@kensio/pinyinjs 1.7.3 → 1.7.5

package/dist/decode/numbers.js

@@ -0,0 +1,69 @@
+import { isSpaced, numberPieces, runEdge } from "./number-pieces.js";
+/**
+ * The numbers in a text, said out loud.
+ *
+ * A digit is not Han and does not reach the decoder, so reading 1998年 as
+ * `yī jiǔ jiǔ bā nián` means working out what the run around it says and
+ * whether a space belongs between them. That is all this does.
+ */
+import { saidNumeral } from "../numerals/text.js";
+import { surroundingCharacters } from "./runs.js";
+import { plainPiece, sourcePiece, } from "./pieces.js";
+/**
+ * What surrounds a run once it has been decoded: the character after it, and
+ * the syllable that character is read as.
+ *
+ * The syllable is the one part of a number's context that a decode has to
+ * supply, and it is what a 一 ending the number assimilates to.
+ */
+export function surrounding(runs, decoded, at) {
+    return {
+        after: {
+            character: surroundingCharacters(runs, at).following,
+            syllable: decoded[at + 1]?.[0]?.word.reading[0],
+        },
+        isAfterHan: runs[at - 1]?.isHan === true,
+    };
+}
+/**
+ * The 汉字 a number in front of a Han run stands for, for that run's decode.
+ *
+ * Only the last segment of the run before, because only that one touches the
+ * Han: the D of 3D银行 comes between them, and a decode of 银行 that saw 三
+ * beside it would be reading a text nobody wrote.
+ */
+export function numeralBefore(segments) {
+    return segments.at(-1)?.hanzi ?? "";
+}
+/**
+ * Write a stretch that was never Han, reading the numbers in it.
+ *
+ * Everything that is not a number goes through exactly as written, which is
+ * what this always did: digits are the only part of a non-Han run this package
+ * has anything to say about. Once a number *has* been read, though, the whole
+ * stretch is being said rather than shown, so its parts take the spacing of
+ * words — 3D打印 is `sān D dǎyìn` — and punctuation still takes none.
+ */
+export function writeNumbers(text, segments, context, written, options) {
+    if (segments.every((segment) => segment.reading === undefined)) {
+        return [sourcePiece(text)];
+    }
+    const pieces = [];
+    let before = runEdge(context.isAfterHan);
+    for (const segment of segments) {
+        if (isSpaced(before, segment.text)) {
+            pieces.push(plainPiece(" "));
+        }
+        pieces.push(...(segment.reading === undefined
+            ? [sourcePiece(segment.text)]
+            : numberPieces(saidNumeral(segment, context.after.syllable, options.sandhi), segment, written)));
+        // What decides the next space is what was *written*, not what was read:
+        // 95% ends in a sign and `bǎifēnzhījiǔshíwǔ` ends in a letter.
+        before = pieces.at(-1)?.text ?? before;
+    }
+    if (isSpaced(before, runEdge(context.after.character !== ""))) {
+        pieces.push(plainPiece(" "));
+    }
+    return pieces;
+}
+//# sourceMappingURL=numbers.js.map

package/dist/decode/numbers.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"numbers.js","sourceRoot":"","sources":["../../src/decode/numbers.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,QAAQ,EAAE,YAAY,EAAE,OAAO,EAAE,MAAM,oBAAoB,CAAC;AACrE;;;;;;GAMG;AACH,OAAO,EAAuB,WAAW,EAAE,MAAM,qBAAqB,CAAC;AACvE,OAAO,EAAE,qBAAqB,EAAgB,MAAM,WAAW,CAAC;AAGhE,OAAO,EAEL,UAAU,EACV,WAAW,GAEZ,MAAM,aAAa,CAAC;AAcrB;;;;;;GAMG;AACH,MAAM,UAAU,WAAW,CACzB,IAAwB,EACxB,OAA2C,EAC3C,EAAU;IAEV,OAAO;QACL,KAAK,EAAE;YACL,SAAS,EAAE,qBAAqB,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,SAAS;YACpD,QAAQ,EAAE,OAAO,CAAC,EAAE,GAAG,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC;SAChD;QACD,UAAU,EAAE,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC,EAAE,KAAK,KAAK,IAAI;KACzC,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,aAAa,CAAC,QAAmC;IAC/D,OAAO,QAAQ,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,IAAI,EAAE,CAAC;AACtC,CAAC;AAED;;;;;;;;GAQG;AACH,MAAM,UAAU,YAAY,CAC1B,IAAY,EACZ,QAAmC,EACnC,OAAmB,EACnB,OAAgB,EAChB,OAAuD;IAEvD,IAAI,QAAQ,CAAC,KAAK,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,CAAC,OAAO,KAAK,SAAS,CAAC,EAAE,CAAC;QAC/D,OAAO,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC,CAAC;IAC7B,CAAC;IAED,MAAM,MAAM,GAAqB,EAAE,CAAC;IACpC,IAAI,MAAM,GAAG,OAAO,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC;IAEzC,KAAK,MAAM,OAAO,IAAI,QAAQ,EAAE,CAAC;QAC/B,IAAI,QAAQ,CAAC,MAAM,EAAE,OAAO,CAAC,IAAI,CAAC,EAAE,CAAC;YACnC,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,CAAC;QAC/B,CAAC;QACD,MAAM,CAAC,IAAI,CACT,GAAG,CAAC,OAAO,CAAC,OAAO,KAAK,SAAS;YAC/B,CAAC,CAAC,CAAC,WAAW,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;YAC7B,CAAC,CAAC,YAAY,CACV,WAAW,CAAC,OAAO,EAAE,OAAO,CAAC,KAAK,CAAC,QAAQ,EAAE,OAAO,CAAC,MAAM,CAAC,EAC5D,OAAO,EACP,OAAO,CACR,CAAC,CACP,CAAC;QACF,wEAAwE;QACxE,+DAA+D;QAC/D,MAAM,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,IAAI,MAAM,CAAC;IACzC,CAAC;IACD,IAAI,QAAQ,CAAC,MAAM,EAAE,OAAO,CAAC,OAAO,CAAC,KAAK,CAAC,SAAS,KAAK,EAAE,CAAC,CAAC,EAAE,CAAC;QAC9D,MAAM,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,CAAC;IAC/B,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC"}

