; - i->seqno = 0; - i->cmd = -1; + zebra_unlock (zh->lock_normal); + zebra_unlock (zh->lock_shadow); } -char *encode_key_int (int d, char *bp) +void zebra_begin_trans (ZebraHandle zh) { - if (d <= 63) - *bp++ = d; - else if (d <= 16383) - { - *bp++ = 64 + (d>>8); - *bp++ = d & 255; - } - else if (d <= 4194303) - { - *bp++ = 128 + (d>>16); - *bp++ = (d>>8) & 255; - *bp++ = d & 255; - } - else - { - *bp++ = 192 + (d>>24); - *bp++ = (d>>16) & 255; - *bp++ = (d>>8) & 255; - *bp++ = d & 255; - } - return bp; -} + int pass; + int seqno = 0; + char val = '?'; + const char *rval; -void encode_key_write (char *k, struct encode_info *i, FILE *outf) -{ - struct it_key key; - char *bp = i->buf; - - while ((*bp++ = *k++)) - ; - memcpy (&key, k+1, sizeof(struct it_key)); - bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp); - if (i->sysno != key.sysno) - { - i->sysno = key.sysno; - i->seqno = 0; - } - 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) + (zh->trans_no++); + if (zh->trans_no != 1) { - logf (LOG_FATAL|LOG_ERRNO, "fwrite"); - exit (1); + return; } -} -static void extract_flushWriteKeys (ZebraHandle zh) -{ - FILE *outf; - char out_fname[200]; - char *prevcp, *cp; - struct encode_info encode_info; - int ptr_i = zh->ptr_i; -#if SORT_EXTRA - int i; + yaz_log (LOG_LOG, "zebra_begin_trans"); +#if HAVE_SYS_TIMES_H + times (&zh->tms1); #endif - if (!zh->key_buf || ptr_i <= 0) - return; - (zh->key_file_no)++; - logf (LOG_LOG, "sorting section %d", (zh->key_file_no)); -#if !SORT_EXTRA - qsort (zh->key_buf + zh->ptr_top - ptr_i, ptr_i, sizeof(char*), - key_qsort_compare); - extract_get_fname_tmp (zh, out_fname, zh->key_file_no); + /* lock */ + rval = res_get (zh->service->res, "shadow"); - if (!(outf = fopen (out_fname, "wb"))) - { - logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname); - exit (1); - } - logf (LOG_LOG, "writing section %d", zh->key_file_no); - prevcp = cp = (zh->key_buf)[zh->ptr_top - ptr_i]; - - encode_key_init (&encode_info); - encode_key_write (cp, &encode_info, outf); - - while (--ptr_i > 0) + for (pass = 0; pass < 2; pass++) { - cp = (zh->key_buf)[zh->ptr_top - ptr_i]; - if (strcmp (cp, prevcp)) + if (rval) { - encode_key_init (&encode_info); - encode_key_write (cp, &encode_info, outf); - prevcp = cp; + zebra_lock_r (zh->lock_normal); + zebra_lock_w (zh->lock_shadow); } 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) + zebra_lock_w (zh->lock_normal); + zebra_lock_w (zh->lock_shadow); + } + + zebra_get_state (&val, &seqno); + if (val == 'c') + { + yaz_log (LOG_LOG, "previous transaction didn't finish commit"); + zebra_unlock (zh->lock_shadow); + zebra_unlock (zh->lock_normal); + zebra_commit (zh); + continue; + } + else if (val == 'd') + { + if (rval) { - cp = key_buf[ptr_top-ptr_i]; - encode_key_write (cp+key_y_len, &encode_info, outf); + BFiles bfs = bfs_create (res_get (zh->service->res, "shadow")); + yaz_log (LOG_LOG, "previous transaction didn't reach commit"); + bf_commitClean (bfs, rval); + bfs_destroy (bfs); + } + else + { + yaz_log (LOG_WARN, "your previous transaction didn't finish"); } - 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); + break; } - logf (LOG_LOG, "finished section %d", zh->key_file_no); - zh->ptr_i = 0; - zh->key_buf_used = 0; -} - -static void extract_flushRecordKeys (ZebraHandle zh, SYSNO sysno, - int cmd, struct recKeys *reckeys) -{ -#if SU_SCHEME -#else - unsigned char attrSet = (unsigned char) -1; - unsigned short attrUse = (unsigned short) -1; -#endif - int seqno = 0; - int off = 0; - int ch = 0; - ZebraExplainInfo zei = zh->service->zei; - - if (!zh->key_buf) + if (pass == 2) { - int mem = 8*1024*1024; - zh->key_buf = (char**) xmalloc (mem); - zh->ptr_top = mem/sizeof(char*); - zh->ptr_i = 0; - zh->key_buf_used = 0; - zh->key_file_no = 0; + yaz_log (LOG_FATAL, "zebra_begin_trans couldn't finish commit"); + abort(); + return; } - zebraExplain_recordCountIncrement (zei, cmd ? 1 : -1); - while (off < reckeys->buf_used) - { - const char *src = reckeys->buf + off; - struct it_key key; - int lead; - - lead = *src++; + zebra_set_state ('d', seqno); -#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->key_buf_used + 1024 > (zh->ptr_top-zh->ptr_i)*sizeof(char*)) - extract_flushWriteKeys (zh); - ++(zh->ptr_i); - (zh->key_buf)[zh->ptr_top - zh->ptr_i] = - (char*)zh->key_buf + zh->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->key_buf_used += - key_SU_encode (ch,((char*)zh->key_buf) + zh->key_buf_used); - - while (*src) - ((char*)zh->key_buf) [(zh->key_buf_used)++] = *src++; - src++; - ((char*)(zh->key_buf))[(zh->key_buf_used)++] = '\0'; - ((char*)(zh->key_buf))[(zh->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->key_buf + zh->key_buf_used, &key, sizeof(key)); - (zh->key_buf_used) += sizeof(key); - off = src - reckeys->buf; - } - assert (off == reckeys->buf_used); + zebra_register_activate (zh, 1, rval ? 