Elasticsearch 5.x 关于term query和match query的认识

时间:2023-03-08 17:15:45
Elasticsearch 5.x 关于term query和match query的认识

http://blog.csdn.net/yangwenbo214/article/details/54142786

一、基本情况

前言:term query和match query牵扯的东西比较多,例如分词器、mapping、倒排索引等。我结合官方文档中的一个实例,谈谈自己对此处的理解

string类型在es5.*分为text和keyword。text是要被分词的,整个字符串根据一定规则分解成一个个小写的term,keyword类似es2.3中not_analyzed的情况。
string数据put到elasticsearch中,默认是text。 NOTE:默认分词器为standard analyzer。”Quick Brown Fox!”会被分解成[quick,brown,fox]写入倒排索引 term query会去倒排索引中寻找确切的term,它并不知道分词器的存在。这种查询适合keyword 、numeric、date
match query知道分词器的存在。并且理解是如何被分词的
总的来说有如下:
- term query 查询的是倒排索引中确切的term
- match query 会对filed进行分词操作,然后在查询 二、测试(1) 准备数据:
POST /termtest/termtype/1
{
"content":"Name"
} POST /termtest/termtype/2
{
"content":"name city"
} 查看数据是否导入
GET /termtest/_search
{
"query":
{
"match_all": {}
}
} 结果:
{
"took": 1,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 2,
"max_score": 1,
"hits": [
{
"_index": "termtest",
"_type": "termtype",
"_id": "",
"_score": 1,
"_source": {
"content": "name city"
}
},
{
"_index": "termtest",
"_type": "termtype",
"_id": "",
"_score": 1,
"_source": {
"content": "Name"
}
}
]
}
} 如上说明,数据已经被导入。该处字符串类型是text,也就是默认被分词了 做如下查询:
POST /termtest/_search
{
"query":{
"term":{
"content":"Name"
}
}
} 结果
{
"took": 1,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 0,
"max_score": null,
"hits": []
}
} 分析结果:因为是默认被standard analyzer分词器分词,大写字母全部转为了小写字母,并存入了倒排索引以供搜索。term是确切查询,
必须要匹配到大写的Name。所以返回结果为空 POST /termtest/_search
{
"query":{
"match":{
"content":"Name"
}
}
} 结果
{
"took": 2,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 2,
"max_score": 0.2876821,
"hits": [
{
"_index": "termtest",
"_type": "termtype",
"_id": "",
"_score": 0.2876821,
"_source": {
"content": "Name"
}
},
{
"_index": "termtest",
"_type": "termtype",
"_id": "",
"_score": 0.25811607,
"_source": {
"content": "name city"
}
}
]
}
} 分析结果: 原因(1):默认被standard analyzer分词器分词,大写字母全部转为了小写字母,并存入了倒排索引以供搜索,
原因(2):match query先对filed进行分词,分词为”name”,再去匹配倒排索引中的term 三、测试(2) 下面是官网实例官网实例
1. 导入数据 PUT my_index
{
"mappings": {
"my_type": {
"properties": {
"full_text": {
"type": "text"
},
"exact_value": {
"type": "keyword"
}
}
}
}
} PUT my_index/my_type/1
{
"full_text": "Quick Foxes!",
"exact_value": "Quick Foxes!"
} 先指定类型,再导入数据 full_text: 指定类型为text,是会被分词
exact_value: 指定类型为keyword,不会被分词
full_text: 会被standard analyzer分词为如下terms [quick,foxes],存入倒排索引
exact_value: 只有[Quick Foxes!]这一个term会被存入倒排索引 做如下查询
GET my_index/my_type/_search
{
"query": {
"term": {
"exact_value": "Quick Foxes!"
}
}
} 结果: {
"took": 1,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 1,
"max_score": 0.2876821,
"hits": [
{
"_index": "my_index",
"_type": "my_type",
"_id": "",
"_score": 0.2876821,
"_source": {
"full_text": "Quick Foxes!",
"exact_value": "Quick Foxes!"
}
}
]
}
} exact_value包含了确切的Quick Foxes!,因此被查询到 GET my_index/my_type/_search
{
"query": {
"term": {
"full_text": "Quick Foxes!"
}
}
}
结果: {
"took": 4,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 0,
"max_score": null,
"hits": []
}
} full_text被分词了,倒排索引中只有quick和foxes。没有Quick Foxes! GET my_index/my_type/_search
{
"query": {
"term": {
"full_text": "foxes"
}
}
} 结果: {
"took": 2,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 1,
"max_score": 0.25811607,
"hits": [
{
"_index": "my_index",
"_type": "my_type",
"_id": "",
"_score": 0.25811607,
"_source": {
"full_text": "Quick Foxes!",
"exact_value": "Quick Foxes!"
}
}
]
}
} full_text被分词,倒排索引中只有quick和foxes,因此查询foxes能成功 GET my_index/my_type/_search
{
"query": {
"match": {
"full_text": "Quick Foxes!"
}
}
} 结果: {
"took": 3,
"timed_out": false,
"_shards": {
"total": 5,
"successful": 5,
"failed": 0
},
"hits": {
"total": 1,
"max_score": 0.51623213,
"hits": [
{
"_index": "my_index",
"_type": "my_type",
"_id": "",
"_score": 0.51623213,
"_source": {
"full_text": "Quick Foxes!",
"exact_value": "Quick Foxes!"
}
}
]
}
} match query会先对自己的query string进行分词。也就是”Quick Foxes!”先分词为quick和foxes。然后在去倒排索引中查询,此处full_text是text类型,被分词为quick和foxes
因此能匹配上。
参考文献:http://blog.csdn.net/yangwenbo214/article/details/54142786