自学内容网 自学内容网

ElasticSearch 搜索、排序、分页功能

一、DSL 查询文档

ElasticSearch 的查询依然是基于 json 风格的 DSL 来实现的。

官方文档:https://www.elastic.co/guide/en/elasticsearch/reference/8.15/query-dsl.html

1.1 DSL 查询分类

常见的查询类型包括:

  • 查询所有:查询出所有数据,一般测试用。如:
    • match_all
  • 全文检索(full text)查询:利用分词器对用户输入内容分词,然后去倒排索引库中匹配,如:
    • match
    • multi_match
  • 精确查询:根据精确词条值查找数据,一般是查找keyword、数值、日期、boolean等类型字段。如:
    • ids
    • range
    • term、terms
  • 地理(geo)查询:根据经纬度查询,如:
    • geo_distance
    • geo_bounding_box
  • 复合(compound)查询:复合查询可以将上述各种查询条件组合起来,合并查询条件。如:
    • bool
    • function_score

1.2 DSL 基本语法

GET /索引库/_search
{
"query":{
"查询类型":{
"查询字段": "值"
}
}
}

查询所有,示例:

GET /hotel/_search
{
"query":{
"match_all":{}
}
}

查询“速8”酒店,示例:

GET /hotel/_search
{
  "query":{
    "match": {
      "name": "速8"
    }
  }
}

二、全文检索查询

2.1 使用场景

  • 商城的输入框搜索
  • 百度输入框搜索

2.2 基本流程

  • 对用户搜索的内容做分词,得到词条
  • 根据词条去倒排索引库中匹配,得到文档id
  • 根据文档id找到文档,返回给用户

说明:因为是拿着词条去匹配,因此参与搜索的字段也必须是可分词的 tex t类型的字段。

2.3 基本语法

常见的全文检索查询包括:

  • match 查询:单字段查询
  • multi_match查询:多字段查询,任意一个字段符合条件就算符合查询条件;字段越多,性能越差。

match 基本语法:

GET /索引名/_search
{
  "query": {
    "match": {
      "FIELD": "TEXT"
    }
  }
}

kibanna 测试示例:

GET /hotel/_search
{
  "query":{
    "match": {
      "name": "速8"
    }
  }
}

# 响应结果:
{"took":4,"timed_out":false,"_shards":{"total":1,"successful":1,"skipped":0,"failed":0},"hits":{"total":{"value":15,"relation":"eq"},"max_score":5.8927264,"hits":[{"_index":"hotel","_id":"1637944903","_score":5.8927264,"_source":{"id":1637944903,"name":"速8酒店北京后海店","address":"西城北京市西城区德胜门内大街兴华胡同五福里2号","price":213,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"后海","location":"39.934452,116.38184","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/48/0C/Cii9EVk1JNuILdBWAAHv5O89TjMAALrFgJ8bwcAAe_8197_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"38609","_score":5.5926995,"_source":{"id":38609,"name":"速8酒店(上海赤峰路店)","address":"广灵二路126号","price":249,"score":35,"brand":"速8","city":"上海","starName":"二钻","business":"四川北路商业区","location":"31.282444,121.479385","pic":"https://m.tuniucdn.com/fb2/t1/G2/M00/DF/96/Cii-TFkx0ImIQZeiAAITil0LM7cAALCYwKXHQ4AAhOi377_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"711837","_score":5.3217444,"_source":{"id":711837,"name":"速8酒店(北京立水桥店)","address":"朝阳安立路3号1幢3层","price":268,"score":36,"brand":"速8","city":"北京","starName":"二钻","business":"亚运村、奥体中心地区","location":"40.043717,116.410962","pic":"https://m2.tuniucdn.com/filebroker/cdn/res/b3/87/b3876eaf16af62521cf6fb474504b8ca_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197516492","_score":5.3217444,"_source":{"id":197516492,"name":"速8酒店(北京南苑东高地店)","address":"丰台南大红门路东营房15号","price":651,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"永定门、南站、大红门、南苑地区","location":"39.78996,116.42081","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/3B/D8/Cii-U1kxKGWIQlaxAAIdkjkSALkAALXDQMFbTsAAh2q158_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197492277","_score":5.075831,"_source":{"id":197492277,"name":"速8酒店(北京平谷兴谷环岛店)","address":"平谷平谷大街31号","price":614,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"平谷城区","location":"40.159255,117.12401","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/38/D5/Cii9EFkwFCiII79zAAHKsXy_LAoAALQuQEmEZ4AAcrJ339_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197496980","_score":5.075831,"_source":{"id":197496980,"name":"速8酒店(北京温都水城王府店)","address":"昌平北七家镇平西府村(温都水城东200米)","price":585,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"小汤山温泉度假区","location":"40.10144,116.380641","pic":"https://m.tuniucdn.com/fb2/t1/G2/M00/C7/CB/Cii-T1km_5eICnpJAAHOWN1GylMAAKYJwF0Hp8AAc5w000_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"47478","_score":4.8516397,"_source":{"id":47478,"name":"速8酒店(上海松江中心店)","address":"松江荣乐东路677号","price":428,"score":35,"brand":"速8","city":"上海","starName":"二钻","business":"佘山、松江大学城","location":"31.016712,121.261606","pic":"https://m.tuniucdn.com/filebroker/cdn/res/07/36/073662e1718fccefb7130a9da44ddf5c_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"5873072","_score":4.8516397,"_source":{"id":5873072,"name":"速8酒店(上海火车站北广场店)","address":"闸北芷江西路796号","price":190,"score":41,"brand":"速8","city":"上海","starName":"二钻","business":"上海火车站地区","location":"31.255579,121.452903","pic":"https://m2.tuniucdn.com/filebroker/cdn/res/96/6d/966d6596e6cb7b48c9cc1d7da79b57c8_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197488318","_score":4.8516397,"_source":{"id":197488318,"name":"速8酒店(北京立水桥地铁南站店)","address":"朝阳北苑路18号院3号楼4层","price":344,"score":36,"brand":"速8","city":"北京","starName":"二钻","business":"亚运村、奥体中心地区","location":"40.043689,116.414138","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/36/4D/Cii9EVkvP72IYYjgAAF7yZeWV-wAALMQACOARMAAXvh983_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"234719728","_score":4.8516397,"_source":{"id":234719728,"name":"速8酒店(北京房山城关店)","address":"房山城关镇城隍庙街10号(原房山老公安局)","price":392,"score":47,"brand":"速8","city":"北京","starName":"二钻","business":"","location":"39.705216,115.981904","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/3F/66/Cii9EFkyeImIB3ZVAAHcTtTFt4oAALdsgICDO0AAdxm378_w200_h200_c1_t0.jpg"}}]}}

mulit_match 基本语法:

GET /indexName/_search
{
  "query": {
    "multi_match": {
      "query": "TEXT",
      "fields": ["FIELD1", " FIELD12"]
    }
  }
}

kibana 测试示例:

GET /hotel/_search
{
  "query":{
    "multi_match": {
      "query": "北京速8",
      "fields": ["name","city"]
    }
  }
}
# 响应结果:
{"took":18,"timed_out":false,"_shards":{"total":1,"successful":1,"skipped":0,"failed":0},"hits":{"total":{"value":67,"relation":"eq"},"max_score":7.23897,"hits":[{"_index":"hotel","_id":"1637944903","_score":7.23897,"_source":{"id":1637944903,"name":"速8酒店北京后海店","address":"西城北京市西城区德胜门内大街兴华胡同五福里2号","price":213,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"后海","location":"39.934452,116.38184","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/48/0C/Cii9EVk1JNuILdBWAAHv5O89TjMAALrFgJ8bwcAAe_8197_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"711837","_score":6.5375423,"_source":{"id":711837,"name":"速8酒店(北京立水桥店)","address":"朝阳安立路3号1幢3层","price":268,"score":36,"brand":"速8","city":"北京","starName":"二钻","business":"亚运村、奥体中心地区","location":"40.043717,116.410962","pic":"https://m2.tuniucdn.com/filebroker/cdn/res/b3/87/b3876eaf16af62521cf6fb474504b8ca_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197516492","_score":6.5375423,"_source":{"id":197516492,"name":"速8酒店(北京南苑东高地店)","address":"丰台南大红门路东营房15号","price":651,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"永定门、南站、大红门、南苑地区","location":"39.78996,116.42081","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/3B/D8/Cii-U1kxKGWIQlaxAAIdkjkSALkAALXDQMFbTsAAh2q158_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197492277","_score":6.235448,"_source":{"id":197492277,"name":"速8酒店(北京平谷兴谷环岛店)","address":"平谷平谷大街31号","price":614,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"平谷城区","location":"40.159255,117.12401","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/38/D5/Cii9EFkwFCiII79zAAHKsXy_LAoAALQuQEmEZ4AAcrJ339_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197496980","_score":6.235448,"_source":{"id":197496980,"name":"速8酒店(北京温都水城王府店)","address":"昌平北七家镇平西府村(温都水城东200米)","price":585,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"小汤山温泉度假区","location":"40.10144,116.380641","pic":"https://m.tuniucdn.com/fb2/t1/G2/M00/C7/CB/Cii-T1km_5eICnpJAAHOWN1GylMAAKYJwF0Hp8AAc5w000_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"197488318","_score":5.960038,"_source":{"id":197488318,"name":"速8酒店(北京立水桥地铁南站店)","address":"朝阳北苑路18号院3号楼4层","price":344,"score":36,"brand":"速8","city":"北京","starName":"二钻","business":"亚运村、奥体中心地区","location":"40.043689,116.414138","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/36/4D/Cii9EVkvP72IYYjgAAF7yZeWV-wAALMQACOARMAAXvh983_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"234719728","_score":5.960038,"_source":{"id":234719728,"name":"速8酒店(北京房山城关店)","address":"房山城关镇城隍庙街10号(原房山老公安局)","price":392,"score":47,"brand":"速8","city":"北京","starName":"二钻","business":"","location":"39.705216,115.981904","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/3F/66/Cii9EFkyeImIB3ZVAAHcTtTFt4oAALdsgICDO0AAdxm378_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"1714520967","_score":5.960038,"_source":{"id":1714520967,"name":"速8酒店(北京安华桥黄寺大街店)","address":"黄寺大街12号院16号楼","price":559,"score":43,"brand":"速8","city":"北京","starName":"二钻","business":"马甸、安贞地区","location":"39.962742,116.388431","pic":"https://m.tuniucdn.com/fb2/t1/G1/M00/4A/21/Cii-U1k1o-uIdcUZAAIbmIKVlKAAALtvQGBb6kAAhuw170_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"706343","_score":5.7079287,"_source":{"id":706343,"name":"速8酒店(北京西客站北广场店)","address":"丰台莲花池东路126号","price":268,"score":39,"brand":"速8","city":"北京","starName":"二钻","business":"北京西站、丽泽商务区","location":"39.896623,116.315586","pic":"https://m.tuniucdn.com/fb2/t1/G2/M00/E3/46/Cii-TlkzMXWIL0sAAAGG8a3YwiwAALJlgG-r5YAAYcJ067_w200_h200_c1_t0.jpg"}},{"_index":"hotel","_id":"38609","_score":5.5926995,"_source":{"id":38609,"name":"速8酒店(上海赤峰路店)","address":"广灵二路126号","price":249,"score":35,"brand":"速8","city":"上海","starName":"二钻","business":"四川北路商业区","location":"31.282444,121.479385","pic":"https://m.tuniucdn.com/fb2/t1/G2/M00/DF/96/Cii-TFkx0ImIQZeiAAITil0LM7cAALCYwKXHQ4AAhOi377_w200_h200_c1_t0.jpg"}}]}}

前面我们将 brand、name、business 值利用 copy_to 复制到了 all 字段中,比较
match + all 与 multi_match brand,name,business查询结果:
在这里插入图片描述
在这里插入图片描述
通过比较,我们发现两次的查询结果是一样的。但是,搜索字段越多,对查询性能影响越大,因此建议采用 copy_to,然后单字段查询的方式。

三、精确查询

精确查询一般是查找 keyword、数值、日期、boolean 等类型字段。所以不会对搜索条件分词。

常见的全文检索查询包括:

  • term:根据词条精确值查询
  • range:根据值的范围查询

3.1 term 查询

说明:

  • 因为精确查询的字段搜是不分词的字段,因此查询的条件也必须是不分词的词条。
  • 用户输入的内容跟文档值完全匹配时才认为符合条件。

基本语法:

