X-Git-Url: http://git.indexdata.com/?a=blobdiff_plain;f=index%2Finvstat.c;h=69973eb0bb65e24426a89f48a884eb77f11af639;hb=47ea1fc957c7b97bb30a26698f072109cae275e4;hp=afd0147830db19877c1abef9361d746c5cce6536;hpb=801ef4b322574c793abdc1ba81dea2da62ca60bc;p=idzebra-moved-to-github.git diff --git a/index/invstat.c b/index/invstat.c index afd0147..69973eb 100644 --- a/index/invstat.c +++ b/index/invstat.c @@ -1,35 +1,8 @@ /* - * Copyright (C) 1994-1998, Index Data I/S + * Copyright (C) 1994-1999, Index Data * All rights reserved. - * Sebastian Hammer, Adam Dickmeiss - * - * $Log: invstat.c,v $ - * Revision 1.7 1998-03-13 15:30:50 adam - * New functions isc_block_used and isc_block_size. Fixed 'leak' - * in isc_alloc_block. - * - * Revision 1.6 1998/03/06 13:54:02 adam - * Fixed two nasty bugs in isc_merge. - * - * Revision 1.5 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 1.4 1996/11/08 11:10:21 adam - * Buffers used during file match got bigger. - * Compressed ISAM support everywhere. - * Bug fixes regarding masking characters in queries. - * Redesigned Regexp-2 queries. - * - * Revision 1.3 1996/06/04 10:18:58 adam - * Minor changes - removed include of ctype.h. - * - * Revision 1.2 1996/05/22 08:25:56 adam - * Minor change. - * - * Revision 1.1 1996/05/14 14:04:34 adam - * In zebraidx, the 'stat' command is improved. Statistics about ISAM/DICT - * is collected. + * Sebastian Hammer, Adam Dickmeiss, Heikki Levanto + * log at eof * */ #include @@ -37,110 +10,241 @@ #include #include "index.h" -#include "recindex.h" +#include "../isamc/isamd-p.h" struct inv_stat_info { - ISAM isam; - ISAMC isamc; - int no_isam_entries[8]; + ZebraHandle zh; + int no_isam_entries[9]; int no_dict_entries; int no_dict_bytes; int isam_bounds[20]; int isam_occurrences[20]; char tmp[128]; + int isamb_levels[10][5]; + int isamb_sizes[10]; + int isamb_blocks[10]; + unsigned long cksum; + int dumpwords; }; +#define SINGLETON_TYPE 8 /* the type to use for singletons that */ + /* have no block and no block type */ + +static void print_dict_item (ZebraMaps zm, const char *s, int count, + int firstsys, int firstseq, int lastsys, int lastseq ) +{ + int reg_type = s[1]; + char keybuf[IT_MAX_WORD+1]; + char *to = keybuf; + const char *from = s + 2; + + while (*from) + { + const char *res = zebra_maps_output (zm, reg_type, &from); + if (!res) + *to++ = *from++; + else + while (*res) + *to++ = *res++; + } + *to = '\0'; + /* yaz_log (LOG_LOG, "%s", keybuf); */ + printf("%10d %s %d.%d - %d.%d\n",count, keybuf, + firstsys,firstseq, lastsys,lastseq); +} + static int inv_stat_handle (char *name, const char *info, int pos, void *client) { int occur = 0; int i = 0; struct inv_stat_info *stat_info = (struct inv_stat_info*) client; - ISAM_P isam_p; + ISAMS_P isam_p; + int firstsys=-1; + int firstseq=-1; + int lastsys=-1; + int lastseq=-1; stat_info->no_dict_entries++; stat_info->no_dict_bytes += strlen(name); - assert (*info == sizeof(ISAM_P)); - memcpy (&isam_p, info+1, sizeof(ISAM_P)); + if (!stat_info->zh->reg->isamd) + { + assert (*info == sizeof(ISAMS_P)); + memcpy (&isam_p, info+1, sizeof(ISAMS_P)); + } - if (stat_info->isam) + if (stat_info->zh->reg->isams) + { + ISAMS_PP pp; + int occurx = 0; + struct it_key key; + + pp = isams_pp_open (stat_info->zh->reg->isams, isam_p); + occur = isams_pp_num (pp); + while (isams_pp_read(pp, &key)) + { + stat_info->cksum = stat_info->cksum * 65509 + + key.sysno + 11 * key.seqno; + occurx++; + if (-1==firstsys) + { + firstseq=key.seqno; + firstsys=key.sysno; + } + lastsys=key.sysno; + lastseq=key.seqno; + } + assert (occurx == occur); + stat_info->no_isam_entries[0] += occur; + isams_pp_close (pp); + } + if (stat_info->zh->reg->isam) { ISPT ispt; - ispt = is_position (stat_info->isam, isam_p); + ispt = is_position (stat_info->zh->reg->isam, isam_p); occur = is_numkeys (ispt); + stat_info->no_isam_entries[is_type(isam_p)] += occur; is_pt_free (ispt); } - if (stat_info->isamc) + if (stat_info->zh->reg->isamc) { ISAMC_PP pp; int occurx = 0; - char buf[128]; + struct it_key key; - pp = isc_pp_open (stat_info->isamc, isam_p); + pp = isc_pp_open (stat_info->zh->reg->isamc, isam_p); occur = isc_pp_num (pp); - while (isc_pp_read(pp, buf)) + while (isc_pp_read(pp, &key)) + { + stat_info->cksum = stat_info->cksum * 65509 + + key.sysno + 11 * key.seqno; occurx++; + if (-1==firstsys) + { + firstseq=key.seqno; + firstsys=key.sysno; + } + lastsys=key.sysno; + lastseq=key.seqno; + } assert (occurx == occur); stat_info->no_isam_entries[isc_type(isam_p)] += occur; isc_pp_close (pp); } + if (stat_info->zh->reg->isamd) + { + ISAMD_PP pp; + int occurx = 0; + struct it_key key; + /* printf("[%d: %d %d %d %d %d %d] ", */ + /* info[0], info[1], info[2], info[3], info[4], info[5], info[7]);*/ + pp = isamd_pp_open (stat_info->zh->reg->isamd, info+1, info[0]); + + occur = isamd_pp_num (pp); + while (isamd_pp_read(pp, &key)) + { + stat_info->cksum = stat_info->cksum * 65509 + + key.sysno + 11 * key.seqno; + occurx++; + /* printf("%d.%d ", key.sysno, key.seqno); */ /*!*/ + if (-1==firstsys) + { + firstseq=key.seqno; + firstsys=key.sysno; + } + lastsys=key.sysno; + lastseq=key.seqno; + if ( pp->is->method->debug >8 ) + logf (LOG_LOG,"sysno=%d seqno=%d (%x/%x) oc=%d/%d ofs=%d ", + key.sysno, key.seqno, + key.sysno, key.seqno, + occur,occurx, pp->offset); + } + /* printf("\n"); */ /*!*/ +#ifdef SKIPTHIS + if ( pp->is->method->debug >7 ) + logf(LOG_LOG,"item %d=%d:%d says %d keys, counted %d", + isam_p, isamd_type(isam_p), isamd_block(isam_p), + occur, occurx); +#endif + if (occurx != occur) + logf(LOG_LOG,"Count error!!! read %d, counted %d", occur, occurx); + assert (occurx == occur); + i = pp->cat; + if (info[1]) + i=SINGLETON_TYPE; + stat_info->no_isam_entries[i] += occur; + isamd_pp_close (pp); + } + if (stat_info->zh->reg->isamb) + { + ISAMB_PP pp; + struct it_key key; + int cat = isam_p & 3; + int level; + int size; + int blocks; + + pp = isamb_pp_open_x(stat_info->zh->reg->isamb, isam_p, &level); + while (isamb_pp_read(pp, &key)) + { + stat_info->cksum = stat_info->cksum * 65509 + + key.sysno + 11 * key.seqno; + occur++; + if (-1==firstsys) + { + firstseq=key.seqno; + firstsys=key.sysno; + } + lastsys=key.sysno; + lastseq=key.seqno; + } + isamb_pp_close_x (pp, &size, &blocks); + stat_info->isamb_blocks[cat] += blocks; + stat_info->isamb_sizes[cat] += size; + if (level > 4) + level = 4; + stat_info->isamb_levels[cat][level] ++; + stat_info->no_isam_entries[cat] += occur; + } + i=0; while (occur > stat_info->isam_bounds[i] && stat_info->isam_bounds[i]) i++; ++(stat_info->isam_occurrences[i]); + if (stat_info->dumpwords) + print_dict_item(stat_info->zh->reg->zebra_maps, name, occur, + firstsys,firstseq, lastsys, lastseq); return 0; } -void inv_prstat (BFiles bfs) +void zebra_register_statistics (ZebraHandle zh, int dumpdict) { - Dict dict; - ISAM isam = NULL; - ISAMC isamc = NULL; - Records records; + int blocks; + int size; + int count; int i, prev; int before = 0; + int occur; int after = 1000000000; struct inv_stat_info stat_info; char term_dict[2*IT_MAX_WORD+2]; + if (zebra_begin_read (zh)) + return; + + stat_info.zh = zh; + stat_info.dumpwords=dumpdict; + term_dict[0] = 1; term_dict[1] = 0; - dict = dict_open (bfs, FNAME_DICT, 100, 0); - if (!dict) - { - logf (LOG_FATAL, "dict_open fail"); - exit (1); - } - if (!res_get_match (common_resource, "isam", "i", NULL)) - { - isamc = isc_open (bfs, FNAME_ISAMC, 0, key_isamc_m (common_resource)); - if (!isamc) - { - logf (LOG_FATAL, "isc_open fail"); - exit (1); - } - } - else - { - isam = is_open (bfs, FNAME_ISAM, key_compare, 0, - sizeof(struct it_key), common_resource); - if (!isam) - { - logf (LOG_FATAL, "is_open fail"); - exit (1); - } - } - records = rec_open (bfs, 0); - - for (i = 0; i<8; i++) + for (i = 0; i<=SINGLETON_TYPE; i++) stat_info.no_isam_entries[i] = 0; stat_info.no_dict_entries = 0; stat_info.no_dict_bytes = 0; - stat_info.isam = isam; - stat_info.isamc = isamc; stat_info.isam_bounds[0] = 1; stat_info.isam_bounds[1] = 2; stat_info.isam_bounds[2] = 3; @@ -161,50 +265,228 @@ void inv_prstat (BFiles bfs) stat_info.isam_bounds[17] = 1000000; stat_info.isam_bounds[18] = 0; + stat_info.cksum = 0; + for (i = 0; i<20; i++) stat_info.isam_occurrences[i] = 0; - dict_scan (dict, term_dict, &before, &after, &stat_info, inv_stat_handle); + for (i = 0; i<10; i++) + { + int j; + for (j = 0; j<5; j++) + stat_info.isamb_levels[i][j] = 0; + stat_info.isamb_sizes[i] = 0; + stat_info.isamb_blocks[i] = 0; + } + + dict_scan (zh->reg->dict, term_dict, &before, &after, &stat_info, + inv_stat_handle); - if (isamc) + if (zh->reg->isamc) { - fprintf (stderr, " Blocks Occur Size KB Bytes/Entry\n"); - for (i = 0; isc_block_used (isamc, i) >= 0; i++) + fprintf (stdout, " Blocks Occur Size KB Bytes/Entry\n"); + for (i = 0; isc_block_used (zh->reg->isamc, i) >= 0; i++) { - fprintf (stderr, " %8d %8d", isc_block_used (isamc, i), + fprintf (stdout, " %8d %8d", isc_block_used (zh->reg->isamc, i), stat_info.no_isam_entries[i]); if (stat_info.no_isam_entries[i]) - fprintf (stderr, " %8d %f", - (int) ((1023.0 + (double) isc_block_used(isamc, i) * - isc_block_size(isamc,i))/1024), - ((double) isc_block_used(isamc, i) * - isc_block_size(isamc,i))/ + fprintf (stdout, " %8d %f", + (int) ((1023.0 + (double) + isc_block_used(zh->reg->isamc, i) * + isc_block_size(zh->reg->isamc,i))/1024), + ((double) isc_block_used(zh->reg->isamc, i) * + isc_block_size(zh->reg->isamc,i))/ stat_info.no_isam_entries[i]); - fprintf (stderr, "\n"); + fprintf (stdout, "\n"); } } - - fprintf (stderr, "\n%d words using %d bytes\n", + if (zh->reg->isamd) + { + fprintf (stdout, " Blocks Occur KB Bytes/Entry\n"); + if (zh->reg->isamd->method->debug >0) + logf(LOG_LOG," Blocks Occur KB Bytes/Entry"); + for (i = 0; i<=SINGLETON_TYPE; i++) + { + blocks= isamd_block_used(zh->reg->isamd,i); + size= isamd_block_size(zh->reg->isamd,i); + count=stat_info.no_isam_entries[i]; + if (i==SINGLETON_TYPE) + blocks=size=0; + if (stat_info.no_isam_entries[i]) + { + fprintf (stdout, "%c %7d %7d %7d %5.2f\n", + (i==SINGLETON_TYPE)?('z'):('A'+i), + blocks, + count, + (int) ((1023.0 + (double) blocks * size)/1024), + ((double) blocks * size)/count); + if (zh->reg->isamd->method->debug >0) + logf(LOG_LOG, "%c %7d %7d %7d %5.2f", + (i==SINGLETON_TYPE)?('z'):('A'+i), + blocks, + count, + (int) ((1023.0 + (double) blocks * size)/1024), + ((double) blocks * size)/count); + } /* entries */ + } /* for */ + } /* isamd */ + if ( (zh->reg->isamd) && (zh->reg->isamd->method->debug>0)) + fprintf (stdout, "\n%d words using %d bytes\n", stat_info.no_dict_entries, stat_info.no_dict_bytes); - fprintf (stderr, " Occurrences Words\n"); + + if (zh->reg->isamb) + { + for (i = 0; i<4; i++) + { + int j; + int bsize = isamb_block_info(zh->reg->isamb, i); + if (bsize < 0) + break; + fprintf (stdout, "Category %d\n", i); + fprintf (stdout, "Block size %d\n", bsize); + fprintf (stdout, "Blocks: %d\n", stat_info.isamb_blocks[i]); + fprintf (stdout, "Size: %d\n", stat_info.isamb_sizes[i]); + fprintf (stdout, "Entries: %d\n", stat_info.no_isam_entries[i]); + fprintf (stdout, "Total %d\n", stat_info.isamb_blocks[i]* + bsize); + for (j = 0; j<5; j++) + if (stat_info.isamb_levels[i][j]) + fprintf (stdout, "Level%d %d\n", j, + stat_info.isamb_levels[i][j]); + fprintf (stdout, "\n"); + } + } + fprintf (stdout, "Checksum %08lX\n", stat_info.cksum); + + fprintf (stdout, "Distinct words %d\n", stat_info.no_dict_entries); + occur = 0; + for (i = 0; i<9; i++) + occur += stat_info.no_isam_entries[i]; + fprintf (stdout, "Word pos %d\n", occur); + fprintf (stdout, " Occurrences Words\n"); prev = 1; for (i = 0; stat_info.isam_bounds[i]; i++) { int here = stat_info.isam_bounds[i]; - fprintf (stderr, "%7d-%-7d %7d\n", + fprintf (stdout, "%7d-%-7d %7d\n", prev, here, stat_info.isam_occurrences[i]); prev = here+1; } - fprintf (stderr, "%7d- %7d\n", + fprintf (stdout, "%7d- %7d\n", prev, stat_info.isam_occurrences[i]); + xmalloc_trav("unfreed"); /*! while hunting memory leaks */ + zebra_end_read (zh); +} - rec_close (&records); - dict_close (dict); - if (isam) - is_close (isam); - if (isamc) - isc_close (isamc); - -} +/* + * + * $Log: invstat.c,v $ + * Revision 1.32 