Elamur Groupantsworld-lessons-draw
The groupant beside world/src/lessons.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 source model | forward |
| ast | partial | ../syntax.json |
| composition | partial source model | composition |
| cost | partial source model | cost |
| descriptant | partial source model | epoch-memory |
| invariant | partial source model | invariant |
| tests | partial source model | checks |
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 |
|---|---|---|---|
world-lessons-draw-behavior-complement | behavior_complement | partial source model | behavior-complement |
world-lessons-draw-reverse-reconstruction | reverse_reconstruction | partial source model | reverse-reconstruction |
world-lessons-draw-coupling | constraint_relation | partial source model | coupling |
Model text
forward
At most 2000 attempts: bind variables in order, derive lets in order, require all needs nonzero, evaluate nonnegative answer, fill selected template k modulo template count. Empty templates fail before RNG mutation. Exhausted unique names and invalid/reversed ranges fail. Inclusive numeric sampling uses i128 width and u128 residue, including full i64 domain. Wrapping seed addition defines large lesson numbers.
behavior-complement
U is lesson/RNG/template/index inputs and output/final-RNG pairs with bounded expressions and helper domains. Absent success includes negative answer, unsatisfied conditions, exhausted names, reversed intervals and empty templates. Exhaustion returns None rather than guaranteeing a sample.
reverse-reconstruction
B(d,s1) includes all initial seeds and rejected-attempt histories that produce d and final RNG s1. Modulo reductions and dropped variables lose information. Output alone cannot recover attempt count or template identity.
coupling
Same seed and supplied lesson reproduce text/answer. Exercises/test/exam choose distinct containers and seed constants, but neither seed separation nor containers prove content disjointness. RNG(0) is an absorbing state; no uniformity promise. Failed attempts consume RNG state except early structural rejection.
invariant
Successful observations obey the declared forward relation with the same supplied state and order. Failure is explicit; no hidden label or retained-state guarantee is introduced.
cost
Bound 2000 times variable binding, expression and render costs. Name filtering bounded by 16-name table. Batch cost multiplies by requested count; accepted count can be smaller. No live learning or persistence modeled.
composition
Evaluation and substitution share numeric bindings; field labels choose template containers. Child relations must agree on intermediate state, not merely final counts. Partial syntax binds the source; macro contents are opaque.
epoch-memory
Initial review predicted arithmetic overflow and finite-pool boundary failures. Old native code failed six named witnesses: overflow, empty templates, name exhaustion, full-width range, reversed range and seed overflow. Narrow repairs retain public signatures; finite observations do not establish arbitrary-input totality. See file evidence.md for hashes and repeated results.
checks
Eight wave12_models tests plus two existing lesson tests. Independent 19 by 19 by 7 arithmetic cases, precedence/eager failure literals, partition/substitution fixtures and exactly 2000 rejected RNG draws. Runtime guard mutations must fail named assertions. Evidence is finite; no whole-world or live course claim.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/world/src/lessons.rs.groupant/draw.groupant/