@kensio/pinyinjs 1.7.4 → 1.7.5

package/dist/dictionary/dictionary.js

@@ -1,67 +1,9 @@
-import { toCharacters } from "../script/characters.js";
 import { toCanonicalGlyphs } from "../script/glyphs.js";
-import { decodeReading } from "./artifact.js";
+import { decodeEntry } from "./dictionary-entry.js";
+import { alternateReadings, readingsCursor, storedReadingOf, } from "./dictionary-readings.js";
+import { EntryLines } from "./entry-lines.js";
 import { FrequencyTable } from "./frequency-table.js";
 import { KeyIndex } from "./key-index.js";
-/**
- * Separator between a line's columns, matching what the artifact writes.
- */
-const COLUMN = "\t";
-const LINE = "\n";
-const ALTERNATE = ",";
-const PROPER_NOUN_FLAG = "p";
-/**
- * Where a proper name divides, after the proper-noun flag: `p1`, `p2.4`. *
- * `security/detect-unsafe-regex` flags the nested quantifier, and there is no
- * ambiguity behind it: a literal `.` separates `\d+` from `(?:\.\d+)*`, so no
- * input can be split between them two ways. Timed on a string of digits and
- * dots that cannot match, it is linear — 8k to 128k characters doubles the time
- * for each doubling of the input, 39µs to 607µs.
- */
-const NAME_BOUNDARIES = /^p(?<at>\d+(?:\.\d+)*)$/u;
-/**
- * The lowest code unit that begins a surrogate pair.
- */
-const HIGH_SURROGATE = 0x1_00_00;
-/**
- * Whether a key is written with exactly one character, without allocating.
- *
- * Asked of every key in the dictionary when {@link Dictionary.readingsInOrder}
- * gathers the character defaults, which is where splitting each key into
- * characters would cost an array per key for an answer that is a length check.
- */
-function isOneCharacter(key) {
-    if (key.length === 1) {
-        return true;
-    }
-    return key.length === 2 && (key.codePointAt(0) ?? 0) >= HIGH_SURROGATE;
-}
-/**
- * A word's reading, joined from its characters' default readings.
- *
- * Walked by code point rather than split into characters first, for the same
- * reason as {@link isOneCharacter}: this runs over every key in the dictionary
- * three times during a reverse index build, and the array `toCharacters`
- * allocates per key is 20% of it.
- *
- * Returns the empty string where any character has no default, which is what a
- * tier missing a character would look like. No tier does: `selectTier` never
- * drops a character some word is written with.
- */
-function derivedReading(key, defaults) {
-    let joined = "";
-    let at = 0;
-    while (at < key.length) {
-        const width = (key.codePointAt(at) ?? 0) >= HIGH_SURROGATE ? 2 : 1;
-        const value = defaults.get(key.slice(at, at + width));
-        if (value === undefined) {
-            return "";
-        }
-        joined = joined === "" ? value : `${joined} ${value}`;
-        at += width;
-    }
-    return joined;
-}
 /**
  * A compiled dictionary, queried without being parsed.
  *
@@ -89,91 +31,17 @@ export class Dictionary {
         return new Dictionary(artifact);
     }
     #index;
-    #entries;
-    #lineStarts;
+    #lines;
     #frequencies;
     #decoded = new Map();
     /**
-     * Scan the entry blob for its line starts. Use {@link Dictionary.from}.
+     * Wrap an artifact's blobs. Use {@link Dictionary.from}.
      */
     constructor(artifact) {
         this.#index = KeyIndex.from(artifact.keys);
-        this.#entries = artifact.entries;
-        const starts = [0];
-        let separator = this.#entries.indexOf(LINE);
-        while (separator !== -1) {
-            starts.push(separator + 1);
-            separator = this.#entries.indexOf(LINE, separator + 1);
-        }
-        starts.push(this.#entries.length + 1);
-        this.#lineStarts = Uint32Array.from(starts);
+        this.#lines = EntryLines.from(artifact.entries);
         this.#frequencies = FrequencyTable.from(artifact.frequencies, this.#index.size);
     }
-    /**
-     * The columns of the line at a position.
-     */
-    #columns(at) {
-        const start = this.#lineStarts[at];
-        const nextStart = this.#lineStarts[at + 1];
-        /* c8 ignore next 3 -- every position comes from a successful key lookup */
-        if (start === undefined || nextStart === undefined) {
-            return [];
-        }
-        return this.#entries.slice(start, nextStart - 1).split(COLUMN);
-    }
-    /**
-     * The stored reading of a key, or undefined when it is not a key.
-     *
-     * Read from the blob without decoding the rest of the line, since deriving a
-     * word's reading needs its characters' readings and nothing else about them.
-     */
-    #storedReading(word) {
-        const found = this.#index.lookup(word);
-        return found.isKey ? (this.#columns(found.index)[0] ?? "") : undefined;
-    }
-    /**
-     * Decode one line into an entry.
-     */
-    #decode(word, at) {
-        const [reading = "", taiwan = "", partOfSpeech = "", flags = ""] = this.#columns(at);
-        // An empty reading means the word is read as its characters' defaults,
-        // which is how 87.8% of multi-character entries are stored.
-        const encoded = reading === ""
-            ? toCharacters(word)
-                .map((character) => this.#storedReading(character) ?? "")
-                .join(" ")
-            : reading;
-        const decoded = decodeReading(encoded) ?? [];
-        const taiwanReading = decodeReading(taiwan);
-        const stated = NAME_BOUNDARIES.exec(flags)?.groups?.["at"];
-        const nameBoundaries = stated === undefined ? [] : stated.split(".").map(Number);
-        return {
-            word,
-            reading: decoded,
-            ...(taiwanReading !== undefined && { taiwanReading }),
-            partOfSpeech,
-            isProperNoun: flags.includes(PROPER_NOUN_FLAG),
-            ...(nameBoundaries.length > 0 && { nameBoundaries }),
-            cost: this.#frequencies.costOf(at),
-        };
-    }
-    /**
-     * The reading column of a line, without slicing or splitting the rest of it.
-     *
-     * Deriving a word's reading needs its characters' readings and nothing else
-     * about them, and this is asked of every line in the dictionary during a
-     * reverse index build, so the other four columns are never touched.
-     */
-    #readingColumn(at) {
-        const start = this.#lineStarts[at];
-        const nextStart = this.#lineStarts[at + 1];
-        if (start === undefined || nextStart === undefined) {
-            return "";
-        }
-        const end = nextStart - 1;
-        const column = this.#entries.indexOf(COLUMN, start);
-        return this.#entries.slice(start, column === -1 || column > end ? end : column);
-    }
     /**
      * How many words the dictionary holds, counting each script's key.
      */
@@ -204,30 +72,9 @@ export class Dictionary {
     }
     /**
      * A cursor over every key's reading, in key order.
-     *
-     * Two passes, mirroring `readArtifact`: the character defaults are gathered
-     * first, because 83.25% of the full tier's keys store no reading and mean
-     * "the characters' defaults, in order". Holding the cursor holds those
-     * defaults, so build what needs it, use it, and let it go.
      */
     readingsInOrder() {
-        const defaults = new Map();
-        for (let at = 0; at < this.#index.size; at++) {
-            const key = this.#index.keyAt(at);
-            const reading = isOneCharacter(key) ? this.#readingColumn(at) : "";
-            if (reading !== "") {
-                defaults.set(key, reading);
-            }
-        }
-        return {
-            size: this.#index.size,
-            readingAt: (at) => {
-                const stored = this.#readingColumn(at);
-                return stored === ""
-                    ? derivedReading(this.#index.keyAt(at), defaults)
-                    : stored;
-            },
-        };
+        return readingsCursor(this.#index, this.#lines);
     }
     /**
      * Whether any word begins with this text. An exact match counts.
@@ -256,16 +103,13 @@ export class Dictionary {
         if (held !== undefined) {
             return held;
         }
-        const entry = this.#decode(canonical, found.index);
+        const entry = decodeEntry(canonical, this.#lines.columnsAt(found.index), (character) => storedReadingOf(this.#index, this.#lines, character), this.#frequencies.costOf(found.index));
         this.#decoded.set(found.index, entry);
         return entry;
     }
     /**
      * Every reading a single character is known to take, most likely first.
      *
-     * The lattice needs these for its single-character fallback edges: 行 has to
-     * offer `xíng`, `háng` and `héng` at a position no word covers.
-     *
      * The alternates are read off the canonical form's line, the same form
      * {@link Dictionary.lookup} found the entry under. Searching the index for
      * the raw character instead would land on the insertion point for a Hong Kong
@@ -278,11 +122,7 @@ export class Dictionary {
             return [];
         }
         const found = this.#index.lookup(toCanonicalGlyphs(character));
-        const alternates = (this.#columns(found.index)[4] ?? "")
-            .split(ALTERNATE)
-            .map((text) => decodeReading(text))
-            .filter((reading) => reading !== undefined);
-        return [entry.reading, ...alternates];
+        return [entry.reading, ...alternateReadings(this.#lines, found.index)];
     }
 }
 //# sourceMappingURL=dictionary.js.map

