-/*
- * Copyright (C) 1994-1999, Index Data
- * All rights reserved.
- * Sebastian Hammer, Adam Dickmeiss
- *
- * $Log: recindex.c,v $
- * Revision 1.31 2001-02-26 22:14:59 adam
- * Updated for BZIP2 1.0.X. Configure script doesn't enable 64 bit LFS
- * on broken glibc on Redhat 7.0.
- *
- * Revision 1.30 2000/07/13 10:14:20 heikki
- * Removed compiler warnings when making zebra
- *
- * Revision 1.29 2000/04/05 09:49:35 adam
- * On Unix, zebra/z'mbol uses automake.
- *
- * Revision 1.28 1999/12/08 22:44:45 adam
- * Zebra/Z'mbol dependencies added.
- *
- * Revision 1.27 1999/10/29 10:02:33 adam
- * Fixed decompression buffer overflow.
- *
- * Revision 1.26 1999/07/06 13:34:57 adam
- * Fixed bug (introduced by previous commit).
- *
- * Revision 1.25 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 1.24 1999/06/25 13:48:02 adam
- * Updated MSVC project files.
- * Added BZIP2 record compression (not very well tested).
- *
- * Revision 1.23 1999/05/26 07:49:13 adam
- * C++ compilation.
- *
- * Revision 1.22 1999/02/18 12:49:34 adam
- * Changed file naming scheme for register files as well as record
- * store/index files.
- *
- * Revision 1.21 1999/02/02 14:51:03 adam
- * Updated WIN32 code specific sections. Changed header.
- *
- * Revision 1.20 1998/01/12 15:04:08 adam
- * The test option (-s) only uses read-lock (and not write lock).
- *
- * Revision 1.19 1997/09/17 12:19:16 adam
- * Zebra version corresponds to YAZ version 1.4.
- * Changed Zebra server so that it doesn't depend on global common_resource.
- *
- * Revision 1.18 1997/07/15 16:28:42 adam
- * Bug fix: storeData didn't work with files with multiple records.
- * Bug fix: fixed memory management with records; not really well
- * thought through.
- *
- * Revision 1.17 1997/02/12 20:39:46 adam
- * Implemented options -f <n> that limits the log to the first <n>
- * records.
- * Changed some log messages also.
- *
- * Revision 1.16 1996/06/04 10:19:00 adam
- * Minor changes - removed include of ctype.h.
- *
- * Revision 1.15 1996/05/13 14:23:06 adam
- * Work on compaction of set/use bytes in dictionary.
- *
- * Revision 1.14 1996/02/01 20:48:15 adam
- * The total size of records are always checked in rec_cache_insert to
- * reduce memory usage.
- *
- * Revision 1.13 1995/12/11 09:12:49 adam
- * The rec_get function returns NULL if record doesn't exist - will
- * happen in the server if the result set records have been deleted since
- * the creation of the set (i.e. the search).
- * The server saves a result temporarily if it is 'volatile', i.e. the
- * set is register dependent.
- *
- * Revision 1.12 1995/12/07 17:38:47 adam
- * Work locking mechanisms for concurrent updates/commit.
- *
- * Revision 1.11 1995/12/06 13:58:26 adam
- * Improved flushing of records - all flushes except the last one
- * don't write the last accessed. Also flush takes place if record
- * info occupy more than about 256k.
- *
- * Revision 1.10 1995/12/06 12:41:24 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 1995/11/30 08:34:33 adam
- * Started work on commit facility.
- * Changed a few malloc/free to xmalloc/xfree.
- *
- * Revision 1.8 1995/11/28 14:26:21 adam
- * Bug fix: recordId with constant wasn't right.
- * Bug fix: recordId dictionary entry wasn't deleted when needed.
- *
- * Revision 1.7 1995/11/28 09:09:43 adam
- * Zebra config renamed.
- * Use setting 'recordId' to identify record now.
- * Bug fix in recindex.c: rec_release_blocks was invokeded even
- * though the blocks were already released.
- * File traversal properly deletes records when needed.
- *
- * Revision 1.6 1995/11/25 10:24:06 adam
- * More record fields - they are enumerated now.
- * New options: flagStoreData flagStoreKey.
- *
- * Revision 1.5 1995/11/22 17:19:18 adam
- * Record management uses the bfile system.
- *
- * Revision 1.4 1995/11/20 16:59:46 adam
- * New update method: the 'old' keys are saved for each records.
- *
- * Revision 1.3 1995/11/16 15:34:55 adam
- * Uses new record management system in both indexer and server.
- *
- * Revision 1.2 1995/11/15 19:13:08 adam
- * Work on record management.
- *
- * Revision 1.1 1995/11/15 14:46:20 adam
- * Started work on better record management system.
- *
- */
+/* $Id: recindex.c,v 1.50 2006-06-07 10:14:41 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 Zebra; see the file LICENSE.zebra. If not, write to the
+Free Software Foundation, 59 Temple Place - Suite 330, Boston, MA
+02111-1307, USA.
