@kensio/pinyinjs 1.7.6 → 1.7.8

package/dist/decode/chang-complement.js

@@ -0,0 +1,113 @@
+/**
+ * What can follow a 长 that is a length and cannot follow one that is growing.
+ *
+ * The near-side contexts are in `chang-context.ts`, and each of them has to
+ * check this side as well. These two do the opposite: they settle a 长 from what
+ * comes after it and ask nothing at all about what comes before.
+ */
+import { toCharacters } from "../script/characters.js";
+import { tagOf, wordsStartingAt } from "./rules.js";
+/**
+ * What can only follow a 长 that is a length.
+ *
+ * A quality can be compared and intensified, and a growing cannot: 长一点 is
+ * longer by a bit, 长极了 is extremely long, and 长而美丽 conjoins two
+ * adjectives. None of them is something a verb governs, so the near side needs
+ * no test at all and these settle a 长 on their own.
+ *
+ * 得 is pointedly absent even though 长得很 looks like the same shape, because
+ * 长得很快 is the growing with its complement and 得 is where that complement
+ * hangs. {@link GROWING} in `chang-rule.ts` guards it from the other side.
+ */
+const COMPARED = new Set([
+    "而",
+    "一点",
+    "一點",
+    "点",
+    "點",
+    "一些",
+    "些",
+    "极了",
+    "極了",
+    "得多",
+]);
+/**
+ * Whether what follows a 长 is a comparison or an intensifier.
+ */
+export function isCompared(context) {
+    const { characters, edge } = context;
+    for (let length = 1; length <= 2; length++) {
+        const after = characters.slice(edge.to, edge.to + length).join("");
+        if (COMPARED.has(after)) {
+            return true;
+        }
+    }
+    return false;
+}
+/**
+ * The tag jieba gives a numeral.
+ */
+const NUMERAL = "m";
+/**
+ * The units a length or a duration is given in.
+ *
+ * Written out rather than taken from the 量词 tag, and the corpus says why: a
+ * numeral after a 长 is far more often counting something else. 学校现有通榆和新
+ * 长两个校区 counts campuses and 竞争马华总会长一职 counts posts, and both read
+ * `zhǎng`. What a 长 can be measured in is a distance or a stretch of time, and
+ * that is a closed set.
+ */
+const MEASURED_IN = new Set([
+    "米",
+    "公里",
+    "千米",
+    "厘米",
+    "毫米",
+    "公分",
+    "英里",
+    "英尺",
+    "英寸",
+    "尺",
+    "寸",
+    "里",
+    "码",
+    "秒",
+    "分钟",
+    "分鐘",
+    "小时",
+    "小時",
+    "天",
+    "周",
+    "週",
+    "年",
+    "世纪",
+    "世紀",
+]);
+/**
+ * Whether a 长 is being given a measurement.
+ *
+ * 那条河长三百公里, 隧道长五公里. A quantity after a 长 is how long the thing is,
+ * and a growing takes an aspect marker before it can take one — 长了三厘米 —
+ * which is the 了 this never sees.
+ *
+ * A number written in digits reads the same, though nothing here knows the
+ * difference. 那条河长300公里 is four runs and the 长 ends the first of them, so
+ * the numeral and the 公里 are one run and two runs away; the decode is handed
+ * the 汉字 they stand for before this is ever asked. See `convert-pipeline.ts`.
+ */
+export function isMeasured(context) {
+    const { characters, edge } = context;
+    for (const number of wordsStartingAt(context, edge.to)) {
+        if (!tagOf(context, number).startsWith(NUMERAL)) {
+            continue;
+        }
+        const at = edge.to + toCharacters(number).length;
+        for (let length = 1; length <= 2; length++) {
+            if (MEASURED_IN.has(characters.slice(at, at + length).join(""))) {
+                return true;
+            }
+        }
+    }
+    return false;
+}
+//# sourceMappingURL=chang-complement.js.map

package/dist/decode/chang-complement.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"chang-complement.js","sourceRoot":"","sources":["../../src/decode/chang-complement.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAoB,KAAK,EAAE,eAAe,EAAE,MAAM,YAAY,CAAC;AAEtE;;;;;;;;;;;GAWG;AACH,MAAM,QAAQ,GAAG,IAAI,GAAG,CAAC;IACvB,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,GAAG;IACH,IAAI;IACJ,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;CACL,CAAC,CAAC;AAEH;;GAEG;AACH,MAAM,UAAU,UAAU,CAAC,OAAoB;IAC7C,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;IACrC,KAAK,IAAI,MAAM,GAAG,CAAC,EAAE,MAAM,IAAI,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;QAC3C,MAAM,KAAK,GAAG,UAAU,CAAC,KAAK,CAAC,IAAI,CAAC,EAAE,EAAE,IAAI,CAAC,EAAE,GAAG,MAAM,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACnE,IAAI,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,CAAC;YACxB,OAAO,IAAI,CAAC;QACd,CAAC;IACH,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC;AAED;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;;;;GAQG;AACH,MAAM,WAAW,GAAG,IAAI,GAAG,CAAC;IAC1B,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,IAAI;IACJ,IAAI;CACL,CAAC,CAAC;AAEH;;;;;;;;;;;GAWG;AACH,MAAM,UAAU,UAAU,CAAC,OAAoB;IAC7C,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;IACrC,KAAK,MAAM,MAAM,IAAI,eAAe,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,EAAE,CAAC;QACvD,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,UAAU,CAAC,OAAO,CAAC,EAAE,CAAC;YAChD,SAAS;QACX,CAAC;QACD,MAAM,EAAE,GAAG,IAAI,CAAC,EAAE,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC;QACjD,KAAK,IAAI,MAAM,GAAG,CAAC,EAAE,MAAM,IAAI,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;YAC3C,IAAI,WAAW,CAAC,GAAG,CAAC,UAAU,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,GAAG,MAAM,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,EAAE,CAAC;gBAChE,OAAO,IAAI,CAAC;YACd,CAAC;QACH,CAAC;IACH,CAAC;IACD,OAAO,KAAK,CAAC;AACf,CAAC"}

