Changeset 1662 for trunk/gsdl

Timestamp:

2000-11-09T15:18:58+13:00 (24 years ago)

Author:

nzdl

Message:

Fixed another small bug in the way phrase searching works and made it
always use a boolean and query when phrase searching. It also now forces
the search onto the index with the finest granularity when it detects
a phrase search is being done. I'm not entirely convinced that this is
best as doing a document level phrase search can now return a whole
bunch of section level documents.

Location:

trunk/gsdl/src/colservr

Files:

• mgqueryfilter.cpp (modified) (6 diffs)
• mgqueryfilter.h (modified) (1 diff)

trunk/gsdl/src/colservr/mgqueryfilter.cpp

@@ -22 +22 @@
  * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  *
- * $Id$
- *
  *********************************************************************/
-
-/*
-   $Log$
-   Revision 1.1  2000/08/02 00:58:27  kjm18
-   mgpp incorporated. the old mgsearchclass and queryfilterclass are changed.
-   Have a base searchclass, from which mgsearchclass and mgppsearchclass inherit.
-   Have a base queryfilterclass, from which mgqueryfilterclass and
-   mgppqueryfilterclass inherit. librarymain in recpt should choose the appropriate
-   type (mg vs mgpp) for each collection.
-
-   Revision 1.21  1999/11/25 02:21:13  sjboddie
-   fixed bug in phrasematch stuff
-
-   Revision 1.20  1999/11/01 22:06:06  sjboddie
-   Added filter option to remove documents not matching a phrase match.
-   This used to be done in the receptionist.
-
-   Revision 1.19  1999/10/19 03:23:40  davidb
-   Collection building support through web pages
-   and internal and external link handling for collection documents
-
-   Revision 1.18  1999/09/22 03:43:18  sjboddie
-   Endresults queryfilter option may now take '-1' for 'all'
-
-   Revision 1.17  1999/09/21 12:01:07  sjboddie
-   added Maxdocs queryfilter option (which may be -1 for 'all')
-
-   Revision 1.16  1999/09/07 04:57:24  sjboddie
-   added gpl notice
-
-   Revision 1.15  1999/08/31 22:47:09  rjmcnab
-   Added matchmode option for some and all.
-
-   Revision 1.14  1999/07/16 03:42:21  sjboddie
-   changed isApprox
-
-   Revision 1.13  1999/07/16 00:17:06  sjboddie
-   got using phrasesearch for post-processing
-
-   Revision 1.12  1999/07/09 02:19:43  rjmcnab
-   Fixed a couple of compiler conflicts
-
-   Revision 1.11  1999/07/08 20:49:44  rjmcnab
-   Added result_num to the ResultDocInto_t structure.
-
-   Revision 1.10  1999/07/07 06:19:46  rjmcnab
-   Added ability to combine two or more independant queries.
-
-   Revision 1.9  1999/07/01 09:29:20  rjmcnab
-   Changes for better reporting of number documents which match a query. Changes
-   should still work as before with older versions of mg.
-
-   Revision 1.8  1999/07/01 03:59:54  rjmcnab
-   reduced MAXDOCS to 200 (more reasonable ???). I also added a virtual
-   method for post-processing the query.
-
-   Revision 1.7  1999/06/30 04:04:13  rjmcnab
-   made stemming functions available from mgsearch and made the stems
-   for the query terms available in queryinfo
-
-   Revision 1.6  1999/06/29 22:06:23  rjmcnab
-   Added a couple of fields to queryinfo to handle a special version
-   of mg.
-
-   Revision 1.5  1999/06/27 22:08:48  sjboddie
-   now check for defaultindex, defaultsubcollection, and defaultlanguage
-   entries in config files
-
-   Revision 1.4  1999/06/16 02:03:25  sjboddie
-   fixed bug in isApprox and set MAXDOCS to always be 500
-
-   Revision 1.3  1999/04/19 23:56:09  rjmcnab
-   Finished the gdbm metadata stuff
-
-   Revision 1.2  1999/04/12 03:45:03  rjmcnab
-   Finished the query filter.
-
-   Revision 1.1  1999/04/06 22:22:09  rjmcnab
-   Initial revision.
-
- */
-
 
 #include "mgqueryfilter.h"
@@ -273 +189 @@
   while (query_here != query_end) {
     queryresultsclass thisqueryresults;
-    
+
     if (!mgsearchptr->search(*query_here, thisqueryresults)) {
       // most likely a system problem
@@ -401 +317 @@
   vector<queryparamclass> queryfilterparams;
   parse_query_params (request, queryfilterparams, startresults, 
-              endresults, phrasematch, logout);  
-  
+              endresults, phrasematch, logout);
+  // do any mg specific diddling with query parameters that may be required
+  mg_parse_query_params (request, queryfilterparams, startresults, 
+             endresults, phrasematch, logout);
+
+
   // do query
   queryresultsclass queryresults;
@@ -419 +339 @@
     vector<int>::iterator docorder_end = queryresults.docs.docorder.end();
 
