chaffjs 0.17.0 → 0.18.0

package/src/detectors/long-vowel.ts

@@ -0,0 +1,66 @@
+import type { Detector, Finding, OptionValue, ProseDocument, Sentence } from "../plugin.ts";
+import { isKatakanaWord, oddLongVowels, stemOf, type Ending, type KanaWord } from "../long-vowel.ts";
+import { nameSpans, touchesAny } from "../team-names.ts";
+
+const ENDINGS: readonly Ending[] = ["consistent", "drop", "keep"];
+
+const endingOf = (value: OptionValue | undefined): Ending => ENDINGS.find((ending) => ending === value) ?? "consistent";
+
+const wordsOf = (value: OptionValue | undefined): readonly string[] =>
+  Array.isArray(value) ? value.filter((entry): entry is string => typeof entry === "string") : [];
+
+/** The minimum when no setting gives one: three morae, as the guidelines that drop the ー count it. */
+const DEFAULT_MIN_MORAE = 3;
+
+/** A listed word is exempt written either way: listing コンピューター exempts コンピュータ too. */
+const exemptStems = (except: readonly string[]): ReadonlySet<string> => new Set(except.map((word) => stemOf(word.trim())));
+
+/**
+ * The katakana nouns of a sentence, as the language adapter split them: ユーザー inside ユーザーインターフェース is its own word.
+ * Proper nouns (PROPN) and the team's names keep their own spelling (ディズニー).
+ */
+const kanaWordsOf = (sentence: Sentence, names: readonly string[], exempt: ReadonlySet<string>): KanaWord[] => {
+  const named = nameSpans(sentence.text, names);
+  return (sentence.tokens ?? [])
+    .filter((token) => token.pos === "NOUN" && isKatakanaWord(token.surface) && !exempt.has(stemOf(token.surface)))
+    .filter((token) => !touchesAny(named, token.span.start - sentence.span.start, token.span.end - sentence.span.start))
+    .map((token) => ({
+      surface: token.surface,
+      offset: token.span.start,
+      long: token.surface.endsWith("ー"),
+      dropped: token.features?.["LongVowelEnding"] === "Dropped",
+    }));
+};
+
+const quoteAt = (doc: ProseDocument, offset: number): string =>
+  doc.sentences.find((sentence) => sentence.span.start <= offset && offset < sentence.span.end)?.text.trim() ?? "";
+
+/**
+ * The final ー of katakana loanwords (コンピューター / コンピュータ). With no setting, it takes no side: it points at a word written
+ * both ways, and at the way the document writes less. Set to drop or keep, it points at every word the other way.
+ */
+export const katakanaLongVowel: Detector = (doc, options): Finding[] => {
+  const settings = options.settings ?? {};
+  const minMorae = typeof settings["min_morae"] === "number" ? settings["min_morae"] : DEFAULT_MIN_MORAE;
+  const exempt = exemptStems(wordsOf(settings["except"]));
+  const words = doc.sentences.flatMap((sentence) => kanaWordsOf(sentence, doc.names ?? [], exempt));
+  const odd = oddLongVowels(words, endingOf(settings["ending"]), minMorae);
+  if (odd.length === 0 || odd.length < options.limit) return [];
+  return odd.map((entry) => ({
+    rule: "",
+    severity: "warning",
+    line: 0,
+    column: 0,
+    quote: quoteAt(doc, entry.word.offset),
+    variant: entry.variant,
+    values: {
+      matched: entry.word.surface,
+      preferred: entry.preferred,
+      count: entry.count,
+      of: entry.of,
+      min_morae: minMorae,
+      limit: options.limit,
+      offset: entry.word.offset,
+    },
+  }));
+};

package/src/detectors/markup-finding.ts

@@ -0,0 +1,24 @@
+import type { Finding, Markup, ProseDocument, Span } from "../plugin.ts";
+
+// 記法の rule（見出し・画像・リンク・URL）が指摘を作る形。
+
+/** 引いて見せる記法の長さの上限。長い URL をまるごと見せても、どこが悪いかは伝わらない。 */
+const QUOTE_LIMIT = 80;
+
+export const quoteOf = (source: string, span: Span): string => {
+  const written = source.slice(span.start, span.end).replace(/\s+/gu, " ").trim();
+  return written.length > QUOTE_LIMIT ? `${written.slice(0, QUOTE_LIMIT)}…` : written;
+};
+
+export const findingAt = (doc: ProseDocument, span: Span, values: Readonly<Record<string, string | number>>, variant?: string): Finding => ({
+  rule: "",
+  severity: "warning",
+  line: 0,
+  column: 0,
+  quote: quoteOf(doc.source, span),
+  values: { ...values, offset: span.start },
+  ...(variant === undefined ? {} : { variant }),
+});
+
+/** Markdown の文書の記法。Markdown でなければ undefined。 */
+export const markupOf = (doc: ProseDocument): Markup | undefined => (doc.markup?.markdown === true ? doc.markup : undefined);

