How to filter nested aggregations in ElasticSearch?

Question

For example, let's assume we have a product index with the following mapping:

 {
  "product": {
    "mappings": {
      "producttype": {
        "properties": {
          "id": {
            "type": "keyword"
          },
          "productAttributes": {
            "type": "nested",
            "properties": {
              "name": {
                "type": "keyword"
              },
              "value": {
                "type": "keyword"
              }
            }
          },
          "title": {
            "type": "text",
            "fields": {
              "keyword": {
                "type": "text",
                "analyzer": "keyword"
              }
            },
            "analyzer": "standard"
          }
        }
      }
    }
  }
}

I am trying to find how many products which have specific product attributes using the following query(I am using a fuzzy query to allow some edit distance):

  {
  "size": 0,
  "query": {
    "nested": {
      "query": {
        "fuzzy": {
          "productAttributes.name": {
            "value": "SSD"
          }
        }
      },
      "path": "productAttributes"
    }
  },
  "aggs": {
    "product_attribute_nested_agg": {
      "nested": {
        "path": "productAttributes"
      },
      "aggs": {
        "terms_nested_agg": {
          "terms": {
            "field": "productAttributes.name"
          }
        }
      }
    }
  }
}

But it returns all product attributes for each matched document and here is the response I get.

  "aggregations" : {
    "product_attribute_nested_agg" : {
      "doc_count" : 6,
      "terms_nested_agg" : {
        "doc_count_error_upper_bound" : 0,
        "sum_other_doc_count" : 0,
        "buckets" : [
          {
            "key" : "SSD",
            "doc_count" : 3
          },
          {
            "key" : "USB 2.0",
            "doc_count" : 3
          }
        ]
      }
    }
  }

Could you please guide me to how to filter buckets to only return matched attributes?

Edit: Here are some document samples:

  "hits" : {
    "total" : 12,
    "max_score" : 1.0,
    "hits" : [
      {
        "_index" : "product",
        "_type" : "producttype",
        "_id" : "677d1164-c401-4d36-8a08-6aa14f7f32bb",
        "_score" : 1.0,
        "_source" : {
          "title" : "Dell laptop",
          "productAttributes" : [
            {
              "name" : "USB 2.0",
              "value" : "4"
            },
            {
              "name" : "SSD",
              "value" : "250 GB"
            }
          ]
        }
      },
      {
        "_index" : "product",
        "_type" : "producttype",
        "_id" : "2954935a-7f60-437a-8a54-00da2d71da46",
        "_score" : 1.0,
        "_source" : {
          "productAttributes" : [
            {
              "name" : "USB 2.0",
              "value" : "3"
            },
            {
              "name" : "SSD",
              "value" : "500 GB"
            }
          ],
          "title" : "HP laptop"
        }
      },
    ]
  }

Can you provide a sample doc and can you specify what needs tp be filtered from buckets? — Gibbs
– Gibbs, Commented Jul 10, 2020 at 9:12
For example, if I search "SSD" I want to return how many products(documents) have this attribute's name — Simple Code
– Simple Code, Commented Jul 10, 2020 at 9:20

Simple Code · Accepted Answer · 2020-07-10 11:41:43Z

1

To filter only specific, you can use filter queries.

Query:

{
  "size": 0,
  "aggs": {
    "product_attribute_nested_agg": {
      "nested": {
        "path": "productAttributes"
      },
      "aggs": {
        "inner": {
          "filter": {
            "terms": {
              "productAttributes.name": [
                "SSD"
              ]
            }
          },
          "aggs": {
            "terms_nested_agg": {
              "terms": {
                "field": "productAttributes.name"
              }
            }
          }
        }
      }
    }
  }
}

This is what it does the trick:

"filter": {
  "terms": {
    "productAttributes.name": [
      "SSD"
    ]
  }
}

You need to do filter part of the aggregation.

Output:

"aggregations": {
  "product_attribute_nested_agg": {
    "doc_count": 4,
    "inner": {
      "doc_count": 2,
      "terms_nested_agg": {
        "doc_count_error_upper_bound": 0,
        "sum_other_doc_count": 0,
        "buckets": [
          {
            "key": "SSD",
            "doc_count": 2
          }
        ]
      }
    }
  }
}

Filtering using Fuzziness :

GET /product/_search
{
  "size": 0,
  "aggs": {
    "product_attribute_nested_agg": {
      "nested": {
        "path": "productAttributes"
      },
      "aggs": {
        "inner": {
          "filter": {
           "fuzzy": {
                "productAttributes.name": {
                  "value": "SSt",//here will match SSD
                  "fuzziness": 3//you can remove it to be as Auto
                }
              }
          },
          "aggs": {
            "terms_nested_agg": {
              "terms": {
                "field": "productAttributes.name"
              }
            }
          }
        }
      }
    }
  }
}

edited Jul 10, 2020 at 11:41

Simple Code

2,6543 gold badges35 silver badges65 bronze badges

answered Jul 10, 2020 at 9:22

Gibbs

23.1k14 gold badges84 silver badges147 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Simple Code Over a year ago

Thanks. it solved my problem I edited your answer to support my use case(fuzziness). Could you please upvote the question if you found it helpful?

H_H Over a year ago

@Gibbs Can you post here the bodybuilder.js syntax for the first query which is in your answer?

Collectives™ on Stack Overflow

How to filter nested aggregations in ElasticSearch?

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related