1 : 0); + zh->service->matchDict = dict_open (zh->service->bfs, GMATCH_DICT, + 50, 1, 0); + zh->seqno = seqno; } -static void extract_index (ZebraHandle zh) +void zebra_end_trans (ZebraHandle zh) { - extract_flushWriteKeys (zh); - zebra_index_merge (zh); -} + char val; + int seqno; + const char *rval; -static int explain_extract (void *handle, Record rec, data1_node *n) -{ - ZebraHandle zh = (ZebraHandle) handle; - struct recExtractCtrl extractCtrl; - int i; + zh->trans_no--; + if (zh->trans_no != 0) + return; - if (zebraExplain_curDatabase (zh->service->zei, - rec->info[recInfo_databaseName])) - { - abort(); - if (zebraExplain_newDatabase (zh->service->zei, - rec->info[recInfo_databaseName], 0)) - abort (); - } + yaz_log (LOG_LOG, "zebra_end_trans"); + rval = res_get (zh->service->res, "shadow"); - zh->keys.buf_used = 0; - zh->keys.prevAttrUse = -1; - zh->keys.prevAttrSet = -1; - zh->keys.prevSeqNo = 0; - zh->sortKeys = 0; - - extractCtrl.init = extract_init; - extractCtrl.tokenAdd = extract_token_add; - extractCtrl.schemaAdd = extract_schema_add; - extractCtrl.dh = zh->service->dh; - for (i = 0; i<256; i++) - extractCtrl.seqno[i] = 0; - extractCtrl.zebra_maps = zh->service->zebra_maps; - extractCtrl.flagShowRecords = 0; - extractCtrl.handle = handle; - - grs_extract_tree(&extractCtrl, n); + zebra_flush_reg (zh); - logf (LOG_LOG, "flush explain record, sysno=%d", rec->sysno); + zebra_register_deactivate (zh); - if (rec->size[recInfo_delKeys]) + zebra_get_state (&val, &seqno); + if (val != 'd') { - 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); + BFiles bfs = bfs_create (res_get (zh->service->res, "shadow")); + bf_commitClean (bfs, rval); + bfs_destroy (bfs); } - extract_flushRecordKeys (zh, rec->sysno, 1, &zh->keys); - extract_flushSortKeys (zh, rec->sysno, 1, &zh->sortKeys); + if (!rval) + seqno++; + zebra_set_state ('o', seqno); - xfree (rec->info[recInfo_delKeys]); - rec->size[recInfo_delKeys] = zh->keys.buf_used; - rec->info[recInfo_delKeys] = zh->keys.buf; - zh->keys.buf = NULL; - zh->keys.buf_max = 0; - return 0; + zebra_unlock (zh->lock_shadow); + zebra_unlock (zh->lock_normal); + +#if HAVE_SYS_TIMES_H + times (&zh->tms2); + logf (LOG_LOG, "user/system: %ld/%ld", + (long) (zh->tms2.tms_utime - zh->tms1.tms_utime), + (long) (zh->tms2.tms_stime - zh->tms1.tms_stime)); + +#endif } -static 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 zebra_repository_update (ZebraHandle zh) { - RecordAttr *recordAttr; - struct recExtractCtrl extractCtrl; - int i, r; - RecType recType; - char subType[1024]; - void *clientData; - const char *fname = ""; - 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; + zebra_begin_trans (zh); + logf (LOG_LOG, "updating %s", zh->rGroup.path); + repositoryUpdate (zh); + zebra_end_trans (zh); +} - /* announce database */ - if (zebraExplain_curDatabase (zh->service->zei, databaseName)) - { - if (zebraExplain_newDatabase (zh->service->zei, databaseName, 0)) - return 0; - } - if (!(recType = - recType_byName (zh->service->recTypes, recordType, subType, - &clientData))) - { - logf (LOG_WARN, "No such record type: %s", recordType); - return 0; - } +void zebra_repository_delete (ZebraHandle zh) +{ + logf (LOG_LOG, "deleting %s", zh->rGroup.path); + repositoryDelete (zh); +} - zh->keys.buf_used = 0; - zh->keys.prevAttrUse = -1; - zh->keys.prevAttrSet = -1; - zh->keys.prevSeqNo = 0; - zh->sortKeys = 0; - - extractCtrl.subType = subType; - extractCtrl.init = extract_init; - extractCtrl.tokenAdd = extract_token_add; - extractCtrl.schemaAdd = extract_schema_add; - extractCtrl.dh = zh->service->dh; - extractCtrl.handle = zh; - extractCtrl.zebra_maps = zh->service->zebra_maps; - extractCtrl.flagShowRecords = 0; - for (i = 0; i<256; i++) - { - if (zebra_maps_is_positioned(zh->service->zebra_maps, i)) - extractCtrl.seqno[i] = 1; - else - extractCtrl.seqno[i] = 0; - } +void zebra_repository_show (ZebraHandle zh) +{ + repositoryShow (zh); +} - r = (*recType->extract)(clientData, &extractCtrl); +void zebra_commit (ZebraHandle zh) +{ + int seqno; + char val; + const char *rval = res_get (zh->service->res, "shadow"); + BFiles bfs; - 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->keys.buf_used == 0) + if (!rval) { - /* 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; + logf (LOG_WARN, "Cannot perform commit"); + logf (LOG_WARN, "No shadow area defined"); + return; } - /* 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->service->records); + zebra_lock_w (zh->lock_normal); + zebra_lock_r (zh->lock_shadow); - *sysno = rec->sysno; + bfs = bfs_create (res_get (zh->service->res, "register")); - recordAttr = rec_init_attr (zh->service->zei, rec); + zebra_get_state (&val, &seqno); -#if 0 - if (matchStr) - { - dict_insert (matchDict, matchStr, sizeof(*sysno), sysno); - } -#endif - extract_flushRecordKeys (zh, *sysno, 1, &zh->keys); - extract_flushSortKeys (zh, *sysno, 1, &zh->sortKeys); - } - else + if (rval && *rval) + bf_cache (bfs, rval); + if (bf_commitExists (bfs)) { - /* record already exists */ - struct recKeys delkeys; - - rec = rec_get (zh->service->records, *sysno); - assert (rec); - - recordAttr = rec_init_attr (zh->service->zei, rec); + zebra_set_state ('c', seqno); - if (recordAttr->runNumber == - zebraExplain_runNumberIncrement (zh->service->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->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); + logf (LOG_LOG, "commit start"); + bf_commitExec (bfs); +#ifndef WIN32 + sync (); #endif - rec_del (zh->service->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->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->keys.buf_used > 0 && store_keys == 1) - { - rec->size[recInfo_delKeys] = zh->keys.buf_used; - rec->info[recInfo_delKeys] = zh->keys.buf; - zh->keys.buf = NULL; - zh->keys.buf_max = 0; + logf (LOG_LOG, "commit clean"); + bf_commitClean (bfs, rval); + seqno++; + zebra_set_state ('o', seqno); } else { - rec->info[recInfo_delKeys] = NULL; - rec->size[recInfo_delKeys] = 0; + logf (LOG_LOG, "nothing to commit"); } + bfs_destroy (bfs); - /* save file size of original record */ - zebraExplain_recordBytesIncrement (zh->service->zei, - - recordAttr->recordSize); -#if 0 - recordAttr->recordSize = fi->file_moffset - recordOffset; - if (!recordAttr->recordSize) - recordAttr->recordSize = fi->file_max - recordOffset; -#else - recordAttr->recordSize = buf_size; -#endif - zebraExplain_recordBytesIncrement (zh->service->zei, - recordAttr->recordSize); - - /* set run-number for this record */ - recordAttr->runNumber = - zebraExplain_runNumberIncrement (zh->service->zei, 0); + zebra_unlock (zh->lock_shadow); + zebra_unlock (zh->lock_normal); +} - /* update store data */ - xfree (rec->info[recInfo_storeData]); - if (store_data == 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 - 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 - { - rec->info[recInfo_storeData] = NULL; - rec->size[recInfo_storeData] = 0; - } - /* update database name */ - xfree (rec->info[recInfo_databaseName]); - rec->info[recInfo_databaseName] = - rec_strdup (databaseName, &rec->size[recInfo_databaseName]); +void zebra_init (ZebraHandle zh) +{ + const char *rval = res_get (zh->service->res, "shadow"); + BFiles bfs = 0; - /* update offset */ - recordAttr->recordOffset = recordOffset; + bfs = bfs_create (res_get (zh->service->res, "register")); + if (rval && *rval) + bf_cache (bfs, rval); - /* commit this record */ - rec_put (zh->service->records, &rec); + bf_reset (bfs); + bfs_destroy (bfs); + zebra_set_state ('o', 0); +} - return 0; +void zebra_compact (ZebraHandle zh) +{ + BFiles bfs = bfs_create (res_get (zh->service->res, "register")); + inv_compact (bfs); + bfs_destroy (bfs); +} + +int zebra_record_insert (ZebraHandle zh, const char *buf, int len) +{ + int sysno = 0; + zebra_begin_trans (zh); + extract_rec_in_mem (zh, "grs.sgml", + buf, len, + "Default", /* database */ + 0 /* delete_flag */, + 0 /* test_mode */, + &sysno /* sysno */, + 1 /* store_keys */, + 1 /* store_data */, + 0 /* match criteria */); + zebra_end_trans (zh); + return sysno; } diff --git a/index/zebraapi.h b/index/zebraapi.h index 2935fef..8fe0dad 100644 --- a/index/zebraapi.h +++ b/index/zebraapi.h @@ -1,49 +1,14 @@ /* - * Copyright (C) 1994-1998, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * - * $Log: zebraapi.h,v $ - * Revision 1.12 2000-04-05 09:49:35 adam - * On Unix, zebra/z'mbol uses automake. - * - * Revision 1.11 2000/03/20 19:08:36 adam - * Added remote record import using Z39.50 extended services and Segment - * Requests. - * - * Revision 1.10 2000/03/15 15:00:31 adam - * First work on threaded version. - * - * Revision 1.9 2000/02/24 12:31:17 adam - * Added zebra_string_norm. - * - * Revision 1.8 1999/11/30 13:48:03 adam - * Improved installation. Updated for inclusion of YAZ header files. - * - * Revision 1.7 1999/11/04 15:00:45 adam - * Implemented delete result set(s). - * - * Revision 1.6 1999/02/17 11:29:57 adam - * Fixed in record_fetch. Minor updates to API. - * - * Revision 1.5 1998/09/22 10:48:19 adam - * Minor changes in search API. - * - * Revision 1.4 1998/09/02 13:53:18 adam - * Extra parameter decode added to search routines to implement - * persistent queries. - * - * Revision 1.3 1998/06/22 11:36:48 adam - * Added authentication check facility to zebra. - * - * Revision 1.2 1998/06/13 00:14:09 adam - * Minor changes. - * - * Revision 