Access Result in a MultiResult by job name and not index

[cdtomkins]

This was originally proposed by @gbieli, @ubaumann as an outcome of discussions at AutoCon1 - see point 8 here (#911).

I would like to pick this up.

However, it's already possible to access Results in a MultiResult by job name using a generator, e.g.:

    print("Using a generator to pick out a result with a specific job name:")
    wanted = "Subtask NAME 2"
    subresult = next((r for r in v if r.name == wanted), None)
    if subresult:
        print(f"Found '{wanted}': {subresult.result}")
    else:
        print(f"'{wanted}' not found in results.")

So, can I confirm the acceptance criteria for this - would it be:

result["host1"].get_results_by_task_name("my_task")

aka:

def get_results_by_task_name(self, name: str) -> MultiResult:
    ...

My definition proposed about returns a MultiResult so it can handle multiple results with the same task name.

If we can confirm that is correct, I'll try to work up the change.

[dbarrosop]

However, it's already possible to access Results in a MultiResult by job name using a generator, e.g.:

I agree with this and that's the reason why I think this is a bit redundant/unnecessary, adding more surface to the API without any meaningful gain. Instead, I think it would be more interesting to document a few examples of common generators to filter the results and showcase how people can use them or build their custom ones.

[ubaumann]

I see your point. I think some helper functions (could be in nornir-utils) would be nice.

I am just thinking out loud: Would it be interesting to have the same filter logic as we have in the inventory?

[dbarrosop]

I don't have any opinion about nornir-utils.

I am just thinking out loud: Would it be interesting to have the same filter logic as we have in the inventory?

Yeah, if it is simple enough and doesn't add a lot of complexity that would be a much better solution IMO, at least more consistent. But I think it depends on if it adds a lot of complexity or not. The main difference between using the F filter in the hosts and the tasks is that while one is in the critical path (filtering hosts is something extremely common), filtering tasks isn't so much so complexity needs to be managed differently IMO.

[cdtomkins]

Thanks for the comments. I will look into how applying the filter logic for inventory to results and report back how complex that looks to mirror.