Different meanings set on values when reading null

Hey, we have a list in our datastore with ids and sometimes nulls in it. Like here:

["null","null","a953c730fc2e4647b3a5b4e88fa4f0ce","59ce754ddf03401eb09abd50c977cf67","null"]    

When we try to fetch them, we get the “Different meanings set on values within a list_value” Exception.

Is this a bug, that the method does not handles null?

_get_meaning Method:

def _get_meaning(value_pb, is_list=False):
    meaning = None
    if is_list:
        # An empty list will have no values, hence no shared meaning
        # set among them.
        if len(value_pb.list_value) == 0:
            return None

        # We check among all the meanings, some of which may be None,
        # the rest which may be enum/int values.
        all_meanings = set(_get_meaning(sub_value_pb)
                           for sub_value_pb in value_pb.list_value)
        meaning = all_meanings.pop()
        # The value we popped off should have been unique. If not
        # then we can't handle a list with values that have more
        # than one meaning.
        if all_meanings:
            raise ValueError('Different meanings set on values '
                             'within a list_value')
    elif value_pb.meaning:  # Simple field (int32)
        meaning = value_pb.meaning

    return meaning

[dhermes]

Thanks. I’ll try to dig into it today (I am running a training so I may not get to it until the evening).

[dhermes]

Also, for context, _get_meaning and associated code were added in #1241.

Also original design discussion (go to comment, issue is long). I’m trying to track down the discussion of the decision to use a single type for an entire list, but can’t find it.

@xamoom-raphael Can you provide a tiny ndb snippet? I’ll store my own data with the remote API and then check the raw output to see what would need to be done to accommodate.