value-set-transition

The groupant beside world/src/values.rs in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.

Facts

Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.

Components

component	label	where
approximant	partial	Forward
ast	partial	`../../syntax.json`
composition	partial	Composition
cost	symbolic unmeasured	Cost
descriptant	present	`model.md`
invariant	partial	Invariant
tests	bounded evidence	tests

Obernants

The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.

id	type	label	where
`nonbehavior`	behavior_complement	partial source derived model	nonbehavior
`antecedents`	reverse_reconstruction	partial source derived model	antecedents
`command-snapshot`	snapshot_boundary	partial source derived model	command-snapshot
`lifetime-idempotence`	history_idempotence	partial source derived model	lifetime-idempotence
`representative-selection`	lexical_minimum	partial source derived model	representative-selection
`field-permutation`	input_equivalence	partial source derived model	field-permutation
`reciprocity-time`	temporal_precondition	partial source derived model	reciprocity-time
`single-field-refinement`	restricted_refinement	partial source derived model	single-field-refinement

Model text

descriptant

External causal set transition

Forward

Freeze h=prior harmed entry exists, s=current agent belongs to prior harm history, p=current directed help pair exists, q=earlier reverse help exists. Eligibility requires a resolved event, character target and agent, and distinct identities. Ignore cause/goal/belief/name/trait keys. H means a harm field exists; R means a relief-but-not-harm field exists, preserving harm-first classification. Then dH=eligible AND H AND NOT s; dP=eligible AND R AND h AND NOT p; mercy=dP AND s; reciprocity=dP AND q.

Append the harm agent iff dH; append (agent,target,current line) to helped iff dP. These updates do not require principle rules to exist. Emit each judgment only if its rule exists. The first lexical qualifying key selects each explanation; merge the harm emission and help bundle according to their key positions. The help bundle orders relief, mercy, reciprocity.

Invariant

Histories are monotone and unique per target/agent or directed help pair. New same-command harm cannot qualify as prior harm. Reachable harmed entries are nonempty; arbitrary fabricated empty entries are outside this model's domain.

Composition

The parent records goals and self-achievement before this child's external-cause guard. This child freezes eligibility before ordered field traversal. Names/events supply explanations; global state classes and principle availability remain external dependencies. Reverse reconstruction retains every reachable prior history satisfying successor union/append and ordered emission equations. If only judgments are observed, successor history is existentially hidden; empty output can conceal rule-independent history changes.

Cost

Snapshot membership costs O(H_x). A conservative bound for F fields and P helped tuples is O(F(H_x+P)), excluding string and rule lookup costs. Parsing and achievement are parent work, not additional child work. No timing or speedup claim.

Epoch memory

2026-10-29: the historical mixed-field witness showed differing membership, even after sorting. Canonical order alone would select one accidental within-command causal interpretation. Frozen command-start history repairs that boundary while preserving lifetime suppression. Syntax is a current source-bound partial projection; semantic AST analysis and universal correspondence remain absent.

catalog

Command-start set-transition typed models

Current revision is bound by manifest.json. Domain: one external causal set command, reachable histories, successful allocation, fixed global tables and finite nonoverflowing inputs. Source relations are partial; bounded correspondence does not establish universal semantics.

nonbehavior

Type: behavior_complement. U contains reachable prior histories, a command and successor histories/ordered observations. Excluded transitions include new help without prior target harm, mercy without prior harm by this agent, repeated helped pairs and history changes under ineligible cause. This is a candidate inner bound on U minus actual behavior.