/*
- * Copyright (C) 1995, Index Data I/S
+ * Copyright (C) 1994-1995, Index Data I/S
* All rights reserved.
* Sebastian Hammer, Adam Dickmeiss
*
* $Log: extract.c,v $
- * Revision 1.1 1995-09-01 14:06:35 adam
+ * Revision 1.21 1995-10-10 12:24:38 adam
+ * Temporary sort files are compressed.
+ *
+ * Revision 1.20 1995/10/06 13:52:05 adam
+ * Bug fixes. Handler may abort further scanning.
+ *
+ * Revision 1.19 1995/10/04 12:55:16 adam
+ * Bug fix in ranked search. Use=Any keys inserted.
+ *
+ * Revision 1.18 1995/10/04 09:37:08 quinn
+ * Fixed bug.
+ *
+ * Revision 1.17 1995/10/03 14:28:57 adam
+ * Buffered read in extract works.
+ *
+ * Revision 1.16 1995/10/03 14:28:45 adam
+ * Work on more effecient read handler in extract.
+ *
+ * Revision 1.15 1995/10/02 15:42:53 adam
+ * Extract uses file descriptors instead of FILE pointers.
+ *
+ * Revision 1.14 1995/10/02 15:29:13 adam
+ * More logging in file_extract.
+ *
+ * Revision 1.13 1995/09/29 14:01:39 adam
+ * Bug fixes.
+ *
+ * Revision 1.12 1995/09/28 14:22:56 adam
+ * Sort uses smaller temporary files.
+ *
+ * Revision 1.11 1995/09/28 12:10:31 adam
+ * Bug fixes. Field prefix used in queries.
+ *
+ * Revision 1.10 1995/09/28 09:19:41 adam
+ * xfree/xmalloc used everywhere.
+ * Extract/retrieve method seems to work for text records.
+ *
+ * Revision 1.9 1995/09/27 12:22:28 adam
+ * More work on extract in record control.
+ * Field name is not in isam keys but in prefix in dictionary words.
+ *
+ * Revision 1.8 1995/09/14 07:48:22 adam
+ * Record control management.
+ *
+ * Revision 1.7 1995/09/11 13:09:32 adam
+ * More work on relevance feedback.
+ *
+ * Revision 1.6 1995/09/08 14:52:27 adam
+ * Minor changes. Dictionary is lower case now.
+ *
+ * Revision 1.5 1995/09/06 16:11:16 adam
+ * Option: only one word key per file.
+ *
+ * Revision 1.4 1995/09/05 15:28:39 adam
+ * More work on search engine.
+ *
+ * Revision 1.3 1995/09/04 12:33:41 adam
+ * Various cleanup. YAZ util used instead.
+ *
+ * Revision 1.2 1995/09/04 09:10:34 adam
+ * More work on index add/del/update.
+ * Merge sort implemented.
+ * Initial work on z39 server.
+ *
+ * Revision 1.1 1995/09/01 14:06:35 adam
* Split of work into more files.
