This is a major release with many new features and several breaking changes.

Major Changes

• All queries now use a new implementation, using a job and polling for results.
• The copy, load, extract methods now all have high-level and low-level versions, similar to query and query_job.
• Added asynchronous row insertion, allowing data to be collected and inserted in batches.
• Support external data sources for both queries and table views.
• Added create-on-insert support for tables.
• Allow for customizing job IDs to aid in organizing jobs.

Change Details

• Update high-level queries as follows:
  • Update QueryJob#wait_until_done! to use getQueryResults.
  • Update Project#query and Dataset#query with breaking changes:
    • Remove timeout and dryrun parameters.
    • Change return type from QueryData to Data.
  • Add QueryJob#data
  • Alias QueryJob#query_results to QueryJob#data with breaking changes:
    • Remove the timeout parameter.
    • Change the return type from QueryData to Data.
  • Update View#data with breaking changes:
    • Remove the timeout and dryrun parameters.
    • Change the return type from QueryData to Data.
  • Remove QueryData.
  • Update Project#query and Dataset#query with improved errors, replacing the previous simple error with one that contains all available information for why the job failed.
• Rename Dataset#load to Dataset#load_job; add high-level, synchronous version as Dataset#load.
• Rename Table#copy to Table#copy_job; add high-level, synchronous version as Table#copy.
• Rename Table#extract to Table#extract_job; add high-level, synchronous version as Table#extract.
• Rename Table#load to Table#load_job; add high-level, synchronous version as Table#load.
• Add support for querying external data sources with External.
• Add Table::AsyncInserter, Dataset#insert_async and Table#insert_async to collect and insert rows in batches.
• Add Dataset#insert to support creating a table while inserting rows if the table does not exist.
• Update retry logic to conform to the BigQuery SLA.
  • Use a minimum back-off interval of 1 second; for each consecutive error, increase the back-off interval exponentially up to 32 seconds.
  • Retry if all error reasons are retriable, not if any of the error reasons are retriable.
• Add support for labels to Dataset, Table, View and Job.
  • Add filter option to Project#datasets and Project#jobs.
• Add support for user-defined functions to Project#query_job, Dataset#query_job, QueryJob and View.
• In Dataset, Table, and View updates, add the use of ETags for optimistic concurrency control.
• Update Dataset#load and Table#load:
  • Add null_marker option and LoadJob#null_marker.
  • Add autodetect option and LoadJob#autodetect?.
• Fix the default value for LoadJob#quoted_newlines?.
• Add job_id and prefix options for controlling client-side job ID generation to Project#query_job, Dataset#load, Dataset#query_job, Table#copy, Table#extract, and Table#load.
• Add Job#user_email.
• Set the maximum delay of Job#wait_until_done! polling to 60 seconds.
• Automatically retry Job#cancel.
• Allow users to specify if a View query is using Standard vs. Legacy SQL.
• Add project option to Project#query_job.
• Add QueryJob#query_plan, QueryJob::Stage and QueryJob::Step to expose query plan information.
• Add Table#buffer_bytes, Table#buffer_rows and Table#buffer_oldest_at to expose streaming buffer information.
• Update Dataset#insert and Table#insert to raise an error if rows is empty.
• Update Error with a mapping from code 412 to FailedPreconditionError.
• Update Data#schema to freeze the returned Schema object (as in View and LoadJob.)