Should functions check to see if boolean inputs are actually boolean?

[williamhobbs]

I noticed that passing

backtracking = 'false'

in pvlib.tracking.singleaxis() results in backtracking being included. Should this, and other functions, check to make sure that boolean inputs are actually boolean and not a string that is not being interpreted as the user might have expected?

I'm guessing this is a general Python best practices question, but I honestly don't know.

[cwhanse]

We should could restructure that check. Any "truthy" value (string, non-zero number, etc.) will be interpreted as True.

[echedey-ls]

I understand your point @cwhanse . But if the problem is exclusively with this function, then the error is relatively easy to debug with plots (and also there many examples using backtrack=False). IDEs should also help with that. At least in VSCode (sorry for spamming again) with Pylance extension you get the option to opt-in for type checking:

And that's by default, I don't need to specify explicitly backtrack: bool = True in tracking.singleaxis signature.

[williamhobbs]

I don't think pvlib should fight python semantics.

@wholmgren, I thought this might be the case, i just didn't know what standard Python semantics were in this situation.

[mikofski]

According to the docs it is unpythonic to “look before you leap” and pythonic to follow “easier to ask forgiveness than permission”. This is explained in Microsoft developer Brett Cannon’s blogpost EAFP vs LBYL.

[cwhanse]

All I'm suggesting is that we could consider replacing

if backtrack:

with

if backtrack is True:

That would probably alarm the linter, but it would restrict when backtracking code is executed to only backtrack=True.

In any case, I'm not invested in making a change here, only wanting to help pvlib be easier for new users.

[adriesse]

In my own code I have a habit of using 0 and 1 to set boolean parameters, but I have noticed it doesn't always work with other packages. I have a very slight preference for leaving it the way it is.

[adriesse]

Is there perhaps an inherent benefit in having default values be False or None?

[echedey-ls]

You are certainly right @adriesse !
In my comment up there my IDE/static code checker detects it should be a bool value because the default is False. If it had been None then the inference would be of type Any, and no warning would have been emitted.