chaffjs 0.1.0 → 0.3.0

package/dist/detectors/signals.js

@@ -0,0 +1,183 @@
+import { proseText } from "../measure.js";
+import { wordsOf } from "./structure.js";
+const PER = 1000;
+/**
+ * これより短い文書では密度が暴れる。1 個で「1000 あたり 100」になる。
+ * 単位は言語で違うので、床も分ける。英語の 200 語と日本語の 200 文字では長さが桁で違う。
+ */
+const FLOOR = { word: 200, char: 500 };
+const bodyOf = (doc) => doc.sentences.map(proseText).join(" ");
+const findIn = (doc, pattern) => doc.sentences.flatMap((sentence) => [...sentence.text.matchAll(pattern)].map((match) => ({ sentence, offset: sentence.span.start + (match.index ?? 0) })));
+const density = (doc, count) => {
+    const length = wordsOf(doc);
+    return length === 0 ? 0 : Math.round((count / length) * PER);
+};
+/**
+ * 絵文字と装飾記号。1 つなら親しみだが、並ぶと文字そのものが読めなくなる。
+ * 件数ではなく密度で見る。長い文書ほど当たるのを避けるため（bold-density と同じ）。
+ */
+const EMOJI = /\p{Extended_Pictographic}|[←-⇿☀-➿]/gu;
+export const emojiDensity = (doc, options) => {
+    const hits = findIn(doc, EMOJI);
+    const first = hits[0];
+    const rate = density(doc, hits.length);
+    if (wordsOf(doc) < FLOOR[doc.lengthUnit] || first === undefined || rate <= options.limit)
+        return [];
+    return [
+        {
+            rule: "emoji-density",
+            severity: "info",
+            line: 0,
+            column: 0,
+            quote: first.sentence.text.trim(),
+            values: { count: hits.length, density: rate, limit: options.limit, offset: first.offset },
+        },
+    ];
+};
+/**
+ * 同じ言い回しの繰り返し。character n-gram で見るので、語の区切りが要らない。
+ * spec §10 が n-gram を選んだのはこのためで、新しい言語でそのまま動く。
+ */
+const GRAM = 8;
+/** 短い n-gram は偶然一致する。8 文字は「〜することができます」のような言い回しの長さ。 */
+const countGrams = (text) => {
+    const counts = new Map();
+    Array.from({ length: Math.max(0, text.length - GRAM + 1) }).forEach((_, index) => {
+        const gram = text.slice(index, index + GRAM);
+        counts.set(gram, (counts.get(gram) ?? 0) + 1);
+    });
+    return counts;
+};
+/**
+ * 名前ではなく言い回しだけを数える。実文書で測ったら、上位は
+ * 「AGENTS.m」「シンギュラリティ」のような固有名詞と識別子だった。名前は繰り返して当たり前。
+ *
+ * 言い回しは、その言語の「つなぎの материал」を含む。日本語ならひらがな、英語なら語の切れ目。
+ * どちらを見るかは adapter が宣言する単位で決める。ひらがなを持たない char 単位の言語が
+ * 来たら、この見分けは効かなくなる（そのときは adapter 側に判定を移す）。
+ */
+const CONNECTIVE = { char: /[ぁ-ゖ]/u, word: / /u };
+const isPhrasing = (gram, unit) => CONNECTIVE[unit].test(gram);
+/** 文をまたいで数えない。文の終わりと次の文の始まりが繋がると、固有名詞が言い回しに見える。 */
+const gramsOf = (doc) => {
+    const counts = new Map();
+    doc.sentences.forEach((sentence) => {
+        const text = doc.lengthUnit === "char" ? proseText(sentence).replace(/\s+/gu, "") : proseText(sentence);
+        countGrams(text).forEach((count, gram) => counts.set(gram, (counts.get(gram) ?? 0) + count));
+    });
+    return counts;
+};
+export const ngramRepetition = (doc, options) => {
+    if (wordsOf(doc) < FLOOR[doc.lengthUnit])
+        return [];
+    const worst = [...gramsOf(doc).entries()]
+        .filter(([gram]) => isPhrasing(gram, doc.lengthUnit))
+        .reduce((best, [gram, count]) => (count > best.count ? { gram, count } : best), { gram: "", count: 0 });
+    if (worst.count <= options.limit)
+        return [];
+    const at = doc.sentences.find((sentence) => sentence.text.includes(worst.gram));
+    return [
+        {
+            rule: "ngram-repetition",
+            severity: "info",
+            line: 0,
+            column: 0,
+            quote: at?.text.trim() ?? worst.gram,
+            values: { word: worst.gram, count: worst.count, limit: options.limit, offset: at?.span.start ?? 0 },
+        },
+    ];
+};
+/**
+ * 略語が初出で展開されているか。「CI」だけでは、読む人によって指すものが違う。
+ * 展開は「略語の直前か直後の括弧」か「括弧の中の略語」で書かれる。
+ */
+const ACRONYM = /\b[A-Z]{2,6}\b/gu;
+/** 読み手が説明なしで通じると見なしてよい語。展開すると逆に読みにくい。 */
+const COMMON = new Set([
+    "OK",
+    "NG",
+    "URL",
+    "API",
+    "CSS",
+    "HTML",
+    "JSON",
+    "YAML",
+    "HTTP",
+    "HTTPS",
+    "PDF",
+    "CPU",
+    "GPU",
+    "RAM",
+    "USB",
+    "AI",
+    "ID",
+    "FAQ",
+    "PR",
+    "OS",
+    "CLI",
+    "UI",
+    "UX",
+    "SQL",
+    "CSV",
+    "XML",
+    "TODO",
+    "NOTE",
+]);
+/**
+ * 展開は略語の**すぐ隣**にあるときだけ認める。
+ * 60 文字も見ると、同じ文のどこかに括弧があるだけで「説明済み」になり、1 件も出なくなる。
+ */
+const AFTER = /^\s*[(（]/u;
+const BEFORE = /[)）]\s*$/u;
+const isExpanded = (body, acronym) => {
+    const at = body.indexOf(acronym);
+    if (at === -1)
+        return false;
+    return AFTER.test(body.slice(at + acronym.length, at + acronym.length + 3)) || BEFORE.test(body.slice(Math.max(0, at - 3), at));
+};
+export const undefinedAcronym = (doc, options) => {
+    const body = bodyOf(doc);
+    const seen = new Map();
+    findIn(doc, ACRONYM).forEach((hit) => {
+        const text = /^[A-Z]{2,6}/u.exec(doc.source.slice(hit.offset, hit.offset + 6))?.[0] ?? "";
+        if (text.length > 0 && !COMMON.has(text) && !seen.has(text))
+            seen.set(text, hit);
+    });
+    const bare = [...seen.entries()].filter(([acronym]) => !isExpanded(body, acronym));
+    if (bare.length < options.limit)
+        return [];
+    return bare.map(([acronym, hit]) => ({
+        rule: "undefined-acronym",
+        severity: "info",
+        line: 0,
+        column: 0,
+        quote: hit.sentence.text.trim(),
+        values: { word: acronym, count: bare.length, limit: options.limit, offset: hit.offset },
+    }));
+};
+/**
+ * 具体物の密度。数値・コード・リンク・引用が 1 つも無い節は、
+ * 読み終えても何も持ち帰れない。**少ないほうを指摘する**ので、他の密度 rule と向きが逆。
+ */
+const CONCRETE = /\d|`|https?:\/\//u;
+export const concreteEvidence = (doc, options) => {
+    const sections = doc.sections.filter((section) => section.sentences.length >= 3);
+    const empty = sections.filter((section) => !CONCRETE.test(doc.source.slice(section.span.start, section.span.end)));
+    if (empty.length < options.limit)
+        return [];
+    return empty.map((section) => ({
+        rule: "concrete-evidence-density",
+        severity: "info",
+        line: 0,
+        column: 0,
+        quote: section.heading.length > 0 ? section.heading : (section.sentences[0]?.text.trim() ?? ""),
+        values: {
+            word: section.heading.length > 0 ? section.heading : "この節",
+            count: empty.length,
+            total: sections.length,
+            limit: options.limit,
+            offset: section.span.start,
+        },
+    }));
+};
+//# sourceMappingURL=signals.js.map

