IssueTracker

the current version of Shapes (alpha04) has a calculateBounds() function on RoundedPolygon that estimates the bounds based on the (x,y) location of all anchors and control points, where the returned boujnds are the min/max bounding box of all of those points. This gives a result which is correct or close in most cases, but can be a large overestimate of the bounds in some cases. Specifically, if the control points of a shape are quite far away from the outline of the shape, this can result in a much larger bounds than the shape actually has. For example, a clover shape with rounded corners that are very far away from the vertices they are rounding will result in the attached screenshot (where the green square is the estimated bounds).

Meanwhile, the bounds are also limited to whatever the screen orientation of the shape is at the time. So if the caller was hoping to get the bounds once to calculate the size of the bounding view/composable and then rotate the shape within that area, they may find that the shape exceeds those bounds as it changes orientation. For a simple example, imagine a rectangle that is long in the x direction but short in the y direction, which would result in bounds that are similarly wide horizontally but short vertically, Rotating that rectangle would exceed the bounds in the Y direction quickly as the corners of the rectangle rotated outside of the original bounds.

It seems like we should provide more functionality for bounds calculations in two ways:

1) Exact bounds: Approximate bounds are sufficient for many purposes, and have the advantage of being cheap to compute (we are simply doing a bunch of float compares against all points in a shape). But if a caller wants the exact bounds for some reason, we should be able to calculate those by checking the bounds of the actual curves instead of just the points that created those curves. For example, in the attached screenshot of the clover, exact bounds would look closer to the yellow box than the green box, where the bounds met the outline of the clover's curves.

2) Max bounds: If a caller intends to rotate or otherwise change the orientation of a shape, we could calculate the max bounds. This is essentially the farthest point on any of the curves from the center, doubled, in both directions. For example, if the farthest point on a curve were (100, 100) away from the center, then the max bounds would be (-100, -100), (100, 100).

The caller can come up with these values already; they have access to the same underlying cubics that we would use to calculate these new bounds values. But it seems like the kind of utility that the library should provide directly to make it easier to position these shapes in a UI, where bounds help determine placement and sizing of the views/composables which contain these shapes.