* added another auto-fix for collection index inconsitency check

* fixed words size computation for collection index


git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@2368 6c8d7289-2bf4-0310-a012-ef5d649a1542

source/de/anomic/kelondro/kelondroCollectionIndex.java

@@ -315,8 +315,18 @@ public class kelondroCollectionIndex {
         // read the row and define a collection
         kelondroRowSet collection = new kelondroRowSet(this.rowdef, arrayrow.getColBytes(1)); // FIXME: this does not yet work with different rowdef in case of several rowdef.objectsize()
         if (index.order().compare(arrayrow.getColBytes(0), indexrow.getColBytes(idx_col_key)) != 0) {
-            // check if we got the right row; this row is wrong
+            // check if we got the right row; this row is wrong. Fix it:
-            throw new kelondroException(arrayFile(this.path, this.filenameStub, this.loadfactor, chunksize, partitionnumber, serialnumber).toString(), "array contains wrong row '" + new String(arrayrow.getColBytes(0)) + "', expected is '" + new String(indexrow.getColBytes(idx_col_key)) + "'");
+            index.remove(indexrow.getColBytes(idx_col_key)); // the wrong row cannot be fixed
+            // store the row number in the index; this may be a double-entry, but better than nothing
+            kelondroRow.Entry indexEntry = index.row().newEntry();
+            indexEntry.setCol(idx_col_key, arrayrow.getColBytes(0));
+            indexEntry.setColLongB256(idx_col_chunksize, this.rowdef.objectsize());
+            indexEntry.setColLongB256(idx_col_chunkcount, collection.size());
+            indexEntry.setColLongB256(idx_col_indexpos, (long) rownumber);
+            indexEntry.setColLongB256(idx_col_lastread, kelondroRowCollection.daysSince2000(System.currentTimeMillis()));
+            indexEntry.setColLongB256(idx_col_lastwrote, kelondroRowCollection.daysSince2000(System.currentTimeMillis()));
+            index.put(indexEntry);
+            throw new kelondroException(arrayFile(this.path, this.filenameStub, this.loadfactor, chunksize, partitionnumber, serialnumber).toString(), "array contains wrong row '" + new String(arrayrow.getColBytes(0)) + "', expected is '" + new String(indexrow.getColBytes(idx_col_key)) + "', the row has been fixed");
         }
         int chunkcountInArray = collection.size();
         if (chunkcountInArray != chunkcount) {

source/de/anomic/plasma/plasmaWordIndex.java

@@ -379,6 +379,11 @@ public final class plasmaWordIndex extends indexAbstractRI implements indexRI {
     }
 
     public int size() {
+        if (useCollectionIndex)
+            return java.lang.Math.max(collections.size(),
+                    java.lang.Math.max(assortmentCluster.size(),
+                     java.lang.Math.max(backend.size(), ramCache.size())));
+        else
             return java.lang.Math.max(assortmentCluster.size(),
                         java.lang.Math.max(backend.size(), ramCache.size()));
     }

yacy.init

@@ -484,7 +484,7 @@ xpstopw=true
 80_indexing_busysleep=100
 80_indexing_memprereq=2097152
 82_crawlstack_idlesleep=5000
-82_crawlstack_busysleep=0
+82_crawlstack_busysleep=10
 82_crawlstack_memprereq=1048576
 90_cleanup_idlesleep=300000
 90_cleanup_busysleep=300000