fix: remove jobstep

[BEFH]

Features:

• Better support for MPI and span strings

Fixes:

• Remove unnecessary jobstep executor
• Change generator type hint
• Move some code out of the main Executor method

[BEFH]

@cmeesters I have integrated (and tested) some of our suggested changes. Could you please particularly take a look at what I've done for the span arguments?

[cmeesters]

Hi Brian,

pew, sorry, I was busy.

I think there was a misunderstanding: You still need the jobstep executor. Just like in lines 58 here.

The Jobstep Executor in turn calls Snakemake again. Except for the MPI case, where merely the shell command is executed. What you do not need in the LSF case is a handling of a true jobstep like in SLURM.

Does this make sense to you?

[BEFH]

I tested this and it seems to be working without the jobstep. I'll post the test results shortly.

[cmeesters]

I doubt this: for ordinary SMP jobs and group jobs, yes. But for MPI Jobs?

[BEFH]

On LSF, MPI commands are just run from within jobs like any other commands, so I think it should be fine for MPI.

But if you point me to your pi test workflow, I can get confirmation one way or another.