✅(e2e) add test to compare generated PDF against reference template

CHANGELOG.md

@@ -10,6 +10,7 @@ and this project adheres to
 
 - ✨(export) enable ODT export for documents #1524
 - ✨(frontend) improve mobile UX by showing subdocs count #1540
+- ✅(e2e) add test to compare generated PDF against reference template #1648
 
 ### Fixed
 

src/frontend/apps/e2e/__tests__/app-impress/doc-export.spec.ts

@@ -1,3 +1,4 @@
+import fs from 'fs';
 import path from 'path';
 
 import { expect, test } from '@playwright/test';
@@ -13,6 +14,35 @@ import {
 import { openSuggestionMenu, writeInEditor } from './utils-editor';
 import { createRootSubPage } from './utils-sub-pages';
 
+const REGRESSION_FIXTURE_CONTENT = fs.readFileSync(
+  path.join(__dirname, 'assets/content.txt'),
+  'utf-8',
+);
+const REGRESSION_SNAPSHOT_NAME = 'doc-export-regression.pdf';
+const REGRESSION_DOC_TITLE = 'doc-export-regression-reference';
+
+/**
+ * Playwright snapshots store the raw PDF bytes. However, each export embeds
+ * dynamic metadata (timestamps, font-subset identifiers, etc.) that would make
+ * the snapshot differ at every run. To ensure deterministic comparisons we
+ * strip/neutralize those fields before matching against the reference PDF.
+ */
+const sanitizePdfBuffer = (buffer: Buffer) => {
+  const pdfText = buffer.toString('latin1');
+  const neutralized = pdfText
+    // Remove per-export timestamps
+    .replace(/\/CreationDate\s*\(.*?\)/g, '/CreationDate ()')
+    .replace(/\/ModDate\s*\(.*?\)/g, '/ModDate ()')
+    // Remove file identifiers
+    .replace(/\/ID\s*\[<[^>]+>\s*<[^>]+>\]/g, '/ID [<0><0>]')
+    .replace(/D:\d{14}Z/g, 'D:00000000000000Z')
+    // Remove subset font prefixes generated by PDF renderer
+    .replace(/\b[A-Z]{6}\+(Inter18pt-[A-Za-z]+)\b/g, 'STATIC+$1')
+    .replace(/\b[A-Z]{6}\+(GeistMono-[A-Za-z]+)\b/g, 'STATIC+$1');
+
+  return Buffer.from(neutralized, 'latin1');
+};
+
 test.beforeEach(async ({ page }) => {
   await page.goto('/');
 });
@@ -551,4 +581,118 @@ test.describe('Doc Export', () => {
     const download = await downloadPromise;
     expect(download.suggestedFilename()).toBe(`${docChild}.odt`);
   });
+
+  /**
+   * Regression guard for the full PDF export pipeline.
+   *
+   * Usage reminder:
+   * 1. `npx playwright test __tests__/app-impress/doc-export.spec.ts --update-snapshots -g "full document" --project=chromium`
+   *    -> refresh the reference PDF whenever we intentionally change the export output.
+   * 2. `npx playwright test __tests__/app-impress/doc-export.spec.ts -g "full document" --project=chromium`
+   *    -> CI (and local runs without --update-snapshots) will compare the PDF to the reference
+   *       and fail on any byte-level difference once the dynamic metadata has been sanitized.
+   */
+  test('it keeps the full document PDF export identical to the reference snapshot', async ({
+    page,
+    browserName,
+  }, testInfo) => {
+    // PDF generation for a large, image-heavy document can be slow in CI.
+    // Give this regression test a higher timeout budget than the default.
+    testInfo.setTimeout(120000);
+    const snapshotPath = testInfo.snapshotPath(REGRESSION_SNAPSHOT_NAME);
+
+    test.skip(
+      !fs.existsSync(snapshotPath) &&
+        testInfo.config.updateSnapshots === 'none',
+      `Missing PDF snapshot at ${snapshotPath}. Run Playwright with --update-snapshots to record it.`,
+    );
+
+    // We must use a deterministic title so that block content (and thus the
+    // exported PDF) stays identical between runs.
+    await createDoc(page, 'doc-export-regression', browserName, 1);
+    const titleInput = page.getByRole('textbox', { name: 'Document title' });
+    await expect(titleInput).toBeVisible();
+    await titleInput.fill(REGRESSION_DOC_TITLE);
+    await titleInput.blur();
+    await verifyDocName(page, REGRESSION_DOC_TITLE);
+    const regressionDoc = REGRESSION_DOC_TITLE;
+
+    const docId = page
+      .url()
+      .split('/docs/')[1]
+      ?.split('/')
+      .filter(Boolean)
+      .shift();
+
+    expect(docId).toBeTruthy();
+
+    // Inject the pre-crafted blocknote document via the REST API to avoid
+    // rebuilding it through the UI (which would be slow and flaky).
+    const cookies = await page.context().cookies();
+    const csrfToken = cookies.find(
+      (cookie) => cookie.name === 'csrftoken',
+    )?.value;
+
+    const headers: Record<string, string> = {
+      'content-type': 'application/json',
+    };
+
+    if (csrfToken) {
+      headers['X-CSRFToken'] = csrfToken;
+    }
+
+    const updateResponse = await page.request.patch(
+      `http://localhost:8071/api/v1.0/documents/${docId}/`,
+      {
+        headers,
+        data: {
+          content: REGRESSION_FIXTURE_CONTENT,
+          websocket: true,
+        },
+      },
+    );
+
+    if (!updateResponse.ok()) {
+      throw new Error(
+        `Failed to seed document content. Status: ${updateResponse.status()}, body: ${await updateResponse.text()}`,
+      );
+    }
+
+    await page.reload();
+    // After reloading, just ensure the editor container is present before exporting.
+    await expect(page.locator('.--docs--editor-container')).toBeVisible({
+      timeout: 15000,
+    });
+
+    await page
+      .getByRole('button', {
+        name: 'Export the document',
+      })
+      .click();
+
+    await expect(
+      page.getByTestId('doc-open-modal-download-button'),
+    ).toBeVisible();
+
+    await expect(page.getByTestId('doc-export-download-button')).toBeEnabled({
+      timeout: 15000,
+    });
+
+    // Export to PDF and confirm the generated bytes match the reference file.
+    const downloadPromise = page.waitForEvent('download', {
+      timeout: 60000,
+      predicate: (download) =>
+        download.suggestedFilename().includes(`${regressionDoc}.pdf`),
+    });
+
+    void page.getByTestId('doc-export-download-button').click();
+
+    const download = await downloadPromise;
+    expect(download.suggestedFilename()).toBe(`${regressionDoc}.pdf`);
+
+    const pdfBuffer = await cs.toBuffer(await download.createReadStream());
+    const normalizedPdfBuffer = sanitizePdfBuffer(pdfBuffer);
+
+    expect(normalizedPdfBuffer).toMatchSnapshot(REGRESSION_SNAPSHOT_NAME);
+  });
 });