package/dist/detectors/signals.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"signals.js","sourceRoot":"","sources":["../../src/detectors/signals.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,SAAS,EAAE,MAAM,eAAe,CAAC;AAC1C,OAAO,EAAE,OAAO,EAAE,MAAM,gBAAgB,CAAC;AAGzC,MAAM,GAAG,GAAG,IAAI,CAAC;AAEjB;;;GAGG;AACH,MAAM,KAAK,GAAG,EAAE,IAAI,EAAE,GAAG,EAAE,IAAI,EAAE,GAAG,EAAE,CAAC;AAEvC,MAAM,MAAM,GAAG,CAAC,GAAkB,EAAU,EAAE,CAAC,GAAG,CAAC,SAAS,CAAC,GAAG,CAAC,SAAS,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC;AAItF,MAAM,MAAM,GAAG,CAAC,GAAkB,EAAE,OAAe,EAAS,EAAE,CAC5D,GAAG,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC,GAAG,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,CAAC,EAAE,QAAQ,EAAE,MAAM,EAAE,QAAQ,CAAC,IAAI,CAAC,KAAK,GAAG,CAAC,KAAK,CAAC,KAAK,IAAI,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC;AAE7J,MAAM,OAAO,GAAG,CAAC,GAAkB,EAAE,KAAa,EAAU,EAAE;IAC5D,MAAM,MAAM,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC;IAC5B,OAAO,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,KAAK,GAAG,MAAM,CAAC,GAAG,GAAG,CAAC,CAAC;AAC/D,CAAC,CAAC;AAEF;;;GAGG;AACH,MAAM,KAAK,GAAG,sCAAsC,CAAC;AAErD,MAAM,CAAC,MAAM,YAAY,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IAChE,MAAM,IAAI,GAAG,MAAM,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;IAChC,MAAM,KAAK,GAAG,IAAI,CAAC,CAAC,CAAC,CAAC;IACtB,MAAM,IAAI,GAAG,OAAO,CAAC,GAAG,EAAE,IAAI,CAAC,MAAM,CAAC,CAAC;IACvC,IAAI,OAAO,CAAC,GAAG,CAAC,GAAG,KAAK,CAAC,GAAG,CAAC,UAAU,CAAC,IAAI,KAAK,KAAK,SAAS,IAAI,IAAI,IAAI,OAAO,CAAC,KAAK;QAAE,OAAO,EAAE,CAAC;IACpG,OAAO;QACL;YACE,IAAI,EAAE,eAAe;YACrB,QAAQ,EAAE,MAAM;YAChB,IAAI,EAAE,CAAC;YACP,MAAM,EAAE,CAAC;YACT,KAAK,EAAE,KAAK,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,EAAE;YACjC,MAAM,EAAE,EAAE,KAAK,EAAE,IAAI,CAAC,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,MAAM,EAAE;SAC1F;KACF,CAAC;AACJ,CAAC,CAAC;AAEF;;;GAGG;AACH,MAAM,IAAI,GAAG,CAAC,CAAC;AAEf,sDAAsD;AACtD,MAAM,UAAU,GAAG,CAAC,IAAY,EAAuB,EAAE;IACvD,MAAM,MAAM,GAAG,IAAI,GAAG,EAAkB,CAAC;IACzC,KAAK,CAAC,IAAI,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,MAAM,GAAG,IAAI,GAAG,CAAC,CAAC,EAAE,CAAC,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,KAAK,EAAE,EAAE;QAC/E,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,KAAK,GAAG,IAAI,CAAC,CAAC;QAC7C,MAAM,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC;IAChD,CAAC,CAAC,CAAC;IACH,OAAO,MAAM,CAAC;AAChB,CAAC,CAAC;AAEF;;;;;;;GAOG;AACH,MAAM,UAAU,GAAG,EAAE,IAAI,EAAE,QAAQ,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC;AAElD,MAAM,UAAU,GAAG,CAAC,IAAY,EAAE,IAAiC,EAAW,EAAE,CAAC,UAAU,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;AAE7G,mDAAmD;AACnD,MAAM,OAAO,GAAG,CAAC,GAAkB,EAAuB,EAAE;IAC1D,MAAM,MAAM,GAAG,IAAI,GAAG,EAAkB,CAAC;IACzC,GAAG,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC,QAAQ,EAAE,EAAE;QACjC,MAAM,IAAI,GAAG,GAAG,CAAC,UAAU,KAAK,MAAM,CAAC,CAAC,CAAC,SAAS,CAAC,QAAQ,CAAC,CAAC,OAAO,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,CAAC,CAAC,SAAS,CAAC,QAAQ,CAAC,CAAC;QACxG,UAAU,CAAC,IAAI,CAAC,CAAC,OAAO,CAAC,CAAC,KAAK,EAAE,IAAI,EAAE,EAAE,CAAC,MAAM,CAAC,GAAG,CAAC,IAAI,EAAE,CAAC,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,GAAG,KAAK,CAAC,CAAC,CAAC;IAC/F,CAAC,CAAC,CAAC;IACH,OAAO,MAAM,CAAC;AAChB,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,eAAe,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IACnE,IAAI,OAAO,CAAC,GAAG,CAAC,GAAG,KAAK,CAAC,GAAG,CAAC,UAAU,CAAC;QAAE,OAAO,EAAE,CAAC;IACpD,MAAM,KAAK,GAAG,CAAC,GAAG,OAAO,CAAC,GAAG,CAAC,CAAC,OAAO,EAAE,CAAC;SACtC,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,EAAE,CAAC,UAAU,CAAC,IAAI,EAAE,GAAG,CAAC,UAAU,CAAC,CAAC;SACpD,MAAM,CAAkC,CAAC,IAAI,EAAE,CAAC,IAAI,EAAE,KAAK,CAAC,EAAE,EAAE,CAAC,CAAC,KAAK,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC,EAAE,EAAE,IAAI,EAAE,EAAE,EAAE,KAAK,EAAE,CAAC,EAAE,CAAC,CAAC;IAC3I,IAAI,KAAK,CAAC,KAAK,IAAI,OAAO,CAAC,KAAK;QAAE,OAAO,EAAE,CAAC;IAC5C,MAAM,EAAE,GAAG,GAAG,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,QAAQ,EAAE,EAAE,CAAC,QAAQ,CAAC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,CAAC;IAChF,OAAO;QACL;YACE,IAAI,EAAE,kBAAkB;YACxB,QAAQ,EAAE,MAAM;YAChB,IAAI,EAAE,CAAC;YACP,MAAM,EAAE,CAAC;YACT,KAAK,EAAE,EAAE,EAAE,IAAI,CAAC,IAAI,EAAE,IAAI,KAAK,CAAC,IAAI;YACpC,MAAM,EAAE,EAAE,IAAI,EAAE,KAAK,CAAC,IAAI,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,EAAE,EAAE,IAAI,CAAC,KAAK,IAAI,CAAC,EAAE;SACpG;KACF,CAAC;AACJ,CAAC,CAAC;AAEF;;;GAGG;AACH,MAAM,OAAO,GAAG,kBAAkB,CAAC;AAEnC,yCAAyC;AACzC,MAAM,MAAM,GAAG,IAAI,GAAG,CAAC;IACrB,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,KAAK;IACL,KAAK;IACL,MAAM;IACN,MAAM;IACN,MAAM;IACN,MAAM;IACN,OAAO;IACP,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,IAAI;IACJ,IAAI;IACJ,KAAK;IACL,KAAK;IACL,KAAK;IACL,MAAM;IACN,MAAM;CACP,CAAC,CAAC;AAEH;;;GAGG;AACH,MAAM,KAAK,GAAG,WAAW,CAAC;AAC1B,MAAM,MAAM,GAAG,WAAW,CAAC;AAE3B,MAAM,UAAU,GAAG,CAAC,IAAY,EAAE,OAAe,EAAW,EAAE;IAC5D,MAAM,EAAE,GAAG,IAAI,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC;IACjC,IAAI,EAAE,KAAK,CAAC,CAAC;QAAE,OAAO,KAAK,CAAC;IAC5B,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,GAAG,OAAO,CAAC,MAAM,EAAE,EAAE,GAAG,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,IAAI,MAAM,CAAC,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,EAAE,GAAG,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC;AAClI,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,gBAAgB,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IACpE,MAAM,IAAI,GAAG,MAAM,CAAC,GAAG,CAAC,CAAC;IACzB,MAAM,IAAI,GAAG,IAAI,GAAG,EAAe,CAAC;IACpC,MAAM,CAAC,GAAG,EAAE,OAAO,CAAC,CAAC,OAAO,CAAC,CAAC,GAAG,EAAE,EAAE;QACnC,MAAM,IAAI,GAAG,cAAc,CAAC,IAAI,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,GAAG,CAAC,MAAM,EAAE,GAAG,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;QAC1F,IAAI,IAAI,CAAC,MAAM,GAAG,CAAC,IAAI,CAAC,MAAM,CAAC,GAAG,CAAC,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC;YAAE,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,GAAG,CAAC,CAAC;IACnF,CAAC,CAAC,CAAC;IACH,MAAM,IAAI,GAAG,CAAC,GAAG,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,OAAO,CAAC,EAAE,EAAE,CAAC,CAAC,UAAU,CAAC,IAAI,EAAE,OAAO,CAAC,CAAC,CAAC;IACnF,IAAI,IAAI,CAAC,MAAM,GAAG,OAAO,CAAC,KAAK;QAAE,OAAO,EAAE,CAAC;IAC3C,OAAO,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,OAAO,EAAE,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC;QACnC,IAAI,EAAE,mBAAmB;QACzB,QAAQ,EAAE,MAAM;QAChB,IAAI,EAAE,CAAC;QACP,MAAM,EAAE,CAAC;QACT,KAAK,EAAE,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,IAAI,EAAE;QAC/B,MAAM,EAAE,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,IAAI,CAAC,MAAM,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,CAAC,MAAM,EAAE;KACxF,CAAC,CAAC,CAAC;AACN,CAAC,CAAC;AAEF;;;GAGG;AACH,MAAM,QAAQ,GAAG,mBAAmB,CAAC;AAErC,MAAM,CAAC,MAAM,gBAAgB,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IACpE,MAAM,QAAQ,GAAG,GAAG,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,CAAC,SAAS,CAAC,MAAM,IAAI,CAAC,CAAC,CAAC;IACjF,MAAM,KAAK,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,QAAQ,CAAC,IAAI,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,OAAO,CAAC,IAAI,CAAC,KAAK,EAAE,OAAO,CAAC,IAAI,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;IACnH,IAAI,KAAK,CAAC,MAAM,GAAG,OAAO,CAAC,KAAK;QAAE,OAAO,EAAE,CAAC;IAC5C,OAAO,KAAK,CAAC,GAAG,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC;QAC7B,IAAI,EAAE,2BAA2B;QACjC,QAAQ,EAAE,MAAM;QAChB,IAAI,EAAE,CAAC;QACP,MAAM,EAAE,CAAC;QACT,KAAK,EAAE,OAAO,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,SAAS,CAAC,CAAC,CAAC,EAAE,IAAI,CAAC,IAAI,EAAE,IAAI,EAAE,CAAC;QAC/F,MAAM,EAAE;YACN,IAAI,EAAE,OAAO,CAAC,OAAO,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,KAAK;YAC1D,KAAK,EAAE,KAAK,CAAC,MAAM;YACnB,KAAK,EAAE,QAAQ,CAAC,MAAM;YACtB,KAAK,EAAE,OAAO,CAAC,KAAK;YACpB,MAAM,EAAE,OAAO,CAAC,IAAI,CAAC,KAAK;SAC3B;KACF,CAAC,CAAC,CAAC;AACN,CAAC,CAAC"}