*
*/
#include <fcntl.h>
#include <ctype.h>
-#include <util.h>
+#include <alexutil.h>
+#include <recctrl.h>
#include "index.h"
-#define KEY_BUF_SIZE 100000
-
static Dict file_idx;
static SYSNO sysno_next;
-static int key_fd = -1;
-static char *key_buf;
-static int key_offset;
+static int sys_idx_fd = -1;
-void key_open (const char *fname)
+static int key_cmd;
+static int key_sysno;
+static char **key_buf;
+static size_t ptr_top;
+static size_t ptr_i;
+static size_t kused;
+static int key_file_no;
+
+void key_open (int mem)
{
void *file_key;
- if (key_fd != -1)
- return;
- if ((key_fd = open (fname, O_RDWR|O_CREAT, 0666)) == -1)
- {
- log (LOG_FATAL|LOG_ERRNO, "Creat %s", fname);
- exit (1);
- }
- if (!(key_buf = malloc (KEY_BUF_SIZE)))
- {
- log (LOG_FATAL|LOG_ERRNO, "malloc");
- exit (1);
- }
- key_offset = 0;
- if (!(file_idx = dict_open ("fileidx", 10, 1)))
+
+ if (mem < 50000)
+ mem = 50000;
+ key_buf = xmalloc (mem);
+ ptr_top = mem/sizeof(char*);
+ ptr_i = 0;
+ kused = 0;
+ key_file_no = 0;
+
+ if (!(file_idx = dict_open (FNAME_FILE_DICT, 40, 1)))
{
- log (LOG_FATAL, "dict_open fail of %s", "fileidx");
+ logf (LOG_FATAL, "dict_open fail of %s", "fileidx");
exit (1);
}
file_key = dict_lookup (file_idx, ".");
memcpy (&sysno_next, (char*)file_key+1, sizeof(sysno_next));
else
sysno_next = 1;
+ if ((sys_idx_fd = open (FNAME_SYS_IDX, O_RDWR|O_CREAT, 0666)) == -1)
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "open %s", FNAME_SYS_IDX);
+ exit (1);
+ }
}
-void key_close (void)
+struct encode_info {
+ int sysno;
+ int seqno;
+ char buf[512];
+};
+
+void encode_key_init (struct encode_info *i)
{
- if (key_fd == -1)
- return;
- close (key_fd);
- dict_insert (file_idx, ".", sizeof(sysno_next), &sysno_next);
- dict_close (file_idx);
- key_fd = -1;
+ i->sysno = 0;
+ i->seqno = 0;
}
+char *encode_key_int (int d, char *bp)
+{
+ if (d <= 63)
+ *bp++ = d;
+ else if (d <= 16383)
+ {
+ *bp++ = 64 + (d>>8);
+ *bp++ = d & 255;
+ }
+ else if (d <= 4194303)
+ {
+ *bp++ = 128 + (d>>16);
+ *bp++ = (d>>8) & 255;
+ *bp++ = d & 255;
+ }
+ else
+ {
+ *bp++ = 192 + (d>>24);
+ *bp++ = (d>>16) & 255;
+ *bp++ = (d>>8) & 255;
+ *bp++ = d & 255;
+ }
+ return bp;
+}
+
+void encode_key_write (char *k, struct encode_info *i, FILE *outf)
+{
+ struct it_key key;
+ char *bp = i->buf;
+
+ while ((*bp++ = *k++))
+ ;
+ memcpy (&key, k+1, sizeof(struct it_key));
+ bp = encode_key_int ( (key.sysno - i->sysno) * 2 + *k, bp);
+ if (i->sysno != key.sysno)
+ {
+ i->sysno = key.sysno;
+ i->seqno = 0;
+ }
+ bp = encode_key_int (key.seqno - i->seqno, bp);
+ i->seqno = key.seqno;
+ if (fwrite (i->buf, bp - i->buf, 1, outf) != 1)
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "fwrite");
+ exit (1);
+ }
+}
+
void key_flush (void)
{
- size_t i = 0;
- int w;
+ FILE *outf;
+ char out_fname[200];
+ char *prevcp, *cp;
+ struct encode_info encode_info;
+
+ if (ptr_i <= 0)
+ return;
+
+ key_file_no++;
+ logf (LOG_LOG, "sorting section %d", key_file_no);
