BigQuery TypeError: to_pandas() got an unexpected keyword argument ‘timestamp_as_object’

Environment details

  • OS type and version: 1.5.29-debian10
  • Python version: 3.7
  • google-cloud-bigquery version: 2.8.0

I’m provisioning a dataproc cluster which gets the data from BigQuery into a pandas dataframe. As my data is growing I was looking to boost the performance and heard about using the BigQuery storage client.

I had the same problem in the past and this was solved by setting the google-cloud-bigquery to version 1.26.1. If I use that version I get the following message.

/opt/conda/default/lib/python3.7/site-packages/google/cloud/bigquery/client.py:407: UserWarning: Cannot create BigQuery Storage client, the dependency google-cloud-bigquery-storage is not installed.
 "Cannot create BigQuery Storage client, the dependency " 

The code snippet executes but at a way slower rate. If I do not specify the pip version, I encounter this error.

Steps to reproduce

  1. Cluster creation on dataproc
gcloud dataproc clusters create testing-cluster  --region=europe-west1  --zone=europe-west1-b  --master-machine-type n1-standard-16  --single-node  --image-version 1.5-debian10  --initialization-actions gs://dataproc-initialization-actions/python/pip-install.sh  --metadata 'PIP_PACKAGES=elasticsearch google-cloud-bigquery google-cloud-bigquery-storage pandas pandas_gbq'
  1. Execute the Following script on the cluster
bqclient = bigquery.Client(project=project)
    job_config = bigquery.QueryJobConfig(
        query_parameters=[
            bigquery.ScalarQueryParameter("query_start", "STRING", str('2021-02-09 00:00:00')),
            bigquery.ScalarQueryParameter("query_end", "STRING", str('2021-02-09 23:59:59.99')),
        ]
    )
    df = bqclient.query(query, job_config=job_config).to_dataframe(create_bqstorage_client=True)
2021-02-11 10:10:14,069 - preprocessing logger initialized
2021-02-11 10:10:14,069 - arguments = [file, arg1, arg2, arg3, arg4, project_id, arg5, arg6]
Traceback (most recent call last):
  File "/tmp/782503bcc80246258560a07d2179891f/immo_preprocessing-pageviews_kyero.py", line 104, in <module>
    df = bqclient.query(base_query, job_config=job_config).to_dataframe(create_bqstorage_client=True)
  File "/opt/conda/default/lib/python3.7/site-packages/google/cloud/bigquery/job/query.py", line 1333, in to_dataframe
    date_as_object=date_as_object,
  File "/opt/conda/default/lib/python3.7/site-packages/google/cloud/bigquery/table.py", line 1793, in to_dataframe
    df = record_batch.to_pandas(date_as_object=date_as_object, **extra_kwargs)
  File "pyarrow/array.pxi", line 414, in pyarrow.lib._PandasConvertible.to_pandas
TypeError: to_pandas() got an unexpected keyword argument 'timestamp_as_object'

Using the pandas-gbq version gives exaclty the same error

    query_config = {
        'query': {
            'parameterMode': 'NAMED',
            'queryParameters': [
                {
                    'name': 'query_start',
                    'parameterType': {'type': 'STRING'},
                    'parameterValue': {'value': str('2021-02-09 00:00:00')}
                },
                {
                    'name': 'query_end',
                    'parameterType': {'type': 'STRING'},
                    'parameterValue': {'value': str('2021-02-09 23:59:59.99')}
                },
            ]
        }
    }
df = pd.read_gbq(base_query, configuration=query_config, progress_bar_type='tqdm',
                             use_bqstorage_api=True)
2021-02-11 09:21:19,532 - preprocessing logger initialized
2021-02-11 09:21:19,532 - arguments = [file, arg1, arg2, arg3, arg4, project_id, arg5, arg6]
started
Downloading: 100%|██████████| 3107858/3107858 [00:14<00:00, 207656.33rows/s]
Traceback (most recent call last):
  File "/tmp/1830d5bcf198440e9e030c8e42a1b870/immo_preprocessing-pageviews.py", line 98, in <module>
    use_bqstorage_api=True)
  File "/opt/conda/default/lib/python3.7/site-packages/pandas/io/gbq.py", line 193, in read_gbq
    **kwargs,
  File "/opt/conda/default/lib/python3.7/site-packages/pandas_gbq/gbq.py", line 977, in read_gbq
    dtypes=dtypes,
  File "/opt/conda/default/lib/python3.7/site-packages/pandas_gbq/gbq.py", line 536, in run_query
    user_dtypes=dtypes,
  File "/opt/conda/default/lib/python3.7/site-packages/pandas_gbq/gbq.py", line 590, in _download_results
    **to_dataframe_kwargs
  File "/opt/conda/default/lib/python3.7/site-packages/google/cloud/bigquery/table.py", line 1793, in to_dataframe
    df = record_batch.to_pandas(date_as_object=date_as_object, **extra_kwargs)
  File "pyarrow/array.pxi", line 414, in pyarrow.lib._PandasConvertible.to_pandas
TypeError: to_pandas() got an unexpected keyword argument 'timestamp_as_object'

https://github.com/googleapis/python-bigquery/issues/519

Answer

Dataproc installs by default pyarrow 0.15.0 while the bigquery-storage-api needs a more recent version. Manually setting pyarrow to 3.0.0 at install solved the issue. That being said, PySpark has a compability setting for Pyarrow >= 0.15.0 https://spark.apache.org/docs/3.0.0-preview/sql-pyspark-pandas-with-arrow.html#apache-arrow-in-spark I’ve taken a look at the release notes of dataproc and this env variable is set as default since May 2020.