ADDED: yacy.badwords list to filter the topwords

git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@1711 6c8d7289-2bf4-0310-a012-ef5d649a1542

ChangeLog

@@ -5,6 +5,7 @@ version 0.44svn
 	* ADDED: Show public Bookmarks in Bookmarks.html, private ones, if the user is logged in. (Allo)
 	* FIXED: /xml/bookmarks/* now uses one file for private/public entries. private only with password.
 	* ADDED: possibility to get the ranking for a url. (Allo)
+	* ADDED: yacy.badwords list to filter the topwords. (Borg-0300)
 	
 version 0.43
 	* UPDATED: new database handling of index entry objects, less IO overhead (Orbiter)
@@ -19,7 +20,7 @@ version 0.43
 	* UPDATED: Memorysettings now also working for Windows (Galaxis, VT100)
 	* UPDATED: added more informations to network.xml, added possibility to limit number of results (hendi)
 	* FIXED: some Filemodes were set wrong (theli)
-	* FIXED: minor bug-fix in Cache for some rare URLs (borg-0300)
+	* FIXED: minor bug-fix in Cache for some rare URLs (Borg-0300)
 	* ADDED: The YaCy-Bookmarkssystem (allo)
 	* UPDATED: New Searchstyle
 	* FIXED: notifier.gif works even when only DATA is writable (allo)

htroot/index.java

@@ -50,10 +50,12 @@ import java.io.IOException;
 import java.net.MalformedURLException;
 import java.net.URL;
 import java.util.HashMap;
+import java.util.Iterator;
 import java.util.TreeSet;
 
 import de.anomic.http.httpHeader;
 import de.anomic.kelondro.kelondroMSetTools;
+import de.anomic.kelondro.kelondroNaturalOrder;
 import de.anomic.plasma.plasmaSearchRankingProfile;
 import de.anomic.plasma.plasmaSearchTimingProfile;
 import de.anomic.plasma.plasmaSwitchboard;
@@ -67,6 +69,8 @@ import de.anomic.yacy.yacyCore;
 
 public class index {
 
+    public static final int MAX_TOPWORDS = 16;
+
     public static serverObjects respond(httpHeader header, serverObjects post, serverSwitch env) {
         final plasmaSwitchboard sb = (plasmaSwitchboard) env;
 
@@ -223,20 +227,35 @@ public class index {
                 prop.put("num-results_totalcount", totalcount);
                 int hintcount = references.length;
                 if (hintcount > 0) {
-                    if (hintcount > 16) { hintcount = 16; }
+
                     prop.put("combine", 1);
-                    String word;
+
+                    final TreeSet topwords = new TreeSet(kelondroNaturalOrder.naturalOrder);
                     for (int i = 0; i < hintcount; i++) {
-                        word = (String) references[i];
+                        topwords.add(references[i]);
+                    }
+
+                    // filter out the badwords
+                    final TreeSet filteredtopwords = kelondroMSetTools.joinConstructive(topwords, plasmaSwitchboard.badwords);
+                    if (filteredtopwords.size() > 0) {
+                        kelondroMSetTools.excludeDestructive(topwords, plasmaSwitchboard.badwords);
+                    }
+
+                    String word;
+                    hintcount = 0;
+                    final Iterator iter = topwords.iterator();
+                    while (iter.hasNext()) {
+                        word = (String) iter.next();
                         if (word != null) {
-                            prop.put("combine_words_" + i + "_word", word);
-                            prop.put("combine_words_" + i + "_newsearch", post.get("search", "").replace(' ', '+') + "+" + word);
-                            prop.put("combine_words_" + i + "_count", count);
-                            prop.put("combine_words_" + i + "_order", order);
-                            prop.put("combine_words_" + i + "_resource", ((global) ? "global" : "local"));
-                            prop.put("combine_words_" + i + "_time", (searchtime / 1000));
+                            prop.put("combine_words_" + hintcount + "_word", word);
+                            prop.put("combine_words_" + hintcount + "_newsearch", post.get("search", "").replace(' ', '+') + "+" + word);
+                            prop.put("combine_words_" + hintcount + "_count", count);
+                            prop.put("combine_words_" + hintcount + "_order", order);
+                            prop.put("combine_words_" + hintcount + "_resource", ((global) ? "global" : "local"));
+                            prop.put("combine_words_" + hintcount + "_time", (searchtime / 1000));
                         }
-                        prop.put("combine_words", i);
+                        prop.put("combine_words", hintcount);
+                        if (hintcount++ > MAX_TOPWORDS) { break; }
                     }
                 }
             } else {
@@ -312,6 +331,4 @@ public class index {
         return prop;
     }
 
-
-
 }

source/de/anomic/plasma/plasmaSwitchboard.java

@@ -161,8 +161,9 @@ public final class plasmaSwitchboard extends serverAbstractSwitch implements ser
     public static int maxCRGDump = 200000;
     
     // couloured list management
+    public static TreeSet badwords = null;
     public static TreeSet blueList = null;
-    public static TreeSet stopwords = null;
+    public static TreeSet stopwords = null;    
     public static plasmaURLPattern urlBlacklist;
     
     // storage management
@@ -283,7 +284,16 @@ public final class plasmaSwitchboard extends serverAbstractSwitch implements ser
             urlBlacklist.size() + " entries, " +
             ppRamString(ulrBlackListFile.length()/1024));
         }
-        
+
+        // load badwords (to filter the topwords)
+        if (badwords == null) {
+            File badwordsFile = new File(rootPath, "yacy.badwords");
+            badwords = kelondroMSetTools.loadList(badwordsFile, kelondroNaturalOrder.naturalOrder);
+            this.log.logConfig("loaded badwords from file " + badwordsFile.getName() +
+                               ", " + badwords.size() + " entries, " +
+                               ppRamString(badwordsFile.length()/1024));
+        }
+
         // load stopwords
         if (stopwords == null) {
             File stopwordsFile = new File(rootPath, "yacy.stopwords");
@@ -292,7 +302,7 @@ public final class plasmaSwitchboard extends serverAbstractSwitch implements ser
             stopwords.size() + " entries, " +
             ppRamString(stopwordsFile.length()/1024));
         }
-        
+
         // load ranking tables
         File YBRPath = new File(rootPath, "ranking/YBR");
         if (YBRPath.exists()) {

yacy.badwords

@@ -0,0 +1,29 @@
+ads
+ads43
+dir
+jhtml
+shtml
+nach
+auf
+mit
+zum
+bei
+ueber
+all
+app
+ist
+eine
+auch
+aber
+und
+der
+die
+das
+dass
+oder
+from
+kein
+weg
+site=all
+referrer=default
+>>>