+*/
+
+#define RIDX_CHUNK 128
/*
* Format of first block
- * next (4 bytes)
- * ref_count (4 bytes)
- * block (504 bytes)
+ * next (8 bytes)
+ * ref_count (2 bytes)
+ * block (500 bytes)
*
* Format of subsequent blocks
- * next (4 bytes)
- * block (508 bytes)
+ * next (8 bytes)
+ * block (502 bytes)
*
* Format of each record
* sysno
* length = 0 if same as previous
*/
#include <stdio.h>
+#include <stdlib.h>
#include <assert.h>
#include <string.h>
+#include <yaz/yaz-util.h>
#include "recindxp.h"
#if HAVE_BZLIB_H
#include <bzlib.h>
#endif
-static void rec_write_head (Records p)
+
+/* Modify argument to if below: 1=normal, 0=sysno testing */
+#if 1
+/* If this is used sysno are not converted (no testing) */
+#define FAKE_OFFSET 0
+#define USUAL_RANGE 6000000000LL
+
+#else
+/* Use a fake > 2^32 offset so we can test for proper 64-bit handling */
+#define FAKE_OFFSET 6000000000LL
+#define USUAL_RANGE 2000000000LL
+#endif
+
+static SYSNO rec_sysno_to_ext(SYSNO sysno)
+{
+ assert(sysno >= 0 && sysno <= USUAL_RANGE);
+ return sysno + FAKE_OFFSET;
+}
+
+SYSNO rec_sysno_to_int(SYSNO sysno)
+{
+ assert(sysno >= FAKE_OFFSET && sysno <= FAKE_OFFSET + USUAL_RANGE);
+ return sysno - FAKE_OFFSET;
+}
+
+static ZEBRA_RES rec_write_head(Records p)
{
int r;
- assert (p);
- assert (p->index_BFile);
+ assert(p);
+ assert(p->index_BFile);
- r = bf_write (p->index_BFile, 0, 0, sizeof(p->head), &p->head);
+ r = bf_write(p->index_BFile, 0, 0, sizeof(p->head), &p->head);
if (r)
{
- logf (LOG_FATAL|LOG_ERRNO, "write head of %s", p->index_fname);
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "write head of %s", p->index_fname);
+ return ZEBRA_FAIL;
}
+ return ZEBRA_OK;
}
-static void rec_tmp_expand (Records p, int size)
+static void rec_tmp_expand(Records p, int size)
{
if (p->tmp_size < size + 2048 ||
p->tmp_size < p->head.block_size[REC_BLOCK_TYPES-1]*2)
{
- xfree (p->tmp_buf);
- p->tmp_size = size + p->head.block_size[REC_BLOCK_TYPES-1]*2 + 2048;
- p->tmp_buf = (char *) xmalloc (p->tmp_size);
+ xfree(p->tmp_buf);
+ p->tmp_size = size + (int)
+ (p->head.block_size[REC_BLOCK_TYPES-1])*2 + 2048;
+ p->tmp_buf = (char *) xmalloc(p->tmp_size);
}
}
-static int read_indx (Records p, int sysno, void *buf, int itemsize,
+static int read_indx(Records p, SYSNO sysno, void *buf, int itemsize,
int ignoreError)
{
int r;
- int pos = (sysno-1)*itemsize;
+ zint pos = (sysno-1)*itemsize;
+ int off = CAST_ZINT_TO_INT(pos%RIDX_CHUNK);
+ int sz1 = RIDX_CHUNK - off; /* sz1 is size of buffer to read.. */
+
+ if (sz1 > itemsize)
+ sz1 = itemsize; /* no more than itemsize bytes */
- r = bf_read (p->index_BFile, 1+pos/128, pos%128, itemsize, buf);
+ r = bf_read(p->index_BFile, 1+pos/RIDX_CHUNK, off, sz1, buf);
+ if (r == 1 && sz1 < itemsize) /* boundary? - must read second part */
+ r = bf_read(p->index_BFile, 2+pos/RIDX_CHUNK, 0, itemsize - sz1,
+ (char*) buf + sz1);
if (r != 1 && !ignoreError)
{
- logf (LOG_FATAL|LOG_ERRNO, "read in %s at pos %ld",
- p->index_fname, (long) pos);
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "read in %s at pos %ld",
+ p->index_fname, (long) pos);
}
return r;
}
-static void write_indx (Records p, int sysno, void *buf, int itemsize)
+static void write_indx(Records p, SYSNO sysno, void *buf, int itemsize)
{
- int pos = (sysno-1)*itemsize;
+ zint pos = (sysno-1)*itemsize;
+ int off = CAST_ZINT_TO_INT(pos%RIDX_CHUNK);
+ int sz1 = RIDX_CHUNK - off; /* sz1 is size of buffer to read.. */
+
+ if (sz1 > itemsize)
+ sz1 = itemsize; /* no more than itemsize bytes */
- bf_write (p->index_BFile, 1+pos/128, pos%128, itemsize, buf);
+ bf_write(p->index_BFile, 1+pos/RIDX_CHUNK, off, sz1, buf);
+ if (sz1 < itemsize) /* boundary? must write second part */
+ bf_write(p->index_BFile, 2+pos/RIDX_CHUNK, 0, itemsize - sz1,
+ (char*) buf + sz1);
}
-static void rec_release_blocks (Records p, int sysno)
+static ZEBRA_RES rec_release_blocks(Records p, SYSNO sysno)
{
struct record_index_entry entry;
- int freeblock;
- char block_and_ref[sizeof(short) + sizeof(int)];
+ zint freeblock;
+ char block_and_ref[sizeof(zint) + sizeof(short)];
int dst_type;
int first = 1;
- if (read_indx (p, sysno, &entry, sizeof(entry), 1) != 1)
- return ;
+ if (read_indx(p, sysno, &entry, sizeof(entry), 1) != 1)
+ return ZEBRA_FAIL;
freeblock = entry.next;
- assert (freeblock > 0);
- dst_type = freeblock & 7;
- assert (dst_type < REC_BLOCK_TYPES);
+ assert(freeblock > 0);
+ dst_type = CAST_ZINT_TO_INT(freeblock & 7);
+ assert(dst_type < REC_BLOCK_TYPES);
freeblock = freeblock / 8;
while (freeblock)
{
- if (bf_read (p->data_BFile[dst_type], freeblock, 0,
- sizeof(block_and_ref), block_and_ref) != 1)
+ if (bf_read(p->data_BFile[dst_type], freeblock, 0,
+ first ? sizeof(block_and_ref) : sizeof(zint),
+ block_and_ref) != 1)
{
- logf (LOG_FATAL|LOG_ERRNO, "read in rec_del_single");
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "read in rec_del_single");
+ return ZEBRA_FAIL;
}
if (first)
{
short ref;
- memcpy (&ref, block_and_ref + sizeof(int), sizeof(ref));
+ memcpy(&ref, block_and_ref + sizeof(freeblock), sizeof(ref));
--ref;
- memcpy (block_and_ref + sizeof(int), &ref, sizeof(ref));
+ memcpy(block_and_ref + sizeof(freeblock), &ref, sizeof(ref));
if (ref)
{
- if (bf_write (p->data_BFile[dst_type], freeblock, 0,
+ if (bf_write(p->data_BFile[dst_type], freeblock, 0,
sizeof(block_and_ref), block_and_ref))
{
- logf (LOG_FATAL|LOG_ERRNO, "write in rec_del_single");
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "write in rec_del_single");
+ return ZEBRA_FAIL;
}
- return;
+ return ZEBRA_OK;
}
first = 0;
}
- if (bf_write (p->data_BFile[dst_type], freeblock, 0, sizeof(freeblock),
+ if (bf_write(p->data_BFile[dst_type], freeblock, 0, sizeof(freeblock),
&p->head.block_free[dst_type]))
{
- logf (LOG_FATAL|LOG_ERRNO, "write in rec_del_single");
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "write in rec_del_single");
+ return ZEBRA_FAIL;
}
p->head.block_free[dst_type] = freeblock;
- memcpy (&freeblock, block_and_ref, sizeof(int));
+ memcpy(&freeblock, block_and_ref, sizeof(freeblock));
p->head.block_used[dst_type]--;
}
p->head.total_bytes -= entry.size;
+ return ZEBRA_OK;
}
-static void rec_delete_single (Records p, Record rec)
+static ZEBRA_RES rec_delete_single(Records p, Record rec)
{
struct record_index_entry entry;
- rec_release_blocks (p, rec->sysno);
+ if (rec_release_blocks(p, rec_sysno_to_int(rec->sysno)) != ZEBRA_OK)
+ return ZEBRA_FAIL;
entry.next = p->head.index_free;
entry.size = 0;
- p->head.index_free = rec->sysno;
- write_indx (p, rec->sysno, &entry, sizeof(entry));
+ p->head.index_free = rec_sysno_to_int(rec->sysno);
+ write_indx(p, rec_sysno_to_int(rec->sysno), &entry, sizeof(entry));
+ return ZEBRA_OK;
}
-static void rec_write_tmp_buf (Records p, int size, int *sysnos)
+static ZEBRA_RES rec_write_tmp_buf(Records p, int size, SYSNO *sysnos)
{
struct record_index_entry entry;
int no_written = 0;
char *cptr = p->tmp_buf;
- int block_prev = -1, block_free;
+ zint block_prev = -1, block_free;
int dst_type = 0;
int i;
block_free = p->head.block_free[dst_type];
if (block_free)
{
- if (bf_read (p->data_BFile[dst_type],
+ if (bf_read(p->data_BFile[dst_type],
block_free, 0, sizeof(*p->head.block_free),
&p->head.block_free[dst_type]) != 1)
{
- logf (LOG_FATAL|LOG_ERRNO, "read in %s at free block %d",
- p->data_fname[dst_type], block_free);
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "read in %s at free block "
+ ZINT_FORMAT,
+ p->data_fname[dst_type], block_free);
+ return ZEBRA_FAIL;
}
}
else
p->head.total_bytes += size;
while (*sysnos > 0)
{
- write_indx (p, *sysnos, &entry, sizeof(entry));
+ write_indx(p, *sysnos, &entry, sizeof(entry));
sysnos++;
}
}
else
{
- memcpy (cptr, &block_free, sizeof(int));
- bf_write (p->data_BFile[dst_type], block_prev, 0, 0, cptr);
+ memcpy(cptr, &block_free, sizeof(block_free));
+ bf_write(p->data_BFile[dst_type], block_prev, 0, 0, cptr);
cptr = p->tmp_buf + no_written;
}
block_prev = block_free;
- no_written += p->head.block_size[dst_type] - sizeof(int);
+ no_written += CAST_ZINT_TO_INT(p->head.block_size[dst_type])
