Elamur Groupantstranslation-flow-many
The groupant beside tools/translate/src/many.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 | symbolic source derived | invariant-and-cost |
| descriptant | present | descriptant |
| invariant | partial | invariant-and-cost |
| tests | bounded helpers only | 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 |
|---|---|---|---|
behavior_complement | behavior_complement | partial source derived model | behavior_complement |
reverse_reconstruction | reverse_reconstruction | partial source derived model | reverse_reconstruction |
normalization | normalization | partial source derived model | normalization |
alignment_threshold | alignment_threshold | partial source derived model | alignment_threshold |
membership_alias | membership_alias | partial source derived model | membership_alias |
tie_order | tie_order | partial source derived model | tie_order |
Parts
- translation-flow-many-agreement
- translation-flow-many-storage
- translation-flow-many-stream_measure
- translation-flow-many-clustering
Model text
descriptant
translation-flow-many
versification averages same-verse Dice versus maximum adjacent-verse Dice and keeps strict superiority. agree builds punctuation-free concept/edge multisets with ROOT for head0 or out-of-range. pairing rotates common keys by floor(n/2). measure_lex trains state then streams each dev block, combines reader and target autoencoder tail, records missing whole blocks separately from cross residual, optionally retains dev metas. block_meta interns concepts and ORs translation bits. Reports average normalized pair scores and order squares by tradition then UPGMA.
Epoch 2026-10-31
Prospective predictions saved before edits in the external wave32 tr_flow receipts. Original production bytes preserved; only appended synthetic test modules added. Selected actual helpers exercised; five compiled behavior controls rejected with exact byte restoration and positive reruns. Source syntax projection is partial, not expanded semantic Rust. Diagnostic reconstruction with residual is not semantic understanding or persistent learning.
For agreement domain finite admitted counts with shared<=min(side totals), Dice lies[0,1], empty=0. UPGMA finite symmetric distances and representable sums, unique input indices. Membership interpretation requires distinct indices in0..64; modulo64 aliases outside it. Interner len<u32::MAX, arithmetic within u64, token head coordinates local to verse. Corpus execution source-derived only.
approximant
Forward relation
versification averages same-verse Dice versus maximum adjacent-verse Dice and keeps strict superiority. agree builds punctuation-free concept/edge multisets with ROOT for head0 or out-of-range. pairing rotates common keys by floor(n/2). measure_lex trains state then streams each dev block, combines reader and target autoencoder tail, records missing whole blocks separately from cross residual, optionally retains dev metas. block_meta interns concepts and ORs translation bits. Reports average normalized pair scores and order squares by tradition then UPGMA.
Scope: source-derived conditional transition model at the manifest SHA. For agreement domain finite admitted counts with shared<=min(side totals), Dice lies[0,1], empty=0. UPGMA finite symmetric distances and representable sums, unique input indices. Membership interpretation requires distinct indices in0..64; modulo64 aliases outside it. Interner len<u32::MAX, arithmetic within u64, token head coordinates local to verse. Corpus execution source-derived only.
composition
Composition
Scoped children are projections of this file, not independent executables. Main population filtering feeds aggregation; order/text helpers feed measure's exact decoder; many's agreement and storage read the same MTok with intentionally different punctuation treatment, streaming accounting consumes reader costs, clustering consumes normalized PairM. Uncovered branches remain source-derived; finite helper checks do not establish orchestration wiring.
catalog
Typed models
All entries bind the source SHA and optional byte range in manifest.json. Observation is one admitted helper call or the explicitly described orchestration trace. Relations are partial source-derived models, not universal proofs. Finite test names and unexecuted effects are in tests.md.
behavior_complement
Empty pairing cannot divide by zero because no mapped item executes; it emits no pairs. Strict versification never keeps tied mean scores. Agreement excludes punctuation whereas storage includes it, so counts are not interchangeable. Whole absent blocks do not inflate cross verse counts.
Domain and held-fixed assumptions: the invariant/forward documents for this exact scope. Status: partial source-derived relation; complement is a conditional inner exclusion, reverse retains ambiguity, other types specify the named projection/intervention. Oracle: relevant named synthetic tests when listed; otherwise source inspection only.
reverse_reconstruction
Agreement ratios admit many bag triples and graphs; even exact concepts/edges counts lose token order and punctuation. Pairing is invertible for fixed ordered keys and rotation length; ratios alone do not recover keys. Storage counts admit graphs with identical cardinalities; masks alias indices separated by64. UPGMA leaf order admits many distance matrices satisfying the same sequence of strict-min/tie comparisons.
