/*
- * Copyright (C) 1994-1995, Index Data I/S
+ * Copyright (C) 1994-1998, Index Data I/S
* All rights reserved.
* Sebastian Hammer, Adam Dickmeiss
*
* $Log: extract.c,v $
- * Revision 1.45 1995-12-15 12:37:41 adam
+ * Revision 1.78 1998-02-10 12:03:05 adam
+ * Implemented Sort.
+ *
+ * Revision 1.77 1998/01/12 15:04:08 adam
+ * The test option (-s) only uses read-lock (and not write lock).
+ *
+ * Revision 1.76 1997/10/27 14:33:04 adam
+ * Moved towards generic character mapping depending on "structure"
+ * field in abstract syntax file. Fixed a few memory leaks. Fixed
+ * bug with negative integers when doing searches with relational
+ * operators.
+ *
+ * Revision 1.75 1997/09/17 12:19:12 adam
+ * Zebra version corresponds to YAZ version 1.4.
+ * Changed Zebra server so that it doesn't depend on global common_resource.
+ *
+ * Revision 1.74 1997/09/09 13:38:06 adam
+ * Partial port to WIN95/NT.
+ *
+ * Revision 1.73 1997/09/04 13:57:20 adam
+ * New file extract/retrieve method tellf (added).
+ * Added O_BINARY for open calls.
+ *
+ * Revision 1.72 1997/07/15 16:32:29 adam
+ * Bug fix: Match handler didn't terminate the resulting string!
+ *
+ * Revision 1.71 1997/07/15 16:28:41 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 1.70 1997/07/01 13:00:42 adam
+ * Bug fix in routine searchRecordKey: uninitialized variables.
+ *
+ * Revision 1.69 1997/04/29 09:26:03 adam
+ * Bug fix: generic recordId handling didn't work for compressed internal
+ * keys.
+ *
+ * Revision 1.68 1997/02/12 20:39:45 adam
+ * Implemented options -f <n> that limits the log to the first <n>
+ * records.
+ * Changed some log messages also.
+ *
+ * Revision 1.67 1996/11/15 15:02:14 adam
+ * Minor changes regarding logging.
+ *
+ * Revision 1.66 1996/11/14 09:52:21 adam
+ * Strings in record keys bound by IT_MAX_WORD.
+ *
+ * Revision 1.65 1996/11/14 08:57:56 adam
+ * Reduction of storeKeys area.
+ *
+ * Revision 1.64 1996/11/08 11:10:16 adam
+ * Buffers used during file match got bigger.
+ * Compressed ISAM support everywhere.
+ * Bug fixes regarding masking characters in queries.
+ * Redesigned Regexp-2 queries.
+ *
+ * Revision 1.63 1996/10/29 14:09:39 adam
+ * Use of cisam system - enabled if setting isamc is 1.
+ *
+ * Revision 1.62 1996/10/11 10:57:01 adam
+ * New module recctrl. Used to manage records (extract/retrieval).
+ * Several files have been moved to the recctrl sub directory.
+ *
+ * Revision 1.61 1996/06/06 12:08:37 quinn
+ * Added showRecord function
+ *
+ * Revision 1.60 1996/06/04 10:18:12 adam
+ * Search/scan uses character mapping module.
+ *
+ * Revision 1.59 1996/05/14 15:47:07 adam
+ * Cleanup of various buffer size entities.
+ *
+ * Revision 1.58 1996/05/14 06:16:38 adam
+ * Compact use/set bytes used in search service.
+ *
+ * Revision 1.57 1996/05/13 14:23:04 adam
+ * Work on compaction of set/use bytes in dictionary.
+ *
+ * Revision 1.56 1996/05/09 09:54:42 adam
+ * Server supports maps from one logical attributes to a list of physical
+ * attributes.
+ * The extraction process doesn't make space consuming 'any' keys.
+ *
+ * Revision 1.55 1996/05/09 07:28:55 quinn
+ * Work towards phrases and multiple registers
+ *
+ * Revision 1.54 1996/05/01 13:46:35 adam
+ * First work on multiple records in one file.
+ * New option, -offset, to the "unread" command in the filter module.
+ *
+ * Revision 1.53 1996/04/26 12:09:43 adam
+ * Added a few comments.
+ *
+ * Revision 1.52 1996/04/25 13:27:57 adam
+ * Function recordExtract modified so that files with no keys (possibly empty)
+ * are ignored.
+ *
+ * Revision 1.51 1996/03/19 11:08:42 adam
+ * Bug fix: Log preamble wasn't always turned off after recordExtract.
+ *
+ * Revision 1.50 1996/02/12 18:45:36 adam
+ * New fileVerboseFlag in record group control.