+ qsort (key_buf + ptr_top-ptr_i, ptr_i, sizeof(char*), key_qsort_compare);
+ sprintf (out_fname, TEMP_FNAME, key_file_no);
+
+ if (!(outf = fopen (out_fname, "w")))
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "fopen (4) %s", out_fname);
+ exit (1);
+ }
+ logf (LOG_LOG, "writing section %d", key_file_no);
+ prevcp = cp = key_buf[ptr_top-ptr_i];
- while (i < key_offset)
+ encode_key_init (&encode_info);
+ encode_key_write (cp, &encode_info, outf);
+ while (--ptr_i > 0)
{
- w = write (key_fd, key_buf + i, key_offset - i);
- if (w == -1)
+ cp = key_buf[ptr_top-ptr_i];
+ if (strcmp (cp, prevcp))
{
- log (LOG_FATAL|LOG_ERRNO, "Write key fail");
- exit (1);
+ encode_key_init (&encode_info);
+ encode_key_write (cp, &encode_info, outf);
+ prevcp = cp;
}
- i += w;
+ else
+ encode_key_write (cp + strlen(cp), &encode_info, outf);
}
- key_offset = 0;
+ if (fclose (outf))
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "fclose %s", out_fname);
+ exit (1);
+ }
+ logf (LOG_LOG, "finished section %d", key_file_no);
+ ptr_i = 0;
+ kused = 0;
}
-void key_write (int cmd, struct it_key *k, const char *str)
+int key_close (void)
{
- char x = cmd;
- size_t slen = strlen(str);
+ key_flush ();
+ xfree (key_buf);
+ close (sys_idx_fd);
+ dict_insert (file_idx, ".", sizeof(sysno_next), &sysno_next);
+ dict_close (file_idx);
+ return key_file_no;
+}
+
+static void wordInit (RecWord *p)
+{
+ p->attrSet = 1;
+ p->attrUse = 1016;
+ p->which = Word_String;
+}
+
+static void wordAdd (const RecWord *p)
+{
+ struct it_key key;
+ size_t i;
- if (key_offset + sizeof(*k) + slen >= KEY_BUF_SIZE - 2)
+ if (kused + 1024 > (ptr_top-ptr_i)*sizeof(char*))
key_flush ();
- memcpy (key_buf + key_offset, &x, 1);
- key_offset++;
- memcpy (key_buf + key_offset, k, sizeof(*k));
- key_offset += sizeof(*k);
- memcpy (key_buf + key_offset, str, slen+1);
- key_offset += slen+1;
+ ++ptr_i;
+ key_buf[ptr_top-ptr_i] = (char*)key_buf + kused;
+ kused += index_word_prefix ((char*)key_buf + kused,
+ p->attrSet, p->attrUse);
+ switch (p->which)
+ {
+ case Word_String:
+ for (i = 0; p->u.string[i]; i++)
+ ((char*)key_buf) [kused++] = index_char_cvt (p->u.string[i]);
+ ((char*)key_buf) [kused++] = '\0';
+ break;
+ default:
+ return ;
+ }
+ ((char*) key_buf)[kused++] = ((key_cmd == 'a') ? 1 : 0);
+ key.sysno = key_sysno;
+ key.seqno = p->seqno;
+ memcpy ((char*)key_buf + kused, &key, sizeof(key));
+ kused += sizeof(key);
}
-void text_extract (SYSNO sysno, int cmd, const char *fname)
+static void wordAddAny (const RecWord *p)
{
- FILE *inf;
- struct it_key k;
- int seqno = 1;
- int c;
- char w[256];
-
- log (LOG_DEBUG, "Text extract of %d", sysno);
- k.sysno = sysno;
- inf = fopen (fname, "r");
- if (!inf)
+ if (p->attrSet != 1 || p->attrUse != 1016)
{
- log (LOG_WARN|LOG_ERRNO, "open %s", fname);
- return;
+ RecWord w;
+
+ memcpy (&w, p, sizeof(w));
+ w.attrSet = 1;
+ w.attrUse = 1016;
+ wordAdd (&w);
}
- while ((c=getc (inf)) != EOF)
+ wordAdd (p);
+}
+
+