package/dist/decode/chang-context.d.ts

@@ -1,10 +1,3 @@
-/**
- * The two 长 contexts a degree adverb does not cover.
- *
- * 很 or 太 in front of a 长 settles it on its own, and `chang-rule.ts` says why.
- * These two do not: each names a near side that a verb 长 shares, so each has
- * to check the far side as well.
- */
 import { type EdgeContext } from "./rules.js";
 /**
  * Whether a 长 is the whole of what an adverb is scoping.
@@ -22,6 +15,15 @@ export declare function isPredicated(context: EdgeContext): boolean;
  * near side has to say the rest. A 量词 or 有 in front of a 长 leaves it nothing
  * to be the verb of: 一个长头发的女生 and 兔子有长耳朵 have their subject
  * already spoken for, and 他长毛了 and 树长叶子 do not.
+ *
+ * A verb in front is **not** enough, though 他留长头发 wears long hair where
+ * 他长头发 grows it. jieba tags 树 a verb and 习惯 a noun, so the tag names
+ * neither set: taking every verb read 树长叶子 as an adjective and 教育长邓演达
+ * with it.
+ *
+ * Every word ending at the position is asked rather than the longest, because
+ * the longest is 一个 where the 量词 is the 个 inside it. See
+ * {@link wordsEndingAt}.
  */
 export declare function isAttributive(context: EdgeContext): boolean;
 //# sourceMappingURL=chang-context.d.ts.map

package/dist/decode/chang-context.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"chang-context.d.ts","sourceRoot":"","sources":["../../src/decode/chang-context.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EACL,KAAK,WAAW,EAIjB,MAAM,YAAY,CAAC;AAqBpB;;;;;;GAMG;AACH,wBAAgB,YAAY,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAO1D;AAYD;;;;;;;;GAQG;AACH,wBAAgB,aAAa,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAO3D"}
+{"version":3,"file":"chang-context.d.ts","sourceRoot":"","sources":["../../src/decode/chang-context.ts"],"names":[],"mappings":"AAQA,OAAO,EACL,KAAK,WAAW,EAIjB,MAAM,YAAY,CAAC;AAqBpB;;;;;;GAMG;AACH,wBAAgB,YAAY,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAO1D;AAmCD;;;;;;;;;;;;;;;;;GAiBG;AACH,wBAAgB,aAAa,CAAC,OAAO,EAAE,WAAW,GAAG,OAAO,CAW3D"}

package/dist/decode/chang-context.js

@@ -5,7 +5,8 @@
  * These two do not: each names a near side that a verb 长 shares, so each has
  * to check the far side as well.
  */
-import { tagOf, wordEndingAt, wordStartingAt, } from "./rules.js";
+import { QUANTITY_CHARACTERS } from "../numerals/characters.js";
+import { tagOf, wordsEndingAt, wordStartingAt, } from "./rules.js";
 /**
  * The adverbs {@link DEGREE} leaves out, because they scope a verb just as well.
  *
@@ -46,6 +47,27 @@ const MEASURE = "q";
  * The tag prefix a noun carries, which is what an attributive 长 modifies.
  */
 const NOUN_TAG = "n";
+/**
+ * The mark that makes the number after it a position rather than a count.
+ */
+const ORDINAL = "第";
+/**
+ * Whether the 量词 in front of a position is counting an ordinal.
+ *
+ * 一个长头发的女生 has its subject spoken for by the 一个, and 第一次长胡子 does
+ * not: 第一次 is when the growing happened rather than what is being counted.
+ * The walk back over the quantity characters is the same one
+ * {@link import("./reading-rules.js").COUNTED_MEASURE} makes, and for the same
+ * reason.
+ */
+function isOrdinalPhrase(context, at) {
+    const { characters } = context;
+    let from = at - 1;
+    while (from > 0 && QUANTITY_CHARACTERS.has(characters[from - 1] ?? "")) {
+        from--;
+    }
+    return characters[from - 1] === ORDINAL;
+}
 /**
  * Whether a 长 is an adjective in front of the noun it modifies.
  *
@@ -54,13 +76,24 @@ const NOUN_TAG = "n";
  * near side has to say the rest. A 量词 or 有 in front of a 长 leaves it nothing
  * to be the verb of: 一个长头发的女生 and 兔子有长耳朵 have their subject
  * already spoken for, and 他长毛了 and 树长叶子 do not.
+ *
+ * A verb in front is **not** enough, though 他留长头发 wears long hair where
+ * 他长头发 grows it. jieba tags 树 a verb and 习惯 a noun, so the tag names
+ * neither set: taking every verb read 树长叶子 as an adjective and 教育长邓演达
+ * with it.
+ *
+ * Every word ending at the position is asked rather than the longest, because
+ * the longest is 一个 where the 量词 is the 个 inside it. See
+ * {@link wordsEndingAt}.
  */
 export function isAttributive(context) {
     const { edge } = context;
     if (!tagOf(context, wordStartingAt(context, edge.to)).startsWith(NOUN_TAG)) {
         return false;
     }
-    const before = wordEndingAt(context, edge.from);
-    return before === "有" || tagOf(context, before) === MEASURE;
+    if (isOrdinalPhrase(context, edge.from)) {
+        return false;
+    }
+    return wordsEndingAt(context, edge.from).some((word) => word === "有" || tagOf(context, word) === MEASURE);
 }
 //# sourceMappingURL=chang-context.js.map

