@kensio/pinyinjs 1.7.14 → 1.8.0

package/dist/decode/sandhi-grouping.d.ts

@@ -0,0 +1,55 @@
+/**
+ * How the syllables group, and the two things sandhi reads out of a grouping.
+ *
+ * Third-tone sandhi wants the junctions, since its domain is the prosodic foot
+ * and a foot is built from structure. 一 wants the word ends, since a 一 that
+ * closes a longer word counts nothing after it. Both are answered here so that
+ * `sandhi.ts` holds the rules rather than the bookkeeping.
+ */
+/**
+ * How the syllables group, for sandhi to read.
+ *
+ * One entry per word: its syllable count, or the counts of the constituents it
+ * divides into. 老板 很 好 is `[2, 1, 1]` and 纸老虎 is `[[1, 2]]`.
+ *
+ * Third-tone sandhi's domain is the prosodic foot rather than the word, and the
+ * foot is built from structure — which is why this is worth carrying. See
+ * {@link import("./sandhi.js").applySandhi}.
+ */
+export type SandhiGrouping = readonly (number | readonly number[])[];
+/**
+ * Where sandhi may apply, as the index of the syllable that would lower.
+ *
+ * Every junction is between adjacent syllables — words and their constituents
+ * cover the reading end to end — so one index says all of it, and the three
+ * lists are the three passes third-tone sandhi makes.
+ */
+export interface SandhiJunctions {
+    /** Inside a constituent. */
+    readonly inner: readonly number[];
+    /** Between the constituents of a word. */
+    readonly parts: readonly number[];
+    /** Between a single-syllable word and the word after it. */
+    readonly words: readonly number[];
+}
+/**
+ * Read a grouping as the junctions it puts in the syllable array.
+ *
+ * The whole array is treated as one undivided word where no grouping is given,
+ * or where the one given does not account for exactly the syllables there are —
+ * a grouping that does not fit is describing some other text, and guessing
+ * which syllables it meant would be worse than ignoring it.
+ */
+export declare function junctionsOf(length: number, grouping?: SandhiGrouping): SandhiJunctions;
+/**
+ * Where a word of more than one syllable ends, as indices into the array.
+ *
+ * Empty without a grouping, and empty for a grouping that does not fit, which
+ * is the judgement {@link junctionsOf} makes and for the same reason.
+ *
+ * A one-syllable word is deliberately not an ending here. 一 on its own is a
+ * word and it counts the word after it: see
+ * {@link import("./sandhi.js").applySandhi}.
+ */
+export declare function endingsOf(length: number, grouping?: SandhiGrouping): ReadonlySet<number>;
+//# sourceMappingURL=sandhi-grouping.d.ts.map

package/dist/decode/sandhi-grouping.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"sandhi-grouping.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi-grouping.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH;;;;;;;;;GASG;AACH,MAAM,MAAM,cAAc,GAAG,SAAS,CAAC,MAAM,GAAG,SAAS,MAAM,EAAE,CAAC,EAAE,CAAC;AAErE;;;;;;GAMG;AACH,MAAM,WAAW,eAAe;IAC9B,4BAA4B;IAC5B,QAAQ,CAAC,KAAK,EAAE,SAAS,MAAM,EAAE,CAAC;IAClC,0CAA0C;IAC1C,QAAQ,CAAC,KAAK,EAAE,SAAS,MAAM,EAAE,CAAC;IAClC,4DAA4D;IAC5D,QAAQ,CAAC,KAAK,EAAE,SAAS,MAAM,EAAE,CAAC;CACnC;AASD;;;;;;;GAOG;AACH,wBAAgB,WAAW,CACzB,MAAM,EAAE,MAAM,EACd,QAAQ,CAAC,EAAE,cAAc,GACxB,eAAe,CAiCjB;AAED;;;;;;;;;GASG;AACH,wBAAgB,SAAS,CACvB,MAAM,EAAE,MAAM,EACd,QAAQ,CAAC,EAAE,cAAc,GACxB,WAAW,CAAC,MAAM,CAAC,CAerB"}

package/dist/decode/sandhi-grouping.js

@@ -0,0 +1,81 @@
+/**
+ * How the syllables group, and the two things sandhi reads out of a grouping.
+ *
+ * Third-tone sandhi wants the junctions, since its domain is the prosodic foot
+ * and a foot is built from structure. 一 wants the word ends, since a 一 that
+ * closes a longer word counts nothing after it. Both are answered here so that
+ * `sandhi.ts` holds the rules rather than the bookkeeping.
+ */
+/**
+ * The syllable counts one grouping entry holds, as a flat list of its parts.
+ */
+function partsOf(word) {
+    return typeof word === "number" ? [word] : word;
+}
+/**
+ * Read a grouping as the junctions it puts in the syllable array.
+ *
+ * The whole array is treated as one undivided word where no grouping is given,
+ * or where the one given does not account for exactly the syllables there are —
+ * a grouping that does not fit is describing some other text, and guessing
+ * which syllables it meant would be worse than ignoring it.
+ */
+export function junctionsOf(length, grouping) {
+    const undivided = {
+        inner: Array.from({ length: Math.max(length - 1, 0) }, (_, at) => at),
+        parts: [],
+        words: [],
+    };
+    if (grouping === undefined) {
+        return undivided;
+    }
+    const inner = [];
+    const parts = [];
+    const words = [];
+    let at = 0;
+    for (const [index, word] of grouping.entries()) {
+        const divisions = partsOf(word);
+        const from = at;
+        for (const [division, size] of divisions.entries()) {
+            for (let step = 1; step < size; step++) {
+                inner.push(at + step - 1);
+            }
+            at += size;
+            // The junction closing a word's last constituent is the one *around* the
+            // word rather than one inside it.
+            if (division < divisions.length - 1) {
+                parts.push(at - 1);
+            }
+        }
+        if (at - from === 1 && index < grouping.length - 1) {
+            words.push(at - 1);
+        }
+    }
+    return at === length ? { inner, parts, words } : undivided;
+}
+/**
+ * Where a word of more than one syllable ends, as indices into the array.
+ *
+ * Empty without a grouping, and empty for a grouping that does not fit, which
+ * is the judgement {@link junctionsOf} makes and for the same reason.
+ *
+ * A one-syllable word is deliberately not an ending here. 一 on its own is a
+ * word and it counts the word after it: see
+ * {@link import("./sandhi.js").applySandhi}.
+ */
+export function endingsOf(length, grouping) {
+    const endings = new Set();
+    if (grouping === undefined) {
+        return endings;
+    }
+    let at = 0;
+    for (const word of grouping) {
+        const size = partsOf(word).reduce((total, part) => total + part, 0);
+        at += size;
+        if (size > 1) {
+            endings.add(at - 1);
+        }
+    }
+    return at === length ? endings : new Set();
+}
+//# sourceMappingURL=sandhi-grouping.js.map

