Elastic script_score的使用

script_score介绍

在Elasticsearch中,script_score是在function_score查询中的一种功能强大的方式,允许用户使用内置Painless脚本语言或者其他支持的语言来动态计算每个文档的评分

script_score语法

bash 复制代码
GET /<索引名>/_search
{
  "query": {
    "function_score": {
      "query": { "match_all": {} }, // 或者其它查询条件
      "functions": [
        {
          "script_score": {
            "script": {
              "source": """
                double customScore = 0;
                if (doc['field1'].value > params.threshold1) {
                  customScore += doc['field1'].value * params.multiplier1;
                }
                customScore += doc['field2'].value;
                return customScore;
              """,
              "params": {
                "threshold1": 50,
                "multiplier1": 0.5
              }
            }
          }
        }
      ],
      "score_mode": "sum", // 或者其它score_mode
      "boost_mode": "replace" // 或者其它boost_mode
    }
  }
}
  • script_score被用来定义一个脚本,该脚本计算文档的自定义评分
  • source字段内是Painless脚本,它可以访问文档中的字段值(如doc'field1'.value和doc'field2'.value)并对它们进行计算
  • params是一个键值对对象,用于传递给脚本的参数,此处定义了两个参数:threshold1和multiplier1

script_score 案例

场景

假设我们有一个问答论坛索引,需要基于回答数量和点赞数查找高质量

索引创建

bash 复制代码
PUT /forum_questions
{
  "mappings": {
    "properties": {
      "question": {
        "type": "text"
      },
      "answer_count": {
        "type": "long"
      },
      "upvotes": {
        "type": "long"
      }
    }
  }
}

文档插入

bash 复制代码
POST /forum_questions/_doc/
{
  "question": "What is Elasticsearch?",
  "answer_count": 5,
  "upvotes": 20
}

POST /forum_questions/_doc/
{
  "question": "How to configure Elasticsearch for production?",
  "answer_count": 3,
  "upvotes": 15
}


POST /forum_questions/_doc/
{
  "question": "Best practices for indexing data in Elasticsearch?",
  "answer_count": 10,
  "upvotes": 30
}

POST /forum_questions/_doc/
{
  "question": "How to optimize Elasticsearch performance?",
  "answer_count": 8,
  "upvotes": 18
}

POST /forum_questions/_doc/
{
  "question": "What are shards and replicas in Elasticsearch?",
  "answer_count": 6,
  "upvotes": 25
}

POST /forum_questions/_doc/
{
  "question": "How to handle time-based data in Elasticsearch?",
  "answer_count": 4,
  "upvotes": 12
}

POST /forum_questions/_doc/
{
  "question": "What is the difference between match and term queries?",
  "answer_count": 7,
  "upvotes": 23
}

POST /forum_questions/_doc/
{
  "question": "How to set up Elasticsearch clusters?",
  "answer_count": 9,
  "upvotes": 16
}

查询

bash 复制代码
GET /forum_questions/_search
{
  "query": {
    "function_score": {
      "query": { "match_all": {} }, // 或者使用具体查询条件
      "functions": [
        {
          "script_score": {
            "script": {
              "source": """
                double score = 0;
                score += doc['answer_count'].value * params.answer_weight;
                score += doc['upvotes'].value * params.upvote_weight;
                return score;
              """,
              "params": {
                "answer_weight": 0.7,
                "upvote_weight": 0.3
              }
            }
          }
        }
      ],
      "score_mode": "sum"
    }
  }
}
相关推荐
Elasticsearch6 小时前
训练量仅占 0.35%,竞争力却达到 100%:jina-embeddings-v5-omni 背后的冻结塔架构
elasticsearch
扶苏100210 小时前
同一个 Git 项目整出两份,切分支互不影响?两种方案实测
大数据·git·elasticsearch
Elasticsearch12 小时前
用两行 JSON 替换你的 ILM 策略:数据流生命周期新增冻结层支持
elasticsearch
AI大模型-小雄1 天前
Codex写分页接口为什么越翻越慢?用Cursor Pagination解决重复与漏数据
大数据·数据库·elasticsearch·搜索引擎·chatgpt·后端开发·codex
Elasticsearch1 天前
深入浅出 Elastic 工作流(Workflows):核心架构与十一大顶级字段详解
elasticsearch
Elasticsearch1 天前
在 Elasticsearch 中构建上下文:AI Indices 如何使用更少的 tokens 为更智能的 agent 提供支持
elasticsearch
淮北4941 天前
ubuntu22 默认输入法调整频率
运维·服务器·git·ubuntu·elasticsearch
Elasticsearch2 天前
Elasticsearch:使用 AI Agent 来创建 workflows
elasticsearch
gll7732 天前
RAG 检索层实战:Redis 缓存 + ES 混合检索 + BGE-Rerank 重排全链路落地与踩坑
elasticsearch
Elasticsearch2 天前
ES 存日志很贵?我用 ES 9.5 把日志从 4.5G 压到 412M 压缩比11.2倍,日志硬扫每秒128万行!
elasticsearch