package/dist/decode/chang-context.js.map

@@ -1 +1 @@
-{"version":3,"file":"chang-context.js","sourceRoot":"","sources":["../../src/decode/chang-context.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EAEL,KAAK,EACL,YAAY,EACZ,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB;;;;;;;;GAQG;AACH,MAAM,OAAO,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC;AAE1D;;;;;GAKG;AACH,MAAM,OAAO,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAE7D;;;;;;GAMG;AACH,MAAM,UAAU,YAAY,CAAC,OAAoB;IAC/C,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;IACrC,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;QAClD,OAAO,KAAK,CAAC;IACf,CAAC;IACD,MAAM,KAAK,GAAG,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;IAClC,OAAO,KAAK,KAAK,SAAS,IAAI,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;AACnD,CAAC;AAED;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;GAEG;AACH,MAAM,QAAQ,GAAG,GAAG,CAAC;AAErB;;;;;;;;GAQG;AACH,MAAM,UAAU,aAAa,CAAC,OAAoB;IAChD,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;IACzB,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,cAAc,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC3E,OAAO,KAAK,CAAC;IACf,CAAC;IACD,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;IAChD,OAAO,MAAM,KAAK,GAAG,IAAI,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC,KAAK,OAAO,CAAC;AAC9D,CAAC"}
+{"version":3,"file":"chang-context.js","sourceRoot":"","sources":["../../src/decode/chang-context.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EAAE,mBAAmB,EAAE,MAAM,2BAA2B,CAAC;AAChE,OAAO,EAEL,KAAK,EACL,aAAa,EACb,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB;;;;;;;;GAQG;AACH,MAAM,OAAO,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC;AAE1D;;;;;GAKG;AACH,MAAM,OAAO,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAE7D;;;;;;GAMG;AACH,MAAM,UAAU,YAAY,CAAC,OAAoB;IAC/C,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;IACrC,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,EAAE,CAAC;QAClD,OAAO,KAAK,CAAC;IACf,CAAC;IACD,MAAM,KAAK,GAAG,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;IAClC,OAAO,KAAK,KAAK,SAAS,IAAI,OAAO,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;AACnD,CAAC;AAED;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;GAEG;AACH,MAAM,QAAQ,GAAG,GAAG,CAAC;AAErB;;GAEG;AACH,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB;;;;;;;;GAQG;AACH,SAAS,eAAe,CAAC,OAAoB,EAAE,EAAU;IACvD,MAAM,EAAE,UAAU,EAAE,GAAG,OAAO,CAAC;IAC/B,IAAI,IAAI,GAAG,EAAE,GAAG,CAAC,CAAC;IAClB,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,OAAO,UAAU,CAAC,IAAI,GAAG,CAAC,CAAC,KAAK,OAAO,CAAC;AAC1C,CAAC;AAED;;;;;;;;;;;;;;;;;GAiBG;AACH,MAAM,UAAU,aAAa,CAAC,OAAoB;IAChD,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;IACzB,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,cAAc,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC3E,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IAAI,eAAe,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;QACxC,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,aAAa,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC,IAAI,CAC3C,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,KAAK,GAAG,IAAI,KAAK,CAAC,OAAO,EAAE,IAAI,CAAC,KAAK,OAAO,CAC3D,CAAC;AACJ,CAAC"}

package/dist/decode/chang-rule.d.ts

@@ -11,24 +11,34 @@ import { type EdgeRule } from "./rules.js";
  * 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`.
  *
- * Three contexts settle it, and the first of them needs only one side. 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 whatever follows. {@link DEGREE} is that set.
+ * Five contexts settle it, and one of them needs a single side. 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 whatever follows. {@link DEGREE} is that set.
  *
- * The other two need both sides, and they are in `chang-context.ts`:
- * {@link isPredicated} for 那座桥不长, {@link isAttributive} for 他有长头发.
- * What they have in common is that the near side alone would be wrong. 不长 is
- * 不长在耳朵外面 as readily as it is 不长, and a noun after a 长 is 长知识 as
- * readily as 长头发.
+ * Two more read the far side alone, in `chang-complement.ts`. A quality can be
+ * compared, intensified and conjoined where a growing cannot, so 长一点, 长极了
+ * and 长而美丽 need nothing in front of them ({@link isCompared}); and a length
+ * can be given in metres, so 长三百公里 settles itself ({@link isMeasured}).
+ *
+ * The last two need both sides, in `chang-context.ts`: {@link isPredicated} for
+ * 那座桥不长, {@link isAttributive} for 他有长头发. What those two have in common
+ * is that the near side alone would be wrong. 不长 is 不长在耳朵外面 as readily
+ * as it is 不长, and a noun after a 长 is 长知识 as readily as 长头发.
  *
  * 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. The degree
- * adverbs move 75 of them to `cháng`, all 75 correctly; the two contexts added
- * since move 8 more, also all correct — 那座桥不长, 生命不长, 射程不长, 还长,
- * 生得又瘦又长, 他有长头发, 兔子有长耳朵 and 那位长胡子德国人. On CPP's 40
- * hand-labelled 长 the character stands at 90.00%, against 85.00% with no rule
- * and 87.50% with the degree adverbs alone.
+ * rules were sized against, 379 长 and 長 decode as a word of their own and this
+ * moves 125 of them to `cháng`. Nine of the 125 are 长 the run they end could
+ * not see past until the decode was given the 汉字 behind the digits after it,
+ * as 那条河长300公里 is. Every conversion the last three contexts changed is
+ * right: 一个长头发的女生, 一个长毛绒玩具, 没有长胡子的舅舅, 要长一点, 活长点,
+ * 已经长极了, 神经棘长而狭窄 and 拉得很长. On CPP's 40 hand-labelled 长 the
+ * character stands at 92.50%, against 85.00% with no rule at all.
+ *
+ * **A verb in front is not one of the contexts**, though 他留长头发 wears long
+ * hair where 他长头发 grows it. jieba tags 树 a verb and 习惯 a noun, so no tag
+ * names the set, and taking every verb read 树长叶子 and 教育长邓演达 as
+ * adjectives.
  *
  * The last piece is {@link GROWN_INTO}, which pushes the opposite way on the one
  * shape the sources get wrong.

package/dist/decode/chang-rule.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"chang-rule.d.ts","sourceRoot":"","sources":["../../src/decode/chang-rule.ts"],"names":[],"mappings":"AASA,OAAO,EAAoB,KAAK,QAAQ,EAAgB,MAAM,YAAY,CAAC;AA2H3E;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAiCG;AACH,eAAO,MAAM,gBAAgB,EAAE,QAmB9B,CAAC"}
+{"version":3,"file":"chang-rule.d.ts","sourceRoot":"","sources":["../../src/decode/chang-rule.ts"],"names":[],"mappings":"AAUA,OAAO,EAAoB,KAAK,QAAQ,EAAiB,MAAM,YAAY,CAAC;AA+H5E;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;AACH,eAAO,MAAM,gBAAgB,EAAE,QAyB9B,CAAC"}