+ *
+ * Revision 1.49 1996/02/05 12:29:57 adam
+ * Logging reduced a bit.
+ * The remaining running time is estimated during register merge.
+ *
+ * Revision 1.48 1996/02/01 20:53:26 adam
+ * The temporary per-record keys are compacted a little, and duplication
+ * of the per-records keys are avoided when they are saved in the record
+ * information buffer.
+ *
+ * Revision 1.47 1996/01/17 14:57:48 adam
+ * Prototype changed for reader functions in extract/retrieve. File
+ * is identified by 'void *' instead of 'int.
+ *
+ * Revision 1.46 1995/12/15 14:57:16 adam
+ * Bug fix.
+ *
+ * Revision 1.45 1995/12/15 12:37:41 adam
* In addRecordKeyAny: Writes key only when attrSet != -1.
*
* Revision 1.44 1995/12/12 16:00:54 adam
*/
#include <stdio.h>
#include <assert.h>
+#ifdef WINDOWS
+#include <io.h>
+#else
#include <unistd.h>
+#endif
#include <fcntl.h>
-#include <ctype.h>
-#include <alexutil.h>
#include <recctrl.h>
+#include <charmap.h>
+#include <sortidx.h>
#include "index.h"
-#include "recindex.h"
+#include "zinfo.h"
static Dict matchDict;
static Records records = NULL;
+static SortIdx sortIdx = NULL;
static char **key_buf;
static size_t ptr_top;
static int records_inserted = 0;
static int records_updated = 0;
static int records_deleted = 0;
+static int records_processed = 0;
-void key_open (int mem)
+static ZebTargetInfo *zti = NULL;
+
+static void logRecord (int showFlag)
{
+ if (!showFlag)
+ ++records_processed;
+ if (showFlag || !(records_processed % 1000))
+ {
+ logf (LOG_LOG, "Records: %7d i/u/d %d/%d/%d",
+ records_processed, records_inserted, records_updated,
+ records_deleted);
+ }
+}
+
+int key_open (BFiles bfs, int mem, int rw)
+{
+ if (!mem)
+ mem = atoi(res_get_def (common_resource, "memMax", "4"))*1024*1024;
if (mem < 50000)
mem = 50000;
key_buf = xmalloc (mem);
key_buf_used = 0;
key_file_no = 0;
- if (!(matchDict = dict_open (GMATCH_DICT, 50, 1)))
+ if (!(matchDict = dict_open (bfs, GMATCH_DICT, 50, rw)))
{
logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT);
- exit (1);
+ return -1;
}
assert (!records);
- records = rec_open (1);
+ records = rec_open (bfs, rw);
+ if (!records)
+ {
+ dict_close (matchDict);
+ return -1;
+ }
+ zti = zebTargetInfo_open (records, rw);
+ if (!zti)
+ {
+ rec_close (&records);
+ dict_close (matchDict);
+ return -1;
+ }
+ sortIdx = sortIdx_open (bfs, 1);
+ return 0;
}
struct encode_info {
int sysno;
int seqno;
- char buf[512];
+ char buf[768];
};
void encode_key_init (struct encode_info *i)
}
}
+#define SORT_EXTRA 0
+
+#if SORT_EXTRA
+static int key_y_len;
+
+static int key_y_compare (const void *p1, const void *p2)
+{
+ int r;
+
+ if ((r = key_compare (*(char**) p1 + key_y_len + 1,
+ *(char**) p2 + key_y_len + 1)))
+ return r;
+ return *(*(char**) p1 + key_y_len) - *(*(char**) p2 + key_y_len);
+}
+
+static int key_x_compare (const void *p1, const void *p2)
+{
+ return strcmp (*(char**) p1, *(char**) p2);
+}
+#endif
+
void key_flush (void)
{
FILE *outf;
char out_fname[200];
char *prevcp, *cp;
struct encode_info encode_info;
+#if SORT_EXTRA
+ int i;
+#endif
if (ptr_i <= 0)
return;
key_file_no++;
logf (LOG_LOG, "sorting section %d", key_file_no);
+#if !SORT_EXTRA
qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare);
- sprintf (out_fname, TEMP_FNAME, key_file_no);
+ getFnameTmp (out_fname, key_file_no);
if (!(outf = fopen (out_fname, "w")))
{
- logf (LOG_FATAL|LOG_ERRNO, "fopen (4) %s", out_fname);
+ logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname);
exit (1);
}
logf (LOG_LOG, "writing section %d", key_file_no);
else
