Lucene 1.9 多目录搜索的的一个bug

   这个问题解决有一段时间了。刚才在Lucene的群有朋友提问了这个问题。所以方便其他朋友遇到此问题时,好参考下。特贴出来

在需要搜索多个索引目录,多个字段的时候,发现有错误,提示的大概意思是(具体忘记了),关键词已经存在字典中。研究Lucene的源代码发现,是Lucene.net 1.9 rc1的一个bug :

具体代码在Lucene.Net.Search命名空间下的Query类的Combine method :

修改前的代码如下:

public virtual Query Combine(Query[] queries)
  {
   System.Collections.Hashtable uniques = new System.Collections.Hashtable();
   for (int i = 0; i < queries.Length; i++)
   {
    Query query = queries[i];
    BooleanClause[] clauses = null;
    // check if we can split the query into clauses
    bool splittable = (query is BooleanQuery);
    if (splittable)
    {
     BooleanQuery bq = (BooleanQuery) query;
     splittable = bq.IsCoordDisabled();
     clauses = bq.GetClauses();
     for (int j = 0; splittable && j < clauses.Length; j++)
     {
      splittable = (clauses[j].GetOccur() == BooleanClause.Occur.SHOULD);
     }
    }
    if (splittable)
    {
     for (int j = 0; j < clauses.Length; j++)
     {
                        Query tmp = clauses[j].GetQuery();
      uniques.Add(tmp, tmp);
     }
    }
    else
    {
                    uniques.Add(query, query);
    }
   }
   // optimization: if we have just one query, just return it
   if (uniques.Count == 1)
   {
                return (Query) uniques.GetEnumerator().Current;
   }
   System.Collections.IEnumerator it = uniques.GetEnumerator();
   BooleanQuery result = new BooleanQuery(true);
   while (it.MoveNext())
   {
    result.Add((Query) it.Current, BooleanClause.Occur.SHOULD);
   }
   return result;
  }

 

修改后的代码(飘红的地方是更改过的):

public virtual Query Combine(Query[] queries)
  {
   System.Collections.Hashtable uniques = new System.Collections.Hashtable();
   for (int i = 0; i < queries.Length; i++)
   {
    Query query = queries[i];
    BooleanClause[] clauses = null;
    // check if we can split the query into clauses
    bool splittable = (query is BooleanQuery);
    if (splittable)
    {
     BooleanQuery bq = (BooleanQuery) query;
     splittable = bq.IsCoordDisabled();
     clauses = bq.GetClauses();
     for (int j = 0; splittable && j < clauses.Length; j++)
     {
      splittable = (clauses[j].GetOccur() == BooleanClause.Occur.SHOULD);
     }
    }
    if (splittable)
    {
     for (int j = 0; j < clauses.Length; j++)
     {
                        Query tmp = clauses[j].GetQuery();
      uniques.Add(tmp, tmp);
     }
    }
    else
    {
                    //uniques.Add(query, query);

                    //modify by kwklover 2008/1/17 bug fixed for 防止加入多个重复的字典项
                    if (uniques.Contains(query) == false)
                    {
                        uniques.Add(query, query);
                    }

    }
   }
   // optimization: if we have just one query, just return it
   if (uniques.Count == 1)
   {
               //return (Query) uniques.GetEnumerator().Current;

                //modify by kwklover 2008/1/17
                System.Collections.IDictionaryEnumerator iter = uniques.GetEnumerator();
                iter.MoveNext();
                return iter.Value as Query;

   }
   System.Collections.IEnumerator it = uniques.GetEnumerator();
   BooleanQuery result = new BooleanQuery(true);
   while (it.MoveNext())
   {
    result.Add((Query) it.Current, BooleanClause.Occur.SHOULD);
   }
   return result;
  }

 

另外在Lucene 2.0版已经修复了此问题。不过Lucene 2.0 和 Lucene 1.9的 API有比较大的变化

posted @ 2008-03-11 14:41 kwklover 阅读(...) 评论(...) 编辑 收藏