+ - sizeof(zint);
p->head.block_used[dst_type]++;
}
- assert (block_prev != -1);
+ assert(block_prev != -1);
block_free = 0;
- memcpy (cptr, &block_free, sizeof(int));
- bf_write (p->data_BFile[dst_type], block_prev, 0,
- sizeof(int) + (p->tmp_buf+size) - cptr, cptr);
+ memcpy(cptr, &block_free, sizeof(block_free));
+ bf_write(p->data_BFile[dst_type], block_prev, 0,
+ sizeof(block_free) + (p->tmp_buf+size) - cptr, cptr);
+ return ZEBRA_OK;
}
-Records rec_open (BFiles bfs, int rw, int compression_method)
+Records rec_open(BFiles bfs, int rw, int compression_method)
{
Records p;
int i, r;
int version;
+ ZEBRA_RES ret = ZEBRA_OK;
- p = (Records) xmalloc (sizeof(*p));
+ p = (Records) xmalloc(sizeof(*p));
p->compression_method = compression_method;
p->rw = rw;
p->tmp_size = 1024;
- p->tmp_buf = (char *) xmalloc (p->tmp_size);
p->index_fname = "reci";
- p->index_BFile = bf_open (bfs, p->index_fname, 128, rw);
+ p->index_BFile = bf_open(bfs, p->index_fname, RIDX_CHUNK, rw);
if (p->index_BFile == NULL)
{
- logf (LOG_FATAL|LOG_ERRNO, "open %s", p->index_fname);
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "open %s", p->index_fname);
+ xfree(p);
+ return 0;
}
- r = bf_read (p->index_BFile, 0, 0, 0, p->tmp_buf);
+ p->tmp_buf = (char *) xmalloc(p->tmp_size);
+ r = bf_read(p->index_BFile, 0, 0, 0, p->tmp_buf);
switch (r)
{
case 0:
- memcpy (p->head.magic, REC_HEAD_MAGIC, sizeof(p->head.magic));
- sprintf (p->head.version, "%3d", REC_VERSION);
+ memcpy(p->head.magic, REC_HEAD_MAGIC, sizeof(p->head.magic));
+ sprintf(p->head.version, "%3d", REC_VERSION);
p->head.index_free = 0;
p->head.index_last = 1;
p->head.no_records = 0;
p->head.block_move[i] = p->head.block_size[i] * 24;
}
if (rw)
- rec_write_head (p);
+ {
+ if (rec_write_head(p) != ZEBRA_OK)
+ ret = ZEBRA_FAIL;
+ }
break;
case 1:
- memcpy (&p->head, p->tmp_buf, sizeof(p->head));
- if (memcmp (p->head.magic, REC_HEAD_MAGIC, sizeof(p->head.magic)))
+ memcpy(&p->head, p->tmp_buf, sizeof(p->head));
+ if (memcmp(p->head.magic, REC_HEAD_MAGIC, sizeof(p->head.magic)))
{
- logf (LOG_FATAL, "file %s has bad format", p->index_fname);
- exit (1);
+ yaz_log(YLOG_FATAL, "file %s has bad format", p->index_fname);
+ ret = ZEBRA_FAIL;
}
- version = atoi (p->head.version);
+ version = atoi(p->head.version);
if (version != REC_VERSION)
{
- logf (LOG_FATAL, "file %s is version %d, but version"
+ yaz_log(YLOG_FATAL, "file %s is version %d, but version"
" %d is required", p->index_fname, version, REC_VERSION);
- exit (1);
+ ret = ZEBRA_FAIL;
}
break;
}
for (i = 0; i<REC_BLOCK_TYPES; i++)
{
char str[80];
- sprintf (str, "recd%c", i + 'A');
- p->data_fname[i] = (char *) xmalloc (strlen(str)+1);
- strcpy (p->data_fname[i], str);
+ sprintf(str, "recd%c", i + 'A');
+ p->data_fname[i] = (char *) xmalloc(strlen(str)+1);
+ strcpy(p->data_fname[i], str);
p->data_BFile[i] = NULL;
}
for (i = 0; i<REC_BLOCK_TYPES; i++)
{
- if (!(p->data_BFile[i] = bf_open (bfs, p->data_fname[i],
- p->head.block_size[i],
- rw)))
+ if (!(p->data_BFile[i] =
+ bf_open(bfs, p->data_fname[i],
+ CAST_ZINT_TO_INT(p->head.block_size[i]), rw)))
{
- logf (LOG_FATAL|LOG_ERRNO, "bf_open %s", p->data_fname[i]);
- exit (1);
+ yaz_log(YLOG_FATAL|YLOG_ERRNO, "bf_open %s", p->data_fname[i]);
+ ret = ZEBRA_FAIL;
}
}
p->cache_max = 400;
p->cache_cur = 0;
p->record_cache = (struct record_cache_entry *)
- xmalloc (sizeof(*p->record_cache)*p->cache_max);
- zebra_mutex_init (&p->mutex);
+ xmalloc(sizeof(*p->record_cache)*p->cache_max);
+ zebra_mutex_init(&p->mutex);
+ if (ret == ZEBRA_FAIL)
+ rec_close(&p);
return p;
}
-static void rec_encode_unsigned (unsigned n, unsigned char *buf, int *len)
+static void rec_encode_unsigned(unsigned n, unsigned char *buf, int *len)
{
(*len) = 0;
while (n > 127)
*np = n;
}
-static void rec_cache_flush_block1 (Records p, Record rec, Record last_rec,
- char **out_buf, int *out_size,
- int *out_offset)
+static void rec_encode_zint(zint n, unsigned char *buf, int *len)
+{
+ (*len) = 0;
+ while (n > 127)
+ {
