22 match_phrase_prefix实现search-time搜索推荐
构造数据
PUT /my_index1/1 { "content":"hello Jack" } PUT /my_index1/2 { "content":"hello John" } PUT /my_index1/3 { "content":"hello Jose" } PUT /my_index1/4 { "content":"hello Dave" }
查询
GET /my_index1/my_type1/_search { "query": { "match_phrase_prefix": { "content": "hello J" } } }
返回
{ "took": 38, "timed_out": false, "_shards": { "total": 5, "successful": 5, "skipped": 0, "failed": 0 }, "hits": { "total": 3, "max_score": 1.7509375, "hits": [ { "_index": "my_index1", "_type": "my_type1", "_id": "2", "_score": 1.7509375, "_source": { "content": "hello John" } }, { "_index": "my_index1", "_type": "my_type1", "_id": "1", "_score": 1.1507283, "_source": { "content": "hello Jack" } }, { "_index": "my_index1", "_type": "my_type1", "_id": "3", "_score": 1.1507283, "_source": { "content": "hello Jose" } } ] } }
match_phrase_prefix原理跟match_phrase类似,唯一的区别,就是把最后一个term作为前缀去搜索
- hello就是去进行match,搜索对应的doc
- w,会作为前缀,去扫描整个倒排索引,找到所有w开头的doc
- 然后找到所有doc中,即包含hello,又包含w开头的字符的doc
- 根据slop去计算,看在slop范围内,能不能让hello w,正好跟doc中的hello和w开头的单词的position相匹配
- 也可以指定slop,但是只有最后一个term会作为前缀
- max_expansions:指定prefix最多匹配多少个term,超过这个数量就不继续匹配了,限定性能
- 默认情况下,前缀要扫描所有的倒排索引中的term,去查找w打头的单词,但是这样性能太差。可以用max_expansions限定,w前缀最多匹配多少个term,就不再继续搜索倒排索引了。
尽量不要用,因为,最后一个前缀始终要去扫描大量的索引,性能可能会很差
立志如山 静心求实
浙公网安备 33010602011771号