Error running in node.js environment

[amserra]

Hey!
I'm trying to use pdf.js in a node.js environment with Tanstack Start + Nitro to read a pdf content. I have installed pdfjs-dist version 5.4.530.

Here's my file

import { getDocument } from 'pdfjs-dist/legacy/build/pdf.mjs';

export async function parsePdf(data: ArrayBuffer) {
  try {
    const loadingTask = getDocument({ data });

    const pdf = await loadingTask.promise;
    let fullText = '';

    for (let pageNum = 1; pageNum <= pdf.numPages; pageNum++) {
      const page = await pdf.getPage(pageNum);
      const textContent = await page.getTextContent();
      const pageText = textContent.items
        .map(item => ('str' in item ? item.str : '').trim())
        .filter(Boolean)
        .join(' ');
      fullText += pageText + '\n\n';
    }

    return fullText.trim();
  } catch (e) {
    console.error('Failed to parse PDF:', e);
    throw e;
  }
}

Whenever I run this function, at runtime, I get the following error.

Warning: Cannot load "@napi-rs/canvas" package: "Error: Cannot find module '@napi-rs/canvas'
Require stack:
- /Users/amserra/Documents/Repos/jurisfacil/apps/web/.output/server/chunks/_/add-file-DKqFpuv9.mjs".
Warning: Cannot polyfill `DOMMatrix`, rendering may be broken.
Warning: Cannot polyfill `ImageData`, rendering may be broken.
Warning: Cannot polyfill `Path2D`, rendering may be broken.
ReferenceError: DOMMatrix is not defined

It is my belief that in node.js environments the @napi-rs/canvas would not be required, right? But it is still being called. Anyone knows how to fix this?

[AKzar1el]

You're probably hitting a Node-side dependency/runtime issue rather than anything wrong with your text-extraction loop.

The giveaway is the exact chain:

• Cannot load @napi-rs/canvas
• cannot polyfill DOMMatrix / ImageData / Path2D
• ReferenceError: DOMMatrix is not defined

That usually means the Node legacy build is still trying to initialize its Node polyfill/canvas path, so even if you're only calling getTextContent(), the import/runtime can still trip over missing canvas-related pieces.

What I'd try first:

1. Install @napi-rs/canvas in the package that actually executes the Nitro server bundle (not only at a workspace root).
2. Make sure Nitro isn't bundling/stripping it from the server output; if needed, mark it external so it can be resolved at runtime.
3. Since you're on pdfjs-dist 5.4.530, also test the latest 5.5.x release before digging deeper.

So to your direct question: in the current Node legacy path, yes, @napi-rs/canvas can still matter even if your end goal is only text extraction.

If adding @napi-rs/canvas still doesn't fix it, I'd next inspect Nitro externals/server bundling, because this looks more like runtime packaging than a problem with the getTextContent() loop itself.