Elamur Groupantsworld-idioms-mining-export
The groupant beside world/src/idioms.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-idioms-mining-export-behavior-complement | behavior_complement | partial source model | behavior-complement |
world-idioms-mining-export-reverse-reconstruction | reverse_reconstruction | partial source model | reverse-reconstruction |
world-idioms-mining-export-dependency-feedback | constraint_relation | partial source model | dependency-feedback |
world-idioms-mining-export-export-collision | constraint_relation | partial source model | export-collision |
Model text
forward
Merge dictionary evidence, accumulate OBJ marginals, retain last duplicate pair count/books, gate count/books with dictionary bypass but count>=3 always, classify by matrix and PMI. Aggregate two-word events and export sorted rows then final map.
invariant
OBJ rows require count>=3 even with dictionary evidence. Particle rows require min only. Final map is pair-unique, detailed output need not be. Successful writes do not imply atomic publication.
cost
For B bytes, U pairs, E event rows and R retained rows: O(B+(U+E)log(U+E)+R log R), plus matrix lookup and output bytes. Sorting is separate from accumulation; no timing claim.
composition
This fragment shares source and state with its siblings; root composition specifies ordering and latent dependencies. It is not an independent replacement API.
epoch-memory
Predicted first minimum dictionary rank and first MWE row; finite tests confirm. Limitations: substring flags, unconditional first-line discard, duplicate last pair versus summed marginals, absent phrasal book gate and unused k%1 set. Production unchanged.
checks
Five native tests: 64 rank triples, eight phrase/POS contexts plus header, 80 MWE flag/type/title combinations, reversed order, four pair-key preimages. Full run/exports not executed. Fragment scopes are defined by source byte ranges; root evidence distinguishes executed helper tests from inspection-only branches.
behavior-complement
Within the scoped input/output universe, violating dependency-feedback or export-collision is excluded by the source model. This inner constraint family does not claim a complete complement.
reverse-reconstruction
Converse retains all inputs compatible with the scoped output and fixed dependencies; projections and discarded candidates generally create multiple antecedents. No single inverse is claimed.
dependency-feedback
Matrix score consults existing global idiom table, clamps known literal OBJ scores, and can consult known layer. Freeze prior snapshot.
export-collision
Particle rows use min only, no min_books; insertion overwrites same OBJ key in global map although detailed output can contain both. Two-file output is not atomic.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/world/src/idioms.rs.groupant/mining-export.groupant/