package/dist/detectors/structure.d.ts

@@ -0,0 +1,24 @@
+import type { Detector, ProseDocument } from "../plugin.ts";
+/**
+ * ばらつきは変動係数（標準偏差 ÷ 平均）で測る。
+ *
+ * 標準偏差そのものを見ると、長い文書ほど大きくなって閾値が効かない。
+ * 平均で割ると長さに依らなくなり、「揃いすぎ」と「ばらつきすぎ」を同じ尺度で言える。
+ */
+export declare const coefficientOfVariation: (values: readonly number[]) => number | undefined;
+/** 1 段落に文を詰めすぎると、読み手はどこで息継ぎしていいか分からなくなる。 */
+export declare const paragraphLength: Detector;
+/** 段落の長さが揃いすぎている。人が書くと、言いたいことの重さで段落の長さは揺れる。 */
+export declare const paragraphVariance: Detector;
+/** 節の長さが揃いすぎている。見出しごとに機械的に埋めた文書で起きる。 */
+export declare const sectionUniformity: Detector;
+export declare const ruleOfThree: Detector;
+/**
+ * 本題に入るまでが長い。業務文書では、読み手は結論を探しに来ている。
+ *
+ * 「本題」は最初の中見出し（深さ 2 以上）とする。表題（深さ 1）の直後から数えると、
+ * 表題しか無い文書で全文が前置きになる。中見出しが無い文書では何も言わない。
+ */
+export declare const preambleLength: Detector;
+export declare const wordsOf: (doc: ProseDocument) => number;
+//# sourceMappingURL=structure.d.ts.map