GET /索引库/_search
{
  "query":{
    "term": {
      "FIELD": {
        "value": "VALUE"
      }
    }
  }
}

Kibana 测试:
在这里插入图片描述

3.2 range 查询

范围查询,一般应用在对数值、日期类型做范围过滤。

基本语法:

GET /索引库/_search
{
  "query": {
    "range": {
      "FIELD": {
        "gte": 10,
        "lte": 20
      }
    }
  }
}

Kibana 测试:
在这里插入图片描述

四、地理坐标查询

所谓的地理坐标查询,其实就是根据经纬度查询。

常见的使用场景:

  • 搜索我附近的酒店
  • 搜索我附近的出租车
  • 搜索我附近的人

4.1 geo_bounding_box 矩形范围查询

说明:指定矩形的左上、右下两个点的坐标,然后画出一个矩形,落在该矩形内的都是符合条件的点。
基本语法:

GET /hotel/_search
{
  "query": {
    "geo_bounding_box": {
      "FIELD": {
        "top_left": {
          "lat": 40.73,
          "lon": -74.1
        },
        "bottom_right": {
          "lat": 40.717,
          "lon": -73.99
        }
      }
    }
  }

Kibana 测试:
在这里插入图片描述

4.2 geo_distance 距离查询

说明:查询到指定中心点小于某个距离值的所有文档。换句话来说,在地图上找一个点作为圆心,以指定距离为半径,画一个圆,落在圆内的坐标都算符合条件。

基本语法:

GET /hotel/_search
{
  "query": {
    "geo_distance": {
      "distance": "10km", // 半径
      "FIELD": { // 圆心
        "lat": 40.73,// 纬度
        "lon": -74.1// 经度
      }
    }
  }
}

Kibana 测试:
在这里插入图片描述

五、compound 复合查询

说明:复合查询可以将其它简单查询组合起来,实现更复杂的搜索逻辑。常见的有两种:

  • bool 查询:布尔查询,利用逻辑关系组合多个其它的查询,实现复杂搜索
  • function_score 查询:算分函数查询,可以控制文档相关性算分,控制文档排名

相关性算分:
当我们利用match查询时,文档结果会根据与搜索词条的关联度打分(_score),返回结果时按照分值降序排列。
在这里插入图片描述
打分算法是TF-IDF算法(5.1+):
在这里插入图片描述

5.1 bool 查询

布尔查询是一个或多个查询子句的组合,每一个子句就是一个子查询。

子查询的组合方式有:

  • must:必须匹配每个子查询,类似 “与” 查询
  • should:选择性匹配子查询,类似 “或” 查询
  • must_not:必须不匹配,类似 “非” 查询,不参与算分
  • filter:必须匹配,不参与算分

说明:参与打分的字段越多,查询性能越差。

建议:

  • 搜索框的关键字搜索,是全文检索查询,使用 must 查询,参与算分
  • 其它过滤条件,采用filter查询。不参与算分

语法示例:

POST /索引库/_search
{
  "query": {
    "bool" : {
      "must" : {
        "term" : { "user.id" : "kimchy" }
      },
      "filter": {
        "term" : { "tags" : "production" }
      },
      "must_not" : {
        "range" : {
          "age" : { "gte" : 10, "lte" : 20 }
        }
      },
      "should" : [
        { "term" : { "tags" : "env1" } },
        { "term" : { "tags" : "deployed" } }
      ],
      "minimum_should_match" : 1
    }
  }
}

测试示例1:
需求:搜索名字包含“如家”,价格不高于 400,在坐标 31.21,121.5 周围 10km 范围内的酒店。

分析:

  • 名称搜索,属于全文检索查询,应该参与算分。放到 must 中
  • 价格不高于400,用 range 查询,属于过滤条件,不参与算分。放到 must_not 中
  • 周围 10km 范围内,用 geo_distance 查询,属于过滤条件,不参与算分。放到filter中

语法结构:

GET /hotel/_search
{
  "query": {
    "bool": {
      "must": [
        {
          "match": {
            "name": "如家"
          }
        }
      ],
      "must_not": [
        {
          "range": {
            "price": {
              "gt": 400
            }
          }
        }
      ],
      "filter": [
        {
          "geo_distance": {
            "distance": "10km",
            "location": {
              "lat": 31.21,
              "lon": 121.5
            }
          }
        }
      ]
    }
  }
}

