欢迎您访问程序员文章站本站旨在为大家提供分享程序员计算机编程知识!
您现在的位置是: 首页

Elasticsearch增删改查

程序员文章站 2022-07-05 07:57:34
...

面向文档

document数据格式

  1. 应用系统的数据结构都是面向对象的,复杂的
  2. 对象数据存储到数据库中,只能拆解开来,变为扁平的多张表,每次查询的时候还得还原回对象格式,相当麻烦
  3. ES是面向文档的,文档中存储的数据结构,与面向对象的数据结构是一样的,基于这种文档数据结构,es可以提供复杂的索引,全文检索,分析聚合等功能
  4. es的document用json数据格式来表达

Java数据

public class Employee {

  private String email;
  private String firstName;
  private String lastName;
  private EmployeeInfo info;
  private Date joinDate;

}

private class EmployeeInfo {
  
  private String bio; // 性格
  private Integer age;
  private String[] interests; // 兴趣爱好

}
EmployeeInfo info = new EmployeeInfo();
info.setBio("curious and modest");
info.setAge(30);
info.setInterests(new String[]{"bike", "climb"});

Employee employee = new Employee();
employee.setEmail("[email protected]");
employee.setFirstName("san");
employee.setLastName("zhang");
employee.setInfo(info);
employee.setJoinDate(new Date());

 数据库数据

employee
id 	email 	first_name 	last_name 	join_date
001 	[email protected] 	san 	zhang 	2017/01/01
employee_info
employee_id 	bio 	age 	interests
001 	curious and modest 	30 	bike, climb

 Json数据

{
    "email":      "[email protected]",
    "first_name": "san",
    "last_name": "zhang",
    "info": {
        "bio":         "curious and modest",
        "age":         30,
        "interests": [ "bike", "climb" ]
    },
    "join_date": "2017/01/01"
}

 集群管理

GET /_cat/health?v

 

green:每个索引的primary shard和replica shard都是active状态的
yellow:每个索引的primary shard都是active状态的,但是部分replica shard不是active状态,处于不可用的状态
red:不是所有索引的primary shard都是active状态的,部分索引有数据丢失了

现在只启动动了一个es进程,相当于就只有一个node。现在es中有一个index,就是kibana自己内置建立的index。由于默认的配置是给每个index分配5个primary shard和5个replica shard,而且primary shard和replica shard不能在同一台机器上(为了容错)。现在kibana自己建立的index是1个primary shard和1个replica shard。当前就一个node,所以只有1个primary shard被分配了和启动了,但是一个replica shard没有第二台机器去启动。只要启动第二个es进程,就会在es集群中有2个node,然后那1个replica shard就会自动分配过去,然后cluster status就会变成green状态。

新增

#语法
PUT /index/type/id
{
  "json数据"
}
# 添加商品1
PUT /ecommerce/product/1
{
    "name" : "gaolujie yagao",                        #商品名称
    "desc" :  "gaoxiao meibai",                       #商品描述
    "price" :  30,								   #商品价格
    "producer" :      "gaolujie producer",            #生厂厂家
    "tags": [ "meibai", "fangzhu" ]                   #产品标签
}
#添加商品2
PUT /ecommerce/product/2
{
    "name" : "jiajieshi yagao",
    "desc" :  "youxiao fangzhu",
    "price" :  25,
    "producer" :      "jiajieshi producer",
    "tags": [ "fangzhu" ]
}
#添加商品3
PUT /ecommerce/product/3
{
    "name" : "zhonghua yagao",
    "desc" :  "caoben zhiwu",
    "price" :  40,
    "producer" :      "zhonghua producer",
    "tags": [ "qingxin" ]
}

 es会自动建立index和type,不需要提前创建,而且es默认会对document每个field都建立倒排索引,让其可以被搜索

查询

#语法
GET /index/type/id
GET /ecommerce/product/1

 

{
  "_index": "ecommerce",
  "_type": "product",
  "_id": "1",
  "_version": 1,
  "found": true,
  "_source": {
    "name": "gaolujie yagao",
    "desc": "gaoxiao meibai",
    "price": 30,
    "producer": "gaolujie producer",
    "tags": [
      "meibai",
      "fangzhu"
    ]
  }
}

