`collection.iterator` created with `return_properties` fails when using properties with `object[]` `dataType`

[anakin87]

Description

This issue was found in the Haystack integration. Corresponding Haystack issue: deepset-ai/haystack-core-integrations#1172

weaviate-client==4.9.0

When a property with object[] dataType is present,
we get an error when iterating on a collection.iterator created with return_properties.

If we create a similar iterator but without passing return_properties, all works correctly.

Reproducible code example

import weaviate

# I'm running a Docker instance of Weaviate (1.24.5; same behavior with 1.27.2)

client = weaviate.WeaviateClient(
   connection_params=(weaviate.connect.base.ConnectionParams.from_url(url="http://localhost:8080", grpc_port=50051))
)

client.connect()

DOCUMENT_COLLECTION_PROPERTIES = [
   {"name": "_original_id", "dataType": ["text"]},
   {"name": "content", "dataType": ["text"]},
   {"name": "dataframe", "dataType": ["text"]},
   {"name": "blob_data", "dataType": ["blob"]},
   {"name": "blob_mime_type", "dataType": ["text"]},
   {"name": "score", "dataType": ["number"]},
   # the following properties can be present or not. Weaviate shows the same behavior:
   # documents are correctly written but not correctly returned
   # {
   #         'name': 'mylistofobjects', 'dataType': ['object[]'],
   #         'nestedProperties': [
   #             {'dataType': ['text'], 'name': 'doc_id'},
   #             {'dataType': ['number[]'], 'name': 'range'}],
   # }
]


collection_settings = {
   "class": "Default",
   "invertedIndexConfig": {"indexNullState": True},
   "properties": DOCUMENT_COLLECTION_PROPERTIES,
}


collection = client.collections.create_from_dict(collection_settings)

properties = {
   'content': 'This is a test document',
   'dataframe': None,
   'score': None,
   'mylistofobjects': [{'doc_id': '1', 'range': [1, 2]}],
   '_original_id': '3972bbfa2c09af05a7118ed4233124582a138dd83e3de1db3ff742f810df4c41',
}

collection.data.insert(
   properties=properties,
   vector=[0.1] * 300,
)

# this works and returns all properties except byte
# (in this case byte properties are not present, but they are not returned even if present)
it = collection.iterator(include_vector=True)
for i in it:
   print(i)

# this fails
it = collection.iterator(include_vector=True, return_properties=["content", "mylistofobjects"])
for i in it:
   print(i)

Error:

  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/.hatch/weaviate-haystack/lib/python3.10/site-packages/weaviate/collections/grpc/query.py", line 798, in __call
    res = await self._connection.grpc_stub.Search(
  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/.hatch/weaviate-haystack/lib/python3.10/site-packages/grpc/aio/_call.py", line 327, in __await__
    raise _create_rpc_error(
grpc.aio._call.AioRpcError: <AioRpcError of RPC that terminated with:
        status = StatusCode.UNKNOWN
        details = "creating primitive value for mylistofobjects: proto: invalid type: []interface {}"
        debug_error_string = "UNKNOWN:Error received from peer  {created_time:"2024-11-11T17:27:11.360212037+01:00", grpc_status:2, grpc_message:"creating primitive value for mylistofobjects: proto: invalid type: []interface {}"}"
>

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/newtryweaviate.py", line 55, in <module>
    for i in it:
  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/.hatch/weaviate-haystack/lib/python3.10/site-packages/weaviate/collections/iterator.py", line 59, in __next__
    res = self.__query.fetch_objects(
  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/.hatch/weaviate-haystack/lib/python3.10/site-packages/weaviate/syncify.py", line 23, in sync_method
    return _EventLoopSingleton.get_instance().run_until_complete(
  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/.hatch/weaviate-haystack/lib/python3.10/site-packages/weaviate/event_loop.py", line 40, in run_until_complete
    return fut.result()
  File "/home/anakin87/.pyenv/versions/3.10.13/lib/python3.10/concurrent/futures/_base.py", line 458, in result
    return self.__get_result()
  File "/home/anakin87/.pyenv/versions/3.10.13/lib/python3.10/concurrent/futures/_base.py", line 403, in __get_result
    raise self._exception
  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/.hatch/weaviate-haystack/lib/python3.10/site-packages/weaviate/collections/queries/fetch_objects/query.py", line 65, in fetch_objects
    res = await self._query.get(
  File "/home/anakin87/apps/haystack-core-integrations/integrations/weaviate/.hatch/weaviate-haystack/lib/python3.10/site-packages/weaviate/collections/grpc/query.py", line 805, in __call
    raise WeaviateQueryError(str(e), "GRPC search")  # pyright: ignore
weaviate.exceptions.WeaviateQueryError: Query call with protocol GRPC search failed with message <AioRpcError of RPC that terminated with:
        status = StatusCode.UNKNOWN
        details = "creating primitive value for mylistofobjects: proto: invalid type: []interface {}"
        debug_error_string = "UNKNOWN:Error received from peer  {created_time:"2024-11-11T17:27:11.360212037+01:00", grpc_status:2, grpc_message:"creating primitive value for mylistofobjects: proto: invalid type: []interface {}"}"
>.```

[tsmith023]

Hi @anakin87, thanks for raising this one! The exact syntax that you're attempting here actually isn't supported currently but I think it should be, it seems an oversight that it was missed

In short, if you specify the object or object[] property that you want returned then you have to also specify the exact nested properties that you want back. In your case, this would look like:

it = collection.iterator(
    include_vector=True,
    return_properties=[
        "content",
        wvc.query.QueryNested(name="mylistofobjects", properties=["doc_id", "range"])
    ]
)

However, the use-case you show with the Python client appears natural so it is certainly a surprise that it fails with such a cryptic error! I will look into how we can support this in the gRPC API on the server. If we can do it easily, we'll be able to release a patch version to all currently supported minors including the fix 😁

[davidsbatista]

Hi @tsmith023! Did you have the chance to work on this issue?

[tsmith023]

Hi @davidsbatista, and others watching, I've been able to revisit the stale Python client issues and have introduced a change to the server here that will allow this syntax moving forward. It will be available in all the latest patches of v1.33 -> v1.36 once merged and released!