1.4.3

Dropped support for end-of-life Python 3.8.
Dependency updates.
Removed built-in registration and discovery.

1.4.2

Small bugfix in error handling.

1.4.1

Fixed an issue with the progress bar not showing up.
Fixed an issue with pipeline error handling / logging on python 3.8.

Release 1.4.0

What's Changed

• Fixed an AttributeError in FilesInDirectoryProcessingSource by @benknoll-umn in #400
• Added pipeline hosting and tests by @benknoll-umn in #464
• Dependency updates

Full Changelog: v1.3.0...v1.4.0

Release 1.3.0

What's Changed

MTAP now has a minimum supported Python of 3.8 and a minimum supported Java of 11.

• Small fixes and synchronization with documents by @benknoll-umn in #301
• Added a PipelineTerminated exception that is thrown on termination by error handler by @benknoll-umn in #312
• Added an "enabled" key to pipeline components by @benknoll-umn in #313
• 314 move serializer and error handler registries by @benknoll-umn in #315
• Added a default deployment timeout of 30 seconds by @benknoll-umn in #319
• Prefix deployment log messages with the service name by @benknoll-umn in #367
• Added configurable timeout for pipeline remote processor calls by @benknoll-umn in #376
• Various dependency updates

Full Changelog: v1.2.1...v1.3.0

Breaking Changes

• In the Python RemoteProcessor class the processor_name argument is now just "name".

1.2.1

What's Changed

• Added functionality for advanced, customizable error handling in the pipeline.
• Restructuring and cleaning up the type annotations
• Updating out-of-date documentation

Breaking Changes

There are a number of breaking changes between MTAP 1.1.0 and 1.2.1

Python

• Functionality for pipelines was split from mtap.processing to the mtap.pipeline package.

• EventsClient is now instantiated via the mtap.events_client function.

• RemoteProcessor and LocalProcessor added to default mtap package namespace to ease Pipeline construction

• The Pipeline object is no longer a context manager and no longer an events_client property available.
  This changes instantiation of pipelines to something like:

  pipeline = Pipeline.from_yaml_file(pipeline_conf)
  with events_client(address=pipeline.events_address) as events:
      source = FilesInDirectoryProcessingSource(directory=conf.directory, client=events)
      pipeline.run_multithread(source=source)

  See here for a full example.

• Functionality for time results on the pipeline have been moved to a results object returned by Pipeline.run_multithread

• ProcessingSource is now in mtap.pipeline and its provide method has been renamed produce.

• get_serializers has been removed from mtap.serialization, instead use the get method on SerializerRegistry.

• The methods on Serializer have been changed to class methods to better reflect the intended statelessness.

• Processor descriptors like labels and parameter have been moved from mtap.processing.descriptions to mtap.descriptors.

• The ability to add arbitrary key values to the processor annotation has been replaced with an additional_data dictionary attribute.

• The mtap.data package has been removed. Public access to types in this package that are not already in the mtap namespace have been moved to mtap.types.

1.1.0

What's Changed

Reworked dependencies in Python and Java projects to be tested ranges automatically updated by dependabot. This should promote release stability since broken dependencies will no longer be automatically pulled in by open-ended version ranges.

Full Changelog: v1.0.0...v1.1.0

1.0.0

First public stable release of MTAP. Available on PyPi and Maven Central.

Release 0.6.0

Small release containing updates for label accuracy testing and reporting.

0.5.0

• Fixed an issue with events not being cleaned up when multithreaded pipeline processing was cancelled.
• Fixed an issue with events and processor services not being terminated correctly on Ctrl+C on Windows.
• Label Reference Fields.
• Various bug fixes.