@kensio/pinyinjs 1.6.0 → 1.7.1

package/dist/dictionary/corpus-mass.d.ts

@@ -0,0 +1,68 @@
+import type { CedictEntry } from "../sources/cedict.js";
+import type { JiebaEntry } from "../sources/jieba.js";
+import { type ReadingField } from "../sources/unihan.js";
+import type { Syllable } from "../syllable/syllable.js";
+import type { TraditionalTable } from "./traditional.js";
+/**
+ * The sources a character's corpus mass is counted over.
+ *
+ * The same three the merge already holds: the words, their readings, and how
+ * often jieba's corpus saw each word.
+ */
+export interface MassSources {
+    readonly phrase: ReadonlyMap<string, readonly string[]>;
+    readonly cedict: readonly CedictEntry[];
+    readonly jieba: ReadonlyMap<string, JiebaEntry>;
+}
+/**
+ * How much corpus each of a character's readings carries, keyed by reading.
+ */
+export type CorpusMass = ReadonlyMap<string, ReadonlyMap<string, number>>;
+/**
+ * Count how much of jieba's corpus each reading of each character carries.
+ *
+ * Every word the dictionary knows votes for the reading it gives each of its
+ * characters, weighted by how often jieba's corpus saw that word. It is the one
+ * piece of evidence about a polyphone that is neither a lexicographer's pick nor
+ * a dictionary's page order: 行 is `xíng` because 进行, 行政 and 行为 outweigh
+ * 行业 and 银行, and nothing had to say so.
+ *
+ * **Single characters do not vote.** A one-character entry's reading *is* the
+ * default being ranked — it comes from Unihan, or from a source that picked one
+ * reading for the character — so counting it would have the ranking confirm
+ * itself. Only a word puts a character in a context that chose between its
+ * readings.
+ *
+ * **繁體 is counted through the script pairing.** jieba's corpus is 简体 and the
+ * phrase corpus is 简体, so a traditional character appears in none of the words
+ * counted here and would rank on nothing at all. Each vote is therefore also
+ * cast for whichever 繁體 form the word's reading picks — 髮 gets 头发's `fà` and
+ * 發 gets 发送's `fā` — which is the same evidence reaching the same character
+ * under the other script.
+ */
+export declare function countCorpusMass(sources: MassSources, traditional: TraditionalTable): CorpusMass;
+/**
+ * Re-rank one character's readings by how much corpus each carries.
+ *
+ * The sort is stable and the mass of an unseen reading is zero, so a character
+ * no word covers keeps Unihan's order exactly, and so does the tail of readings
+ * below the ones words attest. The corpus only ever reorders what it has
+ * evidence about, and never past a leading 轻声 — see {@link rankingMass}.
+ *
+ * **Only where `kHanyuPinlu` did not already rank the character**, which is the
+ * whole of what this instrument is for. `kMandarin` names one reading and orders
+ * nothing, so where it decides the default there is nothing behind it, and that
+ * is what the corpus supplies. Where `kHanyuPinlu` lists several readings there
+ * already is a corpus ranking, and it is the better one: **this mass counts a
+ * character only inside words, so a reading used mostly as a bare character is
+ * invisible to it.** 呢 is 呢子 and 呢绒 in the dictionary and 呢 on the end of a
+ * question everywhere else; 都 is 首都 and 都市 against a bare 都 that is simply
+ * `dōu`; 将, 与, 处, 喝, 蒙 and 吗 are the same shape. `kHanyuPinlu` counts
+ * occurrences rather than words and has seen exactly what this cannot.
+ *
+ * Measured over CPP, that division is the whole difference between the rule
+ * paying and not: where `kHanyuPinlu` is silent or alone the mass fixes 438 and
+ * breaks 256, and where it ranks the character it fixes 201 and breaks 263.
+ */
+export declare function rankByCorpusMass(character: string, readings: readonly Syllable[], fields: ReadonlyMap<ReadingField, readonly string[]>, mass: CorpusMass): readonly Syllable[];
+//# sourceMappingURL=corpus-mass.d.ts.map

package/dist/dictionary/corpus-mass.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"corpus-mass.d.ts","sourceRoot":"","sources":["../../src/dictionary/corpus-mass.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AACxD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,qBAAqB,CAAC;AACtD,OAAO,EAAqB,KAAK,YAAY,EAAE,MAAM,sBAAsB,CAAC;AAC5E,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAGxD,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,kBAAkB,CAAC;AAEzD;;;;;GAKG;AACH,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,MAAM,EAAE,WAAW,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,CAAC,CAAC;IACxD,QAAQ,CAAC,MAAM,EAAE,SAAS,WAAW,EAAE,CAAC;IACxC,QAAQ,CAAC,KAAK,EAAE,WAAW,CAAC,MAAM,EAAE,UAAU,CAAC,CAAC;CACjD;AAED;;GAEG;AACH,MAAM,MAAM,UAAU,GAAG,WAAW,CAAC,MAAM,EAAE,WAAW,CAAC,MAAM,EAAE,MAAM,CAAC,CAAC,CAAC;AAiC1E;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,wBAAgB,eAAe,CAC7B,OAAO,EAAE,WAAW,EACpB,WAAW,EAAE,gBAAgB,GAC5B,UAAU,CAyCZ;AA2CD;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,wBAAgB,gBAAgB,CAC9B,SAAS,EAAE,MAAM,EACjB,QAAQ,EAAE,SAAS,QAAQ,EAAE,EAC7B,MAAM,EAAE,WAAW,CAAC,YAAY,EAAE,SAAS,MAAM,EAAE,CAAC,EACpD,IAAI,EAAE,UAAU,GACf,SAAS,QAAQ,EAAE,CAYrB"}

package/dist/dictionary/corpus-mass.js

@@ -0,0 +1,163 @@
+import { toCharacters } from "../script/characters.js";
+import { isFrequencyRanked } from "../sources/unihan.js";
+import { NEUTRAL_TONE } from "../tone/tone.js";
+import { readAlignedReading } from "./reading.js";
+/**
+ * A reading key that ignores 儿化, which is a suffix rather than a reading.
+ */
+function readingKey(syllable) {
+    return `${syllable.initial}|${syllable.final}|${String(syllable.tone)}`;
+}
+/**
+ * One reading per word, resolved the way the merge resolves it.
+ *
+ * The phrase corpus leads and CC-CEDICT fills its gaps, which is the precedence
+ * the merge itself applies. Taking both would let a word CC-CEDICT lists under
+ * several senses vote several times, and vote for readings the dictionary does
+ * not give it: 重点 is `zhòngdiǎn` in the dictionary, and counting every
+ * CC-CEDICT sense would have it also vote for `chóng`.
+ */
+function readingsByWord(sources) {
+    const readings = new Map();
+    for (const entry of sources.cedict) {
+        if (!readings.has(entry.simplified)) {
+            readings.set(entry.simplified, entry.readings);
+        }
+    }
+    for (const [word, reading] of sources.phrase) {
+        readings.set(word, reading);
+    }
+    return readings;
+}
+/**
+ * Count how much of jieba's corpus each reading of each character carries.
+ *
+ * Every word the dictionary knows votes for the reading it gives each of its
+ * characters, weighted by how often jieba's corpus saw that word. It is the one
+ * piece of evidence about a polyphone that is neither a lexicographer's pick nor
+ * a dictionary's page order: 行 is `xíng` because 进行, 行政 and 行为 outweigh
+ * 行业 and 银行, and nothing had to say so.
+ *
+ * **Single characters do not vote.** A one-character entry's reading *is* the
+ * default being ranked — it comes from Unihan, or from a source that picked one
+ * reading for the character — so counting it would have the ranking confirm
+ * itself. Only a word puts a character in a context that chose between its
+ * readings.
+ *
+ * **繁體 is counted through the script pairing.** jieba's corpus is 简体 and the
+ * phrase corpus is 简体, so a traditional character appears in none of the words
+ * counted here and would rank on nothing at all. Each vote is therefore also
+ * cast for whichever 繁體 form the word's reading picks — 髮 gets 头发's `fà` and
+ * 發 gets 发送's `fā` — which is the same evidence reaching the same character
+ * under the other script.
+ */
+export function countCorpusMass(sources, traditional) {
+    const mass = new Map();
+    const cast = (character, key, weight) => {
+        const byReading = mass.get(character) ?? new Map();
+        byReading.set(key, (byReading.get(key) ?? 0) + weight);
+        mass.set(character, byReading);
+    };
+    for (const [word, reading] of readingsByWord(sources)) {
+        const weight = sources.jieba.get(word)?.frequency ?? 0;
+        if (weight <= 0 || toCharacters(word).length < 2) {
+            continue;
+        }
+        const aligned = readAlignedReading(word, reading);
+        if (aligned === undefined) {
+            continue;
+        }
+        for (const read of aligned) {
+            // Punctuation is written and unread, and so votes for nothing.
+            if (read.syllable === undefined) {
+                continue;
+            }
+            // A syllable covering two characters — 儿化 — reads the first of them and
+            // says nothing about the second, which is the same rule the 繁體
+            // derivation applies to the same alignment.
+            const character = toCharacters(read.characters)[0];
+            /* c8 ignore next -- an aligned group always names at least one character */
+            if (character === undefined) {
+                continue;
+            }
+            const key = readingKey(read.syllable);
+            cast(character, key, weight);
+            const hant = traditional.convertCharacter(character, read.syllable);
+            if (hant !== character) {
+                cast(hant, key, weight);
+            }
+        }
+    }
+    return mass;
+}
+/**
+ * The mass each reading is ranked on, with a leading 轻声 held in place.
+ *
+ * A 轻声 the fields put ahead of every full-tone reading ranks on the mass of
+ * whichever reading carries the most, so the corpus can reorder the readings
+ * behind it but never take the default away from it.
+ *
+ * 轻声 is the unstressed reading — a 语气词, a suffix, the second half of a
+ * reduplication — and unstressed is precisely the use this mass cannot see. It
+ * counts a character only inside the words the dictionary holds, and a particle
+ * spends its life on the end of a sentence rather than inside a word: every
+ * occurrence of 吧 the corpus can reach is 酒吧, 网吧 or 吧台, so `bā` carries
+ * 1027 of it and the `ba` that ends four sentences in five carries 225. The
+ * sources are not silent about that reading — `kMandarin` names `ba` and
+ * `kHanyuPinlu` counted it 2073 times — they say it about the bare character,
+ * which is the one place a word cannot vote.
+ *
+ * Only a 轻声 the fields already lead with is held, and only against being
+ * displaced. Where the corpus reaches a neutral reading the fields rank lower
+ * it still promotes it — 蔔 takes 萝卜's `bo` over `bó` — and where they lead
+ * with a full tone it ranks the readings as it finds them: 哩 is `lī` in
+ * `kMandarin`, and the corpus is free to prefer the `lǐ` of 哩程 and 平方哩.
+ */
+function rankingMass(readings, massOf) {
+    const ranked = readings.map((syllable) => ({
+        syllable,
+        mass: massOf(syllable),
+    }));
+    const heaviest = Math.max(0, ...ranked.map((reading) => reading.mass));
+    for (const reading of ranked) {
+        if (reading.syllable.tone !== NEUTRAL_TONE) {
+            break;
+        }
+        reading.mass = heaviest;
+    }
+    return ranked;
+}
+/**
+ * Re-rank one character's readings by how much corpus each carries.
+ *
+ * The sort is stable and the mass of an unseen reading is zero, so a character
+ * no word covers keeps Unihan's order exactly, and so does the tail of readings
+ * below the ones words attest. The corpus only ever reorders what it has
+ * evidence about, and never past a leading 轻声 — see {@link rankingMass}.
+ *
+ * **Only where `kHanyuPinlu` did not already rank the character**, which is the
+ * whole of what this instrument is for. `kMandarin` names one reading and orders
+ * nothing, so where it decides the default there is nothing behind it, and that
+ * is what the corpus supplies. Where `kHanyuPinlu` lists several readings there
+ * already is a corpus ranking, and it is the better one: **this mass counts a
+ * character only inside words, so a reading used mostly as a bare character is
+ * invisible to it.** 呢 is 呢子 and 呢绒 in the dictionary and 呢 on the end of a
+ * question everywhere else; 都 is 首都 and 都市 against a bare 都 that is simply
+ * `dōu`; 将, 与, 处, 喝, 蒙 and 吗 are the same shape. `kHanyuPinlu` counts
+ * occurrences rather than words and has seen exactly what this cannot.
+ *
+ * Measured over CPP, that division is the whole difference between the rule
+ * paying and not: where `kHanyuPinlu` is silent or alone the mass fixes 438 and
+ * breaks 256, and where it ranks the character it fixes 201 and breaks 263.
+ */
+export function rankByCorpusMass(character, readings, fields, mass) {
+    const byReading = mass.get(character);
+    if (byReading === undefined || isFrequencyRanked(fields)) {
+        return readings;
+    }
+    const ranking = rankingMass(readings, (syllable) => byReading.get(readingKey(syllable)) ?? 0);
+    return ranking
+        .toSorted((left, right) => right.mass - left.mass)
+        .map((ranked) => ranked.syllable);
+}
+//# sourceMappingURL=corpus-mass.js.map

package/dist/dictionary/corpus-mass.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"corpus-mass.js","sourceRoot":"","sources":["../../src/dictionary/corpus-mass.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAGvD,OAAO,EAAE,iBAAiB,EAAqB,MAAM,sBAAsB,CAAC;AAE5E,OAAO,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAC;AAC/C,OAAO,EAAE,kBAAkB,EAAE,MAAM,cAAc,CAAC;AAoBlD;;GAEG;AACH,SAAS,UAAU,CAAC,QAAkB;IACpC,OAAO,GAAG,QAAQ,CAAC,OAAO,IAAI,QAAQ,CAAC,KAAK,IAAI,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,EAAE,CAAC;AAC1E,CAAC;AAED;;;;;;;;GAQG;AACH,SAAS,cAAc,CACrB,OAAoB;IAEpB,MAAM,QAAQ,GAAG,IAAI,GAAG,EAA6B,CAAC;IACtD,KAAK,MAAM,KAAK,IAAI,OAAO,CAAC,MAAM,EAAE,CAAC;QACnC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,UAAU,CAAC,EAAE,CAAC;YACpC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,UAAU,EAAE,KAAK,CAAC,QAAQ,CAAC,CAAC;QACjD,CAAC;IACH,CAAC;IACD,KAAK,MAAM,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,OAAO,CAAC,MAAM,EAAE,CAAC;QAC7C,QAAQ,CAAC,GAAG,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC;IAC9B,CAAC;IACD,OAAO,QAAQ,CAAC;AAClB,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,MAAM,UAAU,eAAe,CAC7B,OAAoB,EACpB,WAA6B;IAE7B,MAAM,IAAI,GAAG,IAAI,GAAG,EAA+B,CAAC;IAEpD,MAAM,IAAI,GAAG,CAAC,SAAiB,EAAE,GAAW,EAAE,MAAc,EAAQ,EAAE;QACpE,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,IAAI,GAAG,EAAkB,CAAC;QACnE,SAAS,CAAC,GAAG,CAAC,GAAG,EAAE,CAAC,SAAS,CAAC,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,GAAG,MAAM,CAAC,CAAC;QACvD,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,SAAS,CAAC,CAAC;IACjC,CAAC,CAAC;IAEF,KAAK,MAAM,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,cAAc,CAAC,OAAO,CAAC,EAAE,CAAC;QACtD,MAAM,MAAM,GAAG,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,SAAS,IAAI,CAAC,CAAC;QACvD,IAAI,MAAM,IAAI,CAAC,IAAI,YAAY,CAAC,IAAI,CAAC,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;YACjD,SAAS;QACX,CAAC;QACD,MAAM,OAAO,GAAG,kBAAkB,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC;QAClD,IAAI,OAAO,KAAK,SAAS,EAAE,CAAC;YAC1B,SAAS;QACX,CAAC;QACD,KAAK,MAAM,IAAI,IAAI,OAAO,EAAE,CAAC;YAC3B,+DAA+D;YAC/D,IAAI,IAAI,CAAC,QAAQ,KAAK,SAAS,EAAE,CAAC;gBAChC,SAAS;YACX,CAAC;YACD,wEAAwE;YACxE,+DAA+D;YAC/D,4CAA4C;YAC5C,MAAM,SAAS,GAAG,YAAY,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC,CAAC;YACnD,4EAA4E;YAC5E,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;gBAC5B,SAAS;YACX,CAAC;YACD,MAAM,GAAG,GAAG,UAAU,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC;YACtC,IAAI,CAAC,SAAS,EAAE,GAAG,EAAE,MAAM,CAAC,CAAC;YAC7B,MAAM,IAAI,GAAG,WAAW,CAAC,gBAAgB,CAAC,SAAS,EAAE,IAAI,CAAC,QAAQ,CAAC,CAAC;YACpE,IAAI,IAAI,KAAK,SAAS,EAAE,CAAC;gBACvB,IAAI,CAAC,IAAI,EAAE,GAAG,EAAE,MAAM,CAAC,CAAC;YAC1B,CAAC;QACH,CAAC;IACH,CAAC;IAED,OAAO,IAAI,CAAC;AACd,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,SAAS,WAAW,CAClB,QAA6B,EAC7B,MAAsC;IAEtC,MAAM,MAAM,GAAG,QAAQ,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC;QACzC,QAAQ;QACR,IAAI,EAAE,MAAM,CAAC,QAAQ,CAAC;KACvB,CAAC,CAAC,CAAC;IACJ,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC;IACvE,KAAK,MAAM,OAAO,IAAI,MAAM,EAAE,CAAC;QAC7B,IAAI,OAAO,CAAC,QAAQ,CAAC,IAAI,KAAK,YAAY,EAAE,CAAC;YAC3C,MAAM;QACR,CAAC;QACD,OAAO,CAAC,IAAI,GAAG,QAAQ,CAAC;IAC1B,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,MAAM,UAAU,gBAAgB,CAC9B,SAAiB,EACjB,QAA6B,EAC7B,MAAoD,EACpD,IAAgB;IAEhB,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,SAAS,CAAC,CAAC;IACtC,IAAI,SAAS,KAAK,SAAS,IAAI,iBAAiB,CAAC,MAAM,CAAC,EAAE,CAAC;QACzD,OAAO,QAAQ,CAAC;IAClB,CAAC;IACD,MAAM,OAAO,GAAG,WAAW,CACzB,QAAQ,EACR,CAAC,QAAQ,EAAE,EAAE,CAAC,SAAS,CAAC,GAAG,CAAC,UAAU,CAAC,QAAQ,CAAC,CAAC,IAAI,CAAC,CACvD,CAAC;IACF,OAAO,OAAO;SACX,QAAQ,CAAC,CAAC,IAAI,EAAE,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC;SACjD,GAAG,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;AACtC,CAAC"}

package/dist/dictionary/dictionary.d.ts

@@ -1,5 +1,28 @@
 import type { Syllable } from "../syllable/syllable.js";
 import { type DictionaryArtifact } from "./artifact.js";
+/**
+ * Every key's reading, in key order, at the string level.
+ *
+ * The seam a second index over the same artifact is built through. A reading is
+ * `yin2 hang2` on the way in and on the way out, and no {@link Syllable} is
+ * constructed anywhere along it — which is what makes a pass over all 723,147
+ * keys affordable at load time rather than a decode of the whole dictionary.
+ *
+ * Handed out as a cursor rather than as an array because the array is 39 MB on
+ * the full tier. What the cursor holds instead is the character defaults a
+ * derived reading is assembled from, and dropping it releases them.
+ */
+export interface DictionaryReadings {
+    /** How many positions there are, which is the dictionary's own size. */
+    readonly size: number;
+    /**
+     * The reading at a position, in the artifact's tone-numbered spelling.
+     *
+     * Empty where the position is out of range, or where a character the word is
+     * written with is not itself a key.
+     */
+    readonly readingAt: (at: number) => string;
+}
 /**
  * What the dictionary knows about one word.
  */
@@ -60,6 +83,33 @@ export declare class Dictionary {
      * How many words the dictionary holds, counting each script's key.
      */
     get size(): number;
+    /**
+     * The word at a position, or the empty string where there is none.
+     *
+     * Positions are what a second index over the dictionary stores, since a
+     * position is one number where a word is a string — the reverse index in
+     * `src/search/` is built on that. The same position indexes
+     * {@link Dictionary.frequencyAt} and {@link DictionaryReadings.readingAt}.
+     */
+    wordAt(at: number): string;
+    /**
+     * The frequency bucket of the word at a position, 0 rarest to 15 commonest.
+     *
+     * The bucket rather than {@link WordEntry.cost}, because ranking a candidate
+     * list wants the sixteen-value scale a counting sort runs on — and because
+     * reading it costs no decode, where asking for the entry would decode every
+     * candidate to learn one number. See {@link FrequencyTable.bucketOf}.
+     */
+    frequencyAt(at: number): number;
+    /**
+     * A cursor over every key's reading, in key order.
+     *
+     * Two passes, mirroring `readArtifact`: the character defaults are gathered
+     * first, because 83.25% of the full tier's keys store no reading and mean
+     * "the characters' defaults, in order". Holding the cursor holds those
+     * defaults, so build what needs it, use it, and let it go.
+     */
+    readingsInOrder(): DictionaryReadings;
     /**
      * Whether any word begins with this text. An exact match counts.
      *
@@ -81,6 +131,12 @@ export declare class Dictionary {
      *
      * The lattice needs these for its single-character fallback edges: 行 has to
      * offer `xíng`, `háng` and `héng` at a position no word covers.
+     *
+     * The alternates are read off the canonical form's line, the same form
+     * {@link Dictionary.lookup} found the entry under. Searching the index for
+     * the raw character instead would land on the insertion point for a Hong Kong
+     * glyph form that is not itself a key — 裏 sits between whatever keys
+     * surround it — and take some unrelated entry's alternates.
      */
     readingsOf(character: string): readonly (readonly Syllable[])[];
 }

package/dist/dictionary/dictionary.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"dictionary.d.ts","sourceRoot":"","sources":["../../src/dictionary/dictionary.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AACxD,OAAO,EAAiB,KAAK,kBAAkB,EAAE,MAAM,eAAe,CAAC;AAyBvE;;GAEG;AACH,MAAM,WAAW,SAAS;IACxB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,mBAAmB;IACnB,QAAQ,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,CAAC;IACtC,kDAAkD;IAClD,QAAQ,CAAC,aAAa,CAAC,EAAE,SAAS,QAAQ,EAAE,CAAC;IAC7C,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;IAC9B,QAAQ,CAAC,YAAY,EAAE,OAAO,CAAC;IAC/B;;;;;;;OAOG;IACH,QAAQ,CAAC,cAAc,CAAC,EAAE,SAAS,MAAM,EAAE,CAAC;IAC5C;;;;OAIG;IACH,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;CACvB;AAED;;;;;;;;;;;;;;;;;;GAkBG;AACH,qBAAa,UAAU;;IACrB;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,QAAQ,EAAE,kBAAkB,GAAG,UAAU,CAEpD;IAQD;;OAEG;IACH,OAAO,eAiBN;IA0DD;;OAEG;IACH,IAAI,IAAI,IAAI,MAAM,CAEjB;IAED;;;;;OAKG;IACH,SAAS,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAE/B;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAI,EAAE,MAAM,GAAG,SAAS,GAAG,SAAS,CAa1C;IAED;;;;;OAKG;IACH,UAAU,CAAC,SAAS,EAAE,MAAM,GAAG,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,CAW9D;CACF"}
+{"version":3,"file":"dictionary.d.ts","sourceRoot":"","sources":["../../src/dictionary/dictionary.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AACxD,OAAO,EAAiB,KAAK,kBAAkB,EAAE,MAAM,eAAe,CAAC;AA0EvE;;;;;;;;;;;GAWG;AACH,MAAM,WAAW,kBAAkB;IACjC,wEAAwE;IACxE,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB;;;;;OAKG;IACH,QAAQ,CAAC,SAAS,EAAE,CAAC,EAAE,EAAE,MAAM,KAAK,MAAM,CAAC;CAC5C;AAED;;GAEG;AACH,MAAM,WAAW,SAAS;IACxB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,mBAAmB;IACnB,QAAQ,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,CAAC;IACtC,kDAAkD;IAClD,QAAQ,CAAC,aAAa,CAAC,EAAE,SAAS,QAAQ,EAAE,CAAC;IAC7C,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;IAC9B,QAAQ,CAAC,YAAY,EAAE,OAAO,CAAC;IAC/B;;;;;;;OAOG;IACH,QAAQ,CAAC,cAAc,CAAC,EAAE,SAAS,MAAM,EAAE,CAAC;IAC5C;;;;OAIG;IACH,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;CACvB;AAED;;;;;;;;;;;;;;;;;;GAkBG;AACH,qBAAa,UAAU;;IACrB;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,QAAQ,EAAE,kBAAkB,GAAG,UAAU,CAEpD;IAQD;;OAEG;IACH,OAAO,eAiBN;IA+ED;;OAEG;IACH,IAAI,IAAI,IAAI,MAAM,CAEjB;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,EAAE,EAAE,MAAM,GAAG,MAAM,CAEzB;IAED;;;;;;;OAOG;IACH,WAAW,CAAC,EAAE,EAAE,MAAM,GAAG,MAAM,CAE9B;IAED;;;;;;;OAOG;IACH,eAAe,IAAI,kBAAkB,CAmBpC;IAED;;;;;OAKG;IACH,SAAS,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAE/B;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAI,EAAE,MAAM,GAAG,SAAS,GAAG,SAAS,CAa1C;IAED;;;;;;;;;;;OAWG;IACH,UAAU,CAAC,SAAS,EAAE,MAAM,GAAG,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,CAW9D;CACF"}

package/dist/dictionary/dictionary.js

@@ -19,6 +19,49 @@ const PROPER_NOUN_FLAG = "p";
  * for each doubling of the input, 39µs to 607µs.
  */
 const NAME_BOUNDARIES = /^p(?<at>\d+(?:\.\d+)*)$/u;
+/**
+ * The lowest code unit that begins a surrogate pair.
+ */
+const HIGH_SURROGATE = 0x1_00_00;
+/**
+ * Whether a key is written with exactly one character, without allocating.
+ *
+ * Asked of every key in the dictionary when {@link Dictionary.readingsInOrder}
+ * gathers the character defaults, which is where splitting each key into
+ * characters would cost an array per key for an answer that is a length check.
+ */
+function isOneCharacter(key) {
+    if (key.length === 1) {
+        return true;
+    }
+    return key.length === 2 && (key.codePointAt(0) ?? 0) >= HIGH_SURROGATE;
+}
+/**
+ * A word's reading, joined from its characters' default readings.
+ *
+ * Walked by code point rather than split into characters first, for the same
+ * reason as {@link isOneCharacter}: this runs over every key in the dictionary
+ * three times during a reverse index build, and the array `toCharacters`
+ * allocates per key is 20% of it.
+ *
+ * Returns the empty string where any character has no default, which is what a
+ * tier missing a character would look like. No tier does: `selectTier` never
+ * drops a character some word is written with.
+ */
+function derivedReading(key, defaults) {
+    let joined = "";
+    let at = 0;
+    while (at < key.length) {
+        const width = (key.codePointAt(at) ?? 0) >= HIGH_SURROGATE ? 2 : 1;
+        const value = defaults.get(key.slice(at, at + width));
+        if (value === undefined) {
+            return "";
+        }
+        joined = joined === "" ? value : `${joined} ${value}`;
+        at += width;
+    }
+    return joined;
+}
 /**
  * A compiled dictionary, queried without being parsed.
  *
@@ -114,12 +157,78 @@ export class Dictionary {
             cost: this.#frequencies.costOf(at),
         };
     }
+    /**
+     * The reading column of a line, without slicing or splitting the rest of it.
+     *
+     * Deriving a word's reading needs its characters' readings and nothing else
+     * about them, and this is asked of every line in the dictionary during a
+     * reverse index build, so the other four columns are never touched.
+     */
+    #readingColumn(at) {
+        const start = this.#lineStarts[at];
+        const nextStart = this.#lineStarts[at + 1];
+        if (start === undefined || nextStart === undefined) {
+            return "";
+        }
+        const end = nextStart - 1;
+        const column = this.#entries.indexOf(COLUMN, start);
+        return this.#entries.slice(start, column === -1 || column > end ? end : column);
+    }
     /**
      * How many words the dictionary holds, counting each script's key.
      */
     get size() {
         return this.#index.size;
     }
+    /**
+     * The word at a position, or the empty string where there is none.
+     *
+     * Positions are what a second index over the dictionary stores, since a
+     * position is one number where a word is a string — the reverse index in
+     * `src/search/` is built on that. The same position indexes
+     * {@link Dictionary.frequencyAt} and {@link DictionaryReadings.readingAt}.
+     */
+    wordAt(at) {
+        return this.#index.keyAt(at);
+    }
+    /**
+     * The frequency bucket of the word at a position, 0 rarest to 15 commonest.
+     *
+     * The bucket rather than {@link WordEntry.cost}, because ranking a candidate
+     * list wants the sixteen-value scale a counting sort runs on — and because
+     * reading it costs no decode, where asking for the entry would decode every
+     * candidate to learn one number. See {@link FrequencyTable.bucketOf}.
+     */
+    frequencyAt(at) {
+        return this.#frequencies.bucketOf(at);
+    }
+    /**
+     * A cursor over every key's reading, in key order.
+     *
+     * Two passes, mirroring `readArtifact`: the character defaults are gathered
+     * first, because 83.25% of the full tier's keys store no reading and mean
+     * "the characters' defaults, in order". Holding the cursor holds those
+     * defaults, so build what needs it, use it, and let it go.
+     */
+    readingsInOrder() {
+        const defaults = new Map();
+        for (let at = 0; at < this.#index.size; at++) {
+            const key = this.#index.keyAt(at);
+            const reading = isOneCharacter(key) ? this.#readingColumn(at) : "";
+            if (reading !== "") {
+                defaults.set(key, reading);
+            }
+        }
+        return {
+            size: this.#index.size,
+            readingAt: (at) => {
+                const stored = this.#readingColumn(at);
+                return stored === ""
+                    ? derivedReading(this.#index.keyAt(at), defaults)
+                    : stored;
+            },
+        };
+    }
     /**
      * Whether any word begins with this text. An exact match counts.
      *
@@ -156,13 +265,19 @@ export class Dictionary {
      *
      * The lattice needs these for its single-character fallback edges: 行 has to
      * offer `xíng`, `háng` and `héng` at a position no word covers.
+     *
+     * The alternates are read off the canonical form's line, the same form
+     * {@link Dictionary.lookup} found the entry under. Searching the index for
+     * the raw character instead would land on the insertion point for a Hong Kong
+     * glyph form that is not itself a key — 裏 sits between whatever keys
+     * surround it — and take some unrelated entry's alternates.
      */
     readingsOf(character) {
         const entry = this.lookup(character);
         if (entry === undefined) {
             return [];
         }
-        const found = this.#index.lookup(character);
+        const found = this.#index.lookup(toCanonicalGlyphs(character));
         const alternates = (this.#columns(found.index)[4] ?? "")
             .split(ALTERNATE)
             .map((text) => decodeReading(text))

package/dist/dictionary/dictionary.js.map

@@ -1 +1 @@
-{"version":3,"file":"dictionary.js","sourceRoot":"","sources":["../../src/dictionary/dictionary.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAE,iBAAiB,EAAE,MAAM,qBAAqB,CAAC;AAExD,OAAO,EAAE,aAAa,EAA2B,MAAM,eAAe,CAAC;AACvE,OAAO,EAAE,cAAc,EAAE,MAAM,sBAAsB,CAAC;AACtD,OAAO,EAAE,QAAQ,EAAE,MAAM,gBAAgB,CAAC;AAE1C;;GAEG;AACH,MAAM,MAAM,GAAG,IAAI,CAAC;AAEpB,MAAM,IAAI,GAAG,IAAI,CAAC;AAElB,MAAM,SAAS,GAAG,GAAG,CAAC;AAEtB,MAAM,gBAAgB,GAAG,GAAG,CAAC;AAE7B;;;;;;;GAOG;AACH,MAAM,eAAe,GAAG,0BAA0B,CAAC;AA8BnD;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,OAAO,UAAU;IACrB;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,QAA4B;QACtC,OAAO,IAAI,UAAU,CAAC,QAAQ,CAAC,CAAC;IAClC,CAAC;IAEQ,MAAM,CAAW;IACjB,QAAQ,CAAS;IACjB,WAAW,CAAc;IACzB,YAAY,CAAiB;IAC7B,QAAQ,GAAG,IAAI,GAAG,EAAqB,CAAC;IAEjD;;OAEG;IACH,YAAoB,QAA4B;QAC9C,IAAI,CAAC,MAAM,GAAG,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;QAC3C,IAAI,CAAC,QAAQ,GAAG,QAAQ,CAAC,OAAO,CAAC;QAEjC,MAAM,MAAM,GAAa,CAAC,CAAC,CAAC,CAAC;QAC7B,IAAI,SAAS,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;QAC5C,OAAO,SAAS,KAAK,CAAC,CAAC,EAAE,CAAC;YACxB,MAAM,CAAC,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC;YAC3B,SAAS,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC;QACzD,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QACtC,IAAI,CAAC,WAAW,GAAG,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAE5C,IAAI,CAAC,YAAY,GAAG,cAAc,CAAC,IAAI,CACrC,QAAQ,CAAC,WAAW,EACpB,IAAI,CAAC,MAAM,CAAC,IAAI,CACjB,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,QAAQ,CAAC,EAAU;QACjB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAC3C,2EAA2E;QAC3E,IAAI,KAAK,KAAK,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YACnD,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IACjE,CAAC;IAED;;;;;OAKG;IACH,cAAc,CAAC,IAAY;QACzB,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;QACvC,OAAO,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IACzE,CAAC;IAED;;OAEG;IACH,OAAO,CAAC,IAAY,EAAE,EAAU;QAC9B,MAAM,CAAC,OAAO,GAAG,EAAE,EAAE,MAAM,GAAG,EAAE,EAAE,YAAY,GAAG,EAAE,EAAE,KAAK,GAAG,EAAE,CAAC,GAC9D,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;QAEpB,uEAAuE;QACvE,4DAA4D;QAC5D,MAAM,OAAO,GACX,OAAO,KAAK,EAAE;YACZ,CAAC,CAAC,YAAY,CAAC,IAAI,CAAC;iBACf,GAAG,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,IAAI,CAAC,cAAc,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC;iBACxD,IAAI,CAAC,GAAG,CAAC;YACd,CAAC,CAAC,OAAO,CAAC;QACd,MAAM,OAAO,GAAG,aAAa,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;QAC7C,MAAM,aAAa,GAAG,aAAa,CAAC,MAAM,CAAC,CAAC;QAC5C,MAAM,MAAM,GAAG,eAAe,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,CAAC,IAAI,CAAC,CAAC;QAC3D,MAAM,cAAc,GAClB,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;QAE5D,OAAO;YACL,IAAI;YACJ,OAAO,EAAE,OAAO;YAChB,GAAG,CAAC,aAAa,KAAK,SAAS,IAAI,EAAE,aAAa,EAAE,CAAC;YACrD,YAAY;YACZ,YAAY,EAAE,KAAK,CAAC,QAAQ,CAAC,gBAAgB,CAAC;YAC9C,GAAG,CAAC,cAAc,CAAC,MAAM,GAAG,CAAC,IAAI,EAAE,cAAc,EAAE,CAAC;YACpD,IAAI,EAAE,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,EAAE,CAAC;SACnC,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;IAC1B,CAAC;IAED;;;;;OAKG;IACH,SAAS,CAAC,IAAY;QACpB,OAAO,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC;IACxD,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAY;QACjB,MAAM,SAAS,GAAG,iBAAiB,CAAC,IAAI,CAAC,CAAC;QAC1C,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QAC5C,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,CAAC;YACjB,OAAO,SAAS,CAAC;QACnB,CAAC;QACD,MAAM,IAAI,GAAG,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;QAC5C,IAAI,IAAI,KAAK,SAAS,EAAE,CAAC;YACvB,OAAO,IAAI,CAAC;QACd,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC;QACnD,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;QACtC,OAAO,KAAK,CAAC;IACf,CAAC;IAED;;;;;OAKG;IACH,UAAU,CAAC,SAAiB;QAC1B,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QACrC,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;YACxB,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QAC5C,MAAM,UAAU,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;aACrD,KAAK,CAAC,SAAS,CAAC;aAChB,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,aAAa,CAAC,IAAI,CAAC,CAAC;aAClC,MAAM,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,KAAK,SAAS,CAAC,CAAC;QAC9C,OAAO,CAAC,KAAK,CAAC,OAAO,EAAE,GAAG,UAAU,CAAC,CAAC;IACxC,CAAC;CACF"}
+{"version":3,"file":"dictionary.js","sourceRoot":"","sources":["../../src/dictionary/dictionary.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAE,iBAAiB,EAAE,MAAM,qBAAqB,CAAC;AAExD,OAAO,EAAE,aAAa,EAA2B,MAAM,eAAe,CAAC;AACvE,OAAO,EAAE,cAAc,EAAE,MAAM,sBAAsB,CAAC;AACtD,OAAO,EAAE,QAAQ,EAAE,MAAM,gBAAgB,CAAC;AAE1C;;GAEG;AACH,MAAM,MAAM,GAAG,IAAI,CAAC;AAEpB,MAAM,IAAI,GAAG,IAAI,CAAC;AAElB,MAAM,SAAS,GAAG,GAAG,CAAC;AAEtB,MAAM,gBAAgB,GAAG,GAAG,CAAC;AAE7B;;;;;;;GAOG;AACH,MAAM,eAAe,GAAG,0BAA0B,CAAC;AAEnD;;GAEG;AACH,MAAM,cAAc,GAAG,SAAS,CAAC;AAEjC;;;;;;GAMG;AACH,SAAS,cAAc,CAAC,GAAW;IACjC,IAAI,GAAG,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACrB,OAAO,IAAI,CAAC;IACd,CAAC;IACD,OAAO,GAAG,CAAC,MAAM,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,cAAc,CAAC;AACzE,CAAC;AAED;;;;;;;;;;;GAWG;AACH,SAAS,cAAc,CACrB,GAAW,EACX,QAAqC;IAErC,IAAI,MAAM,GAAG,EAAE,CAAC;IAChB,IAAI,EAAE,GAAG,CAAC,CAAC;IACX,OAAO,EAAE,GAAG,GAAG,CAAC,MAAM,EAAE,CAAC;QACvB,MAAM,KAAK,GAAG,CAAC,GAAG,CAAC,WAAW,CAAC,EAAE,CAAC,IAAI,CAAC,CAAC,IAAI,cAAc,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;QACnE,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,GAAG,KAAK,CAAC,CAAC,CAAC;QACtD,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;YACxB,