General Regression Analysis Wrapper

[drewrip]

No description provided.

[kpouget]

thanks @drewrip ,

could you add a pydantic model to do define the regression result object?

the way I prefer to generate the object looks like this: (model definition, object generation)

def generate_lts_payload(results, import_settings, must_validate=False):

    lts_metadata = generate_lts_metadata(results, import_settings)
    lts_results = generate_lts_results(results)

    # ---

    lts_payload = types.SimpleNamespace()
    lts_payload.__dict__["$schema"] = f"urn:rhods-notebooks-perf:{models_lts.VERSION}"
    lts_payload.metadata = lts_metadata
    lts_payload.results = lts_results

    lts_payload.kpis = lts.generate_lts_kpis(lts_payload)
    lts_payload.regression_results = results.regression_results

    # ---

    json_lts = json.dumps(lts_payload, indent=4, default=functools.partial(json_dumper, strict=False))
    parsed_lts = json.loads(json_lts)

    return models.lts.Payload.parse_obj(parsed_lts)

can you make regression a package (a directory), and ZScoreIndicator a module (a file), and everything in __init__.py.
Not sure if we want the not implemented classes or not 🤔

[kpouget]

can you try to have a single return here, and use variables to save the explanation/status
that will make the code more readable (=always return a RegressionStatus, but if above threshold, set a different status/explanation)

[kpouget]

can you save the result of this call into a variable, and use the variable here?
this is a key the important part of the function body, shouldn't be "hidden" in the middle of the object creation

[drewrip]

Sure I'll make those more clear

[drewrip]

@kpouget How do you mean: and everything in __init__.py?

[kpouget]

in the regression directory, you'll create some modules/files,
and what does not belong to a module, you put it in regression/__init__.py, so that it's reachable with eg regression.RegressionStatus(