lucene-搜索过程源码解析-Weight树

最新推荐文章于 2023-12-12 10:13:46 发布

flyingpig4

最新推荐文章于 2023-12-12 10:13:46 发布

阅读量2.3k

点赞数

分类专栏：搜索引擎：Lucene|solr 文章标签： query iterator filter search string

本文链接：https://blog.csdn.net/flyingpig4/article/details/6634823

版权

1.总述  IndexSearch全过程源码分析--->生成weight树:负责的是query权重及生成score树
IndexSearch ---> search(createWeight(query), filter, n, sort)
	|--createWeight(query)  |实际为生成Weight树
		 |--return query.weight(this);
	
2.创建weight树总过程
query.weight(this)
	|--Query query = searcher.rewrite(this);  |重新解析Query，将Query生成为单个可供直接查询的Query
	|--Weight weight = query.createWeight(searcher);
	|--float sum = weight.sumOfSquaredWeights();  |计算sum分值
	|--float norm = getSimilarity(searcher).queryNorm(sum); |获取标准因子
	|--weight.normalize(norm);   |标准化
	|--return weight;  |返回weight权值树

3.重写Query对象，生成Query树
IndexSearcher.rewrite(Query original) |重写Query对象，主要实现拆分
	|--for (Query rewrittenQuery = query.rewrite(reader); rewrittenQuery != query; rewrittenQuery = query.rewrite(reader)) |重写直至不能再拆分
		 |--query = rewrittenQuery;
		 		|--eg1:BooleanQuery.rewrite(reader) 实现
		 		 	 |--for (int i = 0 ; i < clauses.size(); i++)