09-03-2019 14:33 PDT

Implementation Changes

• Raise when unexpected fields are present in the LoadJobConfig.schema when calling load_table_from_dataframe. (#9096)
• Determine the schema in load_table_from_dataframe based on dtypes. (#9049)
• Raise helpful error when loading table from dataframe with STRUCT columns. (#9053)
• Fix schema recognition of struct field types. (#9001)
• Fix deserializing None in QueryJob for queries with parameters. (#9029)

New Features

• Include indexes in table written by load_table_from_dataframe, only if
  fields corresponding to indexes are present in LoadJobConfig.schema.
  (#9084)
• Add client_options to constructor. (#8999)
• Add --dry_run option to %%bigquery magic. (#9067)
• Add load_table_from_json() method to create a table from a list of dictionaries. (#9076)
• Allow subset of schema to be passed into load_table_from_dataframe. (#9064)
• Add support for unsetting LoadJobConfig.schema. (#9077)
• Add support to Dataset for project IDs containing an org prefix. (#8877)
• Add enum with SQL type names allowed to be used in SchemaField. (#9040)

Documentation

• Fix the reference URL for Client.create_dataset(). (#9149)
• Update code samples to use strings for table names instead of client.dataset(). (#9032)
• Remove compatability badges from READMEs. (#9035)
• Fix Pandas DataFrame load example under Python 2.7. (#9022)

Internal / Testing Changes

• Disable failing snippets test for copying CMEK-protected tables. (#9156)
• Fix BigQuery client unit test assertions (#9112)
• Replace avro with arrow schemas in test_table.py (#9056)