zitadel/apps/docs/scripts/check-links.ts

import { register } from 'node:module';
import { pathToFileURL } from 'node:url';
import { join, dirname, resolve } from 'node:path';
import { writeFileSync, existsSync, writeSync } from 'node:fs';
import { readFile, stat } from 'node:fs/promises';

register('./scripts/loader.mjs', pathToFileURL('./'));
register('fumadocs-mdx/node/loader', import.meta.url);

import {
  type FileObject,
  printErrors,
  validateFiles,
} from 'next-validate-link';

try {
  // Dynamic import to ensure loader is registered before importing source
  const { source, versionSource } = await import('../lib/source');

  const PUBLIC_ROOT = resolve('public');

  // Simple p-limit implementation to avoid adding dependency
  function pLimit(concurrency: number) {
    const queue: (() => void)[] = [];
    let active = 0;

    const next = () => {
      active--;
      if (queue.length > 0) {
        queue.shift()!();
      }
    };

    const run = async <T,>(fn: () => Promise<T>): Promise<T> => {
      if (active >= concurrency) {
        await new Promise<void>((resolve) => queue.push(resolve));
      }
      active++;
      try {
        return await fn();
      } finally {
        next();
      }
    };

    return run;
  }

  async function checkLinks() {
    const pages = [...source.getPages(), ...versionSource.getPages()];
    console.log(`Total pages found: ${pages.length}`);

    const allFiles = await getFiles();
    const filesToCheck = allFiles.filter((f) => !f.isVersioned);
    const allFilesAsTargets = allFiles;

    // Load redirects and add them to the scanned URLs
    let redirects: any[] = [];
    if (existsSync('redirects.json')) {
      redirects = JSON.parse(await readFile('redirects.json', 'utf-8'));
    }

    const scanned = {
      urls: new Map<string, { hashes: string[] }>([
        // Add /docs specifically as it's the home page and might be linked to
        ['/docs', { hashes: [] }],
        // Load redirects first
        ...redirects.map((r: any): [string, { hashes: string[] }] => [r.source, { hashes: [] }]),
        // Load files second, so they overwrite redirects with actual headings
        ...allFilesAsTargets.map((f): [string, { hashes: string[] }] => [f.url!, { hashes: getHeadings(f) }]),
      ]),
      fallbackUrls: [],
    };

    writeFileSync('scanned-urls.json', JSON.stringify(Array.from(scanned.urls.keys()), null, 2));

    console.log(`Manually populated URLs count: ${scanned.urls.size}`);

    console.time('validateFiles');
    let linkErrors: any[] = [];
    let currentFile = '';

    process.on('uncaughtException', (err) => {
      writeSync(1, `\n\n!!! CRITICAL ERROR in ${currentFile} !!!\n`);
      console.error(err);
      process.exit(1);
    });

    for (const file of filesToCheck) {
      currentFile = file.path;
      writeSync(1, `Checking: ${file.path}\n`);
      try {
        const errors = await validateFiles([file], {
          scanned,
          baseUrl: '/docs',
          markdown: {
            components: {
              Card: { attributes: ['href'] },
            },
          },
          checkRelativePaths: 'as-url',
        });
        linkErrors.push(...errors);
      } catch (err: any) {
        console.error(`\nFAILED file: ${file.path}`);
        console.error('Error Details:', err);
        if (file.content) {
          console.error('First 15 lines of content:');
          console.error(file.content.split('\n').slice(0, 15).join('\n'));
        }
        throw err;
      }
    }
    console.timeEnd('validateFiles');

    printErrors(linkErrors, false);

    console.time('checkImages');
    const imageErrors = await checkImages(filesToCheck);
    console.timeEnd('checkImages');

    // Check for .md files
    const mdFiles = filesToCheck.filter(f => f.path.endsWith('.md'));
    if (mdFiles.length > 0) {
      console.error('\n❌ Error: Found .md files. All content files must use .mdx extension:');
      mdFiles.forEach(f => console.error(` - ${f.path}`));
      process.exit(1);
    }

    if (linkErrors.length === 0 && !imageErrors) {
      console.log('\n✅ All checks passed: No broken links or images found.');
    }

    if (linkErrors.length > 0 || imageErrors) {
      process.exit(1);
    }
  }

  async function checkImages(files: (FileObject & { isVersioned?: boolean })[]): Promise<boolean> {
    let hasErrors = false;
    const limit = pLimit(50); // Limit to 50 concurrent file checks

