[ElasticSearch] Using facets when querying parent/child documents

gistfile1.sh

# Create an index:
#
curl -XDELETE 'http://127.0.0.1:9200/articles'
curl -XPUT 'http://127.0.0.1:9200/articles'

# Insert the action mapping, so that actions are children of articles:
#
curl -XPUT 'http://127.0.0.1:9200/articles/action/_mapping' -d '
{
  "action": {
    "_parent": {
      "type": "article"
    }
  }
}
'

# Insert some articles:
#
curl -XPUT 'http://127.0.0.1:9200/articles/article/1' -d '{"title": "One"}'
curl -XPUT 'http://127.0.0.1:9200/articles/article/2' -d '{"title": "Two"}'
curl -XPUT 'http://127.0.0.1:9200/articles/article/3' -d '{"title": "Three"}'

# Insert some actions that happened to those articles:
#
curl -XPOST 'http://127.0.0.1:9200/articles/action?parent=1' -d '{"date": "2011-11-18", "created": 1}'
curl -XPOST 'http://127.0.0.1:9200/articles/action?parent=2' -d '{"date": "2011-11-18", "created": 1}'
curl -XPOST 'http://127.0.0.1:9200/articles/action?parent=2' -d '{"date": "2011-11-23", "published": 1}'
curl -XPOST 'http://127.0.0.1:9200/articles/action?parent=3' -d '{"date": "2011-11-23", "created": 1, "published": 1}'

# Ensure the index is up-to-date:
#
curl -XPOST 'http://127.0.0.1:9200/articles/_refresh'

# Now search for all actions that took place between 15th and 30th of November, on an article that has 'One'
# in the title. Create facets for those actions:
#
curl -XPOST 'http://127.0.0.1:9200/articles/action/_search?pretty=true' -d '{
  "query": {
    "filtered": {
      "query": {
        "match_all": {}
      },
      "filter": {
        "and": [
          {
            "range": {
              "date": {
                "to": "2011-11-30",
                "from": "2011-11-15"
              }
            }
          },
          {
            "has_parent": {
              "type": "article",
              "query": {
                "text": {
                  "title": "One or Three"
                }
              }
            }
          }
        ]
      }
    }
  },
  "facets": {
    "published_facet": {
      "date_histogram": {
        "key_field": "date",
        "value_script": "1",
        "interval": "day"
      },
      "facet_filter": {
        "term": {
          "published": 1
        }
      }
    },
    "created_facet": {
      "date_histogram": {
        "key_field": "date",
        "value_script": "1",
        "interval": "day"
      },
      "facet_filter": {
        "term": {
          "created": 1
        }
      }
    }
  }
}'


# This time search for all actions where the article was created and published on the same day:
#
curl -XPOST 'http://127.0.0.1:9200/articles/action/_search?pretty=true' -d '{
  "query": {
    "filtered": {
      "query": {
        "match_all": {}
      },
      "filter": {
        "and": [
          {
            "range": {
              "date": {
                "to": "2011-11-30",
                "from": "2011-11-15"
              }
            }
          },
          {
            "script": {
                "script": "doc[\"created\"].value == 1 && doc[\"published\"].value == 1"
            }
          },
          {
            "has_parent": {
              "type": "article",
              "query": {
                "text": {
                  "title": "One or Three"
                }
              }
            }
          }
        ]
      }
    }
  }
}'

Just discovered that people are referencing this Gist, so I should say that it does not give the results you'd expect!

Ok...now it works properly. The key thing is to use has_parent to limit the records over which the facets are calculated. Note also that the value_script construct I was trying to use before (doc['action'] == 'published' ? 1 : 0) doesn't work. The solution to that is to usefacet_filter`.