package/dist/detectors/structure.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"structure.d.ts","sourceRoot":"","sources":["../../src/detectors/structure.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAc,QAAQ,EAAsB,aAAa,EAAW,MAAM,cAAc,CAAC;AAErG;;;;;GAKG;AACH,eAAO,MAAM,sBAAsB,GAAI,QAAQ,SAAS,MAAM,EAAE,KAAG,MAAM,GAAG,SAM3E,CAAC;AAsBF,4CAA4C;AAC5C,eAAO,MAAM,eAAe,EAAE,QAUvB,CAAC;AAIR,+CAA+C;AAC/C,eAAO,MAAM,iBAAiB,EAAE,QAI/B,CAAC;AAIF,wCAAwC;AACxC,eAAO,MAAM,iBAAiB,EAAE,QAI/B,CAAC;AAQF,eAAO,MAAM,WAAW,EAAE,QAgBzB,CAAC;AAEF;;;;;GAKG;AACH,eAAO,MAAM,cAAc,EAAE,QAkB5B,CAAC;AAEF,eAAO,MAAM,OAAO,GAAI,KAAK,aAAa,KAAG,MAA8F,CAAC"}

package/dist/detectors/structure.js

@@ -0,0 +1,111 @@
+import { charLength, lengthOf } from "../measure.js";
+/**
+ * ばらつきは変動係数（標準偏差 ÷ 平均）で測る。
+ *
+ * 標準偏差そのものを見ると、長い文書ほど大きくなって閾値が効かない。
+ * 平均で割ると長さに依らなくなり、「揃いすぎ」と「ばらつきすぎ」を同じ尺度で言える。
+ */
+export const coefficientOfVariation = (values) => {
+    if (values.length < 2)
+        return undefined;
+    const mean = values.reduce((sum, value) => sum + value, 0) / values.length;
+    if (mean === 0)
+        return undefined;
+    const variance = values.reduce((sum, value) => sum + (value - mean) ** 2, 0) / values.length;
+    return Math.sqrt(variance) / mean;
+};
+/** 変動係数は 0..1 では読めないので、パーセントの整数にして閾値と突き合わせる。 */
+const PERCENT = 100;
+const asPercent = (value) => Math.round(value * PERCENT);
+const uniformity = (doc, values, span, limit, rule) => {
+    const cv = coefficientOfVariation(values);
+    if (cv === undefined || asPercent(cv) >= limit)
+        return [];
+    return [
+        {
+            rule,
+            severity: "info",
+            line: 0,
+            column: 0,
+            quote: doc.source.slice(span.start, span.start + 80).trim(),
+            values: { count: values.length, variance: asPercent(cv), limit, offset: span.start },
+        },
+    ];
+};
+/** 1 段落に文を詰めすぎると、読み手はどこで息継ぎしていいか分からなくなる。 */
+export const paragraphLength = (doc, options) => doc.paragraphs
+    .filter((paragraph) => paragraph.sentences.length > options.limit)
+    .map((paragraph) => ({
+    rule: "max-paragraph-length",
+    severity: "warning",
+    line: 0,
+    column: 0,
+    quote: doc.source.slice(paragraph.span.start, paragraph.span.start + 80).trim(),
+    values: { count: paragraph.sentences.length, limit: options.limit, offset: paragraph.span.start },
+}));
+const charsOf = (paragraph) => paragraph.sentences.reduce((sum, sentence) => sum + charLength(sentence), 0);
+/** 段落の長さが揃いすぎている。人が書くと、言いたいことの重さで段落の長さは揺れる。 */
+export const paragraphVariance = (doc, options) => {
+    const lengths = doc.paragraphs.map(charsOf).filter((length) => length > 0);
+    const first = doc.paragraphs[0];
+    return first === undefined ? [] : uniformity(doc, lengths, first.span, options.limit, "paragraph-length-variance");
+};
+const sectionChars = (section) => section.sentences.reduce((sum, sentence) => sum + charLength(sentence), 0);
+/** 節の長さが揃いすぎている。見出しごとに機械的に埋めた文書で起きる。 */
+export const sectionUniformity = (doc, options) => {
+    const lengths = doc.sections.map(sectionChars).filter((length) => length > 0);
+    const first = doc.sections[0];
+    return first === undefined ? [] : uniformity(doc, lengths, first.span, options.limit, "section-length-uniformity");
+};
+/**
+ * 3 項目の箇条書きばかりになっている。
+ * 3 つ並べると収まりがよいので、言いたいことが 2 つでも 4 つでも 3 つにされる。
+ */
+const isThree = (list) => list.items.length === 3;
+export const ruleOfThree = (doc, options) => {
+    const lists = doc.lists;
+    const threes = lists.filter(isThree);
+    const share = lists.length === 0 ? 0 : Math.round((threes.length / lists.length) * PERCENT);
+    const first = threes[0];
+    if (lists.length < 3 || first === undefined || share <= options.limit)
+        return [];
+    return [
+        {
+            rule: "rule-of-three",
+            severity: "info",
+            line: 0,
+            column: 0,
+            quote: doc.source.slice(first.span.start, first.span.start + 80).trim(),
+            values: { count: threes.length, total: lists.length, share, limit: options.limit, offset: first.span.start },
+        },
+    ];
+};
+/**
+ * 本題に入るまでが長い。業務文書では、読み手は結論を探しに来ている。
+ *
+ * 「本題」は最初の中見出し（深さ 2 以上）とする。表題（深さ 1）の直後から数えると、
+ * 表題しか無い文書で全文が前置きになる。中見出しが無い文書では何も言わない。
+ */
+export const preambleLength = (doc, options) => {
+    const body = doc.sections.find((section) => section.depth >= 2);
+    if (body === undefined)
+        return [];
+    const before = doc.paragraphs.filter((paragraph) => paragraph.span.start < body.span.start);
+    if (before.length <= options.limit)
+        return [];
+    const first = before[0];
+    return first === undefined
+        ? []
+        : [
+            {
+                rule: "preamble-length",
+                severity: "info",
+                line: 0,
+                column: 0,
+                quote: doc.source.slice(first.span.start, first.span.start + 80).trim(),
+                values: { count: before.length, limit: options.limit, offset: first.span.start },
+            },
+        ];
+};
+export const wordsOf = (doc) => doc.sentences.reduce((sum, sentence) => sum + lengthOf(sentence, doc.lengthUnit), 0);
+//# sourceMappingURL=structure.js.map

