Consider working with TopoJSON.
Before the clipping algorithm is used, the two collections of polygons should represent a topology, and clipping should produce a topology. The clipping algorithm should be designed to preserve topology in the face of floating point arithmetic errors.
It might still make sense to return an array of polygons as the API does currently. It is unlikely consumers would like to work with TopoJSON.
What we are aiming for is robustness in the clipping over overlaid squares. See Chris Tralie's explanation. The current approach simply doesn't do enough to guarantee that clipping works well with floating point vertex coordinates.
Consider working with TopoJSON.
Before the clipping algorithm is used, the two collections of polygons should represent a topology, and clipping should produce a topology. The clipping algorithm should be designed to preserve topology in the face of floating point arithmetic errors.
It might still make sense to return an array of polygons as the API does currently. It is unlikely consumers would like to work with TopoJSON.
What we are aiming for is robustness in the clipping over overlaid squares. See Chris Tralie's explanation. The current approach simply doesn't do enough to guarantee that clipping works well with floating point vertex coordinates.