package/dist/decode/chang-rule.js

@@ -1,5 +1,6 @@
+import { isCompared, isMeasured } from "./chang-complement.js";
 import { isAttributive, isPredicated } from "./chang-context.js";
-import { wordEndingAt } from "./rules.js";
+import { wordsEndingAt } from "./rules.js";
 /**
  * Whether a syllable is `cháng`, the stative reading of 长.
  */
@@ -22,6 +23,10 @@ function isChang(syllable) {
  * 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.
+ *
+ * Matched against every word ending in front of the 长 rather than the longest
+ * one, since 得很 is a word and hides the 很 inside it. See
+ * {@link wordsEndingAt}.
  */
 const DEGREE = new Set([
     "一样",
@@ -124,24 +129,34 @@ function isGrowthCorrelative(context) {
  * 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`.
  *
- * Three contexts settle it, and the first of them needs only one side. 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 whatever follows. {@link DEGREE} is that set.
+ * Five contexts settle it, and one of them needs a single side. 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 whatever follows. {@link DEGREE} is that set.
+ *
+ * Two more read the far side alone, in `chang-complement.ts`. A quality can be
+ * compared, intensified and conjoined where a growing cannot, so 长一点, 长极了
+ * and 长而美丽 need nothing in front of them ({@link isCompared}); and a length
+ * can be given in metres, so 长三百公里 settles itself ({@link isMeasured}).
  *
- * The other two need both sides, and they are in `chang-context.ts`:
- * {@link isPredicated} for 那座桥不长, {@link isAttributive} for 他有长头发.
- * What they have in common is that the near side alone would be wrong. 不长 is
- * 不长在耳朵外面 as readily as it is 不长, and a noun after a 长 is 长知识 as
- * readily as 长头发.
+ * The last two need both sides, in `chang-context.ts`: {@link isPredicated} for
+ * 那座桥不长, {@link isAttributive} for 他有长头发. What those two have in common
+ * is that the near side alone would be wrong. 不长 is 不长在耳朵外面 as readily
+ * as it is 不长, and a noun after a 长 is 长知识 as readily as 长头发.
  *
  * 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. The degree
- * adverbs move 75 of them to `cháng`, all 75 correctly; the two contexts added
- * since move 8 more, also all correct — 那座桥不长, 生命不长, 射程不长, 还长,
- * 生得又瘦又长, 他有长头发, 兔子有长耳朵 and 那位长胡子德国人. On CPP's 40
- * hand-labelled 长 the character stands at 90.00%, against 85.00% with no rule
- * and 87.50% with the degree adverbs alone.
+ * rules were sized against, 379 长 and 長 decode as a word of their own and this
+ * moves 125 of them to `cháng`. Nine of the 125 are 长 the run they end could
+ * not see past until the decode was given the 汉字 behind the digits after it,
+ * as 那条河长300公里 is. Every conversion the last three contexts changed is
+ * right: 一个长头发的女生, 一个长毛绒玩具, 没有长胡子的舅舅, 要长一点, 活长点,
+ * 已经长极了, 神经棘长而狭窄 and 拉得很长. On CPP's 40 hand-labelled 长 the
+ * character stands at 92.50%, against 85.00% with no rule at all.
+ *
+ * **A verb in front is not one of the contexts**, though 他留长头发 wears long
+ * hair where 他长头发 grows it. jieba tags 树 a verb and 习惯 a noun, so no tag
+ * names the set, and taking every verb read 树长叶子 and 教育长邓演达 as
+ * adjectives.
  *
  * The last piece is {@link GROWN_INTO}, which pushes the opposite way on the one
  * shape the sources get wrong.
@@ -159,11 +174,15 @@ export const ADJECTIVAL_CHANG = {
         if (GROWING.has(characters[edge.to] ?? "")) {
             return "keep";
         }
-        if (isPredicated(context) || isAttributive(context)) {
+        if (isPredicated(context) ||
+            isAttributive(context) ||
+            isCompared(context) ||
+            isMeasured(context)) {
             return "force";
         }
-        const before = wordEndingAt(context, edge.from);
-        return before !== undefined && DEGREE.has(before) ? "force" : "keep";
+        return wordsEndingAt(context, edge.from).some((word) => DEGREE.has(word))
+            ? "force"
+            : "keep";
     },
 };
 //# sourceMappingURL=chang-rule.js.map

package/dist/decode/chang-rule.js.map

@@ -1 +1 @@
-{"version":3,"file":"chang-rule.js","sourceRoot":"","sources":["../../src/decode/chang-rule.ts"],"names":[],"mappings":"AAQA,OAAO,EAAE,aAAa,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACjE,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;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAiCG;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,IAAI,YAAY,CAAC,OAAO,CAAC,IAAI,aAAa,CAAC,OAAO,CAAC,EAAE,CAAC;YACpD,OAAO,OAAO,CAAC;QACjB,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":"chang-rule.js","sourceRoot":"","sources":["../../src/decode/chang-rule.ts"],"names":[],"mappings":"AAQA,OAAO,EAAE,UAAU,EAAE,UAAU,EAAE,MAAM,uBAAuB,CAAC;AAC/D,OAAO,EAAE,aAAa,EAAE,YAAY,EAAE,MAAM,oBAAoB,CAAC;AACjE,OAAO,EAAmC,aAAa,EAAE,MAAM,YAAY,CAAC;AAE5E;;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;;;;;;;;;;;;;;;;;;GAkBG;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;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2CG;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,IACE,YAAY,CAAC,OAAO,CAAC;YACrB,aAAa,CAAC,OAAO,CAAC;YACtB,UAAU,CAAC,OAAO,CAAC;YACnB,UAAU,CAAC,OAAO,CAAC,EACnB,CAAC;YACD,OAAO,OAAO,CAAC;QACjB,CAAC;QACD,OAAO,aAAa,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC;YACvE,CAAC,CAAC,OAAO;YACT,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC"}

package/dist/decode/convert-pipeline.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"convert-pipeline.d.ts","sourceRoot":"","sources":["../../src/decode/convert-pipeline.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAU9D,OAAO,KAAK,EAAE,cAAc,EAAE,cAAc,EAAW,MAAM,aAAa,CAAC;AAE3E,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,eAAe,CAAC;AAI5C;;GAEG;AACH,wBAAgB,WAAW,CACzB,MAAM,EAAE,MAAM,EACd,UAAU,EAAE,UAAU,EACtB,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,cAAc,GACtB,SAAS,cAAc,EAAE,CAiF3B"}
+{"version":3,"file":"convert-pipeline.d.ts","sourceRoot":"","sources":["../../src/decode/convert-pipeline.ts"],"names":[],"mappings":"AAAA;;;;;;;;;GASG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAU9D,OAAO,KAAK,EAAE,cAAc,EAAE,cAAc,EAAW,MAAM,aAAa,CAAC;AAG3E,OAAO,KAAK,EAAE,MAAM,EAAE,MAAM,eAAe,CAAC;AAI5C;;GAEG;AACH,wBAAgB,WAAW,CACzB,MAAM,EAAE,MAAM,EACd,UAAU,EAAE,UAAU,EACtB,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,cAAc,GACtB,SAAS,cAAc,EAAE,CAgF3B"}