+    // documents containing matching phrases will be sorted to the top so
+    // we can break out once we're past those that match the PhraseMatch
+    // option -- "all_phrases" = return only those documents containing all
+    //                       phrases in query string
+    //           "some_phrases" = return only those documents containing 
+    //                            at least 1 of the phrases in the document
+    //           "all_docs" = return all documents regardless
+    if (num_phrases > 0) {
+      int numdocs = 0;
+      while (docorder_here != docorder_end) {
+    docresultmap::iterator docset_here = queryresults.docs.docset.find (*docorder_here);
+    
+    if (((phrasematch == "all_phrases") && ((*docset_here).second.num_phrase_match < num_phrases)) ||
+        ((phrasematch == "some_phrases") && ((*docset_here).second.num_phrase_match < 1))) {
+      queryresults.docs_matched = numdocs;
+      break;
+    }
+    numdocs ++;
+    docorder_here ++;
+      }
+    }
+
     if (endresults == -1) endresults = MAXNUMDOCS;
+    docorder_here = queryresults.docs.docorder.begin();
     while (docorder_here != docorder_end) {
-      if (resultnum > endresults) break;
+      if (resultnum > endresults || resultnum > queryresults.docs_matched) break;
       
       // translate the document number
@@ -431 +374 @@
       } else {
     docresultmap::iterator docset_here = queryresults.docs.docset.find (*docorder_here);
-    
-    // documents containing matching phrases will be sorted to the top so
-    // we can break out once we're past those that match the PhraseMatch
-    // option -- "all_phrases" = return only those documents containing all
-    //                       phrases in query string
-    //           "some_phrases" = return only those documents containing 
-    //                            at least 1 of the phrases in the document
-    //           "all_docs" = return all documents regardless
-    if (num_phrases > 0) {
-      if ((phrasematch == "all_phrases") && ((*docset_here).second.num_phrase_match < num_phrases)) {
-        queryresults.docs_matched = response.docInfo.size();
-        break;
-      }
-      if ((phrasematch == "some_phrases") && ((*docset_here).second.num_phrase_match < 1)) {
-        queryresults.docs_matched = response.docInfo.size();
-        break;
-      }
-    }
 
     // see if there is a result for this number,
@@ -508 +433 @@
 }
 
+void mgqueryfilterclass::mg_parse_query_params (const FilterRequest_t &/*request*/,
+                        vector<queryparamclass> &query_params,
+                        int &/*startresults*/, int &/*endresults*/,
+                        text_t &/*phrasematch*/, ostream &/*logout*/) {
+
+  //  outconvertclass text_t2ascii;
+  
+  vector<queryparamclass>::iterator query_here = query_params.begin();
+  vector<queryparamclass>::iterator query_end = query_params.end();
+  while (query_here != query_end) {
+
+    // if we're doing a phrase search we want to maximise hits by making it a boolean
+    // search on the index with the finest granularity
+    // we're deciding it's a phrase search based on if the querystring 
+    // contains at least 2 double quotes (not very scientific but 
+    // then neither is the rest of the mg phrase searching functionality :-)
+    if (countchar ((*query_here).querystring.begin(), (*query_here).querystring.end(), '"') > 1) {
+      (*query_here).search_type = 0;
+
+      // Get the long version of the index and test to see if any indexes with
+      // finer granularity exist. Indexes must be the same type (i.e. same metadata
+      // or "text").
+      text_t longindex; text_tarray splitindex;
+      indexmap.to2from ((*query_here).index, longindex);
+      splitchar (longindex.begin(), longindex.end(), ':', splitindex);
+      text_t &granularity = splitindex[0];
+      text_t &indextype = splitindex[1];
+      bool found = false;
+      // currently supported granularity options are "document", "section" and "paragraph"
+      if (granularity == "document" || granularity == "section") {
+    text_t shortindex;
+    if (indexmap.fromexists ("paragraph:" + indextype)) {
+      //      logout << text_t2ascii << "changing index from " << longindex << " to " << ("paragraph:" + indextype) << "\n";
+      indexmap.from2to ("paragraph:" + indextype, shortindex);
+      (*query_here).index = shortindex;
+      found = true;
+    }
+    if (!found && granularity == "document" && indexmap.fromexists ("section:" + indextype)) {
+      //      logout << text_t2ascii << "changing index from " << longindex << " to " << ("section:" + indextype) << "\n";
+      indexmap.from2to ("section:" + indextype, shortindex);
+      (*query_here).index = shortindex;
+    }
+      }
+    }
+
+    query_here ++;
+  }
+}
+    

trunk/gsdl/src/colservr/mgqueryfilter.h

@@ -69 +69 @@
                  docresultsclass &docs);
 
+  void mg_parse_query_params (const FilterRequest_t &request,
+                  vector<queryparamclass> &query_params,
+                  int &startresults, int &endresults,
+                  text_t &phrasematch, ostream &logout);
+
   
 public: