@kensio/pinyinjs 1.7.7 → 1.7.9

package/dist/dictionary/source.d.ts

@@ -2,6 +2,7 @@ import { type ScriptTables } from "../script/conversion.js";
 import type { DictionaryArtifact } from "./artifact.js";
 import { Dictionary } from "./dictionary.js";
 import type { Tier } from "./tiers.js";
+import { WordCounts } from "./word-counts.js";
 /**
  * Where a dictionary's files are read from.
  *
@@ -43,6 +44,27 @@ export declare function loadArtifact(source: DictionarySource, tier: Tier): Prom
  * Load a tier and wrap it in a {@link Dictionary}.
  */
 export declare function loadDictionary(source: DictionarySource, tier: Tier): Promise<Dictionary>;
+/**
+ * The file the raw corpus counts live in.
+ *
+ * `full` alone, and outside the tiers rather than one per tier: a caller
+ * ranking words wants the whole vocabulary ordered, and a ranking over part of
+ * it answers a different question.
+ */
+export declare const COUNTS_FILE = "full.counts";
+/**
+ * Load the raw corpus counts for the `full` tier.
+ *
+ * Separate from the dictionary for the same reason {@link loadScriptTables} is.
+ * Nothing on the decoding path reads counts, and the quantised frequencies the
+ * decoder does read already ship inside every tier. This file is 243 KB brotli
+ * against `full.freq`'s 148 KB, and only a caller that asks for it pays.
+ *
+ * The counts are positional over the `full` tier's keys, so pair them with a
+ * `full` {@link Dictionary} and nothing else. {@link WordCounts.size} and
+ * {@link Dictionary.size} agree where the two match.
+ */
+export declare function loadWordCounts(source: DictionarySource): Promise<WordCounts>;
 /**
  * The file the script conversion tables live in.
  *

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

@@ -1 +1 @@
-{"version":3,"file":"source.d.ts","sourceRoot":"","sources":["../../src/dictionary/source.ts"],"names":[],"mappings":"AAAA,OAAO,EAAoB,KAAK,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAC9E,OAAO,KAAK,EAAE,kBAAkB,EAAE,MAAM,eAAe,CAAC;AACxD,OAAO,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC7C,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAEvC;;;;;;;GAOG;AACH,MAAM,WAAW,gBAAgB;IAC/B,0CAA0C;IAC1C,QAAQ,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,OAAO,CAAC,MAAM,CAAC,CAAC;IACjD,2CAA2C;IAC3C,QAAQ,CAAC,KAAK,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,OAAO,CAAC,UAAU,CAAC,CAAC;CACvD;AAED;;GAEG;AACH,wBAAgB,SAAS,CAAC,IAAI,EAAE,IAAI,GAAG;IACrC,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;CAC9B,CAMA;AAED;;;;;;GAMG;AACH,wBAAgB,WAAW,CAAC,OAAO,EAAE,MAAM,GAAG,gBAAgB,CAqB7D;AAED;;;;;GAKG;AACH,wBAAsB,YAAY,CAChC,MAAM,EAAE,gBAAgB,EACxB,IAAI,EAAE,IAAI,GACT,OAAO,CAAC,kBAAkB,CAAC,CAQ7B;AAED;;GAEG;AACH,wBAAsB,cAAc,CAClC,MAAM,EAAE,gBAAgB,EACxB,IAAI,EAAE,IAAI,GACT,OAAO,CAAC,UAAU,CAAC,CAErB;AAED;;;;;;GAMG;AACH,eAAO,MAAM,WAAW,eAAe,CAAC;AAExC;;;;;;;GAOG;AACH,wBAAsB,gBAAgB,CACpC,MAAM,EAAE,gBAAgB,GACvB,OAAO,CAAC,YAAY,CAAC,CAEvB"}
+{"version":3,"file":"source.d.ts","sourceRoot":"","sources":["../../src/dictionary/source.ts"],"names":[],"mappings":"AAAA,OAAO,EAAoB,KAAK,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAC9E,OAAO,KAAK,EAAE,kBAAkB,EAAE,MAAM,eAAe,CAAC;AACxD,OAAO,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAC7C,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AACvC,OAAO,EAAE,UAAU,EAAE,MAAM,kBAAkB,CAAC;AAE9C;;;;;;;GAOG;AACH,MAAM,WAAW,gBAAgB;IAC/B,0CAA0C;IAC1C,QAAQ,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,OAAO,CAAC,MAAM,CAAC,CAAC;IACjD,2CAA2C;IAC3C,QAAQ,CAAC,KAAK,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,OAAO,CAAC,UAAU,CAAC,CAAC;CACvD;AAED;;GAEG;AACH,wBAAgB,SAAS,CAAC,IAAI,EAAE,IAAI,GAAG;IACrC,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,OAAO,EAAE,MAAM,CAAC;IACzB,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;CAC9B,CAMA;AAED;;;;;;GAMG;AACH,wBAAgB,WAAW,CAAC,OAAO,EAAE,MAAM,GAAG,gBAAgB,CAqB7D;AAED;;;;;GAKG;AACH,wBAAsB,YAAY,CAChC,MAAM,EAAE,gBAAgB,EACxB,IAAI,EAAE,IAAI,GACT,OAAO,CAAC,kBAAkB,CAAC,CAQ7B;AAED;;GAEG;AACH,wBAAsB,cAAc,CAClC,MAAM,EAAE,gBAAgB,EACxB,IAAI,EAAE,IAAI,GACT,OAAO,CAAC,UAAU,CAAC,CAErB;AAED;;;;;;GAMG;AACH,eAAO,MAAM,WAAW,gBAAgB,CAAC;AAEzC;;;;;;;;;;;GAWG;AACH,wBAAsB,cAAc,CAClC,MAAM,EAAE,gBAAgB,GACvB,OAAO,CAAC,UAAU,CAAC,CAErB;AAED;;;;;;GAMG;AACH,eAAO,MAAM,WAAW,eAAe,CAAC;AAExC;;;;;;;GAOG;AACH,wBAAsB,gBAAgB,CACpC,MAAM,EAAE,gBAAgB,GACvB,OAAO,CAAC,YAAY,CAAC,CAEvB"}

package/dist/dictionary/source.js

@@ -1,5 +1,6 @@
 import { readScriptTables } from "../script/conversion.js";
 import { Dictionary } from "./dictionary.js";
+import { WordCounts } from "./word-counts.js";
 /**
  * The three files a tier is made of.
  */
@@ -58,6 +59,29 @@ export async function loadArtifact(source, tier) {
 export async function loadDictionary(source, tier) {
     return Dictionary.from(await loadArtifact(source, tier));
 }
+/**
+ * The file the raw corpus counts live in.
+ *
+ * `full` alone, and outside the tiers rather than one per tier: a caller
+ * ranking words wants the whole vocabulary ordered, and a ranking over part of
+ * it answers a different question.
+ */
+export const COUNTS_FILE = "full.counts";
+/**
+ * Load the raw corpus counts for the `full` tier.
+ *
+ * Separate from the dictionary for the same reason {@link loadScriptTables} is.
+ * Nothing on the decoding path reads counts, and the quantised frequencies the
+ * decoder does read already ship inside every tier. This file is 243 KB brotli
+ * against `full.freq`'s 148 KB, and only a caller that asks for it pays.
+ *
+ * The counts are positional over the `full` tier's keys, so pair them with a
+ * `full` {@link Dictionary} and nothing else. {@link WordCounts.size} and
+ * {@link Dictionary.size} agree where the two match.
+ */
+export async function loadWordCounts(source) {
+    return WordCounts.from(await source.bytes(COUNTS_FILE));
+}
 /**
  * The file the script conversion tables live in.
  *

package/dist/dictionary/source.js.map

@@ -1 +1 @@
-{"version":3,"file":"source.js","sourceRoot":"","sources":["../../src/dictionary/source.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,gBAAgB,EAAqB,MAAM,yBAAyB,CAAC;AAE9E,OAAO,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAkB7C;;GAEG;AACH,MAAM,UAAU,SAAS,CAAC,IAAU;IAKlC,OAAO;QACL,IAAI,EAAE,GAAG,IAAI,OAAO;QACpB,OAAO,EAAE,GAAG,IAAI,UAAU;QAC1B,WAAW,EAAE,GAAG,IAAI,OAAO;KAC5B,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,WAAW,CAAC,OAAe;IACzC,MAAM,EAAE,GAAG,CAAC,IAAY,EAAU,EAAE,CAAC,GAAG,OAAO,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,IAAI,IAAI,EAAE,CAAC;IAC9E,MAAM,GAAG,GAAG,KAAK,EAAE,IAAY,EAAqB,EAAE;QACpD,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC;QACvC,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,KAAK,CACb,YAAY,EAAE,CAAC,IAAI,CAAC,YAAY,MAAM,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,QAAQ,CAAC,UAAU,EAAE,CACjF,CAAC;QACJ,CAAC;QACD,OAAO,QAAQ,CAAC;IAClB,CAAC,CAAC;IACF,OAAO;QACL,IAAI,EAAE,KAAK,EAAE,IAAY,EAAmB,EAAE;YAC5C,MAAM,QAAQ,GAAG,MAAM,GAAG,CAAC,IAAI,CAAC,CAAC;YACjC,OAAO,QAAQ,CAAC,IAAI,EAAE,CAAC;QACzB,CAAC;QACD,KAAK,EAAE,KAAK,EAAE,IAAY,EAAuB,EAAE;YACjD,MAAM,QAAQ,GAAG,MAAM,GAAG,CAAC,IAAI,CAAC,CAAC;YACjC,OAAO,IAAI,UAAU,CAAC,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC,CAAC;QACtD,CAAC;KACF,CAAC;AACJ,CAAC;AAED;;;;;GAKG;AACH,MAAM,CAAC,KAAK,UAAU,YAAY,CAChC,MAAwB,EACxB,IAAU;IAEV,MAAM,KAAK,GAAG,SAAS,CAAC,IAAI,CAAC,CAAC;IAC9B,MAAM,CAAC,IAAI,EAAE,OAAO,EAAE,WAAW,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;QACrD,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC;QACvB,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC;QAC1B,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,WAAW,CAAC;KAChC,CAAC,CAAC;IACH,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,WAAW,EAAE,CAAC;AACxC,CAAC;AAED;;GAEG;AACH,MAAM,CAAC,KAAK,UAAU,cAAc,CAClC,MAAwB,EACxB,IAAU;IAEV,OAAO,UAAU,CAAC,IAAI,CAAC,MAAM,YAAY,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC,CAAC;AAC3D,CAAC;AAED;;;;;;GAMG;AACH,MAAM,CAAC,MAAM,WAAW,GAAG,YAAY,CAAC;AAExC;;;;;;;GAOG;AACH,MAAM,CAAC,KAAK,UAAU,gBAAgB,CACpC,MAAwB;IAExB,OAAO,gBAAgB,CAAC,MAAM,MAAM,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC;AAC1D,CAAC"}
+{"version":3,"file":"source.js","sourceRoot":"","sources":["../../src/dictionary/source.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,gBAAgB,EAAqB,MAAM,yBAAyB,CAAC;AAE9E,OAAO,EAAE,UAAU,EAAE,MAAM,iBAAiB,CAAC;AAE7C,OAAO,EAAE,UAAU,EAAE,MAAM,kBAAkB,CAAC;AAiB9C;;GAEG;AACH,MAAM,UAAU,SAAS,CAAC,IAAU;IAKlC,OAAO;QACL,IAAI,EAAE,GAAG,IAAI,OAAO;QACpB,OAAO,EAAE,GAAG,IAAI,UAAU;QAC1B,WAAW,EAAE,GAAG,IAAI,OAAO;KAC5B,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,WAAW,CAAC,OAAe;IACzC,MAAM,EAAE,GAAG,CAAC,IAAY,EAAU,EAAE,CAAC,GAAG,OAAO,CAAC,OAAO,CAAC,MAAM,EAAE,EAAE,CAAC,IAAI,IAAI,EAAE,CAAC;IAC9E,MAAM,GAAG,GAAG,KAAK,EAAE,IAAY,EAAqB,EAAE;QACpD,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,EAAE,CAAC,IAAI,CAAC,CAAC,CAAC;QACvC,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;YACjB,MAAM,IAAI,KAAK,CACb,YAAY,EAAE,CAAC,IAAI,CAAC,YAAY,MAAM,CAAC,QAAQ,CAAC,MAAM,CAAC,IAAI,QAAQ,CAAC,UAAU,EAAE,CACjF,CAAC;QACJ,CAAC;QACD,OAAO,QAAQ,CAAC;IAClB,CAAC,CAAC;IACF,OAAO;QACL,IAAI,EAAE,KAAK,EAAE,IAAY,EAAmB,EAAE;YAC5C,MAAM,QAAQ,GAAG,MAAM,GAAG,CAAC,IAAI,CAAC,CAAC;YACjC,OAAO,QAAQ,CAAC,IAAI,EAAE,CAAC;QACzB,CAAC;QACD,KAAK,EAAE,KAAK,EAAE,IAAY,EAAuB,EAAE;YACjD,MAAM,QAAQ,GAAG,MAAM,GAAG,CAAC,IAAI,CAAC,CAAC;YACjC,OAAO,IAAI,UAAU,CAAC,MAAM,QAAQ,CAAC,WAAW,EAAE,CAAC,CAAC;QACtD,CAAC;KACF,CAAC;AACJ,CAAC;AAED;;;;;GAKG;AACH,MAAM,CAAC,KAAK,UAAU,YAAY,CAChC,MAAwB,EACxB,IAAU;IAEV,MAAM,KAAK,GAAG,SAAS,CAAC,IAAI,CAAC,CAAC;IAC9B,MAAM,CAAC,IAAI,EAAE,OAAO,EAAE,WAAW,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;QACrD,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC;QACvB,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC;QAC1B,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,WAAW,CAAC;KAChC,CAAC,CAAC;IACH,OAAO,EAAE,IAAI,EAAE,OAAO,EAAE,WAAW,EAAE,CAAC;AACxC,CAAC;AAED;;GAEG;AACH,MAAM,CAAC,KAAK,UAAU,cAAc,CAClC,MAAwB,EACxB,IAAU;IAEV,OAAO,UAAU,CAAC,IAAI,CAAC,MAAM,YAAY,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC,CAAC;AAC3D,CAAC;AAED;;;;;;GAMG;AACH,MAAM,CAAC,MAAM,WAAW,GAAG,aAAa,CAAC;AAEzC;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,KAAK,UAAU,cAAc,CAClC,MAAwB;IAExB,OAAO,UAAU,CAAC,IAAI,CAAC,MAAM,MAAM,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC,CAAC;AAC1D,CAAC;AAED;;;;;;GAMG;AACH,MAAM,CAAC,MAAM,WAAW,GAAG,YAAY,CAAC;AAExC;;;;;;;GAOG;AACH,MAAM,CAAC,KAAK,UAAU,gBAAgB,CACpC,MAAwB;IAExB,OAAO,gBAAgB,CAAC,MAAM,MAAM,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC,CAAC;AAC1D,CAAC"}