package/dist/decode/convert-pipeline.js

@@ -4,7 +4,8 @@ import { readNumbersIn } from "../numerals/text.js";
 import { toCharacters } from "../script/characters.js";
 import { hintsWithin, resolveHints } from "./hints.js";
 import { splitRuns, surroundingCharacters } from "./runs.js";
-import { numeralBefore, surrounding, writeNumbers } from "./numbers.js";
+import { surrounding, writeNumbers } from "./numbers.js";
+import { contextAround } from "./run-context.js";
 import { wordsOf } from "./run-words.js";
 import { writeRun, rewrite } from "./write-run.js";
 /**
@@ -42,8 +43,7 @@ export function convertWith(decode, dictionary, text, options) {
         const within = hints === undefined
             ? undefined
             : hintsWithin(hints, starts[at] ?? 0, toCharacters(run.text).length);
-        const said = numeralBefore(read[at - 1] ?? []);
-        return wordsOf(decode(dictionary, run.text, said, within), dictionary, grouping);
+        return wordsOf(decode(dictionary, run.text, contextAround(runs, read, at), within), dictionary, grouping);
     });
     for (const [at, run] of runs.entries()) {
         const words = decoded[at] ?? [];

package/dist/decode/convert-pipeline.js.map

@@ -1 +1 @@
-{"version":3,"file":"convert-pipeline.js","sourceRoot":"","sources":["../../src/decode/convert-pipeline.ts"],"names":[],"mappings":"AAWA,OAAO,EACL,uBAAuB,EACvB,UAAU,GACX,MAAM,4BAA4B,CAAC;AACpC,OAAO,EAAE,uBAAuB,EAAE,MAAM,+BAA+B,CAAC;AACxE,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AACpD,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,YAAY,CAAC;AACvD,OAAO,EAAE,SAAS,EAAE,qBAAqB,EAAE,MAAM,WAAW,CAAC;AAE7D,OAAO,EAAE,aAAa,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AAExE,OAAO,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AACzC,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAEnD;;GAEG;AACH,MAAM,UAAU,WAAW,CACzB,MAAc,EACd,UAAsB,EACtB,IAAY,EACZ,OAAuB;IAEvB,MAAM,EACJ,MAAM,GAAG,OAAO,EAChB,QAAQ,GAAG,OAAO,EAClB,UAAU,GAAG,QAAQ,EACrB,QAAQ,GAAG,MAAM,EACjB,WAAW,GAAG,OAAO,EACrB,QAAQ,GAAG,IAAI,EACf,OAAO,GAAG,MAAM,EAChB,MAAM,EACN,QAAQ,GACT,GAAG,OAAO,CAAC;IACZ,MAAM,OAAO,GAAY,EAAE,QAAQ,EAAE,UAAU,EAAE,QAAQ,EAAE,CAAC;IAC5D,MAAM,SAAS,GAAqB,EAAE,CAAC;IAEvC,MAAM,IAAI,GAAG,CAAC,GAAG,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;IAClC,0EAA0E;IAC1E,6DAA6D;IAC7D,MAAM,KAAK,GAAG,QAAQ,KAAK,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,YAAY,CAAC,QAAQ,CAAC,CAAC;IAC1E,6EAA6E;IAC7E,4DAA4D;IAC5D,MAAM,MAAM,GAAa,EAAE,CAAC;IAC5B,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACnB,KAAK,IAAI,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC;IACzC,CAAC;IACD,uEAAuE;IACvE,0EAA0E;IAC1E,mEAAmE;IACnE,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,EAAE,EAAE,CAChC,GAAG,CAAC,KAAK,IAAI,OAAO,KAAK,MAAM;QAC7B,CAAC,CAAC,EAAE;QACJ,CAAC,CAAC,aAAa,CAAC,GAAG,CAAC,IAAI,EAAE,qBAAqB,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,CAC7D,CAAC;IACF,0EAA0E;IAC1E,4EAA4E;IAC5E,+DAA+D;IAC/D,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,EAAE,EAAE;QACnC,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,CAAC;YACf,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,MAAM,GACV,KAAK,KAAK,SAAS;YACjB,CAAC,CAAC,SAAS;YACX,CAAC,CAAC,WAAW,CAAC,KAAK,EAAE,MAAM,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC,CAAC;QACzE,MAAM,IAAI,GAAG,aAAa,CAAC,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC;QAC/C,OAAO,OAAO,CACZ,MAAM,CAAC,UAAU,EAAE,GAAG,CAAC,IAAI,EAAE,IAAI,EAAE,MAAM,CAAC,EAC1C,UAAU,EACV,QAAQ,CACT,CAAC;IACJ,CAAC,CAAC,CAAC;IAEH,KAAK,MAAM,CAAC,EAAE,EAAE,GAAG,CAAC,IAAI,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC;QACvC,MAAM,KAAK,GAAG,OAAO,CAAC,EAAE,CAAC,IAAI,EAAE,CAAC;QAChC,IAAI,GAAG,CAAC,KAAK,EAAE,CAAC;YACd,SAAS,CAAC,IAAI,CAAC,GAAG,QAAQ,CAAC,UAAU,EAAE,KAAK,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC;YACxE,SAAS;QACX,CAAC;QACD,SAAS,CAAC,IAAI,CACZ,GAAG,YAAY,CACb,GAAG,CAAC,IAAI,EACR,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE,EACd,WAAW,CAAC,IAAI,EAAE,OAAO,EAAE,EAAE,CAAC,EAC9B,OAAO,EACP,EAAE,MAAM,EAAE,CACX,CACF,CAAC;IACJ,CAAC;IACD,IAAI,MAAM,GAA8B,SAAS,CAAC;IAElD,0EAA0E;IAC1E,6EAA6E;IAC7E,qEAAqE;IACrE,IAAI,QAAQ,KAAK,MAAM,IAAI,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC;QAC5C,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,uBAAuB,CAAC,CAAC;IACpD,CAAC;IACD,OAAO,WAAW,KAAK,OAAO;QAC5B,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,uBAAuB,CAAC;QAC1C,CAAC,CAAC,MAAM,CAAC;AACb,CAAC"}
