Module candidates 

Candidate generators.

Each generator proposes a bounded, deterministic reconstruction hypothesis. Generators are cheap to decline and never trusted because their logic “looks obvious”: every candidate reaches the common court.

Structs

Candidate

A proposed descriptor together with its originating family.

Enums

CandidateKind

Stable candidate identity. The discriminant order is the final deterministic tie-breaker when prices and work are equal.

Functions

propose

Generate the bounded candidate set for input.

propose_all

Generate the complete bounded candidate set for input.

propose_byte_rans

Propose an order-0 byte-rANS representation of the whole input as one entropy channel.

propose_each

Generate the candidate set one candidate at a time, invoking emit for each. Same set, same order as propose_all, but the streaming form lets the court keep only one unpriced candidate’s payload resident, which bounds encoder memory on large inputs (Phase 14).

propose_forced

Generate exactly the candidate of kind, or Ok(None) when it does not apply.

propose_rle

Propose an inline-run + REPEAT_LAST representation of input.