I think an initializer may be a more natural choice here than a render transform. Render transforms are the most general, but are necessarily tied to the specific rendered form of the underlying mark (e.g., the dy attribute). An initializer in contrast can operate solely on the x and y channels without needing to know how the mark is rendered. Also, we already have the dodge initializer, and this is closely related.

yes, that's even simpler, thanks!