1.1 1998/06/12 12:22:13 adam - * Work on Zebra API. - * + * $Id: zebraapi.h,v 1.13 2002-02-20 17:30:01 adam Exp $ */ +#ifndef ZEBRAAPI_H +#define ZEBRAAPI_H + #include #include #include @@ -137,4 +102,18 @@ YAZ_EXPORT void zebra_admin_import_segment (ZebraHandle zh, Z_Segment *segment); void zebra_admin_import_end (ZebraHandle zh); + +void zebra_begin_trans (ZebraHandle zh); +void zebra_end_trans (ZebraHandle zh); + +void zebra_commit (ZebraHandle zh); + +void zebra_init (ZebraHandle zh); +void zebra_compact (ZebraHandle zh); +void zebra_repository_update (ZebraHandle zh); +void zebra_repository_delete (ZebraHandle zh); +void zebra_repository_show (ZebraHandle zh); +int zebra_record_insert (ZebraHandle zh, const char *buf, int len); + YAZ_END_CDECL +#endif diff --git a/index/zinfo.c b/index/zinfo.c index 0565a00..d91b11e 100644 --- a/index/zinfo.c +++ b/index/zinfo.c @@ -1,85 +1,9 @@ /* - * Copyright (C) 1994-2000, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * - * $Log: zinfo.c,v $ - * Revision 1.22 2001-10-15 19:53:43 adam - * POSIX thread updates. First work on term sets. - * - * Revision 1.21 2000/12/05 10:01:44 adam - * Fixed bug regarding user-defined attribute sets. - * - * Revision 1.20 2000/11/29 14:24:01 adam - * Script configure uses yaz pthreads options. Added locking for - * zebra_register_{lock,unlock}. - * - * Revision 1.19 2000/07/07 12:49:20 adam - * Optimized resultSetInsert{Rank,Sort}. - * - * Revision 1.18 2000/03/20 19:08:36 adam - * Added remote record import using Z39.50 extended services and Segment - * Requests. - * - * Revision 1.17 1999/07/14 10:53:51 adam - * Updated various routines to handle missing explain schema. - * - * Revision 1.16 1999/05/26 07:49:13 adam - * C++ compilation. - * - * Revision 1.15 1999/01/25 13:47:54 adam - * Fixed bug. - * - * Revision 1.14 1998/11/04 16:31:32 adam - * Fixed bug regarding recordBytes in databaseInfo. - * - * Revision 1.13 1998/11/03 10:17:09 adam - * Fixed bug regarding creation of some data1 nodes for Explain records. - * - * Revision 1.12 1998/10/13 20:37:11 adam - * Changed the way attribute sets are saved in Explain database to - * reflect "dynamic" OIDs. - * - * Revision 1.11 1998/06/09 12:16:48 adam - * Implemented auto-generation of CategoryList records. - * - * Revision 1.10 1998/06/08 14:43:15 adam - * Added suport for EXPLAIN Proxy servers - added settings databasePath - * and explainDatabase to facilitate this. Increased maximum number - * of databases and attributes in one register. - * - * Revision 1.9 1998/06/02 12:10:27 adam - * Fixed bug related to attributeDetails. - * - * Revision 1.8 1998/05/20 10:12:20 adam - * Implemented automatic EXPLAIN database maintenance. - * Modified Zebra to work with ASN.1 compiled version of YAZ. - * - * Revision 1.7 1998/03/05 08:45:13 adam - * New result set model and modular ranking system. Moved towards - * descent server API. System information stored as "SGML" records. - * - * Revision 1.6 1998/02/17 10:29:27 adam - * Moved towards 'automatic' EXPLAIN database. - * - * Revision 1.5 1997/10/27 14:33:05 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 1.4 1997/09/25 14:57:08 adam - * Added string.h. - * - * Revision 1.3 1996/05/22 08:21:59 adam - * Added public ZebDatabaseInfo structure. - * - * Revision 1.2 1996/05/14 06:16:41 adam - * Compact use/set bytes used in search service. - * - * Revision 1.1 1996/05/13 14:23:07 adam - * Work on compaction of set/use bytes in dictionary. - * + * $Id: zinfo.c,v 1.23 2002-02-20 17:30:01 adam Exp $ */ #include diff --git a/index/zinfo.h b/index/zinfo.h index c17271e..4195dd1 100644 --- a/index/zinfo.h +++ b/index/zinfo.h @@ -4,7 +4,10 @@ * Sebastian Hammer, Adam Dickmeiss * * $Log: zinfo.h,v $ - * Revision 1.11 2001-10-15 19:53:43 adam + * Revision 1.12 2002-02-20 17:30:01 adam + * Work on new API. Locking system re-implemented + * + * Revision 1.11 2001/10/15 19:53:43 adam * POSIX thread updates. First work on term sets. * * Revision 1.10 2000/05/15 12:56:37 adam @@ -47,6 +50,7 @@ #define ZINFO_H #include +#include "zebraapi.h" #include "recindex.h" YAZ_BEGIN_CDECL diff --git a/index/zserver.c b/index/zserver.c index c2642c7..1df6afd 100644 --- a/index/zserver.c +++ b/index/zserver.c @@ -2,288 +2,7 @@ * Copyright (C) 1995-2000, Index Data * All rights reserved. * - * $Log: zserver.c,v $ - * Revision 1.82 2001-07-09 23:40:09 adam - * WIN32 fix. - * - * Revision 1.81 2001/02/14 21:42:07 adam - * Fixed versions as returned in INIT response. - * - * Revision 1.80 2000/09/05 14:04:05 adam - * Updates for prefix 'yaz_' for YAZ log functions. - * - * Revision 1.79 2000/05/09 10:56:50 adam - * Added call to xmalloc/nmem debugging functions. - * - * Revision 1.78 2000/04/05 09:49:35 adam - * On