2002-07-12 18:12:22 heikki + * Isam-D now stores small entries directly in the dictionary. + * Needs more tuning and cleaning... + * + * Revision 1.31 2002/07/11 16:16:00 heikki + * Fixed a bug in isamd, failed to store a single key when its bits + * did not fit into a singleton. + * + * Revision 1.30 2002/07/11 13:03:01 heikki + * Added dumpdict command line option to dump the + * dictionary before doing the usual stats + * + * Revision 1.29 2002/06/19 10:29:17 adam + * align block sizes for isam sys. Better plot for test + * + * Revision 1.28 2002/04/30 19:31:09 adam + * isamb delete; more statistics + * + * Revision 1.27 2002/04/30 08:28:37 adam + * isamb fixes for pp_read. Statistics + * + * Revision 1.26 2002/04/29 18:03:46 adam + * More isamb statistics + * + * Revision 1.25 2002/04/26 08:44:47 adam + * Index statistics working again + * + * Revision 1.24 2002/04/05 08:46:26 adam + * Zebra with full functionality + * + * Revision 1.23 2002/04/04 14:14:13 adam + * Multiple registers (alpha early) + * + * 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 + * Updated configure for Zmbol/Zebra dependent settings. + * + * Revision 1.19 1999/11/30 13:48:03 adam + * Improved installation. Updated for inclusion of YAZ header files. + * + * Revision 1.18 1999/10/06 11:46:36 heikki + * mproved statistics on isam-d + * + * Revision 1.17 1999/08/20 08:28:37 heikki + * Log levels + * + * Revision 1.16 1999/08/18 08:38:22 heikki + * Memory leak hunting + * + * Revision 1.15 1999/08/18 08:34:53 heikki + * isamd + * + * Revision 1.14 1999/07/14 10:59:26 adam + * Changed functions isc_getmethod, isams_getmethod. + * Improved fatal error handling (such as missing EXPLAIN schema). + * + * Revision 1.13 1999/07/08 14:23:27 heikki + * Fixed a bug in isamh_pp_read and cleaned up a bit + * + * Revision 1.12 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 1.11 1999/05/15 14:36:38 adam + * Updated dictionary. Implemented "compression" of dictionary. + * + * Revision 1.10 1999/05/12 13:08:06 adam + * First version of ISAMS. + * + * Revision 1.9 1999/02/12 13:29:23 adam + * Implemented position-flag for registers. + * + * Revision 1.8 1999/02/02 14:50:53 adam + * Updated WIN32 code specific sections. Changed header. + * + * Revision 1.7 1998/03/13 15:30:50 adam + * New functions isc_block_used and isc_block_size. Fixed 'leak' + * in isc_alloc_block. + * + * Revision 1.6 1998/03/06 13:54:02 adam + * Fixed two nasty bugs in isc_merge. + * + * Revision 1.5 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 1.4 1996/11/08 11:10:21 adam + * Buffers used during file match got bigger. + * Compressed ISAM support everywhere. + * Bug fixes regarding masking characters in queries. + * Redesigned Regexp-2 queries. + * + * Revision 1.3 1996/06/04 10:18:58 adam + * Minor changes - removed include of ctype.h. + * + * Revision 1.2 1996/05/22 08:25:56 adam + * Minor change. + * + * Revision 1.1 1996/05/14 14:04:34 adam + * In zebraidx, the 'stat' command is improved. Statistics about ISAM/DICT + * is collected. + */