Reverse Nested Aggregations With Filters

Elastic Stack Elasticsearch
1

Hi,

I'm currently looking at using reverse nested aggregations with filters and have a question regarding the filter 'scope' within the aggregations. It's probably best explained through examples:

Create an index with the following mapping:

	"_doc": {
		"properties": {
			"firstNested": {
				"type": "nested",
				"properties": {
					"firstNestedName": {
						"type": "keyword"
					},
					"count": {
						"type": "long"
					},
					"secondNested": {
						"type": "nested",
						"properties": {
							"secondNestedName": {
								"type": "keyword"
							},
							"count": {
								"type": "long"
							}
						}
					}
				}
			},
			"count": {
				"type": "long"
			},
			"rootName": {
				"type": "keyword"
			}
		}
	}
}

Index a document:

  "count" : 100,
  "rootName" : 10000,
  "firstNested" : [
    {
      "firstNestedName" : "Nested1",
      "count" : 20,
	  "secondNested" : [
	  	{
		  "secondNestedName" : "Jim", 
		  "count" : 10
		},
	  	{
		  "secondNestedName" : "Bill", 
		  "count" : 10
		}
	  ]
    },
    {
      "firstNestedName" : "Nested2",
      "count" : 80,
	  "secondNested" : [
	  	{
		  "secondNestedName" : "Jim", 
		  "count" : 50
		},
	  	{
		  "secondNestedName" : "Bill", 
		  "count" : 30
		}
	  ]
    }
  ] 
}

Now run an aggregation which should filter on the secondNestedName field for the value "Jim" and run a terms aggregation on the rootName with the count field from the secondNested document being summed:

	"size": 0,
	"query": {
		"match_all": {}
	},
	"aggregations": {
		"firstNested": {
			"nested": {
				"path": "firstNested"
			},
			"aggregations": {

				"secondNested": {
					"nested": {
						"path": "firstNested.secondNested"
					},
					"aggregations": {
						"secondNestedNameFilter": {
							"filter": {
								"match": {
									"firstNested.secondNested.secondNestedName": "Jim"
								}
							},
							"aggregations": {
								"rootName_reverse": {
									"reverse_nested": {},
									"aggregations": {
										"rootName_values": {
											"terms": {
												"field": "rootName"
											},
											"aggregations": {
												"secondNestedSum": {
													"nested": {
														"path": "firstNested.secondNested"
													},
													"aggregations": {
														"secondNestedNameFilterRepeated": {
															"filter": {
																"match": {
																	"firstNested.secondNested.secondNestedName": "Jim"
																}
															},
															"aggregations": {
																"totalCount": {
																	"sum": {
																		"field": "firstNested.secondNested.count"
																	}
																}
															}
														}
													}
												}
											}
										}
									}
								}
							}
						}
					}
				}
			}
		}
	}
}

The response from this is:

    "took": 5,
    "timed_out": false,
    "_shards": {
        "total": 5,
        "successful": 5,
        "skipped": 0,
        "failed": 0
    },
    "hits": {
        "total": 1,
        "max_score": 0,
        "hits": []
    },
    "aggregations": {
        "firstNested": {
            "doc_count": 2,
            "secondNested": {
                "doc_count": 4,
                "secondNestedNameFilter": {
                    "doc_count": 2,
                    "rootName_reverse": {
                        "doc_count": 1,
                        "rootName_values": {
                            "doc_count_error_upper_bound": 0,
                            "sum_other_doc_count": 0,
                            "buckets": [
                                {
                                    "key": "10000",
                                    "doc_count": 1,
                                    "secondNestedSum": {
                                        "doc_count": 4,
                                        "totalCount": {
                                            "value": 100
                                        }
                                    }
                                }
                            ]
                        }
                    }
                }
            }
        }
    }
}

I'd expect the secondNestedSum value to be 60 with 2 matching documents, not 4. I can repeat the filter in the "secondNestedSum" aggregation to get the correct result however this means duplication. I'm assuming that the reverse nested aggregation is causing the sub-aggregations to lose the current filter context? Is there any other way to resolve this than duplicating the filter(s)?

Thanks,

Brent

2

Just to add, this is using Elasticsearch 6.1.1.