package/dist/dictionary/dictionary.js.map

@@ -1 +1 @@
-{"version":3,"file":"dictionary.js","sourceRoot":"","sources":["../../src/dictionary/dictionary.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,yBAAyB,CAAC;AACvD,OAAO,EAAE,iBAAiB,EAAE,MAAM,qBAAqB,CAAC;AAExD,OAAO,EAAE,aAAa,EAA2B,MAAM,eAAe,CAAC;AACvE,OAAO,EAAE,cAAc,EAAE,MAAM,sBAAsB,CAAC;AACtD,OAAO,EAAE,QAAQ,EAAE,MAAM,gBAAgB,CAAC;AAE1C;;GAEG;AACH,MAAM,MAAM,GAAG,IAAI,CAAC;AAEpB,MAAM,IAAI,GAAG,IAAI,CAAC;AAElB,MAAM,SAAS,GAAG,GAAG,CAAC;AAEtB,MAAM,gBAAgB,GAAG,GAAG,CAAC;AAE7B;;;;;;;GAOG;AACH,MAAM,eAAe,GAAG,0BAA0B,CAAC;AAEnD;;GAEG;AACH,MAAM,cAAc,GAAG,SAAS,CAAC;AAEjC;;;;;;GAMG;AACH,SAAS,cAAc,CAAC,GAAW;IACjC,IAAI,GAAG,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACrB,OAAO,IAAI,CAAC;IACd,CAAC;IACD,OAAO,GAAG,CAAC,MAAM,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,cAAc,CAAC;AACzE,CAAC;AAED;;;;;;;;;;;GAWG;AACH,SAAS,cAAc,CACrB,GAAW,EACX,QAAqC;IAErC,IAAI,MAAM,GAAG,EAAE,CAAC;IAChB,IAAI,EAAE,GAAG,CAAC,CAAC;IACX,OAAO,EAAE,GAAG,GAAG,CAAC,MAAM,EAAE,CAAC;QACvB,MAAM,KAAK,GAAG,CAAC,GAAG,CAAC,WAAW,CAAC,EAAE,CAAC,IAAI,CAAC,CAAC,IAAI,cAAc,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;QACnE,MAAM,KAAK,GAAG,QAAQ,CAAC,GAAG,CAAC,GAAG,CAAC,KAAK,CAAC,EAAE,EAAE,EAAE,GAAG,KAAK,CAAC,CAAC,CAAC;QACtD,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;YACxB,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,GAAG,MAAM,KAAK,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,GAAG,MAAM,IAAI,KAAK,EAAE,CAAC;QACtD,EAAE,IAAI,KAAK,CAAC;IACd,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC;AAsDD;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,OAAO,UAAU;IACrB;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,QAA4B;QACtC,OAAO,IAAI,UAAU,CAAC,QAAQ,CAAC,CAAC;IAClC,CAAC;IAEQ,MAAM,CAAW;IACjB,QAAQ,CAAS;IACjB,WAAW,CAAc;IACzB,YAAY,CAAiB;IAC7B,QAAQ,GAAG,IAAI,GAAG,EAAqB,CAAC;IAEjD;;OAEG;IACH,YAAoB,QAA4B;QAC9C,IAAI,CAAC,MAAM,GAAG,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;QAC3C,IAAI,CAAC,QAAQ,GAAG,QAAQ,CAAC,OAAO,CAAC;QAEjC,MAAM,MAAM,GAAa,CAAC,CAAC,CAAC,CAAC;QAC7B,IAAI,SAAS,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;QAC5C,OAAO,SAAS,KAAK,CAAC,CAAC,EAAE,CAAC;YACxB,MAAM,CAAC,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC;YAC3B,SAAS,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,IAAI,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC;QACzD,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QACtC,IAAI,CAAC,WAAW,GAAG,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;QAE5C,IAAI,CAAC,YAAY,GAAG,cAAc,CAAC,IAAI,CACrC,QAAQ,CAAC,WAAW,EACpB,IAAI,CAAC,MAAM,CAAC,IAAI,CACjB,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,QAAQ,CAAC,EAAU;QACjB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAC3C,2EAA2E;QAC3E,IAAI,KAAK,KAAK,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YACnD,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IACjE,CAAC;IAED;;;;;OAKG;IACH,cAAc,CAAC,IAAY;QACzB,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC;QACvC,OAAO,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC;IACzE,CAAC;IAED;;OAEG;IACH,OAAO,CAAC,IAAY,EAAE,EAAU;QAC9B,MAAM,CAAC,OAAO,GAAG,EAAE,EAAE,MAAM,GAAG,EAAE,EAAE,YAAY,GAAG,EAAE,EAAE,KAAK,GAAG,EAAE,CAAC,GAC9D,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;QAEpB,uEAAuE;QACvE,4DAA4D;QAC5D,MAAM,OAAO,GACX,OAAO,KAAK,EAAE;YACZ,CAAC,CAAC,YAAY,CAAC,IAAI,CAAC;iBACf,GAAG,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,IAAI,CAAC,cAAc,CAAC,SAAS,CAAC,IAAI,EAAE,CAAC;iBACxD,IAAI,CAAC,GAAG,CAAC;YACd,CAAC,CAAC,OAAO,CAAC;QACd,MAAM,OAAO,GAAG,aAAa,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;QAC7C,MAAM,aAAa,GAAG,aAAa,CAAC,MAAM,CAAC,CAAC;QAC5C,MAAM,MAAM,GAAG,eAAe,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,CAAC,IAAI,CAAC,CAAC;QAC3D,MAAM,cAAc,GAClB,MAAM,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,CAAC;QAE5D,OAAO;YACL,IAAI;YACJ,OAAO,EAAE,OAAO;YAChB,GAAG,CAAC,aAAa,KAAK,SAAS,IAAI,EAAE,aAAa,EAAE,CAAC;YACrD,YAAY;YACZ,YAAY,EAAE,KAAK,CAAC,QAAQ,CAAC,gBAAgB,CAAC;YAC9C,GAAG,CAAC,cAAc,CAAC,MAAM,GAAG,CAAC,IAAI,EAAE,cAAc,EAAE,CAAC;YACpD,IAAI,EAAE,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,EAAE,CAAC;SACnC,CAAC;IACJ,CAAC;IAED;;;;;;OAMG;IACH,cAAc,CAAC,EAAU;QACvB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAC3C,IAAI,KAAK,KAAK,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YACnD,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,GAAG,GAAG,SAAS,GAAG,CAAC,CAAC;QAC1B,MAAM,MAAM,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;QACpD,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,CACxB,KAAK,EACL,MAAM,KAAK,CAAC,CAAC,IAAI,MAAM,GAAG,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,CAC7C,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;IAC1B,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,EAAU;QACf,OAAO,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IAC/B,CAAC;IAED;;;;;;;OAOG;IACH,WAAW,CAAC,EAAU;QACpB,OAAO,IAAI,CAAC,YAAY,