Unix, zebra/z'mbol uses automake. - * - * Revision 1.77 2000/03/20 19:08:36 adam - * Added remote record import using Z39.50 extended services and Segment - * Requests. - * - * Revision 1.76 2000/03/15 15:00:31 adam - * First work on threaded version. - * - * Revision 1.75 1999/11/30 13:48:04 adam - * Improved installation. Updated for inclusion of YAZ header files. - * - * Revision 1.74 1999/11/29 15:13:26 adam - * Server sets implementationName - and Version. - * - * Revision 1.73 1999/11/04 15:00:45 adam - * Implemented delete result set(s). - * - * Revision 1.71 1999/07/14 10:59:26 adam - * Changed functions isc_getmethod, isams_getmethod. - * Improved fatal error handling (such as missing EXPLAIN schema). - * - * Revision 1.70 1999/06/10 12:14:56 adam - * Fixed to use bend_start instead of pre_init. - * - * Revision 1.69 1999/06/10 09:20:03 adam - * Minor change to pre_init handler. - * - * Revision 1.68 1999/05/26 07:49:13 adam - * C++ compilation. - * - * Revision 1.67 1999/02/02 14:51:14 adam - * Updated WIN32 code specific sections. Changed header. - * - * Revision 1.66 1998/10/28 10:54:41 adam - * SDRKit integration. - * - * Revision 1.65 1998/10/18 07:54:54 adam - * Additional info added for diagnostics 114 (Unsupported use attribute) and - * 121 (Unsupported attribute set). - * - * Revision 1.64 1998/09/22 10:48:21 adam - * Minor changes in search API. - * - * Revision 1.63 1998/09/02 13:53:21 adam - * Extra parameter decode added to search routines to implement - * persistent queries. - * - * Revision 1.62 1998/08/06 14:35:28 adam - * Routine bend_deleterequest removed. - * - * Revision 1.61 1998/06/24 12:16:15 adam - * Support for relations on text operands. Open range support in - * DFA module (i.e. [-j], [g-]). - * - * Revision 1.60 1998/06/22 11:36:49 adam - * Added authentication check facility to zebra. - * - * Revision 1.59 1998/06/12 12:22:13 adam - * Work on Zebra API. - * - * Revision 1.58 1998/05/27 16:57:46 adam - * Zebra returns surrogate diagnostic for single records when - * appropriate. - * - * Revision 1.57 1998/04/03 14:45:18 adam - * Fixed setting of last_in_set in bend_fetch. - * - * Revision 1.56 1998/03/05 08:45:13 adam - * New result set model and modular ranking system. Moved towards - * descent server API. System information stored as "SGML" records. - * - * Revision 1.55 1998/02/10 12:03:06 adam - * Implemented Sort. - * - * Revision 1.54 1998/01/29 13:39:13 adam - * Compress ISAM is default. - * - * Revision 1.53 1998/01/12 15:04:09 adam - * The test option (-s) only uses read-lock (and not write lock). - * - * Revision 1.52 1997/11/18 10:05:08 adam - * Changed character map facility so that admin can specify character - * mapping files for each register type, w, p, etc. - * - * Revision 1.51 1997/10/27 14:33:06 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 1.50 1997/09/29 09:08:36 adam - * Revised locking system to be thread safe for the server. - * - * Revision 1.49 1997/09/25 14:57:23 adam - * Windows NT port. - * - * Revision 1.48 1997/09/17 12:19:19 adam - * Zebra version corresponds to YAZ version 1.4. - * Changed Zebra server so that it doesn't depend on global common_resource. - * - * Revision 1.47 1997/09/04 13:58:36 adam - * New retrieve/extract method tellf (added). - * Added O_BINARY for open calls. - * - * Revision 1.46 1997/07/28 08:30:47 adam - * Server returns diagnostic 14 when record doesn't exist. - * - * Revision 1.45 1996/12/23 15:30:45 adam - * Work on truncation. - * Bug fix: result sets weren't deleted after server shut down. - * - * Revision 1.44 1996/12/11 12:08:01 adam - * Added better compression. - * - * Revision 1.43 1996/11/15 15:03:58 adam - * Logging of execution speed by using the times(2) call. - * - * Revision 1.42 1996/11/08 11:10:36 adam - * Buffers used during file match got bigger. - * Compressed ISAM support everywhere. - * Bug fixes regarding masking characters in queries. - * Redesigned Regexp-2 queries. - * - * Revision 1.41 1996/10/29 14:09:56 adam - * Use of cisam system - enabled if setting isamc is 1. - * - * Revision 1.40 1996/06/04 10:19:02 adam - * Minor changes - removed include of ctype.h. - * - * Revision 1.39 1996/05/31 09:07:05 quinn - * Work on character-set handling - * - * Revision 1.38 1996/05/14 11:34:01 adam - * Scan support in multiple registers/databases. - * - * Revision 1.37 1996/05/14 06:16:48 adam - * Compact use/set bytes used in search service. - * - * Revision 1.36 1996/05/01 13:46:37 adam - * First work on multiple records in one file. - * New option, -offset, to the "unread" command in the filter module. - * - * Revision 1.35 1996/03/26 16:01:14 adam - * New setting lockPath: directory of various lock files. - * - * Revision 1.34 1996/03/20 09:36:46 adam - * Function dict_lookup_grep got extra parameter, init_pos, which marks - * from which position in pattern approximate pattern matching should occur. - * Approximate pattern matching is used in relevance=re-2. - * - * Revision 1.33 1996/01/17 14:57:56 adam - * Prototype changed for reader functions in extract/retrieve. File - * is identified by 'void *' instead of 'int. - * - * Revision 1.32 1995/12/11 09:12:58 adam - * The rec_get function returns NULL if record doesn't exist - will - * happen in the server if the result set records have been deleted since - * the creation of the set (i.e. the search). - * The server saves a result temporarily if it is 'volatile', i.e. the - * set is register dependent. - * - * Revision 1.31 1995/12/08 16:22:56 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 1.30 1995/12/07 17:38:48 adam - * Work locking mechanisms for concurrent updates/commit. - * - * Revision 1.29 1995/12/04 14:22:32 adam - * Extra arg to recType_byName. - * Started work on new regular expression parsed input to - * structured records. - * - * Revision 1.28 1995/11/28 09:09:48 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 1.27 1995/11/27 13:58:54 adam - * New option -t. storeStore data implemented in server. - * - * Revision 1.26 1995/11/25 10:24:07 adam - * More record fields - they are enumerated now. - * New options: flagStoreData flagStoreKey. - * - * Revision 1.25 1995/11/21 15:29:13 adam - * Config file 'base' read by default by both indexer and server. - * - * Revision 1.24 1995/11/20 16:59:47 adam - * New update method: the 'old' keys are saved for each records. - * - * Revision 1.23 1995/11/16 17:00:56 adam - * Better logging of rpn query. - * - * Revision 1.22 1995/11/16 15:34:55 adam - * Uses new record management system in both indexer and server. - * - * Revision 1.21 1995/11/01 16:25:52 quinn - * *** empty log message *** - * - * Revision 1.20 1995/10/27 14:00:12 adam - * Implemented detection of database availability. - * - * Revision 1.19 1995/10/17 18:02:11 adam - * New feature: databases. Implemented as prefix to words in dictionary. - * - * Revision 1.18 1995/10/16 14:03:09 quinn - * Changes to support element set names and espec1 - * - * Revision 1.17 1995/10/16 09:32:40 adam - * More work on relational op. - * - * Revision 1.16 1995/10/13 12:26:44 adam - * Optimization of truncation. - * - * Revision 1.15 1995/10/12 12:40:55 adam - * Bug fixes in rpn_prox. - * - * Revision 1.14 1995/10/09 16:18:37 adam - * Function dict_lookup_grep got extra client data parameter. - * - * Revision 1.13 1995/10/06 14:38:00 adam - * New result set method: r_score. - * Local no (sysno) and score is transferred to retrieveCtrl. - * - * Revision 1.12 1995/10/06 13:52:06 adam - * Bug fixes. Handler may abort further scanning. - * - * Revision 1.11 1995/10/06 10:43:57 adam - * Scan added. 'occurrences' in scan entries not set yet. - * - * Revision 1.10 1995/10/02 16:43:32 quinn - * Set default resulting record type in fetch. - * - * Revision 1.9 1995/10/02 15:18:52 adam - * New member in recRetrieveCtrl: diagnostic. - * - * Revision 1.8 1995/09/28 09:19:47 adam - * xfree/xmalloc used everywhere. - * Extract/retrieve method seems to work for text records. - * - * Revision 1.7 1995/09/27 16:17:32 adam - * More work on retrieve. - * - * Revision 1.6 1995/09/08 08:53:22 adam - * Record buffer maintained in server_info. - * - * Revision 1.5 1995/09/06 16:11:18 adam - * Option: only one word key per file. - * - * Revision 1.4 1995/09/06 10:33:04 adam - * More work on present. Some log messages removed. - * - * Revision 1.3 1995/09/05 15:28:40 adam - * More work on search engine. - * - * Revision 1.2 1995/09/04 12:33:43 adam - * Various cleanup. YAZ util used instead. - * - * Revision 1.1 1995/09/04 09:10:41 adam - * More work on index add/del/update. - * Merge sort implemented. - * Initial work on z39 server. - * + * $Id: zserver.c,v 1.83 2002-02-20 17:30:01 adam Exp $ */ #include diff --git a/index/zserver.h b/index/zserver.h index d0c936b..77c7db6 100644 --- a/index/zserver.h +++ b/index/zserver.h @@ -1,9 +1,9 @@ /* - * Copyright (C) 1994-2001, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * - * $Id: zserver.h,v 1.51 2001-10-15 19:53:43 adam Exp $ + * $Id: zserver.h,v 1.52 2002-02-20 17:30:01 adam Exp $ */ #if HAVE_SYS_TIMES_H @@ -47,6 +47,7 @@ struct zebra_service { ISAMD isamd; #endif Dict dict; + Dict matchDict; SortIdx sortIdx; int registerState; /* 0 (no commit pages), 1 (use commit pages) */ time_t registerChange; @@ -56,6 +57,7 @@ struct zebra_service { Res res; ZebraLockHandle server_lock_cmt; ZebraLockHandle server_lock_org; + char *server_path_prefix; data1_handle dh; ZebraMaps zebra_maps; @@ -98,6 +100,12 @@ struct zebra_session { int