-/*
- * Copyright (C) 1994-2002, Index Data
- * All rights reserved.
- * Sebastian Hammer, Adam Dickmeiss, Heikki Levanto
- *
- * $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.
-
- */
+/* $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>
-#else
+#endif
+#if HAVE_UNISTD_H
#include <unistd.h>
#endif
#include <stdlib.h>
#include <assert.h>
#include "index.h"
-#include "zserver.h"
#define KEY_SIZE (1+sizeof(struct it_key))
#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);
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;
{
const char *pre;
- pre = res_get_def (zh->service->res, "keyTmpDir", ".");
+ pre = res_get_def (zh->res, "keyTmpDir", ".");
sprintf (fname, "%s/key%d.tmp", pre, no);
}
f->buf_size = 0;
if (fd == -1)
{
- logf (LOG_WARN|LOG_ERRNO, "cannot open %s", fname);
+ 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_WARN|LOG_ERRNO, "cannot seek %s", fname);
+ yaz_log (YLOG_WARN|YLOG_ERRNO, "cannot seek %s", fname);
close (fd);
return ;
}
}
if (lseek (fd, f->offset, SEEK_SET) == -1)
{
- logf (LOG_WARN|LOG_ERRNO, "cannot seek %s", fname);
+ yaz_log (YLOG_WARN|YLOG_ERRNO, "cannot seek %s", fname);
close(fd);
return ;
}
}
if (r == -1)
{
- logf (LOG_WARN|LOG_ERRNO, "read of %s", fname);
+ yaz_log (YLOG_WARN|YLOG_ERRNO, "read of %s", fname);
close (fd);
return;
}
close (fd);
}
+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 = (struct key_file *) xmalloc (sizeof(*f));
f->res = res;
- f->sysno = 0;
- f->seqno = 0;
+ f->decode_handle = iscz1_start();
f->no = no;
f->chunk = chunk;
f->offset = 0;
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);
- Dict dict;
- ISAMS isams;
-#if ZMBOL
- ISAM isam;
- ISAMC isamc;
- ISAMD isamd;
-#endif
+ 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 = (struct heap_info *) xmalloc (sizeof(*hi));
+ hi = key_heap_malloc();
+ hi->zh = zh;
hi->info.file = (struct key_file **)
- xmalloc (sizeof(*hi->info.file) * (1+nkeys));
+ xmalloc (sizeof(*hi->info.file) * (1+nkeys));
hi->info.buf = (char **) xmalloc (sizeof(*hi->info.buf) * (1+nkeys));
- hi->heapnum = 0;
hi->ptr = (int *) xmalloc (sizeof(*hi->ptr) * (1+nkeys));
hi->cmp = cmp;
for (i = 0; i<= nkeys; i++)
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[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;
}
+#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 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 mode;
+ int first_in_list;
int more;
+ 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 ? */
+ {
+ 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 (level)
+ break;
+ 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;
+ /* 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->mode == 1)
+ 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->mode = 2;
+ p->first_in_list = 0;
return 1;
}
strcpy (p->prev_name, p->cur_name);
return 0;
if (*p->cur_name && strcmp (p->cur_name, p->prev_name))
{
- p->mode = 1;
+ 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;
}
-#if ZMBOL
-int heap_inpc (struct heap_info *hi)
+int heap_inpc (struct heap_cread_info *hci, struct heap_info *hi)
{
- struct heap_cread_info hci;
- ISAMC_I isamc_i = (ISAMC_I) xmalloc (sizeof(*isamc_i));
+ ISAMC_I *isamc_i = (ISAMC_I *) xmalloc (sizeof(*isamc_i));
- hci.key = (char *) xmalloc (KEY_SIZE);
- hci.mode = 1;
- hci.hi = hi;
- hci.more = heap_read_one (hi, hci.cur_name, hci.key);
+ isamc_i->clientData = hci;
+ isamc_i->read_item = heap_cread_item2;
- isamc_i->clientData = &hci;
- isamc_i->read_item = heap_cread_item;
-
- while (hci.more)
+ while (hci->more)
{
char this_name[INP_NAME_MAX];
- ISAMC_P isamc_p, isamc_p2;