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;IACxC,CAAC;IAED;;;;;;;OAOG;IACH,eAAe;QACb,MAAM,QAAQ,GAAG,IAAI,GAAG,EAAkB,CAAC;QAC3C,KAAK,IAAI,EAAE,GAAG,CAAC,EAAE,EAAE,GAAG,IAAI,CAAC,MAAM,CAAC,IAAI,EAAE,EAAE,EAAE,EAAE,CAAC;YAC7C,MAAM,GAAG,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;YAClC,MAAM,OAAO,GAAG,cAAc,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,cAAc,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;YACnE,IAAI,OAAO,KAAK,EAAE,EAAE,CAAC;gBACnB,QAAQ,CAAC,GAAG,CAAC,GAAG,EAAE,OAAO,CAAC,CAAC;YAC7B,CAAC;QACH,CAAC;QAED,OAAO;YACL,IAAI,EAAE,IAAI,CAAC,MAAM,CAAC,IAAI;YACtB,SAAS,EAAE,CAAC,EAAU,EAAU,EAAE;gBAChC,MAAM,MAAM,GAAG,IAAI,CAAC,cAAc,CAAC,EAAE,CAAC,CAAC;gBACvC,OAAO,MAAM,KAAK,EAAE;oBAClB,CAAC,CAAC,cAAc,CAAC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,EAAE,QAAQ,CAAC;oBACjD,CAAC,CAAC,MAAM,CAAC;YACb,CAAC;SACF,CAAC;IACJ,CAAC;IAED;;;;;OAKG;IACH,SAAS,CAAC,IAAY;QACpB,OAAO,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC;IACxD,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAY;QACjB,MAAM,SAAS,GAAG,iBAAiB,CAAC,IAAI,CAAC,CAAC;QAC1C,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QAC5C,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,CAAC;YACjB,OAAO,SAAS,CAAC;QACnB,CAAC;QACD,MAAM,IAAI,GAAG,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;QAC5C,IAAI,IAAI,KAAK,SAAS,EAAE,CAAC;YACvB,OAAO,IAAI,CAAC;QACd,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC;QACnD,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;QACtC,OAAO,KAAK,CAAC;IACf,CAAC;IAED;;;;;;;;;;;OAWG;IACH,UAAU,CAAC,SAAiB;QAC1B,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QACrC,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;YACxB,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,iBAAiB,CAAC,SAAS,CAAC,CAAC,CAAC;QAC/D,MAAM,UAAU,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;aACrD,KAAK,CAAC,SAAS,CAAC;aAChB,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,aAAa,CAAC,IAAI,CAAC,CAAC;aAClC,MAAM,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,KAAK,SAAS,CAAC,CAAC;QAC9C,OAAO,CAAC,KAAK,CAAC,OAAO,EAAE,GAAG,UAAU,CAAC,CAAC;IACxC,CAAC;CACF"}
+{"version":3,"file":"dictionary.js","sourceRoot":"","sources":["../../src/dictionary/dictionary.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,iBAAiB,EAAE,MAAM,qBAAqB,CAAC;AAGxD,OAAO,EAAE,WAAW,EAAkB,MAAM,uBAAuB,CAAC;AACpE,OAAO,EACL,iBAAiB,EAEjB,cAAc,EACd,eAAe,GAChB,MAAM,0BAA0B,CAAC;AAClC,OAAO,EAAE,UAAU,EAAE,MAAM,kBAAkB,CAAC;AAC9C,OAAO,EAAE,cAAc,EAAE,MAAM,sBAAsB,CAAC;AACtD,OAAO,EAAE,QAAQ,EAAE,MAAM,gBAAgB,CAAC;AAK1C;;;;;;;;;;;;;;;;;;GAkBG;AACH,MAAM,OAAO,UAAU;IACrB;;OAEG;IACH,MAAM,CAAC,IAAI,CAAC,QAA4B;QACtC,OAAO,IAAI,UAAU,CAAC,QAAQ,CAAC,CAAC;IAClC,CAAC;IAEQ,MAAM,CAAW;IACjB,MAAM,CAAa;IACnB,YAAY,CAAiB;IAC7B,QAAQ,GAAG,IAAI,GAAG,EAAqB,CAAC;IAEjD;;OAEG;IACH,YAAoB,QAA4B;QAC9C,IAAI,CAAC,MAAM,GAAG,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC;QAC3C,IAAI,CAAC,MAAM,GAAG,UAAU,CAAC,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;QAChD,IAAI,CAAC,YAAY,GAAG,cAAc,CAAC,IAAI,CACrC,QAAQ,CAAC,WAAW,EACpB,IAAI,CAAC,MAAM,CAAC,IAAI,CACjB,CAAC;IACJ,CAAC;IAED;;OAEG;IACH,IAAI,IAAI;QACN,OAAO,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC;IAC1B,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,EAAU;QACf,OAAO,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IAC/B,CAAC;IAED;;;;;;;OAOG;IACH,WAAW,CAAC,EAAU;QACpB,OAAO,IAAI,CAAC,YAAY,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC;IACxC,CAAC;IAED;;OAEG;IACH,eAAe;QACb,OAAO,cAAc,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;IAClD,CAAC;IAED;;;;;OAKG;IACH,SAAS,CAAC,IAAY;QACpB,OAAO,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,iBAAiB,CAAC,IAAI,CAAC,CAAC,CAAC;IACxD,CAAC;IAED;;;;;;;OAOG;IACH,MAAM,CAAC,IAAY;QACjB,MAAM,SAAS,GAAG,iBAAiB,CAAC,IAAI,CAAC,CAAC;QAC1C,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QAC5C,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,CAAC;YACjB,OAAO,SAAS,CAAC;QACnB,CAAC;QACD,MAAM,IAAI,GAAG,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;QAC5C,IAAI,IAAI,KAAK,SAAS,EAAE,CAAC;YACvB,OAAO,IAAI,CAAC;QACd,CAAC;QACD,MAAM,KAAK,GAAG,WAAW,CACvB,SAAS,EACT,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,KAAK,CAAC,KAAK,CAAC,EAClC,CAAC,SAAS,EAAE,EAAE,CAAC,eAAe,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,MAAM,EAAE,SAAS,CAAC,EACnE,IAAI,CAAC,YAAY,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,CACtC,CAAC;QACF,IAAI,CAAC,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC;QACtC,OAAO,KAAK,CAAC;IACf,CAAC;IAED;;;;;;;;OAQG;IACH,UAAU,CAAC,SAAiB;QAC1B,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,SAAS,CAAC,CAAC;QACrC,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;YACxB,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,KAAK,GAAG,IAAI,CAAC,MAAM,CAAC,MAAM,CAAC,iBAAiB,CAAC,SAAS,CAAC,CAAC,CAAC;QAC/D,OAAO,CAAC,KAAK,CAAC,OAAO,EAAE,GAAG,iBAAiB,CAAC,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC;IACzE,CAAC;CACF"}

