Filtering for top hits aggregation not working as expected-CodePudding

On a parent document I have a list of nested items, which I need to query separately (including paging, sorting and so on). I think the Top hits aggregation is what I need. Unfortunately I don't get the filtering in a way I would expect:

DELETE /test_index

PUT /test_index?pretty=true
{
  "mappings": {
    "properties": {
      "items": {
        "dynamic": false,
        "properties": {
          "text": {
            "type": "keyword"
          },
          "visible": {
            "type": "boolean"
          }
        },
        "type": "nested"
      }
    }
  }
}

POST /test_index/_doc?pretty=true&refresh=wait_for
{
  "description": "Parent 1",
  "items": [
    {
      "text": "Foo A",
      "visible": true
    },
    {
      "text": "Foo invisible",   // <-- this one is found, though visible should be true
      "visible": false
    }
  ]
}

POST /test_index/_doc?pretty=true&refresh=wait_for
{
  "description": "Parent 2",
  "items": [
    {
      "text": "Foo B",
      "visible": true
    }
  ]
}

POST /test_index/_doc?pretty=true&refresh=wait_for
{
  "description": "Parent 3",
  "items": [
    {
      "text": "Foo",     // <-- this one is not found, which is correct
      "visible": false
    }
  ]
}

POST /test_index/_search?pretty=true&typed_keys=true
{
  "aggs": {
    "by_items": {
      "aggs": {
        "top_item_hits": {
          "top_hits": {
            "from": 0,
            "size": 20,
            "sort": [
              {
                "items.text": {
                  "order": "asc"
                }
              }
            ]
          }
        }
      },
      "nested": {
        "path": "items"
      }
    }
  },
  "query": {
    "nested": {
      "path": "items",
      "query": {
        "term": {
          "items.visible": {
            "value": true
          }
        }
      }
    }
  },
  "size": 0,
  "_source": false
}

As mentioned in the comment, one item is found which I wouldn't expect. I know it is related to the sibling in the list which is found, but how can I change the request to filter out all invisible items?

CodePudding user response：

Great progress so far!! You're almost there!

You simply need to add the same filter you have in the query part in the nested aggregation as well and top_hits must be a sub-aggregation of that filter aggregation, like this:

{
  "aggs": {
    "by_items": {
      "nested": {
        "path": "items"
      },
      "aggs": {
        "filter": {                 <--- add this filter
          "term": {
            "items.visible": {
              "value": true
            }
          }
        },
        "aggs": {
          "top_item_hits": {        <--- top_hits should be nested below
            "top_hits": {
              "from": 0,
              "size": 20,
              "sort": [
                {
                  "items.text": {
                    "order": "asc"
                  }
                }
              ]
            }
          }
        }
      }
    }
  },
  "query": {
    "nested": {
      "path": "items",
      "query": {
        "term": {
          "items.visible": {
            "value": true
          }
        }
      }
    }
  },
  "size": 0,
  "_source": false
}