@kensio/pinyinjs 1.7.5 → 1.7.7

package/dist/decode/edge-context.js

@@ -19,6 +19,30 @@ export function wordEndingAt(context, at, longest = 4) {
     }
     return undefined;
 }
+/**
+ * Every word the dictionary has ending at a position, longest first.
+ *
+ * {@link wordEndingAt} answers with the longest, which is what a rule asking
+ * "what word is this" wants. A rule asking whether a *particular* word stands
+ * here wants all of them, because the longest match is often a word the rule
+ * has no opinion about wrapped around the one it does: 很 is a degree adverb
+ * and 得很 is the longest word ending at the same position, so 留得很长 hid its
+ * 很 and read `zhǎng`. 一个 hides 个 the same way.
+ */
+export function wordsEndingAt(context, at, longest = 4) {
+    const held = [];
+    for (let length = longest; length > 0; length--) {
+        const from = at - length;
+        if (from < 0) {
+            continue;
+        }
+        const text = context.characters.slice(from, at).join("");
+        if (context.dictionary.lookup(text) !== undefined) {
+            held.push(text);
+        }
+    }
+    return held;
+}
 /**
  * The word starting at a position, longest first, or undefined.
  */
@@ -33,6 +57,23 @@ export function wordStartingAt(context, at, longest = 4) {
     }
     return undefined;
 }
+/**
+ * Every word the dictionary has starting at a position, longest first.
+ *
+ * The mirror of {@link wordsEndingAt}, and needed for the same reason: 五公 is a
+ * word and hides the 五 inside it, so a rule asking whether a numeral starts
+ * here has to be shown all of them.
+ */
+export function wordsStartingAt(context, at, longest = 4) {
+    const held = [];
+    for (let length = longest; length > 0; length--) {
+        const text = context.characters.slice(at, at + length).join("");
+        if (text !== "" && context.dictionary.lookup(text) !== undefined) {
+            held.push(text);
+        }
+    }
+    return held;
+}
 /**
  * The part of speech the dictionary gives a word, or the empty string.
  */

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

@@ -1 +1 @@
-{"version":3,"file":"edge-context.js","sourceRoot":"","sources":["../../src/decode/edge-context.ts"],"names":[],"mappings":"AAQA;;;;;;;GAOG;AACH,MAAM,UAAU,YAAY,CAC1B,OAAoB,EACpB,EAAU,EACV,OAAO,GAAG,CAAC;IAEX,KAAK,IAAI,MAAM,GAAG,OAAO,EAAE,MAAM,GAAG,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;QAChD,MAAM,IAAI,GAAG,EAAE,GAAG,MAAM,CAAC;QACzB,IAAI,IAAI,GAAG,CAAC,EAAE,CAAC;YACb,SAAS;QACX,CAAC;QACD,MAAM,IAAI,GAAG,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACzD,IAAI,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC;YAClD,OAAO,IAAI,CAAC;QACd,CAAC;IACH,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,cAAc,CAC5B,OAAoB,EACpB,EAAU,EACV,OAAO,GAAG,CAAC;IAEX,KAAK,IAAI,MAAM,GAAG,OAAO,EAAE,MAAM,GAAG,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;QAChD,mEAAmE;QACnE,sBAAsB;QACtB,MAAM,IAAI,GAAG,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,GAAG,MAAM,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QAChE,IAAI,IAAI,KAAK,EAAE,IAAI,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC;YACjE,OAAO,IAAI,CAAC;QACd,CAAC;IACH,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,KAAK,CAAC,OAAoB,EAAE,IAAwB;IAClE,OAAO,IAAI,KAAK,SAAS;QACvB,CAAC,CAAC,EAAE;QACJ,CAAC,CAAC,CAAC,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,YAAY,IAAI,EAAE,CAAC,CAAC;AAC5D,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,CAAC,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAE9D;;GAEG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,GAAG,CAAC"}
+{"version":3,"file":"edge-context.js","sourceRoot":"","sources":["../../src/decode/edge-context.ts"],"names":[],"mappings":"AAQA;;;;;;;GAOG;AACH,MAAM,UAAU,YAAY,CAC1B,OAAoB,EACpB,EAAU,EACV,OAAO,GAAG,CAAC;IAEX,KAAK,IAAI,MAAM,GAAG,OAAO,EAAE,MAAM,GAAG,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;QAChD,MAAM,IAAI,GAAG,EAAE,GAAG,MAAM,CAAC;QACzB,IAAI,IAAI,GAAG,CAAC,EAAE,CAAC;YACb,SAAS;QACX,CAAC;QACD,MAAM,IAAI,GAAG,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACzD,IAAI,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC;YAClD,OAAO,IAAI,CAAC;QACd,CAAC;IACH,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED;;;;;;;;;GASG;AACH,MAAM,UAAU,aAAa,CAC3B,OAAoB,EACpB,EAAU,EACV,OAAO,GAAG,CAAC;IAEX,MAAM,IAAI,GAAa,EAAE,CAAC;IAC1B,KAAK,IAAI,MAAM,GAAG,OAAO,EAAE,MAAM,GAAG,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;QAChD,MAAM,IAAI,GAAG,EAAE,GAAG,MAAM,CAAC;QACzB,IAAI,IAAI,GAAG,CAAC,EAAE,CAAC;YACb,SAAS;QACX,CAAC;QACD,MAAM,IAAI,GAAG,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QACzD,IAAI,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC;YAClD,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAClB,CAAC;IACH,CAAC;IACD,OAAO,IAAI,CAAC;AACd,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,cAAc,CAC5B,OAAoB,EACpB,EAAU,EACV,OAAO,GAAG,CAAC;IAEX,KAAK,IAAI,MAAM,GAAG,OAAO,EAAE,MAAM,GAAG,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;QAChD,mEAAmE;QACnE,sBAAsB;QACtB,MAAM,IAAI,GAAG,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,GAAG,MAAM,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QAChE,IAAI,IAAI,KAAK,EAAE,IAAI,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC;YACjE,OAAO,IAAI,CAAC;QACd,CAAC;IACH,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,eAAe,CAC7B,OAAoB,EACpB,EAAU,EACV,OAAO,GAAG,CAAC;IAEX,MAAM,IAAI,GAAa,EAAE,CAAC;IAC1B,KAAK,IAAI,MAAM,GAAG,OAAO,EAAE,MAAM,GAAG,CAAC,EAAE,MAAM,EAAE,EAAE,CAAC;QAChD,MAAM,IAAI,GAAG,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,GAAG,MAAM,CAAC,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;QAChE,IAAI,IAAI,KAAK,EAAE,IAAI,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,SAAS,EAAE,CAAC;YACjE,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAClB,CAAC;IACH,CAAC;IACD,OAAO,IAAI,CAAC;AACd,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,KAAK,CAAC,OAAoB,EAAE,IAAwB;IAClE,OAAO,IAAI,KAAK,SAAS;QACvB,CAAC,CAAC,EAAE;QACJ,CAAC,CAAC,CAAC,OAAO,CAAC,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,YAAY,IAAI,EAAE,CAAC,CAAC;AAC5D,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,CAAC,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC;AAE9D;;GAEG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,GAAG,CAAC"}

package/dist/decode/hint-lattice.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"hint-lattice.d.ts","sourceRoot":"","sources":["../../src/decode/hint-lattice.ts"],"names":[],"mappings":"AAQA,OAAO,EACL,KAAK,OAAO,EAIb,MAAM,cAAc,CAAC;AACtB,OAAO,KAAK,EAAE,aAAa,EAAY,MAAM,YAAY,CAAC;AAE1D;;;;;;;GAOG;AACH,wBAAgB,UAAU,CAAC,KAAK,EAAE,aAAa,EAAE,EAAE,EAAE,MAAM,GAAG,aAAa,CAK1E;AAED;;;;;;;GAOG;AACH,wBAAgB,WAAW,CACzB,KAAK,EAAE,aAAa,EACpB,IAAI,EAAE,MAAM,EACZ,MAAM,EAAE,MAAM,GACb,aAAa,CAQf;AAeD;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,wBAAgB,UAAU,CAAC,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,aAAa,GAAG,OAAO,CAyF1E"}
+{"version":3,"file":"hint-lattice.d.ts","sourceRoot":"","sources":["../../src/decode/hint-lattice.ts"],"names":[],"mappings":"AAQA,OAAO,EAAE,KAAK,OAAO,EAAoC,MAAM,cAAc,CAAC;AAC9E,OAAO,KAAK,EAAE,aAAa,EAAY,MAAM,YAAY,CAAC;AAiB1D;;;;;;;GAOG;AACH,wBAAgB,UAAU,CAAC,KAAK,EAAE,aAAa,EAAE,EAAE,EAAE,MAAM,GAAG,aAAa,CAK1E;AAED;;;;;;;GAOG;AACH,wBAAgB,WAAW,CACzB,KAAK,EAAE,aAAa,EACpB,IAAI,EAAE,MAAM,EACZ,MAAM,EAAE,MAAM,GACb,aAAa,CAQf;AAeD;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,wBAAgB,UAAU,CAAC,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,aAAa,GAAG,OAAO,CAyF1E"}

