Add missing map2/apply2 functions to `Pixel` trait about image HOT 4 OPEN

The Pixel trait already having a lot of functions is a reason to avoid adding more. Could you share a bit more on what cases these methods would be needed for, and whether there are alternatives ways to implement those?

from image.

The description mentions an example, which is image math (e.g. taking a Laplacian transform and subtracting that from an image's original content, where we want to preserve alpha but operate on the color channels. These seem like elementary operations that are missing, not exactly extra cruft. Is there another way to perform those operations with comparable performance? If so, why are any of the pixel-wise operations present?

from image.

Yes there are other ways to get as good, if not better, performance. The trait exposes the underlying components and if you don't need to be generic over all implementations of Pixel you can rely on additional assumptions about the layout. For best performance, you probably want to operate on directly slices and fine-tune your code until it auto-vectorizes.

From looking closer at the PR, it would actually require a breaking change to add these methods because the Pixel trait isn't sealed. The whole Pixel trait / generic operations over images probably warrants a more substantial overhaul that includes breaking changes. But it would be a lot of work and I don't have the time to work on that

from image.

Can you give a discrete example of how to accomplish map2_without_alpha using existing functions only?

from image.