package/dist/decode/pieces.d.ts

@@ -0,0 +1,78 @@
+/**
+ * What a conversion is made of: the options it takes, and the pieces it
+ * hands back.
+ *
+ * The leaf layer of the conversion. Everything here builds or describes a
+ * single piece of output, and none of it knows how a text is decoded — which
+ * is what lets `convert.ts` sit on top of it.
+ */
+import type { ConvertedPiece, Written } from "./convert-options.js";
+import type { DecodedWord } from "./word.js";
+import type { Dictionary } from "../dictionary/dictionary.js";
+export type { ConvertedPiece, ConvertOptions, NumberStyle, Written, } from "./convert-options.js";
+import type { Locale } from "../script/script.js";
+import { type Syllable } from "../syllable/syllable.js";
+import type { ReadingConfidence } from "./confidence.js";
+/**
+ * The reading a word takes in a locale.
+ *
+ * `zh-TW` is stored as a delta, so a word with no Taiwan reading simply reads
+ * the same in both.
+ *
+ * The delta is measured against the entry's own 普通话 reading, so it only
+ * applies where the decode settled on that reading. A polyphone the decode read
+ * some other way — 长 as `cháng` where its entry reads `zhǎng`, 差 as `chā`
+ * where its entry reads `chà` — is a different sense of the word, and the delta
+ * beside the entry says nothing about it. Replacing the reading there would
+ * discard what the lattice worked out in favour of a reading of the wrong
+ * syllable entirely.
+ */
+export declare function readingFor(dictionary: Dictionary, word: DecodedWord, locale: Locale): readonly Syllable[];
+/**
+ * Text that writes no syllable: a space, or a run that was never Han.
+ */
+export declare function plainPiece(text: string): ConvertedPiece;
+/**
+ * Source text that is written out but read as nothing.
+ *
+ * The difference from {@link plainPiece} is what the text *is*. Both write no
+ * syllable, but a run that was never Han — punctuation, a Latin word, a digit
+ * left unread — stands for something the author wrote, whereas the space
+ * between two words and the hyphen inside 干干净净 are pinyin orthography that
+ * the source has no trace of.
+ *
+ * A conversion joins them into one string and the distinction does not arise.
+ * An annotation writes the source and the reading in different places, and
+ * putting a hyphen the pinyin needs into the hanzi it annotates would be
+ * inventing text: 干干净净 is `gāngān-jìngjìng` and is not written 干干-净净.
+ *
+ * The text and the source can differ, and the punctuation pass is why: it
+ * rewrites 。 to a full stop in the `text` a conversion writes, and leaves the
+ * `source` as the mark the author actually typed.
+ */
+export declare function sourcePiece(text: string): ConvertedPiece;
+/**
+ * Which characters of a word each of its syllables reads.
+ *
+ * One per syllable, in order, and undefined where a syllable reads on into the
+ * characters the one before it took. Only 儿化 does that: measured over the
+ * committed dictionary, 5,283 of 723,147 keys have fewer syllables than
+ * characters, 4,024 of them because a final 儿 folds into the syllable before
+ * it, and the remaining 1,259 because the word is written with punctuation —
+ * which cannot reach here, since punctuation ends a Han run before the decode
+ * ever sees it.
+ *
+ * The deficit is spent from the right, which is where 儿化 always is. 打哪儿指
+ * 哪儿 is the case that needs it: two syllables short over six characters, and
+ * both 儿 attach to the character in front of them rather than one taking two.
+ */
+export declare function sourcesOf(word: string, reading: readonly Syllable[]): readonly (string | undefined)[];
+/**
+ * Write one decoded word as one piece per syllable.
+ *
+ * Only proper nouns are capitalised here; a sentence capital is applied to the
+ * whole conversion afterwards, since it belongs to whichever run happens to
+ * start the sentence.
+ */
+export declare function writeWord(reading: readonly Syllable[], confidence: readonly ReadingConfidence[], word: DecodedWord, written: Written): readonly ConvertedPiece[];
+//# sourceMappingURL=pieces.d.ts.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"pieces.d.ts","sourceRoot":"","sources":["../../src/decode/pieces.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,KAAK,EAAE,cAAc,EAAE,OAAO,EAAE,MAAM,sBAAsB,CAAC;AACpE,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,WAAW,CAAC;AAG7C,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D,YAAY,EACV,cAAc,EACd,cAAc,EACd,WAAW,EACX,OAAO,GACR,MAAM,sBAAsB,CAAC;AAI9B,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,qBAAqB,CAAC;AAClD,OAAO,EAAE,KAAK,QAAQ,EAAiB,MAAM,yBAAyB,CAAC;AACvE,OAAO,KAAK,EAAE,iBAAiB,EAAE,MAAM,iBAAiB,CAAC;AACzD;;;;;;;;;;;;;GAaG;AACH,wBAAgB,UAAU,CACxB,UAAU,EAAE,UAAU,EACtB,IAAI,EAAE,WAAW,EACjB,MAAM,EAAE,MAAM,GACb,SAAS,QAAQ,EAAE,CAWrB;AAED;;GAEG;AACH,wBAAgB,UAAU,CAAC,IAAI,EAAE,MAAM,GAAG,cAAc,CAOvD;AAED;;;;;;;;;;;;;;;;;GAiBG;AACH,wBAAgB,WAAW,CAAC,IAAI,EAAE,MAAM,GAAG,cAAc,CAOxD;AAED;;;;;;;;;;;;;;GAcG;AACH,wBAAgB,SAAS,CACvB,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,SAAS,QAAQ,EAAE,GAC3B,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE,CA2BjC;AAED;;;;;;GAMG;AACH,wBAAgB,SAAS,CACvB,OAAO,EAAE,SAAS,QAAQ,EAAE,EAC5B,UAAU,EAAE,SAAS,iBAAiB,EAAE,EACxC,IAAI,EAAE,WAAW,EACjB,OAAO,EAAE,OAAO,GACf,SAAS,cAAc,EAAE,CAqB3B"}