package/dist/decode/sandhi-grouping.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"sandhi-grouping.js","sourceRoot":"","sources":["../../src/decode/sandhi-grouping.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AA8BH;;GAEG;AACH,SAAS,OAAO,CAAC,IAAgC;IAC/C,OAAO,OAAO,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC;AAClD,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,WAAW,CACzB,MAAc,EACd,QAAyB;IAEzB,MAAM,SAAS,GAAoB;QACjC,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC;QACrE,KAAK,EAAE,EAAE;QACT,KAAK,EAAE,EAAE;KACV,CAAC;IACF,IAAI,QAAQ,KAAK,SAAS,EAAE,CAAC;QAC3B,OAAO,SAAS,CAAC;IACnB,CAAC;IAED,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,IAAI,EAAE,GAAG,CAAC,CAAC;IACX,KAAK,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC,IAAI,QAAQ,CAAC,OAAO,EAAE,EAAE,CAAC;QAC/C,MAAM,SAAS,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC;QAChC,MAAM,IAAI,GAAG,EAAE,CAAC;QAChB,KAAK,MAAM,CAAC,QAAQ,EAAE,IAAI,CAAC,IAAI,SAAS,CAAC,OAAO,EAAE,EAAE,CAAC;YACnD,KAAK,IAAI,IAAI,GAAG,CAAC,EAAE,IAAI,GAAG,IAAI,EAAE,IAAI,EAAE,EAAE,CAAC;gBACvC,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,IAAI,GAAG,CAAC,CAAC,CAAC;YAC5B,CAAC;YACD,EAAE,IAAI,IAAI,CAAC;YACX,yEAAyE;YACzE,kCAAkC;YAClC,IAAI,QAAQ,GAAG,SAAS,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;gBACpC,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;YACrB,CAAC;QACH,CAAC;QACD,IAAI,EAAE,GAAG,IAAI,KAAK,CAAC,IAAI,KAAK,GAAG,QAAQ,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YACnD,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACrB,CAAC;IACH,CAAC;IACD,OAAO,EAAE,KAAK,MAAM,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC;AAC7D,CAAC;AAED;;;;;;;;;GASG;AACH,MAAM,UAAU,SAAS,CACvB,MAAc,EACd,QAAyB;IAEzB,MAAM,OAAO,GAAG,IAAI,GAAG,EAAU,CAAC;IAClC,IAAI,QAAQ,KAAK,SAAS,EAAE,CAAC;QAC3B,OAAO,OAAO,CAAC;IACjB,CAAC;IAED,IAAI,EAAE,GAAG,CAAC,CAAC;IACX,KAAK,MAAM,IAAI,IAAI,QAAQ,EAAE,CAAC;QAC5B,MAAM,IAAI,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE,CAAC,KAAK,GAAG,IAAI,EAAE,CAAC,CAAC,CAAC;QACpE,EAAE,IAAI,IAAI,CAAC;QACX,IAAI,IAAI,GAAG,CAAC,EAAE,CAAC;YACb,OAAO,CAAC,GAAG,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACtB,CAAC;IACH,CAAC;IACD,OAAO,EAAE,KAAK,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,GAAG,EAAE,CAAC;AAC7C,CAAC"}

package/dist/decode/sandhi-input.d.ts

@@ -1,29 +1,34 @@
 /**
  * What the sandhi pass needs from a decoded run, beyond the syllables.
  *
- * Two questions sandhi cannot answer from a reading alone. Third-tone sandhi
- * applies inside a prosodic foot, so it has to be told where the words and
- * their parts are. 一 sandhi asks what stands in front of the 一, and the answer
- * is a character rather than a spelling. Both are things the decode knows and
- * the syllable array does not.
+ * Two questions sandhi cannot answer from a reading alone. Where the words and
+ * their parts are, which third-tone sandhi needs for the prosodic foot it
+ * applies inside and 一 needs for the word ends. And which 汉字 each syllable
+ * reads, which is what says a `yī` is 一 rather than 医 and what stands in front
+ * of it. Both are things the decode knows and the syllable array does not.
  */
 import type { Dictionary } from "../dictionary/dictionary.js";
 import type { Syllable } from "../syllable/syllable.js";
 import type { SandhiGrouping } from "./sandhi.js";
 import type { ScoredWord } from "./word.js";
 /**
- * Where the words and their constituents fall, for third-tone sandhi.
+ * Where the words and their constituents fall.
  *
- * Undefined unless third-tone sandhi was asked for, since dividing a word costs
- * dictionary lookups and nothing else reads the answer.
+ * One entry per word always, because 一 sandhi reads the word ends: a 一 that
+ * closes a longer word is not counting the word after it, so 唯一 stays `wéiyī`
+ * in front of anything.
+ *
+ * Where a word *divides* is only worked out for third-tone sandhi, which is the
+ * only pass that reads it and costs a dictionary lookup per word to answer.
  */
-export declare function groupingOf(dictionary: Dictionary, words: readonly ScoredWord[], readings: readonly (readonly Syllable[])[], isThirdTone: boolean): SandhiGrouping | undefined;
+export declare function groupingOf(dictionary: Dictionary, words: readonly ScoredWord[], readings: readonly (readonly Syllable[])[], isThirdTone: boolean): SandhiGrouping;
 /**
  * One 汉字 per syllable, where a word offers one.
  *
  * A word whose reading is a different length from its text — 玩儿 as `wánr` —
  * has no character to give any one of its syllables, so its syllables get none
  * and the pass falls back to their spellings. See
+ * {@link import("./sandhi-tones.js").isYi} and
  * {@link import("./sandhi-tones.js").isCounting}.
  */
 export declare function charactersPerSyllable(words: readonly ScoredWord[], readings: readonly (readonly Syllable[])[]): readonly (string | undefined)[];

