- make check job under control of the same submit system
- good abstraction on resource management (tmp files can also be a resource!)
- slurm plugin for check and submit
- exit of main if all jobs are checked
- docs for code
- manuals for cli
- more sophisticated print or logger system
- increase task coverage for future CI and collaboration
- CI slurm setup
- add pre and post hooks for main
- more neat exception hierachy and error code assignment
- more on fault tolerance and compatible beyond check_, (eg, resolve_)
- make slurmoo component more general and robust
- plugins support linux background job
- powerful report summary within subway query interface
- absorted state and absorting_ts for tasks in history.json
- excuatble version tracking
- strict version fix checking by hashing the file
- default job_count limitation (one line impl in sslurm)
- add option on sslurm plugin for directly rendering files
_from="template" - node list tracking for slurmoo
- input and output information: write in history.json or keep the convention, which is better?
- is there, by any chance, keyboard interrupt of main_rt would make things confused when run main_rt again? (seems the answer is yes, an ugly way out: increase the sleep time interval which is ok for long running task)
- add mechanism that one can manually add job naturally (seems current setup is good enough for this, just run kickstart again at anytime)
- dot in query and ~ in render, make they same in API and deeper level or leave it as it is