Changeset 16947

Timestamp:

2008-08-21T15:13:41+12:00 (16 years ago)

Author:

mdewsnip

Message:

Changed the Lucene code to use the Greenstone document OIDs directly, instead of creating its own numeric IDs and then mapping them to the Greenstone OIDs in the GDBM file. As well as being simpler and more space and speed efficient (the mapping no longer needs to be stored in the GDBM file, and no lookup needs to be done for each search result), this is another important step along the road to true incremental building.

Files:

• gsdl/trunk/runtime-src/src/colservr/expat_resultset.cpp (modified) (2 diffs)
• gsdl/trunk/runtime-src/src/colservr/lucenequeryfilter.cpp (modified) (1 diff)
• gsdl/trunk/runtime-src/src/colservr/queryinfo.cpp (modified) (4 diffs)
• gsdl/trunk/runtime-src/src/colservr/queryinfo.h (modified) (1 diff)
• indexers/trunk/lucene-gs/src/org/greenstone/LuceneWrapper/GS2LuceneQuery.java (modified) (2 diffs)

gsdl/trunk/runtime-src/src/colservr/expat_resultset.cpp

@@ -91 +91 @@
 
     char* id = get_attribute(attributes,"id");
-    if (id!=NULL) {
-      int id_num = atoi(id);
-    
+    if (id != NULL)
+    {
       docresultclass doc;
       doc.clear();
-      doc.docnum = id_num;
+      doc.docid = id;
       doc.docweight = qrpack_ptr->match_count;
 
@@ -105 +104 @@
       }
 
-      queryresult_ptr->docs.docset[doc.docnum] = doc;
-      queryresult_ptr->docs.docorder.push_back(doc.docnum);
+      queryresult_ptr->docs.docset[doc.docid] = doc;
+      queryresult_ptr->docs.docorder.push_back(doc.docid);
       ++qrpack_ptr->match_count;
 

gsdl/trunk/runtime-src/src/colservr/lucenequeryfilter.cpp

@@ -163 +163 @@
 
   // assemble document results
-  if (need_matching_docs (request.filterResultOptions)) {
-    
+  if (need_matching_docs (request.filterResultOptions))
+  {
+    // Loop through the query results (ordered by ranking)
     int resultnum = 1;
-    ResultDocInfo_t resultdoc;
-    text_t trans_OID;
-    vector<text_t>::iterator docorder_here = queryresults.docs.docorder.begin();
-    vector<text_t>::iterator docorder_end = queryresults.docs.docorder.end();
-
-    // Now handled by Lucene directly
-    //if (endresults == -1) endresults = MAXNUMDOCS;
-
-    while (docorder_here != docorder_end)
+    vector<text_t>::iterator docorder_iterator = queryresults.docs.docorder.begin();
+    while (docorder_iterator != queryresults.docs.docorder.end())
+    {
+      text_t doc_OID = (*docorder_iterator);
+      // logout << "Matching doc OID: " << doc_OID << endl;
+
+      // Make sure this result is in the docset, and either in the request set or the request set is empty
+      docresultmap::iterator doc_result = queryresults.docs.docset.find (doc_OID);
+      if (doc_result != queryresults.docs.docset.end() && (request.docSet.empty() || in_set(request.docSet, doc_OID)))
       {
-        // Now handled by Lucene directly
-        //if (resultnum > endresults) break;
-      
-        // translate the document number
-        if (!translate(db_ptr, *docorder_here, trans_OID))
-          {
-            logout << text_t2ascii
-                   << "warning: could not translate lucene document number \""
-                   << *docorder_here << "\" to OID.\n\n";
-            
-          } 
-        else 
-          {
-            docresultmap::iterator docset_here = queryresults.docs.docset.find (*docorder_here);
-
-            // see if there is a result for this number,
-            // if it is in the request set (or the request set is empty)
-            if (docset_here != queryresults.docs.docset.end() && (request.docSet.empty() || in_set(request.docSet, trans_OID)))
-              {
-                // Now handled by Lucene directly
-                //if (resultnum >= startresults) {
-
-                // add this document
-                resultdoc.OID = trans_OID;
-                resultdoc.result_num = resultnum;
-                resultdoc.ranking = (int)((*docset_here).second.docweight * 10000.0 + 0.5);
-                resultdoc.num_terms_matched = (*docset_here).second.num_query_terms_matched;
-                
-                response.docInfo.push_back (resultdoc);
-                //}
-                ++resultnum;
-              }
-          } // else
-        
-        ++docorder_here;
-      }
-  } // if need matching docs
+    // Add the matching document
+    ResultDocInfo_t resultdoc;
+    resultdoc.OID = doc_OID;
+    resultdoc.result_num = resultnum;
+    resultdoc.ranking = (int)((*doc_result).second.docweight * 10000.0 + 0.5);
+    resultdoc.num_terms_matched = (*doc_result).second.num_query_terms_matched;
+    response.docInfo.push_back (resultdoc);
+
+    resultnum++;
+      }
+
+      docorder_iterator++;
+    }
+  }
   