package/dist/decode/sandhi-input.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"sandhi-input.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi-input.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAClD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AAE5C;;;;;GAKG;AACH,wBAAgB,UAAU,CACxB,UAAU,EAAE,UAAU,EACtB,KAAK,EAAE,SAAS,UAAU,EAAE,EAC5B,QAAQ,EAAE,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,EAC1C,WAAW,EAAE,OAAO,GACnB,cAAc,GAAG,SAAS,CAQ5B;AAED;;;;;;;GAOG;AACH,wBAAgB,qBAAqB,CACnC,KAAK,EAAE,SAAS,UAAU,EAAE,EAC5B,QAAQ,EAAE,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,GACzC,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE,CAQjC"}
+{"version":3,"file":"sandhi-input.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi-input.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAClD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AAE5C;;;;;;;;;GASG;AACH,wBAAgB,UAAU,CACxB,UAAU,EAAE,UAAU,EACtB,KAAK,EAAE,SAAS,UAAU,EAAE,EAC5B,QAAQ,EAAE,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,EAC1C,WAAW,EAAE,OAAO,GACnB,cAAc,CAQhB;AAED;;;;;;;;GAQG;AACH,wBAAgB,qBAAqB,CACnC,KAAK,EAAE,SAAS,UAAU,EAAE,EAC5B,QAAQ,EAAE,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,GACzC,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE,CAQjC"}

package/dist/decode/sandhi-input.js

@@ -1,17 +1,21 @@
 import { toCharacters } from "../script/characters.js";
 import { divisionOf } from "./constituents.js";
 /**
- * Where the words and their constituents fall, for third-tone sandhi.
+ * Where the words and their constituents fall.
  *
- * Undefined unless third-tone sandhi was asked for, since dividing a word costs
- * dictionary lookups and nothing else reads the answer.
+ * One entry per word always, because 一 sandhi reads the word ends: a 一 that
+ * closes a longer word is not counting the word after it, so 唯一 stays `wéiyī`
+ * in front of anything.
+ *
+ * Where a word *divides* is only worked out for third-tone sandhi, which is the
+ * only pass that reads it and costs a dictionary lookup per word to answer.
  */
 export function groupingOf(dictionary, words, readings, isThirdTone) {
-    if (!isThirdTone) {
-        return undefined;
-    }
     return words.map((scored, index) => {
         const reading = readings[index] ?? [];
+        if (!isThirdTone) {
+            return reading.length;
+        }
         return divisionOf(dictionary, scored.word.text, reading) ?? reading.length;
     });
 }
@@ -21,6 +25,7 @@ export function groupingOf(dictionary, words, readings, isThirdTone) {
  * A word whose reading is a different length from its text — 玩儿 as `wánr` —
  * has no character to give any one of its syllables, so its syllables get none
  * and the pass falls back to their spellings. See
+ * {@link import("./sandhi-tones.js").isYi} and
  * {@link import("./sandhi-tones.js").isCounting}.
  */
 export function charactersPerSyllable(words, readings) {

package/dist/decode/sandhi-input.js.map

@@ -1 +1 @@
-{"version":3,"file":"sandhi-input.js","sourceRoot":"","sources":["../../src/decode/sandhi-input.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAEvD,OAAO,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAI/C;;;;;GAKG;AACH,MAAM,UAAU,UAAU,CACxB,UAAsB,EACtB,KAA4B,EAC5B,QAA0C,EAC1C,WAAoB;IAEpB,IAAI,CAAC,WAAW,EAAE,CAAC;QACjB,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,KAAK,EAAE,EAAE;QACjC,MAAM,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC;QACtC,OAAO,UAAU,CAAC,UAAU,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,OAAO,CAAC,MAAM,CAAC;IAC7E,CAAC,CAAC,CAAC;AACL,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,qBAAqB,CACnC,KAA4B,EAC5B,QAA0C;IAE1C,OAAO,KAAK,CAAC,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,EAAmC,EAAE;QACtE,MAAM,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC;QACtC,MAAM,IAAI,GAAG,YAAY,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAC5C,OAAO,IAAI,CAAC,MAAM,KAAK,OAAO,CAAC,MAAM;YACnC,CAAC,CAAC,IAAI;YACN,CAAC,CAAC,KAAK,CAAC,IAAI,CAAqB,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;IACjE,CAAC,CAAC,CAAC;AACL,CAAC"}
+{"version":3,"file":"sandhi-input.js","sourceRoot":"","sources":["../../src/decode/sandhi-input.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAEvD,OAAO,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAI/C;;;;;;;;;GASG;AACH,MAAM,UAAU,UAAU,CACxB,UAAsB,EACtB,KAA4B,EAC5B,QAA0C,EAC1C,WAAoB;IAEpB,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,KAAK,EAAE,EAAE;QACjC,MAAM,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC;QACtC,IAAI,CAAC,WAAW,EAAE,CAAC;YACjB,OAAO,OAAO,CAAC,MAAM,CAAC;QACxB,CAAC;QACD,OAAO,UAAU,CAAC,UAAU,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,OAAO,CAAC,MAAM,CAAC;IAC7E,CAAC,CAAC,CAAC;AACL,CAAC;AAED;;;;;;;;GAQG;AACH,MAAM,UAAU,qBAAqB,CACnC,KAA4B,EAC5B,QAA0C;IAE1C,OAAO,KAAK,CAAC,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,EAAmC,EAAE;QACtE,MAAM,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC;QACtC,MAAM,IAAI,GAAG,YAAY,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAC5C,OAAO,IAAI,CAAC,MAAM,KAAK,OAAO,CAAC,MAAM;YACnC,CAAC,CAAC,IAAI;YACN,CAAC,CAAC,KAAK,CAAC,IAAI,CAAqB,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;IACjE,CAAC,CAAC,CAAC;AACL,CAAC"}

package/dist/decode/sandhi-tones.d.ts

@@ -7,13 +7,29 @@
 import { type Syllable } from "../syllable/syllable.js";
 import { type Tone } from "../tone/tone.js";
 /**
- * Whether a syllable is 一, which is the only thing spelled `yi` that sandhis.
+ * The 汉字 一.
  */
-export declare function isYi(syllable: Syllable): boolean;
+export declare const YI_CHARACTER = "\u4E00";
+/**
+ * The 汉字 不.
+ */
+export declare const BU_CHARACTER = "\u4E0D";
+/**
+ * Whether a syllable is 一.
+ *
+ * **Answered from the 汉字 where the caller has one**, and from the spelling
+ * where it does not, the way {@link isCounting} is. The spelling is not the
+ * morpheme and never was: 医, 依, 衣 and 伊 are all `yī` and none of them
+ * changes tone for what follows it. This rule is a fact about one character.
+ */
+export declare function isYi(syllable: Syllable, character?: string): boolean;
 /**
  * Whether a syllable is 不.
+ *
+ * Read from the 汉字 or the spelling on the same terms as {@link isYi}, and for
+ * the same reason: 部, 布, 步 and 埠 are all `bù` and none of them flattens.
  */
-export declare function isBu(syllable: Syllable): boolean;
+export declare function isBu(syllable: Syllable, character?: string): boolean;
 /**
  * The tone 一 takes before a given following tone.
  *

package/dist/decode/sandhi-tones.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"sandhi-tones.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi-tones.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAE,KAAK,QAAQ,EAAyB,MAAM,yBAAyB,CAAC;AAC/E,OAAO,EAAgB,KAAK,IAAI,EAAE,MAAM,iBAAiB,CAAC;AAE1D;;GAEG;AACH,wBAAgB,IAAI,CAAC,QAAQ,EAAE,QAAQ,GAAG,OAAO,CAEhD;AAED;;GAEG;AACH,wBAAgB,IAAI,CAAC,QAAQ,EAAE,QAAQ,GAAG,OAAO,CAEhD;AAED;;;;;;GAMG;AACH,wBAAgB,YAAY,CAAC,SAAS,EAAE,IAAI,GAAG,SAAS,GAAG,IAAI,CAK9D;AAED;;;;;;;;;;;;GAYG;AACH,eAAO,MAAM,aAAa,aAAyC,CAAC;AAEpE;;;;;GAKG;AACH,eAAO,MAAM,gBAAgB,aAAsC,CAAC;AAEpE;;GAEG;AACH,eAAO,MAAM,cAAc,OAAO,CAAC;AAEnC;;;;;;;;;;;;GAYG;AACH,eAAO,MAAM,kBAAkB,aAS7B,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,2BAA2B,aAItC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,iBAAiB,WAAM,CAAC;AAErC;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,wBAAgB,UAAU,CACxB,SAAS,EAAE,SAAS,QAAQ,EAAE,EAC9B,EAAE,EAAE,MAAM,EACV,UAAU,GAAE,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAO,GAC/C,OAAO,CAwBT"}
+{"version":3,"file":"sandhi-tones.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi-tones.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAE,KAAK,QAAQ,EAAyB,MAAM,yBAAyB,CAAC;AAC/E,OAAO,EAAgB,KAAK,IAAI,EAAE,MAAM,iBAAiB,CAAC;AAE1D;;GAEG;AACH,eAAO,MAAM,YAAY,WAAM,CAAC;AAEhC;;GAEG;AACH,eAAO,MAAM,YAAY,WAAM,CAAC;AAEhC;;;;;;;GAOG;AACH,wBAAgB,IAAI,CAAC,QAAQ,EAAE,QAAQ,EAAE,SAAS,CAAC,EAAE,MAAM,GAAG,OAAO,CAIpE;AAED;;;;;GAKG;AACH,wBAAgB,IAAI,CAAC,QAAQ,EAAE,QAAQ,EAAE,SAAS,CAAC,EAAE,MAAM,GAAG,OAAO,CAIpE;AAED;;;;;;GAMG;AACH,wBAAgB,YAAY,CAAC,SAAS,EAAE,IAAI,GAAG,SAAS,GAAG,IAAI,CAK9D;AAED;;;;;;;;;;;;GAYG;AACH,eAAO,MAAM,aAAa,aAAyC,CAAC;AAEpE;;;;;GAKG;AACH,eAAO,MAAM,gBAAgB,aAAsC,CAAC;AAEpE;;GAEG;AACH,eAAO,MAAM,cAAc,OAAO,CAAC;AAEnC;;;;;;;;;;;;GAYG;AACH,eAAO,MAAM,kBAAkB,aAS7B,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,2BAA2B,aAItC,CAAC;AAEH;;GAEG;AACH,eAAO,MAAM,iBAAiB,WAAM,CAAC;AAErC;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,wBAAgB,UAAU,CACxB,SAAS,EAAE,SAAS,QAAQ,EAAE,EAC9B,EAAE,EAAE,MAAM,EACV,UAAU,GAAE,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAO,GAC/C,OAAO,CAwBT"}

package/dist/decode/sandhi-tones.js

@@ -7,16 +7,36 @@
 import { writeSyllableSpelling } from "../syllable/syllable.js";
 import { NEUTRAL_TONE } from "../tone/tone.js";
 /**
- * Whether a syllable is 一, which is the only thing spelled `yi` that sandhis.
+ * The 汉字 一.
  */
-export function isYi(syllable) {
-    return syllable.initial === "" && syllable.final === "i";
+export const YI_CHARACTER = "一";
+/**
+ * The 汉字 不.
+ */
+export const BU_CHARACTER = "不";
+/**
+ * Whether a syllable is 一.
+ *
+ * **Answered from the 汉字 where the caller has one**, and from the spelling
+ * where it does not, the way {@link isCounting} is. The spelling is not the
+ * morpheme and never was: 医, 依, 衣 and 伊 are all `yī` and none of them
+ * changes tone for what follows it. This rule is a fact about one character.
+ */
+export function isYi(syllable, character) {
+    return character === undefined
+        ? syllable.initial === "" && syllable.final === "i"
+        : character === YI_CHARACTER;
 }
 /**
  * Whether a syllable is 不.
+ *
+ * Read from the 汉字 or the spelling on the same terms as {@link isYi}, and for
+ * the same reason: 部, 布, 步 and 埠 are all `bù` and none of them flattens.
  */
-export function isBu(syllable) {
-    return syllable.initial === "b" && syllable.final === "u";
+export function isBu(syllable, character) {
+    return character === undefined
+        ? syllable.initial === "b" && syllable.final === "u"
+        : character === BU_CHARACTER;
 }
 /**
  * The tone 一 takes before a given following tone.