package/dist/detectors/structure.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"structure.js","sourceRoot":"","sources":["../../src/detectors/structure.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,QAAQ,EAAE,MAAM,eAAe,CAAC;AAGrD;;;;;GAKG;AACH,MAAM,CAAC,MAAM,sBAAsB,GAAG,CAAC,MAAyB,EAAsB,EAAE;IACtF,IAAI,MAAM,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,SAAS,CAAC;IACxC,MAAM,IAAI,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,EAAE,CAAC,CAAC,GAAG,MAAM,CAAC,MAAM,CAAC;IAC3E,IAAI,IAAI,KAAK,CAAC;QAAE,OAAO,SAAS,CAAC;IACjC,MAAM,QAAQ,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,CAAC,KAAK,GAAG,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC,GAAG,MAAM,CAAC,MAAM,CAAC;IAC7F,OAAO,IAAI,CAAC,IAAI,CAAC,QAAQ,CAAC,GAAG,IAAI,CAAC;AACpC,CAAC,CAAC;AAEF,gDAAgD;AAChD,MAAM,OAAO,GAAG,GAAG,CAAC;AAEpB,MAAM,SAAS,GAAG,CAAC,KAAa,EAAU,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,GAAG,OAAO,CAAC,CAAC;AAEzE,MAAM,UAAU,GAAG,CAAC,GAAkB,EAAE,MAAyB,EAAE,IAAuB,EAAE,KAAa,EAAE,IAAY,EAAa,EAAE;IACpI,MAAM,EAAE,GAAG,sBAAsB,CAAC,MAAM,CAAC,CAAC;IAC1C,IAAI,EAAE,KAAK,SAAS,IAAI,SAAS,CAAC,EAAE,CAAC,IAAI,KAAK;QAAE,OAAO,EAAE,CAAC;IAC1D,OAAO;QACL;YACE,IAAI;YACJ,QAAQ,EAAE,MAAM;YAChB,IAAI,EAAE,CAAC;YACP,MAAM,EAAE,CAAC;YACT,KAAK,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE,IAAI,CAAC,KAAK,GAAG,EAAE,CAAC,CAAC,IAAI,EAAE;YAC3D,MAAM,EAAE,EAAE,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,QAAQ,EAAE,SAAS,CAAC,EAAE,CAAC,EAAE,KAAK,EAAE,MAAM,EAAE,IAAI,CAAC,KAAK,EAAE;SACrF;KACF,CAAC;AACJ,CAAC,CAAC;AAEF,4CAA4C;AAC5C,MAAM,CAAC,MAAM,eAAe,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE,CACnE,GAAG,CAAC,UAAU;KACX,MAAM,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,SAAS,CAAC,SAAS,CAAC,MAAM,GAAG,OAAO,CAAC,KAAK,CAAC;KACjE,GAAG,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,CAAC;IACnB,IAAI,EAAE,sBAAsB;IAC5B,QAAQ,EAAE,SAAS;IACnB,IAAI,EAAE,CAAC;IACP,MAAM,EAAE,CAAC;IACT,KAAK,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,SAAS,CAAC,IAAI,CAAC,KAAK,EAAE,SAAS,CAAC,IAAI,CAAC,KAAK,GAAG,EAAE,CAAC,CAAC,IAAI,EAAE;IAC/E,MAAM,EAAE,EAAE,KAAK,EAAE,SAAS,CAAC,SAAS,CAAC,MAAM,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,SAAS,CAAC,IAAI,CAAC,KAAK,EAAE;CAClG,CAAC,CAAC,CAAC;AAER,MAAM,OAAO,GAAG,CAAC,SAAoB,EAAU,EAAE,CAAC,SAAS,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,QAAQ,EAAE,EAAE,CAAC,GAAG,GAAG,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,CAAC;AAE/H,+CAA+C;AAC/C,MAAM,CAAC,MAAM,iBAAiB,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IACrE,MAAM,OAAO,GAAG,GAAG,CAAC,UAAU,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,MAAM,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IAC3E,MAAM,KAAK,GAAG,GAAG,CAAC,UAAU,CAAC,CAAC,CAAC,CAAC;IAChC,OAAO,KAAK,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,GAAG,EAAE,OAAO,EAAE,KAAK,CAAC,IAAI,EAAE,OAAO,CAAC,KAAK,EAAE,2BAA2B,CAAC,CAAC;AACrH,CAAC,CAAC;AAEF,MAAM,YAAY,GAAG,CAAC,OAAgB,EAAU,EAAE,CAAC,OAAO,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,QAAQ,EAAE,EAAE,CAAC,GAAG,GAAG,UAAU,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,CAAC;AAE9H,wCAAwC;AACxC,MAAM,CAAC,MAAM,iBAAiB,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IACrE,MAAM,OAAO,GAAG,GAAG,CAAC,QAAQ,CAAC,GAAG,CAAC,YAAY,CAAC,CAAC,MAAM,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;IAC9E,MAAM,KAAK,GAAG,GAAG,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC;IAC9B,OAAO,KAAK,KAAK,SAAS,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,UAAU,CAAC,GAAG,EAAE,OAAO,EAAE,KAAK,CAAC,IAAI,EAAE,OAAO,CAAC,KAAK,EAAE,2BAA2B,CAAC,CAAC;AACrH,CAAC,CAAC;AAEF;;;GAGG;AACH,MAAM,OAAO,GAAG,CAAC,IAAgB,EAAW,EAAE,CAAC,IAAI,CAAC,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC;AAEvE,MAAM,CAAC,MAAM,WAAW,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IAC/D,MAAM,KAAK,GAAG,GAAG,CAAC,KAAK,CAAC;IACxB,MAAM,MAAM,GAAG,KAAK,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC;IACrC,MAAM,KAAK,GAAG,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,MAAM,CAAC,MAAM,GAAG,KAAK,CAAC,MAAM,CAAC,GAAG,OAAO,CAAC,CAAC;IAC5F,MAAM,KAAK,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC;IACxB,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,IAAI,OAAO,CAAC,KAAK;QAAE,OAAO,EAAE,CAAC;IACjF,OAAO;QACL;YACE,IAAI,EAAE,eAAe;YACrB,QAAQ,EAAE,MAAM;YAChB,IAAI,EAAE,CAAC;YACP,MAAM,EAAE,CAAC;YACT,KAAK,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,KAAK,GAAG,EAAE,CAAC,CAAC,IAAI,EAAE;YACvE,MAAM,EAAE,EAAE,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK,EAAE,KAAK,CAAC,MAAM,EAAE,KAAK,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE;SAC7G;KACF,CAAC;AACJ,CAAC,CAAC;AAEF;;;;;GAKG;AACH,MAAM,CAAC,MAAM,cAAc,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IAClE,MAAM,IAAI,GAAG,GAAG,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,OAAO,CAAC,KAAK,IAAI,CAAC,CAAC,CAAC;IAChE,IAAI,IAAI,KAAK,SAAS;QAAE,OAAO,EAAE,CAAC;IAClC,MAAM,MAAM,GAAG,GAAG,CAAC,UAAU,CAAC,MAAM,CAAC,CAAC,SAAS,EAAE,EAAE,CAAC,SAAS,CAAC,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;IAC5F,IAAI,MAAM,CAAC,MAAM,IAAI,OAAO,CAAC,KAAK;QAAE,OAAO,EAAE,CAAC;IAC9C,MAAM,KAAK,GAAG,MAAM,CAAC,CAAC,CAAC,CAAC;IACxB,OAAO,KAAK,KAAK,SAAS;QACxB,CAAC,CAAC,EAAE;QACJ,CAAC,CAAC;YACE;gBACE,IAAI,EAAE,iBAAiB;gBACvB,QAAQ,EAAE,MAAM;gBAChB,IAAI,EAAE,CAAC;gBACP,MAAM,EAAE,CAAC;gBACT,KAAK,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,KAAK,GAAG,EAAE,CAAC,CAAC,IAAI,EAAE;gBACvE,MAAM,EAAE,EAAE,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE;aACjF;SACF,CAAC;AACR,CAAC,CAAC;AAEF,MAAM,CAAC,MAAM,OAAO,GAAG,CAAC,GAAkB,EAAU,EAAE,CAAC,GAAG,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,QAAQ,EAAE,EAAE,CAAC,GAAG,GAAG,QAAQ,CAAC,QAAQ,EAAE,GAAG,CAAC,UAAU,CAAC,EAAE,CAAC,CAAC,CAAC"}