+ buf[*len] = (unsigned) (128 + (n & 127));
+ n = n >> 7;
+ (*len)++;
+ }
+ buf[*len] = (unsigned) n;
+ (*len)++;
+}
+
+static void rec_decode_zint(zint *np, unsigned char *buf, int *len)
+{
+ zint n = 0;
+ zint w = 1;
+ (*len) = 0;
+
+ while (buf[*len] > 127)
+ {
+ n += w*(buf[*len] & 127);
+ w = w << 7;
+ (*len)++;
+ }
+ n += w * buf[*len];
+ (*len)++;
+ *np = n;
+}
+
+static void rec_cache_flush_block1(Records p, Record rec, Record last_rec,
+ char **out_buf, int *out_size,
+ int *out_offset)
{
int i;
int len;
for (i = 0; i<REC_NO_INFO; i++)
{
- if (*out_offset + (int) rec->size[i] + 20 > *out_size)
+ if (*out_offset + CAST_ZINT_TO_INT(rec->size[i]) + 20 > *out_size)
{
int new_size = *out_offset + rec->size[i] + 65536;
- char *np = (char *) xmalloc (new_size);
+ char *np = (char *) xmalloc(new_size);
if (*out_offset)
- memcpy (np, *out_buf, *out_offset);
- xfree (*out_buf);
+ memcpy(np, *out_buf, *out_offset);
+ xfree(*out_buf);
*out_size = new_size;
*out_buf = np;
}
if (i == 0)
{
- rec_encode_unsigned (rec->sysno, *out_buf + *out_offset, &len);
+ rec_encode_zint(rec_sysno_to_int(rec->sysno),
+ (unsigned char *) *out_buf + *out_offset, &len);
(*out_offset) += len;
}
if (rec->size[i] == 0)
{
- rec_encode_unsigned (1, *out_buf + *out_offset, &len);
+ rec_encode_unsigned(1, (unsigned char *) *out_buf + *out_offset,
+ &len);
(*out_offset) += len;
}
else if (last_rec && rec->size[i] == last_rec->size[i] &&
- !memcmp (rec->info[i], last_rec->info[i], rec->size[i]))
+ !memcmp(rec->info[i], last_rec->info[i], rec->size[i]))
{
- rec_encode_unsigned (0, *out_buf + *out_offset, &len);
+ rec_encode_unsigned(0, (unsigned char *) *out_buf + *out_offset,
+ &len);
(*out_offset) += len;
}
else
{
- rec_encode_unsigned (rec->size[i]+1, *out_buf + *out_offset, &len);
+ rec_encode_unsigned(rec->size[i]+1,
+ (unsigned char *) *out_buf + *out_offset,
+ &len);
(*out_offset) += len;
- memcpy (*out_buf + *out_offset, rec->info[i], rec->size[i]);
+ memcpy(*out_buf + *out_offset, rec->info[i], rec->size[i]);
(*out_offset) += rec->size[i];
}
}
}
-static void rec_write_multiple (Records p, int saveCount)
+static ZEBRA_RES rec_write_multiple(Records p, int saveCount)
{
int i;
short ref_count = 0;
Record last_rec = 0;
int out_size = 1000;
int out_offset = 0;
- char *out_buf = (char *) xmalloc (out_size);
- int *sysnos = (int *) xmalloc (sizeof(*sysnos) * (p->cache_cur + 1));
- int *sysnop = sysnos;
+ char *out_buf = (char *) xmalloc(out_size);
+ SYSNO *sysnos = (SYSNO *) xmalloc(sizeof(*sysnos) * (p->cache_cur + 1));
+ SYSNO *sysnop = sysnos;
+ ZEBRA_RES ret = ZEBRA_OK;
for (i = 0; i<p->cache_cur - saveCount; i++)
{
switch (e->flag)
{
case recordFlagNew:
- rec_cache_flush_block1 (p, e->rec, last_rec, &out_buf,
+ rec_cache_flush_block1(p, e->rec, last_rec, &out_buf,
&out_size, &out_offset);
- *sysnop++ = e->rec->sysno;
+ *sysnop++ = rec_sysno_to_int(e->rec->sysno);
ref_count++;
e->flag = recordFlagNop;
last_rec = e->rec;
break;
case recordFlagWrite:
- rec_release_blocks (p, e->rec->sysno);
- rec_cache_flush_block1 (p, e->rec, last_rec, &out_buf,
+ if (rec_release_blocks(p, rec_sysno_to_int(e->rec->sysno))
+ != ZEBRA_OK)
+ ret = ZEBRA_FAIL;
+
+ rec_cache_flush_block1(p, e->rec, last_rec, &out_buf,
&out_size, &out_offset);
- *sysnop++ = e->rec->sysno;
+ *sysnop++ = rec_sysno_to_int(e->rec->sysno);
ref_count++;
e->flag = recordFlagNop;
last_rec = e->rec;
break;
case recordFlagDelete:
- rec_delete_single (p, e->rec);
+ if (rec_delete_single(p, e->rec) != ZEBRA_OK)
+ ret = ZEBRA_FAIL;
+
e->flag = recordFlagNop;
break;
default:
*sysnop = -1;
if (ref_count)
{
- int csize = 0; /* indicate compression "not performed yet" */
+ unsigned int csize = 0; /* indicate compression "not performed yet" */
compression_method = p->compression_method;
switch (compression_method)
{
case REC_COMPRESS_BZIP2:
#if HAVE_BZLIB_H
csize = out_offset + (out_offset >> 6) + 620;
- rec_tmp_expand (p, csize);
+ rec_tmp_expand(p, csize);
#ifdef BZ_CONFIG_ERROR
i = BZ2_bzBuffToBuffCompress
#else
i = bzBuffToBuffCompress
