@kensio/pinyinjs 1.7.5 → 1.7.6

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

@@ -1 +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"}
+{"version":3,"file":"polyphone-rules.d.ts","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAC;AACnD,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC"}

package/dist/decode/polyphone-rules.js

@@ -1,159 +1,11 @@
-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.
+ * The rules that settle a polyphone from the words around it.
  *
- * 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.
+ * 弹 is `tán` or `dàn`, and which one a text means is a question about the
+ * neighbours rather than about the character. 长 is the same question and has
+ * a file of its own. The other rules in `reading-rules.ts` are about particles
+ * and 儿化, which is a different kind of evidence.
  */
-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";
-    },
-};
+export { ADJECTIVAL_CHANG } from "./chang-rule.js";
+export { PLAYING_TAN } from "./tan-rule.js";
 //# sourceMappingURL=polyphone-rules.js.map

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

@@ -1 +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"}
+{"version":3,"file":"polyphone-rules.js","sourceRoot":"","sources":["../../src/decode/polyphone-rules.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,gBAAgB,EAAE,MAAM,iBAAiB,CAAC;AACnD,OAAO,EAAE,WAAW,EAAE,MAAM,eAAe,CAAC"}

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

@@ -1,34 +1,6 @@
+export { MODAL_DE, PARTICLE_DE } from "./de-rule.js";
 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.
- *
- * One character, three readings, and the dictionary can only carry a default:
- * 得 is stored `de` — the structural particle, which is far the commonest — with
- * `dé` and `děi` as alternates. Nothing in the cost model can prefer an
- * alternate at a position no word covers, so 我得走了 came out `wǒ de zǒule`
- * and every other modal 得 with it. This is the case ROADMAP.md was waiting for
- * before building stage 4: neither harness could see it, since the gold corpus
- * had only 觉得 and CPP's 40 得 cases are labelled `dé` or `de` and never `děi`.
- *
- * The context is what separates the three, and only one side of it carries
- * information. The particle 得 always follows the verb or adjective it attaches
- * to (说得好, 跑得快, 累得要命), so a 得 with a *pronoun, adverb or time word*
- * in front of it is not that particle. Requiring a verb phrase after it as well
- * keeps 得了感冒 and 得了奖 — `dé`, and a real reading of the same character —
- * out of it.
- *
- * Conditioning on what precedes rather than on what follows was measured, not
- * assumed: "the word before is not a verb" fires 309 times over 88,866 lines of
- * Tatoeba and zh.wikipedia and is wrong about four times in five, because
- * jieba tags 下 as `f`, 学 as `n` and 飞 as nothing at all, and 雨下得很大 is a
- * particle whatever the tag says. Naming the small closed set of things that
- * can precede a modal is the same rule from the other side, and over the same
- * text it fires 126 times and is right 122: the four misses are literary 得
- * read `dé` after an adverb — 未必得喜悦, 暂得于己, 仅得约, 未得迁还家乡 —
- * which nothing short of semantics separates from 必须得承认.
- */
-export declare const MODAL_DE: EdgeRule;
 /**
  * 儿 does not stand on its own where the dictionary attests the 儿化.
  *

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,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"}
+{"version":3,"file":"reading-rules.d.ts","sourceRoot":"","sources":["../../src/decode/reading-rules.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AACrD,OAAO,EAAE,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAE/C,OAAO,EAEL,KAAK,QAAQ,EAGd,MAAM,YAAY,CAAC;AAEpB;;;;;;;;;;;;;;;;;;;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,EAQ5C,CAAC"}

package/dist/decode/reading-rules.js

@@ -1,76 +1,10 @@
 import { QUANTITY_CHARACTERS } from "../numerals/characters.js";
+import { MODAL_DE, PARTICLE_DE } from "./de-rule.js";
 import { TEACHING_JIAO } from "./jiao-rule.js";
+export { MODAL_DE, PARTICLE_DE } from "./de-rule.js";
 export { TEACHING_JIAO } from "./jiao-rule.js";
 import { toCharacters } from "../script/characters.js";
-import { tagOf, wordEndingAt, wordStartingAt, } from "./rules.js";
-/**
- * Whether a syllable is `děi`, the modal reading of 得.
- *
- * Compared structurally rather than by spelling it out, since a syllable is
- * stored as its parts and writing one back is the formatting stage's job.
- */
-function isDei(syllable) {
-    return (syllable?.initial === "d" && syllable.final === "ei" && syllable.tone === 3);
-}
-/**
- * What stands in front of a modal 得: a subject, an adverb, or a time word.
- *
- * 我得走, 你就得去, 我今天得走. The particle 得 cannot follow any of these,
- * because it attaches to the predicate it modifies — which is exactly what
- * makes this the discriminating side of the context.
- */
-const SUBJECT_TAGS = new Set(["r", "d", "t"]);
-/**
- * What follows a modal 得: the verb phrase it governs.
- *
- * Prepositions and adverbs are in because 我得在八月, 我得先走 and 我得快点 are
- * all modal, and what follows the 得 in each is the front of a verb phrase
- * rather than a verb.
- */
-const GOVERNED_TAGS = /^(?:v|p|d|a)/u;
-/**
- * 得 read as `děi` where it is a modal rather than a particle.
- *
- * One character, three readings, and the dictionary can only carry a default:
- * 得 is stored `de` — the structural particle, which is far the commonest — with
- * `dé` and `děi` as alternates. Nothing in the cost model can prefer an
- * alternate at a position no word covers, so 我得走了 came out `wǒ de zǒule`
- * and every other modal 得 with it. This is the case ROADMAP.md was waiting for
- * before building stage 4: neither harness could see it, since the gold corpus
- * had only 觉得 and CPP's 40 得 cases are labelled `dé` or `de` and never `děi`.
- *
- * The context is what separates the three, and only one side of it carries
- * information. The particle 得 always follows the verb or adjective it attaches
- * to (说得好, 跑得快, 累得要命), so a 得 with a *pronoun, adverb or time word*
- * in front of it is not that particle. Requiring a verb phrase after it as well
- * keeps 得了感冒 and 得了奖 — `dé`, and a real reading of the same character —
- * out of it.
- *
- * Conditioning on what precedes rather than on what follows was measured, not
- * assumed: "the word before is not a verb" fires 309 times over 88,866 lines of
- * Tatoeba and zh.wikipedia and is wrong about four times in five, because
- * jieba tags 下 as `f`, 学 as `n` and 飞 as nothing at all, and 雨下得很大 is a
- * particle whatever the tag says. Naming the small closed set of things that
- * can precede a modal is the same rule from the other side, and over the same
- * text it fires 126 times and is right 122: the four misses are literary 得
- * read `dé` after an adverb — 未必得喜悦, 暂得于己, 仅得约, 未得迁还家乡 —
- * which nothing short of semantics separates from 必须得承认.
- */
-export const MODAL_DE = {
-    name: "modal-de",
-    verdictFor: (context) => {
-        const { edge } = context;
-        if (edge.text !== "得" || !isDei(edge.reading[0])) {
-            return "keep";
-        }
-        const before = wordEndingAt(context, edge.from);
-        const after = wordStartingAt(context, edge.to);
-        return SUBJECT_TAGS.has(tagOf(context, before)) &&
-            GOVERNED_TAGS.test(tagOf(context, after))
-            ? "force"
-            : "keep";
-    },
-};
+import { tagOf, wordEndingAt, } from "./rules.js";
 /**
  * 儿 does not stand on its own where the dictionary attests the 儿化.
  *
@@ -194,6 +128,7 @@ export { ADJECTIVAL_CHANG, PLAYING_TAN } from "./polyphone-rules.js";
  */
 export const READING_RULES = [
     MODAL_DE,
+    PARTICLE_DE,
     TEACHING_JIAO,
     ATTESTED_ERHUA,
     COUNTED_MEASURE,

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

@@ -1 +1 @@
-{"version":3,"file":"reading-rules.js","sourceRoot":"","sources":["../../src/decode/reading-rules.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,mBAAmB,EAAE,MAAM,2BAA2B,CAAC;AAChE,OAAO,EAAE,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAE/C,OAAO,EAAE,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAC/C,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAEvD,OAAO,EAGL,KAAK,EACL,YAAY,EACZ,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB;;;;;GAKG;AACH,SAAS,KAAK,CAAC,QAA8B;IAC3C,OAAO,CACL,QAAQ,EAAE,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,IAAI,IAAI,QAAQ,CAAC,IAAI,KAAK,CAAC,CAC5E,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,YAAY,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAE9C;;;;;;GAMG;AACH,MAAM,aAAa,GAAG,eAAe,CAAC;AAEtC;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,MAAM,CAAC,MAAM,QAAQ,GAAa;IAChC,IAAI,EAAE,UAAU;IAChB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,IAAI,IAAI,CAAC,IAAI,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACjD,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;QAChD,MAAM,KAAK,GAAG,cAAc,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,CAAC;QAC/C,OAAO,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;YAC7C,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;YACzC,CAAC,CAAC,OAAO;YACT,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC;AAEF;;;;;;;;;;;;;;;;;;;GAmBG;AACH,MAAM,CAAC,MAAM,cAAc,GAAa;IACtC,IAAI,EAAE,gBAAgB;IACtB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,IAAI,IAAI,CAAC,IAAI,KAAK,GAAG,IAAI,IAAI,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC;YACzC,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,wEAAwE;QACxE,MAAM,QAAQ,GAAG,GAAG,UAAU,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,GAAG,CAAC;QACvD,MAAM,OAAO,GAAG,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,QAAQ,CAAC,EAAE,OAAO,CAAC;QAC7D,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,IAAI,KAAK,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC;IAC/D,CAAC;CACF,CAAC;AAEF;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;GAKG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;;;;GAQG;AACH,SAAS,WAAW,CAAC,OAAoB,EAAE,EAAU;IACnD,MAAM,EAAE,UAAU,EAAE,GAAG,OAAO,CAAC;IAC/B,IAAI,IAAI,GAAG,EAAE,CAAC;IACd,OAAO,IAAI,GAAG,CAAC,IAAI,mBAAmB,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;QACvE,IAAI,EAAE,CAAC;IACT,CAAC;IACD,IAAI,IAAI,KAAK,EAAE,IAAI,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,KAAK,OAAO,EAAE,CAAC;QACpD,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;IACzC,MAAM,IAAI,GAAG,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC;IAC1E,OAAO,IAAI,GAAG,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC;AACxC,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,MAAM,CAAC,MAAM,eAAe,GAAa;IACvC,IAAI,EAAE,iBAAiB;IACvB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,yEAAyE;QACzE,yEAAyE;QACzE,yEAAyE;QACzE,WAAW;QACX,IACE,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,GAAG,CAAC;YACvB,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,EAC3C,CAAC;YACD,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,OAAO,GAAG,WAAW,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;QAChD,IAAI,OAAO,KAAK,SAAS,EAAE,CAAC;YAC1B,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,OAAO,KAAK,CAAC,OAAO,EAAE,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,KAAK,OAAO;YACtD,IAAI,CAAC,YAAY,KAAK,OAAO;YAC7B,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC;AAEF,OAAO,EAAE,gBAAgB,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AAErE,OAAO,EAAE,gBAAgB,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AAErE;;GAEG;AACH,MAAM,CAAC,MAAM,aAAa,GAAwB;IAChD,QAAQ;IACR,aAAa;IACb,cAAc;IACd,eAAe;IACf,gBAAgB;IAChB,WAAW;CACZ,CAAC"}
+{"version":3,"file":"reading-rules.js","sourceRoot":"","sources":["../../src/decode/reading-rules.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,mBAAmB,EAAE,MAAM,2BAA2B,CAAC;AAChE,OAAO,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AACrD,OAAO,EAAE,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAE/C,OAAO,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AACrD,OAAO,EAAE,aAAa,EAAE,MAAM,gBAAgB,CAAC;AAC/C,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAGL,KAAK,EACL,YAAY,GACb,MAAM,YAAY,CAAC;AAEpB;;;;;;;;;;;;;;;;;;;GAmBG;AACH,MAAM,CAAC,MAAM,cAAc,GAAa;IACtC,IAAI,EAAE,gBAAgB;IACtB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,IAAI,IAAI,CAAC,IAAI,KAAK,GAAG,IAAI,IAAI,CAAC,IAAI,KAAK,CAAC,EAAE,CAAC;YACzC,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,wEAAwE;QACxE,MAAM,QAAQ,GAAG,GAAG,UAAU,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,GAAG,CAAC;QACvD,MAAM,OAAO,GAAG,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,QAAQ,CAAC,EAAE,OAAO,CAAC;QAC7D,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,IAAI,KAAK,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC;IAC/D,CAAC;CACF,CAAC;AAEF;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;GAKG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;;;;GAQG;AACH,SAAS,WAAW,CAAC,OAAoB,EAAE,EAAU;IACnD,MAAM,EAAE,UAAU,EAAE,GAAG,OAAO,CAAC;IAC/B,IAAI,IAAI,GAAG,EAAE,CAAC;IACd,OAAO,IAAI,GAAG,CAAC,IAAI,mBAAmB,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;QACvE,IAAI,EAAE,CAAC;IACT,CAAC;IACD,IAAI,IAAI,KAAK,EAAE,IAAI,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,KAAK,OAAO,EAAE,CAAC;QACpD,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;IACzC,MAAM,IAAI,GAAG,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC;IAC1E,OAAO,IAAI,GAAG,IAAI,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC;AACxC,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,MAAM,CAAC,MAAM,eAAe,GAAa;IACvC,IAAI,EAAE,iBAAiB;IACvB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,yEAAyE;QACzE,yEAAyE;QACzE,yEAAyE;QACzE,WAAW;QACX,IACE,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,GAAG,CAAC;YACvB,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,EAC3C,CAAC;YACD,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,OAAO,GAAG,WAAW,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;QAChD,IAAI,OAAO,KAAK,SAAS,EAAE,CAAC;YAC1B,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,OAAO,KAAK,CAAC,OAAO,EAAE,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,KAAK,OAAO;YACtD,IAAI,CAAC,YAAY,KAAK,OAAO;YAC7B,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC;AAEF,OAAO,EAAE,gBAAgB,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AAErE,OAAO,EAAE,gBAAgB,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AAErE;;GAEG;AACH,MAAM,CAAC,MAAM,aAAa,GAAwB;IAChD,QAAQ;IACR,WAAW;IACX,aAAa;IACb,cAAc;IACd,eAAe;IACf,gBAAgB;IAChB,WAAW;CACZ,CAAC"}