package/dist/dictionary/entry-invariants.d.ts

@@ -0,0 +1,3 @@
+import type { BuildAssertion } from "./built-dictionary.js";
+export declare const ENTRY_INVARIANTS: readonly BuildAssertion[];
+//# sourceMappingURL=entry-invariants.d.ts.map

package/dist/dictionary/entry-invariants.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"entry-invariants.d.ts","sourceRoot":"","sources":["../../src/dictionary/entry-invariants.ts"],"names":[],"mappings":"AAOA,OAAO,KAAK,EAAE,cAAc,EAAmB,MAAM,uBAAuB,CAAC;AAE7E,eAAO,MAAM,gBAAgB,EAAE,SAAS,cAAc,EAmErD,CAAC"}

package/dist/dictionary/entry-invariants.js

@@ -0,0 +1,67 @@
+/**
+ * What must hold of an entry's own fields.
+ *
+ * The proper noun bit and the 繁體 form are both derived rather than taken from
+ * one source, so each is asserted where the derivation is known to be delicate.
+ */
+import { writeSyllable } from "../syllable/syllable.js";
+export const ENTRY_INVARIANTS = [
+    {
+        description: "北京 is a proper noun (jieba POS carried through)",
+        check: (dictionary) => dictionary.get("北京")?.isProperNoun === true
+            ? undefined
+            : "北京 is not marked as a proper noun",
+    },
+    {
+        // jieba tags all four of these a proper noun; CC-CEDICT writes their
+        // pinyin in lower case, and the decoder capitalises straight off this bit.
+        description: "沙发, 城市, 阿姨 and 长大 are not proper nouns",
+        check: (dictionary) => {
+            const wrong = ["沙发", "城市", "阿姨", "长大"].filter((word) => dictionary.get(word)?.isProperNoun === true);
+            return wrong.length === 0
+                ? undefined
+                : `marked as proper nouns: ${wrong.join(", ")}`;
+        },
+    },
+    {
+        // The veto only demotes. These stay proper nouns, and 人民政府 is the
+        // control that shows the veto still fires on an institution.
+        description: "齐白石, 国务院 and 湖北 stay proper nouns",
+        check: (dictionary) => {
+            const wrong = ["齐白石", "国务院", "湖北"].filter((word) => dictionary.get(word)?.isProperNoun !== true);
+            return wrong.length === 0
+                ? undefined
+                : `no longer proper nouns: ${wrong.join(", ")}`;
+        },
+    },
+    {
+        // The gap composeLocaleDeltas closes, and the two ways it must not close
+        // it. 運行狀況 contains the marked word 行狀 and cuts elsewhere; 一个个地
+        // ends in a character whose delta is a different sense of 地 entirely.
+        description: "垃圾分類 inherits 垃圾's 國語 reading, 運行狀況 inherits nothing",
+        check: (dictionary) => {
+            const taiwan = (word) => dictionary
+                .get(word)
+                ?.readings.tw?.map((syllable) => writeSyllable(syllable))
+                .join(" ");
+            const composed = taiwan("垃圾分类");
+            if (composed !== "lè sè fēn lèi") {
+                return `垃圾分类 reads ${composed ?? "nothing"} in zh-TW, expected lè sè fēn lèi`;
+            }
+            const wrong = ["运行状况", "一个个地", "相亲相爱"].filter((word) => taiwan(word) !== undefined);
+            return wrong.length === 0
+                ? undefined
+                : `composed a zh-TW reading for: ${wrong.join(", ")}`;
+        },
+    },
+    {
+        description: "头发 derives 頭髮 rather than 頭發",
+        check: (dictionary) => {
+            const hant = dictionary.get("头发")?.hant;
+            return hant === "頭髮"
+                ? undefined
+                : `头发 derived ${hant ?? "nothing"}, expected 頭髮`;
+        },
+    },
+];
+//# sourceMappingURL=entry-invariants.js.map

