如何在Elasticsearch中按文档数组中的多个对象字段进行查询？

Question

Hi I am newbie of Elasticsearch. And I want to query documents by the value of its field (type:array). Example like this:

docx=people:[
{id:1,role:admin},
{id:2,role:other}
]
docy=people:[
{id:1,role:other},
{id:2,role:admin}
]

My query is people.id:1 AND people.role:admin . My expected result is only docx, but ES also returns docy. I know it's wrong but how can I do it in ES. So how a query string can be to filter only document like docx. Thanks!

Answer 1

You need to use the nested data-type in your case, as object data-type is flattened and treated separately which is causing both the documents to match the result, will add a working example.

But please be aware that nested data-type becomes costly when you have large data-set and a high number of concurrent queries as explained in go-jek's medium blog.

Please follow this blog which highlights your issue in detail.

Working example with proper mapping

Index mapping

{
    "mappings": {
        "properties": {
            "name" : {
             "type" :   "text"
            },
            "people": {
                "type": "nested"
            }
        }
    }
}

index sample doc

{
    "name": "bar",
    "people": 
    [
        {
            "id": 1,
            "role": "other"
        },
        {
            "id": 2,
            "role": "admin"
        }
    ]
}

And second sample doc

{
    "name": "foo",
    "people": 
    [
        {
            "id": 1,
            "role": "admin"
        },
        {
            "id": 2,
            "role": "other"
        }
    ]
}

Search query

{
    "query": {
        "nested": {
            "path": "people",
            "query": {
                "bool": {
                    "must": [
                        {
                            "match": {
                                "people.id": 1
                            }
                        },
                        {
                            "match": {
                                "people.role": "admin"
                            }
                        }
                    ]
                }
            }
        }
    }
}

And expected search result

"hits": [
            {
                "_index": "matchphrase",
                "_type": "_doc",
                "_id": "1",
                "_score": 1.6931472,
                "_source": {
                    "name": "foo",
                    "people": [
                        {
                            "id": 1,
                            "role": "admin"
                        },
                        {
                            "id": 2,
                            "role": "other"
                        }
                    ]
                }
            }
        ]