OpenSearch嵌套聚合：如何获取指定嵌套属性的分桶
2026-6-25
版本
OpenSearch 2.13
问题描述
我正在尝试使用OpenSearch的搜索API和分桶聚合为索引生成面（Facet），需求是为满足特定条件的嵌套字段生成分桶聚合。
我的索引包含一个nested类型的properties字段（每个文档可包含多个独立查询的属性），还有type、createdBy等object类型字段（每个文档各有一个），映射结构如下：
"idx_container_view_list": { "mappings": { "properties": { "ancestorPath": { "type": "keyword" }, "color": { "type": "keyword" }, "coverFile": { "type": "keyword" }, "createdAt": { "type": "date" }, "createdBy": { "properties": { "email": { "type": "keyword" }, "firstname": { "type": "keyword" }, "lastname": { "type": "keyword" }, "username": { "type": "keyword" }, "uuid": { "type": "keyword" } } }, "deletedAt": { "type": "date" }, "deletedBy": ..., "label": { "type": "keyword" }, "parent": { "type": "keyword" }, "properties": { "type": "nested", "properties": { "propertyName": { "type": "keyword" }, "propertyUuid": { "type": "keyword" }, "propertyValue": { "type": "keyword" }, "schemaPropertyUuid": { "type": "keyword" }, "uuid": { "type": "keyword" }, "valueAsLocation": { "properties": { "latitude": { "type": "long" }, "longitude": { "type": "long" }, "placeName": { "type": "text", "fields": { "keyword": { "type": "keyword", "ignore_above": 256 } } } } }, "valueAsNumber": { "type": "float" } } }, "publishedAt": { "type": "date" }, "publishedBy": ... }, "shared": { "type": "boolean" }, "type": { "properties": { "name": { "type": "keyword" }, "slug": { "type": "keyword" }, "uuid": { "type": "keyword" } } }, "updatedAt": { "type": "date" }, "updatedBy": ..., "uuid": { "type": "keyword" } } } }
我已实现所有属性的通用聚合查询（获取前20个属性的前5个值的计数）：
"size" : 0, "aggs": { "different_container_property_types": { "nested": { "path":"properties" }, "aggs": { "property_uuid_bucket": { "terms": { "field": "properties.propertyUuid", "size": 20 }, "aggs": { "property_value_bucket": { "terms": { "field": "properties.propertyValue", "size": 5 } } } } } } }
但我需要仅针对特定propertyUuid生成分桶聚合。尝试添加嵌套查询过滤文档后，返回的是符合条件文档中所有属性的分桶，而非目标属性的单独分桶：
"size" : 0, "query" : { "nested": { "path": "properties", "query":{ "term": { "properties.propertyUuid": "62329d5b-dbc9-4022-93e7-4690c9069a7e" } } } }, "aggs": { "different_container_property_types": { "nested": { "path":"properties" }, "aggs": { "property_uuid_bucket": { "terms": { "field": "properties.propertyUuid", "size": 20 }, "aggs": { "property_value_bucket": { "terms": { "field": "properties.propertyValue", "size": 5 } } } } } } }
返回结果片段：
"took": 46, "timed_out": false, "_shards": { "total": 1, "successful": 1, "skipped": 0, "failed": 0 }, "hits": { "total": { "value": 352, "relation": "eq" }, "max_score": null, "hits": [] }, "aggregations": { "different_container_property_types": { "doc_count": 7624, "property_uuid_bucket": { "doc_count_error_upper_bound": 0, "sum_other_doc_count": 583, "buckets": [ { "key": "02d39964-e894-4d8e-8fdd-32bf2a9c6eab", "doc_count": 353, "property_value_bucket": { "doc_count_error_upper_bound": 0, "sum_other_doc_count": 84, "buckets": [...] } }, { "key": "03df8c5e-61c1-4485-8978-dd2bb96c6790", ... }, { "key": "62329d5b-dbc9-4022-93e7-4690c9069a7e", "doc_count": 352, "property_value_bucket": { "doc_count_error_upper_bound": 0, "sum_other_doc_count": 77, "buckets": [ { "key": "b0cf8292-6005-4095-946d-51fc908243b8", "doc_count": 138 }, { "key": "3f5230ab-8614-436b-abc3-d5cad7f4b5a7", "doc_count": 75 }, { "key": "25135d79-a4c1-44a9-beed-b926b94a40fb", "doc_count": 34 }, { "key": "ab2807d8-1946-40e4-a156-cf3195438697", "doc_count": 31 }, { "key": "d03982e8-872b-4f02-a6b8-fa461e16389e", "doc_count": 24 } ] } }, ...
我理解这是因为OpenSearch为每个嵌套属性生成隐藏文档，过滤出352个包含目标propertyUuid的文档后，聚合包含这些文档中的全部7624个属性分桶。现询问是否有方法仅获取指定propertyUuid的分桶聚合，以避免不必要的集群负载。
解决方案
要仅针对指定propertyUuid生成分桶聚合，需要在嵌套聚合内部添加过滤条件，直接限制聚合只处理目标嵌套属性，避免遍历无关的嵌套文档，降低集群负载。以下是两种可行实现方式：
方式一：嵌套聚合内使用filter聚合
在嵌套聚合中先过滤出目标propertyUuid，再对其值进行分桶，同时顶层查询保留对含目标属性文档的过滤，进一步缩小处理范围：
"size": 0, "query": { "nested": { "path": "properties", "query": { "term": { "properties.propertyUuid": "62329d5b-dbc9-4022-93e7-4690c9069a7e" } } } }, "aggs": { "target_property": { "nested": { "path": "properties" }, "aggs": { "filter_target_uuid": { "filter": { "term": { "properties.propertyUuid": "62329d5b-dbc9-4022-93e7-4690c9069a7e" } }, "aggs": { "property_value_bucket": { "terms": { "field": "properties.propertyValue", "size": 5 } } } } } } }
方式二：terms聚合使用include参数
如果不需要过滤文档，仅需统计目标propertyUuid的属性值，可直接在terms聚合中用include参数精确匹配目标UUID，省去顶层查询步骤：
"size": 0, "aggs": { "target_property": { "nested": { "path": "properties" }, "aggs": { "property_uuid_bucket": { "terms": { "field": "properties.propertyUuid", "include": "62329d5b-dbc9-4022-93e7-4690c9069a7e", "size": 1 }, "aggs": { "property_value_bucket": { "terms": { "field": "properties.propertyValue", "size": 5 } } } } } } }
方案说明
- 方式一适合需先筛选出含目标属性的文档，再聚合该属性值的场景，双重过滤能最大程度减少集群处理的数据量。
- 方式二更轻量化，无需顶层查询，直接锁定目标UUID进行聚合，性能更优，适合仅需目标属性值统计的场景。
内容的提问来源于stack exchange，提问作者fyts
相关产品推荐
相关产品推荐