package/dist/detectors/token-shape.d.ts

@@ -0,0 +1,4 @@
+import type { Detector } from "../plugin.ts";
+export declare const nounEnding: Detector;
+export declare const doubledParticle: Detector;
+//# sourceMappingURL=token-shape.d.ts.map

package/dist/detectors/token-shape.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"token-shape.d.ts","sourceRoot":"","sources":["../../src/detectors/token-shape.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,QAAQ,EAAwB,MAAM,cAAc,CAAC;AAanE,eAAO,MAAM,UAAU,EAAE,QAmBxB,CAAC;AAiCF,eAAO,MAAM,eAAe,EAAE,QAc7B,CAAC"}

package/dist/detectors/token-shape.js

@@ -0,0 +1,62 @@
+import { hasParticle, isClosed, lastContent } from "../sentence-shape.js";
+/**
+ * 文末が名詞で終わる（体言止め）。箇条書きでは普通だが、本文では文が途中で切れて読める。
+ * 実文書で測ったら、指摘の 9 割が箇条書きの項目だった。箇条書きは除く。
+ */
+const NOMINAL = new Set(["NOUN", "PROPN"]);
+/** 非自立名詞は単独では何も指さない。「回るのか。」の「の」を体言止めと読まない。 */
+const isDependent = (token) => token.features?.["NounType"] === "Dependent";
+const inside = (span, spans) => spans.some((other) => span.start >= other.start && span.start < other.end);
+export const nounEnding = (doc, options) => {
+    const ended = doc.sentences.filter((sentence) => {
+        if (inside(sentence.span, doc.listSpans) || !isClosed(sentence) || !hasParticle(sentence))
+            return false;
+        const last = lastContent(sentence);
+        return last !== undefined && NOMINAL.has(last.pos) && !isDependent(last);
+    });
+    const first = ended[0];
+    // 文書全体の数についての指摘なので、1 文ずつ出さない。13 件並べても直す順番が分からない。
+    if (first === undefined || ended.length <= options.limit)
+        return [];
+    return [
+        {
+            rule: "taigen-dome-in-prose",
+            severity: "info",
+            line: 0,
+            column: 0,
+            quote: first.text.trim(),
+            values: { count: ended.length, limit: options.limit, offset: first.span.start },
+        },
+    ];
+};
+const extend = (runs, token) => {
+    const last = runs.at(-1);
+    if (last !== undefined && last.surface === token.surface)
+        return [...runs.slice(0, -1), { surface: last.surface, tokens: [...last.tokens, token] }];
+    return [...runs, { surface: token.surface, tokens: [token] }];
+};
+const PARTICLE = new Set(["ADP", "SCONJ", "PART", "CCONJ"]);
+const BREAK = { surface: "", tokens: [] };
+/**
+ * 名詞は連なりの中身なので飛ばす。読点と、語彙表に無い助詞は連なりを切る。
+ * 「弊社のAはBの」を 2 回続いたと数えないため。「は」でいちど句が閉じている。
+ */
+const runsOf = (tokens, nesting) => tokens.reduce((acc, token) => {
+    if (nesting.includes(token.surface))
+        return extend(acc, token);
+    return token.pos === "PUNCT" || PARTICLE.has(token.pos) ? [...acc, BREAK] : acc;
+}, []);
+export const doubledParticle = (doc, options) => {
+    const nesting = (options.lexicon ?? []).map((entry) => entry.pattern);
+    return doc.sentences.flatMap((sentence) => runsOf(sentence.tokens ?? [], nesting)
+        .filter((run) => run.tokens.length > options.limit)
+        .map((run) => ({
+        rule: "no-doubled-joshi",
+        severity: "warning",
+        line: 0,
+        column: 0,
+        quote: sentence.text.trim(),
+        values: { word: run.surface, count: run.tokens.length, limit: options.limit, offset: run.tokens[0]?.span.start ?? sentence.span.start },
+    })));
+};
+//# sourceMappingURL=token-shape.js.map