package/dist/decode/sandhi-tones.js.map

@@ -1 +1 @@
-{"version":3,"file":"sandhi-tones.js","sourceRoot":"","sources":["../../src/decode/sandhi-tones.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAiB,qBAAqB,EAAE,MAAM,yBAAyB,CAAC;AAC/E,OAAO,EAAE,YAAY,EAAa,MAAM,iBAAiB,CAAC;AAE1D;;GAEG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB;IACrC,OAAO,QAAQ,CAAC,OAAO,KAAK,EAAE,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG,CAAC;AAC3D,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB;IACrC,OAAO,QAAQ,CAAC,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG,CAAC;AAC5D,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,YAAY,CAAC,SAA2B;IACtD,IAAI,SAAS,KAAK,SAAS,IAAI,SAAS,KAAK,YAAY,EAAE,CAAC;QAC1D,OAAO,CAAC,CAAC;IACX,CAAC;IACD,OAAO,SAAS,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AACjC,CAAC;AAED;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,aAAa,GAAG,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC;AAEpE;;;;;GAKG;AACH,MAAM,CAAC,MAAM,gBAAgB,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,aAAa,EAAE,MAAM,CAAC,CAAC,CAAC;AAEpE;;GAEG;AACH,MAAM,CAAC,MAAM,cAAc,GAAG,IAAI,CAAC;AAEnC;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,kBAAkB,GAAG,IAAI,GAAG,CAAC;IACxC,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;CACJ,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,CAAC,MAAM,2BAA2B,GAAG,IAAI,GAAG,CAAC;IACjD,GAAG,kBAAkB;IACrB,GAAG;IACH,GAAG;CACJ,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,CAAC,MAAM,iBAAiB,GAAG,GAAG,CAAC;AAErC;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,MAAM,UAAU,UAAU,CACxB,SAA8B,EAC9B,EAAU,EACV,UAAU,GAAoC,EAAE;IAEhD,MAAM,KAAK,GAAG,CAAC,KAAa,EAAU,EAAE;QACtC,MAAM,QAAQ,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC;QAClC,OAAO,QAAQ,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,qBAAqB,CAAC,QAAQ,CAAC,CAAC;IACvE,CAAC,CAAC;IACF,MAAM,MAAM,GAAG,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;IAClC,IACE,MAAM,KAAK,SAAS;QAClB,CAAC,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,KAAK,cAAc;QAClC,CAAC,CAAC,MAAM,KAAK,iBAAiB,EAChC,CAAC;QACD,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IACE,MAAM,KAAK,SAAS;QAClB,CAAC,CAAC,CAAC,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACnC,CAAC,CAAC,CAAC,kBAAkB,CAAC,GAAG,CAAC,MAAM,CAAC,EACnC,CAAC;QACD,OAAO,IAAI,CAAC;IACd,CAAC;IACD,MAAM,KAAK,GAAG,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;IACjC,OAAO,KAAK,KAAK,SAAS;QACxB,CAAC,CAAC,gBAAgB,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACrC,CAAC,CAAC,2BAA2B,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;AAC7C,CAAC"}
+{"version":3,"file":"sandhi-tones.js","sourceRoot":"","sources":["../../src/decode/sandhi-tones.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAiB,qBAAqB,EAAE,MAAM,yBAAyB,CAAC;AAC/E,OAAO,EAAE,YAAY,EAAa,MAAM,iBAAiB,CAAC;AAE1D;;GAEG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,GAAG,CAAC;AAEhC;;GAEG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,GAAG,CAAC;AAEhC;;;;;;;GAOG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB,EAAE,SAAkB;IACzD,OAAO,SAAS,KAAK,SAAS;QAC5B,CAAC,CAAC,QAAQ,CAAC,OAAO,KAAK,EAAE,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG;QACnD,CAAC,CAAC,SAAS,KAAK,YAAY,CAAC;AACjC,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB,EAAE,SAAkB;IACzD,OAAO,SAAS,KAAK,SAAS;QAC5B,CAAC,CAAC,QAAQ,CAAC,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG;QACpD,CAAC,CAAC,SAAS,KAAK,YAAY,CAAC;AACjC,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,YAAY,CAAC,SAA2B;IACtD,IAAI,SAAS,KAAK,SAAS,IAAI,SAAS,KAAK,YAAY,EAAE,CAAC;QAC1D,OAAO,CAAC,CAAC;IACX,CAAC;IACD,OAAO,SAAS,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AACjC,CAAC;AAED;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,aAAa,GAAG,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC;AAEpE;;;;;GAKG;AACH,MAAM,CAAC,MAAM,gBAAgB,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,aAAa,EAAE,MAAM,CAAC,CAAC,CAAC;AAEpE;;GAEG;AACH,MAAM,CAAC,MAAM,cAAc,GAAG,IAAI,CAAC;AAEnC;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,kBAAkB,GAAG,IAAI,GAAG,CAAC;IACxC,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;CACJ,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,CAAC,MAAM,2BAA2B,GAAG,IAAI,GAAG,CAAC;IACjD,GAAG,kBAAkB;IACrB,GAAG;IACH,GAAG;CACJ,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,CAAC,MAAM,iBAAiB,GAAG,GAAG,CAAC;AAErC;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,MAAM,UAAU,UAAU,CACxB,SAA8B,EAC9B,EAAU,EACV,UAAU,GAAoC,EAAE;IAEhD,MAAM,KAAK,GAAG,CAAC,KAAa,EAAU,EAAE;QACtC,MAAM,QAAQ,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC;QAClC,OAAO,QAAQ,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,qBAAqB,CAAC,QAAQ,CAAC,CAAC;IACvE,CAAC,CAAC;IACF,MAAM,MAAM,GAAG,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;IAClC,IACE,MAAM,KAAK,SAAS;QAClB,CAAC,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,KAAK,cAAc;QAClC,CAAC,CAAC,MAAM,KAAK,iBAAiB,EAChC,CAAC;QACD,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IACE,MAAM,KAAK,SAAS;QAClB,CAAC,CAAC,CAAC,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACnC,CAAC,CAAC,CAAC,kBAAkB,CAAC,GAAG,CAAC,MAAM,CAAC,EACnC,CAAC;QACD,OAAO,IAAI,CAAC;IACd,CAAC;IACD,MAAM,KAAK,GAAG,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;IACjC,OAAO,KAAK,KAAK,SAAS;QACxB,CAAC,CAAC,gBAAgB,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACrC,CAAC,CAAC,2BAA2B,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;AAC7C,CAAC"}

