Finite stack for one edge-CSP improvement #
Frames retain the prepared original body, its finite occurrence buffer and its actual cycle. Descent performs a fresh native forest search and pushes exactly one frame per contraction. The first improvement unwinds those frames once, lifting an arbitrary improving child buffer. An exhausted child propagates optimality immediately. There is no recursive optimization of children. Clocks use actual occurrence-list lengths, independently of binary label magnitudes.
This program uses the original CSP Table ℕ, existing occurrence buffers, native forest
codes and augmenting-walk codes. Preparation, contraction and lifting execute actual finite
buffer operations, with no implementation or lifting provider as input. Agreement proves
exact mathematical agreement; PolyTime supplies ordinary native certificates with full
encoded storage bounds.
Prepared parent problem and the cycle used for its actual whole-instance contraction.
- parent : EdgeForest.Input
Original body and exact-width prepared buffer.
- cycle : EdgeAugmenting.WalkCode
Binary original base name and binary global occurrence positions.
Instances For
The tuple of the fields of encFrame.
Instances For
Encode the actual positional labeling by scanning actual incidences.
Instances For
Read the actual blossom event, perform its preparatory flip and construct the child.
All old constraint positions and all nullaries are retained by Blossom.contract.
Instances For
Flip the actual decoded native augmentation and return its exact-width buffer.
Instances For
Lift one actual child improvement into its prepared parent. Every outside change in the
comparison is retained; the full liftComparison construction selects and lifts its walk.
Instances For
Finite stack state. Status is running, proved optimal, or an improving finite buffer.
- current : EdgeForest.Input
Current contracted problem; restored to the parent after every pop.
Prepared parents, newest first.
- budget : ℕ
Remaining contraction depth, in unary and funded by the original occurrence list.
nonedescends;some nonestops;some (some bits)unwinds.
Instances For
One descent or one lift. A stopped optimum needs no parent labeling reconstruction: count-preserving preparation and reference optimality transfer certify the original optimum.
Instances For
One-improvement stack program. Each contraction uses one descent and at most one lift. The additional steps fund the terminal search and final status. Clock value is unary and funded by the input occurrence list; no binary name or arity is expanded.
Instances For
Outer optimization state: current original buffer and a flag preserving first exhaustion.
Instances For
Improve the original problem once, stopping permanently after optimality is established.
Instances For
Guarded finite optimization program. The occurrence list pays for all outer iterations. Malformed models reject before any forest search or contraction.
Instances For
Complete guarded finite decision program.