encode_key_write (cp + strlen(cp), &encode_info, outf);
}
+#else
+ qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_x_compare);
+ getFnameTmp (out_fname, key_file_no);
+
+ if (!(outf = fopen (out_fname, "w")))
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "fopen %s", out_fname);
+ exit (1);
+ }
+ logf (LOG_LOG, "writing section %d", key_file_no);
+ i = ptr_i;
+ prevcp = key_buf[ptr_top-i];
+ while (1)
+ if (!--i || strcmp (prevcp, key_buf[ptr_top-i]))
+ {
+ key_y_len = strlen(prevcp)+1;
+#if 0
+ logf (LOG_LOG, "key_y_len: %2d %02x %02x %s",
+ key_y_len, prevcp[0], prevcp[1], 2+prevcp);
+#endif
+ qsort (key_buf + ptr_top-ptr_i, ptr_i - i,
+ sizeof(char*), key_y_compare);
+ cp = key_buf[ptr_top-ptr_i];
+ --key_y_len;
+ encode_key_init (&encode_info);
+ encode_key_write (cp, &encode_info, outf);
+ while (--ptr_i > i)
+ {
+ cp = key_buf[ptr_top-ptr_i];
+ encode_key_write (cp+key_y_len, &encode_info, outf);
+ }
+ if (!i)
+ break;
+ prevcp = key_buf[ptr_top-ptr_i];
+ }
+#endif
if (fclose (outf))
{
logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname);
key_buf_used = 0;
}
-int key_close (void)
+int key_close ()
{
key_flush ();
xfree (key_buf);
+#if 1
+ zebTargetInfo_close (zti, 1);
+#endif
rec_close (&records);
dict_close (matchDict);
+ sortIdx_close (sortIdx);
- logf (LOG_LOG, "Records inserted %6d", records_inserted);
- logf (LOG_LOG, "Records updated %6d", records_updated);
- logf (LOG_LOG, "Records deleted %6d", records_deleted);
+ logRecord (1);
return key_file_no;
}
-static void wordInit (RecWord *p)
+static void wordInit (struct recExtractCtrl *p, RecWord *w)
{
- p->attrSet = 1;
- p->attrUse = 1016;
- p->which = Word_String;
+ w->zebra_maps = p->zebra_maps;
+ w->attrSet = 1;
+ w->attrUse = 1016;
+ w->reg_type = 'w';
}
-struct recKeys {
+static struct sortKey {
+ char *string;
+ int length;
+ int attrSet;
+ int attrUse;
+ struct sortKey *next;
+} *sortKeys = NULL;
+
+static struct recKeys {
int buf_used;
int buf_max;
char *buf;
+ char prevAttrSet;
+ short prevAttrUse;
+ int prevSeqNo;
} reckeys;
-static void addRecordKey (const RecWord *p)
+static void addIndexString (RecWord *p, const char *string, int length)
{
char *dst;
char attrSet;
short attrUse;
- size_t i;
+ int lead = 0;
+ int diff = 0;
if (reckeys.buf_used+1024 > reckeys.buf_max)
{
char *b;
- b = xmalloc (reckeys.buf_max += 65000);
+ b = xmalloc (reckeys.buf_max += 128000);
if (reckeys.buf_used > 0)
memcpy (b, reckeys.buf, reckeys.buf_used);
xfree (reckeys.buf);
reckeys.buf = b;
}
dst = reckeys.buf + reckeys.buf_used;
- switch (p->which)
+
+ attrSet = p->attrSet;
+ if (reckeys.buf_used > 0 && reckeys.prevAttrSet == attrSet)
+ lead |= 1;
+ else
+ reckeys.prevAttrSet = attrSet;
+ attrUse = p->attrUse;
+ if (reckeys.buf_used > 0 && reckeys.prevAttrUse == attrUse)
+ lead |= 2;
+ else
+ reckeys.prevAttrUse = attrUse;
+#if 1
+ diff = 1 + p->seqno - reckeys.prevSeqNo;
+ if (diff >= 1 && diff <= 15)
+ lead |= (diff << 2);
+ else
+ diff = 0;
+#endif
+ reckeys.prevSeqNo = p->seqno;
+
+ *dst++ = lead;
+
+ if (!(lead & 1))
{
- case Word_String:
- attrSet = p->attrSet;
memcpy (dst, &attrSet, sizeof(attrSet));
dst += sizeof(attrSet);
-
- attrUse = p->attrUse;
+ }
+ if (!(lead & 2))
+ {
memcpy (dst, &attrUse, sizeof(attrUse));
dst += sizeof(attrUse);
-
- for (i = 0; p->u.string[i]; i++)
- *dst++ = p->u.string[i];
- *dst++ = '\0';
+ }
+ *dst++ = p->reg_type;
+ memcpy (dst, string, length);
+ dst += length;
+ *dst++ = '\0';
+ if (!diff)
+ {
memcpy (dst, &p->seqno, sizeof(p->seqno));
dst += sizeof(p->seqno);
-
- break;
- default:
- return;