package/dist/decode/pieces.js

@@ -0,0 +1,136 @@
+import { markWord } from "../orthography/apostrophe.js";
+import { capitaliseWord } from "../orthography/capitals.js";
+import { isErCharacter, isErFinal } from "../dictionary/erhua.js";
+import { isSameReading } from "../dictionary/entry.js";
+import { toCharacters } from "../script/characters.js";
+import { writeSyllable } from "../syllable/syllable.js";
+/**
+ * The reading a word takes in a locale.
+ *
+ * `zh-TW` is stored as a delta, so a word with no Taiwan reading simply reads
+ * the same in both.
+ *
+ * The delta is measured against the entry's own 普通话 reading, so it only
+ * applies where the decode settled on that reading. A polyphone the decode read
+ * some other way — 长 as `cháng` where its entry reads `zhǎng`, 差 as `chā`
+ * where its entry reads `chà` — is a different sense of the word, and the delta
+ * beside the entry says nothing about it. Replacing the reading there would
+ * discard what the lattice worked out in favour of a reading of the wrong
+ * syllable entirely.
+ */
+export function readingFor(dictionary, word, locale) {
+    if (locale !== "zh-TW") {
+        return word.reading;
+    }
+    const entry = dictionary.lookup(word.text);
+    if (entry?.taiwanReading === undefined) {
+        return word.reading;
+    }
+    return isSameReading(entry.reading, word.reading)
+        ? entry.taiwanReading
+        : word.reading;
+}
+/**
+ * Text that writes no syllable: a space, or a run that was never Han.
+ */
+export function plainPiece(text) {
+    return {
+        text,
+        syllable: undefined,
+        confidence: undefined,
+        source: undefined,
+    };
+}
+/**
+ * Source text that is written out but read as nothing.
+ *
+ * The difference from {@link plainPiece} is what the text *is*. Both write no
+ * syllable, but a run that was never Han — punctuation, a Latin word, a digit
+ * left unread — stands for something the author wrote, whereas the space
+ * between two words and the hyphen inside 干干净净 are pinyin orthography that
+ * the source has no trace of.
+ *
+ * A conversion joins them into one string and the distinction does not arise.
+ * An annotation writes the source and the reading in different places, and
+ * putting a hyphen the pinyin needs into the hanzi it annotates would be
+ * inventing text: 干干净净 is `gāngān-jìngjìng` and is not written 干干-净净.
+ *
+ * The text and the source can differ, and the punctuation pass is why: it
+ * rewrites 。 to a full stop in the `text` a conversion writes, and leaves the
+ * `source` as the mark the author actually typed.
+ */
+export function sourcePiece(text) {
+    return {
+        text,
+        syllable: undefined,
+        confidence: undefined,
+        source: text,
+    };
+}
+/**
+ * Which characters of a word each of its syllables reads.
+ *
+ * One per syllable, in order, and undefined where a syllable reads on into the
+ * characters the one before it took. Only 儿化 does that: measured over the
+ * committed dictionary, 5,283 of 723,147 keys have fewer syllables than
+ * characters, 4,024 of them because a final 儿 folds into the syllable before
+ * it, and the remaining 1,259 because the word is written with punctuation —
+ * which cannot reach here, since punctuation ends a Han run before the decode
+ * ever sees it.
+ *
+ * The deficit is spent from the right, which is where 儿化 always is. 打哪儿指
+ * 哪儿 is the case that needs it: two syllables short over six characters, and
+ * both 儿 attach to the character in front of them rather than one taking two.
+ */
+export function sourcesOf(word, reading) {
+    const characters = toCharacters(word);
+    if (characters.length === reading.length) {
+        return characters;
+    }
+    // Anything this does not understand is one base for the whole word, which is
+    // always true and merely less useful than naming each syllable's characters.
+    if (characters.length < reading.length || !isErFinal(word)) {
+        return reading.map((_, at) => (at === 0 ? word : undefined));
+    }
+    const sources = [];
+    let deficit = characters.length - reading.length;
+    for (const character of characters) {
+        const previous = sources.at(-1);
+        if (deficit > 0 && previous !== undefined && isErCharacter(character)) {
+            sources[sources.length - 1] = previous + character;
+            deficit--;
+            continue;
+        }
+        sources.push(character);
+    }
+    /* c8 ignore next 3 -- the deficit is 儿 by construction, but a source that
+       wrote one somewhere else would otherwise misalign every later syllable */
+    return sources.length === reading.length
+        ? sources
+        : reading.map((_, at) => (at === 0 ? word : undefined));
+}
+/**
+ * Write one decoded word as one piece per syllable.
+ *
+ * Only proper nouns are capitalised here; a sentence capital is applied to the
+ * whole conversion afterwards, since it belongs to whichever run happens to
+ * start the sentence.
+ */
+export function writeWord(reading, confidence, word, written) {
+    if (reading.length === 0) {
+        return [sourcePiece(word.text)];
+    }
+    // A tone number already ends its syllable, raised or not, so `xi1an1` cannot
+    // be misread and the 隔音符号 would only be noise.
+    const isNumbered = written.notation === "numbers" || written.notation === "superscript";
+    const spellings = markWord(reading.map((syllable) => writeSyllable(syllable, written.notation)), isNumbered ? "never" : written.apostrophe);
+    const isCapitalised = word.isProperNoun && written.capitals !== "none";
+    const sources = sourcesOf(word.text, reading);
+    return spellings.map((spelling, at) => ({
+        text: at === 0 && isCapitalised ? capitaliseWord(spelling) : spelling,
+        syllable: reading[at],
+        confidence: confidence[at],
+        source: sources[at],
+    }));
+}
+//# sourceMappingURL=pieces.js.map

