@kensio/pinyinjs 1.7.5 → 1.7.6

package/data/manifest.json

@@ -21,20 +21,20 @@
       "keys": "full.keys",
       "readings": "full.entries",
       "frequencies": "full.freq",
-      "bytes": 9113585,
-      "compressedBytes": 2396367
+      "bytes": 9140884,
+      "compressedBytes": 2401549
     }
   },
   "script": {
     "file": "script.map",
     "toTraditional": 5944,
     "toSimplified": 6117,
-    "traditionalWords": 2327,
+    "traditionalWords": 2390,
     "simplifiedWords": 849,
     "hansOnly": 5785,
     "hantOnly": 5497,
-    "bytes": 291502,
-    "compressedBytes": 100003
+    "bytes": 292980,
+    "compressedBytes": 100301
   },
   "sources": [
     "https://www.unicode.org/Public/UCD/latest/ucd/Unihan.zip",

package/data/script.map

@@ -273,7 +273,7 @@ t	艺	藝
 t	担	擔
 t	卧	臥
 t	卷	卷		捲
-t	历	歷	li5=曆
+t	历	歷		曆
 t	厢	廂
 t	愿	願
 t	踪	蹤
@@ -521,7 +521,7 @@ t	篑	簣
 t	类	類
 t	护	護
 t	坏	壞
-t	系	系	xi5=係,ji4=繫
+t	系	系	ji4=繫	係
 t	纠	糾
 t	纯	純
 t	络	絡
@@ -643,7 +643,7 @@ t	裤	褲
 t	钢	鋼
 t	镐	鎬
 t	宁	寧
-t	辟	闢	bi4=辟
+t	辟	闢		辟
 t	栎	櫟
 t	珰	璫
 t	税	稅
@@ -12071,6 +12071,7 @@ w	一箭双雕	一箭雙鵰
 w	一见钟情	一見鍾情
 w	丁丁炒面	丁丁炒麵
 w	丁丑	丁丑
+w	丁巳复辟	丁巳復辟
 w	丁零当啷	丁零噹啷
 w	万事俱备，只欠东风	萬事俱備，只欠東風
 w	万历	萬曆
@@ -12102,12 +12103,14 @@ w	不法分子	不法份子
 w	不相干	不相干
 w	不知就里	不知就裡
 w	不知所云	不知所云
+w	不辟斧钺	不辟斧鉞
 w	不锈钢	不銹鋼
 w	丑时	丑時
 w	丑牛	丑牛
 w	丑角	丑角
 w	丑鸭	丑鴨
 w	世上无难事，只怕有心人	世上無難事，只怕有心人
+w	世辟	世辟
 w	丛冢	叢冢
 w	东台	東台
 w	东台市	東台市
@@ -12258,8 +12261,12 @@ w	公布栏	公佈欄
 w	共轭复数	共軛複數
