Elamur Groupantswordgraph-exceptions
The groupant beside wordgraph/src/exceptions.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 | Approximant |
| Ast | partial | syntax.json |
| Composition | partial | Composition |
| Cost | partial | Cost |
| Descriptant | present | Descriptant |
| Invariant | partial | Invariant |
| Tests | executed bounded checks | Tests |
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.
| relation | 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 hand lists: the list of the wrong (seeds/wrong.tsv) and the white list (seeds/right.tsv), compiled into the
binary; parsing with checks, the fingerprint recorded in parts and graphs, the compound-reading lookup and phrase
matching used by the extractor. Hints, never bans. Epoch 2026-11-06: first version, 26 wrong and 5 white entries.
Approximant
parse(text, white) = the rows with 6 tab-separated columns, a known relation and a kind of the right list, refusing duplicates of (head, relation, dependent). compound(h, r, d) = the compound-reading entry with those fields. phrase_ends_at(p, f, e) ⇔ e + 1 ≥ |p| ∧ f[e+1−|p| .. e] = p[.. |p|−1] (the token at e is the edge's end, matched by lemma, so it stands for the last word in any inflection; 2026-11-06c).
Invariant
No duplicate (head, relation, dependent) in a parsed list; a white list holds only correct, a wrong list none.
fingerprint() changes with any byte of either list. phrase_ends_at never indexes before 0.
Cost
O(list bytes) parse once per process (OnceLock); compound lookup O(entries) (26) with integer relation compare first.
Composition
Used by extract (suppression), graph (judged values from the lists embedded in the file), compile (embeds them),
pass (fingerprint) and the CLI (lists).
Tests
seed_lists_parse, bad_lists_are_refused (short row, unknown relation, white kind in the wrong list, duplicate:
red side), phrase_matching (no underflow).
Behavior complement
Does not ban an edge, match lemmas by UPOS, match phrases with gaps or other word order, learn entries by itself, or apply a changed list to parts already counted (they record the fingerprint they were made with).
Reverse reconstruction
From a part's suppression counts the listed phrases that fired are recoverable, not the sentences; from the fingerprint, whether two parts used the same lists, not the lists themselves.
Provenance
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/wordgraph/src/exceptions.rs.groupant/