Elasticsearch源码分析五--调用Lucene查询接口之模糊查询(Fuzzy)

• 简介
• 查询语法
• 源码分析

---

简介

模糊查询是基于编辑距离算法来匹配文档。编辑距离的计算基于我们提供的查询词条和被搜索文档。此查询很占用CPU资源。可以在搜索词的尾部加上字符 “~” 来进行模糊查询。

查询语法

例如，查询语句 “think~” 返回所有包含和 think 类似的关键词的文档。增量因子（boost factor）为0.2。
{
“query” : {
“fuzzy” : {
“title” : {
“value” : “think~”,
“min_similarity” : 0.2
}}}}
min_similarity：指定了一个词条被算作匹配所必须拥有的最小相似度。对字符串字段来说，这个值应该在0到1之间，包含0和1。对于数值型字段，这个值可以大于1，比如查询值是20， 设为3，则可以得到17~23的值。对于日期字段，可以把参数值设为1d、 2d、 1m等，分别表示1天、 2天、 1个月。

源码分析

'''(1)Elasticsearch code'''
public class FuzzyQueryParser implements QueryParser {
public static final String NAME = "fuzzy";

@Override
public Query parse(QueryParseContext parseContext) throws IOException, QueryParsingException {
        XContentParser parser = parseContext.parser();

        XContentParser.Token token = parser.nextToken();
if (token != XContentParser.Token.FIELD_NAME) {
throw new QueryParsingException(parseContext.index(), "[fuzzy] query malformed, no field");
        }
        String fieldName = parser.currentName();

        String value = null;
float boost = 1.0f;
//LUCENE 4 UPGRADE we should find a good default here I'd vote for 1.0 -> 1 edit
        String minSimilarity = "0.5";
int prefixLength = FuzzyQuery.defaultPrefixLength;
int maxExpansions = FuzzyQuery.defaultMaxExpansions;
boolean transpositions = false;
        MultiTermQuery.RewriteMethod rewriteMethod = null;
        token = parser.nextToken();
if (token == XContentParser.Token.START_OBJECT) {
            String currentFieldName = null;
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token == XContentParser.Token.FIELD_NAME) {
                    currentFieldName = parser.currentName();
                } else {
if ("term".equals(currentFieldName)) {
                        value = parser.text();
                    } else if ("value".equals(currentFieldName)) {
                        value = parser.text();
                    } else if ("boost".equals(currentFieldName)) {
                        boost = parser.floatValue();
                    } else if ("min_similarity".equals(currentFieldName) || "minSimilarity".equals(currentFieldName)) {
                        minSimilarity = parser.text();
                    } else if ("prefix_length".equals(currentFieldName) || "prefixLength".equals(currentFieldName)) {
                        prefixLength = parser.intValue();
                    } else if ("max_expansions".equals(currentFieldName) || "maxExpansions".equals(currentFieldName)) {
                        maxExpansions = parser.intValue();
                    } else if ("transpositions".equals(currentFieldName)) {
                      transpositions = parser.booleanValue();
                    } else if ("rewrite".equals(currentFieldName)) {
                        rewriteMethod = QueryParsers.parseRewriteMethod(parser.textOrNull(), null);
                    } else {
throw new QueryParsingException(parseContext.index(), "[fuzzy] query does not support [" + currentFieldName + "]");
                    }
                }
            }
            parser.nextToken();
        } else {
            value = parser.text();
// move to the next token
            parser.nextToken();
        }

if (value == null) {
throw new QueryParsingException(parseContext.index(), "No value specified for fuzzy query");
        }

        Query query = null;
        MapperService.SmartNameFieldMappers smartNameFieldMappers = parseContext.smartFieldMappers(fieldName);
if (smartNameFieldMappers != null) {
if (smartNameFieldMappers.hasMapper()) {
                query = smartNameFieldMappers.mapper().fuzzyQuery(value, minSimilarity, prefixLength, maxExpansions, transpositions);
            }
        }
if (query == null) {
//LUCENE 4 UPGRADE we need to document that this should now be an int rather than a float
int edits = FuzzyQuery.floatToEdits(Float.parseFloat(minSimilarity), 
              value.codePointCount(0, value.length()));

'''构造Lucene的FuzzyQuery对象，参数包括查询Term、编辑距离、匹配的公共前缀长度、查询可被扩展到的最大词条数'''
            query = new FuzzyQuery(new Term(fieldName, value), edits, prefixLength, maxExpansions, transpositions);
        }
if (query instanceof MultiTermQuery) {
            QueryParsers.setRewriteMethod((MultiTermQuery) query, rewriteMethod);
        }
        query.setBoost(boost);

return wrapSmartNameQuery(query, smartNameFieldMappers, parseContext);
    }
}

'''(2)Lucene code'''
'''FuzzyQuery是MultiTermQuery的子类，调用父类的rewrite方法，需要将查询词"think~"重写成think、thinking等索引中存在的词，然后合并这些词的倒排表'''
public class FuzzyQuery extends MultiTermQuery {
   ...
}

public abstract class MultiTermQuery extends Query {
@Override
public final Query rewrite(IndexReader reader) throws IOException {
return rewriteMethod.rewrite(reader, this);
  }

---