package/dist/decode/hint-lattice.js

@@ -1,4 +1,18 @@
-import { READING_CHARGE, UNKNOWN_COST, } from "./lattice.js";
+import { READING_CHARGE } from "./lattice.js";
+/**
+ * What an edge a hint brought with it costs, in both decodes.
+ *
+ * Zero, because a caller naming a reading is the strongest evidence the decode
+ * has. The other branch of {@link applyHints} says so already by overwriting a
+ * dictionary entry's reading where a hint names the same stretch, and an edge
+ * with no entry behind it has to say the same thing through its cost.
+ *
+ * It was {@link import("./lattice-types.js").UNKNOWN_COST} until
+ * {@link READING_CHARGE} came down, which worked only while one edge beat two
+ * whatever they cost: 玩儿 hinted as `wánr` went back to `wán ér` the moment
+ * frequency could speak.
+ */
+const HINT_COST = 0;
 /**
  * The same hints with every position moved along by an offset.
  *
@@ -123,8 +137,8 @@ export function applyHints(lattice, hints) {
                     to,
                     text: spanning.characters.join(""),
                     reading: spanning.reading,
-                    cost: UNKNOWN_COST,
-                    readingCost: UNKNOWN_COST + READING_CHARGE,
+                    cost: HINT_COST,
+                    readingCost: HINT_COST + READING_CHARGE,
                     isProperNoun: false,
                     partOfSpeech: "",
                     // No dictionary entry backs this one — that is the whole reason it

