Expand description
nibli-lexicon — the COMMITTED English corpus (the dictionary IS Rust source;
see corpus and the Dictionary Data section of CLAUDE.md).
Since the committed-corpus milestone there is exactly ONE build mode: the
strongly-typed tables in src/corpus/ are committed, const-validated at
compile time, and identical in local, CI, and deployed builds. The old
build-time parse of dictionary-en.json (and its FULL/FALLBACK dual mode)
is gone — the JSON is now only the INPUT of the tools/lexigen regeneration
tool (just regen-lexicon).
The public surface is functions-only (no pub statics), so the internals can evolve without consumer churn:
alias/compound— KR-input resolution (English names /a+bcompound spellings; gismu do NOT resolve — see the compat note below).get_arity/get_gloss/get_template— IR-relation lookups.canonical_alias/by_provenance— the gismu→English provenance bridge (Predilex gates, render overlays, lexigen drift reports).corpus::corpus_entries/corpus::corpus_compounds— iteration.
Gismu do not resolve anywhere: the ONLY spellings are the English corpus
names (and a+b compound spellings). The source gismu survives solely as
the by_provenance bridge — provenance metadata, never input.
Re-exports§
pub use corpus::CompoundEntry;pub use corpus::CorpusTier;pub use corpus::PredicateEntry;pub use corpus::Swap;
Modules§
- corpus
- The committed English corpus — the successor to the build-time dictionary-en.json parse (the committed-corpus milestone).
- reserved
- The nibli KR reserved-word list — SINGLE SOURCE (NIBLI_KR.md §2).
Functions§
- alias
- Look up an English predicate name (the only resolvable atomic spelling).
Nonemeans the name is unknown — nibli-kr’s resolver FAILS CLOSED (no arity-2 default; NIBLI_KR §13). - by_
provenance - The provenance bridge: gismu → its CANONICAL (unswapped) corpus entry. Permanent API — the Predilex gates key their Lojban-lemma bounds through it, and future ontology-row import uses the same mapping.
- canonical_
alias - The canonical (plain, unswapped) English name for a gismu — the render direction. Converted entries are never canonical.
- compound
- Look up a compound predicate by its
a+bKR spelling. - compound_
by_ relation - Look up a compound by its IR RELATION ident (
computer_user) — the post-emit direction (render/arity/place-name lookups on compound relations). Linear over the curated compound table, which stays small. - get_
arity - Arity of an IR relation (English corpus name or compound relation ident).
- get_
gloss - Primary English gloss of an IR relation.
- get_
template - Curated English place-frame template (
{x1}..{x5}placeholders), e.g.get_template("dog")->Some("{x1} is a dog").None= no curated frame; callers build a generic frame fromget_gloss+get_arity. - relation_
places - English place names of an IR relation: an atomic corpus name or a
compound’s relation ident.
None= unknown relation (callers fall back to positionalx<N>).