Converting to string type in py4j causes AttributeError on _detach

Question

I've got a type WriteableFileName that "acts like a string" in python. I can convert it to a string in java using this custom converter:

class WriteableFileNameConverter(object):

    def can_convert(self, object):
        return isinstance(object, WriteableFileName)

    def convert(self, object, gateway_client):
        str_representation = str(object)
        String = JavaClass("java.lang.String", gateway_client)
        ret_val = String(str(str_representation))
        return ret_val

When I use that converter, I can execute something like df.write.format('csv').save(my_filename) and lazy eval my_filename to a file path. After save is done, the converted type fails with this error:

  File "C:\Users\riguy\Anaconda3\envs\work37\lib\site-packages\pyspark\sql\readwriter.py", line 739, in save
    self._jwrite.save(path)
  File "C:\Users\riguy\Anaconda3\envs\work37\lib\site-packages\py4j\java_gateway.py", line 1260, in __call__
    temp_arg._detach()
AttributeError: 'str' object has no attribute '_detach'

Weirdly (for me!) type(ret_val) above is still 'str' but it has a different id than str_representation.

So the question is (1) what's the detach error, and also (2) why is ret_val a 'str' type not a JavaClass?