scalar inputs yield ndim=0 array outputs

[wholmgren]

Describe the bug
pvlib has a policy that the output type should be as similar as possible to the input type. So, for all scalar input, most functions should return a scalar output. However, some functions return ndim=0 array outputs instead of true scalars. I know this happens with irradiance.perez, but I think it happens with a handful of functions.

To Reproduce

    out = irradiance.perez(40, 180, 118.45831879, 939.95469881,
                           1321.1655834833093, 10.56413562, 144.76567754,
                           1.01688136)
    assert np.isscalar(out)

Expected behavior
The function should return a scalar, not an ndim=0 array.

Versions:

• pvlib.__version__: 0.6.0
• pandas.__version__: many
• numpy.__version__: many
• python: many

Additional context
We've previously discussed creating utility functions and decorators to handle type issues like this one. See #635 for example.

The first thing we need to do is catalog which functions exhibit this behavior.

[markcampanelli]

This is a bit of a PITA: numpy/numpy#12636

[mikofski]

I recently got bit by this too in scipy

[markcampanelli]

@mikofski Thanks for pushing on that vectorized newton in scipy, it has been quite useful. I am a bit concerned about the different output formats in the scalar vs. vectorized cases, however. Seems like that leads to the same issue we are discussing here.

[wholmgren]

Thanks @markcampanelli for raising the issue with numpy and bringing it to my attention here. I'm less convinced that we should put much effort into it within pvlib.