Elamur Groupantsprag-feats-dictionary
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-dictionary-behavior-complement | behavior_complement | partial source model | behavior-complement |
prag-feats-dictionary-reverse-reconstruction | reverse_reconstruction | partial source model | reverse-reconstruction |
prag-feats-dictionary-coupling | constraint_relation | partial source model | coupling |
prag-feats-dictionary-observation-fiber | information_loss | partial source model | observation-fiber |
Model text
Training-order per-slot dictionaries with complete previous-act vocabulary and explicit unseen sentinel.
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
build seeds prev with Act::ALL names then start, inserts each row value once in traversal order. id looks up slot map or UNSEEN; encode projects every supplied coordinate. Serialization skips index; reindex recreates per-slot maps from values. show uses '?' for UNSEEN and SLOTS names.
behavior-complement
Universe: valid-width rows and built/reindexed dictionary plus candidate IDs. Excludes ID reuse for distinct values within a slot, unseen training values after build, missing Act IDs in prev, and encoding an absent value as an ordinary ID. No guarantee for deliberately malformed serialized duplicate values.
reverse-reconstruction
Given values retain every training row history with identical per-slot first-appearance order and arbitrary repeats. Given UNSEEN retain every absent string, not one unknown spelling. Numeric IDs require dictionary and slot to interpret; they alone do not reconstruct rows.
invariant
Slot index must be in dictionary range; non-UNSEEN show ID must be in values range. Rows longer than SLOTS panic; shorter rows encode shorter vectors. Reindex is required after deserialize; no automatic repair.
cost
Expected total input string hashing/copy time and distinct-value storage. reindex copies every retained value; encode expected O(row string bytes); show allocates explanation string. No measured speedup or executable approximation error bound is claimed.
coupling
Dict receives training rows after prev injection; it pre-seeds all acts so predicted but unseen training acts still encode. Persisted model users must reindex before id/encode. Model and eval callers maintain that lifecycle.
observation-fiber
Training counts and repeated-row order are forgotten except first appearances. Serialized values retain decoding data but discard the hash index. UNSEEN collapses all unobserved values and show prints one '?'.
composition
Dict receives training rows after prev injection; it pre-seeds all acts so predicted but unseen training acts still encode. Persisted model users must reindex before id/encode. Model and eval callers maintain that lifecycle. No child fragments; this is one scoped relation.
checks
empty_row_literal_and_dictionary_roundtrip covers every empty feature coordinate, Act coverage, missing value, show, serde omission, reindex, equal encoding and serialized bytes. 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/dictionary.groupant/