tonus 0.9.1 → 0.10.0

package/CHANGELOG.md

@@ -2,6 +2,68 @@
 
 All notable changes to tonus. Newest first.
 
+## 0.10.0 — 2026-08-31
+
+The corpus stores a melody once instead of once per book, the genus labels say
+what GregoBase says, and the root index stops carrying the anatomy of its own
+return values.
+
+### Changed
+
+- **BREAKING: the Kyrie's ordinary code is `ke`, not `ky`.** GregoBase files
+  every ordinary chant under office-part `ky` = Kyriale, which is the roll-up
+  genus, so the Kyrie itself needed a code of its own. `cantus({ ordinary })`
+  and `ordinarium({ ordinary })` both take `ke` now, and `ORDINARIA` is keyed
+  by it. **A stale `"ky"` throws** rather than returning `[]` — an unknown
+  ordinary code is a malformed query, and this is the one most likely to
+  arrive stale, so the error names the replacement.
+- **BREAKING: three office genus labels were wrong and are corrected.** They
+  were read off the two-letter codes rather than off the source; GregoBase's
+  own vocabulary (`include/txt.php`, `$txt['usage']`) is now followed. `or` is
+  **Toni Communes** (it read "Ordinarium"), `tp` is **Tropa** (it read "Tonus
+  Peregrinus"), and `pa` is **Prosa**. `or` was the costly one: tonus stamped
+  an invented genus on a code the database already used for something else,
+  and 132 real Toni Communes chants were dropped at the extractor for want of
+  a mapping. Anything reading `chant.genus`, or filtering `office: "or"`
+  expecting the Mass ordinary, is affected — the ordinary is `ky` (genus) or
+  `ordinary` (part).
+- **Two office codes are new: `ky` and `va`.** `ky` is the Mass ordinary's
+  roll-up genus; `va` ("Varia") is the catch-all a chant falls to when its
+  office-part is unrecognised. It used to fall to `or`, which was harmless
+  while `or` was an invented label and is not now that it means a recitation
+  formula the census excludes.
+- **BREAKING: type-only names moved off the root entry.** Four subpath entries
+  now hold the grain the index was carrying — `tonus/corpus` (the shelf and its
+  vocabulary), `tonus/score` (what a `Score` is made of), `tonus/inscriptio`
+  (the drawing surface), `tonus/census` (one chant against the corpus). Verbs
+  stay on the root namespace, and the types a root signature names stay on the
+  root too. **Every value export the root carried is still exported from it**,
+  so nothing moved out of a JavaScript consumer's reach — though the code
+  renames above do reach one.
+- **A melody printed in several books is stored once.** It was stored once per
+  book. `books` now lists every printing and `pages` is keyed by book code, so
+  `cantus({ source })` presents a shared chant as the book asked for: that
+  book's source record and page numbers, not those of the file the record
+  happens to live in. 7840 chants, 9758 listings.
+
+### Added
+
+- **Four subpath entries surface types that were reachable but unnameable.**
+  `ChantSource`, `OfficeCode`, `OrdinaryCode`, `CanonicalHour`, `Theme`,
+  `ThemeColors`, `TrackName`, `TrackData`, and `CENSUS_BLOCK_FLOATS` — each a
+  value callers had to spell by hand against a query or an option bag.
+- **The two boundaries are gated instead of described.** `npm run gate` holds
+  the rendering boundary (only `inscriptio.ts` and the emitters may import from
+  `emitters/`) and the entry map (every `exports` subpath has a source module,
+  and every entry module is exported). Prose had already failed to hold the
+  rendering boundary once. It runs as part of `npm test`.
+
+### Removed
+
+- **`ams`, `cot`, and `cse` are no longer separate data files.** Their contents
+  were listings on records the other books already own, and are now reachable
+  through `books`/`pages` on those records.
+
 ## 0.9.1 — 2026-08-24
 
 Moderna reads the pitch it was given: the whole page had been sitting a sixth

package/README.md

@@ -72,7 +72,7 @@ its final, dominant, ambitus, and ethos, with its psalm tones.
 **The calendar** is Tridentine, resolved against Easter by
 the Julian or the Gregorian computus.
 
