More work on relevance feedback.

[idzebra-moved-to-github.git] / index / extract.c

diff --git a/index/extract.c b/index/extract.c
index 0d905f5..ce4ed96 100644 (file)
--- a/index/extract.c
+++ b/index/extract.c
@@ -1,10 +1,22 @@
 /*
- * 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 1.3  1995-09-04 12:33:41  adam
+ * 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
@@ -30,6 +42,7 @@
 static Dict file_idx;
 static SYSNO sysno_next;
 static int key_fd = -1;
+static int sys_idx_fd = -1;
 static char *key_buf;
 static int key_offset;
 
@@ -40,7 +53,7 @@ void key_open (const char *fname)
         return;
     if ((key_fd = open (fname, O_RDWR|O_CREAT, 0666)) == -1)
     {
-        logf (LOG_FATAL|LOG_ERRNO, "Creat %s", fname);
+        logf (LOG_FATAL|LOG_ERRNO, "open %s", fname);
         exit (1);
     }
     logf (LOG_DEBUG, "key_open of %s", fname);
@@ -50,7 +63,7 @@ void key_open (const char *fname)
         exit (1);
     }
     key_offset = 0;
-    if (!(file_idx = dict_open ("fileidx", 10, 1)))
+    if (!(file_idx = dict_open (FNAME_FILE_DICT, 40, 1)))
     {
         logf (LOG_FATAL, "dict_open fail of %s", "fileidx");
         exit (1);
@@ -60,6 +73,11 @@ void key_open (const char *fname)
         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);
+    }
 }
 
 int key_close (void)
@@ -70,6 +88,7 @@ int key_close (void)
         return 0;
     }
     close (key_fd);
+    close (sys_idx_fd);
     dict_insert (file_idx, ".", sizeof(sysno_next), &sysno_next);
     dict_close (file_idx);
     key_fd = -1;
@@ -113,13 +132,37 @@ void key_write (int cmd, struct it_key *k, const char *str)
     key_offset += sizeof(*k);
 }
 
-void text_extract (SYSNO sysno, int cmd, const char *fname)
+#if !IT_KEY_HAVE_SEQNO
+void key_write_x (struct strtab *t, int cmd, struct it_key *k, const char *str)
+{
+    void **oldinfo;
+
+    if (strtab_src (t, str, &oldinfo))
+        ((struct it_key *) *oldinfo)->freq++;
+    else
+    {
+        *oldinfo = xmalloc (sizeof(*k));
+        memcpy (*oldinfo, k, sizeof(*k));
+        ((struct it_key *) *oldinfo)->freq = 1;
+    }
+}
+#endif
+
+void key_rec_flush (const char *str, void *info, void *data)
+{
+    key_write (*((int*) data), (struct it_key *)info, str);
+    xfree (info);
+}
+
+void text_extract (struct strtab *t, SYSNO sysno, int cmd, const char *fname)
 {
     FILE *inf;
     struct it_key k;
+#if IT_KEY_HAVE_SEQNO
     int seqno = 1;
+#endif
     int c;
-    char w[256];
+    char w[IT_MAX_WORD];
 
     logf (LOG_DEBUG, "Text extract of %d", sysno);
     k.sysno = sysno;
@@ -132,18 +175,24 @@ void text_extract (SYSNO sysno, int cmd, const char *fname)
     while ((c=getc (inf)) != EOF)
     {
         int i = 0;
-        while (i < 254 && c != EOF && isalnum(c))
+        while (i < IT_MAX_WORD-1 && c != EOF && isalnum(c))
         {
-            w[i++] = c;
+            w[i++] = index_char_cvt (c);
             c = getc (inf);
         }
         if (i)
         {
             w[i] = 0;
             
-            k.seqno = seqno++;
+#if IT_KEY_HAVE_FIELD
             k.field = 0;
+#endif
+#if IT_KEY_HAVE_SEQNO
+            k.seqno = seqno++;
             key_write (cmd, &k, w);
+#else
+            key_write_x (t, cmd, &k, w);
+#endif
         }
         if (c == EOF)
             break;
@@ -159,6 +208,7 @@ void file_extract (int cmd, const char *fname, const char *kname)
     char ext_res[128];
     const char *file_type;
     void *file_info;
+    struct strtab *t;
 
     logf (LOG_DEBUG, "%c %s k=%s", cmd, fname, kname);
     for (i = strlen(fname); --i >= 0; )
@@ -181,11 +231,15 @@ void file_extract (int cmd, const char *fname, const char *kname)
     {
         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, kname, strlen(kname)+1);
     }
     else
         memcpy (&sysno, (char*) file_info+1, sizeof(sysno));
+    t = strtab_mk ();
     if (!strcmp (file_type, "text"))
-        text_extract (sysno, cmd, fname);
+        text_extract (t, sysno, cmd, fname);
+    strtab_del (t, key_rec_flush, &cmd);
 }