-/*
- * Copyright (C) 1994-1995, Index Data I/S
- * All rights reserved.
- * Sebastian Hammer, Adam Dickmeiss
- *
- * $Log: kinput.c,v $
- * 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 1.75 2006-05-17 17:46:45 adam Exp $
+ Copyright (C) 1995-2005
+ 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 Zebra; see the file LICENSE.zebra. If not, write to the
+Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
+02111-1307, USA.
+*/
+
#include <fcntl.h>
+#ifdef WIN32
+#include <io.h>
+#endif
+#if HAVE_UNISTD_H
#include <unistd.h>
+#endif
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
-#include <ctype.h>
#include <assert.h>
#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 */
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;
};
+#if 0
+static void pkey(const char *b, int mode)
+{
+ 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;
- char fname[256];
- sprintf (fname, TEMP_FNAME, f->no);
- fd = open (fname, O_RDONLY);
+ int nr = 0, r = 0, fd;
+ char fname[1024];
+ 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)
+ {
+ yaz_log (YLOG_WARN|YLOG_ERRNO, "cannot seek %s", fname);
+ close (fd);
+ return ;
+ }
}
- logf (LOG_LOG, "reading chunk from %s", fname);
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)
{
}
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->buf = xmalloc (f->chunk);
- f->prev_name = xmalloc (256);
+ f->length = 0;
+ f->readHandler = NULL;
+ 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;
d = ((c&63) << 8) + (key_file_getc (f) & 0xff);
d = (d << 8) + (key_file_getc (f) & 0xff);
break;
- case 192:
+ default: /* 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);
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)
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 {
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 */
+ int no_diffs;
+ int no_updates;
+ int no_deletions;
+ int no_insertions;
+ int 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()
+{ /* 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 (512);
+ 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;
}
}
+static int heap_read_one_raw(struct heap_info *hi, char *name, char *key)
+{
+ ZebraHandle zh = hi->zh;
+ size_t ptr_i = zh->reg->ptr_i;
+ char *cp;
+ if (!ptr_i)
+ return 0;
+ --(zh->reg->ptr_i);
+ cp=(zh->reg->key_buf)[zh->reg->ptr_top - ptr_i];
+ strcpy(name, cp);
+ memcpy(key, cp+strlen(name)+1, KEY_SIZE);
+ hi->no_iterations++;
+ return 1;
+}
+
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->raw_reading)
+ return heap_read_one_raw(hi, name, key);
+
if (!hi->heapnum)
return 0;
n = hi->ptr[1];
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
+
+#if 0
+/* for debugging only */
+static void print_dict_item(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;
+ const char *db = 0;
+
+ if (!zh)
+ yaz_log(YLOG_LOG, "ord=%d", ord);
+ else
+ {
+ zebraExplain_lookup_ord (zh->reg->zei,
+ ord, &index_type, &db, 0, 0, 0);
+
+ zebra_term_untrans(zh, index_type, dst, s + len);
+
+ yaz_log(YLOG_LOG, "ord=%d term=%s", ord, dst);
+ }
+}
+#endif
+
+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)
+ {
+ 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 ? */
{
- int nmemb;
- key_buf_ptr = KEY_SIZE;
- while (1)
+ 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
+ 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
+ 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();
}
- }
- no_diffs++;
- nmemb = key_buf_ptr / KEY_SIZE;
- assert (nmemb*KEY_SIZE == key_buf_ptr);
- if ((info = dict_lookup (dict, cur_name)))
+ else
+ {
+ 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, 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);
+ 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);
+ }
+ }
+ 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
- if (!isam_p2)
+ assert(hi->zh);
+ print_dict_item(hi->zh, hci->cur_name);
+#endif
+ if ((dict_info = dict_lookup (hi->reg->dict, hci->cur_name)))
+ {
+ 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++;
- dict_delete (dict, cur_name);
+ hi->no_deletions++;
+ if (!dict_delete (hi->reg->dict, this_name))
+ abort();
}
else
-#endif
{
- 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;
+ time_t lastTime;
+ off_t totalBytes;
+ off_t totalOffset;
+};
+
+void progressFunc (struct key_file *keyp, void *info)
+{
+ struct progressInfo *p = (struct progressInfo *) info;
+ time_t now, remaining;
+
+ if (keyp->buf_size <= 0 || p->totalBytes <= 0)
+ return ;
+ time (&now);
+
+ if (now >= p->lastTime+10)
+ {
+ p->lastTime = now;
+ remaining = (time_t) ((now - p->startTime)*
+ ((double) p->totalBytes/p->totalOffset - 1.0));
+ if (remaining <= 130)
+ yaz_log (YLOG_LOG, "Merge %2.1f%% completed; %ld seconds remaining",
+ (100.0*p->totalOffset) / p->totalBytes, (long) remaining);
+ else
+ 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 = zh->reg->key_file_no;
+ int usefile;
+ yaz_log (YLOG_DEBUG, " index_merge called with nk=%d b=%p",
+ nkeys, zh->reg->key_buf);
+ if ( (nkeys==0) && (zh->reg->key_buf==0) )
+ return; /* nothing to merge - probably flush after end-trans */
+
+ usefile = (nkeys!=0);
- dict = dict_open (dict_fname, cache, 1);
- if (!dict)
+ if (usefile)
{
- logf (LOG_FATAL, "dict_open fail of `%s'", dict_fname);
- exit (1);
+ if (nkeys < 0)
+ {
+ char fname[1024];
+ nkeys = 0;
+ while (1)
+ {
+ extract_get_fname_tmp (zh, 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;
+ progressInfo.totalOffset = 0;
+ time (&progressInfo.startTime);
+ time (&progressInfo.lastTime);
+ for (i = 1; i<=nkeys; i++)
+ {
+ 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_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]);
+ } /* use file */
+ else
+ { /* do not use file, read straight from buffer */
+ hi = key_heap_init_raw(zh, key_qsort_compare);
+ hi->reg = zh->reg;
}
- isam = is_open (isam_fname, key_compare, 1, sizeof(struct it_key));
- if (!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);
+ }
+
+ if (usefile)
{
- logf (LOG_FATAL, "is_open fail of `%s'", isam_fname);
- exit (1);
+ for (i = 1; i<=nkeys; i++)
+ {
+ extract_get_fname_tmp (zh, rbuf, i);
+ unlink (rbuf);
+ }
+ 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 . . .%7d", hi->no_iterations);
+ yaz_log (YLOG_LOG, "Distinct words .%7d", hi->no_diffs);
+ yaz_log (YLOG_LOG, "Updates. . . . .%7d", hi->no_updates);
+ yaz_log (YLOG_LOG, "Deletions. . . .%7d", hi->no_deletions);
+ yaz_log (YLOG_LOG, "Insertions . . .%7d", hi->no_insertions);
+ }
+ zh->reg->key_file_no = 0;
- kf = xmalloc ((1+nkeys) * sizeof(*kf));
- for (i = 1; i<=nkeys; i++)
- kf[i] = key_file_init (i, 32768);
-
- hi = key_heap_init (nkeys, key_qsort_compare);
- 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);
-
- 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);
+ key_heap_destroy (hi, nkeys);
}
-
+/*
+ * Local variables:
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ * vim: shiftwidth=4 tabstop=8 expandtab
+ */