package/dist/decode/sandhi.d.ts

@@ -1,4 +1,6 @@
 import type { Syllable } from "../syllable/syllable.js";
+import { type SandhiGrouping } from "./sandhi-grouping.js";
+export type { SandhiGrouping } from "./sandhi-grouping.js";
 /**
  * How much of the sandhi to apply.
  */
@@ -19,17 +21,6 @@ export interface SandhiOptions {
      */
     readonly thirdTone?: boolean;
 }
-/**
- * How the syllables group, for third-tone sandhi to read.
- *
- * One entry per word: its syllable count, or the counts of the constituents it
- * divides into. 老板 很 好 is `[2, 1, 1]` and 纸老虎 is `[[1, 2]]`.
- *
- * Sandhi's domain is the prosodic foot rather than the word, and the foot is
- * built from structure — which is why this is worth carrying. See
- * {@link applyThirdTone} for what is done with it.
- */
-export type SandhiGrouping = readonly (number | readonly number[])[];
 /**
  * Apply the tone sandhi that standard orthography writes.
  *
@@ -39,14 +30,24 @@ export type SandhiGrouping = readonly (number | readonly number[])[];
  * contextual tone is put back, over the whole syllable array rather than within
  * a word, which is what lets 不 in one word assimilate to a tone in the next.
  *
- * 不 assimilates to whatever follows it and needs nothing but the syllables.
- * Third-tone sandhi is not like that — its domain is the foot rather than the
- * syllable string — so `grouping` says where the words and their parts are.
- * Without one the whole array is taken for a single word, which is what a
- * caller holding nothing but a reading has. 一 sits between the two: the tone it
- * takes is settled by the syllables, and whether it takes one at all is a
- * question about the 汉字, so `characters` carries them where the caller has
- * them. See {@link isCounting}.
+ * A caller holding nothing but a reading can still use this, and several do:
+ * `pinyinjs sandhi bùshì` and the numeral reader both hand over syllables and
+ * nothing else. `grouping` and `characters` are what a caller that knows more
+ * says so with, and each one is read on its own, since a caller can have one
+ * and not the other.
+ *
+ * `characters` holds one 汉字 per syllable, and is what tells 一 and 不 apart
+ * from the syllables that merely sound like them — 医生 is `yīshēng` and 部队
+ * is `bùduì`. Without it the pass falls back to the spellings, which is all
+ * bare pinyin offers. See {@link isYi} and {@link isCounting}.
+ *
+ * `grouping` says where the words and their parts are. Third-tone sandhi needs
+ * it because its domain is the prosodic foot rather than the syllable string,
+ * and 一 needs the word ends out of it: a 一 that closes a longer word is not
+ * counting the word after it. Without a grouping the whole array is taken for
+ * a single word, which is all a bare reading says.
+ *
+ * 不 is the one rule that needs neither: it assimilates to whatever follows.
  *
  * Never operates on a string. The old project patched output text with regexes,
  * which is what made its rules order-dependent and untestable.

package/dist/decode/sandhi.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"sandhi.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD;;GAEG;AACH,MAAM,WAAW,aAAa;IAC5B;;;;;OAKG;IACH,QAAQ,CAAC,IAAI,CAAC,EAAE,OAAO,CAAC;IACxB;;;;;;OAMG;IACH,QAAQ,CAAC,SAAS,CAAC,EAAE,OAAO,CAAC;CAC9B;AAED;;;;;;;;;GASG;AACH,MAAM,MAAM,cAAc,GAAG,SAAS,CAAC,MAAM,GAAG,SAAS,MAAM,EAAE,CAAC,EAAE,CAAC;AA6HrE;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,wBAAgB,WAAW,CACzB,SAAS,EAAE,SAAS,QAAQ,EAAE,EAC9B,OAAO,GAAE,aAAkB,EAC3B,QAAQ,CAAC,EAAE,cAAc,EACzB,UAAU,CAAC,EAAE,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE,GAC3C,SAAS,QAAQ,EAAE,CA6BrB"}
+{"version":3,"file":"sandhi.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AACxD,OAAO,EAGL,KAAK,cAAc,EACpB,MAAM,sBAAsB,CAAC;AAG9B,YAAY,EAAE,cAAc,EAAE,MAAM,sBAAsB,CAAC;AAE3D;;GAEG;AACH,MAAM,WAAW,aAAa;IAC5B;;;;;OAKG;IACH,QAAQ,CAAC,IAAI,CAAC,EAAE,OAAO,CAAC;IACxB;;;;;;OAMG;IACH,QAAQ,CAAC,SAAS,CAAC,EAAE,OAAO,CAAC;CAC9B;AA+DD;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,wBAAgB,WAAW,CACzB,SAAS,EAAE,SAAS,QAAQ,EAAE,EAC9B,OAAO,GAAE,aAAkB,EAC3B,QAAQ,CAAC,EAAE,cAAc,EACzB,UAAU,CAAC,EAAE,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE,GAC3C,SAAS,QAAQ,EAAE,CAsCrB"}