package/dist/dictionary/traditional-carry.d.ts

@@ -0,0 +1,23 @@
+import type { DictionaryEntry } from "./entry.js";
+/**
+ * Whether a tag says anything about the word.
+ */
+export declare function isTagged(partOfSpeech: string): boolean;
+/**
+ * The entries after the carry, and how many took each field.
+ */
+export interface CarriedEntries {
+    readonly entries: readonly DictionaryEntry[];
+    readonly carriedTags: number;
+    readonly carriedCounts: number;
+    readonly carriedCapitals: number;
+}
+/**
+ * Give every 繁體 headword the tag and the count its 简体 word carries.
+ *
+ * Neither overwrites something the source stated. A spelling jieba classified
+ * keeps its own tag, and a spelling jieba counted more often than its 简体 word
+ * keeps its own count.
+ */
+export declare function carryToTraditional(entries: readonly DictionaryEntry[]): CarriedEntries;
+//# sourceMappingURL=traditional-carry.d.ts.map

package/dist/dictionary/traditional-carry.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"traditional-carry.d.ts","sourceRoot":"","sources":["../../src/dictionary/traditional-carry.ts"],"names":[],"mappings":"AA6BA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,YAAY,CAAC;AAkBlD;;GAEG;AACH,wBAAgB,QAAQ,CAAC,YAAY,EAAE,MAAM,GAAG,OAAO,CAEtD;AAED;;GAEG;AACH,MAAM,WAAW,cAAc;IAC7B,QAAQ,CAAC,OAAO,EAAE,SAAS,eAAe,EAAE,CAAC;IAC7C,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,aAAa,EAAE,MAAM,CAAC;IAC/B,QAAQ,CAAC,eAAe,EAAE,MAAM,CAAC;CAClC;AAqID;;;;;;GAMG;AACH,wBAAgB,kBAAkB,CAChC,OAAO,EAAE,SAAS,eAAe,EAAE,GAClC,cAAc,CA0ChB"}

package/dist/dictionary/traditional-carry.js

@@ -0,0 +1,201 @@
+/**
+ * What a 繁體 spelling takes from the 简体 word it pairs with.
+ *
+ * jieba's dictionary was counted over a 简体 corpus, and all three of the fields
+ * it settles arrive on one script and not the other. 听 is tagged `v` and
+ * counted 20,435 times. 聽 is tagged nothing, counted nowhere, and 麥 is not a
+ * name where 麦 is one. Every rule that asks what the word beside it is decides
+ * on the tag, every path the decoder weighs is priced on the count, and the
+ * capital comes straight off the proper-noun bit, so 繁體 text was read by
+ * rules that could not see it, priced by a model that had never met it and
+ * capitalised by a different answer from the one 简体 got.
+ *
+ * All three travel over the pairing the entries already hold. An entry names
+ * the 繁體 spelling of its own word. The entry keyed on that spelling is
+ * therefore the same word written the other way, and how a word is classed, how
+ * often it is met and whether it is a name are facts about the word rather than
+ * about the spelling. `corpus-mass.ts` makes the same move for the polyphone
+ * priors, where a vote cast for a 简体 word is cast again for whichever 繁體
+ * form the reading picks.
+ *
+ * A 繁體 spelling that is not a headword of its own needs none of this, because
+ * `artifact-claims.ts` lets its 简体 entry claim the key outright and the key
+ * carries that entry's fields. 銀行 is claimed by 银行 and always priced as it.
+ * What is left is the characters Unihan gives an entry to, which is why only
+ * single characters are paired here.
+ */
+import { convertCharacter } from "../script/conversion.js";
+import { isSingleCharacter } from "../script/characters.js";
+import { traditionalCharacterTable } from "./script-pairings.js";
+/**
+ * jieba's tag for a character it has counted and not classified.
+ *
+ * Written on 5,666 entries and on nothing longer than one character, 简体 and
+ * 繁體 alike: 且, 丢, 乎, 井, 今 and 仍 carry it next to 來, 個 and 這. It says
+ * that the character was seen rather than what it is. Read here as no tag at
+ * all, and left in the artifact, because whether jieba wrote `zg` or wrote
+ * nothing is a fact about the source and only this decides what to make of it.
+ *
+ * Nothing longer than one character carries it. That leaves the rules forbidding
+ * an **untagged key** alone, since each of those asks about a word of two
+ * characters or more and no such key is tagged `zg`.
+ */
+const UNCATEGORISED = "zg";
+/**
+ * Whether a tag says anything about the word.
+ */
+export function isTagged(partOfSpeech) {
+    return partOfSpeech !== "" && partOfSpeech !== UNCATEGORISED;
+}
+/**
+ * The 繁體 character a 简体 one is written as, or undefined where it is neither.
+ *
+ * Read off the aggregate table rather than off {@link DictionaryEntry.hant},
+ * and that is the whole of what keeps this honest. An entry's own 繁體 form is
+ * whichever CC-CEDICT sense matched its reading, and for a single character
+ * that can be an oddity: CC-CEDICT holds `旹 时` as an old variant of 時, so
+ * 时's entry can name 旹 while every word 时 appears in says 時. Lending 时's
+ * 103,735 occurrences to 旹 made a character nobody writes the third commonest
+ * way to write `shí`. The table counts words, so 時 wins it thousands to one.
+ */
+function traditionalOf(table, entry) {
+    if (!isSingleCharacter(entry.hans)) {
+        return undefined;
+    }
+    const written = convertCharacter(table, entry.hans, entry.readings.cn[0]);
+    return written === entry.hans ? undefined : written;
+}
+/**
+ * The 简体 entry each 繁體 character should take a field from.
+ *
+ * Where two 简体 characters are written the same way in 繁體 the commoner of
+ * them lends. Both fields are being used to guess at what a reader meant, and
+ * the commoner character is the likelier thing to have been meant. `hasField`
+ * is asked first, so that a commoner character with nothing to lend does not
+ * shut out a rarer one that has something.
+ */
+function lendersByTraditional(entries, table, hasField) {
+    const lenders = new Map();
+    for (const entry of entries) {
+        const written = hasField(entry) ? traditionalOf(table, entry) : undefined;
+        if (written === undefined) {
+            continue;
+        }
+        const held = lenders.get(written);
+        if (held === undefined || held.frequency < entry.frequency) {
+            lenders.set(written, entry);
+        }
+    }
+    return lenders;
+}
+/**
+ * Whether an entry has a tag worth lending.
+ */
+function canLendTag(entry) {
+    return isTagged(entry.partOfSpeech);
+}
+/**
+ * The tag a spelling ends up with, and whether it took one.
+ *
+ * `nr`, `ns`, `nt` and `nz` are lent along with the rest now that
+ * {@link properFor} carries the bit they go with. Lending the tag without it
+ * would have left a spelling claiming a place name the same entry denies is a
+ * proper noun, which is why they were held back until the bit travelled.
+ */
+function tagFor(entry, lender) {
+    if (isTagged(entry.partOfSpeech) || lender === undefined) {
+        return undefined;
+    }
+    return lender.partOfSpeech;
+}
+/**
+ * The count a spelling ends up with, and whether it took one.
+ *
+ * The larger of the two, rather than the lender's outright. A 繁體 spelling
+ * that jieba counted at all was counted incidentally in a corpus written the
+ * other way, and the word occurs at least as often as the commoner spelling of
+ * it was seen. Taking the smaller would price 過 by the handful of 繁體 lines
+ * jieba's corpus happens to contain.
+ *
+ * Only the form the reading picks takes it. 发 read `fā` is written 發, so 發
+ * takes 发's count and 髮 does not, which is the same line the conversion draws
+ * when it writes 出发 as 出發 and 头发 as 頭髮.
+ *
+ * **The count is carried whole and not shared out.** A share would say that
+ * 繁體 text is some fraction as common as 简体 text, and nothing measured here
+ * says what that fraction is. It would also be invisible: the buckets the
+ * decoder compares are log-spaced, one of them is a factor of 2.49, so halving
+ * a count moves it 0.76 of a bucket and a tenth of it moves it 2.5. Only a
+ * share of an order of magnitude would change a decode, and inventing one that
+ * large to fix a decode that is currently wrong would be trading one guess for
+ * another.
+ */
+function countFor(entry, lender) {
+    if (lender === undefined || lender.frequency <= entry.frequency) {
+        return undefined;
+    }
+    return lender.frequency;
+}
+/**
+ * Whether a spelling is a name, and whether that is a change.
+ *
+ * The bit is what survived `properNounOf`, so it is the answer both sources
+ * reached together rather than jieba's proposal, and it is taken whichever way
+ * it points. Demoting matters as much as promoting. 后 is a locative and 後
+ * arrived tagged `nr` with nothing under its own spelling to challenge it, so
+ * 退休後 came out `tuìxiū Hòu` where 退休后 came out `tuìxiū hòu`.
+ *
+ * This carries the 简体 answer whether or not that answer is right. jieba calls
+ * 连 a surname and 連 now agrees, where before the two differed and one of them
+ * happened to be correct. The two scripts saying one thing is what this is for.
+ * How good that one thing is for a bare character is #159.
+ */
+function properFor(entry, lender) {
+    if (lender === undefined || lender.isProperNoun === entry.isProperNoun) {
+        return undefined;
+    }
+    return lender.isProperNoun;
+}
+/**
+ * Give every 繁體 headword the tag and the count its 简体 word carries.
+ *
+ * Neither overwrites something the source stated. A spelling jieba classified
+ * keeps its own tag, and a spelling jieba counted more often than its 简体 word
+ * keeps its own count.
+ */
+export function carryToTraditional(entries) {
+    const table = traditionalCharacterTable(entries);
+    const tagLenders = lendersByTraditional(entries, table, canLendTag);
+    const countLenders = lendersByTraditional(entries, table, (entry) => entry.frequency > 0);
+    let carriedTags = 0;
+    let carriedCounts = 0;
+    let carriedCapitals = 0;
+    const carried = entries.map((entry) => {
+        const tagLender = tagLenders.get(entry.hans);
+        const partOfSpeech = tagFor(entry, tagLender);
+        const frequency = countFor(entry, countLenders.get(entry.hans));
+        const isProperNoun = properFor(entry, tagLender);
+        if (partOfSpeech === undefined &&
+            frequency === undefined &&
+            isProperNoun === undefined) {
+            return entry;
+        }
+        if (partOfSpeech !== undefined) {
+            carriedTags += 1;
+        }
+        if (frequency !== undefined) {
+            carriedCounts += 1;
+        }
+        if (isProperNoun !== undefined) {
+            carriedCapitals += 1;
+        }
+        return {
+            ...entry,
+            ...(partOfSpeech !== undefined && { partOfSpeech }),
+            ...(frequency !== undefined && { frequency }),
+            ...(isProperNoun !== undefined && { isProperNoun }),
+        };
+    });
+    return { entries: carried, carriedTags, carriedCounts, carriedCapitals };
+}
+//# sourceMappingURL=traditional-carry.js.map