 w	关云长	關云長
 w	关征	關征
+w	关系	關係
+w	关系代名词	關係代名詞
 w	关系到	關係到
 w	关系式	關係式
+w	关系户	關係戶
+w	关系网	關係網
 w	兴冲冲	興沖沖
 w	兴高采烈	興高采烈
 w	兼并	兼併
@@ -12369,6 +12376,9 @@ w	划痕	劃痕
 w	划破	劃破
 w	划过	劃過
 w	列别杰夫	列別傑夫
+w	利害关系	利害關係
+w	利害关系人	利害關係人
+w	利害关系方	利害關係方
 w	利欲	利慾
 w	利欲心	利慾心
 w	利欲熏心	利慾薰心
@@ -12417,6 +12427,9 @@ w	制陶工人	製陶工人
 w	制鞋匠	製鞋匠
 w	制鞋工人	製鞋工人
 w	前车之鉴	前車之鑑
+w	割舍了	割舍了
+w	割舍得	割舍得
+w	割舍的	割舍的
 w	力争上游	力爭上游
 w	力征	力征
 w	动荡	動盪
@@ -12518,6 +12531,7 @@ w	发夹	髮夾
 w	发小儿	髮小兒
 w	发布	發佈
 w	发布会	發佈會
+w	发生关系	發生關係
 w	发面	發麵
 w	受托	受託
 w	受托人	受託人
@@ -12533,6 +12547,7 @@ w	只要功夫深，铁杵磨成针	只要功夫深，鐵杵磨成針
 w	只见树木，不见森林	只見樹木，不見森林
 w	只言片语	只言片語
 w	只许州官放火，不许百姓点灯	只許州官放火，不許百姓點燈
+w	只读存储器	隻讀存儲器
 w	叮叮当当	叮叮噹噹
 w	叮当	叮噹
 w	叮当响	叮噹響
@@ -12796,6 +12811,7 @@ w	复合母音	複合母音
 w	复合肥	複合肥
 w	复合词	複合詞
 w	复听	複聽
+w	复子明辟	復子明辟
 w	复平面	複平面
 w	复式	複式
 w	复数	複數
@@ -12820,6 +12836,7 @@ w	复词	複詞
 w	复试	複試
 w	复购	複購
 w	复赛	複賽
+w	复辟	復辟
 w	复迭	複迭
 w	复述	複述
 w	复选框	複選框
@@ -12949,6 +12966,7 @@ w	尸罗	尸羅
 w	尼采	尼采
 w	尼龙搭扣	尼龍搭釦
 w	尾注	尾註
+w	尾随不舍	尾隨不舍
 w	局促	侷促
 w	局限	侷限
 w	层岩	層巖
@@ -13054,6 +13072,7 @@ w	开杆	開杆
 w	开炮	開砲
 w	开诚布公	開誠佈公
 w	引以为鉴	引以為鑑
+w	张勋复辟	張勳復辟
 w	弥漫星云	瀰漫星雲
 w	弩炮	弩砲
 w	弯管面	彎管麵
@@ -13094,6 +13113,7 @@ w	征程	征程
 w	征衣	征衣
 w	征衫	征衫
 w	征讨	征討
+w	征辟	徵辟
 w	征途	征途
 w	征马	征馬
 w	征驾	征駕
@@ -13127,6 +13147,7 @@ w	怀里	懷裡
 w	怒冲冲	怒沖沖
 w	怒气冲冲	怒氣沖沖
 w	性伙伴	性夥伴
+w	性关系	性關係
 w	性欲	性慾
 w	性欲高潮	性慾高潮
 w	怨气冲天	怨氣沖天
@@ -13270,9 +13291,11 @@ w	撒马尔干	撒馬爾干
 w	撤并	撤併
 w	擀面杖	擀麵杖
 w	操作台	操作台
+w	攘辟	攘辟
 w	支努干	支努干
 w	收获节	收穫節
 w	收银台	收銀台
+w	改善关系	改善關係
 w	攻占	攻占