}
reckeys.buf_used = dst - reckeys.buf;
+ (p->seqno)++;
+}
+
+static void addSortString (RecWord *p, const char *string, int length)
+{
+ struct sortKey *sk;
+
+ for (sk = sortKeys; sk; sk = sk->next)
+ if (sk->attrSet == p->attrSet && sk->attrUse == p->attrUse)
+ return;
+
+ sk = xmalloc (sizeof(*sk));
+ sk->next = sortKeys;
+ sortKeys = sk;
+
+ sk->string = xmalloc (p->length);
+ sk->length = p->length;
+ memcpy (sk->string, p->string, p->length);
+ sk->attrSet = p->attrSet;
+ sk->attrUse = p->attrUse;
+}
+
+static void addString (RecWord *p, const char *string, int length)
+{
+ if (zebra_maps_is_sort (p->zebra_maps, p->reg_type))
+ addSortString (p, string, length);
+ else
+ addIndexString (p, string, length);
+}
+
+static void addIncompleteField (RecWord *p)
+{
+ const char *b = p->string;
+ int remain = p->length;
+ const char **map = 0;
+
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+
+ while (map)
+ {
+ char buf[IT_MAX_WORD+1];
+ int i, remain;
+
+ /* Skip spaces */
+ while (map && *map && **map == *CHR_SPACE)
+ {
+ remain = p->length - (b - p->string);
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+ else
+ map = 0;
+ }
+ if (!map)
+ break;
+ i = 0;
+ while (map && *map && **map != *CHR_SPACE)
+ {
+ const char *cp = *map;
+
+ while (i < IT_MAX_WORD && *cp)
+ buf[i++] = *(cp++);
+ remain = p->length - (b - p->string);
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+ else
+ map = 0;
+ }
+ if (!i)
+ return;
+ addString (p, buf, i);
+ }
+}
+
+static void addCompleteField (RecWord *p)
+{
+ const char *b = p->string;
+ char buf[IT_MAX_WORD+1];
+ const char **map = 0;
+ int i = 0, remain = p->length;
+
+ if (remain > 0)
+ map = zebra_maps_input (p->zebra_maps, p->reg_type, &b, remain);
+
+ while (remain > 0 && i < IT_MAX_WORD)
+ {
+ while (map && *map && **map == *CHR_SPACE)
+ {
+ remain = p->length - (b - p->string);
+ if (remain > 0)
+ map = zebra_maps_input(p->zebra_maps, p->reg_type, &b, remain);
+ else
+ map = 0;
+ }
+ if (!map)
+ break;
+
+ if (i && i < IT_MAX_WORD)
+ buf[i++] = *CHR_SPACE;
+ while (map && *map && **map != *CHR_SPACE)
+ {
+ const char *cp = *map;
+
+ if (i >= IT_MAX_WORD)
+ break;
+ while (i < IT_MAX_WORD && *cp)
+ buf[i++] = *(cp++);
+ remain = p->length - (b - p->string);
+ if (remain > 0)
+ map = zebra_maps_input (p->zebra_maps, p->reg_type, &b,
+ remain);
+ else
+ map = 0;
+ }
+ }
+ if (!i)
+ return;
+ addString (p, buf, i);
+}
+
+static void addRecordKey (RecWord *p)
+{
+ if (zebra_maps_is_complete (p->zebra_maps, p->reg_type))
+ addCompleteField (p);
+ else
+ addIncompleteField(p);
+}
+
+static void flushSortKeys (SYSNO sysno, int cmd)
+{
+ struct sortKey *sk = sortKeys;
+
+ sortIdx_sysno (sortIdx, sysno);
+ while (sk)
+ {
+ struct sortKey *sk_next = sk->next;
+ sortIdx_type (sortIdx, sk->attrUse);
+ sortIdx_add (sortIdx, sk->string, sk->length);
+ xfree (sk->string);
+ xfree (sk);
+ sk = sk_next;
+ }
+ sortKeys = NULL;
}
static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys,
const char *databaseName)
{
+ char attrSet = -1;
+ short attrUse = -1;
+ int seqno = 0;
int off = 0;
+
+ if (zebTargetInfo_curDatabase (zti, databaseName))
+ {
+ if (zebTargetInfo_newDatabase (zti, databaseName))
+ abort ();
+ }
while (off < reckeys->buf_used)
{
const char *src = reckeys->buf + off;
- char attrSet;
- short attrUse;
struct it_key key;
-
- memcpy (&attrSet, src, sizeof(attrSet));
- src += sizeof(attrSet);
-
- memcpy (&attrUse, src, sizeof(attrUse));
- src += sizeof(attrUse);
+ int lead, ch;
+
+ lead = *src++;
+ if (!(lead & 1))
+ {
