Elamur Groupants⌕
☰

translation-flow-corpus

The groupant beside tools/translate/src/corpus.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

componentlabelwhere
approximantpartialapproximant
astpartialsyntax.json
costsymbolic source derivedinvariant-and-cost
descriptantpresentdescriptant
invariantpartialinvariant-and-cost
testsbounded helpers onlytests

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.

idtypelabelwhere
behavior_complementbehavior_complementpartial source derived modelbehavior_complement
reverse_reconstructionreverse_reconstructionpartial source derived modelreverse_reconstruction
parallel_orderparallel_orderpartial source derived modelparallel_order
compaction_losscompaction_losspartial source derived modelcompaction_loss
head_transporthead_transportpartial source derived modelhead_transport
anchor_priorityanchor_prioritypartial source derived modelanchor_priority

Model text

descriptant

translation-flow-corpus

PVerse::space derives next-gap nonemptiness and false last bit; meta_words chooses mw iff nonempty; compact drops text/forms/gaps/tail/events/entities while retaining parsed words, mw, facts and sentence lengths. join_trees shifts nonroot heads by preceding token count. anchored applies valid anchor links/roles sequentially unless bounds/self/cycle guard rejects. parse_grouped partitions selected keys and returns indexed thread results in group order; par_map stores each f(items[i]) at index i.

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.

Representable shifted heads, valid initial head indices for anchored walk; no overflow in atomic work indices/counters. Callback must finish without panic for returned Vec guarantee. max(threads,1) means0 still executes. Schedule-independent outputs require a pure, stateless deterministic callback or a separately established state/effect independence condition; Fn+Sync does not ensure purity. Scheduling and callback effect order are not promised.

approximant

Forward relation

PVerse::space derives next-gap nonemptiness and false last bit; meta_words chooses mw iff nonempty; compact drops text/forms/gaps/tail/events/entities while retaining parsed words, mw, facts and sentence lengths. join_trees shifts nonroot heads by preceding token count. anchored applies valid anchor links/roles sequentially unless bounds/self/cycle guard rejects. parse_grouped partitions selected keys and returns indexed thread results in group order; par_map stores each f(items[i]) at index i.

Scope: source-derived conditional transition model at the manifest SHA. Representable shifted heads, valid initial head indices for anchored walk; no overflow in atomic work indices/counters. Callback must finish without panic for returned Vec guarantee. max(threads,1) means0 still executes. Schedule-independent outputs require a pure, stateless deterministic callback or a separately established state/effect independence condition; Fn+Sync does not ensure purity. Scheduling and callback effect order are not promised.

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

par_map cannot reorder successfully returned values relative to input index; no callback result is returned for empty input. compact cannot retain discarded source text. Anchored rejects self-link and proposed cycle through x on valid starting tree.

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

For a fixed pure, stateless deterministic callback f and output y, the input preimage is the product of f^-1(y_i); worker schedule remains ambiguous. Fn+Sync alone does not establish this premise: a shared AtomicUsize callback can return [0,1] on two identical inputs with one worker. Stateful callbacks instead require a joint relation retaining initial shared state and an admissible execution schedule. The squared-input synthetic check covers only the pure case. compact preimage fixes retained trees/facts but admits arbitrary discarded text/gaps/events. join_trees inverse needs sentence lengths, otherwise root positions do not uniquely determine segmentation.

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.

parallel_order

Atomic fetch chooses unique index, mutex writes that slot, final traversal collects0..n. Callback effects may occur in any order; panic propagates through scoped threads.

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.

compaction_loss

Retained sent_lens can sum to nonzero while forms becomes empty; compact output is intended for training representation, not text decoding.

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.

head_transport

Within each sentence head0 remains0 and head h>0 maps to h+offset; inverse subtract offset only when boundary lengths are known.

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.

anchor_priority

Predicate link applied before roles for each anchor; Agent→Nsubj, Patient→Obj, othersObl. Later successful edits can overwrite earlier ones; malformed initial heads can panic despite guard on proposed coordinates.

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

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

Representable shifted heads, valid initial head indices for anchored walk; no overflow in atomic work indices/counters. Callback must finish without panic for returned Vec guarantee. max(threads,1) means0 still executes. Schedule-independent outputs require a pure, stateless deterministic callback or a separately established state/effect independence condition; Fn+Sync does not ensure purity. Scheduling and callback effect order are not promised.

Cost

Let S be sentence vectors, W total words, B cloned string bytes, and A anchor/link/role references visited. par_map uses O(n) result slots, O(max(t,1)) thread startup and the sum of callback work; mutex assignment is serialized. join_trees costs O(S+W+B), including iteration over empty sentence vectors. anchored costs O(W+B+A*(W+1)): initial word/string cloning occurs even when A=0, and guarded head walks contribute the reference-dependent term. parse_book retains frames/world per discourse; actual annotation costs are delegated and unexecuted.

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/corpus.rs.groupant/