    // Flat list of all image checks
    const allChecks: Promise<void>[] = [];

    for (const file of files) {
      allChecks.push(limit(async () => {
        const content = file.content;
        const imageRegex = /!\[.*?\]\((.*?)\)|<img.*?src=[{"'](.*?)["'}]/g;
        const matches = Array.from(content.matchAll(imageRegex));

        for (const match of matches) {
          const isMarkdown = !!match[1];
          let imagePath = match[1] || match[2];

          if (match[1]) {
            imagePath = imagePath.trim().split(/\s+/)[0];
          }

          if (imagePath.startsWith('http') || imagePath.startsWith('https') || imagePath.startsWith('data:')) {
            continue;
          }

          imagePath = imagePath.split('?')[0].split('#')[0];
          imagePath = decodeURIComponent(imagePath);

          // HTML Image Rules
          if (!isMarkdown) {
            if (imagePath.includes('public/')) {
              console.error(`Broken HTML image link in ${file.path}: ${imagePath} (HTML tags cannot reference 'public/' directly, use absolute path '/docs/img/...')`);
              hasErrors = true;
              continue;
            }
            if (!imagePath.startsWith('/docs/')) {
              console.error(`Broken HTML image link in ${file.path}: ${imagePath} (HTML tags must use absolute path starting with '/docs/, e.g. '/docs/img/...')`);
              hasErrors = true;
              continue;
            }
            // Check existence for absolute /docs/ path
            const relativeToPublic = imagePath.slice(5); // Remove /docs
            const fullPath = join(PUBLIC_ROOT, relativeToPublic);
            try {
              await stat(fullPath);
            } catch {
              console.error(`Broken image link in ${file.path}: ${imagePath} (File not found at ${fullPath})`);
              hasErrors = true;
            }
            continue;
          }

          // Markdown Image Rules
          let fullPath;
          if (imagePath.startsWith('/')) {
            // Markdown absolute paths are relative to public (usually) or not supported by webpack import if not aliased
            // We'll assume they map to public if they start with slash, but fumadocs might prefer relative.
            // For now, let's just check existence relative to public if absolute.
            // BUT, if they start with /docs/, that's likely wrong for Markdown import if it expects file path.
            if (imagePath.startsWith('/docs/')) {
              // Be lenient or strict? User had issues with imports.
              // Let's rely on standard resolution.
              const relativeToPublic = imagePath.slice(5);
              fullPath = join(PUBLIC_ROOT, relativeToPublic);
            } else {
              fullPath = join(PUBLIC_ROOT, imagePath.startsWith('/') ? imagePath.slice(1) : imagePath);
            }
          } else {
            fullPath = resolve(dirname(file.path), imagePath);
          }

          try {
            await stat(fullPath);
          } catch {
            console.error(`Broken image link in ${file.path}: ${imagePath}`);
            hasErrors = true;
          }
        }
      }));
    }

    await Promise.all(allChecks);
    return hasErrors;
  }

  function getHeadings({ data, content }: any): string[] {
    const headings = new Set<string>();

    if (data.toc && data.toc.length > 0) {
      data.toc.forEach((item: any) => headings.add(item.url.slice(1)));
    }
    if (data.structuredData?.headings) {
      data.structuredData.headings.forEach((h: any) => headings.add(h.id));
    }

    // Fallback: parse content directly
    if (content) {
      const lines = content.split('\n');
      let inCodeBlock = false;
      for (const line of lines) {
        if (line.trim().startsWith('```')) {
          inCodeBlock = !inCodeBlock;
          continue;
        }
        if (inCodeBlock) continue;

        const match = line.match(/^(#{1,6})\s+(.+)$/);
        if (match) {
          const title = match[2].trim();
          const slug = title
            .toLowerCase()
            .replace(/[^\w\s-]/g, '')
            .replace(/_/g, '-')
            .replace(/\s+/g, '-')
            .replace(/-+/g, '-');
          headings.add(slug);
        }
      }
    }

    return Array.from(headings);
  }

  async function getFiles() {
    const pages = [...source.getPages(), ...versionSource.getPages()];
    const promises = pages.map(
      async (page: any): Promise<FileObject & { isVersioned: boolean }> => {
        const path = page.file?.path || page.absolutePath;
        return {
          path,
          content: await readFile(path, 'utf8'),
          // REMOVED /docs prefix here to ensure strict link checking against actual routes
          url: page.url.startsWith('/') ? page.url : '/' + page.url,
          data: page.data,
          isVersioned: path.includes('/v4.') || page.url.startsWith('v4.'),
        };
      },
    );
    return Promise.all(promises);
  }

  await checkLinks();
} catch (err) {
  writeSync(1, `\n\n!!! TOP-LEVEL ERROR !!!\n`);
  console.error(err);
  process.exit(1);
}