聚合在Elasticsearch中的使用
系统中使用的ES环境不一定每篇文章都有，但是可以在合集中找到，关注《醉鱼Java》一起进步
环境
- elasticsearch 8.1
搭建
version: '3.8' services: cerebro: image: lmenezes/cerebro:0.8.3 container_name: cerebro ports: - "9000:9000" command: - -Dhosts.0.host=http://eshot:9200 networks: - elastic kibana: image: docker.elastic.co/kibana/kibana:8.1.3 container_name: kibana environment: - I18N_LOCALE=zh-CN - XPACK_GRAPH_ENABLED=true - TIMELION_ENABLED=true - XPACK_MONITORING_COLLECTION_ENABLED="true" - ELASTICSEARCH_HOSTS=http://eshot:9200 - server.publicBaseUrl=http://192.168.160.234:5601 ports: - "5601:5601" networks: - elastic eshot: image: elasticsearch:8.1.3 container_name: eshot environment: - node.name=eshot - cluster.name=es-docker-cluster - discovery.seed_hosts=eshot,eswarm,escold - cluster.initial_master_nodes=eshot,eswarm,escold - bootstrap.memory_lock=true - "ES_JAVA_OPTS=-Xms512m -Xmx512m" - xpack.security.enabled=false - node.attr.node_type=hot ulimits: memlock: soft: -1 hard: -1 volumes: - D:\zuiyuftp\docker\es8.1\eshot\data:/usr/share/elasticsearch/data - D:\zuiyuftp\docker\es8.1\eshot\logs:/usr/share/elasticsearch/logs - D:\zuiyuftp\docker\es8.1\eshot\plugins:/usr/share/elasticsearch/plugins ports: - 9200:9200 networks: - elastic eswarm: image: elasticsearch:8.1.3 container_name: eswarm environment: - node.name=eswarm - cluster.name=es-docker-cluster - discovery.seed_hosts=eshot,eswarm,escold - cluster.initial_master_nodes=eshot,eswarm,escold - bootstrap.memory_lock=true - "ES_JAVA_OPTS=-Xms512m -Xmx512m" - xpack.security.enabled=false - node.attr.node_type=warm ulimits: memlock: soft: -1 hard: -1 volumes: - D:\zuiyuftp\docker\es8.1\eswarm\data:/usr/share/elasticsearch/data - D:\zuiyuftp\docker\es8.1\eswarm\logs:/usr/share/elasticsearch/logs - D:\zuiyuftp\docker\es8.1\eshot\plugins:/usr/share/elasticsearch/plugins networks: - elastic escold: image: elasticsearch:8.1.3 container_name: escold environment: - node.name=escold - cluster.name=es-docker-cluster - discovery.seed_hosts=eshot,eswarm,escold - cluster.initial_master_nodes=eshot,eswarm,escold - bootstrap.memory_lock=true - "ES_JAVA_OPTS=-Xms512m -Xmx512m" - xpack.security.enabled=false - node.attr.node_type=cold ulimits: memlock: soft: -1 hard: -1 volumes: - D:\zuiyuftp\docker\es8.1\escold\data:/usr/share/elasticsearch/data - D:\zuiyuftp\docker\es8.1\escold\logs:/usr/share/elasticsearch/logs - D:\zuiyuftp\docker\es8.1\eshot\plugins:/usr/share/elasticsearch/plugins networks: - elastic # volumes: # eshotdata: # driver: local # eswarmdata: # driver: local # escolddata: # driver: local networks: elastic: driver: bridge
什么是聚合？
在Elasticsearch中，聚合是一种功能强大的数据处理技术，它允许我们对索引中的数据进行多种计算和分析操作。聚合可以理解为对数据集进行分组，并在每个分组上执行各种指标计算，类似于SQL中的GROUP BY和聚合函数。
示例数据
为了验证聚合功能，我们将使用一个示例数据集，假设我们有一个存储了商品信息的索引，包含以下字段：
product_name：商品名称category：商品分类price：商品价格quantity：商品数量manufacturer：制造商timestamp：记录时间戳
下面我们导入测试数据
创建索引
