dagster-io/dagster 0.7.9

on GitHub

Breaking Changes

• If you are launching runs using DagsterInstance.launch_run, this method now takes a run id instead of an instance of PipelineRun. Additionally, DagsterInstance.create_run and DagsterInstance.create_empty_run have been replaced by DagsterInstance.get_or_create_run and DagsterInstance.create_run_for_pipeline.
• If you have implemented your own RunLauncher, there are two required changes:
  • RunLauncher.launch_run takes a pipeline run that has already been created. You should remove any calls to instance.create_run in this method.
  • Instead of calling startPipelineExecution (defined in the dagster_graphql.client.query.START_PIPELINE_EXECUTION_MUTATION) in the run launcher, you should call startPipelineExecutionForCreatedRun (defined in dagster_graphql.client.query.START_PIPELINE_EXECUTION_FOR_CREATED_RUN_MUTATION`
  • Refer to the RemoteDagitRunLauncher for an example implementation.

New

• Improvements to preset and solid subselection in the playground. An inline preview of the pipeline instead of a modal when doing subselection, and the correct subselection is chosen when selecting a preset.
• Improvements to the log searching. Tokenization and autocompletion for searching messages types and for specific steps.
• You can now view the structure of pipelines from historical runs, even if that pipeline no longer exists in the loaded repository or has changed structure.
• Historical execution plans are now viewable, even if the pipeline has changed structure.
• Added metadata link to raw compute logs for all StepStart events in PipelineRun view and Step view.
• Improved error handling for the scheduler. If a scheduled run has config errors, the errors are persisted to the event log for the run and can be viewed in Dagit.

Bugfix

• No longer manually dispose sqlalchemy engine in dagster-postgres
• Made boto3 dependency in dagster-aws more flexible (#2418)
• Fixed tooltip UI cleanup in partitioned schedule view

Documentation

• Brand new documentation site, available at https://docs.dagster.io
• The tutorial has been restructured to multiple sections, and the examples in intro_tutorial have been rearranged to separate folders to reflect this.