package/dist/decode/sandhi.js

@@ -1,45 +1,5 @@
+import { endingsOf, junctionsOf, } from "./sandhi-grouping.js";
 import { isBu, isCounting, isYi, yiToneBefore } from "./sandhi-tones.js";
-/**
- * Read a grouping as the junctions it puts in the syllable array.
- *
- * The whole array is treated as one undivided word where no grouping is given,
- * or where the one given does not account for exactly the syllables there are —
- * a grouping that does not fit is describing some other text, and guessing
- * which syllables it meant would be worse than ignoring it.
- */
-function junctionsOf(length, grouping) {
-    const undivided = {
-        inner: Array.from({ length: Math.max(length - 1, 0) }, (_, at) => at),
-        parts: [],
-        words: [],
-    };
-    if (grouping === undefined) {
-        return undivided;
-    }
-    const inner = [];
-    const parts = [];
-    const words = [];
-    let at = 0;
-    for (const [index, word] of grouping.entries()) {
-        const divisions = typeof word === "number" ? [word] : word;
-        const from = at;
-        for (const [division, size] of divisions.entries()) {
-            for (let step = 1; step < size; step++) {
-                inner.push(at + step - 1);
-            }
-            at += size;
-            // The junction closing a word's last constituent is the one *around* the
-            // word rather than one inside it.
-            if (division < divisions.length - 1) {
-                parts.push(at - 1);
-            }
-        }
-        if (at - from === 1 && index < grouping.length - 1) {
-            words.push(at - 1);
-        }
-    }
-    return at === length ? { inner, parts, words } : undivided;
-}
 /**
  * Lower a third tone to a second where the syllable after it is a third too.
  *
@@ -103,14 +63,24 @@ function applyThirdTone(syllables, grouping) {
  * contextual tone is put back, over the whole syllable array rather than within
  * a word, which is what lets 不 in one word assimilate to a tone in the next.
  *
- * 不 assimilates to whatever follows it and needs nothing but the syllables.
- * Third-tone sandhi is not like that — its domain is the foot rather than the
- * syllable string — so `grouping` says where the words and their parts are.
- * Without one the whole array is taken for a single word, which is what a
- * caller holding nothing but a reading has. 一 sits between the two: the tone it
- * takes is settled by the syllables, and whether it takes one at all is a
- * question about the 汉字, so `characters` carries them where the caller has
- * them. See {@link isCounting}.
+ * A caller holding nothing but a reading can still use this, and several do:
+ * `pinyinjs sandhi bùshì` and the numeral reader both hand over syllables and
+ * nothing else. `grouping` and `characters` are what a caller that knows more
+ * says so with, and each one is read on its own, since a caller can have one
+ * and not the other.
+ *
+ * `characters` holds one 汉字 per syllable, and is what tells 一 and 不 apart
+ * from the syllables that merely sound like them — 医生 is `yīshēng` and 部队
+ * is `bùduì`. Without it the pass falls back to the spellings, which is all
+ * bare pinyin offers. See {@link isYi} and {@link isCounting}.
+ *
+ * `grouping` says where the words and their parts are. Third-tone sandhi needs
+ * it because its domain is the prosodic foot rather than the syllable string,
+ * and 一 needs the word ends out of it: a 一 that closes a longer word is not
+ * counting the word after it. Without a grouping the whole array is taken for
+ * a single word, which is all a bare reading says.
+ *
+ * 不 is the one rule that needs neither: it assimilates to whatever follows.
  *
  * Never operates on a string. The old project patched output text with regexes,
  * which is what made its rules order-dependent and untestable.
@@ -120,19 +90,28 @@ export function applySandhi(syllables, options = {}, grouping, characters) {
     const applied = thirdTone
         ? applyThirdTone(syllables, grouping)
         : [...syllables];
+    const endings = endingsOf(syllables.length, grouping);
     for (const [at, syllable] of applied.entries()) {
+        const character = characters?.[at];
         const following = applied[at + 1];
-        if (yiBu && isYi(syllable) && syllable.tone === 1) {
+        if (yiBu && isYi(syllable, character) && syllable.tone === 1) {
             // A 一 that is not counting keeps its citation tone, so it is left
             // exactly as the dictionary stored it.
             if (isCounting(applied, at, characters)) {
-                applied[at] = { ...syllable, tone: yiToneBefore(following?.tone) };
+                // A 一 closing a longer word counts nothing, because what follows it
+                // is outside the word: 唯一 is `wéiyī` whatever comes next, and so are
+                // 之一, 统一 and 星期一. The 一 of 一个 is a word of its own, which is
+                // why it still assimilates across the boundary.
+                const counted = endings.has(at) ? undefined : following;
+                applied[at] = { ...syllable, tone: yiToneBefore(counted?.tone) };
             }
             continue;
         }
         // 不 flattens to second tone before a fourth, and is otherwise unchanged.
+        // Unlike 一 it keeps assimilating at the end of a word, since what it
+        // negates is regularly in the next one: 决不会 is `juébú huì`.
         if (yiBu &&
-            isBu(syllable) &&
+            isBu(syllable, character) &&
             syllable.tone === 4 &&
             following?.tone === 4) {
             applied[at] = { ...syllable, tone: 2 };