#endif
- (p->tmp_buf+sizeof(int)+sizeof(short)+
+ (p->tmp_buf+sizeof(zint)+sizeof(short)+
sizeof(char),
&csize, out_buf, out_offset, 1, 0, 30);
if (i != BZ_OK)
{
- logf (LOG_WARN, "bzBuffToBuffCompress error code=%d", i);
+ yaz_log(YLOG_WARN, "bzBuffToBuffCompress error code=%d", i);
csize = 0;
}
- logf (LOG_LOG, "compress %4d %5d %5d", ref_count, out_offset,
+ yaz_log(YLOG_LOG, "compress %4d %5d %5d", ref_count, out_offset,
csize);
#endif
break;
{
/* either no compression or compression not supported ... */
csize = out_offset;
- rec_tmp_expand (p, csize);
- memcpy (p->tmp_buf + sizeof(int) + sizeof(short) + sizeof(char),
+ rec_tmp_expand(p, csize);
+ memcpy(p->tmp_buf + sizeof(zint) + sizeof(short) + sizeof(char),
out_buf, out_offset);
csize = out_offset;
compression_method = REC_COMPRESS_NONE;
}
- memcpy (p->tmp_buf + sizeof(int), &ref_count, sizeof(ref_count));
- memcpy (p->tmp_buf + sizeof(int)+sizeof(short),
+ memcpy(p->tmp_buf + sizeof(zint), &ref_count, sizeof(ref_count));
+ memcpy(p->tmp_buf + sizeof(zint)+sizeof(short),
&compression_method, sizeof(compression_method));
/* -------- compression */
- rec_write_tmp_buf (p, csize + sizeof(short) + sizeof(char), sysnos);
+ if (rec_write_tmp_buf(p, csize + sizeof(short) + sizeof(char), sysnos)
+ != ZEBRA_OK)
+ ret = ZEBRA_FAIL;
}
- xfree (out_buf);
- xfree (sysnos);
+ xfree(out_buf);
+ xfree(sysnos);
+ return ret;
}
-static void rec_cache_flush (Records p, int saveCount)
+static ZEBRA_RES rec_cache_flush(Records p, int saveCount)
{
int i, j;
+ ZEBRA_RES ret;
if (saveCount >= p->cache_cur)
saveCount = 0;
- rec_write_multiple (p, saveCount);
+ ret = rec_write_multiple(p, saveCount);
for (i = 0; i<p->cache_cur - saveCount; i++)
{
struct record_cache_entry *e = p->record_cache + i;
- rec_rm (&e->rec);
+ rec_rm(&e->rec);
}
/* i still being used ... */
for (j = 0; j<saveCount; j++, i++)
- memcpy (p->record_cache+j, p->record_cache+i,
+ memcpy(p->record_cache+j, p->record_cache+i,
sizeof(*p->record_cache));
p->cache_cur = saveCount;
+ return ret;
}
-static Record *rec_cache_lookup (Records p, int sysno,
- enum recordCacheFlag flag)
+static Record *rec_cache_lookup(Records p, SYSNO sysno,
+ enum recordCacheFlag flag)
{
int i;
for (i = 0; i<p->cache_cur; i++)
return NULL;
}
-static void rec_cache_insert (Records p, Record rec, enum recordCacheFlag flag)
+static ZEBRA_RES rec_cache_insert(Records p, Record rec, enum recordCacheFlag flag)
{
struct record_cache_entry *e;
+ ZEBRA_RES ret = ZEBRA_OK;
if (p->cache_cur == p->cache_max)
- rec_cache_flush (p, 1);
+ ret = rec_cache_flush(p, 1);
else if (p->cache_cur > 0)
{
int i, j;
used += r->size[j];
}
if (used > 90000)
- rec_cache_flush (p, 1);
+ ret = rec_cache_flush(p, 1);
}
- assert (p->cache_cur < p->cache_max);
+ assert(p->cache_cur < p->cache_max);
e = p->record_cache + (p->cache_cur)++;
e->flag = flag;
- e->rec = rec_cp (rec);
+ e->rec = rec_cp(rec);
+ return ret;
}
-void rec_close (Records *pp)
+ZEBRA_RES rec_close(Records *pp)
{
Records p = *pp;
int i;
+ ZEBRA_RES ret = ZEBRA_OK;
- assert (p);
+ if (!p)
+ return ret;
- zebra_mutex_destroy (&p->mutex);
- rec_cache_flush (p, 0);
- xfree (p->record_cache);
+ zebra_mutex_destroy(&p->mutex);
+ if (rec_cache_flush(p, 0) != ZEBRA_OK)
+ ret = ZEBRA_FAIL;
+
+ xfree(p->record_cache);
if (p->rw)
- rec_write_head (p);
+ {
+ if (rec_write_head(p) != ZEBRA_OK)
+ ret = ZEBRA_FAIL;
+ }
if (p->index_BFile)
- bf_close (p->index_BFile);
+ bf_close(p->index_BFile);
for (i = 0; i<REC_BLOCK_TYPES; i++)
{
if (p->data_BFile[i])
- bf_close (p->data_BFile[i]);
- xfree (p->data_fname[i]);
+ bf_close(p->data_BFile[i]);
+ xfree(p->data_fname[i]);
}
- xfree (p->tmp_buf);
- xfree (p);
+ xfree(p->tmp_buf);
+ xfree(p);
*pp = NULL;
+ return ret;
}
-static Record rec_get_int (Records p, int sysno)
+static Record rec_get_int(Records p, SYSNO sysno)
{
int i, in_size, r;
Record rec, *recp;
struct record_index_entry entry;
- int freeblock, dst_type;
+ zint freeblock;
+ int dst_type;
char *nptr, *cptr;
char *in_buf = 0;
char *bz_buf = 0;