 w	放在眼里	放在眼裡
 w	放松	放鬆
@@ -13312,12 +13335,14 @@ w	无干	無干
 w	无筋面粉	無筋麵粉
 w	无精打采	無精打采
 w	无须	無須
+w	日佣人	日佣人
 w	日历	日曆
 w	日志	日誌
 w	日本国志	日本國誌
 w	日里	日裡
 w	旧历	舊曆
 w	旧历年	舊曆年
+w	旧皇历	舊皇曆
 w	昆仑	崑崙
 w	昆仑山	崑崙山
 w	昆仑山脉	崑崙山脈
@@ -13328,6 +13353,7 @@ w	昆山市	崑山市
 w	昆曲	崑曲
 w	昆腔	崑腔
 w	明窗净几	明窗淨几
+w	明辟	明辟
 w	明里暗里	明裡暗裡
 w	星历	星曆
 w	春光乍泄	春光乍泄
@@ -13478,6 +13504,8 @@ w	沉淀	沈澱
 w	沉溺	沈溺
 w	沉鱼落雁	沈魚落雁
 w	沙坪坝	沙坪垻
+w	没关系	沒關係
+w	没有关系	沒有關係
 w	没精打采	沒精打采
 w	河涌	河涌
 w	油泼扯面	油潑扯麵
@@ -13521,9 +13549,11 @@ w	流里流气	流裡流氣
 w	测杆	測杆
 w	浑仪注	渾儀註
 w	浓郁	濃郁
+w	浙江天台县	浙江天臺縣
 w	浮游	浮游
 w	浮游生物	浮游生物
 w	海于格松	海于格松
+w	海峡两岸关系协会	海峽兩岸關係協會
 w	海淀	海淀
 w	海淀区	海淀區
 w	海淀图书城	海淀圖書城
@@ -13657,6 +13687,7 @@ w	生态环境游	生態環境游
 w	生物制剂	生物製劑
 w	生物制品	生物製品
 w	生面团	生麵糰
+w	用尽地	用儘地
 w	由表及里	由表及裡
 w	电子表	電子錶
 w	电影制作	電影製作
@@ -13686,7 +13717,9 @@ w	白腹隼雕	白腹隼鵰
 w	白面	白麵
 w	白面儿	白麵兒
 w	百谷	百穀
+w	百辟	百辟
 w	百里挑一	百裡挑一
+w	皇历	皇曆
 w	皇后	皇后
 w	皇后区	皇后區
 w	皇后镇	皇后鎮
@@ -13719,6 +13752,7 @@ w	瞈蒙	瞈矇
 w	矢志	矢誌
 w	知识分子	知識份子
 w	矬子里拔将军	矬子裡拔將軍
+w	短发地	短發地
 w	短趾雕	短趾鵰
 w	矮子里拔将军	矮子裡拔將軍
 w	石台	石台
@@ -13730,6 +13764,7 @@ w	砚台	硯臺
 w	砷制剂	砷製劑
 w	碑志	碑誌
 w	碱水面包	鹼水麵包
+w	磁流体发电	磁流體髮電
 w	磁轨炮	磁軌砲
 w	磨制石器	磨製石器
 w	示复	示覆
@@ -13754,6 +13789,10 @@ w	禾谷	禾穀
 w	秃发	禿發
 w	秋千	鞦韆
 w	秋征	秋征
+w	科斗书	科鬥書
+w	科斗字	科鬥字
+w	科斗筯	科鬥筯
+w	科斗籍	科鬥籍
 w	秕谷	秕穀
 w	秘传	祕傳
 w	秘制	秘製
@@ -13810,6 +13849,7 @@ w	系囚	繫囚
 w	系放	繫放
 w	系数	係數
 w	素朴	素樸
+w	紧跟不舍	緊跟不舍
 w	紫云苗族布依族自治县	紫云苗族布依族自治縣
 w	紫金山天文台	紫金山天文台
 w	繁复	繁複
@@ -13845,6 +13885,7 @@ w	美中台	美中台
 w	美体小铺	美體小舖
 w	美制	美製
 w	美国存托凭证	美國存託憑證
