tonus 0.10.0 → 0.10.1

package/CHANGELOG.md

@@ -2,6 +2,31 @@
 
 All notable changes to tonus. Newest first.
 
+## 0.10.1 — 2026-08-31
+
+A capital ligature is a vowel like any other.
+
+### Fixed
+
+- **`Æ` and `Œ` reached the nucleus unexpanded.** `selectVowel` expanded the
+  ligatures case-sensitively, and neither has an NFD decomposition (unlike
+  `á` → `a`), so a capital survived the scan and was reported AS the vowel:
+  `vowel: "æ"`, a two-letter answer to a one-letter promise. Consumers that
+  hold tonus to that contract throw on it. 136 chants carry one, every case
+  opening a word — 94 at the start of the lyric, the other 42 inside an
+  all-capital opening word (`SÆpe`, `PRÆ`, `HÆC`), which is exactly where a
+  capital lives. The gated shelf held almost none of them, so 0.10.0's re-cut
+  is what surfaced it; the defect is older.
+- **An accented ligature accented the off-glide.** `œ́` is `œ` + combining
+  acute, with no precomposed form, so expanding it to `oe` + accent marked the
+  `e`: `selectVowel("œ́")` returned nucleus `"e"` where it should return `"o"`.
+  The pair is now rebuilt as `o` + accent + `e`. `ǽ`, which IS precomposed,
+  was always correct — the two agree now.
+
+Note for consumers baking off `vowel`: 136 chants change nucleus, and `Æ`/`Œ`
+now report `diphthong: "ae"`/`"oe"` where they reported `null`. Anything
+derived from the vowel plane wants a re-bake.
+
 ## 0.10.0 — 2026-08-31
 
 The corpus stores a melody once instead of once per book, the genus labels say

package/dist/engines/chant/syllabify.js

@@ -18,7 +18,8 @@
 //      cl, gl, fl) and the Greek digraphs th, ph, ch stay with the following
 //      vowel; all other clusters split after the first consonant
 //
-// Diacritics (áéíóúàèìòùâêîôû etc.) are treated as their base vowel throughout.
+// Diacritics (áéíóúàèìòùâêîôû etc.) are treated as their base vowel throughout,
+// and the ligatures æ/œ expand to their pairs in either case (see selectVowel).
 // ── Character classes ──
 const VOWELS = new Set("aeiouyáéíóúàèìòùâêîôûäëïöüæœý");
 const SOFT_HYPHEN = "\u00ad";
@@ -203,7 +204,25 @@ export function hyphenateWord(word) {
 }
 const ACCENTED = /[\u0301]/; // combining acute accent (NFD form)
 export function selectVowel(text) {
-    const expanded = text.replace(/ǽ/g, "áe").replace(/æ/g, "ae").replace(/œ/g, "oe");
+    // THE LIGATURES EXPAND BEFORE THE SCAN, IN EITHER CASE. æ and œ have no
+    // NFD decomposition (unlike á → a), so a ligature that survives to the loop
+    // below is reported AS the nucleus — a two-letter vowel, which breaks the
+    // one-letter contract `vowel` promises and throws downstream. This was
+    // case-sensitive until 2026-08-31 and so missed every capital: 135 chants,
+    // all of them word-initial (`Ægýptum`, the Nocturnale's Æ responsories),
+    // invisible while the shelf was gated because it carried almost none.
+    //
+    // The accent rides the NUCLEUS, not the glide. ǽ is precomposed, but œ́ is
+    // not — it is œ + combining acute — so the pair is rebuilt as o + accent + e
+    // rather than oe + accent, which would accent the off-glide and invert what
+    // `accent` reports. NFC first, so a decomposed source still matches here.
+    const expanded = text
+        .normalize("NFC")
+        .replace(/ǽ/gi, "áe")
+        .replace(/([æœ])(́?)/gi, (_m, lig, acc) => {
+        const pair = lig.toLowerCase() === "æ" ? "ae" : "oe";
+        return acc ? pair[0] + acc + pair[1] : pair;
+    });
     const nfd = expanded.normalize("NFD");
     let firstVowel = "";
     let accentedVowel = "";

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "tonus",
-  "version": "0.10.0",
+  "version": "0.10.1",
   "type": "module",
   "description": "Medieval music analysis: GABC plainchant exports, liturgical calendar, tuning systems, ephemeris, and the harmony of the spheres",
   "author": "Jeffrey Pierce <jeffrey@jeffreypierce.net>",