Appending an `Encoding` to a `Scene` about vello

[waywardmonkeys]

This might need feedback from @raphlinus or @armansito

[waywardmonkeys]

Some further thoughts:

Doing a Scene::append_encoding() or allowing Scene::from_encoding() and then using Scene::append() would both require that the BumpEstimator on the Scene be updated to include what was in the Encoding.

Doing a Scene::append_encoding() would require that either the Encoding have its own BumpEstimator or that it be repeatedly calculated on every Scene::append_encoding() call.

Doing a Scene::from_encoding() would allow for it to cache that encoding's estimator.

Doing the middle option of a Scene::get_encoding_mut() would complicate bump estimator code given the current design.

[waywardmonkeys]

This is all related to https://xi.zulipchat.com/#narrow/stream/147932-chatter/topic/SVG.20icon.20libraries

[armansito]

I think all of the approaches you described are reasonable. My main concern wrt adding a Scene::get_encoding_mut is that it may become a maintenance issue if we decide to add additional state to Scene that makes it more than a simple wrapper around Encoding.

The BumpEstimator is one such example as you pointed out. I'm not too worried about that particular case though, as I'm planning to move the estimation from scene construction time to resolve time (which happens later in the pipeline, right before instantiating the resources for a render).

[waywardmonkeys]

I have submitted a PR that allows for Scene::from(encoding) or encoding.into() returning a Scene. This would then work with passing the scene wrapping the encoding to Scene::append() without adding extra methods to Scene for append_encoding or the like and it is more typical Rust to do this than Scene::from_encoding().