I'm confused why the transformation above is `i*q + (-q/2)`, but the transformation in this function is effectively `i*q + (-q)`. My mental model is that there's a "sample" domain and then the "pixel" domain, so the transformation function should be the same in both situations. I tried experimenting, and it looks like it's correct (as in if I use the same function in both places, things are offset: https://observablehq.com/@pcarleton/working-with-d3-contours) Am I missing something?