Elamur Groupantsmova-topic-corpus-hardcases
The groupant beside tools/topic-corpus/src/hardcases.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 | Forward approximant |
| ast | partial | syntax.json |
| composition | partial | Composition |
| cost | partial | Cost |
| descriptant | present | Descriptant |
| invariant | partial | Invariant |
| tests | finite evidence | tests.md |
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 |
|---|---|---|---|
behavior_complement | behavior_complement | partial source derived model | behavior_complement |
reverse_reconstruction | reverse_reconstruction | partial source derived model | reverse_reconstruction |
Model text
Descriptant
The topic corpora's side of the one shared hard-cases list (mova-hard-case-list, format mova-hard-case-v2, the same file as the Wikipedia corpus's cases): judges a parse (no words, a tree without exactly one root, a head outside the sentence, no frame for six words or more), records found sentences with the document's licence and a pointer to the document, re-judges only its own open cases, and writes a parsed document as a mova-hard-bundle-v1 bundle for the Fish priority lane. Source: tools/topic-corpus/src/hardcases.rs.
Forward approximant
issues(words, roots per tree, frames) -> reasons. record(found, licence, reader): build v2 Found records (source topic-corpus, id <doc>:<s>, document {doc, s}) -> hard_case_list::record for parser en. recheck: hard_case_list::recheck(source = topic-corpus, parser en) with the sentence check. bundle(corpus, doc, sentences) -> JSON with each sentence's CoNLL-U (forms, lemmas, UPOS, heads, relations; XPOS/FEATS unknown) and empty pragmatics rows.
Invariant
A key is open from its first found until a learned event; found again after learned reopens it; appends are flocked and fsynced.
Cost
O(list) per record/recheck (the whole file is read); bundle O(sentences).
Composition
Parent: topic-corpus.
behavior_complement
Does not delete cases, does not decide what the parser learns, never learns a case another source found (source-scoped recheck, tested). Evidence status: source-derived model with finite checks, not a proof.
reverse_reconstruction
The list keeps text, place and reasons of each case; the parse that failed is not kept. No unique inverse is claimed.
tests.md
hardcases::tests::found_once_open_until_learned (v2 records with licence and document pointer; red control: another source's case is not rechecked; bundle CoNLL-U shape). Live 2026-11-04: 309 cases (all no_skeleton) from the hot neighbourhood of bryant-2017.
cargo test --manifest-path tools/topic-corpus/Cargo.toml --offline with WORLD_EN_MODEL=$MOVA_DATA/download/0/0.4/en.bin (18 unit tests; fixture library, fake papers binary, fake terminal, local Sea Spider store; two identical runs on 2026-11-04). Without the model the model-dependent tests report a skip.
Negative controls 2026-11-04 (each red, then restored): public admission without the origin condition; assert_public without the Sea Spider/draft/note refusals; seeding without the learner lock; shadow-library filter off; public citing that ignores refusals; cache eviction ignoring the hot pin.
These are finite fixtures and live samples; not a proof.
2026-11-04 (branch hard-loop): 17 of 18 pass; seed::tests::seeds_terms_from_neighbour_papers_and_respects_the_learner_lock fails on main as well (unrelated to this change).
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/tools/topic-corpus/src/hardcases.rs.groupant/