X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fkinput.c;h=b9d7f066da1b2e9c7b3cce036f3bc84e1723cb63;hp=9a2d6c1b0b667800bea961b5e9d9f5dc860b8ba7;hb=7e75317bed8eecabcb57e59b16093a32238738e2;hpb=e8f4cf34c75ffdc7bbfaf3fa232f29cc1384b60c diff --git a/index/kinput.c b/index/kinput.c index 9a2d6c1..b9d7f06 100644 --- a/index/kinput.c +++ b/index/kinput.c @@ -1,9 +1,13 @@ /* - * Copyright (C) 1994-1999, 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. + * Doesn't matter, as the program terminates soon after. */ @@ -19,6 +23,7 @@ #include #include "index.h" +#include "zserver.h" #define KEY_SIZE (1+sizeof(struct it_key)) #define INP_NAME_MAX 768 @@ -45,13 +50,22 @@ struct key_file { /* handler invoked in each read */ void (*readHandler)(struct key_file *keyp, void *rinfo); void *readInfo; + Res res; }; -void getFnameTmp (char *fname, int no) +void getFnameTmp (Res res, char *fname, int no) +{ + const char *pre; + + pre = res_get_def (res, "keyTmpDir", "."); + sprintf (fname, "%s/key%d.tmp", pre, no); +} + +void extract_get_fname_tmp (ZebraHandle zh, char *fname, int no) { const char *pre; - pre = res_get_def (common_resource, "keyTmpDir", "."); + pre = res_get_def (zh->service->res, "keyTmpDir", "."); sprintf (fname, "%s/key%d.tmp", pre, no); } @@ -59,25 +73,30 @@ void key_file_chunk_read (struct key_file *f) { int nr = 0, r = 0, fd; char fname[1024]; - getFnameTmp (fname, f->no); + getFnameTmp (f->res, fname, f->no); fd = open (fname, O_BINARY|O_RDONLY); + + f->buf_ptr = 0; + f->buf_size = 0; if (fd == -1) { - logf (LOG_FATAL|LOG_ERRNO, "cannot open %s", fname); - exit (1); + logf (LOG_WARN|LOG_ERRNO, "cannot open %s", fname); + return ; } if (!f->length) { if ((f->length = lseek (fd, 0L, SEEK_END)) == (off_t) -1) { - logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname); - exit (1); + logf (LOG_WARN|LOG_ERRNO, "cannot seek %s", fname); + close (fd); + return ; } } if (lseek (fd, f->offset, SEEK_SET) == -1) { - logf (LOG_FATAL|LOG_ERRNO, "cannot seek %s", fname); - exit (1); + logf (LOG_WARN|LOG_ERRNO, "cannot seek %s", fname); + close(fd); + return ; } while (f->chunk - nr > 0) { @@ -88,21 +107,22 @@ void key_file_chunk_read (struct key_file *f) } if (r == -1) { - logf (LOG_FATAL|LOG_ERRNO, "read of %s", fname); - exit (1); + logf (LOG_WARN|LOG_ERRNO, "read of %s", fname); + close (fd); + return; } f->buf_size = nr; - f->buf_ptr = 0; if (f->readHandler) (*f->readHandler)(f, f->readInfo); close (fd); } -struct key_file *key_file_init (int no, int chunk) +struct key_file *key_file_init (int no, int chunk, Res res) { struct key_file *f; f = (struct key_file *) xmalloc (sizeof(*f)); + f->res = res; f->sysno = 0; f->seqno = 0; f->no = no; @@ -204,11 +224,12 @@ struct heap_info { int *ptr; int (*cmp)(const void *p1, const void *p2); Dict dict; + ISAMS isams; +#if ZMBOL ISAM isam; ISAMC isamc; - ISAMS isams; - ISAMH isamh; ISAMD isamd; +#endif }; struct heap_info *key_heap_init (int nkeys, @@ -342,6 +363,7 @@ int heap_cread_item (void *vp, char **dst, int *insertMode) return 1; } +#if ZMBOL int heap_inpc (struct heap_info *hi) { struct heap_cread_info hci; @@ -390,92 +412,7 @@ int heap_inpc (struct heap_info *hi) } } xfree (isamc_i); - return 0; -} - -int heap_inps (struct heap_info *hi) -{ - struct heap_cread_info hci; - ISAMS_I isams_i = (ISAMS_I) xmalloc (sizeof(*isams_i)); - - hci.key = (char *) xmalloc (KEY_SIZE); - hci.mode = 1; - hci.hi = hi; - hci.more = heap_read_one (hi, hci.cur_name, hci.key); - - isams_i->clientData = &hci; - isams_i->read_item = heap_cread_item; - - while (hci.more) - { - char this_name[INP_NAME_MAX]; - ISAMS_P isams_p; - char *dict_info; - - strcpy (this_name, hci.cur_name); - assert (hci.cur_name[1]); - no_diffs++; - if (!