Elamur Groupantsworld-mwe-cupt
The groupant beside world/src/mwe.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-mwe-cupt-behavior-complement | behavior_complement | partial source model | behavior-complement |
world-mwe-cupt-reverse-reconstruction | reverse_reconstruction | partial source model | reverse-reconstruction |
world-mwe-cupt-head-coordinate | identity_relation | partial source model | head-coordinate |
world-mwe-cupt-gold-order | priority_relation | partial source model | gold-order |
Model text
CUpt accepted rows, annotation grouping and token-ID projection.
forward
Retain rows with >=11 tab columns and positive ASCII-decimal usize IDs; range, decimal, zero, empty and overflowing IDs are ignored. Preserve retained row order. Map each unique numeric ID to row index+1; duplicate numeric IDs map to zero. Parse heads through that map, absent/unparsable heads to zero. Lowercase lemmas, preserve forms/POS/relations. Annotation groups use string-key BTreeMap order; later nonempty category wins.
invariant
Output vectors have identical lengths. Each nonzero head names a retained row with an unambiguous source ID. Gold indices name retained positions; duplicate annotation labels may duplicate indices, and self-heads can remain in Sent but predict ignores them.
cost
O(B+N log N+A log G) plus string copying, where A counts annotation parts. O(B+N+A) storage includes read_to_string and grouped output.
composition
Filtering must precede head remapping. Otherwise a sparse raw head can index an unrelated word or panic. Annotation indexing shares the retained order, not raw numeric IDs. Missing head maps to root-like zero, deliberately losing distinction.
epoch-memory
2026-10-30 first lexical epoch: old native witnesses returned raw heads [0,10,99], accepted noun-only overlap and duplicate token IDs, and panicked on an out-of-range head. Narrow fixes translate unambiguous IDs, omit invalid/missing arcs, and enforce the documented distinct/shared-verb condition. Prediction-order greedy matching remains intentionally unchanged; historical tuning scores can differ. No corpus retuning or measured speed claim.
checks
Native filter mwe::wave10_models; see ../evidence.md for fragments or evidence.md at the root. Finite supplied fixtures and independent oracles; no live parser, corpus, teacher or global dictionary activation. Partial syntax has opaque macro bodies and no semantic AST. No executable percentage surrogate or speed claim.
behavior-complement
Universe: CUpt bytes and parsed outputs after successful I/O. Inner absent relation excludes nonzero attachments to missing or duplicated IDs and retained range/empty-node rows. This is tolerant projection, not a complete CUpt schema validator; comments must begin with # and annotation syntax is permissive.
Subject: current source hash and this node's declared scope. Relation describes one terminating call over the finite inputs stated above, with fixed dependencies and successful allocation. Status: partial source-derived model; named native fixtures support only their finite domain. Unobserved effects remain unknown.
reverse-reconstruction
Given Sent retain all raw files producing its retained rows, including comments, ignored range/decimal rows, capitalization erased in lemmas and renumberings preserving head identity. Zero heads admit root, missing, duplicate and unparsable antecedents; they cannot identify which cause occurred.
Subject: current source hash and this node's declared scope. Relation describes one terminating call over the finite inputs stated above, with fixed dependencies and successful allocation. Status: partial source-derived model; named native fixtures support only their finite domain. Unobserved effects remain unknown.
head-coordinate
For retained row j and unique source ID d, map(d)=j+1. For every row's raw head d this mapping replaces the old raw ID. The native sparse fixture [10,20,30] with heads [0,10,99] must become [0,1,0]; ignored range/decimal rows cannot shift the mapping.
Subject: current source hash and this node's declared scope. Relation describes one terminating call over the finite inputs stated above, with fixed dependencies and successful allocation. Status: partial source-derived model; named native fixtures support only their finite domain. Unobserved effects remain unknown.
gold-order
Group IDs are strings, so BTreeMap sorts lexically, not numerically (10 before 2). This gold order feeds greedy judge. Categories are metadata only for matching; repeated category assignments retain the last nonempty value. No canonical semantic grouping or category agreement is imposed.
Subject: current source hash and this node's declared scope. Relation describes one terminating call over the finite inputs stated above, with fixed dependencies and successful allocation. Status: partial source-derived model; named native fixtures support only their finite domain. Unobserved effects remain unknown.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/world/src/mwe.rs.groupant/cupt.groupant/