Allow task(s) with parsable logs to be configurable in CylcRoseManager

[edoyango]

Currently, CylcRoseManager searches through all task logs and attempts to parse the logs with a configurable known parsers property.

Probably makes more sense to specify which tasks' logs should be parsed by which parsers.

Perhaps could be passed in at initialization?

my_profiling = CylcRoseManagerChild(
    word_dir = Path(),
    task_parsers = {
        "*/*_um_fcst_*/job.out": UMTotalRuntimeParser(), # parse job.out log for task *_um_fcst_* for all cycles
        "2022_*/*_um_fcst_*/job.out": UMProfilingParser(), # parse job.out log for task *_um_fcst_* for only 2022_* cycles
    }
)

[micaeljtoliveira]

Couldn't this be done in a similar way as for the Payu-based models? For those models, the manager requires a function that returns all relevant ProfilingLogs. Those logs are stored in a dict of lists, whose keys are model components, but this could be adapted for Cylc.

[edoyango]

probably! I'm just unsure how fixed the task namings are. WHile payu models and dir structure are fairly fixed, since cylc is more generic - tasks can have arbitrary names so I think when I opened this issue I was hoping we could have some runtime way to specify which tasks we care about.

Additional context is that there are many tasks, and we might not care about having logs about all of them.