   // assemble the term results

gsdl/trunk/runtime-src/src/colservr/queryinfo.cpp

@@ -203 +203 @@
 
 void docresultclass::clear () {
+  docid="";
   docnum=-1;
   docweight=0.0;
@@ -219 +220 @@
 
 docresultclass &docresultclass::operator=(const docresultclass &d) {
+  docid = d.docid;
   docnum = d.docnum;
   docweight = d.docweight;
@@ -229 +231 @@
 
 bool operator==(const docresultclass &x, const docresultclass &y) {
-  return ((x.docnum == y.docnum) && (x.docweight == y.docweight) &&
+  return ((x.docid == y.docid) && (x.docnum == y.docnum) && (x.docweight == y.docweight) &&
       (x.num_query_terms_matched == y.num_query_terms_matched) &&
       (x.num_phrase_match == y.num_phrase_match));
@@ -235 +237 @@
 
 bool operator<(const docresultclass &x, const docresultclass &y) {
-  return ((x.docnum < y.docnum) ||
-      ((x.docnum == y.docnum) &&
-       ((x.docweight < y.docweight) ||
-        ((x.docweight == y.docweight) &&
-         ((x.num_query_terms_matched < y.num_query_terms_matched) ||
-          ((x.num_query_terms_matched == y.num_query_terms_matched) &&
-           ((x.num_phrase_match < y.num_phrase_match))))))));
+  return ((x.docid < y.docid) ||
+      ((x.docid == y.docid) &&
+       ((x.docnum < y.docnum) ||
+        ((x.docnum == y.docnum) &&
+         ((x.docweight < y.docweight) ||
+          ((x.docweight == y.docweight) &&
+           ((x.num_query_terms_matched < y.num_query_terms_matched) ||
+        ((x.num_query_terms_matched == y.num_query_terms_matched) &&
+         ((x.num_phrase_match < y.num_phrase_match))))))))));
 }
 

gsdl/trunk/runtime-src/src/colservr/queryinfo.h

@@ -115 +115 @@
 
 struct docresultclass {
+  text_t docid;  // currently used by Lucene only
   int docnum;
   float docweight;

indexers/trunk/lucene-gs/src/org/greenstone/LuceneWrapper/GS2LuceneQuery.java

@@ -203 +203 @@
             Document doc = hits.doc(i - 1);
             int doc_term_freq = ((Integer) doc_term_freq_map.get(new Integer(lucene_doc_num))).intValue();
-            lucene_query_result.addDoc(doc.get("nodeID").trim(), hits.score(i-1), doc_term_freq);
+            lucene_query_result.addDoc(doc.get("docOID").trim(), hits.score(i-1), doc_term_freq);
         }
         }
@@ -221 +221 @@
             Document doc = reader.document(lucene_doc_num);
             int doc_term_freq = ((Integer) doc_term_freq_map.get(new Integer(lucene_doc_num))).intValue();
-            lucene_query_result.addDoc(doc.get("nodeID").trim(), hits.scoreDocs[i-1].score, doc_term_freq);
+            lucene_query_result.addDoc(doc.get("docOID").trim(), hits.scoreDocs[i-1].score, doc_term_freq);
         }
         }