package/dist/dictionary/entry-invariants.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"entry-invariants.js","sourceRoot":"","sources":["../../src/dictionary/entry-invariants.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AACH,OAAO,EAAE,aAAa,EAAE,MAAM,yBAAyB,CAAC;AAGxD,MAAM,CAAC,MAAM,gBAAgB,GAA8B;IACzD;QACE,WAAW,EAAE,iDAAiD;QAC9D,KAAK,EAAE,CAAC,UAA2B,EAAsB,EAAE,CACzD,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,YAAY,KAAK,IAAI;YACzC,CAAC,CAAC,SAAS;YACX,CAAC,CAAC,mCAAmC;KAC1C;IACD;QACE,qEAAqE;QACrE,2EAA2E;QAC3E,WAAW,EAAE,wCAAwC;QACrD,KAAK,EAAE,CAAC,UAA2B,EAAsB,EAAE;YACzD,MAAM,KAAK,GAAG,CAAC,IAAI,EAAE,IAAI,EAAE,IAAI,EAAE,IAAI,CAAC,CAAC,MAAM,CAC3C,CAAC,IAAI,EAAE,EAAE,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,YAAY,KAAK,IAAI,CACtD,CAAC;YACF,OAAO,KAAK,CAAC,MAAM,KAAK,CAAC;gBACvB,CAAC,CAAC,SAAS;gBACX,CAAC,CAAC,2BAA2B,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;QACpD,CAAC;KACF;IACD;QACE,kEAAkE;QAClE,6DAA6D;QAC7D,WAAW,EAAE,mCAAmC;QAChD,KAAK,EAAE,CAAC,UAA2B,EAAsB,EAAE;YACzD,MAAM,KAAK,GAAG,CAAC,KAAK,EAAE,KAAK,EAAE,IAAI,CAAC,CAAC,MAAM,CACvC,CAAC,IAAI,EAAE,EAAE,CAAC,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,YAAY,KAAK,IAAI,CACtD,CAAC;YACF,OAAO,KAAK,CAAC,MAAM,KAAK,CAAC;gBACvB,CAAC,CAAC,SAAS;gBACX,CAAC,CAAC,2BAA2B,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;QACpD,CAAC;KACF;IACD;QACE,yEAAyE;QACzE,gEAAgE;QAChE,sEAAsE;QACtE,WAAW,EACT,sDAAsD;QACxD,KAAK,EAAE,CAAC,UAA2B,EAAsB,EAAE;YACzD,MAAM,MAAM,GAAG,CAAC,IAAY,EAAsB,EAAE,CAClD,UAAU;iBACP,GAAG,CAAC,IAAI,CAAC;gBACV,EAAE,QAAQ,CAAC,EAAE,EAAE,GAAG,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,aAAa,CAAC,QAAQ,CAAC,CAAC;iBACxD,IAAI,CAAC,GAAG,CAAC,CAAC;YACf,MAAM,QAAQ,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC;YAChC,IAAI,QAAQ,KAAK,eAAe,EAAE,CAAC;gBACjC,OAAO,cAAc,QAAQ,IAAI,SAAS,mCAAmC,CAAC;YAChF,CAAC;YACD,MAAM,KAAK,GAAG,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,CAAC,CAAC,MAAM,CAC3C,CAAC,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,SAAS,CACrC,CAAC;YACF,OAAO,KAAK,CAAC,MAAM,KAAK,CAAC;gBACvB,CAAC,CAAC,SAAS;gBACX,CAAC,CAAC,iCAAiC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;QAC1D,CAAC;KACF;IACD;QACE,WAAW,EAAE,8BAA8B;QAC3C,KAAK,EAAE,CAAC,UAA2B,EAAsB,EAAE;YACzD,MAAM,IAAI,GAAG,UAAU,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,IAAI,CAAC;YACxC,OAAO,IAAI,KAAK,IAAI;gBAClB,CAAC,CAAC,SAAS;gBACX,CAAC,CAAC,cAAc,IAAI,IAAI,SAAS,eAAe,CAAC;QACrD,CAAC;KACF;CACF,CAAC"}

package/dist/dictionary/entry-lines.d.ts

@@ -0,0 +1,35 @@
+/**
+ * The entry blob read as lines and columns, and nothing above that.
+ *
+ * `Dictionary` never parses its artifact; it scans for separators and slices.
+ * That scanning is a concern of its own — it is the only code that knows the
+ * blob is tab-separated columns one line per key — so it sits here, and the
+ * dictionary above it deals in positions and words.
+ */
+/**
+ * A blob of entry lines, indexed by position.
+ */
+export declare class EntryLines {
+    #private;
+    /**
+     * Scan a blob for its line starts.
+     *
+     * The scan is the whole of the load cost: an `indexOf` per line, and no
+     * allocation beyond the offsets themselves. Everything after it is a slice.
+     */
+    static from(entries: string): EntryLines;
+    private constructor();
+    /**
+     * The columns of the line at a position.
+     */
+    columnsAt(at: number): readonly string[];
+    /**
+     * The reading column of a line, without slicing or splitting the rest of it.
+     *
+     * Deriving a word's reading needs its characters' readings and nothing else
+     * about them, and this is asked of every line in the dictionary during a
+     * reverse index build, so the other four columns are never touched.
+     */
+    readingColumnAt(at: number): string;
+}
+//# sourceMappingURL=entry-lines.d.ts.map

package/dist/dictionary/entry-lines.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"entry-lines.d.ts","sourceRoot":"","sources":["../../src/dictionary/entry-lines.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AASH;;GAEG;AACH,qBAAa,UAAU;;IACrB;;;;;OAKG;IACH,MAAM,CAAC,IAAI,CAAC,OAAO,EAAE,MAAM,GAAG,UAAU,CASvC;IAKD,OAAO,eAGN;IAED;;OAEG;IACH,SAAS,CAAC,EAAE,EAAE,MAAM,GAAG,SAAS,MAAM,EAAE,CAQvC;IAED;;;;;;OAMG;IACH,eAAe,CAAC,EAAE,EAAE,MAAM,GAAG,MAAM,CAYlC;CACF"}

package/dist/dictionary/entry-lines.js

@@ -0,0 +1,70 @@
+/**
+ * The entry blob read as lines and columns, and nothing above that.
+ *
+ * `Dictionary` never parses its artifact; it scans for separators and slices.
+ * That scanning is a concern of its own — it is the only code that knows the
+ * blob is tab-separated columns one line per key — so it sits here, and the
+ * dictionary above it deals in positions and words.
+ */
+/**
+ * Separator between a line's columns, matching what the artifact writes.
+ */
+const COLUMN = "\t";
+const LINE = "\n";
+/**
+ * A blob of entry lines, indexed by position.
+ */
+export class EntryLines {
+    /**
+     * Scan a blob for its line starts.
+     *
+     * The scan is the whole of the load cost: an `indexOf` per line, and no
+     * allocation beyond the offsets themselves. Everything after it is a slice.
+     */
+    static from(entries) {
+        const starts = [0];
+        let separator = entries.indexOf(LINE);
+        while (separator !== -1) {
+            starts.push(separator + 1);
+            separator = entries.indexOf(LINE, separator + 1);
+        }
+        starts.push(entries.length + 1);
+        return new EntryLines(entries, Uint32Array.from(starts));
+    }
+    #entries;
+    #lineStarts;
+    constructor(entries, lineStarts) {
+        this.#entries = entries;
+        this.#lineStarts = lineStarts;
+    }
+    /**
+     * The columns of the line at a position.
+     */
+    columnsAt(at) {
+        const start = this.#lineStarts[at];
+        const nextStart = this.#lineStarts[at + 1];
+        /* c8 ignore next 3 -- every position comes from a successful key lookup */
+        if (start === undefined || nextStart === undefined) {
+            return [];
+        }
+        return this.#entries.slice(start, nextStart - 1).split(COLUMN);
+    }
+    /**
+     * The reading column of a line, without slicing or splitting the rest of it.
+     *
+     * Deriving a word's reading needs its characters' readings and nothing else
+     * about them, and this is asked of every line in the dictionary during a
+     * reverse index build, so the other four columns are never touched.
+     */
+    readingColumnAt(at) {
+        const start = this.#lineStarts[at];
+        const nextStart = this.#lineStarts[at + 1];
+        if (start === undefined || nextStart === undefined) {
+            return "";
+        }
+        const end = nextStart - 1;
+        const column = this.#entries.indexOf(COLUMN, start);
+        return this.#entries.slice(start, column === -1 || column > end ? end : column);
+    }
+}
+//# sourceMappingURL=entry-lines.js.map

package/dist/dictionary/entry-lines.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"entry-lines.js","sourceRoot":"","sources":["../../src/dictionary/entry-lines.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH;;GAEG;AACH,MAAM,MAAM,GAAG,IAAI,CAAC;AAEpB,MAAM,IAAI,GAAG,IAAI,CAAC;AAElB;;GAEG;AACH,MAAM,OAAO,UAAU;IACrB;;;;;OAKG;IACH,MAAM,CAAC,IAAI,CAAC,OAAe;QACzB,MAAM,MAAM,GAAa,CAAC,CAAC,CAAC,CAAC;QAC7B,IAAI,SAAS,GAAG,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC;QACtC,OAAO,SAAS,KAAK,CAAC,CAAC,EAAE,CAAC;YACxB,MAAM,CAAC,IAAI,CAAC,SAAS,GAAG,CAAC,CAAC,CAAC;YAC3B,SAAS,GAAG,OAAO,CAAC,OAAO,CAAC,IAAI,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC;QACnD,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QAChC,OAAO,IAAI,UAAU,CAAC,OAAO,EAAE,WAAW,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC;IAC3D,CAAC;IAEQ,QAAQ,CAAS;IACjB,WAAW,CAAc;IAElC,YAAoB,OAAe,EAAE,UAAuB;QAC1D,IAAI,CAAC,QAAQ,GAAG,OAAO,CAAC;QACxB,IAAI,CAAC,WAAW,GAAG,UAAU,CAAC;IAChC,CAAC;IAED;;OAEG;IACH,SAAS,CAAC,EAAU;QAClB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAC3C,2EAA2E;QAC3E,IAAI,KAAK,KAAK,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YACnD,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,KAAK,EAAE,SAAS,GAAG,CAAC,CAAC,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;IACjE,CAAC;IAED;;;;;;OAMG;IACH,eAAe,CAAC,EAAU;QACxB,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,CAAC,CAAC;QACnC,MAAM,SAAS,GAAG,IAAI,CAAC,WAAW,CAAC,EAAE,GAAG,CAAC,CAAC,CAAC;QAC3C,IAAI,KAAK,KAAK,SAAS,IAAI,SAAS,KAAK,SAAS,EAAE,CAAC;YACnD,OAAO,EAAE,CAAC;QACZ,CAAC;QACD,MAAM,GAAG,GAAG,SAAS,GAAG,CAAC,CAAC;QAC1B,MAAM,MAAM,GAAG,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,MAAM,EAAE,KAAK,CAAC,CAAC;QACpD,OAAO,IAAI,CAAC,QAAQ,CAAC,KAAK,CACxB,KAAK,EACL,MAAM,KAAK,CAAC,CAAC,IAAI,MAAM,GAAG,GAAG,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,MAAM,CAC7C,CAAC;IACJ,CAAC;CACF"}

package/dist/dictionary/erhua-repair.d.ts

@@ -0,0 +1,26 @@
+/**
+ * Folding a word's 儿 into the syllable before it.
+ *
+ * Two different repairs, because the sources mark 儿化 in two different
+ * places. A trailing 儿 is settled from the exception list and CC-CEDICT's
+ * `r5` token; one in the *middle* of a word can only come from CC-CEDICT,
+ * since the phrase corpus writes a syllable per character and has no way to
+ * say that two characters are read as one.
+ */
+import type { CedictEntry } from "../sources/cedict.js";
+import type { Syllable } from "../syllable/syllable.js";
+import { type ReadCharacters } from "./reading.js";
+/**
+ * A reading and its alignment, after any repair.
+ */
+export interface ErhuaRepair {
+    readonly reading: readonly Syllable[];
+    readonly aligned: readonly ReadCharacters[] | undefined;
+    /** Whether anything was folded. */
+    readonly isRepaired: boolean;
+}
+/**
+ * Repair one word's reading, or hand back what it was given.
+ */
+export declare function repairErhua(word: string, reading: readonly Syllable[], aligned: readonly ReadCharacters[] | undefined, cedictEntries: readonly CedictEntry[], cedictReadings: readonly (readonly Syllable[])[]): ErhuaRepair;
+//# sourceMappingURL=erhua-repair.d.ts.map

