Find Elasticsearch value of empty object

Question

I would like to run a query to see how many documents have an empty object stored as their value. For example it would return a document like:

"hits": [
  {
    "_source": {
      "otherfield": "abc",
      "somefield": {}
    }
  }
]

But not either no field / the field with an undefined value, or the field with an object containing attributes:

"hits": [
  {
    "_source": {
      "otherfield": "abc",
      // <-- note no "somefield"
    }
  },
  {
    "_source": {
      "otherfield": "abc",
      "somefield": { "field1": "value1" }
    }
  }
]

But the query I have will also return documents where the field is an object with attributes such as "somefield": { "field1": "value1" }

GET /documents/_search
{
  "query": {
    "bool": {
      "must_not": [
        {
          "exists": {
            "field": "somefield.field1"
          }
        },
      ]
      "should": [
        {
          "exists": {
            "field": "somefield"
          }
        }
      ],
      "minimum_should_match": 2
    }
  }
}

Using Elasticsearch 5.4

Answer 1

The following query should be enough to find all documents with empty somefield field:

{
  "query": {
    "bool": {
      "must_not": {
        "exists": {
          "field": "somefield"
        }
      }
    }
  }
}

While your query is a bit confusing for me. First you are trying to find any documents where somefield.field1 does not exist. Then you combine content of must_not clause with mutually exclusive content of should clause that filters documents with non-empty somefield. Actually should clause is translated into

"should": [{
    "exists": {"field": "somefield.field1"}
}]

So your query should not match neither documents with somefield: {} nor documents with somefield: {field1: value1}.