-**The repertoire** is 2,187 chants across ten books, nine Solesmes editions and
+**The repertoire** is 7,840 chants across seven books, six Solesmes editions and
 the _Nocturnale Romanum_ night office in
 [GABC](https://gregorio-project.github.io/) notation: Mass propers, the Kyriale,
 the Office hours, the psalter, and the Matins nocturns.

package/dist/census.d.ts

@@ -0,0 +1,4 @@
+export { getCensus as census } from "./engines/census/census.js";
+export type { Census, CensusQuery, CensusBy, CensusGroup, CensusGroupProfile, CensusNeighbor, } from "./engines/census/types.js";
+export { CENSUS_GROUPS, CENSUS_ORDER, CENSUS_BLOCK_FLOATS, } from "./data/census.js";
+//# sourceMappingURL=census.d.ts.map

package/dist/census.js

@@ -0,0 +1,23 @@
+// ---------------------------------------------------------------------------
+// tonus/census — one chant measured against the corpus that holds it
+// ---------------------------------------------------------------------------
+// The analysis boundary, made addressable. The library never RUNS a census; it
+// CITES one. Everything here reads a baked table — the blocks the corpus
+// pipeline emits — and does arithmetic over it. No mining lives behind this
+// entry and none may be added: a threshold calibrated across many chants is the
+// corpus repo's business, and re-enters as generated data with provenance.
+//
+// Year-shaped questions ("this season," "this manuscript," a window of the
+// calendar) are not here and do not belong here. `data/census.ts` states the
+// same rule from the data side: this is the only census artifact in the
+// package, and the corpus repo's other outputs stay there. Pooling blocks is
+// the caller's to do, under the rule census.md fixes — cosine per field group,
+// never over the flat 221.
+//
+// CENSUS_BLOCK_FLOATS joins the two constants the root already exported. A
+// caller decoding a block had to know the stride and could get it only by
+// counting CENSUS_GROUPS, which is the transcription the appendix rule exists
+// to prevent.
+export { getCensus as census } from "./engines/census/census.js";
+export { CENSUS_GROUPS, CENSUS_ORDER, CENSUS_BLOCK_FLOATS, } from "./data/census.js";
+//# sourceMappingURL=census.js.map

package/dist/corpus.d.ts

@@ -0,0 +1,4 @@
+export { getChants as cantus, getCorpus as corpus, SOURCES } from "./engines/chant/chant.js";
+export { HORAE, OFFICIA, ORDINARIA, MODI } from "./engines/chant/types.js";
+export type { Chant, OrdinaryChant, CantusQuery, CorpusQuery, Corpus, CorpusLedger, CorpusFullCount, GenusCount, ModeCount, SharedCount, ChantSource, OfficeCode, OrdinaryCode, CanonicalHour, } from "./engines/chant/types.js";
+//# sourceMappingURL=corpus.d.ts.map

package/dist/corpus.js

@@ -0,0 +1,17 @@
+// ---------------------------------------------------------------------------
+// tonus/corpus — the shelf: what tonus holds, and how to ask for it
+// ---------------------------------------------------------------------------
+// The repertoire and its vocabulary, separated from the calendar that resolves
+// a DAY. `cantus` and `corpus` ask what is on the shelf; `festum`, `proprium`
+// and `officium` ask what a day calls for. The two questions have been one
+// import since the beginning, which is part of why the shelf reads as though it
+// were a property of the calendar.
+//
+// This entry surfaces four types the root index never exported, each of them a
+// value a caller otherwise had to TYPE OUT against a query: ChantSource (the
+// book codes `cantus({ source })` accepts), OfficeCode and OrdinaryCode (the
+// genus and ordinary codes), and CanonicalHour (what HORAE holds). The appendix
+// rule already admits the tables; the types they range over belong beside them.
+export { getChants as cantus, getCorpus as corpus, SOURCES } from "./engines/chant/chant.js";
+export { HORAE, OFFICIA, ORDINARIA, MODI } from "./engines/chant/types.js";
+//# sourceMappingURL=corpus.js.map