Elamur Groupantsprag-feats-features-projection
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-features-projection-behavior-complement | behavior_complement | partial source model | behavior-complement |
prag-feats-features-projection-reverse-reconstruction | reverse_reconstruction | partial source model | reverse-reconstruction |
prag-feats-features-projection-coupling | constraint_relation | partial source model | coupling |
prag-feats-features-projection-observation-fiber | information_loss | partial source model | observation-fiber |
Model text
Project supplied words, terminal punctuation, UD relations, cues and metadata into 32 coordinates.
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
Content omits PUNCT and wholly nonalphanumeric forms. Terminal punctuation priority is ? then ! then ellipsis then period then dash then comma-like. Root is first head=0; subjects and tense auxiliaries are direct children. Inversion requires AUX before selected subject. Cue evaluation distinguishes pos/neg/mix; first prioritized act-like class becomes cue. ppos is only/first/last/mid; genre maps source.
behavior-complement
Universe: finite Word vectors and valid text/frame/meta to 32-string rows. Excludes len outside six bucket labels, qm=no when a question mark occurs (value is yes or -), unknown source genre, neg=root when only a non-root/non-child negative exists, and terminal ! when terminal tail contains ?.
reverse-reconstruction
A row retains all Word vectors with equal observed first/last/content bucket and selected root/child/cue facts. Length bucket 4-6 admits three content counts; punctuation tokens can vary without changing content slots. Unknown sperson '-' permits absent morphology even for you.
invariant
First root and first qualifying child win; rootless input uses missing root slots. Head values are compared as indices, not validated as a tree. No semantic truth or correct parsing is inferred.
cost
Constant-many linear word scans plus cue matching O(WPL); constant 32-slot output plus copied words and frame. No measured speedup or executable approximation error bound is claimed.
coupling
Word forms determine normalization and lexical cues; UPOS/tag/feats/head/rel independently affect grammar slots. A token form alone cannot imply person morphology. ppos assumes sensible positions; extreme inconsistent public metadata remains outside tested domain.
observation-fiber
Bucket and booleans collapse counts; first-hit wh/modal/subject discard later alternatives. qm scans entire utterance while end scans only final nonalphanumeric tail, so internal ? and terminal . can coexist.
composition
Word forms determine normalization and lexical cues; UPOS/tag/feats/head/rel independently affect grammar slots. A token form alone cannot imply person morphology. ppos assumes sensible positions; extreme inconsistent public metadata remains outside tested domain. No child fragments; this is one scoped relation.
checks
bucket_boundaries_and_terminal_priority checks twelve bucket boundaries and seven terminal strings; supplied UD checks eighteen named feature coordinates and empty-row test checks every slot. 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/features.groupant/projection.groupant/