parent/child mapping does not seem to work with ElasticSearch 6

[aschriner]

# search.py 

class Org(DocType):
    name = Text(required=True)
    state = Keyword()
    employee_size_range = Keyword()
    revenue_range = Keyword()
    lifetime_revenue_range = Keyword()
    street_address = Keyword()
    city = Keyword()
    postal_code = Keyword()
    website = Text()


class Contact(DocType):
    first_name = Keyword()
    last_name = Keyword()
    job_title = Text()
    email = Keyword()

    class Meta:
        parent = MetaField(type='org')


# create an index and register the doc types
index = Index(settings.ES_INDEX)
index.settings(number_of_shards=1, number_of_replicas=0)
index.doc_type(Org)
index.doc_type(Contact)

When trying to fill the index with index_data I get elasticsearch.helpers.BulkIndexError with "reason": "routing is required for [stack]/[doc]/[some-document-id]".

The created mapping looks like

{
  "stack": {
    "mappings": {
      "doc": {
        "_parent": {
          "type": "org"
        },
        "_routing": {
          "required": true
        },
        "properties": {
          "email": {
            "type": "keyword"
          },
          "first_name": {
            "type": "keyword"
          },
          "job_title": {
            "type": "text"
          },
          "last_name": {
            "type": "keyword"
          }
        }
      }
    }
  }
}

which doesn't look right to me! Looks like the Org mapping has been overridden by the Contact mapping.

I've tried adding doc_type='org' doc_type='contact' to the DocType Meta options, but ES 6 does not allow multiple doc types. I assume there is some way to make this work in ES6 but I'm not seeing it. Am I missing something or is this broken for ES6?

[TheCthulhuKid]

You can no longer have multiple doc types in an index. If you are using the latest elasticsearch-dsl you can make a type of InnerDoc for something you want nested.

I think it is now suggested to only use doc for the document type name.

[netoxico]

@aschriner there is currently a pull request with a example on how to do this with the new Join field
https://github.com/elastic/elasticsearch-dsl-py/pull/865/files#diff-caba059a7c2e8c265658722732b4a88bR75