sqlalchemy-bigquery-user-cred 1.0.0a2

Supported Python Versions

Python >= 3.6

Unsupported Python Versions

Python <= 3.5.

Mac/Linux

pip install virtualenv
virtualenv <your-env>
source <your-env>/bin/activate
<your-env>/bin/pip install sqlalchemy-bigquery

Windows

pip install virtualenv
virtualenv <your-env>
<your-env>\Scripts\activate
<your-env>\Scripts\pip.exe install sqlalchemy-bigquery

SQLAlchemy

from sqlalchemy import *
from sqlalchemy.engine import create_engine
from sqlalchemy.schema import *
engine = create_engine('bigquery://project')
table = Table('dataset.table', MetaData(bind=engine), autoload=True)
print(select([func.count('*')], from_obj=table).scalar())

Project

project in bigquery://project is used to instantiate BigQuery client with the specific project ID. To infer project from the environment, use bigquery:// – without project

Authentication

Follow the Google Cloud library guide for authentication. Alternatively, you can provide the path to a service account JSON file in create_engine():

engine = create_engine('bigquery://', credentials_path='/path/to/keyfile.json')

Location

To specify location of your datasets pass location to create_engine():

engine = create_engine('bigquery://project', location="asia-northeast1")

Table names

To query tables from non-default projects or datasets, use the following format for the SQLAlchemy schema name: [project.]dataset, e.g.:

# If neither dataset nor project are the default
sample_table_1 = Table('natality', schema='bigquery-public-data.samples')
# If just dataset is not the default
sample_table_2 = Table('natality', schema='bigquery-public-data')

Batch size

By default, arraysize is set to 5000. arraysize is used to set the batch size for fetching results. To change it, pass arraysize to create_engine():

engine = create_engine('bigquery://project', arraysize=1000)

Page size for dataset.list_tables

By default, list_tables_page_size is set to 1000. list_tables_page_size is used to set the max_results for dataset.list_tables operation. To change it, pass list_tables_page_size to create_engine():

engine = create_engine('bigquery://project', list_tables_page_size=100)

Adding a Default Dataset

If you want to have the Client use a default dataset, specify it as the “database” portion of the connection string.

engine = create_engine('bigquery://project/dataset')

When using a default dataset, don’t include the dataset name in the table name, e.g.:

table = Table('table_name')

Note that specifying a default dataset doesn’t restrict execution of queries to that particular dataset when using raw queries, e.g.:

# Set default dataset to dataset_a
engine = create_engine('bigquery://project/dataset_a')

# This will still execute and return rows from dataset_b
engine.execute('SELECT * FROM dataset_b.table').fetchall()

Connection String Parameters

There are many situations where you can’t call create_engine directly, such as when using tools like Flask SQLAlchemy. For situations like these, or for situations where you want the Client to have a default_query_job_config, you can pass many arguments in the query of the connection string.

The credentials_path, credentials_info, location, arraysize and list_tables_page_size parameters are used by this library, and the rest are used to create a QueryJobConfig

Note that if you want to use query strings, it will be more reliable if you use three slashes, so 'bigquery:///?a=b' will work reliably, but 'bigquery://?a=b' might be interpreted as having a “database” of ?a=b, depending on the system being used to parse the connection string.

Here are examples of all the supported arguments. Any not present are either for legacy sql (which isn’t supported by this library), or are too complex and are not implemented.

engine = create_engine(
    'bigquery://some-project/some-dataset' '?'
    'credentials_path=/some/path/to.json' '&'
    'location=some-location' '&'
    'arraysize=1000' '&'
    'list_tables_page_size=100' '&'
    'clustering_fields=a,b,c' '&'
    'create_disposition=CREATE_IF_NEEDED' '&'
    'destination=different-project.different-dataset.table' '&'
    'destination_encryption_configuration=some-configuration' '&'
    'dry_run=true' '&'
    'labels=a:b,c:d' '&'
    'maximum_bytes_billed=1000' '&'
    'priority=INTERACTIVE' '&'
    'schema_update_options=ALLOW_FIELD_ADDITION,ALLOW_FIELD_RELAXATION' '&'
    'use_query_cache=true' '&'
    'write_disposition=WRITE_APPEND'
)

Creating tables

To add metadata to a table:

table = Table('mytable', ..., bigquery_description='my table description', bigquery_friendly_name='my table friendly name')

To add metadata to a column:

Column('mycolumn', doc='my column description')

Threading and Multiprocessing

Because this client uses the grpc library, it’s safe to share instances across threads.

In multiprocessing scenarios, the best practice is to create client instances after the invocation of os.fork by multiprocessing.pool.Pool or multiprocessing.Process.