key_file_no; char *admin_databaseName; + ZebraLockHandle lock_normal; + ZebraLockHandle lock_shadow; + + int trans_no; + int seqno; + int last_val; int destroyed; ZebraSet sets; int errCode; @@ -107,6 +115,7 @@ struct zebra_session { struct tms tms1; struct tms tms2; #endif + struct recordGroup rGroup; }; struct rank_control { @@ -167,6 +176,7 @@ 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); void zebra_sort (ZebraHandle zh, ODR stream, int num_input_setnames, const char **input_setnames, @@ -202,6 +212,15 @@ 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; @@ -219,4 +238,12 @@ 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); + YAZ_END_CDECL diff --git a/index/zsets.c b/index/zsets.c index dfb73e7..2fed59c 100644 --- a/index/zsets.c +++ b/index/zsets.c @@ -4,7 +4,10 @@ * Sebastian Hammer, Adam Dickmeiss * * $Log: zsets.c,v $ - * Revision 1.31 2001-11-19 23:05:22 adam + * Revision 1.32 2002-02-20 17:30:01 adam + * Work on new API. Locking system re-implemented + * + * Revision 1.31 2001/11/19 23:05:22 adam * Added a few prototypes. * * Revision 1.30 2001/10/15 19:53:43 adam @@ -275,6 +278,7 @@ ZebraSet resultSetGet (ZebraHandle zh, const char *name) if (!s->term_entries && !s->rset && s->rpn) { NMEM nmem = nmem_create (); + yaz_log (LOG_LOG, "research %s", name); s->rset = rpn_search (zh, nmem, s->rpn, s->num_bases, s->basenames, s->name, s); @@ -285,9 +289,21 @@ ZebraSet resultSetGet (ZebraHandle zh, const char *name) return NULL; } +void resultSetInvalidate (ZebraHandle zh) +{ + ZebraSet s = zh->sets; + + for (; s; s = s->next) + { + if (s->rset) + rset_delete (s->rset); + s->rset = 0; + } +} + void resultSetDestroy (ZebraHandle zh, int num, char **names,int *statuses) { - ZebraSet *ss = &zh->sets; + ZebraSet * ss = &zh->sets; int i; if (statuses) diff --git a/recctrl/recgrs.c b/recctrl/recgrs.c index d23a3d7..57b144b 100644 --- a/recctrl/recgrs.c +++ b/recctrl/recgrs.c @@ -3,7 +3,10 @@ * All rights reserved. * * $Log: recgrs.c,v $ - * Revision 1.41 2001-05-22 21:01:47 adam + * Revision 1.42 2002-02-20 17:30:01 adam + * Work on new API. Locking system re-implemented + * + * Revision 1.41 2001/05/22 21:01:47 adam * Removed print of data1 tree on stdout so that inetd works again. * * Revision 1.40 2001/03/29 21:31:31 adam @@ -511,6 +514,9 @@ static int grs_extract_sub(struct grs_handlers *h, struct recExtractCtrl *p, if ((oid_ent_to_oid (&oe, oidtmp))) (*p->schemaAdd)(p, oidtmp); +#if 0 + data1_pr_tree (p->dh, n, stdout); +#endif if (dumpkeys(n, p, 0) < 0) { data1_free_tree(p->dh, n); diff --git a/rset/rstemp.c b/rset/rstemp.c index fe30044..0ab65fc 100644 --- a/rset/rstemp.c +++ b/rset/rstemp.c @@ -1,98 +1,9 @@ /* - * Copyright (C) 1994-1999, Index Data + * Copyright (C) 1994-2002, Index Data * All rights reserved. * Sebastian Hammer, Adam Dickmeiss * - * $Log: rstemp.c,v $ - * Revision 1.26 1999-05-26 07:49:14 adam - * C++ compilation. - * - * Revision 1.25 1999/02/02 14:51:37 adam - * Updated WIN32 code specific sections. Changed header. - * - * Revision 1.24 1998/03/05 08:36:28 adam - * New result set model. - * - * Revision 1.23 1997/12/18 10:54:25 adam - * New method result set method rs_hits that returns the number of - * hits in result-set (if known). The ranked result set returns real - * number of hits but only when not combined with other operands. - * - * Revision 1.22 1997/10/31 12:38:12 adam - * Bug fix: added missing xfree() call. - * - * Revision 1.21 1997/09/17 12:19:23 adam - * Zebra version corresponds to YAZ version 1.4. - * Changed Zebra server so that it doesn't depend on global common_resource. - * - * Revision 1.20 1997/09/09 13:38:17 adam - * Partial port to WIN95/NT. - * - * Revision 1.19 1997/09/04 13:58:57 adam - * Added O_BINARY for open calls. - * - * Revision 1.18 1996/10/29 13:54:52 adam - * Changed name of setting tempSetDir to setTmpDir. - * - * Revision 1.17 1995/12/11 09:15:28 adam - * New set types: sand/sor/snot - ranked versions of and/or/not in - * ranked/semi-ranked result sets. - * Note: the snot not finished yet. - * New rset member: flag. - * Bug fix: r_delete in rsrel.c did free bad memory block. - * - * Revision 1.16 1995/11/28 14:47:02 adam - * New setting: tempSetPath. Location of temporary result sets. - * - * Revision 1.15 1995/10/12 12:41:58 adam - * Private info (buf) moved from struct rset_control to struct rset. - * Bug fixes in relevance. - * - * Revision 1.14 1995/10/10 14:00:04 adam - * Function rset_open changed its wflag parameter to general flags. - * - * Revision 1.13 1995/10/06 14:38:06 adam - * New result set method: r_score. - * Local no (sysno) and score is transferred to retrieveCtrl. - * - * Revision 1.12 1995/09/28 09:52:11 adam - * xfree/xmalloc used everywhere. - * - * Revision 