package/dist/dictionary/traditional-carry.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"traditional-carry.js","sourceRoot":"","sources":["../../src/dictionary/traditional-carry.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;;;;;;;;;GAyBG;AACH,OAAO,EAAE,gBAAgB,EAAE,MAAM,yBAAyB,CAAC;AAE3D,OAAO,EAAE,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AAE5D,OAAO,EAAE,yBAAyB,EAAE,MAAM,sBAAsB,CAAC;AAEjE;;;;;;;;;;;;GAYG;AACH,MAAM,aAAa,GAAG,IAAI,CAAC;AAE3B;;GAEG;AACH,MAAM,UAAU,QAAQ,CAAC,YAAoB;IAC3C,OAAO,YAAY,KAAK,EAAE,IAAI,YAAY,KAAK,aAAa,CAAC;AAC/D,CAAC;AAYD;;;;;;;;;;GAUG;AACH,SAAS,aAAa,CACpB,KAA+C,EAC/C,KAAsB;IAEtB,IAAI,CAAC,iBAAiB,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,CAAC;QACnC,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,MAAM,OAAO,GAAG,gBAAgB,CAAC,KAAK,EAAE,KAAK,CAAC,IAAI,EAAE,KAAK,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IAC1E,OAAO,OAAO,KAAK,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,OAAO,CAAC;AACtD,CAAC;AAED;;;;;;;;GAQG;AACH,SAAS,oBAAoB,CAC3B,OAAmC,EACnC,KAA+C,EAC/C,QAA6C;IAE7C,MAAM,OAAO,GAAG,IAAI,GAAG,EAA2B,CAAC;IACnD,KAAK,MAAM,KAAK,IAAI,OAAO,EAAE,CAAC;QAC5B,MAAM,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,aAAa,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;QAC1E,IAAI,OAAO,KAAK,SAAS,EAAE,CAAC;YAC1B,SAAS;QACX,CAAC;QACD,MAAM,IAAI,GAAG,OAAO,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC;QAClC,IAAI,IAAI,KAAK,SAAS,IAAI,IAAI,CAAC,SAAS,GAAG,KAAK,CAAC,SAAS,EAAE,CAAC;YAC3D,OAAO,CAAC,GAAG,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;QAC9B,CAAC;IACH,CAAC;IACD,OAAO,OAAO,CAAC;AACjB,CAAC;AAED;;GAEG;AACH,SAAS,UAAU,CAAC,KAAsB;IACxC,OAAO,QAAQ,CAAC,KAAK,CAAC,YAAY,CAAC,CAAC;AACtC,CAAC;AAED;;;;;;;GAOG;AACH,SAAS,MAAM,CACb,KAAsB,EACtB,MAAmC;IAEnC,IAAI,QAAQ,CAAC,KAAK,CAAC,YAAY,CAAC,IAAI,MAAM,KAAK,SAAS,EAAE,CAAC;QACzD,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,MAAM,CAAC,YAAY,CAAC;AAC7B,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,SAAS,QAAQ,CACf,KAAsB,EACtB,MAAmC;IAEnC,IAAI,MAAM,KAAK,SAAS,IAAI,MAAM,CAAC,SAAS,IAAI,KAAK,CAAC,SAAS,EAAE,CAAC;QAChE,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,MAAM,CAAC,SAAS,CAAC;AAC1B,CAAC;AAED;;;;;;;;;;;;;GAaG;AACH,SAAS,SAAS,CAChB,KAAsB,EACtB,MAAmC;IAEnC,IAAI,MAAM,KAAK,SAAS,IAAI,MAAM,CAAC,YAAY,KAAK,KAAK,CAAC,YAAY,EAAE,CAAC;QACvE,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,MAAM,CAAC,YAAY,CAAC;AAC7B,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,kBAAkB,CAChC,OAAmC;IAEnC,MAAM,KAAK,GAAG,yBAAyB,CAAC,OAAO,CAAC,CAAC;IACjD,MAAM,UAAU,GAAG,oBAAoB,CAAC,OAAO,EAAE,KAAK,EAAE,UAAU,CAAC,CAAC;IACpE,MAAM,YAAY,GAAG,oBAAoB,CACvC,OAAO,EACP,KAAK,EACL,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,SAAS,GAAG,CAAC,CAC/B,CAAC;IACF,IAAI,WAAW,GAAG,CAAC,CAAC;IACpB,IAAI,aAAa,GAAG,CAAC,CAAC;IACtB,IAAI,eAAe,GAAG,CAAC,CAAC;IAExB,MAAM,OAAO,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE;QACpC,MAAM,SAAS,GAAG,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;QAC7C,MAAM,YAAY,GAAG,MAAM,CAAC,KAAK,EAAE,SAAS,CAAC,CAAC;QAC9C,MAAM,SAAS,GAAG,QAAQ,CAAC,KAAK,EAAE,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC;QAChE,MAAM,YAAY,GAAG,SAAS,CAAC,KAAK,EAAE,SAAS,CAAC,CAAC;QACjD,IACE,YAAY,KAAK,SAAS;YAC1B,SAAS,KAAK,SAAS;YACvB,YAAY,KAAK,SAAS,EAC1B,CAAC;YACD,OAAO,KAAK,CAAC;QACf,CAAC;QACD,IAAI,YAAY,KAAK,SAAS,EAAE,CAAC;YAC/B,WAAW,IAAI,CAAC,CAAC;QACnB,CAAC;QACD,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YAC5B,aAAa,IAAI,CAAC,CAAC;QACrB,CAAC;QACD,IAAI,YAAY,KAAK,SAAS,EAAE,CAAC;YAC/B,eAAe,IAAI,CAAC,CAAC;QACvB,CAAC;QACD,OAAO;YACL,GAAG,KAAK;YACR,GAAG,CAAC,YAAY,KAAK,SAAS,IAAI,EAAE,YAAY,EAAE,CAAC;YACnD,GAAG,CAAC,SAAS,KAAK,SAAS,IAAI,EAAE,SAAS,EAAE,CAAC;YAC7C,GAAG,CAAC,YAAY,KAAK,SAAS,IAAI,EAAE,YAAY,EAAE,CAAC;SACpD,CAAC;IACJ,CAAC,CAAC,CAAC;IAEH,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,WAAW,EAAE,aAAa,EAAE,eAAe,EAAE,CAAC;AAC3E,CAAC"}