+#define FILE_READ_BUF 1
+#if FILE_READ_BUF
+static char *file_buf;
+static int file_offset;
+static int file_bufsize;
+
+static void file_read_start (int fd)
+{
+ file_offset = 0;
+ file_buf = xmalloc (4096);
+ file_bufsize = read (fd, file_buf, 4096);
+}
+
+static void file_read_stop (int fd)
+{
+ xfree (file_buf);
+}
+
+static int file_read (int fd, char *buf, size_t count)
+{
+ int l = file_bufsize - file_offset;
+
+ if (count > l)
{
- int i = 0;
- while (i < 254 && c != EOF && isalnum(c))
+ int r;
+ if (l > 0)
+ memcpy (buf, file_buf + file_offset, l);
+ count = count-l;
+ if (count > file_bufsize)
{
- w[i++] = c;
- c = getc (inf);
+ if ((r = read (fd, buf + l, count)) == -1)
+ {
+ logf (LOG_FATAL|LOG_ERRNO, "read");
+ exit (1);
+ }
+ file_bufsize = 0;
+ file_offset = 0;
+ return r;
}
- if (i)
+ file_bufsize = r = read (fd, file_buf, 4096);
+ if (r == -1)
{
- w[i] = 0;
-
- k.seqno = seqno++;
- k.field = 0;
- key_write (cmd, &k, w);
+ logf (LOG_FATAL|LOG_ERRNO, "read");
+ exit (1);
+ }
+ else if (r <= count)
+ {
+ file_offset = r;
+ memcpy (buf + l, file_buf, r);
+ return l + r;
+ }
+ else
+ {
+ file_offset = count;
+ memcpy (buf + l, file_buf, count - l);
+ return count;
}
- if (c == EOF)
- break;
}
- fclose (inf);
+ memcpy (buf, file_buf + file_offset, count);
+ file_offset += count;
+ return count;
}
-
+#else
+static int file_read (int fd, char *buf, size_t count)
+{
+ return read (fd, buf, count);
+}
+#endif
void file_extract (int cmd, const char *fname, const char *kname)
{
- int i;
+ int i, r;
char ext[128];
SYSNO sysno;
char ext_res[128];
const char *file_type;
void *file_info;
+ struct recExtractCtrl extractCtrl;
+ RecType rt;
- log (LOG_DEBUG, "%c %s k=%s", cmd, fname, kname);
for (i = strlen(fname); --i >= 0; )
if (fname[i] == '/')
{
sprintf (ext_res, "fileExtension.%s", ext);
if (!(file_type = res_get (common_resource, ext_res)))
return;
-
+ if (!(rt = recType_byName (file_type)))
+ return;
+ logf (LOG_DEBUG, "%c %s k=%s", cmd, fname, kname);
file_info = dict_lookup (file_idx, kname);
if (!file_info)
{
sysno = sysno_next++;
dict_insert (file_idx, kname, sizeof(sysno), &sysno);
+ lseek (sys_idx_fd, sysno * SYS_IDX_ENTRY_LEN, SEEK_SET);
+ write (sys_idx_fd, file_type, strlen (file_type)+1);
+ write (sys_idx_fd, kname, strlen(kname)+1);
}
else
memcpy (&sysno, (char*) file_info+1, sizeof(sysno));
- if (!strcmp (file_type, "text"))
- text_extract (sysno, cmd, fname);
-}
-
+ if ((extractCtrl.fd = open (fname, O_RDONLY)) == -1)
+ {
+ logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
+ return;
+ }
+ extractCtrl.subType = "";
+ extractCtrl.init = wordInit;
+ extractCtrl.add = wordAddAny;
+#if FILE_READ_BUF
+ file_read_start (extractCtrl.fd);
+#endif
+ extractCtrl.readf = file_read;
+ key_sysno = sysno;
+ key_cmd = cmd;
+ r = (*rt->extract)(&extractCtrl);
+#if FILE_READ_BUF
+ file_read_stop (extractCtrl.fd);
+#endif
+ close (extractCtrl.fd);
+ if (r)
+ logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);
+}