package/dist/dictionary/erhua-repair.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"erhua-repair.d.ts","sourceRoot":"","sources":["../../src/dictionary/erhua-repair.ts"],"names":[],"mappings":"AAAA;;;;;;;;GAQG;AACH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AACxD,OAAO,KAAK,EAAE,QAAQ,EAAE,MAAM,yBAAyB,CAAC;AAGxD,OAAO,EAGL,KAAK,cAAc,EACpB,MAAM,cAAc,CAAC;AAEtB;;GAEG;AACH,MAAM,WAAW,WAAW;IAC1B,QAAQ,CAAC,OAAO,EAAE,SAAS,QAAQ,EAAE,CAAC;IACtC,QAAQ,CAAC,OAAO,EAAE,SAAS,cAAc,EAAE,GAAG,SAAS,CAAC;IACxD,mCAAmC;IACnC,QAAQ,CAAC,UAAU,EAAE,OAAO,CAAC;CAC9B;AAED;;GAEG;AACH,wBAAgB,WAAW,CACzB,IAAI,EAAE,MAAM,EACZ,OAAO,EAAE,SAAS,QAAQ,EAAE,EAC5B,OAAO,EAAE,SAAS,cAAc,EAAE,GAAG,SAAS,EAC9C,aAAa,EAAE,SAAS,WAAW,EAAE,EACrC,cAAc,EAAE,SAAS,CAAC,SAAS,QAAQ,EAAE,CAAC,EAAE,GAC/C,WAAW,CA6Cb"}

package/dist/dictionary/erhua-repair.js

@@ -0,0 +1,49 @@
+import { attachErhua } from "./erhua.js";
+import { isErhua } from "./reading-agreement.js";
+import { ERHUA_TOKEN, readAlignedReading, } from "./reading.js";
+/**
+ * Repair one word's reading, or hand back what it was given.
+ */
+export function repairErhua(word, reading, aligned, cedictEntries, cedictReadings) {
+    if (isErhua(word, cedictReadings)) {
+        const attached = attachErhua(reading);
+        if (attached !== reading) {
+            // Fold the alignment the same way, so the 繁體 derivation still knows
+            // which character each syllable reads.
+            const last = aligned?.at(-1);
+            const previous = aligned?.at(-2);
+            const folded = aligned !== undefined && last !== undefined && previous !== undefined
+                ? [
+                    ...aligned.slice(0, -2),
+                    {
+                        characters: previous.characters + last.characters,
+                        syllable: attached.at(-1),
+                    },
+                ]
+                : aligned;
+            return { reading: attached, aligned: folded, isRepaired: true };
+        }
+    }
+    else if (reading.every((syllable) => syllable.erhua !== true)) {
+        // 儿化 in the middle of a word, which the trailing repair above cannot
+        // reach: 一点儿事 is `yìdiǎnr shì`, three syllables over four characters,
+        // and the phrase corpus writes four. Only CC-CEDICT's `r5` marks it, and
+        // only it knows where, so where it does the whole reading is taken from
+        // it — 53 of its 683 `r5` entries carry the marker mid-word.
+        const marked = cedictEntries.find((entry) => entry.readings.includes(ERHUA_TOKEN));
+        const markedAligned = marked === undefined
+            ? undefined
+            : readAlignedReading(word, marked.readings);
+        if (markedAligned !== undefined) {
+            return {
+                reading: markedAligned
+                    .map((read) => read.syllable)
+                    .filter((syllable) => syllable !== undefined),
+                aligned: markedAligned,
+                isRepaired: true,
+            };
+        }
+    }
+    return { reading, aligned, isRepaired: false };
+}
+//# sourceMappingURL=erhua-repair.js.map

package/dist/dictionary/erhua-repair.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"erhua-repair.js","sourceRoot":"","sources":["../../src/dictionary/erhua-repair.ts"],"names":[],"mappings":"AAWA,OAAO,EAAE,WAAW,EAAE,MAAM,YAAY,CAAC;AACzC,OAAO,EAAE,OAAO,EAAE,MAAM,wBAAwB,CAAC;AACjD,OAAO,EACL,WAAW,EACX,kBAAkB,GAEnB,MAAM,cAAc,CAAC;AAYtB;;GAEG;AACH,MAAM,UAAU,WAAW,CACzB,IAAY,EACZ,OAA4B,EAC5B,OAA8C,EAC9C,aAAqC,EACrC,cAAgD;IAEhD,IAAI,OAAO,CAAC,IAAI,EAAE,cAAc,CAAC,EAAE,CAAC;QAClC,MAAM,QAAQ,GAAG,WAAW,CAAC,OAAO,CAAC,CAAC;QACtC,IAAI,QAAQ,KAAK,OAAO,EAAE,CAAC;YACzB,oEAAoE;YACpE,uCAAuC;YACvC,MAAM,IAAI,GAAG,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;YAC7B,MAAM,QAAQ,GAAG,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;YACjC,MAAM,MAAM,GACV,OAAO,KAAK,SAAS,IAAI,IAAI,KAAK,SAAS,IAAI,QAAQ,KAAK,SAAS;gBACnE,CAAC,CAAC;oBACE,GAAG,OAAO,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC;oBACvB;wBACE,UAAU,EAAE,QAAQ,CAAC,UAAU,GAAG,IAAI,CAAC,UAAU;wBACjD,QAAQ,EAAE,QAAQ,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;qBAC1B;iBACF;gBACH,CAAC,CAAC,OAAO,CAAC;YACd,OAAO,EAAE,OAAO,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,EAAE,UAAU,EAAE,IAAI,EAAE,CAAC;QAClE,CAAC;IACH,CAAC;SAAM,IAAI,OAAO,CAAC,KAAK,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,KAAK,KAAK,IAAI,CAAC,EAAE,CAAC;QAChE,qEAAqE;QACrE,sEAAsE;QACtE,yEAAyE;QACzE,wEAAwE;QACxE,6DAA6D;QAC7D,MAAM,MAAM,GAAG,aAAa,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAC1C,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC,WAAW,CAAC,CACrC,CAAC;QACF,MAAM,aAAa,GACjB,MAAM,KAAK,SAAS;YAClB,CAAC,CAAC,SAAS;YACX,CAAC,CAAC,kBAAkB,CAAC,IAAI,EAAE,MAAM,CAAC,QAAQ,CAAC,CAAC;QAChD,IAAI,aAAa,KAAK,SAAS,EAAE,CAAC;YAChC,OAAO;gBACL,OAAO,EAAE,aAAa;qBACnB,GAAG,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,QAAQ,CAAC;qBAC5B,MAAM,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,KAAK,SAAS,CAAC;gBAC/C,OAAO,EAAE,aAAa;gBACtB,UAAU,EAAE,IAAI;aACjB,CAAC;QACJ,CAAC;IACH,CAAC;IAED,OAAO,EAAE,OAAO,EAAE,OAAO,EAAE,UAAU,EAAE,KAAK,EAAE,CAAC;AACjD,CAAC"}