#if HAVE_BZLIB_H
- int bz_size;
+ unsigned int bz_size;
#endif
char compression_method;
- assert (sysno > 0);
- assert (p);
+ assert(sysno > 0);
+ assert(p);
- if ((recp = rec_cache_lookup (p, sysno, recordFlagNop)))
- return rec_cp (*recp);
+ if ((recp = rec_cache_lookup(p, sysno, recordFlagNop)))
+ return rec_cp(*recp);
- if (read_indx (p, sysno, &entry, sizeof(entry), 1) < 1)
+ if (read_indx(p, rec_sysno_to_int(sysno), &entry, sizeof(entry), 1) < 1)
return NULL; /* record is not there! */
if (!entry.size)
return NULL; /* record is deleted */
- dst_type = entry.next & 7;
- assert (dst_type < REC_BLOCK_TYPES);
+ dst_type = (int) (entry.next & 7);
+ assert(dst_type < REC_BLOCK_TYPES);
freeblock = entry.next / 8;
- assert (freeblock > 0);
+ assert(freeblock > 0);
- rec_tmp_expand (p, entry.size);
+ rec_tmp_expand(p, entry.size);
cptr = p->tmp_buf;
- r = bf_read (p->data_BFile[dst_type], freeblock, 0, 0, cptr);
+ r = bf_read(p->data_BFile[dst_type], freeblock, 0, 0, cptr);
if (r < 0)
return 0;
- memcpy (&freeblock, cptr, sizeof(freeblock));
+ memcpy(&freeblock, cptr, sizeof(freeblock));
while (freeblock)
{
- int tmp;
+ zint tmp;
cptr += p->head.block_size[dst_type] - sizeof(freeblock);
- memcpy (&tmp, cptr, sizeof(tmp));
- r = bf_read (p->data_BFile[dst_type], freeblock, 0, 0, cptr);
+ memcpy(&tmp, cptr, sizeof(tmp));
+ r = bf_read(p->data_BFile[dst_type], freeblock, 0, 0, cptr);
if (r < 0)
return 0;
- memcpy (&freeblock, cptr, sizeof(freeblock));
- memcpy (cptr, &tmp, sizeof(tmp));
+ memcpy(&freeblock, cptr, sizeof(freeblock));
+ memcpy(cptr, &tmp, sizeof(tmp));
}
- rec = (Record) xmalloc (sizeof(*rec));
+ rec = (Record) xmalloc(sizeof(*rec));
rec->sysno = sysno;
- memcpy (&compression_method, p->tmp_buf + sizeof(int) + sizeof(short),
+ memcpy(&compression_method, p->tmp_buf + sizeof(zint) + sizeof(short),
sizeof(compression_method));
- in_buf = p->tmp_buf + sizeof(int) + sizeof(short) + sizeof(char);
+ in_buf = p->tmp_buf + sizeof(zint) + sizeof(short) + sizeof(char);
in_size = entry.size - sizeof(short) - sizeof(char);
switch (compression_method)
{
bz_size = entry.size * 20 + 100;
while (1)
{
- bz_buf = (char *) xmalloc (bz_size);
+ bz_buf = (char *) xmalloc(bz_size);
#ifdef BZ_CONFIG_ERROR
i = BZ2_bzBuffToBuffDecompress
#else
i = bzBuffToBuffDecompress
#endif
- (bz_buf, &bz_size, in_buf, in_size, 0, 0);
- logf (LOG_LOG, "decompress %5d %5d", in_size, bz_size);
+ (bz_buf, &bz_size, in_buf, in_size, 0, 0);
+ yaz_log(YLOG_LOG, "decompress %5d %5d", in_size, bz_size);
if (i == BZ_OK)
break;
- logf (LOG_LOG, "failed");
- xfree (bz_buf);
+ yaz_log(YLOG_LOG, "failed");
+ xfree(bz_buf);
bz_size *= 2;
}
in_buf = bz_buf;
in_size = bz_size;
#else
- logf (LOG_FATAL, "cannot decompress record(s) in BZIP2 format");
- exit (1);
+ yaz_log(YLOG_FATAL, "cannot decompress record(s) in BZIP2 format");
+ return 0;
#endif
break;
case REC_COMPRESS_NONE:
nptr = in_buf; /* skip ref count */
while (nptr < in_buf + in_size)
{
- int this_sysno;
+ zint this_sysno;
int len;
- rec_decode_unsigned (&this_sysno, nptr, &len);
+ rec_decode_zint(&this_sysno, (unsigned char *) nptr, &len);
nptr += len;
for (i = 0; i < REC_NO_INFO; i++)
{
- int this_size;
- rec_decode_unsigned (&this_size, nptr, &len);
+ unsigned int this_size;
+ rec_decode_unsigned(&this_size, (unsigned char *) nptr, &len);
nptr += len;
if (this_size == 0)
else
rec->info[i] = NULL;
}
- if (this_sysno == sysno)
+ if (this_sysno == rec_sysno_to_int(sysno))
break;
}
for (i = 0; i<REC_NO_INFO; i++)
{
if (rec->info[i] && rec->size[i])
{
- char *np = xmalloc (rec->size[i]);
- memcpy (np, rec->info[i], rec->size[i]);
+ char *np = xmalloc(rec->size[i]+1);
+ memcpy(np, rec->info[i], rec->size[i]);
+ np[rec->size[i]] = '\0';
rec->info[i] = np;
}
else
{
- assert (rec->info[i] == 0);
- assert (rec->size[i] == 0);
+ assert(rec->info[i] == 0);
+ assert(rec->size[i] == 0);
}
}
- xfree (bz_buf);
- rec_cache_insert (p, rec, recordFlagNop);
+ xfree(bz_buf);
+ if (rec_cache_insert(p, rec, recordFlagNop) != ZEBRA_OK)
+ return 0;