package/dist/dictionary/word-counts.d.ts

@@ -0,0 +1,74 @@
+import type { DictionaryEntry } from "./entry.js";
+/**
+ * Corpus counts for every key of a tier, one variable-length integer each.
+ *
+ * Counts are positional in the same sense the frequency table is: entry *n*
+ * describes the key at position *n* of the matching
+ * {@link import("./key-index.js").KeyIndex}, which is what
+ * {@link import("./dictionary.js").Dictionary.wordAt} reads.
+ *
+ * The encoding is LEB128, measured against the alternatives on the full tier's
+ * 723,147 keys. It takes 747 KB, and 243 KB brotli, where a fixed 32-bit array
+ * takes 2,825 KB and 299 KB and decimal text takes 1,586 KB and 281 KB. Two
+ * thirds of the keys are attested nowhere and most of the rest carry a small
+ * count, and a fixed width pays for the maximum on all of them.
+ *
+ * Unlike the rest of the artifact this *is* parsed on load, into one typed
+ * array. Nothing in a browser reads it, so the load cost buys a random access
+ * that a varint stream cannot otherwise give.
+ */
+export declare class WordCounts {
+    #private;
+    /**
+     * Encode raw counts, in key order.
+     *
+     * A count of zero means the word is attested nowhere in the corpus, which is
+     * a fact about the word and is stored as one. Fractions and negatives are
+     * floored into range rather than rejected, since a source that supplies one
+     * has already been through the merge.
+     */
+    static build(counts: readonly number[]): WordCounts;
+    /**
+     * Decode packed bytes read back from an artifact.
+     *
+     * How many counts there are is a property of the stream rather than an
+     * argument, since every count ends at the first byte without a continuation
+     * bit. A truncated stream ends mid-count, and the trailing bytes are dropped
+     * rather than reported as a value the file never held.
+     */
+    static from(packed: Uint8Array): WordCounts;
+    /**
+     * Hold decoded counts beside the bytes they came from. Use
+     * {@link WordCounts.build} or {@link WordCounts.from}.
+     */
+    private constructor();
+    /**
+     * How many keys the counts describe.
+     *
+     * Worth checking against {@link import("./dictionary.js").Dictionary.size}
+     * before pairing the two by position. The counts are built for one tier, and
+     * against any other tier every position would name a different word.
+     */
+    get size(): number;
+    /**
+     * Corpus occurrences of the key at a position.
+     *
+     * Positions outside the table report zero, which is what an unattested word
+     * reports, so a caller ranking by count sorts a missing key to the bottom.
+     */
+    countOf(at: number): number;
+    /**
+     * The packed bytes, ready to be written out.
+     */
+    serialise(): Uint8Array;
+}
+/**
+ * Compile the counts artifact from merged entries.
+ *
+ * The keys are claimed and ordered the way `artifact.ts` claims and orders
+ * them, so position *n* here is position *n* there. Only the `full` tier is
+ * written: a caller ranking words wants the whole vocabulary ordered, and a
+ * ranking over part of it answers a different question.
+ */
+export declare function buildWordCounts(entries: readonly DictionaryEntry[]): WordCounts;
+//# sourceMappingURL=word-counts.d.ts.map

package/dist/dictionary/word-counts.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"word-counts.d.ts","sourceRoot":"","sources":["../../src/dictionary/word-counts.ts"],"names":[],"mappings":"AAgBA,OAAO,KAAK,EAAE,eAAe,EAAE,MAAM,YAAY,CAAC;AAelD;;;;;;;;;;;;;;;;;GAiBG;AACH,qBAAa,UAAU;;IACrB;;;;;;;OAOG;IACH,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,SAAS,MAAM,EAAE,GAAG,UAAU,CAgBlD;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAI,CAAC,MAAM,EAAE,UAAU,GAAG,UAAU,CAkB1C;IAKD;;;OAGG;IACH,OAAO,eAGN;IAED;;;;;;OAMG;IACH,IAAI,IAAI,IAAI,MAAM,CAEjB;IAED;;;;;OAKG;IACH,OAAO,CAAC,EAAE,EAAE,MAAM,GAAG,MAAM,CAE1B;IAED;;OAEG;IACH,SAAS,IAAI,UAAU,CAEtB;CACF;AAED;;;;;;;GAOG;AACH,wBAAgB,eAAe,CAC7B,OAAO,EAAE,SAAS,eAAe,EAAE,GAClC,UAAU,CAIZ"}

