本章开始将介绍Elasticsearch DSL查询语法的复合查询。
复合查询将其他复合查询或叶子查询进行包装,组合它们的结果和分数,以此改变它们的行为,或从查询字句切换到过滤上下文模式。
主要的复合查询包括如下:
常量(score)评分查询,该复合查询将忽略文档本身的匹配相关性评分,而是统一返回请求参数的boost。
举例说明:(JAVA示例)
public static void testConstantScoreQuery() {
RestHighLevelClient client = EsClient.getClient();
try {
SearchRequest searchRequest = new SearchRequest();
searchRequest.indices("twitter");
SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
sourceBuilder.query(
QueryBuilders.constantScoreQuery(QueryBuilders.wildcardQuery("user", "ding*"))
.boost(1.5f)
);
searchRequest.source(sourceBuilder);
SearchResponse result = client.search(searchRequest, RequestOptions.DEFAULT);
System.out.println(result);
} catch (Throwable e) {
e.printStackTrace();
} finally {
EsClient.close(client);
}
}
返回的结果为:为了对比,左边的结果是QueryBuilders.wildcardQuery(“user”, “ding*”)查询,而右边的是constant_score (复合查询)。
{ {
"took":4, "took":2,
"timed_out":false, "timed_out":false,
"_shards":{ "_shards":{
"total":5, "total":5,
"successful":5, "successful":5,
"skipped":0, "skipped":0,
"failed":0 "failed":0
}, },
"hits":{ "hits":{
"total":1, "total":1,
"max_score":0.9808292, "max_score":1.5,
"hits":[ "hits":[
{ {
"_index":"twitter", "_index":"twitter",
"_type":"_doc", "_type":"_doc",
"_id":"12", "_id":"12",
"_score":0.9808292, "_score":1.5,
"_source":{ "_source":{
"post_date":"2009-11-18T14:12:12", "post_date":"2009-11-18T14:12:12",
"message":"test bulk", "message":"test bulk",
"user":"dingw" "user":"dingw"
} }
} }
] ]
} }
} }
布尔查询。bool query里能包含的主要字句类型如下:
字句类型 | 描述 |
---|---|
must | 该字句类型的查询语句,文档必须满足,并对评分产生影响(相关度) |
filter | 子句(查询)必须出现在匹配的文档中。然而,与must不同的是,查询的分数将被忽略。过滤器子句在过滤器上下文中执行,子句被考虑用于缓存。 |
should | 应该匹配;如果没有must和filter,多个should只需要至少一个匹配即可,该数据可以通过参数minimum_should_match控制,如果包含了must或filter,则should不参与实际过滤,但会参与评分。 |
must_not | 查询条件取反,及匹配到的文档必须不符合must_not的条件。 |
在过滤上下文环境的查询字句并不会对相关性产生影响,也就是说过滤上下文中的查询字句,返回的score为0。
例如如下查询示例(使用Java编写):
public static void testBoolQuery_filterContext_score() {
RestHighLevelClient client = EsClient.getClient();
try {
SearchRequest searchRequest = new SearchRequest();
searchRequest.indices("twitter");
SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
sourceBuilder.query(
QueryBuilders.boolQuery()
.filter(QueryBuilders.termQuery("user", "dingw"))
);
searchRequest.source(sourceBuilder);
SearchResponse result = client.search(searchRequest, RequestOptions.DEFAULT);
System.out.println(result);
} catch (Throwable e) {
e.printStackTrace();
} finally {
EsClient.close(client);
}
}
其返回的结果,其score都为0,结果如下:
{
"took":4,
"timed_out":false,
"_shards":{
"total":5,
"successful":5,
"skipped":0,
"failed":0
},
"hits":{
"total":3,
"max_score":0,
"hits":[
{
"_index":"twitter",
"_type":"_doc",
"_id":"22",
"_score":0,
"_source":{
"post_date":"2018-10-31T14:12:10",
"message":"ab and hell",
"user":"dingw"
}
},
{
"_index":"twitter",
"_type":"_doc",
"_id":"12",
"_score":0,
"_source":{
"post_date":"2009-11-18T14:12:12",
"message":"test bulk",
"user":"dingw"
}
},
{
"_index":"twitter",
"_type":"_doc",
"_id":"11",
"_score":0,
"_source":{
"post_date":"2009-11-19T14:12:12",
"message":"test bulk update",
"user":"dingw"
}
}
]
}
}
该查询方式将所有查询字句进行联合查询(union),只需要其中一个条件匹配匹配文档,但在计算相关性时不是将所有条件的匹配度(score)相加,而是使使用评分最高的查询条件的score,如果有指定tie_breaker的话,则为最大score 加上 其他score * tie_breaker。dis max query是实现(match query multi fields best_fields)的核心。每个查询,可以指定其评分因子(权重、boost),dis max query使用示例:
/**
* dis max query
*/
public static void testDisMaxQuery() {
RestHighLevelClient client = EsClient.getClient();
try {
SearchRequest searchRequest = new SearchRequest();
searchRequest.indices("twitter");
SearchSourceBuilder sourceBuilder = new SearchSourceBuilder();
sourceBuilder.query(
QueryBuilders.disMaxQuery()
// .tieBreaker(0.1f)
.add(QueryBuilders.termQuery("user", "dingw").boost(1.2f))
.add(QueryBuilders.termQuery("message", "bulk"))
);
searchRequest.source(sourceBuilder);
SearchResponse result = client.search(searchRequest, RequestOptions.DEFAULT);
System.out.println(result);
} catch (Throwable e) {
e.printStackTrace();
} finally {
EsClient.close(client);
}
}
待研究。
boosting query可以用来提升或降低某些查询条件的权重。举例如下:
GET /_search
{
"query": {
"boosting" : {
"positive" : { // @1
"term" : {
"field1" : "value1"
}
},
"negative" : { // @2
"term" : {
"field2" : "value2"
}
},
"negative_boost" : 0.2
}
}
}
代码@1:积极的作用,提升其权重。
代码@2:负面的影响,希望降低其权重,其权重值通过negative_boost指定。
复合查询就介绍到这里了。