+ memcpy (&attrSet, src, sizeof(attrSet));
+ src += sizeof(attrSet);
+ }
+ if (!(lead & 2))
+ {
+ memcpy (&attrUse, src, sizeof(attrUse));
+ src += sizeof(attrUse);
+ }
if (key_buf_used + 1024 > (ptr_top-ptr_i)*sizeof(char*))
key_flush ();
++ptr_i;
key_buf[ptr_top-ptr_i] = (char*)key_buf + key_buf_used;
- key_buf_used += index_word_prefix ((char*)key_buf + key_buf_used,
- attrSet, attrUse, databaseName);
+
+ ch = zebTargetInfo_lookupSU (zti, attrSet, attrUse);
+ if (ch < 0)
+ ch = zebTargetInfo_addSU (zti, attrSet, attrUse);
+ assert (ch > 0);
+ ((char*) key_buf) [key_buf_used++] = ch;
while (*src)
- ((char*)key_buf) [key_buf_used++] = index_char_cvt (*src++);
+ ((char*)key_buf) [key_buf_used++] = *src++;
src++;
((char*)key_buf) [key_buf_used++] = '\0';
-
((char*) key_buf)[key_buf_used++] = cmd;
- memcpy (&key.seqno, src, sizeof(key.seqno));
- src += sizeof(key.seqno);
+ if (lead & 60)
+ seqno += ((lead>>2) & 15)-1;
+ else
+ {
+ memcpy (&seqno, src, sizeof(seqno));
+ src += sizeof(seqno);
+ }
+ key.seqno = seqno;
key.sysno = sysno;
memcpy ((char*)key_buf + key_buf_used, &key, sizeof(key));
key_buf_used += sizeof(key);
int off = 0;
int startSeq = -1;
int i;
+ short attrUse;
+ char attrSet;
+ int seqno = 0;
for (i = 0; i<32; i++)
ws[i] = NULL;
while (off < reckeys->buf_used)
{
- const char *src = reckeys->buf + off;
- char attrSet;
- short attrUse;
- int seqno;
- const char *wstart;
-
- memcpy (&attrSet, src, sizeof(attrSet));
- src += sizeof(attrSet);
- memcpy (&attrUse, src, sizeof(attrUse));
- src += sizeof(attrUse);
+ const char *src = reckeys->buf + off;
+ const char *wstart;
+ int lead;
+
+ lead = *src++;
+ if (!(lead & 1))
+ {
+ memcpy (&attrSet, src, sizeof(attrSet));
+ src += sizeof(attrSet);
+ }
+ if (!(lead & 2))
+ {
+ memcpy (&attrUse, src, sizeof(attrUse));
+ src += sizeof(attrUse);
+ }
wstart = src;
while (*src++)
;
-
- memcpy (&seqno, src, sizeof(seqno));
- src += sizeof(seqno);
-
-#if 0
- logf (LOG_LOG, "(%d,%d) %d %s", attrSet, attrUse, seqno, wstart);
-#endif
+ if (lead & 60)
+ seqno += ((lead>>2) & 15)-1;
+ else
+ {
+ memcpy (&seqno, src, sizeof(seqno));
+ src += sizeof(seqno);
+ }
if (attrUseS == attrUse && attrSetS == attrSet)
{
int woff;
return ws;
}
-static void addRecordKeyAny (const RecWord *p)
+struct file_read_info {
+ off_t file_max;
+ off_t file_offset;
+ off_t file_moffset;
+ int file_more;
+ int fd;
+};
+
+static struct file_read_info *file_read_start (int fd)
{
- if (p->attrSet != 1 || p->attrUse != 1016)
- {
- RecWord w;
+ struct file_read_info *fi = xmalloc (sizeof(*fi));
- memcpy (&w, p, sizeof(w));
- w.attrSet = 1;
- w.attrUse = 1016;
- addRecordKey (&w);
- }
- else if (p->attrSet != -1)
- addRecordKey (p);
+ fi->fd = fd;
+ fi->file_max = 0;
+ fi->file_moffset = 0;
+ return fi;
}
-
-#define FILE_READ_BUFSIZE 4096
-
-static int file_noread;
-#if FILE_READ_BUFSIZE
-static char *file_buf;
-static int file_offset;
-static int file_bufsize;
-#endif
-
-static void file_read_start (int fd)
+static void file_read_stop (struct file_read_info *fi)
{
- file_noread = 0;
-#if FILE_READ_BUFSIZE
- file_offset = 0;
- file_buf = xmalloc (FILE_READ_BUFSIZE);
- file_bufsize = read (fd, file_buf, FILE_READ_BUFSIZE);
-#endif
+ assert (fi);
+ xfree (fi);
}
-static void file_read_stop (int fd)
+static off_t file_seek (void *handle, off_t offset)
{
-#if FILE_READ_BUFSIZE
- xfree (file_buf);
- file_buf = NULL;
-#endif
+ struct file_read_info *p = handle;
+ p->file_offset = offset;
+ return lseek (p->fd, offset, SEEK_SET);
}
-static int file_read (int fd, char *buf, size_t count)