(dict_info = dict_lookup (hi->dict, hci.cur_name))) - { - isams_p = isams_merge (hi->isams, isams_i); - no_insertions++; - dict_insert (hi->dict, this_name, sizeof(ISAMS_P), &isams_p); - } - else - abort(); - } - xfree (isams_i); - return 0; -} - -int heap_inph (struct heap_info *hi) -{ - struct heap_cread_info hci; - ISAMH_I isamh_i = (ISAMH_I) xmalloc (sizeof(*isamh_i)); - - hci.key = (char *) xmalloc (KEY_SIZE); - hci.mode = 1; - hci.hi = hi; - hci.more = heap_read_one (hi, hci.cur_name, hci.key); - - isamh_i->clientData = &hci; - isamh_i->read_item = heap_cread_item; - - while (hci.more) - { - char this_name[INP_NAME_MAX]; - ISAMH_P isamh_p, isamh_p2; - char *dict_info; - - strcpy (this_name, hci.cur_name); - assert (hci.cur_name[1]); - no_diffs++; - if ((dict_info = dict_lookup (hi->dict, hci.cur_name))) - { - memcpy (&isamh_p, dict_info+1, sizeof(ISAMH_P)); - isamh_p2 = isamh_append (hi->isamh, isamh_p, isamh_i); - if (!isamh_p2) - { - no_deletions++; - if (!dict_delete (hi->dict, this_name)) - abort(); - } - else - { - no_updates++; - if (isamh_p2 != isamh_p) - dict_insert (hi->dict, this_name, - sizeof(ISAMH_P), &isamh_p2); - } - } - else - { - isamh_p = isamh_append (hi->isamh, 0, isamh_i); - no_insertions++; - dict_insert (hi->dict, this_name, sizeof(ISAMH_P), &isamh_p); - } - } - xfree (isamh_i); + xfree (hci.key); return 0; } @@ -530,9 +467,6 @@ int heap_inpd (struct heap_info *hi) return 0; } - - - int heap_inp (struct heap_info *hi) { char *info; @@ -603,6 +537,46 @@ int heap_inp (struct heap_info *hi) return 0; } +#endif + +int heap_inps (struct heap_info *hi) +{ + struct heap_cread_info hci; + ISAMS_I isams_i = (ISAMS_I) xmalloc (sizeof(*isams_i)); + + hci.key = (char *) xmalloc (KEY_SIZE); + hci.mode = 1; + hci.hi = hi; + hci.more = heap_read_one (hi, hci.cur_name, hci.key); + + isams_i->clientData = &hci; + isams_i->read_item = heap_cread_item; + + while (hci.more) + { + char this_name[INP_NAME_MAX]; + ISAMS_P isams_p; + char *dict_info; + + strcpy (this_name, hci.cur_name); + assert (hci.cur_name[1]); + no_diffs++; + if (!(dict_info = dict_lookup (hi->dict, hci.cur_name))) + { + isams_p = isams_merge (hi->isams, isams_i); + no_insertions++; + dict_insert (hi->dict, this_name, sizeof(ISAMS_P), &isams_p); + } + else + { + logf (LOG_FATAL, "isams doesn't support this kind of update"); + break; + } + } + xfree (isams_i); + return 0; +} + struct progressInfo { time_t startTime; time_t lastTime; @@ -638,28 +612,22 @@ void progressFunc (struct key_file *keyp, void *info) #define R_OK 4 #endif -void key_input (BFiles bfs, int nkeys, int cache) - +void zebra_index_merge (ZebraHandle zh) { - Dict dict; - ISAM isam = NULL; - ISAMC isamc = NULL; - ISAMS isams = NULL; - ISAMH isamh = NULL; - ISAMD isamd = NULL; struct key_file **kf; char rbuf[1024]; int i, r; struct heap_info *hi; struct progressInfo progressInfo; - + int nkeys = zh->key_file_no; + if (nkeys < 0) { char fname[1024]; nkeys = 0; while (1) { - getFnameTmp (fname, nkeys+1); + extract_get_fname_tmp (zh, fname, nkeys+1); if (access (fname, R_OK) == -1) break; nkeys++; @@ -667,64 +635,77 @@ void key_input (BFiles bfs, int nkeys, int cache) 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 (common_resource, "isam", "s", NULL)) - { - struct ISAMS_M_s isams_m; - isams = isams_open (bfs, FNAME_ISAMS, 1, - key_isams_m (common_resource, &isams_m)); - if (!isams) - { - logf (LOG_FATAL, "isams_open fail"); - exit (1); - } - } - else if (res_get_match (common_resource, "isam", "i", NULL)) - { - isam = is_open (bfs, FNAME_ISAM, key_compare, 1, - sizeof(struct it_key), common_resource); - if (!isam) - { - logf (LOG_FATAL, "is_open fail"); - exit (1); - } - } - else if (res_get_match (common_resource, "isam", "h", NULL)) + kf = (struct key_file **) xmalloc ((1+nkeys) * sizeof(*kf)); + progressInfo.totalBytes = 0; + progressInfo.totalOffset = 0; + time (&progressInfo.startTime); + time (&progressInfo.lastTime); + for (i = 1; i<=nkeys; i++) { - isamh = isamh_open (bfs, FNAME_ISAMH, 1, - key_isamh_m (common_resource)); - if (!isamh) - { - logf (LOG_FATAL, "isamh_open fail"); - exit (1); - } + kf[i] = key_file_init (i, 8192, zh->service->res); + kf[i]->readHandler = progressFunc; + kf[i]->readInfo = &progressInfo; + progressInfo.totalBytes += kf[i]->length; + progressInfo.totalOffset += kf[i]->buf_size; } - else if (res_get_match (common_resource, "isam", "d", NULL)) + hi = key_heap_init (nkeys, key_qsort_compare); + hi->dict = zh->service->dict; + hi->isams = zh->service->isams; +#if ZMBOL + 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 (zh->service->isams) + heap_inps (hi); +#if ZMBOL + else if (zh->service->isamc) + heap_inpc (hi); + else if (zh->service->isam) + heap_inp (hi); + else if (zh->service->isamd) + heap_inpd (hi); +#endif + + for (i = 1; i<=nkeys; i++) { - struct ISAMD_M_s isamd_m; - isamd = isamd_open (bfs, FNAME_ISAMD, 1, - key_isamd_m (common_resource,&isamd_m)); - if (!isamd) - { - logf (LOG_FATAL, "isamd_open fail"); - exit (1); - } + extract_get_fname_tmp (zh, rbuf, i); + unlink (rbuf); } - else + logf (LOG_LOG, "Iterations . . .%7d", no_iterations); + logf (LOG_LOG, "Distinct words .%7d", no_diffs); + logf (LOG_LOG, "Updates. . . . .%7d", no_updates); + logf (LOG_LOG, "Deletions. . . .%7d", no_deletions); + logf (LOG_LOG, "Insertions . . .%7d", no_insertions); + zh->key_file_no = 0; +} + +void key_input (ZebraHandle zh, int nkeys, int cache, Res res) + +{ + struct key_file **kf; + char rbuf[1024]; + int i, r; + struct heap_info *hi; + struct progressInfo progressInfo; + + if (nkeys < 0) { - struct ISAMC_M_s isamc_m; - isamc = isc_open (bfs, FNAME_ISAMC, 1, - key_isamc_m (common_resource, &isamc_m)); - if (!isamc) + char fname[1024]; + nkeys = 0; + while (1) { - logf (LOG_FATAL, "isc_open fail"); - exit (1); + getFnameTmp (res, fname, nkeys+1); + if (access (fname, R_OK) == -1) + break; + nkeys++; } + if (!nkeys) + return ; } kf = (struct key_file **) xmalloc ((1+nkeys) * sizeof(*kf)); progressInfo.totalBytes = 0; @@ -733,49 +714,38 @@ void key_input (BFiles bfs, int nkeys, int cache) time (&progressInfo.lastTime); for (i = 1; i<=nkeys; i++) { - kf[i] = key_file_init (i, 32768); + kf[i] = key_file_init (i, 8192, res); kf[i]->readHandler = progressFunc; kf[i]->readInfo = &progressInfo; progressInfo.totalBytes += kf[i]->length; progressInfo.totalOffset += kf[i]->buf_size; } hi = key_heap_init (nkeys, key_qsort_compare); - hi->dict = dict; - hi->isam = isam; - hi->isamc = isamc; - hi->isams = isams; - hi->isamh = isamh; - hi->isamd = isamd; + hi->dict = zh->service->dict; + hi->isams = zh->service->isams; +#if ZMBOL + 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 (isamc) - heap_inpc (hi); - else if (isams) + if (hi->isams) heap_inps (hi); - else if (isam) +#if ZMBOL + else if (hi->isamc) + heap_inpc (hi); + else if (hi->isam) heap_inp (hi); - else if (isamh) - heap_inph (hi); - else if (isamd) + else if (hi->isamd) heap_inpd (hi); +#endif - dict_close (dict); - if (isam) - is_close (isam); - if (isamc) - isc_close (isamc); - if (isams) - isams_close (isams); - if (isamh) - isamh_close (isamh); - if (isamd) - isamd_close (isamd); - for (i = 1; i<=nkeys; i++) { - getFnameTmp (rbuf, i); + getFnameTmp (res, rbuf, i); unlink (rbuf); } logf (LOG_LOG, "Iterations . . .%7d", no_iterations); @@ -783,137 +753,7 @@ void key_input (BFiles bfs, int nkeys, int cache) logf (LOG_LOG, "Updates. . . . .%7d", no_updates); logf (LOG_LOG, "Deletions. . . .%7d", no_deletions); logf (LOG_LOG, "Insertions . . .%7d", no_insertions); -} - - -/* - * $Log: kinput.c,v $ - * 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. - * - */ + /* xmalloc_trav("unfreed"); while hunting leaks */ +}