Domain and held-fixed assumptions: the invariant/forward documents for this exact scope. Status: partial source-derived relation; complement is a conditional inner exclusion, reverse retains ambiguity, other types specify the named projection/intervention. Oracle: relevant named synthetic tests when listed; otherwise source inspection only.
normalization
PairM cross_per_kw=1000(reader+tail)/max(words,1), reader_kw=1000reader/max(words,1), per_block=(reader+tail)/max(blocks,1). These are pair-local denominators; mean over pairs is not pooled ratio.
Domain and held-fixed assumptions: the invariant/forward documents for this exact scope. Status: partial source-derived relation; complement is a conditional inner exclusion, reverse retains ambiguity, other types specify the named projection/intervention. Oracle: relevant named synthetic tests when listed; otherwise source inspection only.
alignment_threshold
Within a common verse pair, aligned>reader_bits([],target)+1 selects structural no-match; equality keeps aligned. Whole missing block increments canonical/structural levels and absent_words without cross reader/tail.
Domain and held-fixed assumptions: the invariant/forward documents for this exact scope. Status: partial source-derived relation; complement is a conditional inner exclusion, reverse retains ambiguity, other types specify the named projection/intervention. Oracle: relevant named synthetic tests when listed; otherwise source inspection only.
membership_alias
Storage bit=1<<(translation_index%64). Distinct indices0 and64 collapse membership while tokens still sum. This is a representation limitation, not evidence of64+ unique memberships.
Domain and held-fixed assumptions: the invariant/forward documents for this exact scope. Status: partial source-derived relation; complement is a conditional inner exclusion, reverse retains ambiguity, other types specify the named projection/intervention. Oracle: relevant named synthetic tests when listed; otherwise source inspection only.
tie_order
UPGMA scans cluster vector indices lexicographically and replaces best only on strict lower mean distance; merged right appends to left. Finite equal distances preserve first merge order, not numeric sorting of supplied item labels.
Domain and held-fixed assumptions: the invariant/forward documents for this exact scope. Status: partial source-derived relation; complement is a conditional inner exclusion, reverse retains ambiguity, other types specify the named projection/intervention. Oracle: relevant named synthetic tests when listed; otherwise source inspection only.
tests
Evidence and gaps
Native checks
many::tests::clustering_puts_near_ones_togethermany::tests::agreement_and_its_controlmany::tests::a_block_meta_world_is_stored_oncemany::tests::versification_drops_a_shifted_chaptermany::wave32::empty_ties_and_accounting
These selected names ran twice debug and twice release with exact filtering and one test thread; raw logs and numeric status files are in the external octopus-wave32-tr_flow-2026-10-31 receipt directory. All test bodies inspected before selection. Inline cases use only synthetic in-memory records; the existing blocks filesystem test was not selected. Tests listed at file level do not execute every function or this entire scoped fragment.
Controls
Five compiled source-behavior mutants: main gate counter suppressed; many strict chapter comparison relaxed; measure noncanonical-gap recording disabled; pages Meta filter reversed; blocks literal-change predicate reversed. All failed the corresponding assertion and passed after original bytes/hash restoration. Zero assertion-only mutants; zero surviving exploratory mutants.
Limitations
CLI argument/effect paths, live corpus/model/annotation/training, full Run::verse, page filesystem and reports remain source-derived only. Source syntax projections generated twice and byte-compared. No corpus-quality, persistent-learning, throughput or universal semantic coverage claim.
invariant-and-cost
Domain, invariant and cost
Invariant
For agreement domain finite admitted counts with shared<=min(side totals), Dice lies[0,1], empty=0. UPGMA finite symmetric distances and representable sums, unique input indices. Membership interpretation requires distinct indices in0..64; modulo64 aliases outside it. Interner len<u32::MAX, arithmetic within u64, token head coordinates local to verse. Corpus execution source-derived only.
Cost
For N tokens,V interned concepts,E edges, bags and storage O(N log(V+E)); conceptual compact byte count12C+17E+8T excludes BTreeMap/string/allocator overhead. Pair comparison O(P² times reader/logic work per block); training corpora retained, current block metas transient, optional kept metas cumulative. UPGMA recomputes O(n²) cross-pair visits each of n merges => O(n³) distance calls. Rendering squares O(number of metrics*n²) cells plus sort O(n²log n).
Integer additions/casts require representability; finite floating inputs where comparisons are modeled. These symbolic costs are not measurements of corpus throughput. No useful cheaper executable percentage surrogate established; future supported caps only 1/2/5/10%.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/tools/translate/src/many.rs.groupant/