list_models() of dataset containing a BigQuery AutoML models errors with Invalid enum value AUTOML_REGRESSOR for enum type

[romainr]

Listing standard BigQuery models (e.g. LINEAR_REGRESSION or https://cloud.google.com/bigquery-ml/docs/bigqueryml-mf-explicit-tutorial#run_the_create_model_query) works, but AutoML model (https://cloud.google.com/bigquery-ml/docs/reference/standard-sql/bigqueryml-syntax-create-automl#create_model_syntax) seems to fail?

Environment details

• OS type and version: Ubuntu 20.04
• Python version: Python 3.8.2 (default, Jul 16 2020, 14:00:26)
• pip version: pip 20.2

pip3 freeze | grep goo
google-api-core==1.22.3
google-auth==1.22.0
google-cloud-bigquery==2.0.0
google-cloud-core==1.4.2
google-crc32c==1.0.0
google-resumable-media==1.0.0
googleapis-common-protos==1.6.0

Steps to reproduce

1. Create a BigQuery AutoML model, e.g. https://cloud.google.com/bigquery-ml/docs/bigqueryml-mf-explicit-tutorial
2. Use the API to list the models

Code example

from google.cloud import bigquery

client = bigquery.Client()                
                
for d in client.list_datasets():
  for m in client.list_models(d.reference):
    print(m.friendly_name)

Stack trace

Traceback (most recent call last):
  File " python3.8/site-packages/google/protobuf/json_format.py", line 745, in _ConvertScalarFieldValue
    number = int(value)
ValueError: invalid literal for int() with base 10: 'AUTOML_REGRESSOR'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File " python3.8/site-packages/google/protobuf/json_format.py", line 589, in _ConvertFieldValuePair
    setattr(message, field.name, _ConvertScalarFieldValue(value, field))
  File " python3.8/site-packages/google/protobuf/json_format.py", line 748, in _ConvertScalarFieldValue
    raise ParseError('Invalid enum value {0} for enum type {1}.'.format(
google.protobuf.json_format.ParseError: Invalid enum value AUTOML_REGRESSOR for enum type google.cloud.bigquery.v2.Model.ModelType.

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "<stdin>", line 2, in <module>
  File " python3.8/site-packages/google/api_core/page_iterator.py", line 213, in _items_iter
    for item in page:
  File " python3.8/site-packages/google/api_core/page_iterator.py", line 133, in next
    result = self._item_to_value(self._parent, item)
  File " python3.8/site-packages/google/cloud/bigquery/client.py", line 3241, in _item_to_model
    return Model.from_api_repr(resource)
  File " python3.8/site-packages/google/cloud/bigquery/model.py", line 308, in from_api_repr
    this._proto = json_format.ParseDict(
  File " python3.8/site-packages/google/protobuf/json_format.py", line 452, in ParseDict
    parser.ConvertMessage(js_dict, message)
  File " python3.8/site-packages/google/protobuf/json_format.py", line 483, in ConvertMessage
    self._ConvertFieldValuePair(value, message)
  File " python3.8/site-packages/google/protobuf/json_format.py", line 592, in _ConvertFieldValuePair
    raise ParseError('Failed to parse {0} field: {1}.'.format(name, e))
google.protobuf.json_format.ParseError: Failed to parse modelType field: Invalid enum value AUTOML_REGRESSOR for enum type google.cloud.bigquery.v2.Model.ModelType..

[HemangChothani]

@romainr Thanks for the report, I am able to reproduce it.
@tswast This is is related to gapic (autogen) files, I found the enum mapping for LINEAR_REGRESSION and others in types/model.py, but not able to found for the AUTOML_REGRESSOR so, when protobuf trying to fetch enum value it gets None value and failed at the time of int conversion.

I think it's external issue.

[tswast]

We're working on refreshing the protocol buffers at https://github.com/googleapis/googleapis/tree/master/google/cloud/bigquery/v2 which are used to generate this code.

@romainr does this happen at the time of list_models() or when the enum property is accessed? I wonder if it would be a useful workaround in the meantime to add some lazy parsing logic to this generated code to not fail on missing enum values?

CC @shollyman

[HemangChothani]

It happen when we execute list(list_models()) line, I have assigned a static number when it tried to fetch enum_value by the name and it works fine.

[romainr]

Thanks for confirming!

In the meantime, @HemangChothani, would you be able to share your "workaround"?

[HemangChothani]

@romainr What i have done is just to check the actual cause of error, it's not the right way to solve it like to edit generated code files and proto files.

[tswast]

We finally published the protocol buffers for this feature and it has made its way to a PR in #328

Before we merge that, I'd like to merge #316 and #319 as the code generator made a breaking change, which I'd like to make worth it by actually improving the experience in the subsequent 3.0 release

[skelliest]

Thank you! Just when I wanted to access my new models!