package/dist/dictionary/word-counts.js

@@ -0,0 +1,142 @@
+/**
+ * Raw corpus counts, kept beside the quantised table rather than inside it.
+ *
+ * {@link import("./frequency-table.js").FrequencyTable} packs a count into four
+ * bits, because 16 buckets are all the decoder compares and they cost 9 KB
+ * where raw single-byte counts cost 23 KB. A caller ranking words against each
+ * other needs what that packing throws away. Rank the 120,858 CC-CEDICT
+ * headwords the full tier holds by `cost` and 5,934 of them land on the value
+ * at rank 10,000, so which words make a top-10,000 cut is decided by codepoint
+ * order. Rank the same words on the counts those buckets were quantised from
+ * and the tie at that rank is 16 words wide.
+ *
+ * So this is a second artifact, read by a caller that asks for it and by
+ * nothing on the decoding path. See `docs/dictionaries/` for the shape of it.
+ */
+import { claimKeys, orderedClaims } from "./artifact-claims.js";
+/**
+ * The largest count that survives a round trip, which is 32 bits of varint.
+ *
+ * jieba's busiest word is attested 883,634 times, which leaves the ceiling
+ * three orders of magnitude clear of anything the corpus holds. It bounds the
+ * encoder rather than rejecting data.
+ */
+const MAX_COUNT = 0xff_ff_ff_ff;
+const CONTINUATION = 0x80;
+const PAYLOAD = 0x7f;
+const SHIFT = 7;
+/**
+ * Corpus counts for every key of a tier, one variable-length integer each.
+ *
+ * Counts are positional in the same sense the frequency table is: entry *n*
+ * describes the key at position *n* of the matching
+ * {@link import("./key-index.js").KeyIndex}, which is what
+ * {@link import("./dictionary.js").Dictionary.wordAt} reads.
+ *
+ * The encoding is LEB128, measured against the alternatives on the full tier's
+ * 723,147 keys. It takes 747 KB, and 243 KB brotli, where a fixed 32-bit array
+ * takes 2,825 KB and 299 KB and decimal text takes 1,586 KB and 281 KB. Two
+ * thirds of the keys are attested nowhere and most of the rest carry a small
+ * count, and a fixed width pays for the maximum on all of them.
+ *
+ * Unlike the rest of the artifact this *is* parsed on load, into one typed
+ * array. Nothing in a browser reads it, so the load cost buys a random access
+ * that a varint stream cannot otherwise give.
+ */
+export class WordCounts {
+    /**
+     * Encode raw counts, in key order.
+     *
+     * A count of zero means the word is attested nowhere in the corpus, which is
+     * a fact about the word and is stored as one. Fractions and negatives are
+     * floored into range rather than rejected, since a source that supplies one
+     * has already been through the merge.
+     */
+    static build(counts) {
+        const values = Uint32Array.from(counts, (count) => Number.isFinite(count) ? Math.min(MAX_COUNT, Math.max(0, count)) : 0);
+        const packed = [];
+        for (const value of values) {
+            let remaining = value;
+            do {
+                const byte = remaining & PAYLOAD;
+                remaining >>>= SHIFT;
+                packed.push(remaining > 0 ? byte | CONTINUATION : byte);
+            } while (remaining > 0);
+        }
+        return new WordCounts(values, Uint8Array.from(packed));
+    }
+    /**
+     * Decode packed bytes read back from an artifact.
+     *
+     * How many counts there are is a property of the stream rather than an
+     * argument, since every count ends at the first byte without a continuation
+     * bit. A truncated stream ends mid-count, and the trailing bytes are dropped
+     * rather than reported as a value the file never held.
+     */
+    static from(packed) {
+        const values = [];
+        let value = 0;
+        let shift = 0;
+        for (const byte of packed) {
+            // `+` rather than `|`, which is signed and would wrap the top bit of a
+            // 32-bit count into a negative number.
+            value += (byte & PAYLOAD) * 2 ** shift;
+            if ((byte & CONTINUATION) === 0) {
+                values.push(value);
+                value = 0;
+                shift = 0;
+            }
+            else {
+                shift += SHIFT;
+            }
+        }
+        return new WordCounts(Uint32Array.from(values), packed);
+    }
+    #counts;
+    #packed;
+    /**
+     * Hold decoded counts beside the bytes they came from. Use
+     * {@link WordCounts.build} or {@link WordCounts.from}.
+     */
+    constructor(counts, packed) {
+        this.#counts = counts;
+        this.#packed = packed;
+    }
+    /**
+     * How many keys the counts describe.
+     *
+     * Worth checking against {@link import("./dictionary.js").Dictionary.size}
+     * before pairing the two by position. The counts are built for one tier, and
+     * against any other tier every position would name a different word.
+     */
+    get size() {
+        return this.#counts.length;
+    }
+    /**
+     * Corpus occurrences of the key at a position.
+     *
+     * Positions outside the table report zero, which is what an unattested word
+     * reports, so a caller ranking by count sorts a missing key to the bottom.
+     */
+    countOf(at) {
+        return this.#counts[at] ?? 0;
+    }
+    /**
+     * The packed bytes, ready to be written out.
+     */
+    serialise() {
+        return this.#packed;
+    }
+}
+/**
+ * Compile the counts artifact from merged entries.
+ *
+ * The keys are claimed and ordered the way `artifact.ts` claims and orders
+ * them, so position *n* here is position *n* there. Only the `full` tier is
+ * written: a caller ranking words wants the whole vocabulary ordered, and a
+ * ranking over part of it answers a different question.
+ */
+export function buildWordCounts(entries) {
+    return WordCounts.build(orderedClaims(claimKeys(entries)).map(([, entry]) => entry.frequency));
+}
+//# sourceMappingURL=word-counts.js.map

package/dist/dictionary/word-counts.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"word-counts.js","sourceRoot":"","sources":["../../src/dictionary/word-counts.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;GAcG;AACH,OAAO,EAAE,SAAS,EAAE,aAAa,EAAE,MAAM,sBAAsB,CAAC;AAGhE;;;;;;GAMG;AACH,MAAM,SAAS,GAAG,aAAa,CAAC;AAEhC,MAAM,YAAY,GAAG,IAAI,CAAC;AAC1B,MAAM,OAAO,GAAG,IAAI,CAAC;AACrB,MAAM,KAAK,GAAG,CAAC,CAAC;AAEhB;;;;;;;;;;;;;;;;;GAiBG;AACH,MAAM,OAAO,UAAU;IACrB;;;;;;;OAOG;IACH,MAAM,CAAC,KAAK,CAAC,MAAyB;QACpC,MAAM,MAAM,GAAG,WAAW,CAAC,IAAI,CAAC,MAAM,EAAE,CAAC,KAAK,EAAE,EAAE,CAChD,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,SAAS,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CACrE,CAAC;QAEF,MAAM,MAAM,GAAa,EAAE,CAAC;QAC5B,KAAK,MAAM,KAAK,IAAI,MAAM,EAAE,CAAC;YAC3B,IAAI,SAAS,GAAG,KAAK,CAAC;YACtB,GAAG,CAAC;gBACF,MAAM,IAAI,GAAG,SAAS,GAAG,OAAO,CAAC;gBACjC,SAAS,MAAM,KAAK,CAAC;gBACrB,MAAM,CAAC,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,YAAY,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;YAC1D,CAAC,QAAQ,SAAS,GAAG,CAAC,EAAE;QAC1B,CAAC;QAED,OAAO,IAAI,UAAU,CAAC,MAAM,EAAE,UAAU,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC;IACzD,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAI,CAAC,MAAkB;QAC5B,MAAM,MAAM,GAAa,EAAE,CAAC;QAC5B,IAAI,KAAK,GAAG,CAAC,CAAC;QACd,IAAI,KAAK,GAAG,CAAC,CAAC;QACd,KAAK,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;YAC1B,uEAAuE;YACvE,uCAAuC;YACvC,KAAK,IAAI,CAAC,IAAI,GAAG,OAAO,CAAC,GAAG,CAAC,IAAI,KAAK,CAAC;YACvC,IAAI,CAAC,IAAI,GAAG,YAAY,CAAC,KAAK,CAAC,EAAE,CAAC;gBAChC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;gBACnB,KAAK,GAAG,CAAC,CAAC;gBACV,KAAK,GAAG,CAAC,CAAC;YACZ,CAAC;iBAAM,CAAC;gBACN,KAAK,IAAI,KAAK,CAAC;YACjB,CAAC;QACH,CAAC;QAED,OAAO,IAAI,UAAU,CAAC,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,MAAM,CAAC,CAAC;IAC1D,CAAC;IAEQ,OAAO,CAAc;IACrB,OAAO,CAAa;IAE7B;;;OAGG;IACH,YAAoB,MAAmB,EAAE,MAAkB;QACzD,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC;QACtB,IAAI,CAAC,OAAO,GAAG,MAAM,CAAC;IACxB,CAAC;IAED;;;;;;OAMG;IACH,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC;IAC7B,CAAC;IAED;;;;;OAKG;IACH,OAAO,CAAC,EAAU;QAChB,OAAO,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC,IAAI,CAAC,CAAC;IAC/B,CAAC;IAED;;OAEG;IACH,SAAS;QACP,OAAO,IAAI,CAAC,OAAO,CAAC;IACtB,CAAC;CACF;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,eAAe,CAC7B,OAAmC;IAEnC,OAAO,UAAU,CAAC,KAAK,CACrB,aAAa,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,KAAK,CAAC,EAAE,EAAE,CAAC,KAAK,CAAC,SAAS,CAAC,CACtE,CAAC;AACJ,CAAC"}