package/src/detectors/name-numeral.ts

@@ -1,14 +1,16 @@
+import { romanRangeOf } from "./roman-numeral.ts";
+
 /**
  * 大文字で始まる語のすぐ後ろの、I・V・X だけで書いたローマ数字（Engineer II、Leopold III、World War II）。名前に付く
  * 番号で、略語ではない。語の前に文字が無ければ（文の頭、箇条書きの番号の後ろ）、大文字は名前の印ではなく文頭の書き方なので
  * 外さない（Start IV fluids、1. Give IV fluids）。
- * CD、MD、DC、CI、MIX のように C・D・L・M を含む数字は略語であることが多いので、この位置でも略語のまま数える。
+ * 範囲（Reactions II-VI）も同じ。CD、MD、DC、CI、MIX のように C・D・L・M を含む数字は略語であることが多いので、この位置でも略語のまま数える。
  */
 
 type Span = { readonly start: number; readonly end: number };
 
 const SMALL_NUMERAL = "(?=[IVX])X{0,3}(?:IX|IV|V?I{0,3})";
-const NAMED_NUMERAL = new RegExp(String.raw`(?<![\p{L}\p{N}_])\p{Lu}\p{Ll}+\s+(?<numeral>${SMALL_NUMERAL})(?![\p{L}\p{N}_&])`, "gu");
+const NAMED_NUMERAL = new RegExp(String.raw`(?<![\p{L}\p{N}_])\p{Lu}\p{Ll}+\s+(?<numeral>${romanRangeOf(SMALL_NUMERAL)})(?![\p{L}\p{N}_&])`, "gu");
 const LETTER = /\p{L}/u;
 
 /** 文の中の、名前の後ろの番号の範囲（数字だけ）。 */

package/src/detectors/orthography.ts

@@ -1,8 +1,8 @@
 import type { Detector, Finding, ProseDocument, Sentence, Span } from "../plugin.ts";
-import { calendarRuns, calendarUnitsOf, type CalendarRun, type CalendarUnits } from "../calendar-number.ts";
+import { calendarRuns, calendarUnitsOf, type CalendarUnits } from "../calendar-number.ts";
 import { latinBoundaries, minorityStyle, occurrencesOutside, type Boundary, type SpacingKind } from "../orthography.ts";
 import { isWithinAny, quotedSpans } from "../quoted-span.ts";
-import { digitRunAround, isNumberName, sequenceLabelStarts } from "../number-name.ts";
+import { digitRunAround, endsWithDivisionLabel, isNumberName, sequenceLabelStarts, type NameContext } from "../number-name.ts";
 
 /** チームが chaff.yaml の prefer に書いた「使わない書き方」。書いていなければ何も言わない。 */
 export const preferredTerm: Detector = (doc, options): Finding[] => {
@@ -32,32 +32,36 @@ const digitBeside = (boundary: Boundary): number => {
   return boundary.spaced ? boundary.offset + 1 : boundary.offset;
 };
 
-/** 文書全体で一度だけ読むもの。sequence は続き番号の行の頭、topUnits は都道府県の単位、calendar は日付・時刻の単位。 */
-type NumberContext = { readonly sequence: ReadonlySet<number>; readonly topUnits: ReadonlySet<string>; readonly calendar: CalendarUnits };
+/**
+ * 文書全体で一度だけ読むもの（number-name.ts の NameContext に加えて）。calendar は日付・時刻の単位、
+ * divisions は「第1節」のように「第」と番号の後ろに書く区切りの語。
+ */
+type NumberContext = NameContext & { readonly calendar: CalendarUnits; readonly divisions: ReadonlySet<string> };
 