package/dist/decode/hint-lattice.js.map

@@ -1 +1 @@
-{"version":3,"file":"hint-lattice.js","sourceRoot":"","sources":["../../src/decode/hint-lattice.ts"],"names":[],"mappings":"AAQA,OAAO,EAGL,cAAc,EACd,YAAY,GACb,MAAM,cAAc,CAAC;AAGtB;;;;;;;GAOG;AACH,MAAM,UAAU,UAAU,CAAC,KAAoB,EAAE,EAAU;IACzD,OAAO;QACL,KAAK,EAAE,KAAK,CAAC,KAAK;QAClB,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,GAAG,KAAK,CAAC,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,GAAG,EAAE,EAAE,QAAQ,CAAC,CAAC,CAAC;KACxE,CAAC;AACJ,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,WAAW,CACzB,KAAoB,EACpB,IAAY,EACZ,MAAc;IAEd,MAAM,EAAE,GAAG,IAAI,GAAG,EAAoB,CAAC;IACvC,KAAK,MAAM,CAAC,QAAQ,EAAE,QAAQ,CAAC,IAAI,KAAK,CAAC,EAAE,EAAE,CAAC;QAC5C,IAAI,QAAQ,IAAI,IAAI,IAAI,QAAQ,GAAG,IAAI,GAAG,MAAM,EAAE,CAAC;YACjD,EAAE,CAAC,GAAG,CAAC,QAAQ,GAAG,IAAI,EAAE,QAAQ,CAAC,CAAC;QACpC,CAAC;IACH,CAAC;IACD,OAAO,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,EAAE,EAAE,CAAC;AACpC,CAAC;AAED;;GAEG;AACH,SAAS,SAAS,CAChB,IAAc,EACd,UAA6B,EAC7B,EAAU;IAEV,OAAO,IAAI,CAAC,UAAU,CAAC,KAAK,CAC1B,CAAC,SAAS,EAAE,KAAK,EAAE,EAAE,CAAC,UAAU,CAAC,EAAE,GAAG,KAAK,CAAC,KAAK,SAAS,CAC3D,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,MAAM,UAAU,UAAU,CAAC,OAAgB,EAAE,KAAoB;IAC/D,MAAM,EAAE,UAAU,EAAE,GAAG,OAAO,CAAC;IAE/B,+EAA+E;IAC/E,MAAM,MAAM,GAAG,CAAC,EAAU,EAAgD,EAAE;QAC1E,KAAK,MAAM,IAAI,IAAI,KAAK,CAAC,KAAK,EAAE,CAAC;YAC/B,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,UAAU,CAAC,MAAM,EAAE,CAAC;gBACnD,SAAS;YACX,CAAC;YACD,KAAK,IAAI,IAAI,GAAG,CAAC,EAAE,IAAI,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM,EAAE,IAAI,EAAE,EAAE,CAAC;gBACzD,IAAI,EAAE,GAAG,IAAI,IAAI,CAAC,IAAI,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,EAAE,GAAG,IAAI,CAAC,EAAE,CAAC;oBAC7D,OAAO,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC;gBACxB,CAAC;YACH,CAAC;QACH,CAAC;QACD,OAAO,SAAS,CAAC;IACnB,CAAC,CAAC;IAEF,MAAM,OAAO,GAAG,CAAC,IAAiB,EAAe,EAAE;QACjD,MAAM,IAAI,GAAG,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,CAAC;QAEjC,yEAAyE;QACzE,2EAA2E;QAC3E,qEAAqE;QACrE,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,IAAI,CAC5B,CAAC,IAAI,EAAE,EAAE,CACP,IAAI,CAAC,UAAU,CAAC,MAAM,KAAK,IAAI;YAC/B,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,IAAI,CAAC,IAAI,CAAC,CACzC,CAAC;QACF,IAAI,OAAO,GAAG,KAAK,KAAK,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC;QAEjE,0EAA0E;QAC1E,gBAAgB;QAChB,IAAI,OAAO,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC;YAC5B,OAAO,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,KAAK,EAAE,EAAE;gBACxC,MAAM,EAAE,GAAG,IAAI,CAAC,IAAI,GAAG,KAAK,CAAC;gBAC7B,MAAM,UAAU,GAAG,KAAK,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;gBACpC,IAAI,UAAU,KAAK,SAAS,EAAE,CAAC;oBAC7B,OAAO,UAAU,CAAC;gBACpB,CAAC;gBACD,IAAI,IAAI,GAAG,CAAC,EAAE,CAAC;oBACb,OAAO,QAAQ,CAAC;gBAClB,CAAC;gBACD,sEAAsE;gBACtE,qEAAqE;gBACrE,MAAM,KAAK,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC;gBACzB,OAAO,KAAK,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,QAAQ,CAAC;YACrD,CAAC,CAAC,CAAC;QACL,CAAC;QAED,OAAO,OAAO,KAAK,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,GAAG,IAAI,EAAE,OAAO,EAAE,CAAC;IAChE,CAAC,CAAC;IAEF,OAAO;QACL,UAAU;QACV,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,EAAE,EAAE;YACrC,MAAM,SAAS,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC;YACrD,0EAA0E;YAC1E,sEAAsE;YACtE,MAAM,QAAQ,GAAG,KAAK,CAAC,KAAK,CAAC,IAAI,CAC/B,CAAC,IAAI,EAAE,EAAE,CACP,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,UAAU,CAAC,MAAM;gBAC9C,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,EAAE,CAAC;gBAC/B,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC,CACrE,CAAC;YACF,IAAI,QAAQ,KAAK,SAAS,EAAE,CAAC;gBAC3B,OAAO,SAAS,CAAC;YACnB,CAAC;YACD,MAAM,EAAE,GAAG,EAAE,GAAG,QAAQ,CAAC,UAAU,CAAC,MAAM,CAAC;YAC3C,OAAO;gBACL,GAAG,SAAS;gBACZ;oBACE,IAAI,EAAE,EAAE;oBACR,EAAE;oBACF,IAAI,EAAE,QAAQ,CAAC,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC;oBAClC,OAAO,EAAE,QAAQ,CAAC,OAAO;oBACzB,IAAI,EAAE,YAAY;oBAClB,WAAW,EAAE,YAAY,GAAG,cAAc;oBAC1C,YAAY,EAAE,KAAK;oBACnB,YAAY,EAAE,EAAE;oBAChB,mEAAmE;oBACnE,oEAAoE;oBACpE,sEAAsE;oBACtE,6DAA6D;oBAC7D,OAAO,EAAE,KAAK;iBACf;aACF,CAAC;QACJ,CAAC,CAAC;KACH,CAAC;AACJ,CAAC"}
+{"version":3,"file":"hint-lattice.js","sourceRoot":"","sources":["../../src/decode/hint-lattice.ts"],"names":[],"mappings":"AAQA,OAAO,EAAkC,cAAc,EAAE,MAAM,cAAc,CAAC;AAG9E;;;;;;;;;;;;GAYG;AACH,MAAM,SAAS,GAAG,CAAC,CAAC;AAEpB;;;;;;;GAOG;AACH,MAAM,UAAU,UAAU,CAAC,KAAoB,EAAE,EAAU;IACzD,OAAO;QACL,KAAK,EAAE,KAAK,CAAC,KAAK;QAClB,EAAE,EAAE,IAAI,GAAG,CAAC,CAAC,GAAG,KAAK,CAAC,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,QAAQ,CAAC,EAAE,EAAE,CAAC,CAAC,EAAE,GAAG,EAAE,EAAE,QAAQ,CAAC,CAAC,CAAC;KACxE,CAAC;AACJ,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,WAAW,CACzB,KAAoB,EACpB,IAAY,EACZ,MAAc;IAEd,MAAM,EAAE,GAAG,IAAI,GAAG,EAAoB,CAAC;IACvC,KAAK,MAAM,CAAC,QAAQ,EAAE,QAAQ,CAAC,IAAI,KAAK,CAAC,EAAE,EAAE,CAAC;QAC5C,IAAI,QAAQ,IAAI,IAAI,IAAI,QAAQ,GAAG,IAAI,GAAG,MAAM,EAAE,CAAC;YACjD,EAAE,CAAC,GAAG,CAAC,QAAQ,GAAG,IAAI,EAAE,QAAQ,CAAC,CAAC;QACpC,CAAC;IACH,CAAC;IACD,OAAO,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,EAAE,EAAE,CAAC;AACpC,CAAC;AAED;;GAEG;AACH,SAAS,SAAS,CAChB,IAAc,EACd,UAA6B,EAC7B,EAAU;IAEV,OAAO,IAAI,CAAC,UAAU,CAAC,KAAK,CAC1B,CAAC,SAAS,EAAE,KAAK,EAAE,EAAE,CAAC,UAAU,CAAC,EAAE,GAAG,KAAK,CAAC,KAAK,SAAS,CAC3D,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,MAAM,UAAU,UAAU,CAAC,OAAgB,EAAE,KAAoB;IAC/D,MAAM,EAAE,UAAU,EAAE,GAAG,OAAO,CAAC;IAE/B,+EAA+E;IAC/E,MAAM,MAAM,GAAG,CAAC,EAAU,EAAgD,EAAE;QAC1E,KAAK,MAAM,IAAI,IAAI,KAAK,CAAC,KAAK,EAAE,CAAC;YAC/B,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,UAAU,CAAC,MAAM,EAAE,CAAC;gBACnD,SAAS;YACX,CAAC;YACD,KAAK,IAAI,IAAI,GAAG,CAAC,EAAE,IAAI,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM,EAAE,IAAI,EAAE,EAAE,CAAC;gBACzD,IAAI,EAAE,GAAG,IAAI,IAAI,CAAC,IAAI,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,EAAE,GAAG,IAAI,CAAC,EAAE,CAAC;oBAC7D,OAAO,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC;gBACxB,CAAC;YACH,CAAC;QACH,CAAC;QACD,OAAO,SAAS,CAAC;IACnB,CAAC,CAAC;IAEF,MAAM,OAAO,GAAG,CAAC,IAAiB,EAAe,EAAE;QACjD,MAAM,IAAI,GAAG,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,CAAC;QAEjC,yEAAyE;QACzE,2EAA2E;QAC3E,qEAAqE;QACrE,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,IAAI,CAC5B,CAAC,IAAI,EAAE,EAAE,CACP,IAAI,CAAC,UAAU,CAAC,MAAM,KAAK,IAAI;YAC/B,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,IAAI,CAAC,IAAI,CAAC,CACzC,CAAC;QACF,IAAI,OAAO,GAAG,KAAK,KAAK,SAAS,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK,CAAC,OAAO,CAAC;QAEjE,0EAA0E;QAC1E,gBAAgB;QAChB,IAAI,OAAO,CAAC,MAAM,KAAK,IAAI,EAAE,CAAC;YAC5B,OAAO,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,QAAQ,EAAE,KAAK,EAAE,EAAE;gBACxC,MAAM,EAAE,GAAG,IAAI,CAAC,IAAI,GAAG,KAAK,CAAC;gBAC7B,MAAM,UAAU,GAAG,KAAK,CAAC,EAAE,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;gBACpC,IAAI,UAAU,KAAK,SAAS,EAAE,CAAC;oBAC7B,OAAO,UAAU,CAAC;gBACpB,CAAC;gBACD,IAAI,IAAI,GAAG,CAAC,EAAE,CAAC;oBACb,OAAO,QAAQ,CAAC;gBAClB,CAAC;gBACD,sEAAsE;gBACtE,qEAAqE;gBACrE,MAAM,KAAK,GAAG,MAAM,CAAC,EAAE,CAAC,CAAC;gBACzB,OAAO,KAAK,EAAE,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,QAAQ,CAAC;YACrD,CAAC,CAAC,CAAC;QACL,CAAC;QAED,OAAO,OAAO,KAAK,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,GAAG,IAAI,EAAE,OAAO,EAAE,CAAC;IAChE,CAAC,CAAC;IAEF,OAAO;QACL,UAAU;QACV,KAAK,EAAE,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,EAAE,EAAE;YACrC,MAAM,SAAS,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC;YACrD,0EAA0E;YAC1E,sEAAsE;YACtE,MAAM,QAAQ,GAAG,KAAK,CAAC,KAAK,CAAC,IAAI,CAC/B,CAAC,IAAI,EAAE,EAAE,CACP,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,UAAU,CAAC,MAAM;gBAC9C,SAAS,CAAC,IAAI,EAAE,UAAU,EAAE,EAAE,CAAC;gBAC/B,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,GAAG,IAAI,CAAC,UAAU,CAAC,MAAM,CAAC,CACrE,CAAC;YACF,IAAI,QAAQ,KAAK,SAAS,EAAE,CAAC;gBAC3B,OAAO,SAAS,CAAC;YACnB,CAAC;YACD,MAAM,EAAE,GAAG,EAAE,GAAG,QAAQ,CAAC,UAAU,CAAC,MAAM,CAAC;YAC3C,OAAO;gBACL,GAAG,SAAS;gBACZ;oBACE,IAAI,EAAE,EAAE;oBACR,EAAE;oBACF,IAAI,EAAE,QAAQ,CAAC,UAAU,CAAC,IAAI,CAAC,EAAE,CAAC;oBAClC,OAAO,EAAE,QAAQ,CAAC,OAAO;oBACzB,IAAI,EAAE,SAAS;oBACf,WAAW,EAAE,SAAS,GAAG,cAAc;oBACvC,YAAY,EAAE,KAAK;oBACnB,YAAY,EAAE,EAAE;oBAChB,mEAAmE;oBACnE,oEAAoE;oBACpE,sEAAsE;oBACtE,6DAA6D;oBAC7D,OAAO,EAAE,KAAK;iBACf;aACF,CAAC;QACJ,CAAC,CAAC;KACH,CAAC;AACJ,CAAC"}

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

@@ -25,6 +25,8 @@ export declare const TAUGHT_TAGS: Set<string>;
  * a 教 with a pronoun, a noun or a name after it is governing it, and a 教 with
  * 了, 过, 着 or 得 after it is a verb whatever follows that. A particle in front
  * of it rules it out, since a verb does not follow one.
+ *
+ * {@link GOVERNING} is the case where nothing follows to look at.
  */
 export declare function isTeachingAt(context: EdgeContext, at: number): boolean;
 /**
@@ -46,13 +48,17 @@ export declare function isTeachingAt(context: EdgeContext, at: number): boolean;
  * one that matters: 任教, 宗教, 主教, 佛教, 传教, 执教, 请教, 家教.
  *
  * Measured over 88,866 lines of Tatoeba and zh.wikipedia, 181 教 decode as a
- * word of their own and every one of them read `jiào`. This moves 158 to
+ * word of their own and every one of them read `jiào`. This moves 162 to
  * `jiāo` and is wrong on three: 统一教创始人, 方法教深思 and 做到了教政分离,
  * where a nominal compound takes an object's shape, the last of them because
  * 做到了 is a word and so the 了 in front is not a particle to look at. Of the
- * 21 it leaves, 8 are right — 诸教中, 教外别传, 董教总, 风教, 教青局 twice, a
- * 教室 reached through a bad split, and a 教 used as a bare noun — and 13 are
+ * 17 it leaves, 8 are right — 诸教中, 教外别传, 董教总, 风教, 教青局 twice, a
+ * 教室 reached through a bad split, and a 教 used as a bare noun — and 9 are
  * misses, where jieba tags the object something this does not name.
+ *
+ * {@link GOVERNING} is four of that 162 and was added later: 能教教我,
+ * 要教一節課, 也會教像拉丁語, 這在學校是不教的. 有 is pointedly outside it,
+ * since 有教无类 is `yǒujiào wúlèi`.
  */
 export declare const TEACHING_JIAO: EdgeRule;
 //# sourceMappingURL=jiao-rule.d.ts.map