package/dist/format/annotate.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"annotate.d.ts","sourceRoot":"","sources":["../../src/format/annotate.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAU,KAAK,WAAW,EAAsB,MAAM,WAAW,CAAC;AACzE,OAAO,EAAE,KAAK,cAAc,EAAiB,MAAM,sBAAsB,CAAC;AAC1E,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D;;;;;;;;;;;;;;;GAeG;AACH,eAAO,MAAM,QAAQ,OAAO,CAAC;AAE7B;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,mEAAmE;IACnE,QAAQ,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAC;IACpC,QAAQ,CAAC,MAAM,EAAE,SAAS,cAAc,EAAE,CAAC;CAC5C;AAED;;;;;;;;GAQG;AACH,wBAAgB,aAAa,CAAC,MAAM,EAAE,SAAS,cAAc,EAAE,GAAG,UAAU,EAAE,CAc7E;AA2ED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,wBAAgB,eAAe,CAC7B,MAAM,EAAE,SAAS,cAAc,EAAE,EACjC,OAAO,GAAE,WAAgB,GACxB,MAAM,CAIR;AAED;;;;;GAKG;AACH,wBAAgB,sBAAsB,CACpC,UAAU,EAAE,UAAU,EACtB,IAAI,EAAE,MAAM,EACZ,OAAO,GAAE,WAAgB,GACxB,MAAM,CAER;AAED;;;;;;;;;;;GAWG;AACH,wBAAgB,aAAa,CAC3B,UAAU,EAAE,UAAU,EACtB,IAAI,EAAE,MAAM,EACZ,OAAO,GAAE,WAAgB,GACxB,MAAM,CAER"}
+{"version":3,"file":"annotate.d.ts","sourceRoot":"","sources":["../../src/format/annotate.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAU,KAAK,WAAW,EAAuB,MAAM,WAAW,CAAC;AAC1E,OAAO,EAAE,KAAK,cAAc,EAAiB,MAAM,sBAAsB,CAAC;AAC1E,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D;;;;;;;;;;;;;;;GAeG;AACH,eAAO,MAAM,QAAQ,OAAO,CAAC;AAE7B;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,mEAAmE;IACnE,QAAQ,CAAC,MAAM,EAAE,MAAM,GAAG,SAAS,CAAC;IACpC,QAAQ,CAAC,MAAM,EAAE,SAAS,cAAc,EAAE,CAAC;CAC5C;AAED;;;;;;;;GAQG;AACH,wBAAgB,aAAa,CAAC,MAAM,EAAE,SAAS,cAAc,EAAE,GAAG,UAAU,EAAE,CAc7E;AA2ED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,wBAAgB,eAAe,CAC7B,MAAM,EAAE,SAAS,cAAc,EAAE,EACjC,OAAO,GAAE,WAAgB,GACxB,MAAM,CAIR;AAED;;;;;GAKG;AACH,wBAAgB,sBAAsB,CACpC,UAAU,EAAE,UAAU,EACtB,IAAI,EAAE,MAAM,EACZ,OAAO,GAAE,WAAgB,GACxB,MAAM,CAER;AAED;;;;;;;;;;;GAWG;AACH,wBAAgB,aAAa,CAC3B,UAAU,EAAE,UAAU,EACtB,IAAI,EAAE,MAAM,EACZ,OAAO,GAAE,WAAgB,GACxB,MAAM,CAER"}

package/dist/format/annotate.js

@@ -4,7 +4,7 @@
  * The pinyin sits above the characters rather than beside them, so the pieces
  * are regrouped against the text they annotate.
  */
-import { escape, toHtml, writePiece } from "./html.js";
+import { escape, toHtml, writePieces } from "./html.js";
 import { convertPieces } from "../decode/convert.js";
 /**
  * What the characters being annotated declare themselves to be.
@@ -101,9 +101,9 @@ function writeAnnotation(annotation, options) {
         // characters — but {@link toAnnotatedHtml} takes pieces from the caller,
         // and there is no base to hang an annotation on, so it is written as the
         // plain markup it would have got from {@link toHtml}.
-        return pieces.map((piece) => writePiece(piece, options)).join("");
+        return writePieces(pieces, options);
     }
-    const reading = pieces.map((piece) => writePiece(piece, options)).join("");
+    const reading = writePieces(pieces, options);
     // The tag goes on the <ruby> rather than around the base, and the syllables
     // inside the <rt> override it with their own. `<rb>` would be the obvious
     // place for it and is not an element any more: the WHATWG parser drops it,