-/** 文の中の日付・時刻の数の、並びの頭の位置と、日付の中の位置。 */
-const calendarPlaces = (sentence: Sentence, units: CalendarUnits): ReadonlyMap<number, CalendarRun["place"]> =>
-  new Map(calendarRuns(sentence.text, sentence.tokens, sentence.span.start, units).map(({ run, place }) => [run.start, place]));
+/** 文の中の日付・時刻の数の、並びの頭の位置。 */
+const calendarStarts = (sentence: Sentence, units: CalendarUnits): ReadonlySet<number> =>
+  new Set(calendarRuns(sentence.text, sentence.tokens, sentence.span.start, units).map(({ run }) => run.start));
 
 /**
- * 日付・時刻の中の境目（「9月」の数と単位のあいだ、「2026年9月」の年と 9 のあいだ）は、詰めて書く決まりで好みではない。
- * 日付の前の境目（「は 9月」の空白）は、数の前をどう空けるかという書き手の書き方なので数える。
+ * 番号・識別子として書かれた数（number-name.ts）、日付・時刻の数、「第1節」の後ろの境目は、空け方の好みではないので数えない。
+ * 日付・時刻は前の境目（「は 9月」「午後3時」「令和 3 年」）も数えない。日付はまとめて一つの書き方で、数量の空け方の票にはしない。
  */
-const isInsideCalendar = (boundary: Boundary, place: CalendarRun["place"] | undefined): boolean =>
-  place !== undefined && (boundary.kind === "after-digit" || place === "inner");
-
-/** 番号・識別子として書かれた数（number-name.ts）と日付・時刻の中の境目は、空け方の好みではないので数えない。 */
-const isCounted = (sentence: Sentence, boundary: Boundary, context: NumberContext, calendar: ReadonlyMap<number, CalendarRun["place"]>): boolean => {
+const isCounted = (sentence: Sentence, boundary: Boundary, context: NumberContext, calendar: ReadonlySet<number>): boolean => {
+  if (boundary.kind !== "after-digit" && endsWithDivisionLabel(sentence.text.slice(0, boundary.offset), context.divisions)) return false;
   if (boundary.kind === "letter") return true;
   const run = digitRunAround(sentence.text, digitBeside(boundary));
   if (run === undefined) return true;
-  if (isInsideCalendar(boundary, calendar.get(run.start))) return false;
-  return !isNumberName(sentence.text, run, sentence.tokens, sentence.span.start, context.sequence, context.topUnits);
+  if (calendar.has(run.start)) return false;
+  return !isNumberName(sentence.text, run, sentence.tokens, sentence.span.start, context);
 };
 
 const patternList = (doc: ProseDocument, id: string): string[] => (doc.lexicons[id] ?? []).map((entry) => entry.pattern);
 const patternsOf = (doc: ProseDocument, id: string): ReadonlySet<string> => new Set(patternList(doc, id));
 
+/** 語彙表 numbered-label のうち、番号の position 側に書く語。 */
+const labelsAt = (doc: ProseDocument, position: "before" | "after"): ReadonlySet<string> =>
+  new Set((doc.lexicons["numbered-label"] ?? []).filter((entry) => entry.position === position).map((entry) => entry.pattern));
+
 /**
  * 鉤括弧で引いた題名や発言（「…ガイドライン ver. 1.1」）の中の境目。空け方は引いた元のもので、書き手の書き方ではない。
  * 括弧は日本語の字でも英数字でもないので、境目はまるごと括弧の中か外にある。
@@ -73,11 +77,13 @@ export const latinSpacing: Detector = (doc, options): Finding[] => {
   const context: NumberContext = {
     sequence: sequenceLabelStarts(doc.prose ?? doc.source),
     topUnits: patternsOf(doc, "prefecture-unit"),
+    labelWords: labelsAt(doc, "before"),
     calendar: calendarUnitsOf(doc),
+    divisions: labelsAt(doc, "after"),
   };
   const located: Located[] = doc.sentences.flatMap((sentence) => {
     const quoted = quotedSpans(sentence.text);
-    const calendar = calendarPlaces(sentence, context.calendar);
+    const calendar = calendarStarts(sentence, context.calendar);
     return latinBoundaries(sentence.text, doc.source.slice(sentence.span.start, sentence.span.end))
       .filter((boundary) => !isQuoted(quoted, boundary) && isCounted(sentence, boundary, context, calendar))
       .map((boundary) => ({ sentence, ...boundary }));