PUT /zfc-doc-000001 { "settings": { "index":{ "number_of_shards":3, "number_of_replicas":2 } }, "mappings": { "properties": { "product_name":{ "type":"keyword" }, "category":{ "type":"keyword" }, "price":{ "type": "integer" }, "quantity":{ "type": "integer" }, "manufacturer":{ "type": "keyword" }, "timestamp":{ "type": "date", "format": "yyyy-MM-dd HH:mm:ss||yyyy-MM-dd||epoch_millis" } } } }
添加数据
PUT _bulk {"index":{"_index":"zfc-doc-000002","_id":"1"}} {"product_name": "iPhone 12","category": "Electronics","price": 999,"quantity": 50,"manufacturer": "Apple","timestamp": "2023-07-24 10:00:00"} {"index":{"_index":"zfc-doc-000002","_id":"2"}} {"product_name": "Samsung Galaxy S21","category": "Electronics","price": 799,"quantity": 30,"manufacturer": "Samsung","timestamp": "2023-07-24 11:30:00"} {"index":{"_index":"zfc-doc-000002","_id":"3"}} {"product_name": "Sony Bravia 65-inch TV","category": "Electronics","price": 1499,"quantity": 20,"manufacturer": "Sony","timestamp": "2023-07-24 13:15:00"} {"index":{"_index":"zfc-doc-000002","_id":"4"}} {"product_name": "HP Spectre x360","category": "Electronics","price": 1299,"quantity": 25,"manufacturer": "HP","timestamp": "2023-07-24 15:45:00"} {"index":{"_index":"zfc-doc-000002","_id":"5"}} {"product_name": "Dell XPS 15", "category": "Electronics","price": 1399,"quantity": 15,"manufacturer": "Dell","timestamp": "2023-07-24 17:20:00"} {"index":{"_index":"zfc-doc-000002","_id":"6"}} {"product_name": "Nike Air Zoom Pegasus 38", "category": "Sports","price": 119,"quantity": 100,"manufacturer": "Nike","timestamp": "2023-07-24 09:30:00"} {"index":{"_index":"zfc-doc-000002","_id":"7"}} {"product_name": "Adidas Ultraboost 21","category": "Sports","price": 129,"quantity": 80,"manufacturer": "Adidas","timestamp": "2023-07-24 10:45:00"} {"index":{"_index":"zfc-doc-000002","_id":"8"}} {"product_name": "Canon EOS Rebel T7i","category": "Electronics","price": 699,"quantity": 10,"manufacturer": "Canon","timestamp": "2023-07-24 14:05:00"} {"index":{"_index":"zfc-doc-000002","_id":"9"}} {"product_name": "LG 55-inch 4K TV", "category": "Electronics","price": 899,"quantity": 30,"manufacturer": "LG","timestamp": "2023-07-24 16:30:00"} {"index":{"_index":"zfc-doc-000002","_id":"10"}} {"product_name": "Lenovo ThinkPad X1 Carbon", "category": "Electronics","price": 1599,"quantity": 18,"manufacturer": "Lenovo","timestamp": "2023-07-24 18:10:00"}
聚合示例
1. 词条聚合（Terms Aggregation）
词条聚合是一种用于对文本字段进行分组的聚合方式，它会将相同值的文档分到同一个桶（Bucket）中，并计算每个桶中文档的数量。
示例查询：
GET zfc-doc-000002/_search { "size": 0, "aggs": { "category_count": { "terms": { "field": "category", "size": 10 } } } }
解释：
"size": 0：表示只返回聚合结果，不返回实际文档数据。"aggs"：定义聚合操作。"category_count"：自定义的聚合名称，用于标识结果。"terms"：指定使用词条聚合。"field": "category"：指定要进行聚合的字段。
2. 嵌套聚合（Nested Aggregation）
嵌套聚合允许在一个桶内进行更深层次的聚合操作。例如，我们可以先按分类分组，然后在每个分类内再按制造商进行分组，并计算每个分类下的平均价格。
示例查询：