package/dist/format/annotate.js.map

@@ -1 +1 @@
-{"version":3,"file":"annotate.js","sourceRoot":"","sources":["../../src/format/annotate.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAE,MAAM,EAAoB,MAAM,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AACzE,OAAO,EAAuB,aAAa,EAAE,MAAM,sBAAsB,CAAC;AAG1E;;;;;;;;;;;;;;;GAeG;AACH,MAAM,CAAC,MAAM,QAAQ,GAAG,IAAI,CAAC;AAW7B;;;;;;;;GAQG;AACH,MAAM,UAAU,aAAa,CAAC,MAAiC;IAC7D,MAAM,MAAM,GAAiB,EAAE,CAAC;IAChC,KAAK,MAAM,CAAC,EAAE,EAAE,KAAK,CAAC,IAAI,MAAM,CAAC,OAAO,EAAE,EAAE,CAAC;QAC3C,MAAM,IAAI,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;QAC3B,IAAI,IAAI,EAAE,MAAM,KAAK,SAAS,IAAI,SAAS,CAAC,KAAK,EAAE,MAAM,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC;YACnE,MAAM,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG;gBAC1B,MAAM,EAAE,IAAI,CAAC,MAAM;gBACnB,MAAM,EAAE,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC;aAChC,CAAC;YACF,SAAS;QACX,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,KAAK,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IACzD,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,SAAS,SAAS,CAChB,KAAqB,EACrB,IAAgC;IAEhC,IAAI,KAAK,CAAC,QAAQ,KAAK,SAAS,EAAE,CAAC;QACjC,OAAO,KAAK,CAAC,MAAM,KAAK,SAAS,CAAC;IACpC,CAAC;IACD,IAAI,KAAK,CAAC,MAAM,KAAK,SAAS,EAAE,CAAC;QAC/B,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,CACL,KAAK,CAAC,IAAI,CAAC,IAAI,EAAE,KAAK,EAAE;QACxB,CAAC,IAAI,EAAE,QAAQ,KAAK,SAAS,IAAI,IAAI,CAAC,MAAM,KAAK,SAAS,CAAC,CAC5D,CAAC;AACJ,CAAC;AAED;;GAEG;AACH,SAAS,eAAe,CAAC,UAAsB,EAAE,OAAoB;IACnE,MAAM,EAAE,IAAI,GAAG,IAAI,EAAE,GAAG,OAAO,CAAC;IAChC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,GAAG,UAAU,CAAC;IACtC,IAAI,MAAM,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,QAAQ,KAAK,SAAS,CAAC,EAAE,CAAC;QAC1D,oEAAoE;QACpE,2EAA2E;QAC3E,uEAAuE;QACvE,wEAAwE;QACxE,qDAAqD;QACrD,OAAO,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;IACpD,CAAC;IACD,IAAI,MAAM,KAAK,SAAS,EAAE,CAAC;QACzB,wEAAwE;QACxE,yEAAyE;QACzE,yEAAyE;QACzE,yEAAyE;QACzE,sDAAsD;QACtD,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,UAAU,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;IACpE,CAAC;IAED,MAAM,OAAO,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,UAAU,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;IAC3E,4EAA4E;IAC5E,0EAA0E;IAC1E,2EAA2E;IAC3E,2EAA2E;IAC3E,yBAAyB;IACzB,MAAM,OAAO,GAAG,IAAI,CAAC,CAAC,CAAC,eAAe,QAAQ,IAAI,CAAC,CAAC,CAAC,QAAQ,CAAC;IAC9D,OAAO,GAAG,OAAO,GAAG,MAAM,CAAC,MAAM,CAAC,iBAAiB,OAAO,wBAAwB,CAAC;AACrF,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,MAAM,UAAU,eAAe,CAC7B,MAAiC,EACjC,OAAO,GAAgB,EAAE;IAEzB,OAAO,aAAa,CAAC,MAAM,CAAC;SACzB,GAAG,CAAC,CAAC,UAAU,EAAE,EAAE,CAAC,eAAe,CAAC,UAAU,EAAE,OAAO,CAAC,CAAC;SACzD,IAAI,CAAC,EAAE,CAAC,CAAC;AACd,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,sBAAsB,CACpC,UAAsB,EACtB,IAAY,EACZ,OAAO,GAAgB,EAAE;IAEzB,OAAO,eAAe,CAAC,aAAa,CAAC,UAAU,EAAE,IAAI,EAAE,OAAO,CAAC,EAAE,OAAO,CAAC,CAAC;AAC5E,CAAC;AAED;;;;;;;;;;;GAWG;AACH,MAAM,UAAU,aAAa,CAC3B,UAAsB,EACtB,IAAY,EACZ,OAAO,GAAgB,EAAE;IAEzB,OAAO,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,IAAI,EAAE,OAAO,CAAC,EAAE,OAAO,CAAC,CAAC;AACnE,CAAC"}
+{"version":3,"file":"annotate.js","sourceRoot":"","sources":["../../src/format/annotate.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAE,MAAM,EAAoB,MAAM,EAAE,WAAW,EAAE,MAAM,WAAW,CAAC;AAC1E,OAAO,EAAuB,aAAa,EAAE,MAAM,sBAAsB,CAAC;AAG1E;;;;;;;;;;;;;;;GAeG;AACH,MAAM,CAAC,MAAM,QAAQ,GAAG,IAAI,CAAC;AAW7B;;;;;;;;GAQG;AACH,MAAM,UAAU,aAAa,CAAC,MAAiC;IAC7D,MAAM,MAAM,GAAiB,EAAE,CAAC;IAChC,KAAK,MAAM,CAAC,EAAE,EAAE,KAAK,CAAC,IAAI,MAAM,CAAC,OAAO,EAAE,EAAE,CAAC;QAC3C,MAAM,IAAI,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;QAC3B,IAAI,IAAI,EAAE,MAAM,KAAK,SAAS,IAAI,SAAS,CAAC,KAAK,EAAE,MAAM,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC;YACnE,MAAM,CAAC,MAAM,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG;gBAC1B,MAAM,EAAE,IAAI,CAAC,MAAM;gBACnB,MAAM,EAAE,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC;aAChC,CAAC;YACF,SAAS;QACX,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,KAAK,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IACzD,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,SAAS,SAAS,CAChB,KAAqB,EACrB,IAAgC;IAEhC,IAAI,KAAK,CAAC,QAAQ,KAAK,SAAS,EAAE,CAAC;QACjC,OAAO,KAAK,CAAC,MAAM,KAAK,SAAS,CAAC;IACpC,CAAC;IACD,IAAI,KAAK,CAAC,MAAM,KAAK,SAAS,EAAE,CAAC;QAC/B,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,CACL,KAAK,CAAC,IAAI,CAAC,IAAI,EAAE,KAAK,EAAE;QACxB,CAAC,IAAI,EAAE,QAAQ,KAAK,SAAS,IAAI,IAAI,CAAC,MAAM,KAAK,SAAS,CAAC,CAC5D,CAAC;AACJ,CAAC;AAED;;GAEG;AACH,SAAS,eAAe,CAAC,UAAsB,EAAE,OAAoB;IACnE,MAAM,EAAE,IAAI,GAAG,IAAI,EAAE,GAAG,OAAO,CAAC;IAChC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,GAAG,UAAU,CAAC;IACtC,IAAI,MAAM,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,QAAQ,KAAK,SAAS,CAAC,EAAE,CAAC;QAC1D,oEAAoE;QACpE,2EAA2E;QAC3E,uEAAuE;QACvE,wEAAwE;QACxE,qDAAqD;QACrD,OAAO,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;IACpD,CAAC;IACD,IAAI,MAAM,KAAK,SAAS,EAAE,CAAC;QACzB,wEAAwE;QACxE,yEAAyE;QACzE,yEAAyE;QACzE,yEAAyE;QACzE,sDAAsD;QACtD,OAAO,WAAW,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IACtC,CAAC;IAED,MAAM,OAAO,GAAG,WAAW,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;IAC7C,4EAA4E;IAC5E,0EAA0E;IAC1E,2EAA2E;IAC3E,2EAA2E;IAC3E,yBAAyB;IACzB,MAAM,OAAO,GAAG,IAAI,CAAC,CAAC,CAAC,eAAe,QAAQ,IAAI,CAAC,CAAC,CAAC,QAAQ,CAAC;IAC9D,OAAO,GAAG,OAAO,GAAG,MAAM,CAAC,MAAM,CAAC,iBAAiB,OAAO,wBAAwB,CAAC;AACrF,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,MAAM,UAAU,eAAe,CAC7B,MAAiC,EACjC,OAAO,GAAgB,EAAE;IAEzB,OAAO,aAAa,CAAC,MAAM,CAAC;SACzB,GAAG,CAAC,CAAC,UAAU,EAAE,EAAE,CAAC,eAAe,CAAC,UAAU,EAAE,OAAO,CAAC,CAAC;SACzD,IAAI,CAAC,EAAE,CAAC,CAAC;AACd,CAAC;AAED;;;;;GAKG;AACH,MAAM,UAAU,sBAAsB,CACpC,UAAsB,EACtB,IAAY,EACZ,OAAO,GAAgB,EAAE;IAEzB,OAAO,eAAe,CAAC,aAAa,CAAC,UAAU,EAAE,IAAI,EAAE,OAAO,CAAC,EAAE,OAAO,CAAC,CAAC;AAC5E,CAAC;AAED;;;;;;;;;;;GAWG;AACH,MAAM,UAAU,aAAa,CAC3B,UAAsB,EACtB,IAAY,EACZ,OAAO,GAAgB,EAAE;IAEzB,OAAO,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,IAAI,EAAE,OAAO,CAAC,EAAE,OAAO,CAAC,CAAC;AACnE,CAAC"}