package/dist/decode/run-words.d.ts

@@ -0,0 +1,17 @@
+/**
+ * Recutting a run's words for 分词连写.
+ *
+ * Grouping rewrites word boundaries and never the readings behind them, so the
+ * syllables — and the confidence beside them — survive it in order, and are
+ * simply cut in different places.
+ */
+import type { Dictionary } from "../dictionary/dictionary.js";
+import type { ScoredWord } from "./word.js";
+/**
+ * The words a Han run decodes to, with 分词连写 applied.
+ *
+ * Grouping rewrites word boundaries and never the readings behind them, so the
+ * syllables — and the confidence beside them — survive it in order.
+ */
+export declare function wordsOf(decoded: readonly ScoredWord[], dictionary: Dictionary, isGrouped: boolean): readonly ScoredWord[];
+//# sourceMappingURL=run-words.d.ts.map

package/dist/decode/run-words.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"run-words.d.ts","sourceRoot":"","sources":["../../src/decode/run-words.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D,OAAO,KAAK,EAAe,UAAU,EAAE,MAAM,WAAW,CAAC;AAEzD;;;;;GAKG;AACH,wBAAgB,OAAO,CACrB,OAAO,EAAE,SAAS,UAAU,EAAE,EAC9B,UAAU,EAAE,UAAU,EACtB,SAAS,EAAE,OAAO,GACjB,SAAS,UAAU,EAAE,CAUvB"}

package/dist/decode/run-words.js

@@ -0,0 +1,29 @@
+import { applyGrouping } from "../orthography/grouping.js";
+/**
+ * The words a Han run decodes to, with 分词连写 applied.
+ *
+ * Grouping rewrites word boundaries and never the readings behind them, so the
+ * syllables — and the confidence beside them — survive it in order.
+ */
+export function wordsOf(decoded, dictionary, isGrouped) {
+    return isGrouped
+        ? regroup(decoded, applyGrouping(decoded.map((scored) => scored.word), dictionary))
+        : decoded;
+}
+/**
+ * Redistribute the decode's per-syllable confidence over regrouped words.
+ *
+ * 分词连写 moves word boundaries — 看 and 了 become 看了 — without touching a
+ * reading, so the run's syllables are the same syllables in the same order and
+ * are simply cut in different places.
+ */
+function regroup(decoded, grouped) {
+    const confidence = decoded.flatMap((scored) => [...scored.confidence]);
+    let at = 0;
+    return grouped.map((word) => {
+        const held = confidence.slice(at, at + word.reading.length);
+        at += word.reading.length;
+        return { word, confidence: held };
+    });
+}
+//# sourceMappingURL=run-words.js.map

package/dist/decode/run-words.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"run-words.js","sourceRoot":"","sources":["../../src/decode/run-words.ts"],"names":[],"mappings":"AAQA,OAAO,EAAE,aAAa,EAAE,MAAM,4BAA4B,CAAC;AAG3D;;;;;GAKG;AACH,MAAM,UAAU,OAAO,CACrB,OAA8B,EAC9B,UAAsB,EACtB,SAAkB;IAElB,OAAO,SAAS;QACd,CAAC,CAAC,OAAO,CACL,OAAO,EACP,aAAa,CACX,OAAO,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,EACpC,UAAU,CACX,CACF;QACH,CAAC,CAAC,OAAO,CAAC;AACd,CAAC;AAED;;;;;;GAMG;AACH,SAAS,OAAO,CACd,OAA8B,EAC9B,OAA+B;IAE/B,MAAM,UAAU,GAAG,OAAO,CAAC,OAAO,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC,GAAG,MAAM,CAAC,UAAU,CAAC,CAAC,CAAC;IACvE,IAAI,EAAE,GAAG,CAAC,CAAC;IACX,OAAO,OAAO,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE;QAC1B,MAAM,IAAI,GAAG,UAAU,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,GAAG,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;QAC5D,EAAE,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC;QAC1B,OAAO,EAAE,IAAI,EAAE,UAAU,EAAE,IAAI,EAAE,CAAC;IACpC,CAAC,CAAC,CAAC;AACL,CAAC"}