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,GAAG,MAAM,KAAK,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,GAAG,MAAM,IAAI,KAAK,EAAE,CAAC;QACtD,EAAE,IAAI,KAAK,CAAC;IACd,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC;AAsDD;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,OAAO,UAAU;IACrB;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,QAA4B;QACtC,OAAO,IAAI,UAAU,CAAC,QAAQ,CAAC,CAAC;IAClC,CAAC;IAEQ,MAAM,CAAW;IACjB,QAAQ,CAAS;IACjB,WAAW,CAAc;IACzB,YAAY,CAAiB;IAC7B,QAAQ,GAAG,IAAI,GAAG,EAAqB,CAAC;IAEjD;;OAEG;IACH,YAAoB,QAA4B;QAC9C,IAAI,CAAC,MAAM,GAAG,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;QAC3C,IAAI,CAAC,QAAQ,GAAG,QAAQ,CAAC,OAAO,CAAC;QAEjC,MAAM,MAAM,GAAa,CAAC,CAAC,CAAC,CAAC;QAC7B,IAAI,SAAS,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;QAC5C,OAAO,SAAS,KAAK,CAAC,CAAC,EAAE,CAAC;YACxB,MAAM,CAAC,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC;YAC3B,SAAS,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC;QACzD,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QACtC,IAAI,CAAC,WAAW,GAAG,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAE5C,IAAI,CAAC,YAAY,GAAG,cAAc,CAAC,IAAI,CACrC,QAAQ,CAAC,WAAW,EACpB,IAAI,CAAC,MAAM,CAAC,IAAI,CACjB,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,QAAQ,CAAC,EAAU;QACjB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAC3C,2EAA2E;QAC3E,IAAI,KAAK,KAAK,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YACnD,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IACjE,CAAC;IAED;;;;;OAKG;IACH,cAAc,CAAC,IAAY;QACzB,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;QACvC,OAAO,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IACzE,CAAC;IAED;;OAEG;IACH,OAAO,CAAC,IAAY,EAAE,EAAU;QAC9B,MAAM,CAAC,OAAO,GAAG,EAAE,EAAE,MAAM,GAAG,EAAE,EAAE,YAAY,GAAG,EAAE,EAAE,KAAK,GAAG,EAAE,CAAC,GAC9D,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;QAEpB,uEAAuE;QACvE,4DAA4D;QAC5D,MAAM,OAAO,GACX,OAAO,KAAK,EAAE;YACZ,CAAC,CAAC,YAAY,CAAC,IAAI,CAAC;iBACf,GAAG,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,IAAI,CAAC,cAAc,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC;iBACxD,IAAI,CAAC,GAAG,CAAC;YACd,CAAC,CAAC,OAAO,CAAC;QACd,MAAM,OAAO,GAAG,aAAa,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;QAC7C,MAAM,aAAa,GAAG,aAAa,CAAC,MAAM,CAAC,CAAC;QAC5C,MAAM,MAAM,GAAG,eAAe,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,CAAC,IAAI,CAAC,CAAC;QAC3D,MAAM,cAAc,GAClB,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;QAE5D,OAAO;YACL,IAAI;YACJ,OAAO,EAAE,OAAO;YAChB,GAAG,CAAC,aAAa,KAAK,SAAS,IAAI,EAAE,aAAa,EAAE,CAAC;YACrD,YAAY;YACZ,YAAY,EAAE,KAAK,CAAC,QAAQ,CAAC,gBAAgB,CAAC;YAC9C,GAAG,CAAC,cAAc,CAAC,MAAM,GAAG,CAAC,IAAI,EAAE,cAAc,EAAE,CAAC;YACpD,IAAI,EAAE,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,EAAE,CAAC;SACnC,CAAC;IACJ,CAAC;IAED;;;;;;OAMG;IACH,cAAc,CAAC,EAAU;QACvB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAC3C,IAAI,KAAK,KAAK,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YACnD,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,GAAG,GAAG,SAAS,GAAG,CAAC,CAAC;QAC1B,MAAM,MAAM,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;QACpD,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,CACxB,KAAK,EACL,MAAM,KAAK,CAAC,CAAC,IAAI,MAAM,GAAG,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,CAC7C,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;IAC1B,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,EAAU;QACf,OAAO,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IAC/B,CAAC;IAED;;;;;;;OAOG;IACH,WAAW,CAAC,EAAU;QACpB,OAAO,IAAI,CAAC,YAAY,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;IACxC,CAAC;IAED;;;;;;;OAOG;IACH,eAAe;QACb,MAAM,QAAQ,GAAG,IAAI,GAAG,EAAkB,CAAC;QAC3C,KAAK,IAAI,EAAE,GAAG,CAAC,EAAE,EAAE,GAAG,IAAI,CAAC,MAAM,CAAC,IAAI,EAAE,EAAE,EAAE,EAAE,CAAC;YAC7C,MAAM,GAAG,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;YAClC,MAAM,OAAO,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,cAAc,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACnE,IAAI,OAAO,KAAK,EAAE,EAAE,CAAC;gBACnB,QAAQ,CAAC,GAAG,CAAC,GAAG,EAAE,OAAO,CAAC,CAAC;YAC7B,CAAC;QACH,CAAC;QAED,OAAO;YACL,IAAI,EAAE,IAAI,CAAC,MAAM,CAAC,IAAI;YACtB,SAAS,EAAE,CAAC,EAAU,EAAU,EAAE;gBAChC,MAAM,MAAM,GAAG,IAAI,CAAC,cAAc,CAAC,EAAE,CAAC,CAAC;gBACvC,OAAO,MAAM,KAAK,EAAE;oBAClB,CAAC,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC;oBACjD,CAAC,CAAC,MAAM,CAAC;YACb,CAAC;SACF,CAAC;IACJ,CAAC;IAED;;;;;OAKG;IACH,SAAS,CAAC,IAAY;QACpB,OAAO,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC;IACxD,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAY;QACjB,MAAM,SAAS,GAAG,iBAAiB,CAAC,IAAI,CAAC,CAAC;QAC1C,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QAC5C,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,CAAC;YACjB,OAAO,SAAS,CAAC;QACnB,CAAC;QACD,MAAM,IAAI,GAAG,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;QAC5C,IAAI,IAAI,KAAK,SAAS,EAAE,CAAC;YACvB,OAAO,IAAI,CAAC;QACd,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC;QACnD,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;QACtC,OAAO,KAAK,CAAC;IACf,CAAC;IAED;;;;;;;;;;;OAWG;IACH,UAAU,CAAC,SAAiB;QAC1B,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QACrC,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;YACxB,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,iBAAiB,CAAC,SAAS,CAAC,CAAC,CAAC;QAC/D,MAAM,UAAU,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;aACrD,KAAK,CAAC,SAAS,CAAC;aAChB,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,aAAa,CAAC,IAAI,CAAC,CAAC;aAClC,MAAM,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,KAAK,SAAS,CAAC,CAAC;QAC9C,OAAO,CAAC,KAAK,CAAC,OAAO,EAAE,GAAG,UAAU,CAAC,CAAC;IACxC,CAAC;CACF"}