+static off_t file_tell (void *handle)
{
-#if FILE_READ_BUFSIZE
- int l = file_bufsize - file_offset;
+ struct file_read_info *p = handle;
+ return p->file_offset;
+}
- if (count > l)
- {
- int r;
- if (l > 0)
- memcpy (buf, file_buf + file_offset, l);
- count = count-l;
- if (count > FILE_READ_BUFSIZE)
- {
- if ((r = read (fd, buf + l, count)) == -1)
- {
- logf (LOG_FATAL|LOG_ERRNO, "read");
- exit (1);
- }
- file_bufsize = 0;
- file_offset = 0;
- file_noread += l+r;
- return l+r;
- }
- file_bufsize = r = read (fd, file_buf, FILE_READ_BUFSIZE);
- if (r == -1)
- {
- logf (LOG_FATAL|LOG_ERRNO, "read");
- exit (1);
- }
- else if (r <= count)
- {
- file_offset = r;
- memcpy (buf + l, file_buf, r);
- file_noread += l+r;
- return l+r;
- }
- else
- {
- file_offset = count;
- memcpy (buf + l, file_buf, count - l);
- file_noread += count;
- return count;
- }
- }
- memcpy (buf, file_buf + file_offset, count);
- file_offset += count;
- file_noread += count;
- return count;
-#else
+static int file_read (void *handle, char *buf, size_t count)
+{
+ struct file_read_info *p = handle;
+ int fd = p->fd;
int r;
r = read (fd, buf, count);
if (r > 0)
- file_noread += r;
+ {
+ p->file_offset += r;
+ if (p->file_offset > p->file_max)
+ p->file_max = p->file_offset;
+ }
return r;
-#endif
+}
+
+static void file_begin (void *handle)
+{
+ struct file_read_info *p = handle;
+
+ p->file_offset = p->file_moffset;
+ if (p->file_moffset)
+ lseek (p->fd, p->file_moffset, SEEK_SET);
+ p->file_more = 0;
+}
+
+static void file_end (void *handle, off_t offset)
+{
+ struct file_read_info *p = handle;
+
+ assert (p->file_more == 0);
+ p->file_more = 1;
+ p->file_moffset = offset;
}
static int atois (const char **s)
}
if (first)
{
- logf (LOG_WARN, "Record in file %s didn't contain match"
- " fields in (%d,%d)", fname, attrSet, attrUse);
+ logf (LOG_WARN, "Record didn't contain match"
+ " fields in (%d,%d)", attrSet, attrUse);
return NULL;
}
}
fname, rGroup->groupName ? rGroup->groupName : "none");
return NULL;
}
+ *dst = '\0';
return dstBuf;
}
+struct recordLogInfo {
+ const char *fname;
+ int recordOffset;
+ struct recordGroup *rGroup;
+};
+
+static void recordLogPreamble (int level, const char *msg, void *info)
+{
+ struct recordLogInfo *p = info;
+ FILE *outf = log_file ();
+
+ if (level & LOG_LOG)
+ return ;
+ fprintf (outf, "File %s, offset %d, type %s\n",
+ p->rGroup->recordType, p->recordOffset, p->fname);
+ log_event_start (NULL, NULL);
+}
+
static int recordExtract (SYSNO *sysno, const char *fname,
struct recordGroup *rGroup, int deleteFlag,
- int fd, RecType recType, char *subType)
+ struct file_read_info *fi, RecType recType,
+ char *subType)
{
struct recExtractCtrl extractCtrl;
int r;
char *matchStr;
SYSNO sysnotmp;
+ off_t recordOffset = 0;
Record rec;
+ struct recordLogInfo logInfo;
- if (fd != -1)
+ if (fi->fd != -1)
{
- extractCtrl.fd = fd;
- /* extract keys */
+ /* we are going to read from a file, so prepare the extraction */
+ extractCtrl.fh = fi;
extractCtrl.subType = subType;
extractCtrl.init = wordInit;
- extractCtrl.add = addRecordKeyAny;
+ extractCtrl.add = addRecordKey;
+ extractCtrl.dh = rGroup->dh;
reckeys.buf_used = 0;
+ reckeys.prevAttrUse = -1;
+ reckeys.prevAttrSet = -1;
+ reckeys.prevSeqNo = 0;
+
+ recordOffset = fi->file_moffset;
+ extractCtrl.offset = recordOffset;
extractCtrl.readf = file_read;
+ extractCtrl.seekf = file_seek;
+ extractCtrl.tellf = file_tell;
+ extractCtrl.endf = file_end;
+ extractCtrl.zebra_maps = rGroup->zebra_maps;
+ extractCtrl.flagShowRecords = !rGroup->flagRw;
+ if (!rGroup->flagRw)
