aggregate and resample_by_interpolation

[paulf81]

the WindRose class has two functions for something like resampling, aggregate and resample_by_interpolation. aggregate sums the frequencies together and only allows combining into fewer bins. resample_by_interpolation uses interpolation of frequencies and can be used for upsampling or downsampling. In my testing using resample_by_interpolation for upsampling works in an intuitive way but for downsampling can work, I think basically correctly, but in perhaps surprising way, since the frequency of a larger bin made by combining say 10 bins is set totally by those nearest the center of the new bin. This is the correct behavior I think, but can yield results that might surprise on first inspection.

So wanted to open this discussion, since aggregate only allows downsampling, should resample_by_interpolation only allow upsampling? Should we also push their names slowly (without immediate breaking changes) to upsample and downsample?

[misi9170]

I'd support that. I think that aggregation is the appropriate way to downsample, and interpolating could lead to some odd results. Interpolating can probably also produce odd results when upsampling, but in that case it's a bit to be expected (since one is creating resolution that wasn't originally there)

[paulf81]

What if we add upsample and downsample for now as alias versions of the functions so we support either call, and then one day phase out the old names?

[misi9170]

Yeah, I would say we rename the functions to upsample() and downsample(), so that anyone coming to FLORIS from now sees that those is where the actual code lies, and then add new functions with the old names that are just aliases for the upsample() and downsample() methods... I think that's the way to do it?