Adding iterable and length to SQLAlchemyConnectionField

[yfilali]

It is currently difficult to resolve a total_count field with the SQLAlchemy graphene library.

The example from swapi has a nice little implementation that does not work here since unlike Django, the SQLAlchemyConnectionField does not set neither length nor iterable on the connection returned.

This PR brings length and iterable to SQLAlchemyConnectionField to match DjangoConnectionField.

[coveralls]

Coverage increased (+0.1%) to 91.093% when pulling d17c284 on yfilali:master into 2de70b4 on graphql-python:master.

[richmondwang]

This was really needed.

[sebastiandev]

How would you use the length field to be returned along with the query results?
If I just add length on the query, It fails saying the Connection class doesn't have a length field.

[wichert]

@yfilali Is there any example of how to use this? You are storing this on the connection field instance now, but that does not make it visible anywhere. Extending PageInfo with a new field and populating that is still needed somewhere I suspect?

[yfilali]

I haven't tested this recently, but the original example should still stand: swapi

The official docs show how to add an extra field to a connection subclass: http://docs.graphene-python.org/en/latest/relay/connection/, in that example, you would just add a resolve_extra method there that returns self.length.

[wichert]

@yfilali I guess the part I'm missing is how to use that Connection subclass in combination with SQLAlchemyObjectType. Using the swapi example I tried this:

class LengthConnection(graphene.Connection):
    class Meta:
        abstract = True

    total_count = graphene.Int()

    def resolve_total_count(self, info):
        return self.length


class Article(SQLAlchemyObjectType):
    class Meta:
        model = model.Article
        interfaces = (relay.Node, )
        connection_class = LengthConnection

but that immediately aborts with a TypeError:

  File "src/curvetips/api/admin/graphql/article.py", line 38, in <module>
    class Article(SQLAlchemyObjectType):
  File "lib/python3.6/site-packages/graphene/utils/subclass_with_meta.py", line 48, in __init_subclass__
    super_class.__init_subclass_with_meta__(**options)
  File "lib/python3.6/site-packages/graphene_sqlalchemy/types.py", line 131, in __init_subclass_with_meta__
    super(SQLAlchemyObjectType, cls).__init_subclass_with_meta__(_meta=_meta, interfaces=interfaces, **options)
  File "lib/python3.6/site-packages/graphene/types/objecttype.py", line 61, in __init_subclass_with_meta__
    super(ObjectType, cls).__init_subclass_with_meta__(_meta=_meta, **options)
TypeError: __init_subclass_with_meta__() got an unexpected keyword argument 'connection_class'

[wichert]

The graphene documentation for a custom connection makes you think this should work:

class Article(SQLAlchemyObjectType):
    class Meta:
        model = model.Article
        interfaces = (relay.Node, )


class LengthConnection(graphene.Connection):
    class Meta:
        node = Article

    total_count = graphene.Int()

    def resolve_total_count(self, info):
        return self.length


class Query(graphene.ObjectType):
    articles = SQLAlchemyConnectionField(ArticleConnection)

but that fails because ArticleConnection is not a subtype of ArticleConnection.

[wichert]

I think I am on to something. DjangoObjectType. init_subclass_with_meta has an extra connection_class parameter, and uses that as connection type (if provided) instead of dynamically creating a new connection class. There is no equivalent for that in SQLAlchemyObjectType.