package/dist/decode/pieces.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"pieces.js","sourceRoot":"","sources":["../../src/decode/pieces.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,QAAQ,EAAE,MAAM,8BAA8B,CAAC;AACxD,OAAO,EAAE,cAAc,EAAE,MAAM,4BAA4B,CAAC;AAS5D,OAAO,EAAE,aAAa,EAAE,SAAS,EAAE,MAAM,wBAAwB,CAAC;AAClE,OAAO,EAAE,aAAa,EAAE,MAAM,wBAAwB,CAAC;AACvD,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAEvD,OAAO,EAAiB,aAAa,EAAE,MAAM,yBAAyB,CAAC;AAEvE;;;;;;;;;;;;;GAaG;AACH,MAAM,UAAU,UAAU,CACxB,UAAsB,EACtB,IAAiB,EACjB,MAAc;IAEd,IAAI,MAAM,KAAK,OAAO,EAAE,CAAC;QACvB,OAAO,IAAI,CAAC,OAAO,CAAC;IACtB,CAAC;IACD,MAAM,KAAK,GAAG,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAC3C,IAAI,KAAK,EAAE,aAAa,KAAK,SAAS,EAAE,CAAC;QACvC,OAAO,IAAI,CAAC,OAAO,CAAC;IACtB,CAAC;IACD,OAAO,aAAa,CAAC,KAAK,CAAC,OAAO,EAAE,IAAI,CAAC,OAAO,CAAC;QAC/C,CAAC,CAAC,KAAK,CAAC,aAAa;QACrB,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC;AACnB,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,UAAU,CAAC,IAAY;IACrC,OAAO;QACL,IAAI;QACJ,QAAQ,EAAE,SAAS;QACnB,UAAU,EAAE,SAAS;QACrB,MAAM,EAAE,SAAS;KAClB,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;;;;GAiBG;AACH,MAAM,UAAU,WAAW,CAAC,IAAY;IACtC,OAAO;QACL,IAAI;QACJ,QAAQ,EAAE,SAAS;QACnB,UAAU,EAAE,SAAS;QACrB,MAAM,EAAE,IAAI;KACb,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;GAcG;AACH,MAAM,UAAU,SAAS,CACvB,IAAY,EACZ,OAA4B;IAE5B,MAAM,UAAU,GAAG,YAAY,CAAC,IAAI,CAAC,CAAC;IACtC,IAAI,UAAU,CAAC,MAAM,KAAK,OAAO,CAAC,MAAM,EAAE,CAAC;QACzC,OAAO,UAAU,CAAC;IACpB,CAAC;IACD,6EAA6E;IAC7E,6EAA6E;IAC7E,IAAI,UAAU,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,EAAE,CAAC;QAC3D,OAAO,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC;IAC/D,CAAC;IAED,MAAM,OAAO,GAAa,EAAE,CAAC;IAC7B,IAAI,OAAO,GAAG,UAAU,CAAC,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC;IACjD,KAAK,MAAM,SAAS,IAAI,UAAU,EAAE,CAAC;QACnC,MAAM,QAAQ,GAAG,OAAO,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;QAChC,IAAI,OAAO,GAAG,CAAC,IAAI,QAAQ,KAAK,SAAS,IAAI,aAAa,CAAC,SAAS,CAAC,EAAE,CAAC;YACtE,OAAO,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC,GAAG,QAAQ,GAAG,SAAS,CAAC;YACnD,OAAO,EAAE,CAAC;YACV,SAAS;QACX,CAAC;QACD,OAAO,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;IAC1B,CAAC;IACD;gFAC4E;IAC5E,OAAO,OAAO,CAAC,MAAM,KAAK,OAAO,CAAC,MAAM;QACtC,CAAC,CAAC,OAAO;QACT,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC;AAC5D,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,SAAS,CACvB,OAA4B,EAC5B,UAAwC,EACxC,IAAiB,EACjB,OAAgB;IAEhB,IAAI,OAAO,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACzB,OAAO,CAAC,WAAW,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC;IAClC,CAAC;IACD,6EAA6E;IAC7E,+CAA+C;IAC/C,MAAM,UAAU,GACd,OAAO,CAAC,QAAQ,KAAK,SAAS,IAAI,OAAO,CAAC,QAAQ,KAAK,aAAa,CAAC;IACvE,MAAM,SAAS,GAAG,QAAQ,CACxB,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,aAAa,CAAC,QAAQ,EAAE,OAAO,CAAC,QAAQ,CAAC,CAAC,EACpE,UAAU,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,OAAO,CAAC,UAAU,CAC1C,CAAC;IACF,MAAM,aAAa,GAAG,IAAI,CAAC,YAAY,IAAI,OAAO,CAAC,QAAQ,KAAK,MAAM,CAAC;IACvE,MAAM,OAAO,GAAG,SAAS,CAAC,IAAI,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC;IAE9C,OAAO,SAAS,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,EAAE,EAAE,CAAC,CAAC;QACtC,IAAI,EAAE,EAAE,KAAK,CAAC,IAAI,aAAa,CAAC,CAAC,CAAC,cAAc,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,QAAQ;QACrE,QAAQ,EAAE,OAAO,CAAC,EAAE,CAAC;QACrB,UAAU,EAAE,UAAU,CAAC,EAAE,CAAC;QAC1B,MAAM,EAAE,OAAO,CAAC,EAAE,CAAC;KACpB,CAAC,CAAC,CAAC;AACN,CAAC"}

