dagster-io/dagster 0.15.3

0.15.3 (2022-06-30)

on GitHub

New

• When loading an upstream asset or op output as an input, you can now set custom loading behavior using the input_manager_key argument to AssetIn and In
• The list of objects returned by a repository can now contain nested lists.
• Added a data retention instance setting in dagster.yaml that enables the automatic removal of sensor/schedule ticks after a certain number of days.
• Added a sensor daemon setting in dagster.yaml that enables sensor evaluations to happen in a thread pool to increase throughput.
• materialize_to_memory and materialize now both have the partition_key argument.
• Output and DynamicOutput objects now work with deep equality checks:

Output(value=5, name="foo") == Output(value=5, name="foo") # evaluates to True

• RunRequest can now be returned from run status sensors
• Added resource_defs argument to AssetsDefinition.from_graph. Allows for specifying resources required by constituent ops directly on the asset.
• When adding a tag to the Run search filter in Dagit by clicking the hover menu on the tag, the tag will now be appended to the filter instead of replacing the entire filter state.

Bugfixes

• [dagster-dbt] An exception is now emitted if you attempt to invoke the library without having dbt-core installed. dbt-core is now also added as a dependency to the library.
• Asset group names can now contain reserved python keywords
• Fixed a run config parsing bug that was introduced in 0.15.1 that caused Dagit to interpret datetime strings as datetime objects and octal strings as integers.
• Runs that have failed to start are now represented in the Instance Timeline view on Dagit.
• Fixed an issue where the partition status was missing for partitioned jobs that had no runs.
• Fixed a bug where op/resource invocation would error when resources were required, no context was used in the body of the function, and no context was provided when invoking.
• [dagster-databricks] Fixed an issue where an exception related to the deprecated prior_attempts_count field when using the databricks_pyspark_step_launcher.
• [dagster-databricks] Polling information logged from the databricks_pyspark_step_launcher is now emitted at the DEBUG level instead of INFO.
• In the yaml editor in Dagit, the typeahead feature now correctly shows suggestions for nullable schema types.
• When editing asset configuration in Dagit, the “Scaffold config” button in the Dagit launchpad sometimes showed the scaffold dialog beneath the launchpad. This has been fixed.
• A recent change added execution timezones to some human-readable cron strings on schedules in Dagit. This was added incorrectly in some cases, and has now been fixed.
• In the Dagit launchpad, a config state containing only empty newlines could lead to an error that could break the editor. This has been fixed.
• Fixed issue that could cause partitioned graph-backed assets to attempt to load upstream inputs from the incorrect path when using the fs_io_manager (or other similar io managers).
• [dagster-dbt] Fixed issue where errors generated from issuing dbt cli commands would only show json-formatted output, rather than a parsed, human-readable output.
• [dagster-dbt] By default, dagster will invoke the dbt cli with a --log-format json flag. In some cases, this may cause dbt to report incorrect or misleading error messages. As a workaround, it is now possible to disable this behavior by setting the json_log_format configuration option on the dbt_cli_resource to False.
• materialize_to_memory erroneously allowed non-in-memory io managers to be used. Now, providing io managers to materialize_to_memory will result in an error, and mem_io_manager will be provided to all io manager keys.

All Changes

0.15.2...0.15.3