package/dist/detectors/token-shape.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"token-shape.js","sourceRoot":"","sources":["../../src/detectors/token-shape.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAE,QAAQ,EAAE,WAAW,EAAE,MAAM,sBAAsB,CAAC;AAG1E;;;GAGG;AACH,MAAM,OAAO,GAAG,IAAI,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;AAE3C,8CAA8C;AAC9C,MAAM,WAAW,GAAG,CAAC,KAAY,EAAW,EAAE,CAAC,KAAK,CAAC,QAAQ,EAAE,CAAC,UAAU,CAAC,KAAK,WAAW,CAAC;AAE5F,MAAM,MAAM,GAAG,CAAC,IAAU,EAAE,KAAsB,EAAW,EAAE,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,IAAI,CAAC,KAAK,IAAI,KAAK,CAAC,KAAK,IAAI,IAAI,CAAC,KAAK,GAAG,KAAK,CAAC,GAAG,CAAC,CAAC;AAE3I,MAAM,CAAC,MAAM,UAAU,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IAC9D,MAAM,KAAK,GAAG,GAAG,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,QAAQ,EAAE,EAAE;QAC9C,IAAI,MAAM,CAAC,QAAQ,CAAC,IAAI,EAAE,GAAG,CAAC,SAAS,CAAC,IAAI,CAAC,QAAQ,CAAC,QAAQ,CAAC,IAAI,CAAC,WAAW,CAAC,QAAQ,CAAC;YAAE,OAAO,KAAK,CAAC;QACxG,MAAM,IAAI,GAAG,WAAW,CAAC,QAAQ,CAAC,CAAC;QACnC,OAAO,IAAI,KAAK,SAAS,IAAI,OAAO,CAAC,GAAG,CAAC,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,WAAW,CAAC,IAAI,CAAC,CAAC;IAC3E,CAAC,CAAC,CAAC;IACH,MAAM,KAAK,GAAG,KAAK,CAAC,CAAC,CAAC,CAAC;IACvB,iDAAiD;IACjD,IAAI,KAAK,KAAK,SAAS,IAAI,KAAK,CAAC,MAAM,IAAI,OAAO,CAAC,KAAK;QAAE,OAAO,EAAE,CAAC;IACpE,OAAO;QACL;YACE,IAAI,EAAE,sBAAsB;YAC5B,QAAQ,EAAE,MAAM;YAChB,IAAI,EAAE,CAAC;YACP,MAAM,EAAE,CAAC;YACT,KAAK,EAAE,KAAK,CAAC,IAAI,CAAC,IAAI,EAAE;YACxB,MAAM,EAAE,EAAE,KAAK,EAAE,KAAK,CAAC,MAAM,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,KAAK,CAAC,IAAI,CAAC,KAAK,EAAE;SAChF;KACF,CAAC;AACJ,CAAC,CAAC;AAaF,MAAM,MAAM,GAAG,CAAC,IAAoB,EAAE,KAAY,EAAS,EAAE;IAC3D,MAAM,IAAI,GAAG,IAAI,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;IACzB,IAAI,IAAI,KAAK,SAAS,IAAI,IAAI,CAAC,OAAO,KAAK,KAAK,CAAC,OAAO;QAAE,OAAO,CAAC,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,EAAE,OAAO,EAAE,IAAI,CAAC,OAAO,EAAE,MAAM,EAAE,CAAC,GAAG,IAAI,CAAC,MAAM,EAAE,KAAK,CAAC,EAAE,CAAC,CAAC;IACpJ,OAAO,CAAC,GAAG,IAAI,EAAE,EAAE,OAAO,EAAE,KAAK,CAAC,OAAO,EAAE,MAAM,EAAE,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;AAChE,CAAC,CAAC;AAEF,MAAM,QAAQ,GAAG,IAAI,GAAG,CAAC,CAAC,KAAK,EAAE,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC;AAE5D,MAAM,KAAK,GAAQ,EAAE,OAAO,EAAE,EAAE,EAAE,MAAM,EAAE,EAAE,EAAE,CAAC;AAE/C;;;GAGG;AACH,MAAM,MAAM,GAAG,CAAC,MAAwB,EAAE,OAA0B,EAAS,EAAE,CAC7E,MAAM,CAAC,MAAM,CAAQ,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE;IAClC,IAAI,OAAO,CAAC,QAAQ,CAAC,KAAK,CAAC,OAAO,CAAC;QAAE,OAAO,MAAM,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC;IAC/D,OAAO,KAAK,CAAC,GAAG,KAAK,OAAO,IAAI,QAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,GAAG,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC;AAClF,CAAC,EAAE,EAAE,CAAC,CAAC;AAET,MAAM,CAAC,MAAM,eAAe,GAAa,CAAC,GAAG,EAAE,OAAO,EAAa,EAAE;IACnE,MAAM,OAAO,GAAG,CAAC,OAAO,CAAC,OAAO,IAAI,EAAE,CAAC,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;IACtE,OAAO,GAAG,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC,QAAQ,EAAE,EAAE,CACxC,MAAM,CAAC,QAAQ,CAAC,MAAM,IAAI,EAAE,EAAE,OAAO,CAAC;SACnC,MAAM,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,GAAG,OAAO,CAAC,KAAK,CAAC;SAClD,GAAG,CAAC,CAAC,GAAG,EAAE,EAAE,CAAC,CAAC;QACb,IAAI,EAAE,kBAAkB;QACxB,QAAQ,EAAE,SAAkB;QAC5B,IAAI,EAAE,CAAC;QACP,MAAM,EAAE,CAAC;QACT,KAAK,EAAE,QAAQ,CAAC,IAAI,CAAC,IAAI,EAAE;QAC3B,MAAM,EAAE,EAAE,IAAI,EAAE,GAAG,CAAC,OAAO,EAAE,KAAK,EAAE,GAAG,CAAC,MAAM,CAAC,MAAM,EAAE,KAAK,EAAE,OAAO,CAAC,KAAK,EAAE,MAAM,EAAE,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,IAAI,CAAC,KAAK,IAAI,QAAQ,CAAC,IAAI,CAAC,KAAK,EAAE;KACxI,CAAC,CAAC,CACN,CAAC;AACJ,CAAC,CAAC"}

package/dist/document.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"document.d.ts","sourceRoot":"","sources":["../src/document.ts"],"names":[],"mappings":"AAMA,OAAO,KAAK,EAAE,eAAe,EAAE,aAAa,EAA2B,MAAM,aAAa,CAAC;AAsI3F,eAAO,MAAM,aAAa,GAAI,MAAM,MAAM,EAAE,QAAQ,MAAM,EAAE,SAAS,eAAe,KAAG,aAatF,CAAC"}
+{"version":3,"file":"document.d.ts","sourceRoot":"","sources":["../src/document.ts"],"names":[],"mappings":"AAMA,OAAO,KAAK,EAAc,eAAe,EAAa,aAAa,EAA2B,MAAM,aAAa,CAAC;AAoNlH,eAAO,MAAM,aAAa,GAAI,MAAM,MAAM,EAAE,QAAQ,MAAM,EAAE,SAAS,eAAe,KAAG,aAuBtF,CAAC"}

