fix(files): align PDF export with Markdown editor

Author: Bill Leoutsakos

apps/sim/app/api/files/export/[id]/markdown-pdf.test.ts

@@ -2,12 +2,30 @@
  * @vitest-environment node
  */
 import { PDFDocument } from 'pdf-lib'
+import { getDocument, OPS } from 'pdfjs-dist/legacy/build/pdf.mjs'
 import sharp from 'sharp'
 import { describe, expect, it } from 'vitest'
-import { renderMarkdownPdf } from '@/app/api/files/export/[id]/markdown-pdf'
+import { MarkdownPdfLimitError, renderMarkdownPdf } from '@/app/api/files/export/[id]/markdown-pdf'
+
+async function pdfPagesText(buffer: Buffer): Promise<string[]> {
+  const document = await getDocument({ data: new Uint8Array(buffer), disableWorker: true }).promise
+  try {
+    return await Promise.all(
+      Array.from({ length: document.numPages }, async (_, index) => {
+        const page = await document.getPage(index + 1)
+        const content = await page.getTextContent()
+        return content.items.map((item) => ('str' in item ? item.str : '')).join(' ')
+      })
+    )
+  } finally {
+    await document.destroy()
+  }
+}
 
 describe('Markdown PDF rendering', () => {
   it('creates a valid multi-page PDF with GFM and an embedded image', async () => {
+    const imageKey = 'workspace/ws-1/editor-image.png'
+    const imageUrl = `/api/files/serve/${encodeURIComponent(imageKey)}?context=workspace`
     const image = await sharp({
       create: {
         width: 120,
@@ -26,7 +44,9 @@ describe('Markdown PDF rendering', () => {
 
 > A useful blockquote with a [link](https://sim.ai).
 
-Smart quotes “work”, Greek Ω stays readable, and unsupported emoji 🚀 falls back safely.
+Smart quotes “work” and Greek Ω stays readable.
+
+中文排版应该清晰易读。 العربية يجب أن تكون متصلة ومقروءة. हिन्दी पाठ स्पष्ट और पठनीय होना चाहिए। עברית צריכה להיות ברורה וקריאה.
 
 - First item
 - Second item
@@ -40,14 +60,16 @@ Smart quotes “work”, Greek Ω stays readable, and unsupported emoji 🚀 fal
 const exported = true
 \`\`\`
 
-![Embedded image](/workspace/ws-1/files/image-1)
+![Embedded image](${imageUrl})
+
+<img src="${imageUrl}" alt="Resized image" width="240" height="120">
 
 ${repeatedParagraphs}`
 
     const buffer = await renderMarkdownPdf({
       markdown,
       title: 'Export title',
-      images: new Map([['image-1', image]]),
+      images: new Map([[`key:${imageKey}`, image]]),
     })
 
     expect(buffer.subarray(0, 4).toString()).toBe('%PDF')
@@ -56,6 +78,46 @@ ${repeatedParagraphs}`
     const document = await PDFDocument.load(buffer)
     expect(document.getTitle()).toBe('Export title')
     expect(document.getPageCount()).toBeGreaterThan(1)
+
+    const text = (await pdfPagesText(buffer)).join(' ')
+    expect(text).toContain('中文排版应该清晰易读')
+    expect(text).toContain('العربية')
+    // PDF extractors expose visually positioned Indic vowel marks before their base character.
+    expect(text).toMatch(/[\u0900-\u097f]{4,}/u)
+    expect(text).toContain('עברית')
+    expect(text).not.toContain('Image: Embedded image')
+
+    const parsed = await getDocument({ data: new Uint8Array(buffer), disableWorker: true }).promise
+    try {
+      let imagePaints = 0
+      for (let pageNumber = 1; pageNumber <= parsed.numPages; pageNumber += 1) {
+        const operators = await (await parsed.getPage(pageNumber)).getOperatorList()
+        imagePaints += operators.fnArray.filter(
+          (operator) =>
+            operator === OPS.paintImageXObject || operator === OPS.paintInlineImageXObject
+        ).length
+      }
+      expect(imagePaints).toBeGreaterThanOrEqual(2)
+    } finally {
+      await parsed.destroy()
+    }
+  })
+
+  it('keeps long table rows together and repeats the header across table pages', async () => {
+    const rows = Array.from(
+      { length: 90 },
+      (_, index) =>
+        `| Row ${index + 1} | Description ${index + 1} with enough text to exercise wrapping |`
+    ).join('\n')
+    const buffer = await renderMarkdownPdf({
+      markdown: `# Table report\n\n| Name | Value |\n| --- | --- |\n${rows}`,
+      title: 'Table report',
+    })
+
+    const pages = await pdfPagesText(buffer)
+    const tablePages = pages.filter((page) => page.includes('Row '))
+    expect(tablePages.length).toBeGreaterThan(1)
+    expect(tablePages.every((page) => page.includes('Name') && page.includes('Value'))).toBe(true)
   })
 
   it('falls back instead of decoding an image above the pixel ceiling', async () => {
@@ -66,10 +128,19 @@ ${repeatedParagraphs}`
     const buffer = await renderMarkdownPdf({
       markdown: '![Too large](/api/files/view/image-1)',
       title: 'Bounded image',
-      images: new Map([['image-1', oversizedSvg]]),
+      images: new Map([['id:image-1', oversizedSvg]]),
     })
 
     expect(buffer.subarray(0, 4).toString()).toBe('%PDF')
     expect((await PDFDocument.load(buffer)).getPageCount()).toBe(1)
+    expect((await pdfPagesText(buffer)).join(' ')).toContain('Image: Too large')
+  })
+
+  it('rejects a pathological number of document blocks before PDF layout', async () => {
+    const markdown = Array.from({ length: 3_001 }, (_, index) => `Paragraph ${index}`).join('\n\n')
+
+    await expect(renderMarkdownPdf({ markdown, title: 'Too many blocks' })).rejects.toBeInstanceOf(
+      MarkdownPdfLimitError
+    )
   })
 })