 修改

PUT /ecommerce/product/1
{
    "name" : "jiaqiangban gaolujie yagao",
    "desc" :  "gaoxiao meibai",
    "price" :  30,
    "producer" :      "gaolujie producer",
    "tags": [ "meibai", "fangzhu" ]
}

 删除

DELETE /ecommerce/product/1

 查询

query string search

query string search的由来:因为search参数都是以http请求的query string来附带的

{
  "took": 3,
  "timed_out": false,
  "_shards": {
    "total": 5,
    "successful": 5,
    "failed": 0
  },
    "hits": {
    "total": 3,
    "max_score": 1,
    "hits": 
  ......
  {
        "_index": "ecommerce",
        "_type": "product",
        "_id": "3",
        "_score": 1,
        "_source": {
          "name": "zhonghua yagao",
          "desc": "caoben zhiwu",
          "price": 40,
          "producer": "zhonghua producer",
          "tags": [
            "qingxin"
          ]
       ......
}

 

took:耗费了几毫秒
timed_out:是否超时,这里是没有
_shards:数据拆成了5个分片,所以对于搜索请求,会打到所有的primary shard(或者是它的某个replica shard)
hits.total:查询结果的数量,3个document
hits.max_score:score的含义,就是document对于一个search的相关度的匹配分数,越相关,就越匹配,分数也高
hits.hits:包含了匹配搜索的document的详细数据

按售价降序排列

GET /ecommerce/product/_search?q=name:yagao&sort=price:desc

适用场景

适用于临时的在命令行使用一些工具,比如curl,快速的发出请求,来检索想要的信息;如果查询请求很复杂,是很难去构建的在生产环境中,几乎很少使用query string search

query DSL

DSL:Domain Specified Language,特定领域的语言
http request body:请求体,可以用json的格式来构建查询语法,比较方便,可以构建各种复杂的语法,比query string search肯定强大多了

查询所有

GET /ecommerce/product/_search
{
  "query": { "match_all": {} }
}

条件查询

查询名称包含yagao的商品,同时按照价格降序排序

GET /ecommerce/product/_search
{
    "query" : {
        "match" : {
            "name" : "yagao"
        }
    },
    "sort": [
        { "price": "desc" }
    ]
}

 分页查询

GET /ecommerce/product/_search
{
  "query": { "match_all": {} },
  "from": 1,
  "size": 1
}

 指定查询

更加适合生产环境的使用,可以构建复杂的查询

GET /ecommerce/product/_search
{
  "query": { "match_all": {} },
  "_source": ["name", "price"]
  }

 query filter

过滤查询

搜索商品名称包含yagao,而且售价大于25元的商品

GET /ecommerce/product/_search
{
    "query" : {
        "bool" : {
            "must" : {
                "match" : {
                    "name" : "yagao" 
                }
            },
            "filter" : {
                "range" : {
                    "price" : { "gt" : 25 } 
                }
            }
        }
    }
}

 full-text search(全文检索)

GET /ecommerce/product/_search
{
    "query" : {
        "match" : {
            "producer" : "yagao producer"
        }
    }
}

 

producer这个字段,会先被拆解,建立倒排索引

special   4
yagao   4
producer 1,2,3,4  
gaolujie 1  
zhognhua 3  
jiajieshi 2

yagao producer 会被拆解为 yagao和producer

phrase search(短语搜索)

跟全文检索相对应,相反,全文检索会将输入的搜索串拆解开来,去倒排索引里面去一一匹配,只要能匹配上任意一个拆解后的单词,就可以作为结果返回
phrase search,要求输入的搜索串,必须在指定的字段文本中,完全包含一模一样的,才可以算匹配,才能作为结果返回

GET /ecommerce/product/_search
{
    "query" : {
        "match_phrase" : {
            "producer" : "yagao producer"
        }
    }
}

 highlight search(高亮搜索结果)

GET /ecommerce/product/_search
{
    "query" : {
        "match" : {
            "producer" : "producer"
        }
    },
    "highlight": {
        "fields" : {
            "producer" : {}
        }
    }
}