package/dist/document.js

@@ -4,8 +4,14 @@ import { gfmTableFromMarkdown } from "mdast-util-gfm-table";
 import { frontmatter } from "micromark-extension-frontmatter";
 import { frontmatterFromMarkdown } from "mdast-util-frontmatter";
 import { maskSpans } from "./mask.js";
-/** 本文として数えないもの。コードは文章ではなく、表は文章の形をしていない。 */
-const NOT_PROSE = new Set(["code", "inlineCode", "html", "yaml", "toml", "table", "thematicBreak", "definition", "image", "imageReference"]);
+/**
+ * 本文として数えないもの。
+ *
+ * コードは文章ではなく、表は文章の形をしていない。
+ * **引用は自分の文章ではない。** chaff が言えるのは「2 文に割ってください」までで、
+ * 引用文にそれはできない。直せないものを指摘しても、書いた人は動けない。
+ */
+const NOT_PROSE = new Set(["code", "inlineCode", "html", "yaml", "toml", "table", "blockquote", "thematicBreak", "definition", "image", "imageReference"]);
 const spanOf = (node) => {
     const start = node.position?.start.offset;
     const end = node.position?.end.offset;
@@ -29,8 +35,34 @@ const linkChrome = (node) => {
         { start: inner.end, end: whole.end },
     ];
 };
-const collectMasks = (root) => {
-    const spans = [];
+/**
+ * 強調の記号（`**` `*` `_` `~~`）だけを覆う。囲まれた文字は本文なので残す。
+ *
+ * 残すと解析器が `**。` を 1 語の名詞として拾い、文の終わりが消える。
+ * 実文書で「文末が名詞」の誤検知を追って見つけた。文長にも同じだけ効いている。
+ */
+const MARKER = /^[*_~]+/u;
+const emphasisChrome = (node, source) => {
+    const whole = spanOf(node);
+    if (whole === undefined)
+        return [];
+    const width = MARKER.exec(source.slice(whole.start, whole.end))?.[0].length ?? 0;
+    if (width === 0)
+        return [];
+    return [
+        { start: whole.start, end: whole.start + width },
+        { start: whole.end - width, end: whole.end },
+    ];
+};
+const EMPHASIS = new Set(["strong", "emphasis", "delete"]);
+/**
+ * `:::note` `:::` のようなディレクティブ。Zenn・Docusaurus・VitePress が使う記法で、
+ * 標準の Markdown には無いため段落として解析される。囲みの指定であって文章ではない。
+ */
+const DIRECTIVE = /^:::[^\n]*/gmu;
+const directiveSpans = (source) => [...source.matchAll(DIRECTIVE)].map((match) => ({ start: match.index, end: match.index + match[0].length }));
+const collectMasks = (root, source) => {
+    const spans = [...directiveSpans(source)];
     walk(root, (node) => {
         if (node.type === "link" || node.type === "linkReference") {
             spans.push(...linkChrome(node));
@@ -44,6 +76,14 @@ const collectMasks = (root) => {
     });
     return spans;
 };
+const emphasisSpans = (root, source) => {
+    const spans = [];
+    walk(root, (node) => {
+        if (EMPHASIS.has(node.type))
+            spans.push(...emphasisChrome(node, source));
+    });
+    return spans;
+};
 const textOf = (node, source) => {
     const parts = [];
     walk(node, (child) => {
@@ -67,22 +107,30 @@ const headingsOf = (root, source) => {
     });
     return found;
 };
-const strongSpans = (root) => {
+const startsInside = (span, regions) => regions.some((region) => span.start >= region.start && span.start < region.end);
+/**
+ * 本文の強調だけを数える。
+ *
+ * 表のセルや見出しの中の太字はラベルであって強調ではない。
+ * 「太字は読者の目を止める道具」という bold-density の理屈が当てはまらない。
+ * 「本文でないもの」の定義は覆う範囲（collectMasks）に 1 つだけ置き、ここはそれを使う。
+ */
+const strongSpans = (root, masked) => {
     const found = [];
     walk(root, (node) => {
         if (node.type !== "strong")
             return;
         const span = spanOf(node);
-        if (span !== undefined)
+        if (span !== undefined && !startsInside(span, masked))
             found.push(span);
     });
     return found;
 };
 const within = (span, from, to) => span.start >= from && span.start < to;
-const paragraphSpans = (root) => {
+const spansOfType = (root, type) => {
     const found = [];
     walk(root, (node) => {
-        if (node.type !== "paragraph")
+        if (node.type !== type)
             return;
         const span = spanOf(node);
         if (span !== undefined)
@@ -97,12 +145,14 @@ const paragraphSpans = (root) => {
  * 覆った表やコードブロックを越えて次の句点まで飲み込む。段落は文が跨がない境界なので、
  * ここで切れば構造的に起きない。
  */
+const shift = (span, by) => ({ start: by + span.start, end: by + span.end });
 const sentencesOf = (prose, paragraphs, adapter) => paragraphs.flatMap((paragraph) => adapter
     .segment(prose.slice(paragraph.start, paragraph.end))
     .sentences.filter((sentence) => sentence.text.trim().length > 0)
     .map((sentence) => ({
-    span: { start: paragraph.start + sentence.span.start, end: paragraph.start + sentence.span.end },
+    span: shift(sentence.span, paragraph.start),
     text: sentence.text,
+    ...(sentence.tokens === undefined ? {} : { tokens: sentence.tokens.map((token) => ({ ...token, span: shift(token.span, paragraph.start) })) }),
 })));
 const sectionsOf = (headings, sentences, strongs, length) => {
     const bounds = headings.map((heading, index) => ({ heading, from: heading.end, to: headings[index + 1]?.start ?? length }));
@@ -121,17 +171,50 @@ const sectionsOf = (headings, sentences, strongs, length) => {
         };
     });
 };
+const paragraphsOf = (spans, sentences, listSpans) => spans
+    // 箇条書きの中の段落は「段落」として数えない。項目 1 つを 1 段落と読むと、
+    // 段落あたりの文数も長さのばらつきも、箇条書きの多い文書で壊れる。
+    .filter((span) => !listSpans.some((list) => span.start >= list.start && span.start < list.end))
+    .map((span) => ({ span, sentences: sentences.filter((sentence) => within(sentence.span, span.start, span.end)) }));
+/** 箇条書きは list ノードの直下の項目を数える。入れ子の項目は内側の list のものとして数える。 */
+const listsOf = (root, source) => {
+    const found = [];
+    walk(root, (node) => {
+        if (node.type !== "list")
+            return;
+        const span = spanOf(node);
+        if (span === undefined)
+            return;
+        const items = (node.children ?? []).flatMap((child) => {
+            const item = spanOf(child);
+            return child.type === "listItem" && item !== undefined ? [source.slice(item.start, item.end).replace(/\s+/gu, "").length] : [];
+        });
+        if (items.length > 0)
+            found.push({ span, items });
+    });
+    return found;
+};
 export const buildDocument = (path, source, adapter) => {
     const root = parse(source);
-    const prose = maskSpans(source, collectMasks(root));
-    const sentences = sentencesOf(prose, paragraphSpans(root), adapter);
+    // 強調の記号は「本文でないもの」だが、太字の数を数えるときの「覆われた場所」ではない。
+    // 同じ集合にすると、太字が自分の記号のせいで覆われた場所にあることになり、1 つも数えられなくなる。
+    const blocks = collectMasks(root, source);
+    const masked = [...blocks, ...emphasisSpans(root, source)];
+    const prose = maskSpans(source, masked);
+    const paragraphSpans = spansOfType(root, "paragraph");
+    const listItems = spansOfType(root, "listItem");
+    const sentences = sentencesOf(prose, paragraphSpans, adapter);
     return {
         path,
         source,
         language: adapter.id,
         lengthUnit: adapter.capabilities.lengthUnit,
-        sections: sectionsOf(headingsOf(root, source), sentences, strongSpans(root), source.length),
+        capabilities: adapter.capabilities,
+        sections: sectionsOf(headingsOf(root, source), sentences, strongSpans(root, blocks), source.length),
         sentences,
+        listSpans: listItems,
+        paragraphs: paragraphsOf(paragraphSpans, sentences, listItems),
+        lists: listsOf(root, source),
         lexicons: adapter.lexicons,
     };
 };