src/frontend/apps/e2e/print-datauris.js

@@ -0,0 +1,29 @@
+//utilitary script to print the datauris of the targeted assets
+
+const fs = require('fs');
+const path = require('path');
+
+const ASSETS_ROOT = path.resolve(__dirname, '__tests__/app-impress/assets');
+
+function saveDataUrl(file, mime, outName) {
+    const abs = path.join(ASSETS_ROOT, file);
+    const base64 = fs.readFileSync(abs).toString('base64');
+    const dataUrl = `data:${mime};base64,${base64}`;
+    const outPath = path.join(ASSETS_ROOT, outName);
+    fs.writeFileSync(outPath, dataUrl, 'utf8');
+    console.log(`Wrote ${outName}`);
+}
+
+// PNG
+saveDataUrl('panopng.png', 'image/png', 'pano-png-dataurl.txt');
+
+// JPG
+saveDataUrl('panojpg.jpeg', 'image/jpeg', 'pano-jpg-dataurl.txt');
+
+// SVG
+const svgPath = path.join(ASSETS_ROOT, 'test.svg');
+const svgText = fs.readFileSync(svgPath, 'utf8');
+const svgDataUrl =
+    'data:image/svg+xml;base64,' + Buffer.from(svgText).toString('base64');
+fs.writeFileSync(path.join(ASSETS_ROOT, 'test-svg-dataurl.txt'), svgDataUrl, 'utf8');
+console.log('Wrote test-svg-dataurl.txt');