Guardrails for functions with inputs that have numerical bounds

[kurt-rhee]

Hello pvlib,

I was wondering if it would be useful to have some warnings in functions which have numerical inputs that have bounds, but don't error out if the boundary conditions are not met. Happy to contribute, if it would be useful.

For example, pvlib.tracking.singleaxis takes GCR as an input. In the documentation the function asks for a number from 0 to 1, but any number could theoretically be used as an input. If a number outside of 0 and 1 is used, then the function gives interesting results.

import warnings

def check_bound(value, lower_bound, upper_bound):
    if value < lower_bound or value > upper_bound:
        warnings.warn(f"Value {value} is outside the allowed range [{lower_bound}, {upper_bound}].")

[kandersolar]

It's an interesting idea that comes up now and again. See #433 for general related discussion and more recently #1804 for a specific instance.

I have the sense that we aren't interested in adding this kind of checking in the functions. But there is talk (maybe not public, yet) about adding checks along these lines for ModelChain, maybe as an optional post-processing step on the ModelChainResult or something.

[kurt-rhee]

Thanks @kandersolar, #433's speed comments from Mark are very compelling. I like the idea of a separate "checker" where you could maybe hold all of your inputs in a dictionary and check them separately from the actual runtime of the function.

[adriesse]

Range checks could be made conditional, like assertions, to manage a potential performance hit.