package/dist/decode/polyphone-rules.d.ts

@@ -0,0 +1,31 @@
+export { PLAYING_TAN } from "./tan-rule.js";
+import { type EdgeRule } from "./rules.js";
+/**
+ * 长 read as `cháng` where an adverb of degree measures it.
+ *
+ * The character is stored `zhǎng` with `cháng` as an alternate, and the data
+ * says so about the character in isolation: Unihan ranks the readings
+ * `zhǎng(1879) cháng(1179)` by corpus count and names `zhǎng` in `kMandarin`.
+ * That ranking is doing real work — 署长, 团长, 公安局长, 总会长, 审理长 and
+ * the rest reach a bare 长 at the end of a title and read it correctly — so the
+ * default is not the thing to change. What is left over is the 长 that is an
+ * adjective, and nothing in the cost model can prefer an alternate at a
+ * position no word covers: 这篇文章不太长 came out `bú tài zhǎng`.
+ *
+ * The context that separates them is one-sided, as it was for the modal 得. A
+ * 长 that is *growing* is a verb, and a degree adverb cannot modify one — there
+ * is no 很长 meaning it grew a lot — so 很, 太, 最, 多 and their kin in front of
+ * a bare 长 make it the adjective. The other side carries almost nothing: what
+ * follows an adjectival 长 is a noun (很长时间), a particle (最长的大桥), or the
+ * end of the sentence (他的腿很长), which is also what follows half the verbs.
+ *
+ * Measured over the same 88,866 lines of Tatoeba and zh.wikipedia the other
+ * rules were sized against, 260 长 decode as a word of their own and this moves
+ * 75 of them to `cháng`, all 75 correctly. On CPP's 40 hand-labelled 长 it takes
+ * the character from 85.00% to 87.50%, fixing 时间最长 and breaking nothing.
+ *
+ * The other half of the rule is {@link GROWN_INTO}, which pushes the opposite
+ * way on the one shape the sources get wrong.
+ */
+export declare const ADJECTIVAL_CHANG: EdgeRule;
+//# sourceMappingURL=polyphone-rules.d.ts.map

package/dist/decode/polyphone-rules.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"polyphone-rules.d.ts","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAC5C,OAAO,EAAoB,KAAK,QAAQ,EAAgB,MAAM,YAAY,CAAC;AA2H3E;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,eAAO,MAAM,gBAAgB,EAAE,QAgB9B,CAAC"}