+w	群辟	群辟
 w	羽状复叶	羽狀複葉
 w	翻老皇历	翻老皇曆
 w	老家伙	老傢伙
@@ -13871,6 +13912,7 @@ w	肉汤面	肉湯麵
 w	肉烂在锅里	肉爛在鍋裡
 w	肌肉松弛剂	肌肉鬆弛劑
 w	股栗	股慄
+w	肮脏气	骯臟氣
 w	肺炎霉浆菌	肺炎黴漿菌
 w	背债	揹債
 w	背地里	背地裡
@@ -13883,6 +13925,7 @@ w	胡茬	鬍茬
 w	胡须	鬍鬚
 w	胡髭	鬍髭
 w	脆谷乐	脆穀樂
+w	脏躁症	髒躁症
 w	脑力激荡	腦力激盪
 w	脑震荡	腦震盪
 w	脱肛	脫疘
@@ -13900,6 +13943,7 @@ w	臭熏熏	臭燻燻
 w	致密	緻密
 w	臾须	臾須
 w	舍得一身剐，敢把皇帝拉下马	捨得一身剮，敢把皇帝拉下馬
+w	舍身饲虎	舍身飼虎
 w	舍车保帅	捨車保帥
 w	舒卷	舒捲
 w	舞娘	舞孃
@@ -13996,6 +14040,7 @@ w	表针	錶針
 w	衬里	襯裡
 w	袁于令	袁于令
 w	裁制	裁製
+w	裙带关系	裙帶關係
 w	褟绦子	褟縧子
 w	西历	西曆
 w	西岳	西嶽
@@ -14078,6 +14123,8 @@ w	赞誉	讚譽
 w	赞许	讚許
 w	赞赏	讚賞
 w	赞颂	讚頌
+w	起根发脚	起根發腳
+w	超友谊关系	超友誼關係
 w	趋吉避凶	趨吉避凶
 w	跋杯	跋桮
 w	跑表	跑錶
@@ -14086,6 +14133,7 @@ w	跳梁小丑	跳梁小丑
 w	躲清闲	躲清閑
 w	车厘子	車厘子
 w	轧制	軋製
+w	转关系	轉關係
 w	转制	轉製
 w	转台	轉台
 w	转车台	轉車台
@@ -14097,7 +14145,16 @@ w	轮奸	輪姦
 w	轻松	輕鬆
 w	辛丑	辛丑
 w	辛丑条约	辛丑條約
+w	辟举	辟舉
+w	辟书	辟書
+w	辟召	辟召
+w	辟吏	辟吏
+w	辟引	辟引
+w	辟纑	辟纑
+w	辟言	辟言
 w	辟谷	辟穀
+w	辟邪	辟邪
+w	辟雍	辟雍
 w	辟雍砚	辟雍硯
 w	边反卷地	邊反捲地
 w	辽沈战役	遼瀋戰役
@@ -14110,6 +14167,7 @@ w	连三并四	連三併四
 w	连珠炮	連珠砲
 w	连系	連繫
 w	连系词	連繫詞
+w	连辟公府不就	連辟公府不就
 w	迷奸	迷姦
 w	迷蒙	迷濛
 w	选美皇后	選美皇后
@@ -14318,7 +14376,10 @@ w	面馆	麵館
 w	革制品	革製品
 w	靴隼雕	靴隼鵰
 w	鞣制	鞣製
+w	鞭辟	鞭辟
 w	鞭辟入里	鞭闢入裡
+w	鞭辟向里	鞭辟向里
+w	鞭辟着里	鞭辟著里
 w	韬戈卷甲	韜戈捲甲
 w	韵致	韻緻
 w	页签	頁籤
@@ -14360,6 +14421,7 @@ w	骨坛	骨罈
 w	骨子里	骨子裡
 w	骨质疏松	骨質疏鬆
 w	骨质疏松症	骨質疏鬆症
+w	骯脏气	骯臟氣
 w	骰钟	骰鍾
 w	高台	高台
 w	高台县	高台縣
@@ -14374,6 +14436,7 @@ w	鸡蛋里挑骨头	雞蛋裡挑骨頭
 w	鹊巢鸠占	鵲巢鳩占
 w	麸皮面包	麩皮麵包
 w	麻雀虽小，五脏俱全	麻雀雖小，五臟俱全
+w	黄历	黃曆
 w	黄金周	黃金週
 w	黄钟毁弃瓦釜雷鸣	黃鐘譭棄瓦釜雷鳴
 w	黑嘴松鸡	黑嘴鬆雞