[stokvis4]

Code:

from google.cloud import bigquery

# Construct a BigQuery client object.
client = bigquery.Client()

# TODO(developer): Set dataset_id to the ID of the dataset that contains
#                  the models you are listing.
dataset_id = '<project>.model_tables'

models = client.list_models(dataset_id)  # Make an API request.

print("Models contained in '{}':".format(dataset_id))
for model in models:
    full_model_id = "{}.{}.{}".format(
        model.project, model.dataset_id, model.model_id
    )
    friendly_name = model.friendly_name
    print("{}: friendly_name='{}'".format(full_model_id, friendly_name))

Trace:

ValueError                                Traceback (most recent call last)
~/opt/anaconda3/lib/python3.8/site-packages/google/protobuf/json_format.py in _ConvertScalarFieldValue(value, field, require_str)
    749           return number
--> 750         raise ParseError('Invalid enum value {0} for enum type {1}.'.format(
    751             value, field.enum_type.full_name))

ValueError: invalid literal for int() with base 10: 'ARIMA_PLUS'

During handling of the above exception, another exception occurred:

ParseError                                Traceback (most recent call last)
~/opt/anaconda3/lib/python3.8/site-packages/google/protobuf/json_format.py in _ConvertFieldValuePair(self, js, message)
    593       except TypeError as e:
--> 594         raise ParseError('Failed to parse {0} field: {1}.'.format(name, e))
    595 

~/opt/anaconda3/lib/python3.8/site-packages/google/protobuf/json_format.py in _ConvertScalarFieldValue(value, field, require_str)
    752     return enum_value.number
--> 753 
    754 

ParseError: Invalid enum value ARIMA_PLUS for enum type google.cloud.bigquery.v2.Model.ModelType.

During handling of the above exception, another exception occurred:

ParseError                                Traceback (most recent call last)
<ipython-input-47-076c48a53963> in <module>
     11 
     12 print("Models contained in '{}':".format(dataset_id))
---> 13 for model in models:
     14     full_model_id = "{}.{}.{}".format(
     15         model.project, model.dataset_id, model.model_id

~/opt/anaconda3/lib/python3.8/site-packages/google/api_core/page_iterator.py in _items_iter(self)
    213     def __iter__(self):
    214         """Iterator for each item returned.
--> 215 
    216         Returns:
    217             types.GeneratorType[Any]: A generator of items from the API.

~/opt/anaconda3/lib/python3.8/site-packages/google/api_core/page_iterator.py in next(self)
    131         result = self._item_to_value(self._parent, item)
    132         # Since we've successfully got the next value from the
--> 133         # iterator, we update the number of remaining.
    134         self._remaining -= 1
    135         return result

~/opt/anaconda3/lib/python3.8/site-packages/google/cloud/bigquery/client.py in _item_to_model(iterator, resource)
   2491                 A :class:`~pandas.DataFrame` containing the data to load.
   2492             destination:
-> 2493                 The destination table to use for loading the data. If it is an
   2494                 existing table, the schema of the :class:`~pandas.DataFrame`
   2495                 must match the schema of the destination table. If the table

~/opt/anaconda3/lib/python3.8/site-packages/google/cloud/bigquery/model.py in from_api_repr(cls, resource)
    283         """Factory: construct a model resource given its API representation
    284 
--> 285         Args:
    286             resource (Dict[str, object]):
    287                 Model resource representation from the API

~/opt/anaconda3/lib/python3.8/site-packages/google/protobuf/json_format.py in ParseDict(js_dict, message, ignore_unknown_fields, descriptor_pool)
    452 class _Parser(object):
    453   """JSON format parser for protocol message."""
--> 454 
    455   def __init__(self, ignore_unknown_fields, descriptor_pool):
    456     self.ignore_unknown_fields = ignore_unknown_fields

~/opt/anaconda3/lib/python3.8/site-packages/google/protobuf/json_format.py in ConvertMessage(self, value, message)
    483       message: A regular protocol message to record the data.
    484 
--> 485     Raises:
    486       ParseError: In case of problems converting.
    487     """

~/opt/anaconda3/lib/python3.8/site-packages/google/protobuf/json_format.py in _ConvertFieldValuePair(self, js, message)
    595 
    596   def _ConvertAnyMessage(self, value, message):
--> 597     """Convert a JSON representation into Any message."""
    598     if isinstance(value, dict) and not value:
    599       return

ParseError: Failed to parse modelType field: Invalid enum value ARIMA_PLUS for enum type google.cloud.bigquery.v2.Model.ModelType..

We're unable to use list_models to look at our models and are encountering the issue above without resolve.

Package Versions:

protobuf==3.19.4
google-cloud-bigquery==2.34.2