package/dist/dictionary/erhua.d.ts

@@ -16,6 +16,10 @@ export declare const NON_ERHUA_ER_WORDS: ReadonlySet<string>;
  * Whether a word's last character is 儿, in either script.
  */
 export declare function isErFinal(word: string): boolean;
+/**
+ * Whether a character is the 儿 that 儿化 folds into the syllable before it.
+ */
+export declare function isErCharacter(character: string): boolean;
 /**
  * Add the r suffix of 儿化 to a syllable.
  *

package/dist/dictionary/erhua.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"erhua.d.ts","sourceRoot":"","sources":["../../src/dictionary/erhua.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAYxD;;;;;;;;;;;GAWG;AACH,eAAO,MAAM,kBAAkB,EAAE,WAAW,CAAC,MAAM,CAiCjD,CAAC;AAEH;;GAEG;AACH,wBAAgB,SAAS,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAE/C;AAYD;;;;;;;;;;GAUG;AACH,wBAAgB,SAAS,CAAC,QAAQ,EAAE,QAAQ,GAAG,QAAQ,CAQtD;AAED;;;;;;;GAOG;AACH,wBAAgB,WAAW,CACzB,SAAS,EAAE,SAAS,QAAQ,EAAE,GAC7B,SAAS,QAAQ,EAAE,CAarB"}
+{"version":3,"file":"erhua.d.ts","sourceRoot":"","sources":["../../src/dictionary/erhua.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAYxD;;;;;;;;;;;GAWG;AACH,eAAO,MAAM,kBAAkB,EAAE,WAAW,CAAC,MAAM,CAiCjD,CAAC;AAEH;;GAEG;AACH,wBAAgB,SAAS,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAE/C;AAED;;GAEG;AACH,wBAAgB,aAAa,CAAC,SAAS,EAAE,MAAM,GAAG,OAAO,CAExD;AAYD;;;;;;;;;;GAUG;AACH,wBAAgB,SAAS,CAAC,QAAQ,EAAE,QAAQ,GAAG,QAAQ,CAQtD;AAED;;;;;;;GAOG;AACH,wBAAgB,WAAW,CACzB,SAAS,EAAE,SAAS,QAAQ,EAAE,GAC7B,SAAS,QAAQ,EAAE,CAarB"}

package/dist/dictionary/erhua.js

@@ -58,6 +58,12 @@ export const NON_ERHUA_ER_WORDS = new Set([
 export function isErFinal(word) {
     return word.length > 1 && (word.endsWith(ER) || word.endsWith(ER_HANT));
 }
+/**
+ * Whether a character is the 儿 that 儿化 folds into the syllable before it.
+ */
+export function isErCharacter(character) {
+    return character === ER || character === ER_HANT;
+}
 /**
  * Whether a syllable is a bare `er`, which is how the phrase corpus writes 儿化.
  *

package/dist/dictionary/erhua.js.map

@@ -1 +1 @@
-{"version":3,"file":"erhua.js","sourceRoot":"","sources":["../../src/dictionary/erhua.ts"],"names":[],"mappings":"AAEA;;GAEG;AACH,MAAM,EAAE,GAAG,GAAG,CAAC;AAEf;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,MAAM,kBAAkB,GAAwB,IAAI,GAAG,CAAC;IAC7D,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,MAAM;CACP,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,UAAU,SAAS,CAAC,IAAY;IACpC,OAAO,IAAI,CAAC,MAAM,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,IAAI,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC;AAC1E,CAAC;AAED;;;;;GAKG;AACH,SAAS,QAAQ,CAAC,QAAkB;IAClC,OAAO,QAAQ,CAAC,OAAO,KAAK,EAAE,IAAI,QAAQ,CAAC,KAAK,KAAK,IAAI,CAAC;AAC5D,CAAC;AAED;;;;;;;;;;GAUG;AACH,MAAM,UAAU,SAAS,CAAC,QAAkB;IAC1C,IACE,QAAQ,CAAC,OAAO,KAAK,EAAE;QACvB,CAAC,QAAQ,CAAC,KAAK,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,IAAI,CAAC,EACnD,CAAC;QACD,OAAO,EAAE,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,IAAI,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAI,EAAE,CAAC;IAC3D,CAAC;IACD,OAAO,EAAE,GAAG,QAAQ,EAAE,KAAK,EAAE,IAAI,EAAE,CAAC;AACtC,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,WAAW,CACzB,SAA8B;IAE9B,MAAM,IAAI,GAAG,SAAS,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IAC9B,MAAM,QAAQ,GAAG,SAAS,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IAClC,IACE,SAAS,CAAC,MAAM,GAAG,CAAC;QACpB,IAAI,KAAK,SAAS;QAClB,QAAQ,KAAK,SAAS;QACtB,CAAC,QAAQ,CAAC,IAAI,CAAC;QACf,QAAQ,CAAC,KAAK,KAAK,IAAI,EACvB,CAAC;QACD,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,CAAC,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,CAAC,QAAQ,CAAC,CAAC,CAAC;AAC1D,CAAC"}
+{"version":3,"file":"erhua.js","sourceRoot":"","sources":["../../src/dictionary/erhua.ts"],"names":[],"mappings":"AAEA;;GAEG;AACH,MAAM,EAAE,GAAG,GAAG,CAAC;AAEf;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,MAAM,kBAAkB,GAAwB,IAAI,GAAG,CAAC;IAC7D,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,MAAM;CACP,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,UAAU,SAAS,CAAC,IAAY;IACpC,OAAO,IAAI,CAAC,MAAM,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,IAAI,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC;AAC1E,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,aAAa,CAAC,SAAiB;IAC7C,OAAO,SAAS,KAAK,EAAE,IAAI,SAAS,KAAK,OAAO,CAAC;AACnD,CAAC;AAED;;;;;GAKG;AACH,SAAS,QAAQ,CAAC,QAAkB;IAClC,OAAO,QAAQ,CAAC,OAAO,KAAK,EAAE,IAAI,QAAQ,CAAC,KAAK,KAAK,IAAI,CAAC;AAC5D,CAAC;AAED;;;;;;;;;;GAUG;AACH,MAAM,UAAU,SAAS,CAAC,QAAkB;IAC1C,IACE,QAAQ,CAAC,OAAO,KAAK,EAAE;QACvB,CAAC,QAAQ,CAAC,KAAK,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,IAAI,CAAC,EACnD,CAAC;QACD,OAAO,EAAE,OAAO,EAAE,EAAE,EAAE,KAAK,EAAE,IAAI,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAI,EAAE,CAAC;IAC3D,CAAC;IACD,OAAO,EAAE,GAAG,QAAQ,EAAE,KAAK,EAAE,IAAI,EAAE,CAAC;AACtC,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,WAAW,CACzB,SAA8B;IAE9B,MAAM,IAAI,GAAG,SAAS,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IAC9B,MAAM,QAAQ,GAAG,SAAS,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IAClC,IACE,SAAS,CAAC,MAAM,GAAG,CAAC;QACpB,IAAI,KAAK,SAAS;QAClB,QAAQ,KAAK,SAAS;QACtB,CAAC,QAAQ,CAAC,IAAI,CAAC;QACf,QAAQ,CAAC,KAAK,KAAK,IAAI,EACvB,CAAC;QACD,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,CAAC,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,SAAS,CAAC,QAAQ,CAAC,CAAC,CAAC;AAC1D,CAAC"}