+ ISAM_P isamc_p, isamc_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)))
+ 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(ISAMC_P));
- isamc_p2 = isc_merge (hi->isamc, isamc_p, isamc_i);
+ 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)
{
- no_deletions++;
- if (!dict_delete (hi->dict, this_name))
+ hi->no_deletions++;
+ if (!dict_delete (hi->reg->dict, this_name))
abort();
}
else
{
- no_updates++;
+ hi->no_updates++;
if (isamc_p2 != isamc_p)
- dict_insert (hi->dict, this_name,
- sizeof(ISAMC_P), &isamc_p2);
+ dict_insert (hi->reg->dict, this_name,
+ sizeof(ISAM_P), &isamc_p2);
}
}
else
{
- isamc_p = isc_merge (hi->isamc, 0, isamc_i);
- no_insertions++;
- dict_insert (hi->dict, this_name, sizeof(ISAMC_P), &isamc_p);
+ 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);
- xfree (hci.key);
return 0;
}
-int heap_inpd (struct heap_info *hi)
+int heap_inp0(struct heap_cread_info *hci, struct heap_info *hi)
{
- struct heap_cread_info hci;
- ISAMD_I isamd_i = (ISAMD_I) xmalloc (sizeof(*isamd_i));
+ 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;
+}
+
- hci.key = (char *) xmalloc (KEY_SIZE);
- hci.mode = 1;
- hci.hi = hi;
- hci.more = heap_read_one (hi, hci.cur_name, hci.key);
+int heap_inpb(struct heap_cread_info *hci, struct heap_info *hi)
+{
+ ISAMC_I *isamc_i = (ISAMC_I *) xmalloc (sizeof(*isamc_i));
- isamd_i->clientData = &hci;
- isamd_i->read_item = heap_cread_item;
+ isamc_i->clientData = hci;
+ isamc_i->read_item = heap_cread_item2;
- while (hci.more)
+ while (hci->more)
{
char this_name[INP_NAME_MAX];
- ISAMD_P isamd_p, isamd_p2;
+ ISAM_P isamc_p, isamc_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)))
+ strcpy (this_name, hci->cur_name);
+ assert (hci->cur_name[0]);
+ hi->no_diffs++;
+
+#if 0
+ assert(hi->zh);
+ print_dict_item(hi->zh, hci->cur_name);
+#endif
+ if ((dict_info = dict_lookup (hi->reg->dict, hci->cur_name)))
{
- memcpy (&isamd_p, dict_info+1, sizeof(ISAMD_P));
- isamd_p2 = isamd_append (hi->isamd, isamd_p, isamd_i);
- if (!isamd_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 (hi->dict, this_name))
+ hi->no_deletions++;
+ if (!dict_delete (hi->reg->dict, this_name))
abort();
}
else
{
- no_updates++;
- if (isamd_p2 != isamd_p)
- dict_insert (hi->dict, this_name,
- sizeof(ISAMD_P), &isamd_p2);
+ hi->no_updates++;
+ if (isamc_p2 != isamc_p)
+ dict_insert (hi->reg->dict, this_name,
+ sizeof(ISAM_P), &isamc_p2);
}
}
else
{
- isamd_p = isamd_append (hi->isamd, 0, isamd_i);
- no_insertions++;
- dict_insert (hi->dict, this_name, sizeof(ISAMD_P), &isamd_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);
}
}
- xfree (isamd_i);
+ xfree(isamc_i);
return 0;
}
-int heap_inp (struct heap_info *hi)
-{
- char *info;
- char next_name[INP_NAME_MAX];
- char cur_name[INP_NAME_MAX];
- int key_buf_size = INP_BUF_START;
- int key_buf_ptr;
- char *next_key;
- char *key_buf;
- int more;
-
- next_key = (char *) xmalloc (KEY_SIZE);
- key_buf = (char *) xmalloc (key_buf_size);
- more = heap_read_one (hi, cur_name, key_buf);
- while (more) /* EOF ? */
- {
- int nmemb;
- key_buf_ptr = KEY_SIZE;
- while (1)
- {
- if (!(more = heap_read_one (hi, next_name, next_key)))
- break;
- if (*next_name && strcmp (next_name, cur_name))
- break;
- memcpy (key_buf + key_buf_ptr, next_key, KEY_SIZE);
- key_buf_ptr += KEY_SIZE;
- if (key_buf_ptr+(int) KEY_SIZE >= key_buf_size)
- {
- char *new_key_buf;
- new_key_buf = (char *) 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;
- }
- }
- no_diffs++;
- nmemb = key_buf_ptr / KEY_SIZE;
- assert (nmemb * (int) KEY_SIZE == key_buf_ptr);
- if ((info = dict_lookup (hi->dict, cur_name)))
- {
