How can I use bulkuploader to populate class with a db.SelfReferenceProperty?

Question

I've got a class that is using db.SelfReferenceProperty to create a tree-like structure.

When trying to populate the database using appcfg.py upload_data -- config_file=bulkloader.yaml --kind=Group --filename=group.csv (...) , I'm getting an exception saying BadValueError: name must not be empty. ^{(Full stack below)}

I tried ordering the data to make sure that a Groups that had a foreign key pointing at them were first. That didn't work.

By commenting from the bulkloader.yaml the line making the transformation "import_transform: transform.create_foreign_key('Group')", the data is uploaded, but it stores that property as string, breaking my application logic.

- kind: Group
  connector: csv
  connector_options:
  property_map:
    - property: __key__
      external_name: key
      export_transform: transform.key_id_or_name_as_string

    - property: name
      external_name: name
      # Type: String Stats: 9 properties of this type in this kind.

    - property: section
      external_name: section
      # Type: Key Stats: 6 properties of this type in this kind.
      import_transform: transform.create_foreign_key('Group')
      export_transform: transform.key_id_or_name_as_string

Is there any way to make the bulkloader take into account selfreferences, or should I either make a transformation serverside of the bulkloaded data or implement my own bulk loading algorithm?

----
Traceback (most recent call last):
  File "/home/username/src/google_appengine/google/appengine/tools/adaptive_thread_pool.py", line 150, in WorkOnItems
    status, instruction = item.PerformWork(self.__thread_pool)
  File "/home/username/src/google_appengine/google/appengine/tools/bulkloader.py", line 691, in PerformWork
    transfer_time = self._TransferItem(thread_pool)
  File "/home/username/src/google_appengine/google/appengine/tools/bulkloader.py", line 846, in _TransferItem
    self.content = self.request_manager.EncodeContent(self.rows)
  File "/home/username/src/google_appengine/google/appengine/tools/bulkloader.py", line 1267, in EncodeContent
    entity = loader.create_entity(values, key_name=key, parent=parent)
  File "/home/username/src/google_appengine/google/appengine/ext/bulkload/bulkloader_config.py", line 382, in create_entity
    return self.dict_to_entity(input_dict, self.bulkload_state)
  File "/home/username/src/google_appengine/google/appengine/ext/bulkload/bulkloader_config.py", line 133, in dict_to_entity
    self.__run_import_transforms(input_dict, instance, bulkload_state_copy)
  File "/home/username/src/google_appengine/google/appengine/ext/bulkload/bulkloader_config.py", line 230, in __run_import_transforms
    value = self.__dict_to_prop(transform, input_dict, bulkload_state)
  File "/home/username/src/google_appengine/google/appengine/ext/bulkload/bulkloader_config.py", line 188, in __dict_to_prop
    value = transform.import_transform(value)
  File "/home/username/src/google_appengine/google/appengine/ext/bulkload/bulkloader_parser.py", line 93, in __call__
    return self.method(*args, **kwargs)
  File "/home/username/src/google_appengine/google/appengine/ext/bulkload/transform.py", line 114, in generate_foreign_key_lambda
        return datastore.Key.from_path(kind, value)
  File "/home/username/src/google_appengine/google/appengine/api/datastore_types.py", line 384, in from_path
    ValidateString(id_or_name, 'name')
  File "/home/username/src/google_appengine/google/appengine/api/datastore_types.py", line 109, in ValidateString
    raise exception('%s must not be empty.' % name)
BadValueError: name must not be empty.

Answer 1

Using an answer given to a similar problem as a base, I could successfully solve this by creating a small helpers.py file to do act as a wrapper around transform.create_foreign_key:

from google.appengine.api import datastore

def create_foreign_key(kind, key_is_id=False):
  def generate_foreign_key_lambda(value):
    if value is None:
      return None

    if key_is_id:
      value = int(value)
    try:
      return datastore.Key.from_path(kind, value)
    except:
      return None

  return generate_foreign_key_lambda

Once that file is in place in the same directory as your yaml bolk upload configuration file (bulkloader.yaml), you add this to that file:

python_preamble:
- (...)
- import: helpers

transformers:

- kind: Group
  connector: csv
  connector_options:
  property_map:
    - property: __key__
      external_name: key
      export_transform: transform.key_id_or_name_as_string

    - property: name
      external_name: name

    - property: section
      external_name: section
      import_transform: helpers.create_foreign_key('Group')
                      # ^^^^^^^ we use the wrapper instead
      export_transform: transform.key_id_or_name_as_string

With those changes, the bulk upload is now working correctly.

Before using this, you should definitely modify the catch all except, and replace it with may be except BadValueError.

Answer 2

transform.py (maybe only recently) contains a decorator that solves this issue:

def none_if_empty(fn):
  """A decorator which returns None if its input is empty else fn(x).

  Useful on import.  Can be used in config files
  (e.g. "transform.none_if_empty(int)" or as a decorator.

  Args:
    fn: Single argument transform function.

  Returns:
    Wrapped function.
  """

  def wrapper(value):


    if value == '' or value is None or value == []:
      return None
    return fn(value)

  return wrapper

So using the following will also solve the problem, without the introduction of a custom helpers.py file:

transform.none_if_empty(transform.create_foreign_key('Group'))