package/dist/decode/polyphone-rules.js

@@ -0,0 +1,159 @@
+export { PLAYING_TAN } from "./tan-rule.js";
+import { wordEndingAt } from "./rules.js";
+/**
+ * Whether a syllable is `cháng`, the stative reading of 长.
+ */
+function isChang(syllable) {
+    return (syllable?.initial === "ch" &&
+        syllable.final === "ang" &&
+        syllable.tone === 2);
+}
+/**
+ * Adverbs that measure a quality rather than modify an action.
+ *
+ * The discriminating side of the 长 context, and the reason the set is written
+ * out rather than taken from a tag: jieba calls 很 `zg` and 多 `m`, so the two
+ * commonest members are not adverbs as far as the dictionary is concerned,
+ * while 已经, 也, 却 and 就 — which sit in front of the *verb* 长 — are `d`
+ * along with the ones wanted here. The tag names the wrong set in both
+ * directions, so the set is named directly.
+ *
+ * Everything here is degree or comparison. 还 and 不 are deliberately absent
+ * even though 距离还长 and 那座桥不长 are both `cháng`: they modify verbs as
+ * readily as qualities, and 还长有毛 and 胡子不长在前额上 are the same two
+ * characters around the verb.
+ */
+const DEGREE = new Set([
+    "一样",
+    "十分",
+    "多",
+    "多么",
+    "好",
+    "如此",
+    "尽量",
+    "很",
+    "格外",
+    "极",
+    "极其",
+    "比较",
+    "略",
+    "相当",
+    "真",
+    "稍",
+    "稍微",
+    "越",
+    "越来越",
+    "这么",
+    "这样",
+    "那么",
+    "那样",
+    "更",
+    "最",
+    "有多",
+    "特别",
+    "老",
+    "够",
+    "挺",
+    "太",
+    "非常",
+]);
+/**
+ * What shows a 长 to be growing even with a degree word in front of it.
+ *
+ * {@link DEGREE} already excludes nearly everything that precedes the verb, so
+ * this only has to settle the three shapes reachable from both sets: 真长得很快
+ * and 越长得快, where 得 marks the complement; 长着, where the aspect does; and
+ * 越长越高, where the correlative makes the 长 the thing that increases.
+ *
+ * 了 and 的 are pointedly *not* here, and that is the whole use of conditioning
+ * on the adverb first. Following a bare 长 either would suggest the verb — 长了
+ * 一个大包, 长的很像 — but following an adverb and a 长 they are the sentence
+ * particle and the attributive: 时间太长了 and 很长的道路 are both `cháng`, and
+ * an earlier draft of this rule that guarded on them read both as `zhǎng`.
+ */
+const GROWING = new Set(["得", "着", "越"]);
+/**
+ * What growing produces, as the far half of 越长越X.
+ *
+ * 越长越高, 越长越大, 越长越胖: each names a dimension that increases *because*
+ * something grew, so the 长 in front of it is the verb. The adjectives that go
+ * with a length or a duration are pointedly absent — 时间越长越好, 队伍越长越慢
+ * — because those are the cases the sources already read correctly.
+ */
+const GROWN_INTO = new Set(["高", "大", "胖", "壮", "结实"]);
+/**
+ * Whether an edge is the 越长 of a 越长越X saying something grew.
+ *
+ * This exists because of a defect in the sources rather than a gap in them.
+ * 越长 is a key, read `yuè cháng`, and it is the *only* one of its shape: 越大,
+ * 越高, 越好, 越快 and 越多 are all absent, so 越高越好 decodes as two words and
+ * 越长越高 reaches for a word nothing else in the paradigm has. It carries no
+ * part of speech, which is how the dictionary holds a reading somebody asserted
+ * rather than a word anybody counted — the same category {@link TEACHING_JIAO}
+ * declines to trust — and it comes from one source, `large_pinyin.txt`, with
+ * CC-CEDICT holding no such entry.
+ *
+ * Taking the edge away leaves the character's own default, which is `zhǎng`,
+ * and leaves `cháng` standing as a rival about one bucket dearer. That last
+ * part is the point of forbidding rather than forcing: 越长越X is genuinely
+ * ambiguous — 孩子越长越漂亮 grows and 头发越长越漂亮 lengthens, and only the
+ * subject says which — so the decode should answer with the likelier reading
+ * and still report itself as guessing. A forced edge could not.
+ *
+ * Unlike the rest of this file the shape has no corpus behind it: 越长 occurs
+ * three times in the 88,866 lines, and all three are 越来越长 or 说的越长. The
+ * word list is therefore small on purpose, and everything outside it is left as
+ * the sources have it.
+ */
+function isGrowthCorrelative(context) {
+    const { characters, edge } = context;
+    return (edge.text === "越长" &&
+        edge.partOfSpeech === "" &&
+        characters[edge.to] === "越" &&
+        GROWN_INTO.has(characters[edge.to + 1] ?? ""));
+}
+/**
+ * 长 read as `cháng` where an adverb of degree measures it.
+ *
+ * The character is stored `zhǎng` with `cháng` as an alternate, and the data
+ * says so about the character in isolation: Unihan ranks the readings
+ * `zhǎng(1879) cháng(1179)` by corpus count and names `zhǎng` in `kMandarin`.
+ * That ranking is doing real work — 署长, 团长, 公安局长, 总会长, 审理长 and
+ * the rest reach a bare 长 at the end of a title and read it correctly — so the
+ * default is not the thing to change. What is left over is the 长 that is an
+ * adjective, and nothing in the cost model can prefer an alternate at a
+ * position no word covers: 这篇文章不太长 came out `bú tài zhǎng`.
+ *
+ * The context that separates them is one-sided, as it was for the modal 得. A
+ * 长 that is *growing* is a verb, and a degree adverb cannot modify one — there
+ * is no 很长 meaning it grew a lot — so 很, 太, 最, 多 and their kin in front of
+ * a bare 长 make it the adjective. The other side carries almost nothing: what
+ * follows an adjectival 长 is a noun (很长时间), a particle (最长的大桥), or the
+ * end of the sentence (他的腿很长), which is also what follows half the verbs.
+ *
+ * Measured over the same 88,866 lines of Tatoeba and zh.wikipedia the other
+ * rules were sized against, 260 长 decode as a word of their own and this moves
+ * 75 of them to `cháng`, all 75 correctly. On CPP's 40 hand-labelled 长 it takes
+ * the character from 85.00% to 87.50%, fixing 时间最长 and breaking nothing.
+ *
+ * The other half of the rule is {@link GROWN_INTO}, which pushes the opposite
+ * way on the one shape the sources get wrong.
+ */
+export const ADJECTIVAL_CHANG = {
+    name: "adjectival-chang",
+    verdictFor: (context) => {
+        const { characters, edge } = context;
+        if (isGrowthCorrelative(context)) {
+            return "forbid";
+        }
+        if (edge.text !== "长" || !isChang(edge.reading[0])) {
+            return "keep";
+        }
+        if (GROWING.has(characters[edge.to] ?? "")) {
+            return "keep";
+        }
+        const before = wordEndingAt(context, edge.from);
+        return before !== undefined && DEGREE.has(before) ? "force" : "keep";
+    },
+};
+//# sourceMappingURL=polyphone-rules.js.map