+ printf ("File: %s %ld\n", fname, (long) recordOffset);
+
+ logInfo.fname = fname;
+ logInfo.recordOffset = recordOffset;
+ logInfo.rGroup = rGroup;
+ log_event_start (recordLogPreamble, &logInfo);
+
r = (*recType->extract)(&extractCtrl);
-
+
+ log_event_start (NULL, NULL);
+
if (r)
{
- logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);
+ /* error occured during extraction ... */
+ if (rGroup->flagRw &&
+ records_processed < rGroup->fileVerboseLimit)
+ {
+ logf (LOG_WARN, "fail %s %s %ld code = %d", rGroup->recordType,
+ fname, (long) recordOffset, r);
+ }
+ return 0;
+ }
+ if (reckeys.buf_used == 0)
+ {
+ /* the extraction process returned no information - the record
+ is probably empty - unless flagShowRecords is in use */
+ if (!rGroup->flagRw)
+ return 1;
+ logf (LOG_WARN, "No keys generated for file %s", fname);
+ logf (LOG_WARN, " The file is probably empty");
return 0;
}
}
}
else
{
- logf (LOG_WARN, "Record not inserted");
+ logf (LOG_WARN, "Bad match criteria");
return 0;
}
}
}
- /* new record ? */
if (! *sysno)
{
+ /* new record */
if (deleteFlag)
{
- logf (LOG_LOG, "? %s", fname);
+ logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType,
+ fname, (long) recordOffset);
+ logf (LOG_WARN, "cannot delete record above (seems new)");
return 1;
}
- logf (LOG_LOG, "add %s %s", rGroup->recordType, fname);
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "add %s %s %ld", rGroup->recordType,
+ fname, (long) recordOffset);
rec = rec_new (records);
*sysno = rec->sysno;
dict_insert (matchDict, matchStr, sizeof(*sysno), sysno);
}
flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName);
+ flushSortKeys (*sysno, 1);
records_inserted++;
}
else
{
+ /* record already exists */
struct recKeys delkeys;
rec = rec_get (records, *sysno);
assert (rec);
delkeys.buf_used = rec->size[recInfo_delKeys];
delkeys.buf = rec->info[recInfo_delKeys];
+ flushSortKeys (*sysno, 0);
flushRecordKeys (*sysno, 0, &delkeys, rec->info[recInfo_databaseName]);
if (deleteFlag)
{
+ /* record going to be deleted */
if (!delkeys.buf_used)
{
- logf (LOG_WARN, "cannot delete %s: storeKeys false",
- fname);
+ logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType,
+ fname, (long) recordOffset);
+ logf (LOG_WARN, "cannot delete file above, storeKeys false");
}
else
{
- logf (LOG_LOG, "delete %s %s", rGroup->recordType, fname);
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "delete %s %s %ld", rGroup->recordType,
+ fname, (long) recordOffset);
records_deleted++;
if (matchStr)
dict_delete (matchDict, matchStr);
rec_del (records, &rec);
}
+ logRecord (0);
return 1;
}
else
{
+ /* record going to be updated */
if (!delkeys.buf_used)
{
- logf (LOG_WARN, "cannot update %s: storeKeys false",
- fname);
+ logf (LOG_LOG, "update %s %s %ld", rGroup->recordType,
+ fname, (long) recordOffset);
+ logf (LOG_WARN, "cannot update file above, storeKeys false");
}
else
{
- logf (LOG_LOG, "update %s %s", rGroup->recordType,
- fname);
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "update %s %s %ld", rGroup->recordType,
+ fname, (long) recordOffset);
flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName);
records_updated++;
}
}
}
+ /* update file type */
xfree (rec->info[recInfo_fileType]);
rec->info[recInfo_fileType] =
rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]);
+ /* update filename */
xfree (rec->info[recInfo_filename]);
rec->info[recInfo_filename] =
rec_strdup (fname, &rec->size[recInfo_filename]);
+ /* update delete keys */
xfree (rec->info[recInfo_delKeys]);
