0.0.9
=====

Bug Fix:
* Syntax error on generated migration

Features:
* Upgrading dependency to delayed_job
* Move to gemcutter
* Can be installed as a plugin or gem
* Improved Documentation
* Callbacks before and after stage/pipeline execution
* Association to an external entity through external_id (Please refer to documentation)

0.0.8
=====

Bug Fix:
* Clearing message when restarting a failed stage
* Only saves the pipeline if it's a new record (client code is free to pass an already saved pipeline if needed)
* Capturing generic Exception instead of StandardError (in case of Interrupts, Timeouts, ...)
* Auto-recovering from errors was failing on InvalidStatusError. Added a new :retry status to differentiate between a paused or auto-recoverable error

0.0.7
=====

Features:
* Allowing for user-defined default failure mode: when generic errors are raised, pipeline can be paused (default) or fail. Usage:
  class MySamplePipeline < Pipeline::Base
    self.default_failure_mode = :cancel
  end

Bug Fix:
* Refreshes the pipeline and stage statuses before execution (at runtime) in case of offline status update
* Rescue from ActiveRecord::RecordNotFound instead of relying on ActiveRecord#find(id) returning nil

0.0.6
=====

Bug Fix:
* Logging error details when a stage raises exception

0.0.5
=====

Bug Fix:
* Fixing association Stage -> PipelineInstance to use custom foreign_key

0.0.4
=====

Features:
* Allowing for custom name on stage

0.0.3
=====

Bug Fix:
* Autoloading delayed_job to avoid hardcoded dependency to collectiveidea's fork

Known Issue:
* Rails' tests/specs will only work if delayed_job is installed as a gem, rather than a plugin. Need to investigate further...

0.0.2
=====

Bug Fix:
* Bundling migrations on generator to avoid timestamp clashing

0.0.1
=====

Initial Release

Features:
* Execution of sequential user-defined stages in an asynchronous pipeline
* Persistence of pipeline instances and stages
* Error recovery strategies:
  * Irrecoverable errors fail the entire pipeline
  * Recoverable errors are automatically retried (using dj's exponential retry strategy)
  * Recoverable errors that require user input pause the pipeline for further retry
* Cancelling of a paused pipeline