Elamur Groupantsprag-feats-lexical
The groupant beside prag/src/feats.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 |
|---|---|---|---|
prag-feats-lexical-behavior-complement | behavior_complement | partial source model | behavior-complement |
prag-feats-lexical-reverse-reconstruction | reverse_reconstruction | partial source model | reverse-reconstruction |
prag-feats-lexical-coupling | constraint_relation | partial source model | coupling |
prag-feats-lexical-observation-fiber | information_loss | partial source model | observation-fiber |
Model text
Normalize words and greedily consume cues in stable longest-phrase order.
Source snapshot: 2a1a0bd6e43dd16f7284e993f30d0f4971c447c9e2080616033d19ac8693611f. All relations concern this snapshot on the stated admissible domain. Models are source-derived partial relations; finite fixtures are evidence, not general proofs. Complement is a conservative inner constraint model. Reverse reconstruction preserves all compatible antecedents conceptually; no enumerating solver is implemented.
forward
N lowercases Unicode, replaces curly apostrophe and maps n't,'ll,'m,'re,'ve,'d,ca,wo,sha to not,will,am,are,have,would,can,will,shall. load parses embedded TSV, normalizes phrase tokens and stable-sorts descending length. find scans left-to-right and takes the first matching phrase, advancing by its length; otherwise advances one word.
behavior-complement
Universe: nonempty phrase tables, token vectors, candidate ordered (index,class) lists. Excluded outputs include overlapping consumed hits, decreasing hit positions, hits beyond input bounds and a short match selected before an earlier-listed matching long phrase. This inner bound assumes no empty phrase, as guaranteed by the embedded source.
reverse-reconstruction
For normalized word y retain all Unicode/case/apostrophe variants and contraction antecedents mapped to y. For hit lists retain all token sequences agreeing at consumed phrases plus arbitrary unmatched tokens, subject to no higher-priority match at scanned positions. will admits WO, 'll and will.
invariant
A successful match strictly advances for nonempty phrases. find does not normalize its input; callers supply normalized tokens. Tie priority is embedded file order, not alphabetical class order.
cost
Normalization O(input bytes) allocations; load O(P log P) phrase sort after parsing; find worst O(WPL), output at most W hits. No measured speedup or executable approximation error bound is claimed.
coupling
frame_words normalizes narrator cues; utterance cues use annotation word forms and rewrite let 's to let us. Input tokenization can alter whether a multiword cue is visible.
observation-fiber
Changing consumed phrase interior may remove a hit; changing an unconsumed token can leave the same output. Greedy consumption suppresses alternate overlapping interpretations, so inverse candidates must honor priority.
composition
frame_words normalizes narrator cues; utterance cues use annotation word forms and rewrite let 's to let us. Input tokenization can alter whether a multiword cue is visible. No child fragments; this is one scoped relation.
checks
normalization_and_greedy_cue_loss: ten literal normalizations, idempotence at those outputs, five-token greedy overlap example, embedded thank-you cue and ambiguous will antecedents. See the file-level evidence.md for exact log hashes and repeated run outcomes.
epoch-memory
2026-10-30 wave13: initial source-bound model. Prospective model preceded edits. Existing normalization and quote rules were retained; tests expose their information loss. Source identity witness justified qualifying predecessor keys by Src. Annotation is isolated behind a private closure used by the production row wrapper; fixtures do not call or certify tokenizer/parser. AST is a partial syntax projection, macro bodies opaque; semantic AST, live model/corpus and general proof remain absent.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/prag/src/feats.rs.groupant/lexical.groupant/