package/dist/decode/polyphone-rules.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"polyphone-rules.js","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC;AAC5C,OAAO,EAAmC,YAAY,EAAE,MAAM,YAAY,CAAC;AAE3E;;GAEG;AACH,SAAS,OAAO,CAAC,QAA8B;IAC7C,OAAO,CACL,QAAQ,EAAE,OAAO,KAAK,IAAI;QAC1B,QAAQ,CAAC,KAAK,KAAK,KAAK;QACxB,QAAQ,CAAC,IAAI,KAAK,CAAC,CACpB,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;GAcG;AACH,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC;IACrB,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,IAAI;IACJ,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,IAAI;IACJ,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,IAAI;IACJ,GAAG;IACH,GAAG;IACH,IAAI;IACJ,GAAG;IACH,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,IAAI;CACL,CAAC,CAAC;AAEH;;;;;;;;;;;;;GAaG;AACH,MAAM,OAAO,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAEzC;;;;;;;GAOG;AACH,MAAM,UAAU,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,IAAI,CAAC,CAAC,CAAC;AAEvD;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,SAAS,mBAAmB,CAAC,OAAoB;IAC/C,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;IACrC,OAAO,CACL,IAAI,CAAC,IAAI,KAAK,IAAI;QAClB,IAAI,CAAC,YAAY,KAAK,EAAE;QACxB,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,KAAK,GAAG;QAC3B,UAAU,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,CAC9C,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,MAAM,CAAC,MAAM,gBAAgB,GAAa;IACxC,IAAI,EAAE,kBAAkB;IACxB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,IAAI,mBAAmB,CAAC,OAAO,CAAC,EAAE,CAAC;YACjC,OAAO,QAAQ,CAAC;QAClB,CAAC;QACD,IAAI,IAAI,CAAC,IAAI,KAAK,GAAG,IAAI,CAAC,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACnD,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,IAAI,OAAO,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;YAC3C,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;QAChD,OAAO,MAAM,KAAK,SAAS,IAAI,MAAM,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC;IACvE,CAAC;CACF,CAAC"}

package/dist/decode/reading-rules.d.ts

@@ -1,3 +1,4 @@
+export { TEACHING_JIAO } from "./jiao-rule.js";
 import { type EdgeRule } from "./rules.js";
 /**
  * 得 read as `děi` where it is a modal rather than a particle.
@@ -49,34 +50,6 @@ export declare const MODAL_DE: EdgeRule;
  * 门铃儿 and a row of Mongol names transliterated with 儿.
  */
 export declare const ATTESTED_ERHUA: EdgeRule;
-/**
- * 教 read as `jiāo` where it is teaching rather than a religion.
- *
- * The dictionary stores 教 as `jiào` with `jiāo` as an alternate, and `jiào` is
- * right for the compounds — 教育, 教师, 宗教, 主教 — which are words and reach
- * their reading through the word. What is left is the 教 that stands as a word
- * of its own, and that one is the verb: 他在北京大学教了三年书 came out
- * `jiàole sān nián shū`, and 我教英语, 谁教你法语 and 她教我如何游泳 with it.
- * 教书 is a word and was always `jiāo shū`; nothing carried that across a 了.
- *
- * See {@link isTeachingAt} for what decides it. Forcing the single-character
- * edge is not enough on its own, because a reading spanning two characters
- * carries its own 教 into the position: 王老师教我们汉语 read `jiào` off 师教,
- * and 来教我 off 来教. Those are pairs the dictionary holds with no part of
- * speech and at the cost of a word nothing has ever counted — a reading rather
- * than a word — so a *tagged* word ending in 教 is left alone, which is every
- * one that matters: 任教, 宗教, 主教, 佛教, 传教, 执教, 请教, 家教.
- *
- * Measured over 88,866 lines of Tatoeba and zh.wikipedia, 181 教 decode as a
- * word of their own and every one of them read `jiào`. This moves 158 to
- * `jiāo` and is wrong on three: 统一教创始人, 方法教深思 and 做到了教政分离,
- * where a nominal compound takes an object's shape, the last of them because
- * 做到了 is a word and so the 了 in front is not a particle to look at. Of the
- * 21 it leaves, 8 are right — 诸教中, 教外别传, 董教总, 风教, 教青局 twice, a
- * 教室 reached through a bad split, and a 教 used as a bare noun — and 13 are
- * misses, where jieba tags the object something this does not name.
- */
-export declare const TEACHING_JIAO: EdgeRule;
 /**
  * A 量词 with a number in front of it does not start the next word.
  *
@@ -106,75 +79,7 @@ export declare const TEACHING_JIAO: EdgeRule;
  * corpus and it is a fix, 下了两天雨 having read 天雨 as `tiān yù`.
  */
 export declare const COUNTED_MEASURE: EdgeRule;
-/**
- * 长 read as `cháng` where an adverb of degree measures it.
- *
- * The character is stored `zhǎng` with `cháng` as an alternate, and the data
- * says so about the character in isolation: Unihan ranks the readings
- * `zhǎng(1879) cháng(1179)` by corpus count and names `zhǎng` in `kMandarin`.
- * That ranking is doing real work — 署长, 团长, 公安局长, 总会长, 审理长 and
- * the rest reach a bare 长 at the end of a title and read it correctly — so the
- * default is not the thing to change. What is left over is the 长 that is an
- * adjective, and nothing in the cost model can prefer an alternate at a
- * position no word covers: 这篇文章不太长 came out `bú tài zhǎng`.
- *
- * The context that separates them is one-sided, as it was for the modal 得. A
- * 长 that is *growing* is a verb, and a degree adverb cannot modify one — there
- * is no 很长 meaning it grew a lot — so 很, 太, 最, 多 and their kin in front of
- * a bare 长 make it the adjective. The other side carries almost nothing: what
- * follows an adjectival 长 is a noun (很长时间), a particle (最长的大桥), or the
- * end of the sentence (他的腿很长), which is also what follows half the verbs.
- *
- * Measured over the same 88,866 lines of Tatoeba and zh.wikipedia the other
- * rules were sized against, 260 长 decode as a word of their own and this moves
- * 75 of them to `cháng`, all 75 correctly. On CPP's 40 hand-labelled 长 it takes
- * the character from 85.00% to 87.50%, fixing 时间最长 and breaking nothing.
- *
- * The other half of the rule is {@link GROWN_INTO}, which pushes the opposite
- * way on the one shape the sources get wrong.
- */
-export declare const ADJECTIVAL_CHANG: EdgeRule;
-/**
- * 弹 read as `tán` where it is playing rather than a projectile.
- *
- * The character is stored `dàn` with `tán` as an alternate, and the sources are
- * agreed about it: Unihan's `kHanyuPinlu` counts `dàn(313)` against `tán(50)`
- * and `kMandarin` names `dàn`. That is a fact about a corpus in which nearly
- * every 弹 is ammunition — 子弹, 炮弹, 导弹, 原子弹, 手榴弹 — and every one of
- * those is a word that carries its own reading. What the default is left
- * deciding is the 弹 that stands as a word of its own, and in running text that
- * one is the verb: 他会弹一点儿古筝 came out `dàn`, and 弹吉他, 弹竖琴, 我在弹肖邦
- * and 他钢琴弹得很好 with it.
- *
- * See {@link isPlayingAt} for what decides it. Forcing the single-character
- * edge is not enough on its own, for the reason {@link TEACHING_JIAO} gives: a
- * reading spanning two characters carries its own 弹 into the position, and
- * 我的爱好是开车和弹吉他 read `dàn` off 和弹, a pair the dictionary holds with
- * no part of speech and at the cost of a reading nobody counted. A *tagged*
- * word ending in 弹 is left alone, which is every one that matters — 子弹, 炸弹,
- * 导弹, 反弹, 手榴弹. Nothing has to be done about the other side, since the
- * nominal compounds starting with 弹 are all listed: 弹匣, 弹坑, 弹壳, 弹片,
- * 弹药, 弹道 and 弹头 reach their reading through the word, so a bare 弹 with a
- * noun after it is not the front of one that got away.
- *
- * The 88,866 lines of Tatoeba and zh.wikipedia the other rules were sized
- * against are the wrong corpus for this one on their own: 38 弹 decode as a
- * word of their own in them, and 31 are somebody playing something. So CPP's
- * 20,147 sentences are measured with them as plain text — the benchmark is
- * drawn from zh.wikipedia's military articles, which is precisely where the
- * shapes this rule can break live. Over the 109,013 lines together, 60 弹
- * decode as a word of their own, every one read `dàn`, and 34 of the 60 are
- * wrong.
- *
- * This moves 30 of them, of which 29 are right: the whole of 弹吉他, 弹一曲肖邦
- * and 钢琴弹得很好, and none of 南部弹, 毛瑟弹, 掷弹兵, 供弹爪, 着弹点, 弹洞 or
- * 底排弹时. The one it breaks is 拆弹专家, where a compound has an object's
- * shape on both sides. Five it leaves alone: 开始弹, 四手联弹 and 弹起三次, which
- * have no object to see, 用那种指法弹不会觉得费力, and 反手持法去弹班卓琴, where
- * the decode reaches 班 rather than 班卓琴. On CPP's 40 labelled 弹 nothing
- * moves at all, every one of them having been right through a word.
- */
-export declare const PLAYING_TAN: EdgeRule;
+export { ADJECTIVAL_CHANG, PLAYING_TAN } from "./polyphone-rules.js";
 /**
  * The rules the lattice decode applies, in order.
  */