GET zfc-doc-000002/_search { "size": 0, "aggs": { "category_group": { "terms": { "field": "category", "size": 10 }, "aggs": { "avg_price": { "avg": { "field": "price" } } } } } }
解释：
"aggs"：定义聚合操作。"category_group"：自定义的聚合名称，用于标识结果。"terms"：指定使用词条聚合。"field": "category"：指定要进行聚合的字段。"avg_price"：自定义的聚合名称，用于标识结果。"avg"：指定使用平均值聚合。"field": "price"：指定要进行聚合的数值字段。
3.直方图聚合示例（Histogram）
假设我们希望根据商品价格（price字段）创建一个价格区间的直方图，将商品按照价格范围进行分组，并统计每个价格区间内的商品数量。
示例查询：
GET zfc-doc-000002/_search { "size": 0, "aggs": { "price_histogram": { "histogram": { "field": "price", "interval": 200 } } } }
解释：
"aggs"：定义聚合操作。"price_histogram"：自定义的聚合名称，用于标识结果。"histogram"：指定使用直方图聚合。"field": "price"：指定要进行聚合的数值字段，即商品价格。"interval": 200：指定直方图的间隔大小，这里设置为200表示将价格范围划分为200的区间，例如：0-200、200-400、400-600等。
4.范围聚合示例(Range)
范围聚合允许我们根据指定的范围条件将文档分组，例如：按价格范围进行分组并统计每个价格范围内的商品数量。
示例查询：
GET zfc-doc-000002/_search { "size": 0, "aggs": { "price_ranges": { "range": { "field": "price", "ranges": [ { "from": 0, "to": 200 }, { "from": 200, "to": 500 }, { "from": 500, "to": 1000 }, { "from": 1000 } ] } } } }
解释：
"aggs"：定义聚合操作。"price_ranges"：自定义的聚合名称，用于标识结果。"range"：指定使用范围聚合。"field": "price"：指定要进行聚合的数值字段，即商品价格。"ranges"：指定价格范围的条件数组。
{ "from": 0, "to": 200 }：表示价格从0到200之间的商品。{ "from": 200, "to": 500 }：表示价格从200到500之间的商品。{ "from": 500, "to": 1000 }：表示价格从500到1000之间的商品。{ "from": 1000 }：表示价格大于等于1000的商品。
5. 统计聚合（Stats Aggregation）
统计聚合可以对数值字段进行计算，包括最小值、最大值、平均值、总和和文档数量。
示例查询：
GET zfc-doc-000002/_search { "size": 0, "aggs": { "price_stats": { "stats": { "field": "price" } } } }
解释：
"aggs"：定义聚合操作。"price_stats"：自定义的聚合名称，用于标识结果。"stats"：指定使用统计聚合。"field": "price"：指定要进行聚合的数值字段。
我们上面在统计聚合中可以获取很多值，那么我们也可以细化单独获取某一个的聚合结果。
6. 平均值聚合（Avg Aggregation）
GET zfc-doc-000002/_search { "size": 0, "aggs": { "avg_price": { "avg": { "field": "price" } } } }
7. 总和聚合（Sum Aggregation）
GET zfc-doc-000002/_search { "size": 0, "aggs": { "total_price": { "sum": { "field": "price" } } } }
8. 最小值聚合（Min Aggregation）
GET zfc-doc-000002/_search { "size": 0, "aggs": { "min_price": { "min": { "field": "price" } } } }
9. 最大值聚合（Max Aggregation）
GET zfc-doc-000002/_search { "size": 0, "aggs": { "max_price": { "max": { "field": "price" } } } }
10. 扩展统计聚合（Extended Stats Aggregation）
GET zfc-doc-000002/_search { "size": 0, "aggs": { "price_stats_extended": { "extended_stats": { "field": "price" } } } }
11. 百分位数聚合（Percentiles Aggregation）
GET zfc-doc-000002/_search { "size": 0, "aggs": { "price_percentiles": { "percentiles": { "field": "price", "percents": [25, 50, 75, 90] } } } }
12. 日期直方图聚合（Date Histogram Aggregation）
假设有一个名为timestamp的日期字段，我们可以进行日期直方图聚合，按照日期进行分组并统计每个时间段内的文档数量。
GET zfc-doc-000002/_search { "size": 0, "aggs": { "date_histogram_agg": { "date_histogram": { "field": "timestamp", "fixed_interval": "1h" } } } }
![Elasticsearch Relevance Engine---为AI变革提供高级搜索能力[ES向量搜索、常用配置参数、聚合功能等详解]](https://ucc.alicdn.com/fnj5anauszhew_20231019_05f219d6edca4595abc931668fc0e37f.png?x-oss-process=image/resize,h_160,m_lfit)