+{"version":3,"file":"convert-pipeline.js","sourceRoot":"","sources":["../../src/decode/convert-pipeline.ts"],"names":[],"mappings":"AAWA,OAAO,EACL,uBAAuB,EACvB,UAAU,GACX,MAAM,4BAA4B,CAAC;AACpC,OAAO,EAAE,uBAAuB,EAAE,MAAM,+BAA+B,CAAC;AACxE,OAAO,EAAE,aAAa,EAAE,MAAM,qBAAqB,CAAC;AACpD,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,YAAY,CAAC;AACvD,OAAO,EAAE,SAAS,EAAE,qBAAqB,EAAE,MAAM,WAAW,CAAC;AAE7D,OAAO,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,cAAc,CAAC;AACzD,OAAO,EAAE,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAEjD,OAAO,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AACzC,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAEnD;;GAEG;AACH,MAAM,UAAU,WAAW,CACzB,MAAc,EACd,UAAsB,EACtB,IAAY,EACZ,OAAuB;IAEvB,MAAM,EACJ,MAAM,GAAG,OAAO,EAChB,QAAQ,GAAG,OAAO,EAClB,UAAU,GAAG,QAAQ,EACrB,QAAQ,GAAG,MAAM,EACjB,WAAW,GAAG,OAAO,EACrB,QAAQ,GAAG,IAAI,EACf,OAAO,GAAG,MAAM,EAChB,MAAM,EACN,QAAQ,GACT,GAAG,OAAO,CAAC;IACZ,MAAM,OAAO,GAAY,EAAE,QAAQ,EAAE,UAAU,EAAE,QAAQ,EAAE,CAAC;IAC5D,MAAM,SAAS,GAAqB,EAAE,CAAC;IAEvC,MAAM,IAAI,GAAG,CAAC,GAAG,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC;IAClC,0EAA0E;IAC1E,6DAA6D;IAC7D,MAAM,KAAK,GAAG,QAAQ,KAAK,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,YAAY,CAAC,QAAQ,CAAC,CAAC;IAC1E,6EAA6E;IAC7E,4DAA4D;IAC5D,MAAM,MAAM,GAAa,EAAE,CAAC;IAC5B,IAAI,KAAK,GAAG,CAAC,CAAC;IACd,KAAK,MAAM,GAAG,IAAI,IAAI,EAAE,CAAC;QACvB,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACnB,KAAK,IAAI,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC;IACzC,CAAC;IACD,uEAAuE;IACvE,0EAA0E;IAC1E,mEAAmE;IACnE,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,EAAE,EAAE,CAChC,GAAG,CAAC,KAAK,IAAI,OAAO,KAAK,MAAM;QAC7B,CAAC,CAAC,EAAE;QACJ,CAAC,CAAC,aAAa,CAAC,GAAG,CAAC,IAAI,EAAE,qBAAqB,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,CAC7D,CAAC;IACF,0EAA0E;IAC1E,4EAA4E;IAC5E,+DAA+D;IAC/D,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,EAAE,EAAE;QACnC,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,CAAC;YACf,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,MAAM,GACV,KAAK,KAAK,SAAS;YACjB,CAAC,CAAC,SAAS;YACX,CAAC,CAAC,WAAW,CAAC,KAAK,EAAE,MAAM,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,YAAY,CAAC,GAAG,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC,CAAC;QACzE,OAAO,OAAO,CACZ,MAAM,CAAC,UAAU,EAAE,GAAG,CAAC,IAAI,EAAE,aAAa,CAAC,IAAI,EAAE,IAAI,EAAE,EAAE,CAAC,EAAE,MAAM,CAAC,EACnE,UAAU,EACV,QAAQ,CACT,CAAC;IACJ,CAAC,CAAC,CAAC;IAEH,KAAK,MAAM,CAAC,EAAE,EAAE,GAAG,CAAC,IAAI,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC;QACvC,MAAM,KAAK,GAAG,OAAO,CAAC,EAAE,CAAC,IAAI,EAAE,CAAC;QAChC,IAAI,GAAG,CAAC,KAAK,EAAE,CAAC;YACd,SAAS,CAAC,IAAI,CAAC,GAAG,QAAQ,CAAC,UAAU,EAAE,KAAK,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC;YACxE,SAAS;QACX,CAAC;QACD,SAAS,CAAC,IAAI,CACZ,GAAG,YAAY,CACb,GAAG,CAAC,IAAI,EACR,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE,EACd,WAAW,CAAC,IAAI,EAAE,OAAO,EAAE,EAAE,CAAC,EAC9B,OAAO,EACP,EAAE,MAAM,EAAE,CACX,CACF,CAAC;IACJ,CAAC;IACD,IAAI,MAAM,GAA8B,SAAS,CAAC;IAElD,0EAA0E;IAC1E,6EAA6E;IAC7E,qEAAqE;IACrE,IAAI,QAAQ,KAAK,MAAM,IAAI,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC;QAC5C,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,uBAAuB,CAAC,CAAC;IACpD,CAAC;IACD,OAAO,WAAW,KAAK,OAAO;QAC5B,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,uBAAuB,CAAC;QAC1C,CAAC,CAAC,MAAM,CAAC;AACb,CAAC"}

package/dist/decode/de-rule.d.ts

@@ -28,6 +28,26 @@ import { type EdgeRule } from "./rules.js";
  * which nothing short of semantics separates from 必须得承认.
  */
 export declare const MODAL_DE: EdgeRule;
+/**
+ * A verb and a 得 in front of a complement is not a word reading `dé`.
+ *
+ * 他算得上一个作家 came out `suàn dé shàng`, because 算得 is a key jieba counted
+ * and the phrase corpus reads it `suàn dé`. In a potential complement the 得 is
+ * the particle: 算得上, 吃得了, 买得起 and 看得住 are all `de`, and the reading
+ * the key carries is the one 算得 has standing alone, which is not what a
+ * complement leaves it doing.
+ *
+ * Both sides are needed and neither is enough. A verb before the 得 rules out
+ * 只得上山 and 不得下车, where the 得 belongs to an adverb and the character
+ * after it starts a verb phrase of its own; the complement after it rules out
+ * 取得上级批准, where 上 is the first character of 上级 rather than a complement.
+ * That second test is why the complement has to be the whole of the word
+ * starting there.
+ *
+ * Over the 88,866 lines it changes 5 conversions and all 5 are 算得上, which is
+ * the only word of this shape the corpus holds. CPP does not move.
+ */
+export declare const POTENTIAL_DE: EdgeRule;
 /**
  * A word beginning with 的 does not start where the particle 的 does.
  *

package/dist/decode/de-rule.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"de-rule.d.ts","sourceRoot":"","sources":["../../src/decode/de-rule.ts"],"names":[],"mappings":"AASA,OAAO,EAEL,KAAK,QAAQ,EAId,MAAM,YAAY,CAAC;AAgCpB;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,eAAO,MAAM,QAAQ,EAAE,QActB,CAAC;AAWF;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,eAAO,MAAM,WAAW,EAAE,QAczB,CAAC"}
+{"version":3,"file":"de-rule.d.ts","sourceRoot":"","sources":["../../src/decode/de-rule.ts"],"names":[],"mappings":"AAUA,OAAO,EAEL,KAAK,QAAQ,EAKd,MAAM,YAAY,CAAC;AAgCpB;;;;;;;;;;;;;;;;;;;;;;;;;;;GA2BG;AACH,eAAO,MAAM,QAAQ,EAAE,QActB,CAAC;AAkDF;;;;;;;;;;;;;;;;;;GAkBG;AACH,eAAO,MAAM,YAAY,EAAE,QAoB1B,CAAC;AAWF;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,eAAO,MAAM,WAAW,EAAE,QAczB,CAAC"}

package/dist/decode/de-rule.js

@@ -1,4 +1,13 @@
-import { tagOf, wordEndingAt, wordStartingAt, } from "./rules.js";
+/**
+ * The two rules about 得 and 的, which are one character each and three jobs.
+ *
+ * 得 is stored as the structural particle with `dé` and `děi` as alternates,
+ * and 的 is stored as the particle full stop. Both are among the commonest
+ * characters in the language, so the words they begin and end are numerous and
+ * mostly wrong.
+ */
+import { toCharacters } from "../script/characters.js";
+import { tagOf, wordEndingAt, wordStartingAt, wordsStartingAt, } from "./rules.js";
 /**
  * Whether a syllable is `děi`, the modal reading of 得.
  *
@@ -67,6 +76,83 @@ export const MODAL_DE = {
             : "keep";
     },
 };
+/**
+ * What a potential complement is made of, after the 得 that marks it.
+ *
+ * 算得上, 买得起, 看得住, 说得过去. The complement is a resultative or a
+ * directional, and this is the closed set of the ones that take a 得 in front of
+ * them rather than a verb of their own.
+ *
+ * 了 and 过 are deliberately absent, and for the same reason: both are aspect
+ * markers as well as complements, and the aspect is the commoner by a wide
+ * margin. Over 88,866 lines of Tatoeba and zh.wikipedia, including 了 broke 70
+ * conversions and fixed none — 获得了, 取得了 and 赢得了 all became 获 得 了 —
+ * and 过 broke 赢得过 twice. 吃得了 and 说得过去 are the shapes that gives up.
+ */
+const COMPLEMENTS = new Set(["上", "起", "下", "住"]);
+/**
+ * The tag prefix jieba gives a verb, which is what a complement attaches to.
+ */
+const VERB_TAG = "v";
+/**
+ * Whether a syllable is `dé`, the reading 得 takes as a verb of its own.
+ */
+function isDe(syllable) {
+    return (syllable?.initial === "d" && syllable.final === "e" && syllable.tone === 2);
+}
+/**
+ * Whether a word starting at the complement takes it away from the 得.
+ *
+ * The one thing that separates 算得上一个作家 from 取得上级批准, since both put a
+ * verb, a 得 and a 上 in the same order. 上级 is a word and nothing follows it
+ * that starts another, so the 上 belongs to it; 上道 is also a word, but 道歉
+ * follows and is the better claim on the 道, so the 上 is left to the 得.
+ *
+ * Only multi-character words are asked, because every character is a word of its
+ * own and a test that counted those would answer yes everywhere.
+ */
+function isSwallowed(context, at) {
+    const longer = (from) => wordsStartingAt(context, from).some((word) => toCharacters(word).length > 1);
+    return longer(at) && !longer(at + 1);
+}
+/**
+ * A verb and a 得 in front of a complement is not a word reading `dé`.
+ *
+ * 他算得上一个作家 came out `suàn dé shàng`, because 算得 is a key jieba counted
+ * and the phrase corpus reads it `suàn dé`. In a potential complement the 得 is
+ * the particle: 算得上, 吃得了, 买得起 and 看得住 are all `de`, and the reading
+ * the key carries is the one 算得 has standing alone, which is not what a
+ * complement leaves it doing.
+ *
+ * Both sides are needed and neither is enough. A verb before the 得 rules out
+ * 只得上山 and 不得下车, where the 得 belongs to an adverb and the character
+ * after it starts a verb phrase of its own; the complement after it rules out
+ * 取得上级批准, where 上 is the first character of 上级 rather than a complement.
+ * That second test is why the complement has to be the whole of the word
+ * starting there.
+ *
+ * Over the 88,866 lines it changes 5 conversions and all 5 are 算得上, which is
+ * the only word of this shape the corpus holds. CPP does not move.
+ */
+export const POTENTIAL_DE = {
+    name: "potential-de",
+    verdictFor: (context) => {
+        const { characters, edge } = context;
+        if (edge.to - edge.from !== 2 ||
+            edge.text.at(-1) !== "得" ||
+            edge.reading.length !== 2 ||
+            !isDe(edge.reading[1])) {
+            return "keep";
+        }
+        if (!tagOf(context, characters[edge.from]).startsWith(VERB_TAG)) {
+            return "keep";
+        }
+        return COMPLEMENTS.has(characters[edge.to] ?? "") &&
+            !isSwallowed(context, edge.to)
+            ? "forbid"
+            : "keep";
+    },
+};
 /**
  * The tags a word carries when the 的 after it is the structural particle.
  *