Kibana 测试:
在这里插入图片描述

5.2 function_score 算分函数查询

根据相关度打分是比较合理的需求,但合理的不一定是产品经理需要的。以百度为例,你搜索的结果中,并不是相关度越高排名越靠前,而是谁掏的钱多排名就越靠前。要想认为控制相关性算分,就需要利用 elasticsearch 中的 function_score 查询了。

5.2.1 function score 运行流程:

  1. 根据原始条件查询搜索文档,并且计算相关性算分,称为原始算分(query score)
  2. 根据过滤条件,过滤文档
  3. 符合过滤条件的文档,基于算分函数运算,得到函数算分(function score)
  4. 将原始算分(query score)和函数算分(function score)基于运算模式做运算,得到最终结果,作为相关性算分

5.2.2 相关性算分
当我们利用 match 查询时,文档结果会根据与搜索词条的关联度打分(_score),返回结果时按照分值降序排列。

例如,我们搜索 “如家”,结果如下:
在这里插入图片描述

5.2.3 ElasticSearch(5.1+) BM25 算法公式如下
在这里插入图片描述
5.2.4 function_score 查询语法:

  • query:原始查询条件,基于该条件搜索文档,并基于 BM25 算法为文档打分,所得结果即为原始分

  • functions: 算分函数,算分函数的结果称为 function score,结果将与 query score 运算,得到新算分。

    • 常见的算分函数有:

      • weight:给一个常量值,作为函数结果
      • field_value_factor:用文档中的某个字段值,作为函数结果
      • random_score:随机生成一个值,作为函数结果
      • script_score:自定义计算公式,作为函数结果
    • filter:过滤条件,符合该条件的文档才会重新算分

  • boost_mode:运算模式,算分函数的结果,原始查询的相关性算分,两者之间的运算方式。

    • 常见的运算方式有:
      - multiply:两者相乘(默认)
      - replace:用 function score 替代 query score
      - 其他: sum、avg、max、min

5.2.5 基本语法格式
其他可选参数,可访问官网进行学习。官网地址:https://www.elastic.co/guide/en/elasticsearch/reference/current/query-dsl-function-score-query.html

GET /_search
{
  "query": {
    "function_score": {
      "query": { "match_all": {} },
      "functions": [
        {
          "filter": { "match": { "test": "cat" } },
          "weight": 42
        }
      ],
      "boost_mode": "multiply"
    }
  }
}

5.2.6 测试示例
需求:给价格小于300品牌为“如家”的酒店,排名靠前一些。
整理:

  • 原始条件:价格小于300
  • 过滤条件:品牌为“如家”
  • 算法函数:按需指定,本次就直接固定加分(weight)
  • 运算模式:求和

直接查询查询价格小于300的酒店:

在这里插入图片描述
从图上我们可以看出,直接查询“速8”的酒店靠前。

使用算法函数查询:

定义 DSL 语法:

GET /hotel/_search
{
  "query": {
    "function_score": {
      "query": {
        "range": {
          "price": {
            "lte": 300
          }
        }
      },
      "functions": [
        {
          "filter": {
            "term": {
              "brand": "如家"
            }
          },
          "weight": 2
        }
      ],
      "boost_mode": "sum"
    }
  }
}

测试结果:

在这里插入图片描述
从图中我们可以看出,“如家”酒店已经靠前,且算分结果也是正常加了2分。

六、排序

ElasticSearch 默认是根据相关度算分(_score)来排序的,但是也支持自定义方式对搜索结果排序。可以排序的字段类型有:keyword 类型、数值类型、地理坐标类型、日期类型等。

6.1 普通字段排序

DSL 语法

GET /hotel/_search
{
  "query": {
    "match_all": {}
  },
  "sort": [
    {
      "FIELD": {
        "order": "desc"
      }
    }
  ]
}

说明:支持多字段排序,第一个条件相同时,按第二个排序,以此类推。

案例:查询所有酒店,并按照评分高的在前,价格低的在前
定义 DSL 语法

GET /hotel/_search
{
  "query": {
    "match_all": {}
  },
  "sort": [
    {
      "score": {
        "order": "desc"
      }
    },
    {
      "price": {
        "order": "asc"
      }
    }
  ]
}