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

@@ -1 +1 @@
-{"version":3,"file":"jiao-rule.d.ts","sourceRoot":"","sources":["../../src/decode/jiao-rule.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AACxD,OAAO,EAGL,KAAK,WAAW,EAChB,KAAK,QAAQ,EAId,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,wBAAgB,MAAM,CAAC,QAAQ,EAAE,QAAQ,GAAG,SAAS,GAAG,OAAO,CAI9D;AAED;;;;;;GAMG;AACH,eAAO,MAAM,WAAW,aAA8C,CAAC;AAEvE;;;;;;;GAOG;AACH,wBAAgB,YAAY,CAAC,OAAO,EAAE,WAAW,EAAE,EAAE,EAAE,MAAM,GAAG,OAAO,CAQtE;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,eAAO,MAAM,aAAa,EAAE,QAiB3B,CAAC"}
+{"version":3,"file":"jiao-rule.d.ts","sourceRoot":"","sources":["../../src/decode/jiao-rule.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AACxD,OAAO,EAGL,KAAK,WAAW,EAChB,KAAK,QAAQ,EAId,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,wBAAgB,MAAM,CAAC,QAAQ,EAAE,QAAQ,GAAG,SAAS,GAAG,OAAO,CAI9D;AAED;;;;;;GAMG;AACH,eAAO,MAAM,WAAW,aAA8C,CAAC;AAkCvE;;;;;;;;;GASG;AACH,wBAAgB,YAAY,CAAC,OAAO,EAAE,WAAW,EAAE,EAAE,EAAE,MAAM,GAAG,OAAO,CAUtE;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,eAAO,MAAM,aAAa,EAAE,QAiB3B,CAAC"}

package/dist/decode/jiao-rule.js

@@ -13,6 +13,37 @@ export function isJiao(syllable) {
  * only 教 with a place tag after it over the corpus is 教青局, an office.
  */
 export const TAUGHT_TAGS = new Set(["r", "n", "nz", "nr", "nt", "ng"]);
+/**
+ * The words that can only be governing a verb, whatever comes after them.
+ *
+ * The far side of the context, for the 教 that teaches nothing named:
+ * 他怎么教，我都学不会, 你没能力教, 这在学校是不教的, 有时也会教. A modal or a
+ * negator has to be followed by a verb, so the 教 after one is the verb even
+ * where nothing at all follows it.
+ *
+ * Written out rather than taken from a tag, for the reason the 长 rule gives:
+ * jieba calls 会 `v` along with every other verb and 怎么 `r` along with every
+ * pronoun, so neither tag names this set.
+ */
+const GOVERNING = new Set([
+    "不",
+    "能",
+    "会",
+    "會",
+    "要",
+    "想",
+    "该",
+    "該",
+    "也",
+    "别",
+    "別",
+    "没",
+    "沒",
+    "怎么",
+    "怎麼",
+    "怎样",
+    "怎樣",
+]);
 /**
  * Whether the 教 at a position is teaching something to somebody.
  *
@@ -20,12 +51,16 @@ export const TAUGHT_TAGS = new Set(["r", "n", "nz", "nr", "nt", "ng"]);
  * a 教 with a pronoun, a noun or a name after it is governing it, and a 教 with
  * 了, 过, 着 or 得 after it is a verb whatever follows that. A particle in front
  * of it rules it out, since a verb does not follow one.
+ *
+ * {@link GOVERNING} is the case where nothing follows to look at.
  */
 export function isTeachingAt(context, at) {
-    if (tagOf(context, wordEndingAt(context, at)).startsWith(PARTICLE_TAG)) {
+    const before = wordEndingAt(context, at);
+    if (tagOf(context, before).startsWith(PARTICLE_TAG)) {
         return false;
     }
-    return (ASPECT.has(context.characters[at + 1] ?? "") ||
+    return (GOVERNING.has(before ?? "") ||
+        ASPECT.has(context.characters[at + 1] ?? "") ||
         TAUGHT_TAGS.has(tagOf(context, wordStartingAt(context, at + 1))));
 }
 /**
@@ -47,13 +82,17 @@ export function isTeachingAt(context, at) {
  * one that matters: 任教, 宗教, 主教, 佛教, 传教, 执教, 请教, 家教.
  *
  * Measured over 88,866 lines of Tatoeba and zh.wikipedia, 181 教 decode as a
- * word of their own and every one of them read `jiào`. This moves 158 to
+ * word of their own and every one of them read `jiào`. This moves 162 to
  * `jiāo` and is wrong on three: 统一教创始人, 方法教深思 and 做到了教政分离,
  * where a nominal compound takes an object's shape, the last of them because
  * 做到了 is a word and so the 了 in front is not a particle to look at. Of the
- * 21 it leaves, 8 are right — 诸教中, 教外别传, 董教总, 风教, 教青局 twice, a
- * 教室 reached through a bad split, and a 教 used as a bare noun — and 13 are
+ * 17 it leaves, 8 are right — 诸教中, 教外别传, 董教总, 风教, 教青局 twice, a
+ * 教室 reached through a bad split, and a 教 used as a bare noun — and 9 are
  * misses, where jieba tags the object something this does not name.
+ *
+ * {@link GOVERNING} is four of that 162 and was added later: 能教教我,
+ * 要教一節課, 也會教像拉丁語, 這在學校是不教的. 有 is pointedly outside it,
+ * since 有教无类 is `yǒujiào wúlèi`.
  */
 export const TEACHING_JIAO = {
     name: "teaching-jiao",

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

@@ -1 +1 @@
-{"version":3,"file":"jiao-rule.js","sourceRoot":"","sources":["../../src/decode/jiao-rule.ts"],"names":[],"mappings":"AAOA,OAAO,EACL,MAAM,EACN,YAAY,EAGZ,KAAK,EACL,YAAY,EACZ,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,MAAM,UAAU,MAAM,CAAC,QAA8B;IACnD,OAAO,CACL,QAAQ,EAAE,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,KAAK,IAAI,QAAQ,CAAC,IAAI,KAAK,CAAC,CAC7E,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,CAAC,MAAM,WAAW,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC;AAEvE;;;;;;;GAOG;AACH,MAAM,UAAU,YAAY,CAAC,OAAoB,EAAE,EAAU;IAC3D,IAAI,KAAK,CAAC,OAAO,EAAE,YAAY,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,YAAY,CAAC,EAAE,CAAC;QACvE,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,CACL,MAAM,CAAC,GAAG,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;QAC5C,WAAW,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,EAAE,cAAc,CAAC,OAAO,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,CACjE,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;GA0BG;AACH,MAAM,CAAC,MAAM,aAAa,GAAa;IACrC,IAAI,EAAE,eAAe;IACrB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,IAAI,IAAI,CAAC,IAAI,KAAK,GAAG,EAAE,CAAC;YACtB,OAAO,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC;gBAChE,CAAC,CAAC,OAAO;gBACT,CAAC,CAAC,MAAM,CAAC;QACb,CAAC;QACD,OAAO,IAAI,CAAC,YAAY,KAAK,EAAE;YAC7B,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC;YACvB,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI;YAC3C,CAAC,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;YAC5B,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC;YAClC,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC"}
+{"version":3,"file":"jiao-rule.js","sourceRoot":"","sources":["../../src/decode/jiao-rule.ts"],"names":[],"mappings":"AAOA,OAAO,EACL,MAAM,EACN,YAAY,EAGZ,KAAK,EACL,YAAY,EACZ,cAAc,GACf,MAAM,YAAY,CAAC;AAEpB;;GAEG;AACH,MAAM,UAAU,MAAM,CAAC,QAA8B;IACnD,OAAO,CACL,QAAQ,EAAE,OAAO,KAAK,GAAG,IAAI,QAAQ,CAAC,KAAK,KAAK,KAAK,IAAI,QAAQ,CAAC,IAAI,KAAK,CAAC,CAC7E,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,CAAC,MAAM,WAAW,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,EAAE,GAAG,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC,CAAC;AAEvE;;;;;;;;;;;GAWG;AACH,MAAM,SAAS,GAAG,IAAI,GAAG,CAAC;IACxB,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,GAAG;IACH,IAAI;IACJ,IAAI;IACJ,IAAI;IACJ,IAAI;CACL,CAAC,CAAC;AAEH;;;;;;;;;GASG;AACH,MAAM,UAAU,YAAY,CAAC,OAAoB,EAAE,EAAU;IAC3D,MAAM,MAAM,GAAG,YAAY,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;IACzC,IAAI,KAAK,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,UAAU,CAAC,YAAY,CAAC,EAAE,CAAC;QACpD,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,CACL,SAAS,CAAC,GAAG,CAAC,MAAM,IAAI,EAAE,CAAC;QAC3B,MAAM,CAAC,GAAG,CAAC,OAAO,CAAC,UAAU,CAAC,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC;QAC5C,WAAW,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,EAAE,cAAc,CAAC,OAAO,EAAE,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,CACjE,CAAC;AACJ,CAAC;AAED;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GA8BG;AACH,MAAM,CAAC,MAAM,aAAa,GAAa;IACrC,IAAI,EAAE,eAAe;IACrB,UAAU,EAAE,CAAC,OAAoB,EAAE,EAAE;QACnC,MAAM,EAAE,IAAI,EAAE,GAAG,OAAO,CAAC;QACzB,IAAI,IAAI,CAAC,IAAI,KAAK,GAAG,EAAE,CAAC;YACtB,OAAO,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,IAAI,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,IAAI,CAAC;gBAChE,CAAC,CAAC,OAAO;gBACT,CAAC,CAAC,MAAM,CAAC;QACb,CAAC;QACD,OAAO,IAAI,CAAC,YAAY,KAAK,EAAE;YAC7B,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC;YACvB,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI;YAC3C,CAAC,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;YAC5B,YAAY,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC;YAClC,CAAC,CAAC,QAAQ;YACV,CAAC,CAAC,MAAM,CAAC;IACb,CAAC;CACF,CAAC"}

