cinny/src/app/utils/looksLikeCode.ts

/**
 * [Gitea #107] Decide whether a multi-line paste looks like source code, so the
 * composer can OFFER to turn it into a code block (never converts on its own).
 * Pure and unit-tested; tuned to keep prose, lists, quotes and URL lists out.
 */

export type CodeGuess = {
  /** Offer "Paste as code block?" */
  isCode: boolean;
  /** Language for `class="language-…"`, only when fairly sure. */
  lang?: string;
};

const KEYWORDS = [
  /\bfunction\b/,
  /\bconst\b/,
  /\blet\b/,
  /\bvar\b/,
  /\breturn\b/,
  /\bimport\b/,
  /\bexport\b/,
  /\bclass\b/,
  /\bdef\b/,
  /\bfn\b/,
  /\bpub\b/,
  /\bstruct\b/,
  /\bif \(/,
  /\bfor \(/,
  /\bwhile \(/,
  /#include\b/,
  /\bSELECT\b/i,
  /\bFROM\b/i,
  /<\?php/,
  /<\//,
];

const MONO_HTML = /font-family:[^;"]*(mono|courier|consolas|menlo)|<pre[\s>]|<code[\s>]/i;

const indentOf = (line: string): number => {
  const m = /^[ \t]*/.exec(line);
  return m ? m[0].replace(/\t/g, '    ').length : 0;
};

export function guessLanguage(text: string): string | undefined {
  if (
    /^\s*(SELECT|INSERT|UPDATE|DELETE|CREATE TABLE)\b/im.test(text) &&
    /\bFROM\b|\bINTO\b|\bSET\b|\(/i.test(text)
  )
    return 'sql';
  if (/\bpub fn\b|\bfn \w+\(.*\)\s*(->|\{)|\blet mut\b/.test(text)) return 'rust';
  if (/^\s*#include\s*[<"]/m.test(text)) return 'c';
  if (
    /^\s*def \w+\(.*\):\s*$/m.test(text) ||
    (/^\s*(from \w+ )?import \w+\s*$/m.test(text) && /:\s*$/m.test(text))
  )
    return 'python';
  if (/:\s*(string|number|boolean)\b|\binterface \w+|\btype \w+ =/.test(text) && /[{;]/.test(text))
    return 'ts';
  if (/\b(const|let|var)\b|=>/.test(text) && /[{;]/.test(text)) return 'js';
  if (/^\s*<\?php/m.test(text)) return 'php';
  return undefined;
}

export function looksLikeCode(text: string, html?: string): CodeGuess {
  const lines = text.replace(/\r\n?/g, '\n').split('\n');
  const nonEmpty = lines.filter((l) => l.trim() !== '');
  if (nonEmpty.length < 3) return { isCode: false };
  const n = nonEmpty.length;
  const share = (pred: (l: string) => boolean) => nonEmpty.filter(pred).length / n;

  // ── Prose guards ──
  if (share((l) => /^\s*>/.test(l)) > 0.5) return { isCode: false }; // quoted reply
  if (nonEmpty.every((l) => /^\s*(<)?https?:\/\/\S+(>)?\s*$/.test(l))) return { isCode: false };
  const avgLen = nonEmpty.reduce((sum, l) => sum + l.length, 0) / n;
  if (avgLen > 90 && share((l) => (l.match(/[.!?] /g) ?? []).length >= 2) > 0.5) {
    return { isCode: false };
  }
  if (share((l) => /^[\p{L}\p{N}\s'’",.!?-]*$/u.test(l)) > 0.6) return { isCode: false };
  if (share((l) => /^\s*([-*+]|\d+[.)])\s+\S/.test(l)) > 0.6) return { isCode: false }; // list

  // ── Signals: need two ──
  let signals = 0;
  const indented = share((l) => /^( {2,}|\t)/.test(l));
  let depthChanges = 0;
  for (let i = 1; i < nonEmpty.length; i += 1) {
    if (indentOf(nonEmpty[i]) !== indentOf(nonEmpty[i - 1])) depthChanges += 1;
  }
  if (indented >= 0.4 && depthChanges >= 2) signals += 1;
  if (share((l) => /[;{}):,]\s*$/.test(l)) >= 0.3) signals += 1;
  if (
    nonEmpty.filter((l) => /=>|->|::|\(\)|[^=!<>]=[^=]/.test(l)).length >= 2 &&
    /[{}[\]]/.test(text)
  ) {
    signals += 1;
  }
  if (KEYWORDS.filter((k) => k.test(text)).length >= 2) signals += 1;
  if (html && MONO_HTML.test(html)) signals += 1;
  // SQL has few symbols: lines that open with a clause keyword are its shape.
  if (
    share((l) =>
      /^\s*(SELECT|FROM|WHERE|(LEFT |RIGHT |INNER )?JOIN|ORDER BY|GROUP BY|HAVING|INSERT|VALUES|UPDATE|SET|DELETE|LIMIT|AND|OR)\b/.test(
        l,
      ),
    ) >= 0.5
  ) {
    signals += 1;
  }

  if (signals < 2) return { isCode: false };
  return { isCode: true, lang: guessLanguage(text) };
}