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

@@ -0,0 +1,27 @@
+/**
+ * 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.
+ *
+ * 那座桥不长, 生得又瘦又长, 距离还长. The adverb is the near side and the end of
+ * the clause is the far side, and it takes both: 不长 alone is 不长在耳朵外面
+ * as readily as it is 不长.
+ */
+export declare function isPredicated(context: EdgeContext): boolean;
+/**
+ * Whether a 长 is an adjective in front of the noun it modifies.
+ *
+ * 长头发, 长耳朵, 长胡子, 长裙子. The noun after it says nothing on its own,
+ * since 长毛 and 长知识 are a verb and its object wearing the same shape, so the
+ * 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.
+ */
+export declare function isAttributive(context: EdgeContext): boolean;
+//# sourceMappingURL=chang-context.d.ts.map

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

@@ -0,0 +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"}

package/dist/decode/chang-context.js

@@ -0,0 +1,66 @@
+/**
+ * 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 { tagOf, wordEndingAt, wordStartingAt, } from "./rules.js";
+/**
+ * The adverbs {@link DEGREE} leaves out, because they scope a verb just as well.
+ *
+ * 那座桥不长 and 距离还长 are both `cháng`, and the degree set says why it will
+ * not have them: 胡子不长在前额上 and 还长有毛 put the
+ * same two characters around the verb. What tells the two apart is the other
+ * side. A growing 长 governs something — 在前额上, 有毛, 一个包 — while the
+ * adjective has nothing left to say, so the clause ends.
+ */
+const SCOPING = new Set(["不", "还", "還", "又", "已经", "已經"]);
+/**
+ * What may stand between a predicative 长 and the end of its clause.
+ *
+ * Only the particles that close one. Everything a verb could govern is
+ * therefore excluded by construction, which is the whole of the test.
+ */
+const CLOSING = new Set(["了", "的", "呢", "吗", "嗎", "啊", "呀"]);
+/**
+ * Whether a 长 is the whole of what an adverb is scoping.
+ *
+ * 那座桥不长, 生得又瘦又长, 距离还长. The adverb is the near side and the end of
+ * the clause is the far side, and it takes both: 不长 alone is 不长在耳朵外面
+ * as readily as it is 不长.
+ */
+export function isPredicated(context) {
+    const { characters, edge } = context;
+    if (!SCOPING.has(characters[edge.from - 1] ?? "")) {
+        return false;
+    }
+    const after = characters[edge.to];
+    return after === undefined || CLOSING.has(after);
+}
+/**
+ * The tag jieba gives a 量词.
+ */
+const MEASURE = "q";
+/**
+ * The tag prefix a noun carries, which is what an attributive 长 modifies.
+ */
+const NOUN_TAG = "n";
+/**
+ * Whether a 长 is an adjective in front of the noun it modifies.
+ *
+ * 长头发, 长耳朵, 长胡子, 长裙子. The noun after it says nothing on its own,
+ * since 长毛 and 长知识 are a verb and its object wearing the same shape, so the
+ * 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.
+ */
+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;
+}
+//# sourceMappingURL=chang-context.js.map

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

@@ -0,0 +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"}

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

@@ -0,0 +1,37 @@
+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`.
+ *
+ * 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.
+ *
+ * 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.
+ *
+ * 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.
+ *
+ * The last piece is {@link GROWN_INTO}, which pushes the opposite way on the one
+ * shape the sources get wrong.
+ */
+export declare const ADJECTIVAL_CHANG: EdgeRule;
+//# sourceMappingURL=chang-rule.d.ts.map

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

@@ -0,0 +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"}

package/dist/decode/chang-rule.js

@@ -0,0 +1,169 @@
+import { isAttributive, isPredicated } from "./chang-context.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`.
+ *
+ * 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.
+ *
+ * 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.
+ *
+ * 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.
+ *
+ * The last piece 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";
+        }
+        if (isPredicated(context) || isAttributive(context)) {
+            return "force";
+        }
+        const before = wordEndingAt(context, edge.from);
+        return before !== undefined && DEGREE.has(before) ? "force" : "keep";
+    },
+};
+//# sourceMappingURL=chang-rule.js.map