X-Git-Url: http://git.indexdata.com/?p=idzebra-moved-to-github.git;a=blobdiff_plain;f=index%2Fkinput.c;h=44c44eab07943913987e2bb43927e63e91e7fe11;hp=2ba038dc4bd638e3cb3c75be48c489812b32ea57;hb=ab97a59cbf609730cbc674885832fd5bb1dc64fb;hpb=0c499c33b853a029f00aed4b3630374e955e1c86 diff --git a/index/kinput.c b/index/kinput.c index 2ba038d..44c44ea 100644 --- a/index/kinput.c +++ b/index/kinput.c @@ -1,87 +1,44 @@ -/* - * Copyright (C) 1994-1995, Index Data I/S - * All rights reserved. - * Sebastian Hammer, Adam Dickmeiss - * - * $Log: kinput.c,v $ - * 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. - * - */ +/* $Id: kinput.c,v 1.80 2006-12-06 22:52:38 adam Exp $ + Copyright (C) 1995-2006 + Index Data ApS + +This file is part of the Zebra server. + +Zebra is free software; you can redistribute it and/or modify it under +the terms of the GNU General Public License as published by the Free +Software Foundation; either version 2, or (at your option) any later +version. + +Zebra is distributed in the hope that it will be useful, but WITHOUT ANY +WARRANTY; without even the implied warranty of MERCHANTABILITY or +FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License +for more details. +You should have received a copy of the GNU General Public License +along with this program; if not, write to the Free Software +Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA + +*/ + #include +#ifdef WIN32 +#include +#endif +#if HAVE_UNISTD_H #include +#endif #include #include #include -#include #include #include "index.h" #define KEY_SIZE (1+sizeof(struct it_key)) -#define INP_NAME_MAX 8192 +#define INP_NAME_MAX 768 #define INP_BUF_START 60000 #define INP_BUF_ADD 400000 -static int no_diffs = 0; -static int no_updates = 0; -static int no_deletions = 0; -static int no_insertions = 0; -static int no_iterations = 0; - struct key_file { int no; /* file no */ off_t offset; /* file offset */ @@ -90,42 +47,66 @@ struct key_file { size_t chunk; /* number of bytes allocated */ size_t buf_ptr; /* current position in buffer */ char *prev_name; /* last word read */ - int sysno; /* last sysno */ - int seqno; /* last seqno */ + void *decode_handle; off_t length; /* length of file */ /* handler invoked in each read */ void (*readHandler)(struct key_file *keyp, void *rinfo); void *readInfo; + Res res; }; -void getFnameTmp (char *fname, int no) +#if 1 +static void pkey(const char *b, int mode) { - sprintf (fname, TEMP_FNAME, no); + key_logdump_txt(YLOG_LOG, b, mode ? "i" : "d"); +} +#endif + + +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 (zh->res, "keyTmpDir", "."); + sprintf (fname, "%s/key%d.tmp", pre, no); } void key_file_chunk_read (struct key_file *f) { - int nr = 0, r, fd; + int nr = 0, r = 0, fd; char fname[1024]; - getFnameTmp (fname, f->no); - fd = open (fname, O_RDONLY); + 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); + yaz_log(YLOG_WARN|YLOG_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); + yaz_log(YLOG_WARN|YLOG_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); + yaz_log(YLOG_WARN|YLOG_ERRNO, "cannot seek %s", fname); + close(fd); + return ; } while (f->chunk - nr > 0) { @@ -136,30 +117,38 @@ void key_file_chunk_read (struct key_file *f) } if (r == -1) { - logf (LOG_FATAL|LOG_ERRNO, "read of %s", fname); - exit (1); + yaz_log(YLOG_WARN|YLOG_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) +void key_file_destroy (struct key_file *f) +{ + iscz1_stop(f->decode_handle); + xfree (f->buf); + xfree (f->prev_name); + xfree (f); +} + +struct key_file *key_file_init (int no, int chunk, Res res) { struct key_file *f; - f = xmalloc (sizeof(*f)); - f->sysno = 0; - f->seqno = 0; + f = (struct key_file *) xmalloc (sizeof(*f)); + f->res = res; + f->decode_handle = iscz1_start(); f->no = no; f->chunk = chunk; f->offset = 0; f->length = 0; f->readHandler = NULL; - f->buf = xmalloc (f->chunk); - f->prev_name = xmalloc (512); + f->buf = (unsigned char *) xmalloc (f->chunk); + f->prev_name = (char *) xmalloc (INP_NAME_MAX); *f->prev_name = '\0'; key_file_chunk_read (f); return f; @@ -179,36 +168,13 @@ int key_file_getc (struct key_file *f) return EOF; } -int key_file_decode (struct key_file *f) -{ - int c, d; - - c = key_file_getc (f); - switch (c & 192) - { - case 0: - d = c; - break; - case 64: - d = ((c&63) << 8) + (key_file_getc (f) & 0xff); - break; - case 128: - d = ((c&63) << 8) + (key_file_getc (f) & 0xff); - d = (d << 8) + (key_file_getc (f) & 0xff); - break; - case 192: - d = ((c&63) << 8) + (key_file_getc (f) & 0xff); - d = (d << 8) + (key_file_getc (f) & 0xff); - d = (d << 8) + (key_file_getc (f) & 0xff); - break; - } - return d; -} - int key_file_read (struct key_file *f, char *key) { - int i, d, c; - struct it_key itkey; + int i, c; + char srcbuf[128]; + const char *src = srcbuf; + char *dst; + int j; c = key_file_getc (f); if (c == 0) @@ -225,22 +191,27 @@ int key_file_read (struct key_file *f, char *key) while ((key[i++] = key_file_getc (f))) ; strcpy (f->prev_name, key); - f->sysno = 0; + iscz1_reset(f->decode_handle); } - d = key_file_decode (f); - key[i++] = d & 1; - d = d >> 1; - itkey.sysno = d + f->sysno; - if (d) + c = key_file_getc(f); /* length + insert/delete combined */ + key[i++] = c & 128; + c = c & 127; + for (j = 0; j < c; j++) + srcbuf[j] = key_file_getc(f); + dst = key + i; + iscz1_decode(f->decode_handle, &dst, &src); + +#if 0 + /* debugging */ + if (1) { - f->sysno = itkey.sysno; - f->seqno = 0; + struct it_key k; + memcpy(&k, key+i, sizeof(k)); + if (!k.mem[1]) + yaz_log(YLOG_LOG, "00 KEY"); } - d = key_file_decode (f); - itkey.seqno = d + f->seqno; - f->seqno = itkey.seqno; - memcpy (key + i, &itkey, sizeof(struct it_key)); - return i + sizeof (struct it_key); +#endif + return i + sizeof(struct it_key); } struct heap_info { @@ -251,28 +222,79 @@ struct heap_info { int heapnum; int *ptr; int (*cmp)(const void *p1, const void *p2); + struct zebra_register *reg; + ZebraHandle zh; + int raw_reading; /* 1=raw /mem read. 0=file reading */ + zint no_diffs; + zint no_updates; + zint no_deletions; + zint no_insertions; + zint no_iterations; }; -struct heap_info *key_heap_init (int nkeys, - int (*cmp)(const void *p1, const void *p2)) +static struct heap_info *key_heap_malloc(void) +{ /* malloc and clear it */ + struct heap_info *hi; + hi = (struct heap_info *) xmalloc (sizeof(*hi)); + hi->info.file = 0; + hi->info.buf = 0; + hi->heapnum = 0; + hi->ptr = 0; + hi->raw_reading = 0; + hi->no_diffs = 0; + hi->no_diffs = 0; + hi->no_updates = 0; + hi->no_deletions = 0; + hi->no_insertions = 0; + hi->no_iterations = 0; + return hi; +} + +struct heap_info *key_heap_init_file(ZebraHandle zh, + int nkeys, + int (*cmp)(const void *p1, const void *p2)) { struct heap_info *hi; int i; - hi = xmalloc (sizeof(*hi)); - hi->info.file = xmalloc (sizeof(*hi->info.file) * (1+nkeys)); - hi->info.buf = xmalloc (sizeof(*hi->info.buf) * (1+nkeys)); - hi->heapnum = 0; - hi->ptr = xmalloc (sizeof(*hi->ptr) * (1+nkeys)); + hi = key_heap_malloc(); + hi->zh = zh; + hi->info.file = (struct key_file **) + xmalloc (sizeof(*hi->info.file) * (1+nkeys)); + hi->info.buf = (char **) xmalloc (sizeof(*hi->info.buf) * (1+nkeys)); + hi->ptr = (int *) xmalloc (sizeof(*hi->ptr) * (1+nkeys)); hi->cmp = cmp; for (i = 0; i<= nkeys; i++) { hi->ptr[i] = i; - hi->info.buf[i] = xmalloc (768); + hi->info.buf[i] = (char *) xmalloc (INP_NAME_MAX); } return hi; } +struct heap_info *key_heap_init_raw(ZebraHandle zh, + int (*cmp)(const void *p1, const void *p2)) +{ + struct heap_info *hi=key_heap_malloc(); + hi->cmp = cmp; + hi->zh = zh; + hi->raw_reading = 1; + return hi; +} + +void key_heap_destroy (struct heap_info *hi, int nkeys) +{ + int i; + if (!hi->raw_reading) + for (i = 0; i<=nkeys; i++) + xfree (hi->info.buf[i]); + + xfree (hi->info.buf); + xfree (hi->ptr); + xfree (hi->info.file); + xfree (hi); +} + static void key_heap_swap (struct heap_info *hi, int i1, int i2) { int swap; @@ -330,7 +352,7 @@ static void key_heap_insert (struct heap_info *hi, const char *buf, int nbytes, static int heap_read_one (struct heap_info *hi, char *name, char *key) { int n, r; - char rbuf[512]; + char rbuf[INP_NAME_MAX]; struct key_file *kf; if (!hi->heapnum) @@ -343,81 +365,342 @@ static int heap_read_one (struct heap_info *hi, char *name, char *key) key_heap_delete (hi); if ((r = key_file_read (kf, rbuf))) key_heap_insert (hi, rbuf, r, kf); - no_iterations++; + hi->no_iterations++; return 1; } -int heap_inp (Dict dict, ISAM isam, struct heap_info *hi) +#define PR_KEY_LOW 0 +#define PR_KEY_TOP 0 + +/* for debugging only */ +void zebra_log_dict_entry(ZebraHandle zh, const char *s) { - char *info; - char next_name[INP_NAME_MAX+1]; - char cur_name[INP_NAME_MAX+1]; - int key_buf_size = INP_BUF_START; - int key_buf_ptr; - char *next_key; - char *key_buf; + char dst[IT_MAX_WORD+1]; + int ord; + int len = key_SU_decode(&ord, (const unsigned char *) s); + int index_type; + + if (!zh) + yaz_log(YLOG_LOG, "ord=%d", ord); + else + { + const char *string_index; + const char *db = 0; + zebraExplain_lookup_ord(zh->reg->zei, + ord, &index_type, &db, &string_index); + + zebra_term_untrans(zh, index_type, dst, s + len); + + yaz_log(YLOG_LOG, "ord=%d index_type=%c index=%s term=%s", + ord, index_type, string_index, dst); + } +} + +struct heap_cread_info { + char prev_name[INP_NAME_MAX]; + char cur_name[INP_NAME_MAX]; + char *key; + char *key_1, *key_2; + int mode_1, mode_2; + int sz_1, sz_2; + struct heap_info *hi; + int first_in_list; int more; - - next_key = xmalloc (KEY_SIZE); - key_buf = xmalloc (key_buf_size * (KEY_SIZE)); - more = heap_read_one (hi, cur_name, key_buf); - while (more) /* EOF ? */ + int ret; + int look_level; +}; + +static int heap_cread_item (void *vp, char **dst, int *insertMode); + +int heap_cread_item2(void *vp, char **dst, int *insertMode) +{ + struct heap_cread_info *p = (struct heap_cread_info *) vp; + int level = 0; + + if (p->look_level) { - int nmemb; - key_buf_ptr = KEY_SIZE; - while (1) + if (p->look_level > 0) + { + *insertMode = 1; + p->look_level--; + } + else + { + *insertMode = 0; + p->look_level++; + } + memcpy (*dst, p->key_1, p->sz_1); +#if 0 + yaz_log(YLOG_LOG, "DUP level=%d", p->look_level); + pkey(*dst, *insertMode); +#endif + (*dst) += p->sz_1; + return 1; + } + if (p->ret == 0) /* lookahead was 0?. Return that in read next round */ + { + p->ret = -1; + return 0; + } + else if (p->ret == -1) /* Must read new item ? */ + { + char *dst_1 = p->key_1; + p->ret = heap_cread_item(vp, &dst_1, &p->mode_1); + p->sz_1 = dst_1 - p->key_1; + } + else + { /* lookahead in 2 . Now in 1. */ + p->sz_1 = p->sz_2; + p->mode_1 = p->mode_2; + memcpy (p->key_1, p->key_2, p->sz_2); + } + if (p->mode_1) + level = 1; /* insert */ + else + level = -1; /* delete */ + while(1) + { + char *dst_2 = p->key_2; + p->ret = heap_cread_item(vp, &dst_2, &p->mode_2); + if (!p->ret) { - if (!(more = heap_read_one (hi, next_name, next_key))) + if (level) break; - if (*next_name && strcmp (next_name, cur_name)) + p->ret = -1; + return 0; + } + p->sz_2 = dst_2 - p->key_2; + + if (key_compare(p->key_1, p->key_2) == 0) + { + if (p->mode_2) /* adjust level according to deletes/inserts */ + level++; + else + level--; + } + else + { + if (level) break; - memcpy (key_buf + key_buf_ptr, next_key, KEY_SIZE); - key_buf_ptr += KEY_SIZE; - if (key_buf_ptr+KEY_SIZE >= key_buf_size) + /* all the same. new round .. */ + p->sz_1 = p->sz_2; + p->mode_1 = p->mode_2; + memcpy (p->key_1, p->key_2, p->sz_1); + if (p->mode_1) + level = 1; /* insert */ + else + level = -1; /* delete */ + } + } + /* outcome is insert (1) or delete (0) depending on final level */ + if (level > 0) + { + *insertMode = 1; + level--; + } + else + { + *insertMode = 0; + level++; + } + p->look_level = level; + memcpy (*dst, p->key_1, p->sz_1); +#if 0 + pkey(*dst, *insertMode); +#endif + (*dst) += p->sz_1; + return 1; +} + +int heap_cread_item (void *vp, char **dst, int *insertMode) +{ + struct heap_cread_info *p = (struct heap_cread_info *) vp; + struct heap_info *hi = p->hi; + + if (p->first_in_list) + { + *insertMode = p->key[0]; + memcpy (*dst, p->key+1, sizeof(struct it_key)); +#if PR_KEY_LOW + zebra_log_dict_entry(hi->zh, p->cur_name); + pkey(*dst, *insertMode); +#endif + (*dst) += sizeof(struct it_key); + p->first_in_list = 0; + return 1; + } + strcpy (p->prev_name, p->cur_name); + if (!(p->more = heap_read_one (hi, p->cur_name, p->key))) + return 0; + if (*p->cur_name && strcmp (p->cur_name, p->prev_name)) + { + p->first_in_list = 1; + return 0; + } + *insertMode = p->key[0]; + memcpy (*dst, p->key+1, sizeof(struct it_key)); +#if PR_KEY_LOW + zebra_log_dict_entry(hi->zh, p->cur_name); + pkey(*dst, *insertMode); +#endif + (*dst) += sizeof(struct it_key); + return 1; +} + +int heap_inpc (struct heap_cread_info *hci, struct heap_info *hi) +{ + ISAMC_I *isamc_i = (ISAMC_I *) xmalloc (sizeof(*isamc_i)); + + isamc_i->clientData = hci; + isamc_i->read_item = heap_cread_item2; + + while (hci->more) + { + char this_name[INP_NAME_MAX]; + ISAM_P isamc_p, isamc_p2; + char *dict_info; + + strcpy (this_name, hci->cur_name); + assert (hci->cur_name[0]); + hi->no_diffs++; + if ((dict_info = dict_lookup (hi->reg->dict, hci->cur_name))) + { + memcpy (&isamc_p, dict_info+1, sizeof(ISAM_P)); + isamc_p2 = isamc_p; + isamc_merge (hi->reg->isamc, &isamc_p2, isamc_i); + if (!isamc_p2) { - char *new_key_buf; - new_key_buf = xmalloc (key_buf_size + INP_BUF_ADD); - memcpy (new_key_buf, key_buf, key_buf_size); - key_buf_size += INP_BUF_ADD; - xfree (key_buf); - key_buf = new_key_buf; + hi->no_deletions++; + if (!dict_delete (hi->reg->dict, this_name)) + abort(); } + else + { + hi->no_updates++; + if (isamc_p2 != isamc_p) + dict_insert (hi->reg->dict, this_name, + sizeof(ISAM_P), &isamc_p2); + } + } + else + { + isamc_p = 0; + isamc_merge (hi->reg->isamc, &isamc_p, isamc_i); + hi->no_insertions++; + if (isamc_p) + dict_insert (hi->reg->dict, this_name, + sizeof(ISAM_P), &isamc_p); } - no_diffs++; - nmemb = key_buf_ptr / KEY_SIZE; - assert (nmemb*KEY_SIZE == key_buf_ptr); - if ((info = dict_lookup (dict, cur_name))) + } + xfree (isamc_i); + return 0; +} + +int heap_inp0(struct heap_cread_info *hci, struct heap_info *hi) +{ + while (hci->more) + { + char this_name[INP_NAME_MAX]; + char mybuf[1024]; + char *dst = mybuf; + int mode; + + strcpy (this_name, hci->cur_name); + assert (hci->cur_name[0]); + hi->no_diffs++; + + while (heap_cread_item2(hci, &dst, &mode)) + ; + } + return 0; +} + + +int heap_inpb(struct heap_cread_info *hci, struct heap_info *hi) +{ + ISAMC_I *isamc_i = (ISAMC_I *) xmalloc (sizeof(*isamc_i)); + + isamc_i->clientData = hci; + isamc_i->read_item = heap_cread_item2; + + while (hci->more) + { + char this_name[INP_NAME_MAX]; + ISAM_P isamc_p, isamc_p2; + char *dict_info; + + strcpy (this_name, hci->cur_name); + assert (hci->cur_name[0]); + hi->no_diffs++; + +#if 0 + assert(hi->zh); + zebra_log_dict_entry(hi->zh, hci->cur_name); +#endif + if ((dict_info = dict_lookup (hi->reg->dict, hci->cur_name))) { - ISAM_P isam_p, isam_p2; - logf (LOG_DEBUG, "updating %s", cur_name); - memcpy (&isam_p, info+1, sizeof(ISAM_P)); - isam_p2 = is_merge (isam, isam_p, nmemb, key_buf); - if (!isam_p2) + memcpy (&isamc_p, dict_info+1, sizeof(ISAM_P)); + isamc_p2 = isamc_p; + isamb_merge (hi->reg->isamb, &isamc_p2, isamc_i); + if (!isamc_p2) { - no_deletions++; - if (!dict_delete (dict, cur_name)) - abort (); + hi->no_deletions++; + if (!dict_delete (hi->reg->dict, this_name)) + abort(); } else { - no_updates++; - if (isam_p2 != isam_p) - dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p2); + hi->no_updates++; + if (isamc_p2 != isamc_p) + dict_insert (hi->reg->dict, this_name, + sizeof(ISAM_P), &isamc_p2); } - } + } else { - ISAM_P isam_p; - logf (LOG_DEBUG, "inserting %s", cur_name); - no_insertions++; - isam_p = is_merge (isam, 0, nmemb, key_buf); - dict_insert (dict, cur_name, sizeof(ISAM_P), &isam_p); + isamc_p = 0; + isamb_merge (hi->reg->isamb, &isamc_p, isamc_i); + hi->no_insertions++; + if (isamc_p) + dict_insert (hi->reg->dict, this_name, + sizeof(ISAM_P), &isamc_p); } - memcpy (key_buf, next_key, KEY_SIZE); - strcpy (cur_name, next_name); } + xfree(isamc_i); return 0; -} +} + +int heap_inps (struct heap_cread_info *hci, struct heap_info *hi) +{ + ISAMS_I isams_i = (ISAMS_I) xmalloc (sizeof(*isams_i)); + + isams_i->clientData = hci; + isams_i->read_item = heap_cread_item; + + while (hci->more) + { + char this_name[INP_NAME_MAX]; + ISAM_P isams_p; + char *dict_info; + + strcpy (this_name, hci->cur_name); + assert (hci->cur_name[0]); + hi->no_diffs++; + if (!(dict_info = dict_lookup (hi->reg->dict, hci->cur_name))) + { + isams_p = isams_merge (hi->reg->isams, isams_i); + hi->no_insertions++; + dict_insert (hi->reg->dict, this_name, sizeof(ISAM_P), &isams_p); + } + else + { + yaz_log(YLOG_FATAL, "isams doesn't support this kind of update"); + break; + } + } + xfree (isams_i); + return 0; +} struct progressInfo { time_t startTime; @@ -428,7 +711,7 @@ struct progressInfo { void progressFunc (struct key_file *keyp, void *info) { - struct progressInfo *p = info; + struct progressInfo *p = (struct progressInfo *) info; time_t now, remaining; if (keyp->buf_size <= 0 || p->totalBytes <= 0) @@ -438,37 +721,41 @@ void progressFunc (struct key_file *keyp, void *info) if (now >= p->lastTime+10) { p->lastTime = now; - remaining = (now - p->startTime)* - ((double) p->totalBytes/p->totalOffset - 1.0); + remaining = (time_t) ((now - p->startTime)* + ((double) p->totalBytes/p->totalOffset - 1.0)); if (remaining <= 130) - logf (LOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining", + yaz_log(YLOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining", (100.0*p->totalOffset) / p->totalBytes, (long) remaining); else - logf (LOG_LOG, "Merge %2.1f%% completed; %ld minutes remaining", + yaz_log(YLOG_LOG, "Merge %2.1f%% completed; %ld minutes remaining", (100.0*p->totalOffset) / p->totalBytes, (long) remaining/60); } p->totalOffset += keyp->buf_size; } -void key_input (const char *dict_fname, const char *isam_fname, - int nkeys, int cache) - +#ifndef R_OK +#define R_OK 4 +#endif + +void zebra_index_merge (ZebraHandle zh) { - Dict dict; - ISAM isam; - struct key_file **kf; + struct key_file **kf = 0; char rbuf[1024]; int i, r; struct heap_info *hi; struct progressInfo progressInfo; + int nkeys = key_block_get_no_files(zh->reg->key_block); + if (nkeys == 0) + return; + 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++; @@ -476,51 +763,75 @@ void key_input (const char *dict_fname, const char *isam_fname, if (!nkeys) return ; } - dict = dict_open (dict_fname, cache, 1); - if (!dict) - { - logf (LOG_FATAL, "dict_open fail of `%s'", dict_fname); - exit (1); - } - isam = is_open (isam_fname, key_compare, 1, sizeof(struct it_key)); - if (!isam) - { - logf (LOG_FATAL, "is_open fail of `%s'", isam_fname); - exit (1); - } - - kf = xmalloc ((1+nkeys) * sizeof(*kf)); + 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++) { - kf[i] = key_file_init (i, 32768); + kf[i] = key_file_init (i, 8192, zh->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 = key_heap_init_file(zh, nkeys, key_qsort_compare); + hi->reg = zh->reg; + for (i = 1; i<=nkeys; i++) if ((r = key_file_read (kf[i], rbuf))) key_heap_insert (hi, rbuf, r, kf[i]); - heap_inp (dict, isam, hi); - dict_close (dict); - is_close (isam); - + if (1) + { + struct heap_cread_info hci; + + hci.key = (char *) xmalloc (KEY_SIZE); + hci.key_1 = (char *) xmalloc (KEY_SIZE); + hci.key_2 = (char *) xmalloc (KEY_SIZE); + hci.ret = -1; + hci.first_in_list = 1; + hci.hi = hi; + hci.look_level = 0; + hci.more = heap_read_one (hi, hci.cur_name, hci.key); + + if (zh->reg->isams) + heap_inps(&hci, hi); + if (zh->reg->isamc) + heap_inpc(&hci, hi); + if (zh->reg->isamb) + heap_inpb(&hci, hi); + + xfree (hci.key); + xfree (hci.key_1); + xfree (hci.key_2); + } + for (i = 1; i<=nkeys; i++) { - getFnameTmp (rbuf, i); + extract_get_fname_tmp (zh, rbuf, i); unlink (rbuf); } - 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); + for (i = 1; i<=nkeys; i++) + key_file_destroy (kf[i]); + xfree (kf); + if (hi->no_iterations) + { /* do not log if nothing happened */ + yaz_log(YLOG_LOG, "Iterations %9" ZINT_FORMAT0, hi->no_iterations); + yaz_log(YLOG_LOG, "Distinct words %9" ZINT_FORMAT0, hi->no_diffs); + yaz_log(YLOG_LOG, "Updates %9" ZINT_FORMAT0, hi->no_updates); + yaz_log(YLOG_LOG, "Deletions %9" ZINT_FORMAT0, hi->no_deletions); + yaz_log(YLOG_LOG, "Insertions %9" ZINT_FORMAT0, hi->no_insertions); + } + key_block_destroy(&zh->reg->key_block); + key_heap_destroy(hi, nkeys); } - +/* + * Local variables: + * c-basic-offset: 4 + * indent-tabs-mode: nil + * End: + * vim: shiftwidth=4 tabstop=8 expandtab + */