Should decomposition functions echo GHI in return?

[mikofski]

As proposed in #1179 should decomposition functions like Erbs also echo GHI in the return?

[kandersolar]

I like the idea of the decomposition functions having a common return structure (kind of like #1553). I'd want that structure to include (at least) all three of ghi, dni, and dhi.

One problem is that not all of those functions return frame-like structures which make it easy to add another column; dirindex and dirint just return dni by itself, so it would be breaking to change them to return multi-column structures. I think it's probably worth it, but just wanted to point out that this change wouldn't be purely additive.

[wholmgren]

I think not. I would prefer to see erbs and boland return only dni. I'd also favor a little refactoring to create functions like diffuse_fraction_erbs and diffuse_fraction_boland.

[cwhanse]

Do you mean return only dhi? They are diffuse fraction models.

[wholmgren]

I had dirint on my mind so I was thinking dni. I would be fine with only dhi.

[cwhanse]

Maybe returning all of ghi, dhi and dni should be the job of a general function like get_poa_components

[adriesse]

Maybe returning all of ghi, dhi and dni should be the job of a general function like get_poa_components

And then we could have some simpler base functions that just have the main model equations, like erbs mapping kt to df.