1.11 1995/09/18 14:17:56 adam - * Bug fixes. - * - * Revision 1.10 1995/09/15 14:45:39 adam - * Bug fixes. - * - * Revision 1.9 1995/09/15 09:20:42 adam - * Bug fixes. - * - * Revision 1.8 1995/09/08 14:52:42 adam - * Work on relevance feedback. - * - * Revision 1.7 1995/09/07 13:58:44 adam - * New parameter: result-set file descriptor (RSFD) to support multiple - * positions within the same result-set. - * Boolean operators: and, or, not implemented. - * - * Revision 1.6 1995/09/06 16:11:56 adam - * More work on boolean sets. - * - * Revision 1.5 1995/09/05 16:36:59 adam - * Minor changes. - * - * Revision 1.4 1995/09/05 11:43:24 adam - * Complete version of temporary sets. Not tested yet though. - * - * Revision 1.3 1995/09/04 15:20:40 adam - * More work on temp sets. is_open member removed. - * - * Revision 1.2 1995/09/04 09:10:56 adam - * Minor changes. - * - * Revision 1.1 1994/11/04 13:21:30 quinn - * Working. - * + * $Id: rstemp.c,v 1.27 2002-02-20 17:30:01 adam Exp $ */ #include @@ -215,8 +126,22 @@ static void r_flush (RSFD rfd, int mk) if (!info->fname && mk) { - char *s = (char*) tempnam (info->temp_path, "zrs"); +#if 1 + char template[1024]; + sprintf (template, "%s/zrsXXXXXX", info->temp_path); + + info->fd = mkstemp (template); + + if (info->fd == -1) + { + logf (LOG_FATAL|LOG_ERRNO, "mkstemp %s", template); + exit (1); + } + info->fname = (char *) xmalloc (strlen(template)+1); + strcpy (info->fname, template); +#else + char *s = (char*) tempnam (info->temp_path, "zrs"); info->fname = (char *) xmalloc (strlen(s)+1); strcpy (info->fname, s); @@ -227,6 +152,7 @@ static void r_flush (RSFD rfd, int mk) logf (LOG_FATAL|LOG_ERRNO, "open %s", info->fname); exit (1); } +#endif } if (info->fname && info->fd != -1 && info->dirty) { diff --git a/test/Makefile.am b/test/Makefile.am index 5a75551..2a440d3 100644 --- a/test/Makefile.am +++ b/test/Makefile.am @@ -1,2 +1,2 @@ -SUBDIRS=gils usmarc +SUBDIRS=gils usmarc api diff --git a/test/api/Makefile.am b/test/api/Makefile.am new file mode 100644 index 0000000..9d32563 --- /dev/null +++ b/test/api/Makefile.am @@ -0,0 +1,9 @@ + +noinst_PROGRAMS = t1 t2 + +t1_SOURCES = t1.c +t2_SOURCES = t2.c + +CFLAGS = -I$(top_srcdir)/index $(YAZINC) + +LDADD = ../../index/libzebra.a $(YAZLIB) $(TCL_LIB) diff --git a/test/api/t1.c b/test/api/t1.c new file mode 100644 index 0000000..c276d4d --- /dev/null +++ b/test/api/t1.c @@ -0,0 +1,21 @@ +/* + * $Id: t1.c,v 1.1 2002-02-20 17:30:02 adam Exp $ + */ + +#include + +int main(int argc, char **argv) +{ + ZebraService zs; + ZebraHandle zh; + + nmem_init(); + zs = zebra_start("t1.cfg"); + zh = zebra_open (zs); + + zebra_close (zh); + zebra_stop (zs); + nmem_exit (); + xmalloc_trav ("x"); + exit (0); +} diff --git a/test/api/t1.cfg b/test/api/t1.cfg new file mode 100644 index 0000000..815497d --- /dev/null +++ b/test/api/t1.cfg @@ -0,0 +1,2 @@ +# $Id: t1.cfg,v 1.1 2002-02-20 17:30:02 adam Exp $ +profilepath: ../../../yaz/tab \ No newline at end of file diff --git a/test/api/t2.c b/test/api/t2.c new file mode 100644 index 0000000..dfe8fef --- /dev/null +++ b/test/api/t2.c @@ -0,0 +1,35 @@ +/* + * $Id: t2.c,v 1.1 2002-02-20 17:30:02 adam Exp $ + */ + +#include + +int main(int argc, char **argv) +{ + ZebraService zs; + ZebraHandle zh; + const char *myrec = + "\n" + " My title\n" + "\n"; + ODR odr_input, odr_output; + + nmem_init (); + + odr_input = odr_createmem (ODR_DECODE); + odr_output = odr_createmem (ODR_ENCODE); + + zs = zebra_start("t2.cfg"); + zh = zebra_open (zs); + + zebra_begin_trans (zh); + zebra_record_insert (zh, myrec, strlen(myrec)); + zebra_end_trans (zh); + zebra_commit (zh); + zebra_close (zh); + zebra_stop (zs); + + nmem_exit (); + xmalloc_trav ("x"); + exit (0); +} diff --git a/test/api/t2.cfg b/test/api/t2.cfg new file mode 100644 index 0000000..be03f44 --- /dev/null +++ b/test/api/t2.cfg @@ -0,0 +1,4 @@ +# $Id: t2.cfg,v 1.1 2002-02-20 17:30:02 adam Exp $ +profilepath: ../../../yaz/tab:../../tab + +attset: bib1.att diff --git a/test/gils/Makefile.am b/test/gils/Makefile.am index 56c9752..5cc7812 100644 --- a/test/gils/Makefile.am +++ b/test/gils/Makefile.am @@ -4,3 +4,7 @@ EXTRA_DIST = zebra.cfg test.sh dist-hook: -mkdir $(distdir)/records cp $(srcdir)/records/*.grs $(distdir)/records + +test: + ../../index/zmbolidx init + ../../index/zmbolidx update records diff --git a/zebra.spec b/zebra.spec index 1666b8f..8428878 100644 --- a/zebra.spec +++ b/zebra.spec @@ -1,5 +1,5 @@ Name: zebra -Version: 1.1 +Version: 1.1.1 Release: 1 Requires: yaz Copyright: Distributable diff --git a/zmbol.spec b/zmbol.spec index ec3ae6d..61bd9ec 100644 --- a/zmbol.spec +++ b/zmbol.spec @@ -1,5 +1,5 @@ Name: zmbol -Version: 1.1 +Version: 1.1.1 Release: 1 Requires: yaz Copyright: Commercial -- 1.7.10.4