For example, take the reflective version of the asynchronous π-calculus.
| P,Q | ::= | 0 | // stop, or nil process |
| x[|P|] | // lift | ||
| x(y).P | // blocking input | ||
| P|Q | // parallel composition | ||
| *x | // drop | ||
| x,y | ::= | ^P^ | // quote |
Assign a dimension to each term constructor. Thus, we have
| 0-dim |
| lift-dim |
| input-dim |
| par-dim |
| drop-dim |
| quote-dim |
or 6 dimensions.
We define a recursive function, G[ - ]: L(P)xR2x5 → R6, assigning to each term a shape in 6 dimensions. The additional arguments are scale-factors and offsets for each term-constructor/dimension. For readability we will elide the additional argument, but for access, they are structured as a vector of pairs, ((ls,lo),(is,io),(ps,po),(ds,do),(qs,qo)), scale factor and offset for each dimension but the 0-dim.
| G[0](…) | = | (0,0,0,0,0,0) |
| G[x[|P|]](…) | = | { (r0,...,r5) | let cx = barycenter(G[x](…)) in let cp = barycenter(G[P](…)) in let ct = vertex equidistant to cx, cp and perpendicular to [cx,cp] in lift-direction in (r0,...,r5) in ls[cx,cp,ct] + lo } |
| G[x(y).P](…) | = | { (r0,...,r5) | let cx = barycenter(G[x](…)) in let cy = barycenter(G[y](…)) in let cp = barycenter(G[P](…)) in let ct = vertex equidistant to cx, cy, cp and perpendicular to [cx,cy,cp] in input-direction in (r0,...,r5) in is[cx,cy,cp,ct] + io} |
| G[P|Q](…) | = | { (r0,...,r5) | let cp = barycenter(G[P](…)) in let cq = barycenter(G[Q](…)) in let ct = vertex equidistant to cp, cq and perpendicular to [cp,cq] in the par-direction in (r0,...,r5) in ps[cp,cq,ct] + po } |
| G[*x](…) | = | { (r0,...,r5) | let cx = barycenter(G[x](…)) in let ct = vertex unit length from cx in drop-direction in (r0,...,r5) in ds[cx,ct] + do } |
| G[^P^](…) | = | { (r0,...,r5) | let cp = barycenter(G[P](…)) in let ct = vertex unit length from cp in quote-direction in (r0,...,r5) in qs[cp,ct] + qo } |
i believe that this will yield interesting visualizations of the terms of this calculus if we assign 3 dims to x,y,z and 3 dims to pitch, roll and yaw. The dynamics of execution will yield animations.
Posts
1 comment:
Should each 6-tuple be though of as a point or a vector (if using the (x, y, z, yaw, pitch, roll) coordinates)?
I'm somewhat curious as to what a simple data structure such as a linked-list or tree would look like over time in this kind of geometry.
Post a Comment