package/dist/decode/lattice-decode.d.ts

@@ -0,0 +1,28 @@
+/**
+ * The two sweeps a decode makes over a whole lattice.
+ *
+ * Separate from `decode.ts` because the run-level entry points there build a
+ * lattice from a run and cut it back down again, where these two take a lattice
+ * as given and say nothing about where it came from.
+ */
+import { type Lattice } from "./lattice.js";
+import { type ReadingProjection, type ReadingUnit } from "./locking.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[];
+//# sourceMappingURL=lattice-decode.d.ts.map

package/dist/decode/lattice-decode.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"lattice-decode.d.ts","sourceRoot":"","sources":["../../src/decode/lattice-decode.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EAAa,KAAK,OAAO,EAAE,MAAM,cAAc,CAAC;AACvD,OAAO,EAEL,KAAK,iBAAiB,EACtB,KAAK,WAAW,EAGjB,MAAM,cAAc,CAAC;AAGtB;;;;;;;GAOG;AACH,wBAAgB,cAAc,CAC5B,OAAO,EAAE,OAAO,EAChB,UAAU,EAAE,iBAAiB,GAC5B,SAAS,WAAW,EAAE,CAmBxB;AAED;;;;;;;GAOG;AACH,wBAAgB,aAAa,CAAC,OAAO,EAAE,OAAO,GAAG,SAAS,MAAM,EAAE,CAIjE"}

package/dist/decode/lattice-decode.js

@@ -0,0 +1,48 @@
+/**
+ * The two sweeps a decode makes over a whole lattice.
+ *
+ * Separate from `decode.ts` because the run-level entry points there build a
+ * lattice from a run and cut it back down again, where these two take a lattice
+ * as given and say nothing about where it came from.
+ */
+import { cutPoints } from "./lattice.js";
+import { isSettled, settledUnits, unitsOf, } from "./locking.js";
+import { readingCost, shortestPath, spacingCost } from "./viterbi.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 function decodeReadings(lattice, projection) {
+    const cuts = cutPoints(lattice);
+    const units = [];
+    for (const [index, from] of cuts.entries()) {
+        const to = cuts[index + 1];
+        if (to === undefined) {
+            break;
+        }
+        if (isSettled(projection, from, to)) {
+            units.push(...settledUnits(projection, from, to));
+            continue;
+        }
+        for (const edge of shortestPath(lattice, from, to, readingCost)) {
+            units.push(...unitsOf(edge));
+        }
+    }
+    return units;
+}
+/**
+ * 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 function decodeSpacing(lattice) {
+    return shortestPath(lattice, 0, lattice.characters.length, spacingCost).map((edge) => edge.from);
+}
+//# sourceMappingURL=lattice-decode.js.map

package/dist/decode/lattice-decode.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"lattice-decode.js","sourceRoot":"","sources":["../../src/decode/lattice-decode.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AACH,OAAO,EAAE,SAAS,EAAgB,MAAM,cAAc,CAAC;AACvD,OAAO,EACL,SAAS,EAGT,YAAY,EACZ,OAAO,GACR,MAAM,cAAc,CAAC;AACtB,OAAO,EAAE,WAAW,EAAE,YAAY,EAAE,WAAW,EAAE,MAAM,cAAc,CAAC;AAEtE;;;;;;;GAOG;AACH,MAAM,UAAU,cAAc,CAC5B,OAAgB,EAChB,UAA6B;IAE7B,MAAM,IAAI,GAAG,SAAS,CAAC,OAAO,CAAC,CAAC;IAChC,MAAM,KAAK,GAAkB,EAAE,CAAC;IAEhC,KAAK,MAAM,CAAC,KAAK,EAAE,IAAI,CAAC,IAAI,IAAI,CAAC,OAAO,EAAE,EAAE,CAAC;QAC3C,MAAM,EAAE,GAAG,IAAI,CAAC,KAAK,GAAG,CAAC,CAAC,CAAC;QAC3B,IAAI,EAAE,KAAK,SAAS,EAAE,CAAC;YACrB,MAAM;QACR,CAAC;QACD,IAAI,SAAS,CAAC,UAAU,EAAE,IAAI,EAAE,EAAE,CAAC,EAAE,CAAC;YACpC,KAAK,CAAC,IAAI,CAAC,GAAG,YAAY,CAAC,UAAU,EAAE,IAAI,EAAE,EAAE,CAAC,CAAC,CAAC;YAClD,SAAS;QACX,CAAC;QACD,KAAK,MAAM,IAAI,IAAI,YAAY,CAAC,OAAO,EAAE,IAAI,EAAE,EAAE,EAAE,WAAW,CAAC,EAAE,CAAC;YAChE,KAAK,CAAC,IAAI,CAAC,GAAG,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC;QAC/B,CAAC;IACH,CAAC;IAED,OAAO,KAAK,CAAC;AACf,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,aAAa,CAAC,OAAgB;IAC5C,OAAO,YAAY,CAAC,OAAO,EAAE,CAAC,EAAE,OAAO,CAAC,UAAU,CAAC,MAAM,EAAE,WAAW,CAAC,CAAC,GAAG,CACzE,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,IAAI,CACpB,CAAC;AACJ,CAAC"}

package/dist/decode/lattice-slice.d.ts

@@ -9,11 +9,27 @@ import { type Lattice } from "./lattice.js";
 import type { EdgeRule } from "./rules.js";
 import { type ResolvedHints } from "./hints.js";
 /**
- * A lattice to decode over, and where in it the run itself starts.
+ * The 汉字 either side of a run that the decode should see and not report.
+ *
+ * What a run would have been written with had the digits around it been spelled
+ * out. Both sides are the same kind of claim about the same text, so they travel
+ * together even where the exported signatures keep them apart.
+ */
+export interface DecodeContext {
+    /** 汉字 standing in front of the run. */
+    readonly before: string;
+    /** 汉字 standing after it. */
+    readonly after: string;
+}
+/**
+ * A lattice to decode over, and where in it the run itself lies.
  */
 export interface RunLattice {
     readonly lattice: Lattice;
+    /** Where the run starts, past whatever context stands in front of it. */
     readonly at: number;
+    /** Where the run ends, before whatever context stands after it. */
+    readonly to: number;
 }
 /**
  * Whether any reading in the lattice would hold the two sides of a join
@@ -27,11 +43,11 @@ export interface RunLattice {
  */
 export declare function isJoinedAt(lattice: Lattice, at: number): boolean;
 /**
- * The lattice a run decodes over, with whatever context stands in front of it.
+ * The lattice a run decodes over, with whatever context stands around it.
  *
- * The context is dropped where a reading would hold it to the run, leaving the
- * run decoded on its own, which is what it would have been before there was any
- * context to give.
+ * Each side is dropped where a reading would hold it to the run, and only that
+ * side: a run whose trailing context is joined still gets the leading one, which
+ * is what it would have had before there was any context to give.
  */
-export declare function runLattice(dictionary: Dictionary, run: string, rules: readonly EdgeRule[], before: string, hints: ResolvedHints | undefined): RunLattice;
+export declare function runLattice(dictionary: Dictionary, run: string, rules: readonly EdgeRule[], context: DecodeContext, hints: ResolvedHints | undefined): RunLattice;
 //# sourceMappingURL=lattice-slice.d.ts.map

package/dist/decode/lattice-slice.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"lattice-slice.d.ts","sourceRoot":"","sources":["../../src/decode/lattice-slice.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAC9D,OAAO,EAAY,KAAK,OAAO,EAAE,MAAM,cAAc,CAAC;AACtD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,YAAY,CAAC;AAE3C,OAAO,EAAc,KAAK,aAAa,EAAE,MAAM,YAAY,CAAC;AAG5D;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC;IAC1B,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;CACrB;AAED;;;;;;;;;GASG;AACH,wBAAgB,UAAU,CAAC,OAAO,EAAE,OAAO,EAAE,EAAE,EAAE,MAAM,GAAG,OAAO,CAOhE;AAED;;;;;;GAMG;AACH,wBAAgB,UAAU,CACxB,UAAU,EAAE,UAAU,EACtB,GAAG,EAAE,MAAM,EACX,KAAK,EAAE,SAAS,QAAQ,EAAE,EAC1B,MAAM,EAAE,MAAM,EACd,KAAK,EAAE,aAAa,GAAG,SAAS,GAC/B,UAAU,CAoBZ"}
+{"version":3,"file":"lattice-slice.d.ts","sourceRoot":"","sources":["../../src/decode/lattice-slice.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,6BAA6B,CAAC;AAC9D,OAAO,EAAY,KAAK,OAAO,EAAE,MAAM,cAAc,CAAC;AACtD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,YAAY,CAAC;AAE3C,OAAO,EAAc,KAAK,aAAa,EAAE,MAAM,YAAY,CAAC;AAG5D;;;;;;GAMG;AACH,MAAM,WAAW,aAAa;IAC5B,uCAAuC;IACvC,QAAQ,CAAC,MAAM,EAAE,MAAM,CAAC;IACxB,4BAA4B;IAC5B,QAAQ,CAAC,KAAK,EAAE,MAAM,CAAC;CACxB;AAED;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC;IAC1B,yEAAyE;IACzE,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,mEAAmE;IACnE,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;CACrB;AAED;;;;;;;;;GASG;AACH,wBAAgB,UAAU,CAAC,OAAO,EAAE,OAAO,EAAE,EAAE,EAAE,MAAM,GAAG,OAAO,CAOhE;AAED;;;;;;GAMG;AACH,wBAAgB,UAAU,CACxB,UAAU,EAAE,UAAU,EACtB,GAAG,EAAE,MAAM,EACX,KAAK,EAAE,SAAS,QAAQ,EAAE,EAC1B,OAAO,EAAE,aAAa,EACtB,KAAK,EAAE,aAAa,GAAG,SAAS,GAC/B,UAAU,CA2BZ"}

package/dist/decode/lattice-slice.js

@@ -18,26 +18,32 @@ export function isJoinedAt(lattice, at) {
         edge.reading.length !== edge.to - edge.from);
 }
 /**
- * The lattice a run decodes over, with whatever context stands in front of it.
+ * The lattice a run decodes over, with whatever context stands around it.
  *
- * The context is dropped where a reading would hold it to the run, leaving the
- * run decoded on its own, which is what it would have been before there was any
- * context to give.
+ * Each side is dropped where a reading would hold it to the run, and only that
+ * side: a run whose trailing context is joined still gets the leading one, which
+ * is what it would have had before there was any context to give.
  */
-export function runLattice(dictionary, run, rules, before, hints) {
-    const alone = () => ({
-        lattice: ruledLattice(dictionary, run, rules, hints),
-        at: 0,
-    });
-    if (before === "") {
-        return alone();
+export function runLattice(dictionary, run, rules, context, hints) {
+    const length = toCharacters(run).length;
+    const build = (before, after) => {
+        const at = toCharacters(before).length;
+        // The context is decoded with the run, so every hint position moves along
+        // with it. Shifting a copy keeps the caller's positions relative to the run
+        // they were given for.
+        return {
+            lattice: ruledLattice(dictionary, before + run + after, rules, hints === undefined ? undefined : shiftHints(hints, at)),
+            at,
+            to: at + length,
+        };
+    };
+    const { before, after } = context;
+    const held = build(before, after);
+    const isLeadJoined = before !== "" && isJoinedAt(held.lattice, held.at);
+    const isTailJoined = after !== "" && isJoinedAt(held.lattice, held.to);
+    if (!isLeadJoined && !isTailJoined) {
+        return held;
     }
-    const held = toCharacters(before).length;
-    // The context is decoded with the run, so every hint position moves along
-    // with it. Shifting a copy keeps the caller's positions relative to the run
-    // they were given for.
-    const lattice = ruledLattice(dictionary, before + run, rules, hints === undefined ? undefined : shiftHints(hints, held));
-    const at = held;
-    return isJoinedAt(lattice, at) ? alone() : { lattice, at };
+    return build(isLeadJoined ? "" : before, isTailJoined ? "" : after);
 }
 //# sourceMappingURL=lattice-slice.js.map

package/dist/decode/lattice-slice.js.map

@@ -1 +1 @@
-{"version":3,"file":"lattice-slice.js","sourceRoot":"","sources":["../../src/decode/lattice-slice.ts"],"names":[],"mappings":"AAOA,OAAO,EAAE,QAAQ,EAAgB,MAAM,cAAc,CAAC;AAEtD,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAE,UAAU,EAAsB,MAAM,YAAY,CAAC;AAC5D,OAAO,EAAE,YAAY,EAAE,MAAM,kBAAkB,CAAC;AAUhD;;;;;;;;;GASG;AACH,MAAM,UAAU,UAAU,CAAC,OAAgB,EAAE,EAAU;IACrD,OAAO,QAAQ,CAAC,OAAO,CAAC,CAAC,IAAI,CAC3B,CAAC,IAAI,EAAE,EAAE,CACP,IAAI,CAAC,IAAI,GAAG,EAAE;QACd,IAAI,CAAC,EAAE,GAAG,EAAE;QACZ,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,CAC9C,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,UAAU,CACxB,UAAsB,EACtB,GAAW,EACX,KAA0B,EAC1B,MAAc,EACd,KAAgC;IAEhC,MAAM,KAAK,GAAG,GAAe,EAAE,CAAC,CAAC;QAC/B,OAAO,EAAE,YAAY,CAAC,UAAU,EAAE,GAAG,EAAE,KAAK,EAAE,KAAK,CAAC;QACpD,EAAE,EAAE,CAAC;KACN,CAAC,CAAC;IACH,IAAI,MAAM,KAAK,EAAE,EAAE,CAAC;QAClB,OAAO,KAAK,EAAE,CAAC;IACjB,CAAC;IACD,MAAM,IAAI,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC;IACzC,0EAA0E;IAC1E,4EAA4E;IAC5E,uBAAuB;IACvB,MAAM,OAAO,GAAG,YAAY,CAC1B,UAAU,EACV,MAAM,GAAG,GAAG,EACZ,KAAK,EACL,KAAK,KAAK,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,UAAU,CAAC,KAAK,EAAE,IAAI,CAAC,CAC1D,CAAC;IACF,MAAM,EAAE,GAAG,IAAI,CAAC;IAChB,OAAO,UAAU,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,EAAE,EAAE,CAAC;AAC7D,CAAC"}
+{"version":3,"file":"lattice-slice.js","sourceRoot":"","sources":["../../src/decode/lattice-slice.ts"],"names":[],"mappings":"AAOA,OAAO,EAAE,QAAQ,EAAgB,MAAM,cAAc,CAAC;AAEtD,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAE,UAAU,EAAsB,MAAM,YAAY,CAAC;AAC5D,OAAO,EAAE,YAAY,EAAE,MAAM,kBAAkB,CAAC;AA2BhD;;;;;;;;;GASG;AACH,MAAM,UAAU,UAAU,CAAC,OAAgB,EAAE,EAAU;IACrD,OAAO,QAAQ,CAAC,OAAO,CAAC,CAAC,IAAI,CAC3B,CAAC,IAAI,EAAE,EAAE,CACP,IAAI,CAAC,IAAI,GAAG,EAAE;QACd,IAAI,CAAC,EAAE,GAAG,EAAE;QACZ,IAAI,CAAC,OAAO,CAAC,MAAM,KAAK,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC,IAAI,CAC9C,CAAC;AACJ,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,UAAU,CACxB,UAAsB,EACtB,GAAW,EACX,KAA0B,EAC1B,OAAsB,EACtB,KAAgC;IAEhC,MAAM,MAAM,GAAG,YAAY,CAAC,GAAG,CAAC,CAAC,MAAM,CAAC;IACxC,MAAM,KAAK,GAAG,CAAC,MAAc,EAAE,KAAa,EAAc,EAAE;QAC1D,MAAM,EAAE,GAAG,YAAY,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC;QACvC,0EAA0E;QAC1E,4EAA4E;QAC5E,uBAAuB;QACvB,OAAO;YACL,OAAO,EAAE,YAAY,CACnB,UAAU,EACV,MAAM,GAAG,GAAG,GAAG,KAAK,EACpB,KAAK,EACL,KAAK,KAAK,SAAS,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC,UAAU,CAAC,KAAK,EAAE,EAAE,CAAC,CACxD;YACD,EAAE;YACF,EAAE,EAAE,EAAE,GAAG,MAAM;SAChB,CAAC;IACJ,CAAC,CAAC;IAEF,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,GAAG,OAAO,CAAC;IAClC,MAAM,IAAI,GAAG,KAAK,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;IAClC,MAAM,YAAY,GAAG,MAAM,KAAK,EAAE,IAAI,UAAU,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,CAAC;IACxE,MAAM,YAAY,GAAG,KAAK,KAAK,EAAE,IAAI,UAAU,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,CAAC,EAAE,CAAC,CAAC;IACvE,IAAI,CAAC,YAAY,IAAI,CAAC,YAAY,EAAE,CAAC;QACnC,OAAO,IAAI,CAAC;IACd,CAAC;IACD,OAAO,KAAK,CAAC,YAAY,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,EAAE,YAAY,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC;AACtE,CAAC"}

package/dist/decode/lattice-types.d.ts

@@ -63,15 +63,34 @@ export declare const ALTERNATE_PENALTY = 1;
  * outright that those two characters together read `dì qì`. A stored reading is
  * evidence about the characters; a character's default reading is only a prior.
  *
- * So the reading decode charges per *edge* heavily enough that dictionary
- * evidence outranks frequency: at 16 an attested word always beats splitting it
- * into characters, however common those characters are, since the worst word
- * costs `15 + 16` and the best two-character split costs `0 + 16` twice. Word
- * frequency survives as the tiebreak between paths covering the same span with
- * the same number of words. That ordering — fewest words first, frequency
- * second — is the one ALGORITHM.md already gives its data-free fallback tier.
+ * So the reading decode charges per *edge*, and the charge is what a word is
+ * worth against the frequencies around it. The whole frequency scale is 15
+ * buckets wide, so a charge of 15 already makes the rarest possible word beat
+ * the two commonest possible characters. This was 16 for most of the project's
+ * life, and with `WORD_CHARGE` on top of it the real price of a word boundary
+ * was 20.62 — five buckets past the point where frequency can say anything at
+ * all. Every 那是 in the language read `nà shi` on the strength of one
+ * colloquial CC-CEDICT entry no corpus has ever counted, and 都会, 过得, 得了
+ * and 的哥 went the same way.
+ *
+ * At 6 the price is 10.62 and the arithmetic decides something. A word loses
+ * its span to a two-character split when the two characters are more than
+ * 25.62 buckets better than it, which needs a word nothing attests standing
+ * against two characters near the top of the scale. An attested word survives
+ * whatever is beside it, so 地气 is still `dì qì`.
+ *
+ * Measured over the 88,866 lines of Tatoeba and zh.wikipedia the stage-4 rules
+ * were sized against, moving 16 to 6 changes 822 of 139,682 Han runs. The
+ * changes are 那是 (303), 都 read `dū` (141), 得 read `dé` after a verb (98),
+ * 为 read `wéi` where it means *for* (36), 一个 (27), and a tail of 了, 只, 中,
+ * 说 and 的. Against them stand 21 老是, 11 等地 and 10 倒是, where the word
+ * really was one word. On CPP's 20,139 hand-labelled polyphones it is 91.48%
+ * either way, and the gold corpus does not move.
+ *
+ * Word frequency survives as the tiebreak between paths covering the same span
+ * with the same number of words.
  */
-export declare const READING_CHARGE = 16;
+export declare const READING_CHARGE = 6;
 /**
  * What an edge covering a character no dictionary entry knows costs.
  *

package/dist/decode/lattice-types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"lattice-types.d.ts","sourceRoot":"","sources":["../../src/decode/lattice-types.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD;;;;;;;;GAQG;AACH,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,CAAC;IACtC,2EAA2E;IAC3E,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,2EAA2E;IAC3E,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,YAAY,EAAE,OAAO,CAAC;IAC/B,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;IAC9B,0EAA0E;IAC1E,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,WAAW,OAAO;IACtB,QAAQ,CAAC,UAAU,EAAE,SAAS,MAAM,EAAE,CAAC;IACvC,6DAA6D;IAC7D,QAAQ,CAAC,KAAK,EAAE,SAAS,CAAC,SAAS,WAAW,EAAE,CAAC,EAAE,CAAC;CACrD;AAED;;;;;GAKG;AACH,eAAO,MAAM,YAAY,KAAK,CAAC;AAE/B;;;;;;;GAOG;AACH,eAAO,MAAM,iBAAiB,IAAI,CAAC;AAEnC;;;;;;;;;;;;;;;;;;GAkBG;AACH,eAAO,MAAM,cAAc,KAAK,CAAC;AAEjC;;;;;GAKG;AACH,eAAO,MAAM,YAAY,KAAK,CAAC"}
+{"version":3,"file":"lattice-types.d.ts","sourceRoot":"","sources":["../../src/decode/lattice-types.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD;;;;;;;;GAQG;AACH,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,EAAE,EAAE,MAAM,CAAC;IACpB,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,QAAQ,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,CAAC;IACtC,2EAA2E;IAC3E,QAAQ,CAAC,IAAI,EAAE,MAAM,CAAC;IACtB,2EAA2E;IAC3E,QAAQ,CAAC,WAAW,EAAE,MAAM,CAAC;IAC7B,QAAQ,CAAC,YAAY,EAAE,OAAO,CAAC;IAC/B,QAAQ,CAAC,YAAY,EAAE,MAAM,CAAC;IAC9B,0EAA0E;IAC1E,QAAQ,CAAC,OAAO,EAAE,OAAO,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,WAAW,OAAO;IACtB,QAAQ,CAAC,UAAU,EAAE,SAAS,MAAM,EAAE,CAAC;IACvC,6DAA6D;IAC7D,QAAQ,CAAC,KAAK,EAAE,SAAS,CAAC,SAAS,WAAW,EAAE,CAAC,EAAE,CAAC;CACrD;AAED;;;;;GAKG;AACH,eAAO,MAAM,YAAY,KAAK,CAAC;AAE/B;;;;;;;GAOG;AACH,eAAO,MAAM,iBAAiB,IAAI,CAAC;AAEnC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAqCG;AACH,eAAO,MAAM,cAAc,IAAI,CAAC;AAEhC;;;;;GAKG;AACH,eAAO,MAAM,YAAY,KAAK,CAAC"}

package/dist/decode/lattice-types.js

@@ -25,15 +25,34 @@ export const ALTERNATE_PENALTY = 1;
  * outright that those two characters together read `dì qì`. A stored reading is
  * evidence about the characters; a character's default reading is only a prior.
  *
- * So the reading decode charges per *edge* heavily enough that dictionary
- * evidence outranks frequency: at 16 an attested word always beats splitting it
- * into characters, however common those characters are, since the worst word
- * costs `15 + 16` and the best two-character split costs `0 + 16` twice. Word
- * frequency survives as the tiebreak between paths covering the same span with
- * the same number of words. That ordering — fewest words first, frequency
- * second — is the one ALGORITHM.md already gives its data-free fallback tier.
+ * So the reading decode charges per *edge*, and the charge is what a word is
+ * worth against the frequencies around it. The whole frequency scale is 15
+ * buckets wide, so a charge of 15 already makes the rarest possible word beat
+ * the two commonest possible characters. This was 16 for most of the project's
+ * life, and with `WORD_CHARGE` on top of it the real price of a word boundary
+ * was 20.62 — five buckets past the point where frequency can say anything at
+ * all. Every 那是 in the language read `nà shi` on the strength of one
+ * colloquial CC-CEDICT entry no corpus has ever counted, and 都会, 过得, 得了
+ * and 的哥 went the same way.
+ *
+ * At 6 the price is 10.62 and the arithmetic decides something. A word loses
+ * its span to a two-character split when the two characters are more than
+ * 25.62 buckets better than it, which needs a word nothing attests standing
+ * against two characters near the top of the scale. An attested word survives
+ * whatever is beside it, so 地气 is still `dì qì`.
+ *
+ * Measured over the 88,866 lines of Tatoeba and zh.wikipedia the stage-4 rules
+ * were sized against, moving 16 to 6 changes 822 of 139,682 Han runs. The
+ * changes are 那是 (303), 都 read `dū` (141), 得 read `dé` after a verb (98),
+ * 为 read `wéi` where it means *for* (36), 一个 (27), and a tail of 了, 只, 中,
+ * 说 and 的. Against them stand 21 老是, 11 等地 and 10 倒是, where the word
+ * really was one word. On CPP's 20,139 hand-labelled polyphones it is 91.48%
+ * either way, and the gold corpus does not move.
+ *
+ * Word frequency survives as the tiebreak between paths covering the same span
+ * with the same number of words.
  */
-export const READING_CHARGE = 16;
+export const READING_CHARGE = 6;
 /**
  * What an edge covering a character no dictionary entry knows costs.
  *

package/dist/decode/lattice-types.js.map

@@ -1 +1 @@
-{"version":3,"file":"lattice-types.js","sourceRoot":"","sources":["../../src/decode/lattice-types.ts"],"names":[],"mappings":"AAyCA;;;;;GAKG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,EAAE,CAAC;AAE/B;;;;;;;GAOG;AACH,MAAM,CAAC,MAAM,iBAAiB,GAAG,CAAC,CAAC;AAEnC;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,CAAC,MAAM,cAAc,GAAG,EAAE,CAAC;AAEjC;;;;;GAKG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,EAAE,CAAC"}
+{"version":3,"file":"lattice-types.js","sourceRoot":"","sources":["../../src/decode/lattice-types.ts"],"names":[],"mappings":"AAyCA;;;;;GAKG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,EAAE,CAAC;AAE/B;;;;;;;GAOG;AACH,MAAM,CAAC,MAAM,iBAAiB,GAAG,CAAC,CAAC;AAEnC;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;GAqCG;AACH,MAAM,CAAC,MAAM,cAAc,GAAG,CAAC,CAAC;AAEhC;;;;;GAKG;AACH,MAAM,CAAC,MAAM,YAAY,GAAG,EAAE,CAAC"}

package/dist/decode/numbers.d.ts

@@ -30,14 +30,6 @@ export interface RunContext {
  * supply, and it is what a 一 ending the number assimilates to.
  */
 export declare function surrounding(runs: readonly TextRun[], decoded: readonly (readonly ScoredWord[])[], at: number): RunContext;
-/**
- * The 汉字 a number in front of a Han run stands for, for that run's decode.
- *
- * Only the last segment of the run before, because only that one touches the
- * Han: the D of 3D银行 comes between them, and a decode of 银行 that saw 三
- * beside it would be reading a text nobody wrote.
- */
-export declare function numeralBefore(segments: readonly NumeralSegment[]): string;
 /**
  * Write a stretch that was never Han, reading the numbers in it.
  *

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

@@ -1 +1 @@
-{"version":3,"file":"numbers.d.ts","sourceRoot":"","sources":["../../src/decode/numbers.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD;;;;;;GAMG;AACH,OAAO,EAAE,KAAK,cAAc,EAAe,MAAM,qBAAqB,CAAC;AACvE,OAAO,EAAyB,KAAK,OAAO,EAAE,MAAM,WAAW,CAAC;AAChE,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AAC5C,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AACjD,OAAO,EACL,KAAK,cAAc,EAGnB,KAAK,OAAO,EACb,MAAM,aAAa,CAAC;AAErB;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,QAAQ,CAAC,KAAK,EAAE;QACd,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;QAC3B,QAAQ,CAAC,QAAQ,EAAE,QAAQ,GAAG,SAAS,CAAC;KACzC,CAAC;IACF,8DAA8D;IAC9D,QAAQ,CAAC,UAAU,EAAE,OAAO,CAAC;CAC9B;AAED;;;;;;GAMG;AACH,wBAAgB,WAAW,CACzB,IAAI,EAAE,SAAS,OAAO,EAAE,EACxB,OAAO,EAAE,SAAS,CAAC,SAAS,UAAU,EAAE,CAAC,EAAE,EAC3C,EAAE,EAAE,MAAM,GACT,UAAU,CAQZ;AAED;;;;;;GAMG;AACH,wBAAgB,aAAa,CAAC,QAAQ,EAAE,SAAS,cAAc,EAAE,GAAG,MAAM,CAEzE;AAED;;;;;;;;GAQG;AACH,wBAAgB,YAAY,CAC1B,IAAI,EAAE,MAAM,EACZ,QAAQ,EAAE,SAAS,cAAc,EAAE,EACnC,OAAO,EAAE,UAAU,EACnB,OAAO,EAAE,OAAO,EAChB,OAAO,EAAE;IAAE,QAAQ,CAAC,MAAM,EAAE,aAAa,GAAG,SAAS,CAAA;CAAE,GACtD,SAAS,cAAc,EAAE,CA6B3B"}
+{"version":3,"file":"numbers.d.ts","sourceRoot":"","sources":["../../src/decode/numbers.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAExD;;;;;;GAMG;AACH,OAAO,EAAE,KAAK,cAAc,EAAe,MAAM,qBAAqB,CAAC;AACvE,OAAO,EAAyB,KAAK,OAAO,EAAE,MAAM,WAAW,CAAC;AAChE,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,WAAW,CAAC;AAC5C,OAAO,KAAK,EAAE,aAAa,EAAE,MAAM,aAAa,CAAC;AACjD,OAAO,EACL,KAAK,cAAc,EAGnB,KAAK,OAAO,EACb,MAAM,aAAa,CAAC;AAErB;;GAEG;AACH,MAAM,WAAW,UAAU;IACzB,QAAQ,CAAC,KAAK,EAAE;QACd,QAAQ,CAAC,SAAS,EAAE,MAAM,CAAC;QAC3B,QAAQ,CAAC,QAAQ,EAAE,QAAQ,GAAG,SAAS,CAAC;KACzC,CAAC;IACF,8DAA8D;IAC9D,QAAQ,CAAC,UAAU,EAAE,OAAO,CAAC;CAC9B;AAED;;;;;;GAMG;AACH,wBAAgB,WAAW,CACzB,IAAI,EAAE,SAAS,OAAO,EAAE,EACxB,OAAO,EAAE,SAAS,CAAC,SAAS,UAAU,EAAE,CAAC,EAAE,EAC3C,EAAE,EAAE,MAAM,GACT,UAAU,CAQZ;AAED;;;;;;;;GAQG;AACH,wBAAgB,YAAY,CAC1B,IAAI,EAAE,MAAM,EACZ,QAAQ,EAAE,SAAS,cAAc,EAAE,EACnC,OAAO,EAAE,UAAU,EACnB,OAAO,EAAE,OAAO,EAChB,OAAO,EAAE;IAAE,QAAQ,CAAC,MAAM,EAAE,aAAa,GAAG,SAAS,CAAA;CAAE,GACtD,SAAS,cAAc,EAAE,CA6B3B"}

package/dist/decode/numbers.js

@@ -25,16 +25,6 @@ export function surrounding(runs, decoded, at) {
         isAfterHan: runs[at - 1]?.isHan === true,
     };
 }
-/**
- * The 汉字 a number in front of a Han run stands for, for that run's decode.
- *
- * Only the last segment of the run before, because only that one touches the
- * Han: the D of 3D银行 comes between them, and a decode of 银行 that saw 三
- * beside it would be reading a text nobody wrote.
- */
-export function numeralBefore(segments) {
-    return segments.at(-1)?.hanzi ?? "";
-}
 /**
  * Write a stretch that was never Han, reading the numbers in it.
  *

package/dist/decode/numbers.js.map

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