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

@@ -0,0 +1,30 @@
+/**
+ * What the sandhi pass needs from a decoded run, beyond the syllables.
+ *
+ * Two questions sandhi cannot answer from a reading alone. Third-tone sandhi
+ * applies inside a prosodic foot, so it has to be told where the words and
+ * their parts are. 一 sandhi asks what stands in front of the 一, and the answer
+ * is a character rather than a spelling. Both are things the decode knows and
+ * the syllable array does not.
+ */
+import type { Dictionary } from "../dictionary/dictionary.js";
+import type { Syllable } from "../syllable/syllable.js";
+import type { SandhiGrouping } from "./sandhi.js";
+import type { ScoredWord } from "./word.js";
+/**
+ * Where the words and their constituents fall, for third-tone sandhi.
+ *
+ * Undefined unless third-tone sandhi was asked for, since dividing a word costs
+ * dictionary lookups and nothing else reads the answer.
+ */
+export declare function groupingOf(dictionary: Dictionary, words: readonly ScoredWord[], readings: readonly (readonly Syllable[])[], isThirdTone: boolean): SandhiGrouping | undefined;
+/**
+ * One 汉字 per syllable, where a word offers one.
+ *
+ * A word whose reading is a different length from its text — 玩儿 as `wánr` —
+ * has no character to give any one of its syllables, so its syllables get none
+ * and the pass falls back to their spellings. See
+ * {@link import("./sandhi-tones.js").isCounting}.
+ */
+export declare function charactersPerSyllable(words: readonly ScoredWord[], readings: readonly (readonly Syllable[])[]): readonly (string | undefined)[];
+//# sourceMappingURL=sandhi-input.d.ts.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"sandhi-input.d.ts","sourceRoot":"","sources":["../../src/decode/sandhi-input.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAE9D,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAClD,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AAE5C;;;;;GAKG;AACH,wBAAgB,UAAU,CACxB,UAAU,EAAE,UAAU,EACtB,KAAK,EAAE,SAAS,UAAU,EAAE,EAC5B,QAAQ,EAAE,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,EAC1C,WAAW,EAAE,OAAO,GACnB,cAAc,GAAG,SAAS,CAQ5B;AAED;;;;;;;GAOG;AACH,wBAAgB,qBAAqB,CACnC,KAAK,EAAE,SAAS,UAAU,EAAE,EAC5B,QAAQ,EAAE,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,GACzC,SAAS,CAAC,MAAM,GAAG,SAAS,CAAC,EAAE,CAQjC"}

package/dist/decode/sandhi-input.js

@@ -0,0 +1,35 @@
+import { toCharacters } from "../script/characters.js";
+import { divisionOf } from "./constituents.js";
+/**
+ * Where the words and their constituents fall, for third-tone sandhi.
+ *
+ * Undefined unless third-tone sandhi was asked for, since dividing a word costs
+ * dictionary lookups and nothing else reads the answer.
+ */
+export function groupingOf(dictionary, words, readings, isThirdTone) {
+    if (!isThirdTone) {
+        return undefined;
+    }
+    return words.map((scored, index) => {
+        const reading = readings[index] ?? [];
+        return divisionOf(dictionary, scored.word.text, reading) ?? reading.length;
+    });
+}
+/**
+ * One 汉字 per syllable, where a word offers one.
+ *
+ * A word whose reading is a different length from its text — 玩儿 as `wánr` —
+ * has no character to give any one of its syllables, so its syllables get none
+ * and the pass falls back to their spellings. See
+ * {@link import("./sandhi-tones.js").isCounting}.
+ */
+export function charactersPerSyllable(words, readings) {
+    return words.flatMap((scored, index) => {
+        const reading = readings[index] ?? [];
+        const held = toCharacters(scored.word.text);
+        return held.length === reading.length
+            ? held
+            : Array.from({ length: reading.length });
+    });
+}
+//# sourceMappingURL=sandhi-input.js.map

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

@@ -0,0 +1 @@
+{"version":3,"file":"sandhi-input.js","sourceRoot":"","sources":["../../src/decode/sandhi-input.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAEvD,OAAO,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAC;AAI/C;;;;;GAKG;AACH,MAAM,UAAU,UAAU,CACxB,UAAsB,EACtB,KAA4B,EAC5B,QAA0C,EAC1C,WAAoB;IAEpB,IAAI,CAAC,WAAW,EAAE,CAAC;QACjB,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,KAAK,EAAE,EAAE;QACjC,MAAM,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC;QACtC,OAAO,UAAU,CAAC,UAAU,EAAE,MAAM,CAAC,IAAI,CAAC,IAAI,EAAE,OAAO,CAAC,IAAI,OAAO,CAAC,MAAM,CAAC;IAC7E,CAAC,CAAC,CAAC;AACL,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,qBAAqB,CACnC,KAA4B,EAC5B,QAA0C;IAE1C,OAAO,KAAK,CAAC,OAAO,CAAC,CAAC,MAAM,EAAE,KAAK,EAAmC,EAAE;QACtE,MAAM,OAAO,GAAG,QAAQ,CAAC,KAAK,CAAC,IAAI,EAAE,CAAC;QACtC,MAAM,IAAI,GAAG,YAAY,CAAC,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAC5C,OAAO,IAAI,CAAC,MAAM,KAAK,OAAO,CAAC,MAAM;YACnC,CAAC,CAAC,IAAI;YACN,CAAC,CAAC,KAAK,CAAC,IAAI,CAAqB,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,CAAC,CAAC;IACjE,CAAC,CAAC,CAAC;AACL,CAAC"}

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

@@ -47,6 +47,28 @@ export declare const CONTINUES_NUMBER: Set<string>;
  * The ordinal prefix 第.
  */
 export declare const ORDINAL_PREFIX = "di";
+/**
+ * The 汉字 {@link NUMERAL_WORDS} names, for a caller that has them.
+ *
+ * The spellings are ambiguous and the characters are not, which is the whole
+ * point of carrying both. 是 is spelt `shi` and stands in front of a 一 in
+ * 那是一条狗, 这是一个好主意 and several hundred sentences like them, and every
+ * one of them lost its sandhi to a 十 that was never there. 时, 事, 使 and 试 do
+ * the same.
+ *
+ * The 繁體 forms are here because a 繁體 text reaches this pass with 繁體
+ * characters. 拾, 佰 and 仟 are the 大写 banking forms, which appear on cheques
+ * and in contracts and count exactly as their everyday spellings do.
+ */
+export declare const NUMERAL_CHARACTERS: Set<string>;
+/**
+ * The 汉字 {@link CONTINUES_NUMBER} names.
+ */
+export declare const CONTINUES_NUMBER_CHARACTERS: Set<string>;
+/**
+ * The ordinal prefix as it is written.
+ */
+export declare const ORDINAL_CHARACTER = "\u7B2C";
 /**
  * Whether a 一 is counting something, which is the only time it takes sandhi.
  *
@@ -60,10 +82,16 @@ export declare const ORDINAL_PREFIX = "di";
  *   is counting the ten.
  * - **An ordinal.** 第一个 is `dìyī gè`.
  *
- * **Read off the syllables rather than the characters**, because that is all
- * this pass has: `pinyinjs sandhi shíyī gè` never sees a hanzi. So it cannot
- * tell 十 from 时 or 第 from 地, and `docs/sandhi/` carries what that costs —
- * measured, not estimated.
+ * **Answered from the characters wherever the caller has them**, and from the
+ * spellings where it does not. `pinyinjs sandhi shíyī gè` is handed nothing but
+ * pinyin, so there the pass still cannot tell 十 from 时 or 第 from 地 and
+ * `docs/sandhi/` carries what that costs. A conversion knows which 汉字 it read,
+ * and passing them along is what stops 那是一条狗 losing its sandhi to the 是 in
+ * front of the 一.
+ *
+ * The two sides are decided separately, because a run can know one and not the
+ * other: a word whose reading has a different syllable count from its character
+ * count has no character to offer for any one syllable.
  */
-export declare function isCounting(syllables: readonly Syllable[], at: number): boolean;
+export declare function isCounting(syllables: readonly Syllable[], at: number, characters?: readonly (string | undefined)[]): boolean;
 //# sourceMappingURL=sandhi-tones.d.ts.map

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

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

package/dist/decode/sandhi-tones.js

@@ -56,6 +56,41 @@ export const CONTINUES_NUMBER = new Set([...NUMERAL_WORDS, "ling"]);
  * The ordinal prefix 第.
  */
 export const ORDINAL_PREFIX = "di";
+/**
+ * The 汉字 {@link NUMERAL_WORDS} names, for a caller that has them.
+ *
+ * The spellings are ambiguous and the characters are not, which is the whole
+ * point of carrying both. 是 is spelt `shi` and stands in front of a 一 in
+ * 那是一条狗, 这是一个好主意 and several hundred sentences like them, and every
+ * one of them lost its sandhi to a 十 that was never there. 时, 事, 使 and 试 do
+ * the same.
+ *
+ * The 繁體 forms are here because a 繁體 text reaches this pass with 繁體
+ * characters. 拾, 佰 and 仟 are the 大写 banking forms, which appear on cheques
+ * and in contracts and count exactly as their everyday spellings do.
+ */
+export const NUMERAL_CHARACTERS = new Set([
+    "十",
+    "百",
+    "千",
+    "万",
+    "萬",
+    "拾",
+    "佰",
+    "仟",
+]);
+/**
+ * The 汉字 {@link CONTINUES_NUMBER} names.
+ */
+export const CONTINUES_NUMBER_CHARACTERS = new Set([
+    ...NUMERAL_CHARACTERS,
+    "零",
+    "〇",
+]);
+/**
+ * The ordinal prefix as it is written.
+ */
+export const ORDINAL_CHARACTER = "第";
 /**
  * Whether a 一 is counting something, which is the only time it takes sandhi.
  *
@@ -69,20 +104,36 @@ export const ORDINAL_PREFIX = "di";
  *   is counting the ten.
  * - **An ordinal.** 第一个 is `dìyī gè`.
  *
- * **Read off the syllables rather than the characters**, because that is all
- * this pass has: `pinyinjs sandhi shíyī gè` never sees a hanzi. So it cannot
- * tell 十 from 时 or 第 from 地, and `docs/sandhi/` carries what that costs —
- * measured, not estimated.
+ * **Answered from the characters wherever the caller has them**, and from the
+ * spellings where it does not. `pinyinjs sandhi shíyī gè` is handed nothing but
+ * pinyin, so there the pass still cannot tell 十 from 时 or 第 from 地 and
+ * `docs/sandhi/` carries what that costs. A conversion knows which 汉字 it read,
+ * and passing them along is what stops 那是一条狗 losing its sandhi to the 是 in
+ * front of the 一.
+ *
+ * The two sides are decided separately, because a run can know one and not the
+ * other: a word whose reading has a different syllable count from its character
+ * count has no character to offer for any one syllable.
  */
-export function isCounting(syllables, at) {
+export function isCounting(syllables, at, characters = []) {
     const spelt = (index) => {
         const syllable = syllables[index];
         return syllable === undefined ? "" : writeSyllableSpelling(syllable);
     };
-    const before = spelt(at - 1);
-    if (before === ORDINAL_PREFIX) {
+    const before = characters[at - 1];
+    if (before === undefined
+        ? spelt(at - 1) === ORDINAL_PREFIX
+        : before === ORDINAL_CHARACTER) {
         return false;
     }
-    return !NUMERAL_WORDS.has(before) || CONTINUES_NUMBER.has(spelt(at + 1));
+    if (before === undefined
+        ? !NUMERAL_WORDS.has(spelt(at - 1))
+        : !NUMERAL_CHARACTERS.has(before)) {
+        return true;
+    }
+    const after = characters[at + 1];
+    return after === undefined
+        ? CONTINUES_NUMBER.has(spelt(at + 1))
+        : CONTINUES_NUMBER_CHARACTERS.has(after);
 }
 //# sourceMappingURL=sandhi-tones.js.map

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

@@ -1 +1 @@
-{"version":3,"file":"sandhi-tones.js","sourceRoot":"","sources":["../../src/decode/sandhi-tones.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAiB,qBAAqB,EAAE,MAAM,yBAAyB,CAAC;AAC/E,OAAO,EAAE,YAAY,EAAa,MAAM,iBAAiB,CAAC;AAE1D;;GAEG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB;IACrC,OAAO,QAAQ,CAAC,OAAO,KAAK,EAAE,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG,CAAC;AAC3D,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB;IACrC,OAAO,QAAQ,CAAC,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG,CAAC;AAC5D,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,YAAY,CAAC,SAA2B;IACtD,IAAI,SAAS,KAAK,SAAS,IAAI,SAAS,KAAK,YAAY,EAAE,CAAC;QAC1D,OAAO,CAAC,CAAC;IACX,CAAC;IACD,OAAO,SAAS,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AACjC,CAAC;AAED;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,aAAa,GAAG,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC;AAEpE;;;;;GAKG;AACH,MAAM,CAAC,MAAM,gBAAgB,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,aAAa,EAAE,MAAM,CAAC,CAAC,CAAC;AAEpE;;GAEG;AACH,MAAM,CAAC,MAAM,cAAc,GAAG,IAAI,CAAC;AAEnC;;;;;;;;;;;;;;;;;GAiBG;AACH,MAAM,UAAU,UAAU,CACxB,SAA8B,EAC9B,EAAU;IAEV,MAAM,KAAK,GAAG,CAAC,KAAa,EAAU,EAAE;QACtC,MAAM,QAAQ,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC;QAClC,OAAO,QAAQ,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,qBAAqB,CAAC,QAAQ,CAAC,CAAC;IACvE,CAAC,CAAC;IACF,MAAM,MAAM,GAAG,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;IAC7B,IAAI,MAAM,KAAK,cAAc,EAAE,CAAC;QAC9B,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,CAAC,aAAa,CAAC,GAAG,CAAC,MAAM,CAAC,IAAI,gBAAgB,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC;AAC3E,CAAC"}
+{"version":3,"file":"sandhi-tones.js","sourceRoot":"","sources":["../../src/decode/sandhi-tones.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAiB,qBAAqB,EAAE,MAAM,yBAAyB,CAAC;AAC/E,OAAO,EAAE,YAAY,EAAa,MAAM,iBAAiB,CAAC;AAE1D;;GAEG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB;IACrC,OAAO,QAAQ,CAAC,OAAO,KAAK,EAAE,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG,CAAC;AAC3D,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,IAAI,CAAC,QAAkB;IACrC,OAAO,QAAQ,CAAC,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG,CAAC;AAC5D,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,YAAY,CAAC,SAA2B;IACtD,IAAI,SAAS,KAAK,SAAS,IAAI,SAAS,KAAK,YAAY,EAAE,CAAC;QAC1D,OAAO,CAAC,CAAC;IACX,CAAC;IACD,OAAO,SAAS,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;AACjC,CAAC;AAED;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,aAAa,GAAG,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,CAAC,CAAC;AAEpE;;;;;GAKG;AACH,MAAM,CAAC,MAAM,gBAAgB,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,aAAa,EAAE,MAAM,CAAC,CAAC,CAAC;AAEpE;;GAEG;AACH,MAAM,CAAC,MAAM,cAAc,GAAG,IAAI,CAAC;AAEnC;;;;;;;;;;;;GAYG;AACH,MAAM,CAAC,MAAM,kBAAkB,GAAG,IAAI,GAAG,CAAC;IACxC,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;CACJ,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,CAAC,MAAM,2BAA2B,GAAG,IAAI,GAAG,CAAC;IACjD,GAAG,kBAAkB;IACrB,GAAG;IACH,GAAG;CACJ,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,CAAC,MAAM,iBAAiB,GAAG,GAAG,CAAC;AAErC;;;;;;;;;;;;;;;;;;;;;;;GAuBG;AACH,MAAM,UAAU,UAAU,CACxB,SAA8B,EAC9B,EAAU,EACV,UAAU,GAAoC,EAAE;IAEhD,MAAM,KAAK,GAAG,CAAC,KAAa,EAAU,EAAE;QACtC,MAAM,QAAQ,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC;QAClC,OAAO,QAAQ,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,qBAAqB,CAAC,QAAQ,CAAC,CAAC;IACvE,CAAC,CAAC;IACF,MAAM,MAAM,GAAG,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;IAClC,IACE,MAAM,KAAK,SAAS;QAClB,CAAC,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,KAAK,cAAc;QAClC,CAAC,CAAC,MAAM,KAAK,iBAAiB,EAChC,CAAC;QACD,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IACE,MAAM,KAAK,SAAS;QAClB,CAAC,CAAC,CAAC,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACnC,CAAC,CAAC,CAAC,kBAAkB,CAAC,GAAG,CAAC,MAAM,CAAC,EACnC,CAAC;QACD,OAAO,IAAI,CAAC;IACd,CAAC;IACD,MAAM,KAAK,GAAG,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;IACjC,OAAO,KAAK,KAAK,SAAS;QACxB,CAAC,CAAC,gBAAgB,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QACrC,CAAC,CAAC,2BAA2B,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;AAC7C,CAAC"}