package/dist/decode/reading-rules.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"reading-rules.d.ts","sourceRoot":"","sources":["../../src/decode/reading-rules.ts"],"names":[],"mappings":"AAGA,OAAO,EAEL,KAAK,QAAQ,EAId,MAAM,YAAY,CAAC;AAgCpB;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,eAAO,MAAM,QAAQ,EAAE,QActB,CAAC;AAEF;;;;;;;;;;;;;;;;;;;GAmBG;AACH,eAAO,MAAM,cAAc,EAAE,QAY5B,CAAC;AAqDF;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,eAAO,MAAM,aAAa,EAAE,QAiB3B,CAAC;AAsCF;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,eAAO,MAAM,eAAe,EAAE,QAuB7B,CAAC;AA2HF;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,eAAO,MAAM,gBAAgB,EAAE,QAgB9B,CAAC;AA0EF;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAuCG;AACH,eAAO,MAAM,WAAW,EAAE,QAiBzB,CAAC;AAEF;;GAEG;AACH,eAAO,MAAM,aAAa,EAAE,SAAS,QAAQ,EAO5C,CAAC"}
+{"version":3,"file":"reading-rules.d.ts","sourceRoot":"","sources":["../../src/decode/reading-rules.ts"],"names":[],"mappings":"AAGA,OAAO,EAAE,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAG/C,OAAO,EAEL,KAAK,QAAQ,EAId,MAAM,YAAY,CAAC;AAgCpB;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,eAAO,MAAM,QAAQ,EAAE,QActB,CAAC;AAEF;;;;;;;;;;;;;;;;;;;GAmBG;AACH,eAAO,MAAM,cAAc,EAAE,QAY5B,CAAC;AAsCF;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,eAAO,MAAM,eAAe,EAAE,QAuB7B,CAAC;AAIF,OAAO,EAAE,gBAAgB,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AAErE;;GAEG;AACH,eAAO,MAAM,aAAa,EAAE,SAAS,QAAQ,EAO5C,CAAC"}