chaffjs 0.28.0 → 0.29.0

package/src/warning-action.ts

@@ -0,0 +1,73 @@
+// A warning (Warning:, 注意：, > [!WARNING]) that names a risk and gives the reader nothing to do. Pure: reads a paragraph's
+// text and its sentences; the labels and the words of an action come from the language's warning-label and warning-action
+// lexicons.
+import { escapeRegExp } from "./orthography.ts";
+
+export type WarningWords = {
+  /** What a warning opens with: Warning, Caution, 注意, 警告. */
+  readonly labels: readonly string[];
+  /**
+   * What makes a sentence an action. A Latin entry is matched as whole words anywhere (do not, make sure, must); another is
+   * matched at the end of the sentence (ください, ましょう), where Japanese puts what the reader is to do.
+   */
+  readonly actions: readonly string[];
+};
+
+/** What may come before the label: a quote mark, bold, a GitHub alert's [! (> [!WARNING], **Warning:**). */
+const BEFORE_LABEL = /^(?:>[ \t]*)*(?:\*\*|__)?(?:\[!)?/u;
+/** What closes the label: the end of bold or of an alert, then a colon or an exclamation mark, or the end of the line. */
+const LABEL_CLOSE = /^(?:\*\*|__)?\]?[ \t]*/u;
+const LABEL_MARK = /^[:：!！](?:\*\*|__)?[ \t]*/u;
+const SENTENCE_END_MARKS = new Set(" \t\r\n。．.!！?？」』)）");
+const LATIN_START = /^[A-Za-z]/u;
+const LETTER = /\p{L}/u;
+
+/**
+ * Where the text after a paragraph's warning label starts, or undefined when the paragraph opens with no label. The label
+ * must stand alone: "Warning:", "注意：", "**Caution**", "> [!WARNING]", not "Warnings are logged" or 「注意深く」.
+ */
+export const warningTextStart = (paragraph: string, labels: readonly string[]): number | undefined => {
+  const lead = BEFORE_LABEL.exec(paragraph)?.[0].length ?? 0;
+  const rest = paragraph.slice(lead);
+  const label = labels.find((word) => rest.toLowerCase().startsWith(word.toLowerCase()));
+  if (label === undefined) return undefined;
+  const after = rest.slice(label.length);
+  const close = LABEL_CLOSE.exec(after)?.[0].length ?? 0;
+  const mark = LABEL_MARK.exec(after.slice(close))?.[0].length;
+  if (mark !== undefined) return lead + label.length + close + mark;
+  const next = after.slice(close);
+  return next === "" || next.startsWith("\n") || next.startsWith("\r\n") ? lead + label.length + close : undefined;
+};
+
+const SPACES = /\s+/gu;
+
+/** A word or phrase as whole words, with any run of spaces or a line break between its words ("Do\nnot"). */
+const hasWholeWord = (text: string, word: string): boolean => {
+  const lowered = text.replaceAll(SPACES, " ").toLowerCase();
+  const target = word.toLowerCase();
+  return [...lowered.matchAll(new RegExp(escapeRegExp(target), "gu"))].some(
+    (match) => !LETTER.test(lowered[match.index - 1] ?? "") && !LETTER.test(lowered[match.index + target.length] ?? ""),
+  );
+};
+
+/** Whether a sentence tells the reader what to do, by the words of the warning-action lexicon. */
+export const namesAction = (sentence: string, actions: readonly string[]): boolean => {
+  const chars = [...sentence];
+  const ending = chars.slice(0, chars.findLastIndex((char) => !SENTENCE_END_MARKS.has(char)) + 1).join("");
+  return actions.some((word) => (LATIN_START.test(word) ? hasWholeWord(sentence, word) : ending.endsWith(word)));
+};
+
+const QUOTED = /^[ \t]*>/u;
+
+/**
+ * Where the block quote holding an offset ends, or undefined when its line is not quoted. A GitHub alert (> [!WARNING])
+ * runs over every quoted line after it, blank quoted lines (>) included, so its action may be in a later paragraph.
+ */
+export const quoteBlockEnd = (source: string, offset: number): number | undefined => {
+  const lineStart = source.lastIndexOf("\n", offset - 1) + 1;
+  const lines = source.slice(lineStart).split("\n");
+  const quoted = lines.findIndex((line) => !QUOTED.test(line));
+  const count = quoted < 0 ? lines.length : quoted;
+  if (count === 0) return undefined;
+  return lineStart + lines.slice(0, count).join("\n").length;
+};

package/styles/koyobun.yaml

@@ -5,6 +5,8 @@
 #   意識するとよい」。chaff の上限は「ここを超えたら見直す」所なので、その上の端の 60 字を和文の上限にする。英文には言っていないので、英文の上限は変えない。
 # Ⅲ－1 イ「一つの文書内では、常体と敬体のどちらかで統一する」: no-mixed-desumasu。どちらにするかは文書の相手で決まる（Ⅲ－1 ア）ので、決めない。
 # Ⅰ－3 エ「長音は、原則として長音符号を使って書く」: 英語の語末の -er、-or、-ar などは長音符号を付ける（コンピューター）。katakana-long-vowel の keep。
+# Ⅰ－1（3）ア「常用漢字表に使える漢字があっても仮名で書く場合」: 補助動詞（〜ていただく、〜てくださる）と形式名詞（こと、とき、ため、よう）。
+#   「公用文における漢字使用等について」（平成 22 年内閣訓令第 1 号）1 (2) キ の〜ておく、〜てくださいも同じ。kana-function-word。
 id: koyobun
 
 name:
@@ -12,19 +14,20 @@ name:
   en: 公用文作成の考え方 (Council for Cultural Affairs, 2022)
 
 summary:
-  ja: 一文が 60 字を超えたら見直します。一つの文書の中で「です・ます」と「である」を混ぜません。カタカナ語の語末の「ー」は付けます（コンピューター）。
-  en: Flags a Japanese sentence over 60 characters, a document that mixes です・ます with である, and a katakana word that drops its final ー (コンピューター).
+  ja: 一文が 60 字を超えたら見直します。一つの文書の中で「です・ます」と「である」を混ぜません。カタカナ語の語末の「ー」は付けます（コンピューター）。補助動詞と形式名詞はかなで書きます（〜てください、〜すること）。
+  en: Flags a Japanese sentence over 60 characters, a document that mixes です・ます with である, a katakana word that drops its final ー (コンピューター), and a helper verb or formal noun written in kanji where official style writes kana (〜てください, 〜すること).
 
 source:
   title:
-    ja: 文化審議会「公用文作成の考え方」（令和 4 年 1 月 7 日 建議）Ⅲ－3 ア、Ⅲ－1 イ、Ⅰ－3 エ
-    en: 公用文作成の考え方 (recommendation of the Council for Cultural Affairs, 7 January 2022), III-3 a, III-1 b and I-3 d
+    ja: 文化審議会「公用文作成の考え方」（令和 4 年 1 月 7 日 建議）Ⅲ－3 ア、Ⅲ－1 イ、Ⅰ－3 エ、Ⅰ－1（3）ア
+    en: 公用文作成の考え方 (recommendation of the Council for Cultural Affairs, 7 January 2022), III-3 a, III-1 b, I-3 d and I-1 (3) a
   url: https://www.bunka.go.jp/seisaku/bunkashingikai/kokugo/hokoku/pdf/93651301_01.pdf
 
 rules:
   max-sentence-length: normal
   no-mixed-desumasu: normal
   katakana-long-vowel: normal
+  kana-function-word: normal
 
 limits:
   max-sentence-length: