Changeset 2146

Timestamp:

2001-03-09T16:56:35+13:00 (23 years ago)

Author:

sjboddie

Message:

Fixed a bug that was preventing phrase searching from working on
document level indexes

File:

• trunk/gsdl/src/colservr/phrasesearch.cpp (modified) (5 diffs)

trunk/gsdl/src/colservr/phrasesearch.cpp

@@ -78 +78 @@
 }
 
+static void get_all_docnums (gdbmclass &gdbm, text_t OID, vector<int> &docnum_list) {
+
+  infodbclass OID_info;
+  
+  // get OID
+  if (!gdbm.getinfo (OID, OID_info)) return;
+  if (OID_info["hastxt"] == "1" && !OID_info["docnum"].empty()) {
+    docnum_list.push_back (OID_info["docnum"].getint());
+  }
+
+  // get contents set
+  if (OID_info["contains"].empty()) return;
+  text_tarray contains; text_t tmptext;
+  text_t::iterator contains_here = OID_info["contains"].begin();
+  text_t::iterator contains_end = OID_info["contains"].end();
+  while (contains_here != contains_end) {
+    if (*contains_here == '"') tmptext += OID;
+    else if (*contains_here == ';') {
+      if (!tmptext.empty()) contains.push_back (tmptext);
+      tmptext.clear();
+    } else tmptext.push_back(*contains_here);
+    contains_here++;
+  }
+  if (!tmptext.empty()) contains.push_back (tmptext);
+
+  text_tarray::const_iterator here = contains.begin();
+  text_tarray::const_iterator end = contains.end();
+  while (here != end) {
+    get_all_docnums (gdbm, *here, docnum_list);
+    here ++;
+  }
+}
 
 bool doc_phrase_search (unsigned char *doc, int doclen,
@@ -102 +134 @@
   while (doc_here <= doc_end) {
     first = true;
-    
+
     // there will be at least one member of phrase (see above)
     termfreqclassarray::const_iterator phrase_here = phrase.begin();
@@ -126 +158 @@
   return false;
 }
-
 
 // looks for the stemmed phrase in the metadata or text associated with
@@ -139 +170 @@
             const termfreqclassarray &phrase,
             int docnum) {
-  // disect the long index to find out where the text should come from
-  text_t level, gran;
-  text_t::const_iterator longindex_here = longindex.begin();
-  text_t::const_iterator longindex_end = longindex.end();
-  longindex_here = getdelimitstr (longindex_here, longindex_end, ':', level);
-  longindex_here = getdelimitstr (longindex_here, longindex_end, ':', gran);
-
-  if (gran.empty()) return false;
-  
-  // note that we're treating indexes with granularity of 'all' (i.e. text,Title,Creator)
-  // as if they were simply 'text' indexes
-  if (gran == "text" || gran == "all" || findword(gran.begin(),gran.end(),"text")) {
-    char *doc = NULL;
-    int doclen = 0;
-  
-    // get text from mg.
-    if (!mgsearch.mgdocument (index, subcollection, language, collection,
-                  docnum, doc, doclen)) return false;
-    return doc_phrase_search ((unsigned char *)doc, doclen, phrase);
-  } 
 
   // get OID
-  char *metadata = NULL;
-  text_t::size_type metadata_len = 0;
   infodbclass docnum_info;
-  infodbclass OID_info;
-  
   if (!gdbm.getinfo (docnum, docnum_info)) return false;
   text_t &OID = docnum_info["section"];
   if (OID.empty()) return false;
+
+  // disect the long index to find out where the text should come from
+  text_t gran, type;
+  text_t::const_iterator longindex_here = longindex.begin();
+  text_t::const_iterator longindex_end = longindex.end();
+  longindex_here = getdelimitstr (longindex_here, longindex_end, ':', gran);
+  longindex_here = getdelimitstr (longindex_here, longindex_end, ':', type);
+
+  if (gran.empty()) return false;
+  
+  // note that we're treating indexes of type 'all' (i.e. text,Title,Creator)
+  // or other composite indexes that contain "text" as if they were simply 'text' indexes
+  if (type == "text" || type == "all" || findword(type.begin(),type.end(),"text")) {
+    char *doc = NULL;
+    int doclen = 0;
+  
+    // get text from mg.
+    if (gran == "document") {
+
+      // if this is a document level index (which should only happen if
+      // there are no matching indexes with a finer granularity -- see
+      // mgqueryfilterclass::mg_parse_query_params) then we must do the
+      // phrase search on the entire document (i.e. all the sections)
+      // -- this is going to make a slow process even slower
+      vector<int> docnum_list; text_t fulldoc;
+      get_all_docnums (gdbm, OID, docnum_list);
+      vector<int>::const_iterator this_docnum = docnum_list.begin();
+      vector<int>::const_iterator end_docnum = docnum_list.end();
+      while (this_docnum != end_docnum) {
+    if (mgsearch.mgdocument (index, subcollection, language, collection,
+                 *this_docnum, doc, doclen)) {
+      fulldoc.appendcstr (doc);
+    }
+    this_docnum ++;
+      }
+      doc = fulldoc.getcstr();
+      doclen = fulldoc.size();
+      bool rv = doc_phrase_search ((unsigned char *)doc, doclen, phrase);
+      delete doc;
+      return rv;
+
+    } else {
+
+      if (!mgsearch.mgdocument (index, subcollection, language, collection,
+                docnum, doc, doclen)) return false;
+      return doc_phrase_search ((unsigned char *)doc, doclen, phrase);
+    } 
+  }
+
+  char *metadata = NULL;
+  text_t::size_type metadata_len = 0;
+  infodbclass OID_info;
   
   // get field
@@ -174 +233 @@
 
   bool result = false;
-  text_tarray *tarr_ptr = OID_info.getmultinfo (gran);
+  text_tarray *tarr_ptr = OID_info.getmultinfo (type);
   if (tarr_ptr != NULL ) {
     text_tarray::const_iterator subvalue_here = (*tarr_ptr).begin();