- ISAM_P isam_p, isam_p2;
- memcpy (&isam_p, info+1, sizeof(ISAM_P));
- isam_p2 = is_merge (hi->isam, isam_p, nmemb, key_buf);
- if (!isam_p2)
- {
- no_deletions++;
- if (!dict_delete (hi->dict, cur_name))
- abort ();
- }
- else
- {
- no_updates++;
- if (isam_p2 != isam_p)
- dict_insert (hi->dict, cur_name, sizeof(ISAM_P), &isam_p2);
- }
- }
- else
- {
- ISAM_P isam_p;
- no_insertions++;
- isam_p = is_merge (hi->isam, 0, nmemb, key_buf);
- dict_insert (hi->dict, cur_name, sizeof(ISAM_P), &isam_p);
- }
- memcpy (key_buf, next_key, KEY_SIZE);
- strcpy (cur_name, next_name);
- }
- return 0;
-}
-
-#endif
-
-int heap_inps (struct heap_info *hi)
+int heap_inps (struct heap_cread_info *hci, 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->clientData = hci;
isams_i->read_item = heap_cread_item;
- while (hci.more)
+ while (hci->more)
{
char this_name[INP_NAME_MAX];
- ISAMS_P isams_p;
+ ISAM_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)))
+ 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->isams, isams_i);
- no_insertions++;
- dict_insert (hi->dict, this_name, sizeof(ISAMS_P), &isams_p);
+ 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
{
- logf (LOG_FATAL, "isams doesn't support this kind of update");
+ yaz_log (YLOG_FATAL, "isams doesn't support this kind of update");
break;
}
}
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 zebra_index_merge (ZebraHandle zh)
{
- 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->key_file_no;
+ 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 */
- if (nkeys < 0)
+ usefile = (nkeys!=0);
+
+ if (usefile)
{
- char fname[1024];
- nkeys = 0;
- while (1)
+ if (nkeys < 0)
{
- extract_get_fname_tmp (zh, fname, nkeys+1);
- if (access (fname, R_OK) == -1)
- break;
- nkeys++;
+ 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;
}
- if (!nkeys)
- return ;
+ 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;
}
- 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++)
+
+ if (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;
- }
- 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
+ struct heap_cread_info hci;
- 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
+ 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);
- for (i = 1; i<=nkeys; i++)
- {
- extract_get_fname_tmp (zh, rbuf, i);
- unlink (rbuf);
+ 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);
}
- 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)
+
+ if (usefile)
{
- char fname[1024];
- nkeys = 0;
- while (1)
+ for (i = 1; i<=nkeys; i++)
{
- getFnameTmp (res, fname, nkeys+1);
- if (access (fname, R_OK) == -1)
- break;
- nkeys++;
+ extract_get_fname_tmp (zh, rbuf, i);
+ unlink (rbuf);
}
- 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, res);
- kf[i]->readHandler = progressFunc;
- kf[i]->readInfo = &progressInfo;
- progressInfo.totalBytes += kf[i]->length;
- progressInfo.totalOffset += kf[i]->buf_size;
+ for (i = 1; i<=nkeys; i++)
+ key_file_destroy (kf[i]);
+ xfree (kf);
}
- 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 (hi->isams)
- heap_inps (hi);
-#if ZMBOL
- else if (hi->isamc)
- heap_inpc (hi);
- else if (hi->isam)
- heap_inp (hi);
- else if (hi->isamd)
- heap_inpd (hi);
-#endif
-
- for (i = 1; i<=nkeys; i++)
- {
- getFnameTmp (res, rbuf, i);
- unlink (rbuf);
+ 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);
}
- 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->reg->key_file_no = 0;
- /* xmalloc_trav("unfreed"); while hunting leaks */
+ key_heap_destroy (hi, nkeys);
}
+/*
+ * Local variables:
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ * vim: shiftwidth=4 tabstop=8 expandtab
+ */