Perspective deformation of square vertices yields extra point

[byorgey]

> let sq = unitSquare # rotateBy (1/17) # translate (3 ^& 2) :: Path V2 Double
> let pts = concat $ pathPoints sq
> pts
[P (V2 3.6468569477957544 1.7143847183913987),P (V2 3.2856152816086013 2.6468569477957544),P (V2 2.3531430522042456 2.2856152816086013),P (V2 2.7143847183913983 1.3531430522042456)]
> map (\p -> p ^/ (p ^. _x)) pts
[P (V2 1.0 0.47009925065133495),P (V2 1.0 0.8055894317912602),P (V2 1.0 0.9713031596050272),P (V2 1.0 0.4985082044693159)]
> concat . pathPoints $ deform perspectiveX1 sq
[P (V2 1.0 0.47009925065133495),P (V2 1.0 0.8055894317912602),P (V2 1.0 0.9713031596050272),P (V2 1.0 0.7180539502766538),P (V2 1.0 0.4985082044693159)]

The last two expressions ought to yield equal results, but there is an extra spurious point inserted in the result of calling deform perspectiveX1.

[byorgey]

Maybe this is not actually a bug, but just due to some numerical approximation, i.e. the system thinks it needs to subdivide one of the segments to approximate its deformation accurately enough. Perhaps the real solution is that I should just directly deform points instead of deforming paths where I only care about the vertices.

[byorgey]

I'm not sure whether this is a bug, but I worked around it by directly deforming points, which was probably a better way to do it anyway. I'm going to close this for now.

[bergey]

OK. I can easily believe either that it's a bug or that it's subdivision, as you say.