Bounds check terms based in ICU norm

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

diff --git a/index/extract.c b/index/extract.c
index 4f75f77..a93373c 100644 (file)
--- a/index/extract.c
+++ b/index/extract.c
@@ -149,69 +149,10 @@ static void snippet_add_complete_field(RecWord *p, int ord,
                                        zebra_map_t zm)
 {
     struct snip_rec_info *h = p->extractCtrl->handle;
-
-    const char *b = p->term_buf;
-    char buf[IT_MAX_WORD+1];
-    const char **map = 0;
-    int i = 0, remain = p->term_len;
-    const char *start = b;
-    const char *last = 0;
-
-    if (remain > 0)
-       map = zebra_maps_input(zm, &b, remain, 1);
-
-    while (remain > 0 && i < IT_MAX_WORD)
-    {
-       while (map && *map && **map == *CHR_SPACE)
-       {
-           remain = p->term_len - (b - p->term_buf);
-
-            if (i == 0)
-                start = b;  /* set to first non-ws area */
-           if (remain > 0)
-           {
-               int first = i ? 0 : 1;  /* first position */
-
-               map = zebra_maps_input(zm, &b, remain, first);
-           }
-           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 (**map == *CHR_CUT)
-           {
-               i = 0;
-           }
-           else
-           {
-               if (i >= IT_MAX_WORD)
-                   break;
-               while (i < IT_MAX_WORD && *cp)
-                   buf[i++] = *(cp++);
-           }
-            last = b;
-           remain = p->term_len  - (b - p->term_buf);
-           if (remain > 0)
-           {
-               map = zebra_maps_input(zm, &b, remain, 0);
-           }
-           else
-               map = 0;
-       }
-    }
-    if (!i)
-       return;
-    if (last && start != last && zebra_maps_is_index(zm))
+    if (p->term_len && p->term_buf && zebra_maps_is_index(zm))
         zebra_snippets_appendn(h->snippets, p->seqno, 0, ord,
-                               start, last - start);
+                               p->term_buf, p->term_len);
+    p->seqno++;
 }
 
 static void snippet_add_incomplete_field(RecWord *p, int ord, zebra_map_t zm)
@@ -229,8 +170,7 @@ static void snippet_add_incomplete_field(RecWord *p, int ord, zebra_map_t zm)
 
     while (map)
     {
-       char buf[IT_MAX_WORD+1];
-       int i, remain;
+       int remain;
 
        /* Skip spaces */
        while (map && *map && **map == *CHR_SPACE)
@@ -248,17 +188,10 @@ static void snippet_add_incomplete_field(RecWord *p, int ord, zebra_map_t zm)
         {
             zebra_snippets_appendn(h->snippets, p->seqno, 1, ord,
                                    start, last - start);
-
         }
         start = last;
-
-       i = 0;
        while (map && *map && **map != *CHR_SPACE)
        {
-           const char *cp = *map;
-
-           while (i < IT_MAX_WORD && *cp)
-               buf[i++] = *(cp++);
            remain = p->term_len - (b - p->term_buf);
             last = b;
            if (remain > 0)
@@ -266,8 +199,8 @@ static void snippet_add_incomplete_field(RecWord *p, int ord, zebra_map_t zm)
            else
                map = 0;
        }
-       if (!i)
-           return;
+        if (start == last)
+            return ;
 
         if (first)
         {   
@@ -344,7 +277,6 @@ void extract_snippet(ZebraHandle zh, zebra_snippets *sn,
 {
     struct recExtractCtrl extractCtrl;
     struct snip_rec_info info;
-    int r;
 
     extractCtrl.stream = stream;
     extractCtrl.first_record = 1;
@@ -367,8 +299,7 @@ void extract_snippet(ZebraHandle zh, zebra_snippets *sn,
 
     extractCtrl.setStoreData = 0;
 
-    r = (*rt->extract)(recTypeClientData, &extractCtrl);
-
+    (*rt->extract)(recTypeClientData, &extractCtrl);
 }
 
 static void searchRecordKey(ZebraHandle zh,
@@ -634,7 +565,6 @@ ZEBRA_RES zebra_extract_file(ZebraHandle zh, zint *sysno, const char *fname,
     char gprefix[128];
     char ext[128];
     char ext_res[128];
-    struct file_read_info *fi = 0;
     const char *original_record_type = 0;
     RecType recType;
     void *recTypeClientData;
@@ -700,7 +630,6 @@ ZEBRA_RES zebra_extract_file(ZebraHandle zh, zint *sysno, const char *fname,
     if (sysno && (action == action_delete || action == action_a_delete))
     {
         streamp = 0;
-        fi = 0;
     }
     else
     {
@@ -1858,6 +1787,7 @@ static void extract_add_complete_field(RecWord *p, zebra_map_t zm)
     if (!i)
        return;
     extract_add_string(p, zm, buf, i);
+    p->seqno++;
 }
 
 static void extract_add_icu(RecWord *p, zebra_map_t zm)
@@ -1868,6 +1798,11 @@ static void extract_add_icu(RecWord *p, zebra_map_t zm)
     zebra_map_tokenize_start(zm, p->term_buf, p->term_len);
     while (zebra_map_tokenize_next(zm, &res_buf, &res_len, 0, 0))
     {
+        if (res_len > IT_MAX_WORD)
+        {
+            yaz_log(YLOG_LOG, "Truncating long term %ld", (long) res_len);
+            res_len = IT_MAX_WORD;
+        }
         extract_add_string(p, zm, res_buf, res_len);
         p->seqno++;
     }
@@ -1893,7 +1828,6 @@ static void extract_token_add(RecWord *p)
 {
     ZebraHandle zh = p->extractCtrl->handle;
     zebra_map_t zm = zebra_map_get_or_add(zh->reg->zebra_maps, p->index_type);
-    WRBUF wrbuf;
 
     if (log_level_details)
     {
@@ -1902,11 +1836,6 @@ static void extract_token_add(RecWord *p)
                 p->index_type, p->index_name, 
                 p->seqno, p->term_len, p->term_buf);
     }
-    if ((wrbuf = zebra_replace(zm, 0, p->term_buf, p->term_len)))
-    {
-        p->term_buf = wrbuf_buf(wrbuf);
-        p->term_len = wrbuf_len(wrbuf);
-    }
     if (zebra_maps_is_icu(zm))
     {
         extract_add_icu(p, zm);