return rec;
}
-Record rec_get (Records p, int sysno)
+Record rec_get(Records p, SYSNO sysno)
{
Record rec;
- zebra_mutex_lock (&p->mutex);
+ zebra_mutex_lock(&p->mutex);
- rec = rec_get_int (p, sysno);
- zebra_mutex_unlock (&p->mutex);
+ rec = rec_get_int(p, sysno);
+ zebra_mutex_unlock(&p->mutex);
return rec;
}
-static Record rec_new_int (Records p)
+Record rec_get_root(Records p)
+{
+ return rec_get(p, rec_sysno_to_ext(1));
+}
+
+static Record rec_new_int(Records p)
{
- int sysno, i;
+ int i;
+ SYSNO sysno;
Record rec;
- assert (p);
- rec = (Record) xmalloc (sizeof(*rec));
+ assert(p);
+ rec = (Record) xmalloc(sizeof(*rec));
if (1 || p->head.index_free == 0)
sysno = (p->head.index_last)++;
else
{
struct record_index_entry entry;
- read_indx (p, p->head.index_free, &entry, sizeof(entry), 0);
+ if (read_indx(p, p->head.index_free, &entry, sizeof(entry), 0) < 1)
+ {
+ xfree(rec);
+ return 0;
+ }
sysno = p->head.index_free;
p->head.index_free = entry.next;
}
-#if ZMBOL
-#else
- if (sysno > 100000)
- {
- logf (LOG_FATAL, "100,000 record limit reached");
- exit (1);
- }
-#endif
(p->head.no_records)++;
- rec->sysno = sysno;
+ rec->sysno = rec_sysno_to_ext(sysno);
for (i = 0; i < REC_NO_INFO; i++)
{
rec->info[i] = NULL;
rec->size[i] = 0;
}
- rec_cache_insert (p, rec, recordFlagNew);
+ rec_cache_insert(p, rec, recordFlagNew);
return rec;
}
-Record rec_new (Records p)
+Record rec_new(Records p)
{
Record rec;
- zebra_mutex_lock (&p->mutex);
+ zebra_mutex_lock(&p->mutex);
- rec = rec_new_int (p);
- zebra_mutex_unlock (&p->mutex);
+ rec = rec_new_int(p);
+ zebra_mutex_unlock(&p->mutex);
return rec;
}
-void rec_del (Records p, Record *recpp)
+ZEBRA_RES rec_del(Records p, Record *recpp)
{
Record *recp;
+ ZEBRA_RES ret = ZEBRA_OK;
- zebra_mutex_lock (&p->mutex);
+ zebra_mutex_lock(&p->mutex);
(p->head.no_records)--;
- if ((recp = rec_cache_lookup (p, (*recpp)->sysno, recordFlagDelete)))
+ if ((recp = rec_cache_lookup(p, (*recpp)->sysno, recordFlagDelete)))
{
- rec_rm (recp);
+ rec_rm(recp);
*recp = *recpp;
}
else
{
- rec_cache_insert (p, *recpp, recordFlagDelete);
- rec_rm (recpp);
+ ret = rec_cache_insert(p, *recpp, recordFlagDelete);
+ rec_rm(recpp);
}
- zebra_mutex_unlock (&p->mutex);
+ zebra_mutex_unlock(&p->mutex);
*recpp = NULL;
+ return ret;
}
-void rec_put (Records p, Record *recpp)
+ZEBRA_RES rec_put(Records p, Record *recpp)
{
Record *recp;
+ ZEBRA_RES ret = ZEBRA_OK;
- zebra_mutex_lock (&p->mutex);
- if ((recp = rec_cache_lookup (p, (*recpp)->sysno, recordFlagWrite)))
+ zebra_mutex_lock(&p->mutex);
+ if ((recp = rec_cache_lookup(p, (*recpp)->sysno, recordFlagWrite)))
{
- rec_rm (recp);
+ rec_rm(recp);
*recp = *recpp;
}
else
{
- rec_cache_insert (p, *recpp, recordFlagWrite);
- rec_rm (recpp);
+ ret = rec_cache_insert(p, *recpp, recordFlagWrite);
+ rec_rm(recpp);
}
- zebra_mutex_unlock (&p->mutex);
+ zebra_mutex_unlock(&p->mutex);
*recpp = NULL;
+ return ret;
}
-void rec_rm (Record *recpp)
+void rec_rm(Record *recpp)
{
int i;
if (!*recpp)
return ;
for (i = 0; i < REC_NO_INFO; i++)
- xfree ((*recpp)->info[i]);
- xfree (*recpp);
+ xfree((*recpp)->info[i]);
+ xfree(*recpp);
*recpp = NULL;
}
-Record rec_cp (Record rec)
+Record rec_cp(Record rec)
{
Record n;
int i;
- n = (Record) xmalloc (sizeof(*n));
+ n = (Record) xmalloc(sizeof(*n));
n->sysno = rec->sysno;
for (i = 0; i < REC_NO_INFO; i++)
if (!rec->info[i])
else
{
n->size[i] = rec->size[i];
- n->info[i] = (char *) xmalloc (rec->size[i]);
- memcpy (n->info[i], rec->info[i], rec->size[i]);
+ n->info[i] = (char *) xmalloc(rec->size[i]+1);
+ memcpy(n->info[i], rec->info[i], rec->size[i]);
+ n->info[i][rec->size[i]] = '\0';
}
return n;
}
-char *rec_strdup (const char *s, size_t *len)
+char *rec_strdup(const char *s, size_t *len)
{
char *p;
return NULL;
}
*len = strlen(s)+1;
- p = (char *) xmalloc (*len);
- strcpy (p, s);
+ p = (char *) xmalloc(*len);
+ strcpy(p, s);
return p;
}
+/*
+ * Local variables:
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ * vim: shiftwidth=4 tabstop=8 expandtab
+ */
+