Cookie Cutter

Cookie Cutter is a two-dimensional analog programming language.

A Cookie Cutter program, P, is a set of points in a two-dimensional plane, that is applied to the data, D, which is a set of points in a two-dimensional plane. The input is the initial data.

For purposes of exposition, points in the plane will be specified using Cartesian coordinates.

Definitions
The data, D, can be divided into edge points, E, and interior points, I.

A point, (x,y), is an edge point if there exists an (α,β) for which there exists a δ > 0 for which for all 0 > ε > δ, (x+εα,y+εβ) ∉ D.

A point, (x,y), is an interior point if for all (α,β) there exists a δ > 0 for which for all 0 > ε > δ, (x+εα,y+εβ) ∈ D.

The data, D, can also be divided into 0 or more subregions. Any point in a subregion S can be connected to any other point in S by a continuous path that consists only of points in I. Subregions may overlap, but the overlap can only contain points in E.

The program P can be mapped to P(a,b,s) by scaling and translation, where a point (x,y) is mapped to (sx+a,sy+b).

A mapping P(a,b,s) is contained by a subregion S if all points in P(a,b,s) are in S.

A mapping P(a,b,s) is interior to S if P(a,b,s) is contained by S and all points in P(a,b,s) are interior points.

A mapping P(a,b,s) is an executable mapping in S if it is contained by S but not interior to S and for all (α,β,σ) ≠ (0,0,0), where σ ≥ 0, there exists a δ > 0 for which for all 0 > ε > δ, P(a+εα,b+εβ,s+εσ) either is interior to S or is not contained by S.

Execution
At each step, if D has no subregions or if no subregion of D has an executable mapping, execution terminates.

If there is a single executable mapping, P(a,b,s), the output for the step is s,{(a,b)} and the points in P(a,b,s) are either removed from D or become edge points in D. The new edge points do not necessarily border points not in D, but do serve to divide D into new subregions.

If there are multiple executable mappings, then if there is a single executable mapping P(a,b,s) with the greatest s, the output is s,{(a,b)}, and P(a,b,s) divides D into new subregions.

If there are multiple executable mappings P(ai,bi,s) with the greatest s, and there is no point that is in more than one of the executable mappings, the output is s,{(ai,bi)}, and all the executable mappings with the greatest s divide D into new subregions.

If there are multiple executable mappings P(ai,bi,s) with the greatest s, and points that are in more than one '''executable mapping''' form a 2-dimensional area, execution terminates.

If there are multiple executable mappings P(ai,bi,s) with the greatest s, and points that are in more than one '''executable mapping''' form either 0-dimensional points or 1-dimensional curves, the output is s,{(ai,bi)}, and all the executable mapping with the greatest s divide D into new subregions.

Examples
D is the square, -1 ≤ x ≤ 1 and -1 ≤ y ≤ 1. P is the same square, -1 ≤ x ≤ 1 and -1 ≤ y ≤ 1. The first step is 1,{(0,0)}, then execution terminates because D is empty.

D is the square, -1 ≤ x ≤ 1 and -1 ≤ y ≤ 1. P is the square, -1 < x < 1 and -1 < y < 1. The first step is 1,{(0,0)}, then execution terminates because there are no subregions.

D is the square, -1 < x < 1 and -1 < y < 1. P is the square, -1 ≤ x ≤ 1 and -1 ≤ y ≤ 1. Execution terminates immediately because there are no executable mappings.

D is the square, -1 ≤ x ≤ 1 and -1 ≤ y ≤ 1. P is the line segment, -1 ≤ x ≤ 1 and y = 0. Execution terminates immediately because there are no executable mappings.

D is the square, -1 ≤ x ≤ 1 and -1 ≤ y ≤ 1. P is the cross, -1 ≤ x ≤ 1, y = 0, and x = 0, -1 ≤ y ≤ 1. The first step is 1,{(0,0)}, then the second step is ½,{(-½,-½),(½,-½),(-½,½),(½,½)}, etc.

D is the rectangle, -1 ≤ x ≤ 2 and -1 ≤ y ≤ 1. P is the cross, -1 ≤ x ≤ 1, y = 0, and x = 0, -1 ≤ y ≤ 1. Execution terminates immediately because there are no executable mappings.

D is the square, -1 ≤ x ≤ 1 and -1 ≤ y ≤ 1. P is the annulus, γ ≤ x2 + y2 ≤ 1, where 0 < γ < 1. The first step is 2,{(0,0)}. Subsequent steps depend on γ. Larger values of γ means the second step is 2γ,{(0,0)}, smaller values of γ means there are 4 executable mappings in the second step, and, for one value of γ, there are 5 executable mappings in the second step.

Additional Note
I don't know what would happen if P or D were fractal. If D is fractal, there can be points in D that are neither edge points nor interior points. My guess is that, in most cases, execution would terminate.