if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1)
{
+#if 1
+ rec->size[recInfo_delKeys] = reckeys.buf_used;
+ rec->info[recInfo_delKeys] = reckeys.buf;
+ reckeys.buf = NULL;
+ reckeys.buf_max = 0;
+#else
rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used);
rec->size[recInfo_delKeys] = reckeys.buf_used;
memcpy (rec->info[recInfo_delKeys], reckeys.buf,
rec->size[recInfo_delKeys]);
+#endif
}
else
{
rec->size[recInfo_delKeys] = 0;
}
+ /* update store data */
xfree (rec->info[recInfo_storeData]);
if (rGroup->flagStoreData == 1)
{
- rec->size[recInfo_storeData] = file_noread;
- rec->info[recInfo_storeData] = xmalloc (file_noread);
-#if FILE_READ_BUFSIZE
- if (file_noread < FILE_READ_BUFSIZE)
- memcpy (rec->info[recInfo_storeData], file_buf, file_noread);
- else
-#endif
+ int size = fi->file_moffset - recordOffset;
+ if (!size)
+ size = fi->file_max - recordOffset;
+ rec->size[recInfo_storeData] = size;
+ rec->info[recInfo_storeData] = xmalloc (size);
+ if (lseek (fi->fd, recordOffset, SEEK_SET) < 0)
{
- if (lseek (fd, 0L, SEEK_SET) < 0)
- {
- logf (LOG_ERRNO|LOG_FATAL, "seek to 0 in %s", fname);
- exit (1);
- }
- if (read (fd, rec->info[recInfo_storeData], file_noread)
- < file_noread)
- {
- logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s",
- file_noread, fname);
- exit (1);
- }
+ logf (LOG_ERRNO|LOG_FATAL, "seek to %ld in %s", fname,
+ (long) recordOffset);
+ exit (1);
+ }
+ if (read (fi->fd, rec->info[recInfo_storeData], size) < size)
+ {
+ logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s",
+ fi->file_max, fname);
+ exit (1);
}
}
else
rec->info[recInfo_storeData] = NULL;
rec->size[recInfo_storeData] = 0;
}
+ /* update database name */
xfree (rec->info[recInfo_databaseName]);
rec->info[recInfo_databaseName] =
rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]);
+ /* update offset */
+ xfree (rec->info[recInfo_offset]);
+
+ rec->size[recInfo_offset] = sizeof(recordOffset);
+ rec->info[recInfo_offset] = xmalloc (sizeof(recordOffset));
+ memcpy (rec->info[recInfo_offset], &recordOffset, sizeof(recordOffset));
+
+ /* commit this record */
rec_put (records, &rec);
+ logRecord (0);
return 1;
}
int fileExtract (SYSNO *sysno, const char *fname,
const struct recordGroup *rGroupP, int deleteFlag)
{
- int i, fd;
+ int r, i, fd;
char gprefix[128];
char ext[128];
char ext_res[128];
RecType recType;
struct recordGroup rGroupM;
struct recordGroup *rGroup = &rGroupM;
+ struct file_read_info *fi;
memcpy (rGroup, rGroupP, sizeof(*rGroupP));
sprintf (ext_res, "%srecordType", gprefix);
if (!(rGroup->recordType = res_get (common_resource, ext_res)))
{
- logf (LOG_LOG, "? %s", fname);
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "? %s", fname);
return 0;
}
}
}
if (!rGroup->recordType)
{
- logf (LOG_LOG, "? record %s", fname);
+ if (records_processed < rGroup->fileVerboseLimit)
+ logf (LOG_LOG, "? record %s", fname);
return 0;
}
if (!(recType = recType_byName (rGroup->recordType, subType)))
fd = -1;
else
{
- if ((fd = open (fname, O_RDONLY)) == -1)
+ if ((fd = open (fname, O_BINARY|O_RDONLY)) == -1)
{
logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
return 0;
}
}
- file_read_start (fd);
- recordExtract (sysno, fname, rGroup, deleteFlag, fd, recType, subType);
- file_read_stop (fd);
+ fi = file_read_start (fd);
+ do
+ {
+ file_begin (fi);
+ r = recordExtract (sysno, fname, rGroup, deleteFlag, fi,
+ recType, subType);
+ } while (r && !sysno && fi->file_more);
+ file_read_stop (fi);
if (fd != -1)
close (fd);
- return 1;
+ return r;
}
projects / idzebra-moved-to-github.git / blobdiff