package/dist/dictionary/frequency-tones.d.ts

@@ -0,0 +1,74 @@
+/**
+ * Repairing the tones Unihan's `kHanyuPinlu` frequency field misreports.
+ *
+ * The field counts occurrences in running text, and that gives it two blind
+ * spots about the neutral tone: it drops tone marks it should carry, and it
+ * counts a syllable reduced inside a word as though the bare character read
+ * that way. Both are corrected here, before a character's default is settled.
+ */
+import type { CedictEntry } from "../sources/cedict.js";
+import { type UnihanReadings } from "../sources/unihan.js";
+/**
+ * Put back a tone that `kHanyuPinlu` left off.
+ *
+ * The frequency field writes some readings without their tone mark: 李 is
+ * `li(36)` there where every dictionary field writes `lǐ`. Read as 轻声 — which
+ * is what an unmarked reading means in source data — that makes 李华 come out
+ * `Li Huá`, and it is the only Unihan field this happens in.
+ *
+ * The other fields settle which it is. If they also write the spelling bare,
+ * the neutral tone is real and 么, 们, 吧, 裳 keep it. If every one of them
+ * writes it with a tone, the frequency field simply omitted the mark, and the
+ * tone is taken back. Measured over the real file, 3,799 characters have a
+ * `kHanyuPinlu` entry, 60 of those have only unmarked readings, and the split
+ * is 45 that should carry a tone against 15 genuine 轻声.
+ */
+export declare function resolveFrequencyTones(unihan: UnihanReadings): readonly string[];
+/**
+ * Whether CC-CEDICT reads the bare character 轻声 in its own right.
+ *
+ * The exception to {@link demoteReducedNeutrals}, and the thing that separates
+ * a 语气词 from a syllable that is only ever unstressed inside a word. A
+ * single-character CC-CEDICT headword with a neutral reading is a claim about
+ * the character standing alone, which is exactly what the frequency field's
+ * counting cannot distinguish: 吗 is `(question particle for "yes-no"
+ * questions)`, 啊 is `modal particle ending sentence`, 得 is `structural
+ * particle: used after a verb`. Nine characters carry one — 得, 啊, 吗, 呀, 啦,
+ * 哇, 嘛, 罢 and 子 — and their 繁體 spellings with them.
+ *
+ * **A suffix sense is not such a claim.** 子's neutral headword is `noun
+ * suffix, as in 椅子`, which says the character is unstressed *in words*, and
+ * words are what the corpus can already see. 子 is `zǐ` on its own — 子曰, 母子
+ * — and CPP's human labelling has it `zi3` in 30 of its 38 cases.
+ *
+ * 罢 is the known loss. CC-CEDICT calls its `ba5` a final particle, so it is
+ * held here, and CPP labels the character `ba4` in all 40 of its cases: the
+ * particle is the literary 也罢 and 罢了, while a modern bare 罢 is 罢工. Left
+ * protected because the rule is about what a source claims of the character,
+ * and singling one out because a dataset disagrees is fitting to the dataset.
+ */
+export declare function isNeutralAlone(character: string, entries: readonly CedictEntry[]): boolean;
+/**
+ * Demote a 轻声 the frequency field counted only inside words.
+ *
+ * `kHanyuPinlu` counts occurrences in running text, and a character's neutral
+ * count there mixes two unrelated things: a character that stands alone
+ * unstressed, and one reduced *inside a word*. Where the field writes the same
+ * spelling both bare and toned it is reporting that split with no way to say
+ * which is which — 西 is `xi(902) xī(738)`, and the 902 are 东西 — and since the
+ * field leads {@link import("../sources/unihan.js").READING_FIELDS}, the
+ * reduction became the character's reading. 西 read `xi`, so 往西 was `wǎng xi`;
+ * 夫 read `fu`, 识 `shi`, 拾 `shi`, 候 `hou`, 姐 `jie`, 友 `you`.
+ *
+ * So a bare spelling the field also writes with a tone ranks below that twin —
+ * demoted rather than dropped, keeping it as a candidate without letting it set
+ * the default, exactly as a lone `kHanyuPinlu` reading is. It reaches 50
+ * characters, and {@link isNeutralAlone} holds back the ones a source says are
+ * neutral on their own.
+ *
+ * The other half of the same blind spot is already handled: where the field
+ * marks *no* tone at all it omitted the mark rather than meaning 轻声, which is
+ * {@link resolveFrequencyTones}.
+ */
+export declare function demoteReducedNeutrals(unihan: UnihanReadings, readings: readonly string[]): readonly string[];
+//# sourceMappingURL=frequency-tones.d.ts.map

package/dist/dictionary/frequency-tones.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"frequency-tones.d.ts","sourceRoot":"","sources":["../../src/dictionary/frequency-tones.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AACH,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AACxD,OAAO,EAAmB,KAAK,cAAc,EAAE,MAAM,sBAAsB,CAAC;AAqB5E;;;;;;;;;;;;;;GAcG;AACH,wBAAgB,qBAAqB,CACnC,MAAM,EAAE,cAAc,GACrB,SAAS,MAAM,EAAE,CA2BnB;AAUD;;;;;;;;;;;;;;;;;;;;;;GAsBG;AACH,wBAAgB,cAAc,CAC5B,SAAS,EAAE,MAAM,EACjB,OAAO,EAAE,SAAS,WAAW,EAAE,GAC9B,OAAO,CAST;AAED;;;;;;;;;;;;;;;;;;;;;GAqBG;AACH,wBAAgB,qBAAqB,CACnC,MAAM,EAAE,cAAc,EACtB,QAAQ,EAAE,SAAS,MAAM,EAAE,GAC1B,SAAS,MAAM,EAAE,CAkBnB"}