wordgraph-tamp-build
The groupant beside wordgraph/tamp/src/build.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 | Approximant |
| Ast | partial | syntax.json |
| Composition | partial | Composition |
| Cost | partial | Cost |
| Descriptant | present | Descriptant |
| Invariant | partial | Invariant |
| Tests | executed bounded checks | 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.
| relation | type | label | where |
|---|---|---|---|
| Behavior complement | behavior complement | partial source derived model | Behavior complement |
| Reverse reconstruction | reverse reconstruction | partial source derived model | Reverse reconstruction |
Model text
Descriptant
One build: synonyms, template and anchor nodes (anchor first; same-name anchors merged; templates over the rest), canonical members and names, payload test, importance, per-paper records (important statements, node references, compression ratios), summary, threshold and node samples, anchor diagnostics, coherence; and the controls (shuffled only, mixed real + shuffled copies, foreign statements).
Approximant
nodes = anchor ⊎ template; node_of(u) = its node; boilerplate(u) = node_of(u) ∧ ¬payload(u); words_with_refs(p) = Σ_{u ∉ B} w(u) + Σ_{node} w(ref); ratio_refs = words / words_with_refs.
Invariant
Every reference names a node of the build; outputs are byte-identical across runs and thread counts (measured twice and with 3 threads, 2026-11-06); a payload member is never counted as saved words.
Cost
≈ 60 s and 1.7 GB for 1,763 papers; controls ≈ 140 s and 3.7 GB (three clusterings, one doubled corpus).
Composition
Uses every other module; called by main.rs.
Tests
Real builds and controls (README, 2026-11-06); the whale hook test reads its record format.
Behavior complement
Writes no description, calls no language model, publishes nothing; does not delete earlier outputs.
Reverse reconstruction
Paper records keep every statement's score and node; nodes keep members, fits and papers; the summary keeps all parameters.
Provenance
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/wordgraph/tamp/src/build.rs.groupant/