package/dist/dictionary/merge.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"merge.d.ts","sourceRoot":"","sources":["../../src/dictionary/merge.ts"],"names":[],"mappings":"AAEA,OAAO,EAAE,KAAK,WAAW,EAAoB,MAAM,sBAAsB,CAAC;AAC1E,OAAO,EAAmB,KAAK,UAAU,EAAE,MAAM,qBAAqB,CAAC;AACvE,OAAO,EAEL,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,sBAAsB,CAAC;AAG9B,OAAO,EAAE,KAAK,eAAe,EAAiB,MAAM,YAAY,CAAC;AAejE;;GAEG;AACH,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,cAAc,EAAE,WAAW,CAAC,MAAM,EAAE,cAAc,CAAC,CAAC;IAC7D,QAAQ,CAAC,cAAc,EAAE,cAAc,CAAC;IACxC,QAAQ,CAAC,MAAM,EAAE,WAAW,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,CAAC,CAAC;IACxD,QAAQ,CAAC,MAAM,EAAE,SAAS,WAAW,EAAE,CAAC;IACxC,QAAQ,CAAC,KAAK,EAAE,WAAW,CAAC,MAAM,EAAE,UAAU,CAAC,CAAC;CACjD;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,iEAAiE;IACjE,QAAQ,CAAC,sBAAsB,EAAE,MAAM,CAAC;IACxC,qEAAqE;IACrE,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;IAC9B,wEAAwE;IACxE,QAAQ,CAAC,kBAAkB,EAAE,MAAM,CAAC;IACpC,0DAA0D;IAC1D,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,4DAA4D;IAC5D,QAAQ,CAAC,gBAAgB,EAAE,MAAM,CAAC;IAClC,8DAA8D;IAC9D,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;IAChC,sEAAsE;IACtE,QAAQ,CAAC,sBAAsB,EAAE,MAAM,CAAC;IACxC;;OAEG;IACH,QAAQ,CAAC,gBAAgB,EAAE,MAAM,CAAC;IAClC,8DAA8D;IAC9D,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;IAChC,6DAA6D;IAC7D,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,yCAAyC;IACzC,QAAQ,CAAC,OAAO,EAAE,SAAS,eAAe,EAAE,CAAC;IAC7C,mEAAmE;IACnE,QAAQ,CAAC,QAAQ,EAAE,WAAW,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,CAAC,CAAC;IAC1D,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;CAC5B;AA2WD;;;;;;;;GAQG;AACH,wBAAgB,YAAY,CAAC,OAAO,EAAE,YAAY,GAAG,WAAW,CAoV/D"}
+{"version":3,"file":"merge.d.ts","sourceRoot":"","sources":["../../src/dictionary/merge.ts"],"names":[],"mappings":"AAEA,OAAO,EAAE,KAAK,WAAW,EAAoB,MAAM,sBAAsB,CAAC;AAC1E,OAAO,EAAmB,KAAK,UAAU,EAAE,MAAM,qBAAqB,CAAC;AACvE,OAAO,EAEL,KAAK,cAAc,EACnB,KAAK,cAAc,EACpB,MAAM,sBAAsB,CAAC;AAI9B,OAAO,EAAE,KAAK,eAAe,EAAiB,MAAM,YAAY,CAAC;AAejE;;GAEG;AACH,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,cAAc,EAAE,WAAW,CAAC,MAAM,EAAE,cAAc,CAAC,CAAC;IAC7D,QAAQ,CAAC,cAAc,EAAE,cAAc,CAAC;IACxC,QAAQ,CAAC,MAAM,EAAE,WAAW,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,CAAC,CAAC;IACxD,QAAQ,CAAC,MAAM,EAAE,SAAS,WAAW,EAAE,CAAC;IACxC,QAAQ,CAAC,KAAK,EAAE,WAAW,CAAC,MAAM,EAAE,UAAU,CAAC,CAAC;CACjD;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,QAAQ,CAAC,UAAU,EAAE,MAAM,CAAC;IAC5B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,iEAAiE;IACjE,QAAQ,CAAC,sBAAsB,EAAE,MAAM,CAAC;IACxC,qEAAqE;IACrE,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;IAC9B,wEAAwE;IACxE,QAAQ,CAAC,kBAAkB,EAAE,MAAM,CAAC;IACpC,0DAA0D;IAC1D,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,4DAA4D;IAC5D,QAAQ,CAAC,gBAAgB,EAAE,MAAM,CAAC;IAClC,8DAA8D;IAC9D,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;IAChC,sEAAsE;IACtE,QAAQ,CAAC,sBAAsB,EAAE,MAAM,CAAC;IACxC;;OAEG;IACH,QAAQ,CAAC,gBAAgB,EAAE,MAAM,CAAC;IAClC,8DAA8D;IAC9D,QAAQ,CAAC,cAAc,EAAE,MAAM,CAAC;IAChC,6DAA6D;IAC7D,QAAQ,CAAC,QAAQ,EAAE,MAAM,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,yCAAyC;IACzC,QAAQ,CAAC,OAAO,EAAE,SAAS,eAAe,EAAE,CAAC;IAC7C,mEAAmE;IACnE,QAAQ,CAAC,QAAQ,EAAE,WAAW,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,CAAC,CAAC;IAC1D,QAAQ,CAAC,KAAK,EAAE,UAAU,CAAC;CAC5B;AA2WD;;;;;;;;GAQG;AACH,wBAAgB,YAAY,CAAC,OAAO,EAAE,YAAY,GAAG,WAAW,CA8V/D"}

package/dist/dictionary/merge.js

@@ -4,6 +4,7 @@ import { nameBoundariesOf } from "../sources/cedict.js";
 import { isProperNounTag } from "../sources/jieba.js";
 import { FREQUENCY_FIELD, } from "../sources/unihan.js";
 import { NEUTRAL_TONE } from "../tone/tone.js";
+import { countCorpusMass, rankByCorpusMass } from "./corpus-mass.js";
 import { isSameReading } from "./entry.js";
 import { attachErhua, isErFinal, NON_ERHUA_ER_WORDS } from "./erhua.js";
 import { composeLocaleDeltas } from "./locale.js";
@@ -331,23 +332,30 @@ export function mergeSources(sources) {
     // 誰 is `shéi` under 谁 — so the 简体 index alone cannot say what a character
     // like that already reads in 普通话.
     const cedictByHant = indexCedict(cedict, (entry) => entry.traditional);
+    // ── 繁體 evidence, mined from CC-CEDICT's own pairings ──────
+    // Before the character defaults, because the corpus mass those are ranked by
+    // has to reach 繁體 characters through this table: jieba's corpus and the
+    // phrase corpus are both 简体, so nothing else would count a vote for 髮.
+    const pairings = [];
+    for (const entry of cedict) {
+        const aligned = readAlignedReading(entry.simplified, entry.readings);
+        pairings.push(...pairScripts(entry.simplified, entry.traditional, aligned));
+    }
+    const traditional = TraditionalTable.build(pairings, unihanVariants, unihanReadings);
     // ── Character defaults, which every later step leans on ────
+    // Unihan ranks the readings and the corpus re-ranks the ones it has seen: a
+    // character's default is whichever of its readings the dictionary's own words
+    // spend the most of jieba's corpus on, and Unihan's order decides the rest.
+    const mass = countCorpusMass({ phrase, cedict, jieba }, traditional);
     const defaults = new Map();
     for (const [character, readings] of unihanReadings) {
         const parsed = resolveFrequencyTones(readings)
             .map((reading) => characterSyllable(character, reading))
             .filter((syllable) => syllable !== undefined);
         if (parsed.length > 0) {
-            defaults.set(character, parsed);
+            defaults.set(character, rankByCorpusMass(character, parsed, readings.fields, mass));
         }
     }
-    // ── 繁體 evidence, mined from CC-CEDICT's own pairings ──────
-    const pairings = [];
-    for (const entry of cedict) {
-        const aligned = readAlignedReading(entry.simplified, entry.readings);
-        pairings.push(...pairScripts(entry.simplified, entry.traditional, aligned));
-    }
-    const traditional = TraditionalTable.build(pairings, unihanVariants, unihanReadings);
     const words = new Set([
         ...defaults.keys(),
         ...[...phrase.keys()].filter((word) => !isSpeltTraditionally(word, cedictByWord, cedictByHant)),