测试结果
在这里插入图片描述

6.2 地理坐标排序

DSL 语法

GET /hotel/_search
{
  "query": {
    "match_all": {}
  },
  "sort": [
    {
      "_geo_distance": {
        "FIELD": {
          "lat": 40, // 纬度
          "lon": -70 // 经度
        },
        "order": "asc",
        "unit": "km"
      }
    }
  ]
}

案例:根据自己的位置按照酒店离你的位置升序排序
查询定位网址:https://map.bmcx.com/jingwei_dr__map/
定义 DSL 语法

GET /hotel/_search
{
  "query": {
    "match_all": {}
  },
  "sort": [
    {
      "_geo_distance": {
        "location": {
          "lat": 31.264845, // 纬度
          "lon": 121.658846 // 经度
        },
        "order": "asc",
        "unit": "km"
      }
    }
  ]
}

测试结果:
在这里插入图片描述

七、分页

ElasticSearch 默认情况下只返回10条数据,如果要查询更多数据就需要修改分页参数。

  • from:从第几个文档开始,类似 mysql 的 offset
  • size:查询几个文档,类似 mysql 的 limit

7.1 基本语法

DSL 语法

GET /hotel/_search
{
  "query": {
    "match_all": {}
  },
  "sort": [
    {
      "score": {
        "order": "desc"
      }
    }
  ], 
  "from": 0, 
  "size": 10
}

测试结果:
在这里插入图片描述

7.2 深度分页问题

需求:查询 990 ~ 1000的数据,查询 9900 ~ 10000的数据

问题
ElasticSearch 内部分页时,必须先查询 0~1000 条,然后截取其中的 990 ~ 1000的这 10 条数据。
如果 ES 是单点模式,这并无太大影响。如果是多集群部署,我需要根据条件排序后 查询出 1000 数据,假如有5台节点,并不是每个节点取 200条数据,因为 节点1 的 Top 200,在另一个节点可能排到 10000 名以外了。因此,要想获取整个集群的 Top 1000,必须先查询出每个节点的 Top 1000,汇总结果后,重新排名,重新截取 Top 1000。
那如果我要查询 9900~10000 的数据呢?是不是要先查询 Top 10000 呢?每个节点都要查询 10000 条?汇总到内存中?
当查询分页深度较大时,汇总数据过多,对内存和 CPU 会产生非常大的压力,因此 ES 会禁止 from + size 超过 10000 的请求。

解决方案
search after:分页时需要排序,原理是从上一次的排序值开始,查询下一页数据。官方推荐使用的方式。
scroll:原理将排序后的文档id形成快照,保存在内存。官方已经不推荐使用。

总结

  • from + size

    • 优点:支持随机翻页
    • 缺点:深度分页问题,默认查询上限(from + size)是10000
  • after search

    • 优点:没有查询上限(单次查询的size不超过10000)
    • 缺点:只能向后逐页查询,不支持随机翻页
  • scroll(不推荐)

    • 优点:没有查询上限(单次查询的size不超过10000)
    • 缺点:会有额外内存消耗,并且搜索结果是非实时的

八、高亮显示

DSL 基础语法

GET /hotel/_search
{
  "query": {
    "match": {
      "name": "如家" // 必须指定搜索条件
    }
  },
  "sort": [
    {
      "score": {
        "order": "desc"
      }
    },
    {
      "price": {
        "order": "asc"
      }
    }
  ], 
  "from": 0,
  "size":10,
  "highlight":{
    "fields": {
      "name":{
        "pre_tags": "<em>", // 用来标记高亮字段的前置标签,默认:<em>
        "post_tags": "</em>" // 用来标记高亮字段的后置标签,默认:</em>
      }
    }
  }
}

说明:

  • 高亮是对关键字高亮,因此搜索条件必须带有关键字,而不能是范围这样的查询。
  • 默认情况下,高亮的字段,必须与搜索指定的字段一致,否则无法高亮
  • 如果要对非搜索字段高亮,则需要添加一个属性:required_field_match=false

原文地址:https://blog.csdn.net/qq_29627497/article/details/143937493

免责声明:本站文章内容转载自网络资源,如本站内容侵犯了原著者的合法权益,可联系本站删除。更多内容请关注自学内容网(zxcms.com)!