package/dist/decode/sandhi.js.map

@@ -1 +1 @@
-{"version":3,"file":"sandhi.js","sourceRoot":"","sources":["../../src/decode/sandhi.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,IAAI,EAAE,UAAU,EAAE,IAAI,EAAE,YAAY,EAAE,MAAM,mBAAmB,CAAC;AAkDzE;;;;;;;GAOG;AACH,SAAS,WAAW,CAClB,MAAc,EACd,QAAyB;IAEzB,MAAM,SAAS,GAAoB;QACjC,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC;QACrE,KAAK,EAAE,EAAE;QACT,KAAK,EAAE,EAAE;KACV,CAAC;IACF,IAAI,QAAQ,KAAK,SAAS,EAAE,CAAC;QAC3B,OAAO,SAAS,CAAC;IACnB,CAAC;IAED,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,MAAM,KAAK,GAAa,EAAE,CAAC;IAC3B,IAAI,EAAE,GAAG,CAAC,CAAC;IACX,KAAK,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC,IAAI,QAAQ,CAAC,OAAO,EAAE,EAAE,CAAC;QAC/C,MAAM,SAAS,GAAG,OAAO,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC;QAC3D,MAAM,IAAI,GAAG,EAAE,CAAC;QAChB,KAAK,MAAM,CAAC,QAAQ,EAAE,IAAI,CAAC,IAAI,SAAS,CAAC,OAAO,EAAE,EAAE,CAAC;YACnD,KAAK,IAAI,IAAI,GAAG,CAAC,EAAE,IAAI,GAAG,IAAI,EAAE,IAAI,EAAE,EAAE,CAAC;gBACvC,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,IAAI,GAAG,CAAC,CAAC,CAAC;YAC5B,CAAC;YACD,EAAE,IAAI,IAAI,CAAC;YACX,yEAAyE;YACzE,kCAAkC;YAClC,IAAI,QAAQ,GAAG,SAAS,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;gBACpC,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;YACrB,CAAC;QACH,CAAC;QACD,IAAI,EAAE,GAAG,IAAI,KAAK,CAAC,IAAI,KAAK,GAAG,QAAQ,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YACnD,KAAK,CAAC,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACrB,CAAC;IACH,CAAC;IACD,OAAO,EAAE,KAAK,MAAM,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,SAAS,CAAC;AAC7D,CAAC;AAED;;;;;;GAMG;AACH,SAAS,KAAK,CAAC,OAAmB,EAAE,EAAU;IAC5C,MAAM,QAAQ,GAAG,OAAO,CAAC,EAAE,CAAC,CAAC;IAC7B,IAAI,QAAQ,EAAE,IAAI,KAAK,CAAC,IAAI,OAAO,CAAC,EAAE,GAAG,CAAC,CAAC,EAAE,IAAI,KAAK,CAAC,EAAE,CAAC;QACxD,OAAO,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,QAAQ,EAAE,IAAI,EAAE,CAAC,EAAE,CAAC;IACzC,CAAC;AACH,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,SAAS,cAAc,CACrB,SAA8B,EAC9B,QAAyB;IAEzB,MAAM,OAAO,GAAG,CAAC,GAAG,SAAS,CAAC,CAAC;IAC/B,MAAM,SAAS,GAAG,WAAW,CAAC,SAAS,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;IAE1D,KAAK,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC,KAAK,EAAE,SAAS,CAAC,KAAK,EAAE,SAAS,CAAC,KAAK,CAAC,EAAE,CAAC;QACvE,KAAK,MAAM,EAAE,IAAI,IAAI,EAAE,CAAC;YACtB,KAAK,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;QACrB,CAAC;IACH,CAAC;IAED,OAAO,OAAO,CAAC;AACjB,CAAC;AAED;;;;;;;;;;;;;;;;;;;;GAoBG;AACH,MAAM,UAAU,WAAW,CACzB,SAA8B,EAC9B,OAAO,GAAkB,EAAE,EAC3B,QAAyB,EACzB,UAA4C;IAE5C,MAAM,EAAE,IAAI,GAAG,IAAI,EAAE,SAAS,GAAG,KAAK,EAAE,GAAG,OAAO,CAAC;IACnD,MAAM,OAAO,GAAG,SAAS;QACvB,CAAC,CAAC,cAAc,CAAC,SAAS,EAAE,QAAQ,CAAC;QACrC,CAAC,CAAC,CAAC,GAAG,SAAS,CAAC,CAAC;IAEnB,KAAK,MAAM,CAAC,EAAE,EAAE,QAAQ,CAAC,IAAI,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC;QAC/C,MAAM,SAAS,GAAG,OAAO,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAElC,IAAI,IAAI,IAAI,IAAI,CAAC,QAAQ,CAAC,IAAI,QAAQ,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC;YAClD,kEAAkE;YAClE,uCAAuC;YACvC,IAAI,UAAU,CAAC,OAAO,EAAE,EAAE,EAAE,UAAU,CAAC,EAAE,CAAC;gBACxC,OAAO,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,QAAQ,EAAE,IAAI,EAAE,YAAY,CAAC,SAAS,EAAE,IAAI,CAAC,EAAE,CAAC;YACrE,CAAC;YACD,SAAS;QACX,CAAC;QACD,yEAAyE;QACzE,IACE,IAAI;YACJ,IAAI,CAAC,QAAQ,CAAC;YACd,QAAQ,CAAC,IAAI,KAAK,CAAC;YACnB,SAAS,EAAE,IAAI,KAAK,CAAC,EACrB,CAAC;YACD,OAAO,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,QAAQ,EAAE,IAAI,EAAE,CAAC,EAAE,CAAC;QACzC,CAAC;IACH,CAAC;IAED,OAAO,OAAO,CAAC;AACjB,CAAC"}
+{"version":3,"file":"sandhi.js","sourceRoot":"","sources":["../../src/decode/sandhi.ts"],"names":[],"mappings":"AACA,OAAO,EACL,SAAS,EACT,WAAW,GAEZ,MAAM,sBAAsB,CAAC;AAC9B,OAAO,EAAE,IAAI,EAAE,UAAU,EAAE,IAAI,EAAE,YAAY,EAAE,MAAM,mBAAmB,CAAC;AAyBzE;;;;;;GAMG;AACH,SAAS,KAAK,CAAC,OAAmB,EAAE,EAAU;IAC5C,MAAM,QAAQ,GAAG,OAAO,CAAC,EAAE,CAAC,CAAC;IAC7B,IAAI,QAAQ,EAAE,IAAI,KAAK,CAAC,IAAI,OAAO,CAAC,EAAE,GAAG,CAAC,CAAC,EAAE,IAAI,KAAK,CAAC,EAAE,CAAC;QACxD,OAAO,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,QAAQ,EAAE,IAAI,EAAE,CAAC,EAAE,CAAC;IACzC,CAAC;AACH,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,SAAS,cAAc,CACrB,SAA8B,EAC9B,QAAyB;IAEzB,MAAM,OAAO,GAAG,CAAC,GAAG,SAAS,CAAC,CAAC;IAC/B,MAAM,SAAS,GAAG,WAAW,CAAC,SAAS,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;IAE1D,KAAK,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC,KAAK,EAAE,SAAS,CAAC,KAAK,EAAE,SAAS,CAAC,KAAK,CAAC,EAAE,CAAC;QACvE,KAAK,MAAM,EAAE,IAAI,IAAI,EAAE,CAAC;YACtB,KAAK,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;QACrB,CAAC;IACH,CAAC;IAED,OAAO,OAAO,CAAC;AACjB,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,MAAM,UAAU,WAAW,CACzB,SAA8B,EAC9B,OAAO,GAAkB,EAAE,EAC3B,QAAyB,EACzB,UAA4C;IAE5C,MAAM,EAAE,IAAI,GAAG,IAAI,EAAE,SAAS,GAAG,KAAK,EAAE,GAAG,OAAO,CAAC;IACnD,MAAM,OAAO,GAAG,SAAS;QACvB,CAAC,CAAC,cAAc,CAAC,SAAS,EAAE,QAAQ,CAAC;QACrC,CAAC,CAAC,CAAC,GAAG,SAAS,CAAC,CAAC;IAEnB,MAAM,OAAO,GAAG,SAAS,CAAC,SAAS,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;IACtD,KAAK,MAAM,CAAC,EAAE,EAAE,QAAQ,CAAC,IAAI,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC;QAC/C,MAAM,SAAS,GAAG,UAAU,EAAE,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,OAAO,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAElC,IAAI,IAAI,IAAI,IAAI,CAAC,QAAQ,EAAE,SAAS,CAAC,IAAI,QAAQ,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC;YAC7D,kEAAkE;YAClE,uCAAuC;YACvC,IAAI,UAAU,CAAC,OAAO,EAAE,EAAE,EAAE,UAAU,CAAC,EAAE,CAAC;gBACxC,oEAAoE;gBACpE,qEAAqE;gBACrE,6DAA6D;gBAC7D,gDAAgD;gBAChD,MAAM,OAAO,GAAG,OAAO,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC;gBACxD,OAAO,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,QAAQ,EAAE,IAAI,EAAE,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC;YACnE,CAAC;YACD,SAAS;QACX,CAAC;QACD,yEAAyE;QACzE,qEAAqE;QACrE,4DAA4D;QAC5D,IACE,IAAI;YACJ,IAAI,CAAC,QAAQ,EAAE,SAAS,CAAC;YACzB,QAAQ,CAAC,IAAI,KAAK,CAAC;YACnB,SAAS,EAAE,IAAI,KAAK,CAAC,EACrB,CAAC;YACD,OAAO,CAAC,EAAE,CAAC,GAAG,EAAE,GAAG,QAAQ,EAAE,IAAI,EAAE,CAAC,EAAE,CAAC;QACzC,CAAC;IACH,CAAC;IAED,OAAO,OAAO,CAAC;AACjB,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@kensio/pinyinjs",
-  "version": "1.7.14",
+  "version": "1.8.0",
   "description": "Chinese hanzi to pinyin conversion",
   "repository": "https://github.com/KensioSoftware/pinyinjs",
   "homepage": "https://pinyinjs.dev",