Supported Python Versions
WARNING: Breaking change
The 2.0.0 release requires Python 3.6+.
Import Path
The library’s top-level namespace is google.cloud.bigquery_storage. Importing
from google.cloud.bigquery_storage_v1 still works, but it is advisable to use
the google.cloud.bigquery_storage path in order to reduce the chance of future
compatibility issues should the library be restuctured internally.
Before:
from google.cloud.bigquery_storage_v1 import BigQueryReadClient
After:
from google.cloud.bigquery_storage import BigQueryReadClient
Enum Types
WARNING: Breaking change
Enum types have been moved. Access them through the types module.
Before:
from google.cloud.bigquery_storage_v1 import enums
data_format = enums.DataFormat.ARROW
data_format = BigQueryReadClient.enums.DataFormat.ARROW
After:
from google.cloud.bigquery_storage import types
data_format = types.DataFormat.ARROW
Additionally, enums cannot be accessed through the client anymore. The following code wil not work:
data_format = BigQueryReadClient.enums.DataFormat.ARROW
Clients for Beta APIs
WARNING: Breaking change
Clients for beta APIs have been removed. The following import will not work:
from google.cloud.bigquery_storage_v1beta1 import BigQueryStorageClient
from google.cloud.bigquery_storage_v1beta2.gapic.big_query_read_client import BigQueryReadClient
The beta APIs are still available on the server side, but you will need to use the 1.x version of the library to access them.
Changed Default Value of the read_rows() Method’s metadata Argument
The client.read_rows() method does not accept None anymore as a valid value
for the optional metadata argument. If not given, an empty tuple is used, but
if you want to explicitly pass an “empty” value, you should use an empty tuple, too.
Before:
client.read_rows("stream_name", metadata=None)
After:
client.read_rows("stream_name", metadata=())
OR
client.read_rows("stream_name")
Method Calls
WARNING: Breaking change
Most of the client methods that send requests to the backend expect request objects. We provide a script that will convert most common use cases.
One exception to this is the
BigQueryReadClient.read_rows()which is a hand-written wrapper around the auto-generatedread_rows()method.
- Install the library
python3 -m pip install google-cloud-bigquery-storage
- The script
fixup_bigquery_storage_v1_keywords.pyis shipped with the library. It requireslibcstto be installed. It expects an input directory (with the code to convert) and an empty destination directory.
$ fixup_bigquery_storage_v1_keywords.py --input-directory .samples/ --output-directory samples/
Before:
from google.cloud import bigquery_storage_v1
client = bigquery_storage_v1.BigQueryReadClient()
requested_session = bigquery_storage_v1.types.ReadSession()
requested_session.table = "projects/PROJECT_ID/datasets/DATASET_ID/tables/TABLE_ID"
requested_session.data_format = bigquery_storage_v1.enums.DataFormat.ARROW
session = client.create_read_session(
"projects/parent_project",
requested_session,
max_stream_count=1,
)
After:
from google.cloud import bigquery_storage
client = bigquery_storage.BigQueryReadClient()
requested_session = bigquery_storage.types.ReadSession(
table="projects/PROJECT_ID/datasets/DATASET_ID/tables/TABLE_ID",
data_format=bigquery_storage.types.DataFormat.ARROW,
)
session = client.create_read_session(
request={
"parent": "projects/parent_project",
"read_session": requested_session,
"max_stream_count" 1,
},
)
More Details
In google-cloud-bigquery-storage<2.0.0, parameters required by the API were positional
parameters and optional parameters were keyword parameters.
Before:
def create_read_session(
self,
parent,
read_session,
max_stream_count=None,
retry=google.api_core.gapic_v1.method.DEFAULT,
timeout=google.api_core.gapic_v1.method.DEFAULT,
metadata=None,
):
In the 2.0.0 release, methods that interact with the backend have a single
positional parameter request. Method docstrings indicate whether a parameter is
required or optional.
Some methods have additional keyword only parameters. The available parameters depend
on the google.api.method_signature annotation
specified by the API producer.
After:
def create_read_session(
self,
request: storage.CreateReadSessionRequest = None,
*,
parent: str = None,
read_session: stream.ReadSession = None,
max_stream_count: int = None,
retry: retries.Retry = gapic_v1.method.DEFAULT,
timeout: float = None,
metadata: Sequence[Tuple[str, str]] = (),
) -> stream.ReadSession:
NOTE: The
requestparameter and flattened keyword parameters for the API are mutually exclusive. Passing both will result in an error.
Both of these calls are valid:
session = client.create_read_session(
request={
"parent": "projects/parent_project",
"read_session": requested_session,
"max_stream_count" 1,
},
)
response = client.create_read_session(
parent="projects/parent_project",
read_session=requested_session,
max_stream_count=1,
)
This call is invalid because it mixes request with a keyword argument
max_stream_count. Executing this code will result in an error:
session = client.create_read_session(
request={
"parent": "projects/parent_project",
"read_session": requested_session,
},
max_stream_count=1,
)
NOTE: The
requestparameter of some methods can also contain a more rich set of options that are otherwise not available as explicit keyword only parameters, thus these must be passed throughrequest.
Removed Utility Methods
WARNING: Breaking change
Several utility methods such as project_path() and table_path() have been removed.
These paths must now be constructed manually:
project_path = f"project/{PROJECT_ID}"
table_path = f"projects/{PROJECT_ID}/datasets/{DATASET_ID}/tables/{TABLE_ID}"
The two that remained are read_session_path() and read_stream_path().
Removed client_config and channel Parameter
The client cannot be constructed with channel or client_config arguments anymore,
these deprecated parameters have been removed.
If you used client_config to customize retry and timeout settings for a particular
method, you now need to do it upon method invocation by passing the custom timeout and
retry arguments, respectively.