我们以一个查询的示例开始,我们在student这个type中存储了一些学生的基本信息,我们分别使用match和match_phrase进行查询。
首先,使用match进行检索,关键字是“He is”:
GET /test/student/_search { "query": { "match": { "description": "He is" } } }
执行这条查询,得到的结果如下:
{ "took": 3, "timed_out": false, "_shards": { "total": 5, "successful": 5, "failed": 0 }, "hits": { "total": 4, "max_score": 0.2169777, "hits": [ { "_index": "test", "_type": "student", "_id": "2", "_score": 0.2169777, "_source": { "name": "februus", "sex": "male", "age": 24, "description": "He is passionate.", "interests": "reading, programing" } }, { "_index": "test", "_type": "student", "_id": "1", "_score": 0.16273327, "_source": { "name": "leotse", "sex": "male", "age": 25, "description": "He is a big data engineer.", "interests": "reading, swiming, hiking" } }, { "_index": "test", "_type": "student", "_id": "4", "_score": 0.01989093, "_source": { "name": "pascal", "sex": "male", "age": 25, "description": "He works very hard because he wanna go to Canada.", "interests": "programing, reading" } }, { "_index": "test", "_type": "student", "_id": "3", "_score": 0.016878016, "_source": { "name": "yolovon", "sex": "female", "age": 24, "description": "She is so charming and beautiful.", "interests": "reading, shopping" } } ] } }
而当你执行match_phrase时:
GET /test/student/_search { "query": { "match_phrase": { "description": "He is" } } }
结果如下:
{ "took": 3, "timed_out": false, "_shards": { "total": 5, "successful": 5, "failed": 0 }, "hits": { "total": 2, "max_score": 0.30685282, "hits": [ { "_index": "test", "_type": "student", "_id": "2", "_score": 0.30685282, "_source": { "name": "februus", "sex": "male", "age": 24, "description": "He is passionate.", "interests": "reading, programing" } }, { "_index": "test", "_type": "student", "_id": "1", "_score": 0.23013961, "_source": { "name": "leotse", "sex": "male", "age": 25, "description": "He is a big data engineer.", "interests": "reading, swiming, hiking" } } ] } }
占的篇幅有点长,但是如果能基于此看清这两者之间的区别,那也是值得的。
我们分析一下这两者结果的差别:
1.非常直观的一点,对于同一个数据集,两者检索出来的结果集数量不一样;
2.对于match的结果,我们可以可以看到,结果的Document中description这个field可以包含“He is”,“He”或者“is”;
3.match_phrase的结果中的description字段,必须包含“He is”这一个词组;
4.所有的检索结果都有一个_score字段,看起来是当前这个document在当前搜索条件下的评分,而检索结果也是按照这个得分从高到低进行排序。
我们要想弄清楚match和match_phrase的区别,要先回到他们的用途:match是全文搜索,也就是说这里的搜索条件是针对这个字段的全文,只要发现和搜索条件相关的Document,都会出现在最终的结果集中,事实上,ES会根据结果相关性评分来对结果集进行排序,这个相关性评分也就是我们看到的_score字段;总体上看,description中出现了“He is”的Document的相关性评分高于只出现“He”或“is”的Document。(至于怎么给每一个Document评分,我们会在以后介绍)。
相关性(relevance)的概念在Elasticsearch中非常重要,而这个概念在传统关系型数据库中是不可想象的,因为传统数据库对记录的查询只有匹配或者不匹配。
那么,如果我们不想将我们的查询条件拆分,应该怎么办呢?这时候我们就可以使用match_phrase:
match_phrase是短语搜索,亦即它会将给定的短语(phrase)当成一个完整的查询条件。当使用match_phrase进行搜索的时候,你的结果集中,所有的Document都必须包含你指定的查询词组,在这里是“He is”。这看起来有点像关系型数据库的like查询操作。