Pluggable action engine
The idea for this is to create a new components used by action_engine, for now called the plug-engine-
Certain routines that would be pluggable (to start):
- Compilation (of flows/tasks) -> graph (allowing others to generate new patterns or graph translation techniques)
- Scheduling (allowing others to provide there own techniques for scheduling nodes).
- Completion (allowing others to provider there own techniques for completing nodes/futures).
- The run loop that exists right now in graph_action execute_iter() method.
The existing action_engine would use a standard set of components but allow others to replace attributes or derive from the action_engine to provide there own default sets of plugins (so its interface and usage would still be unaffected).
Blueprint information
- Status:
- Complete
- Approver:
- Joshua Harlow
- Priority:
- Medium
- Drafter:
- Joshua Harlow
- Direction:
- Approved
- Assignee:
- Joshua Harlow
- Definition:
- Approved
- Series goal:
- Accepted for 0.4
- Implementation:
- Implemented
- Milestone target:
- None
- Started by
- Joshua Harlow
- Completed by
- Joshua Harlow
Related branches
Related bugs
Sprints
Whiteboard
Gerrit topic: https:/
Addressed by: https:/
Create and use a new compilation module
Addressed by: https:/
Rename t_storage to engine_storage
Addressed by: https:/
Add a runtime module
Gerrit topic: https:/
Addressed by: https:/
Make the runner a runtime provided property
Addressed by: https:/
Make action handlers extensible and pluggable