package/dist/decode/de-rule.js.map

@@ -1 +1 @@
-{"version":3,"file":"de-rule.js","sourceRoot":"","sources":["../../src/decode/de-rule.ts"],"names":[],"mappings":"AASA,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;;;;;;GAMG;AACH,MAAM,aAAa,GAAG,mBAAmB,CAAC;AAE1C;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,MAAM,CAAC,MAAM,WAAW,GAAa;IACnC,IAAI,EAAE,aAAa;IACnB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,IACE,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,GAAG,CAAC;YACvB,IAAI,CAAC,YAAY,KAAK,EAAE;YACxB,CAAC,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,EAC1B,CAAC;YACD,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;QAChD,OAAO,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC;IACxE,CAAC;CACF,CAAC"}
+{"version":3,"file":"de-rule.js","sourceRoot":"","sources":["../../src/decode/de-rule.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AAEvD,OAAO,EAGL,KAAK,EACL,YAAY,EACZ,cAAc,EACd,eAAe,GAChB,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;;;;;;;;;;;;GAYG;AACH,MAAM,WAAW,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAElD;;GAEG;AACH,MAAM,QAAQ,GAAG,GAAG,CAAC;AAErB;;GAEG;AACH,SAAS,IAAI,CAAC,QAA8B;IAC1C,OAAO,CACL,QAAQ,EAAE,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,GAAG,IAAI,QAAQ,CAAC,IAAI,KAAK,CAAC,CAC3E,CAAC;AACJ,CAAC;AAED;;;;;;;;;;GAUG;AACH,SAAS,WAAW,CAAC,OAAoB,EAAE,EAAU;IACnD,MAAM,MAAM,GAAG,CAAC,IAAY,EAAW,EAAE,CACvC,eAAe,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC,IAAI,CACjC,CAAC,IAAI,EAAE,EAAE,CAAC,YAAY,CAAC,IAAI,CAAC,CAAC,MAAM,GAAG,CAAC,CACxC,CAAC;IACJ,OAAO,MAAM,CAAC,EAAE,CAAC,IAAI,CAAC,MAAM,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;AACvC,CAAC;AAED;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,CAAC,MAAM,YAAY,GAAa;IACpC,IAAI,EAAE,cAAc;IACpB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACrC,IACE,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,KAAK,CAAC;YACzB,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,KAAK,GAAG;YACxB,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,CAAC;YACzB,CAAC,IAAI,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,EACtB,CAAC;YACD,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,UAAU,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC;YAChE,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,OAAO,WAAW,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC,IAAI,EAAE,CAAC;YAC/C,CAAC,WAAW,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC;YAC9B,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC;AAEF;;;;;;GAMG;AACH,MAAM,aAAa,GAAG,mBAAmB,CAAC;AAE1C;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,MAAM,CAAC,MAAM,WAAW,GAAa;IACnC,IAAI,EAAE,aAAa;IACnB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,IACE,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,GAAG,CAAC;YACvB,IAAI,CAAC,YAAY,KAAK,EAAE;YACxB,CAAC,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,GAAG,CAAC,EAC1B,CAAC;YACD,OAAO,MAAM,CAAC;QAChB,CAAC;QACD,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC,CAAC;QAChD,OAAO,aAAa,CAAC,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC,CAAC,CAAC,QAAQ,CAAC,CAAC,CAAC,MAAM,CAAC;IACxE,CAAC;CACF,CAAC"}

package/dist/decode/decode.d.ts

@@ -1,28 +1,15 @@
+/**
+ * Decoding a Han run into words: the two entry points a caller reaches for.
+ *
+ * The sweeps over the lattice itself are in `lattice-decode.ts`, and are
+ * re-exported here so that the decoder's parts stay one import away.
+ */
 import type { Dictionary } from "../dictionary/dictionary.js";
 import type { ResolvedHints } from "./hints.js";
-import { type Lattice } from "./lattice.js";
-import { type ReadingProjection, type ReadingUnit } from "./locking.js";
 import type { EdgeRule } from "./rules.js";
 import type { DecodedWord, ScoredWord } from "./word.js";
-/**
- * Decode the run's readings, scoring only where the readings are still open.
- *
- * The asymmetry ALGORITHM.md calls for. A stretch between two cut points whose
- * positions all locked reads the same way on every path, so its readings are
- * read off the locks and no shortest path is run over it at all. Only the
- * stretches with a genuine choice in them are decoded.
- */
-export declare function decodeReadings(lattice: Lattice, projection: ReadingProjection): readonly ReadingUnit[];
-/**
- * Where a decode would put the word boundaries, as character positions.
- *
- * Run over the whole lattice rather than the open stretches, because spacing
- * stays ambiguous where readings do not: 研究生命起源 reads identically whether
- * it splits 研究生/命 or 研究/生命, so every position in it locks, and the
- * spacing question is still open.
- */
-export declare function decodeSpacing(lattice: Lattice): readonly number[];
-export declare function decodeRun(dictionary: Dictionary, run: string, rules?: readonly EdgeRule[], before?: string, hints?: ResolvedHints): readonly DecodedWord[];
+export { decodeReadings, decodeSpacing } from "./lattice-decode.js";
+export declare function decodeRun(dictionary: Dictionary, run: string, rules?: readonly EdgeRule[], before?: string, hints?: ResolvedHints, after?: string): readonly DecodedWord[];
 /**
  * Decode a Han run, keeping what each reading was chosen over.
  *
@@ -31,9 +18,9 @@ export declare function decodeRun(dictionary: Dictionary, run: string, rules?: r
  * only worth running for a caller that will use the answer — rendering
  * uncertain readings differently, or reporting them.
  *
- * `before` is the context {@link decodeRun} takes, and means the same here: it
- * is decoded with the run and reported with neither words nor confidence of its
- * own.
+ * `before` and `after` are the context {@link decodeRun} takes, and mean the
+ * same here: they are decoded with the run and reported with neither words nor
+ * confidence of their own.
  */
-export declare function decodeRunScored(dictionary: Dictionary, run: string, rules?: readonly EdgeRule[], before?: string, hints?: ResolvedHints): readonly ScoredWord[];
+export declare function decodeRunScored(dictionary: Dictionary, run: string, rules?: readonly EdgeRule[], before?: string, hints?: ResolvedHints, after?: string): readonly ScoredWord[];
 //# sourceMappingURL=decode.d.ts.map