Wow, elasticsearch multi field weight sorting can play like this

background

Readers ask questions ：ES There is no list of the weight ranking of the , Reference ？

I've been in touch with , So I wrote this article , It can be referred to as .

In many complex business scenarios , The rules of sorting will be more complicated , A single descending order , Ascending order can't meet daily needs . however ES Provides a way to weight documents to sort , It still works .

First initialize three test data , Easy to view ：

  

   
{
   
    id: 1,
   
    title: "Java How to learn ",
   
    type: 3,
   
    userId: 1,
   
    tags: [
   
        "java"
   
    ],
   
    textContent: " I have to learn Java",
   
    status: 1,
   
    heat: 80
   
}
   
{
   
    id: 2,
   
    title: "Java How to learn ",
   
    type: 2,
   
    userId: 1,
   
    tags: [
   
        "java"
   
    ],
   
    textContent: " I have to learn Java",
   
    status: 1,
   
    heat: 99
   
}
   
{
   
    id: 3,
   
    title: "Java How to learn ",
   
    type: 1,
   
    userId: 1,
   
    tags: [
   
        "java"
   
    ],
   
    textContent: " I have to learn Java",
   
    status: 1,
   
    heat: 100
   
}
  

type：1 For translation ,2 For reprint ,3 For the original

The requirement is to query userId=1 All articles of , Sort in descending order of heat , But the original type of article should be displayed in the front , Priority over heat .

If we simply sort by heat , So the order must be id by 3（ degree of heat ：100）,2（ degree of heat ：99）,1（ degree of heat ：80） It's arranged like this .

But the original type should be at the front , So the result should be 1（ degree of heat ：80, type ： original ）,3（ degree of heat ：100, type ： translate ）,2（ degree of heat ：99, type ： Reprint ）.

The sorting conditions must be in terms of heat , This is for sure . The only thing that needs to be dealt with is how to put the original type at the top , If only implementation is considered , There are many ways .

such as ： The original type of heat value can be adjusted higher , But what? , The heat value needs a new field , Only for sorting , It shows the user the same heat value as before , This makes sorting simple , Or according to the heat arrangement can achieve the effect .

weightFactorFunction

stay ES In search results _score I believe you are not unfamiliar with this field , This is a ES The score given , We can sort by score , Then improve the score of the original type to achieve the desired effect .

Look directly at Java Code it , adopt FunctionScoreQueryBuilder To build the query .

  

   
@Test
   
public void testSort() {
   
    FunctionScoreQueryBuilder.FilterFunctionBuilder[] filterFunctionBuilders = new FunctionScoreQueryBuilder.FilterFunctionBuilder[]{
   
            new FunctionScoreQueryBuilder.FilterFunctionBuilder(QueryBuilders.termQuery("type", 3), ScoreFunctionBuilders.weightFactorFunction(100)),
   
            new FunctionScoreQueryBuilder.FilterFunctionBuilder(QueryBuilders.termQuery("type", 2), ScoreFunctionBuilders.weightFactorFunction(1)),
   
            new FunctionScoreQueryBuilder.FilterFunctionBuilder(QueryBuilders.matchQuery("type", 1), ScoreFunctionBuilders.weightFactorFunction(1))
   
    };
   
    SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();
   
    BoolQueryBuilder boolQuery = QueryBuilders.boolQuery();
   
    boolQuery.must(QueryBuilders.termQuery("userId", 1));
   
    FunctionScoreQueryBuilder functionScoreQueryBuilder = QueryBuilders.functionScoreQuery(boolQuery, filterFunctionBuilders);
   
    searchSourceBuilder.query(functionScoreQueryBuilder)
   
            .sort("_score", SortOrder.DESC)
   
            .sort("heat", SortOrder.DESC);
   
    SearchRequest searchRequest = new SearchRequest(elasticSearchIndexConfig.getArticleSearchIndexName());
   
    searchRequest.types(EsConstant.DEFAULT_TYPE);
   
    searchRequest.source(searchSourceBuilder);
   

   
    List<ArticleDocument> searchResults = kittyRestHighLevelClient.search(searchRequest, ArticleDocument.class);
   
    searchResults.forEach(doc -> {
   
        System.out.println(doc.getId() + "\t" + doc.getType() + "\t" + doc.getHeat());
   
    });
   
}
  

adopt ScoreFunctionBuilders.weightFactorFunction Set the corresponding weight for the article type , The weight of original article is 100, Everything else 1, In this way, the score of original articles is higher than that of other types of articles .

In order to rank, the score should be ranked first , Then the heat is sorted . We can get the results we want .

scriptFunction

Besides using weightFactorFunction To set the weight , In addition, it introduces a kind of higher flexibility , For more complex sorting scenarios scriptFunction.

scriptFunction It allows us to achieve weights through scripting , Look directly at the code ：

  

   
@Test
   
public void testSort() {
   
    String scoreScript = "if (doc['type'].value == 3) {" +
   
            "   return 100;" +
   
            "} else {" +
   
            "   return 1;" +
   
            "}";
   
    FunctionScoreQueryBuilder.FilterFunctionBuilder[] filterFunctionBuilders = new FunctionScoreQueryBuilder.FilterFunctionBuilder[]{
   
            new FunctionScoreQueryBuilder.FilterFunctionBuilder(QueryBuilders.matchAllQuery(), ScoreFunctionBuilders.scriptFunction(new Script(scoreScript)))
   
    };
   
    SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();
   
    BoolQueryBuilder boolQuery = QueryBuilders.boolQuery();
   
    boolQuery.must(QueryBuilders.termQuery("userId", 1));
   
    FunctionScoreQueryBuilder functionScoreQueryBuilder = QueryBuilders.functionScoreQuery(boolQuery, filterFunctionBuilders);
   
    searchSourceBuilder.query(functionScoreQueryBuilder)
   
            .sort("_score", SortOrder.DESC)
   
            .sort("heat", SortOrder.DESC);
   
    SearchRequest searchRequest = new SearchRequest(elasticSearchIndexConfig.getArticleSearchIndexName());
   
    searchRequest.types(EsConstant.DEFAULT_TYPE);
   
    searchRequest.source(searchSourceBuilder);
   

   
    List<ArticleDocument> searchResults = kittyRestHighLevelClient.search(searchRequest, ArticleDocument.class);
   
    searchResults.forEach(doc -> {
   
        System.out.println(doc.getId() + "\t" + doc.getType() + "\t" + doc.getHeat());
   
    });
   
}
  

scoreScript It's the script that controls the weight , That is, a piece of code （ The default script is groovy）, Is it convenient to do so much .

About author ： Yin Jihuan , Simple technology enthusiasts ,《Spring Cloud Microservices - Full stack technology and case analysis 》, 《Spring Cloud Microservices introduction Actual combat and advanced 》 author , official account Ape world Originator .

I have compiled a complete set of learning materials , Those who are interested can search through wechat 「 Ape world 」, Reply key 「 Learning materials 」 Get what I've sorted out Spring Cloud,Spring Cloud Alibaba,Sharding-JDBC Sub database and sub table , Task scheduling framework XXL-JOB,MongoDB, Reptiles and other related information .