Error during extraction

[bennyzen]

I'm using scribe.js to batch process a large number of PDFs. The error below keeps emerging from time to time. Its appearance is pretty random and is NOT related to specific PDFs. After restarting the batch process, it processes that specific PDF just fine.

file:///home/node/node_modules/.pnpm/scribe.js-ocr@0.5.0/node_modules/scribe.js-ocr/js/generalWorkerMain.js:335
    if (gs.schedulerInner.workers.length > 0) {
                          ^

TypeError: Cannot read properties of null (reading 'workers')
    at gs.initTesseract (file:///home/node/node_modules/.pnpm/scribe.js-ocr@0.5.0/node_modules/scribe.js-ocr/js/generalWorkerMain.js:335:27)
    at async Promise.all (index 2)
    at async init (file:///home/node/node_modules/.pnpm/scribe.js-ocr@0.5.0/node_modules/scribe.js-ocr/scribe.js:77:3)

Node.js v22.11.0

[Balearica]

Can you post a minimal version of the code you are using? This error appears to be caused by resources being cleared while recognition is still running, so it's not possible to answer without knowing the code that is being run.

[bennyzen]

Yes, sure. It's happening even using your code from the nodejs example:

import scribe from 'scribe.js-ocr'

const text = await scribe.extractText(['./image-native.pdf'])
await scribe.terminate()
console.log(text)

I'm not absolutely sure, but in v0.5.0 it seems to happen more often than in previous versions. With certain files it happens always. Unfortunately, I cannot attach the culprit file for you, as it's confidential.

Strange thing: using the scribeocr frontend, it just works without any problem.

Things I've tried so far:

• using different Node versions 20, 22, 23 - always fully rebuilding packages
• using Bun runtime, which does not work at all
• using NPM or PNPM as package manager

With the minimal example above I get this error instead:

Error: TypeError: Cannot read properties of undefined (reading '0')
    at Worker.<anonymous> (/home/ben/repos/scribejs/node_modules/@scribe.js/tesseract.js/src/createWorker.js:283:15)
    at Worker.emit (node:events:518:28)
    at MessagePort.<anonymous> (node:internal/worker:263:53)
    at [nodejs.internal.kHybridDispatch] (node:internal/event_target:826:20)
    at exports.emitMessage (node:internal/per_context/messageport:23:28)

And yes, I've tried the recommended init, import, ... route, but unfortunately it fails there too.

Any hint what I could try next? Any clue on why it is working in the browser (scribeocr) and not on the terminal?

Thank you in advance.

[Balearica]

@bennyzen Thanks for clarifying. I was able to replicate the second error message (Cannot read properties of undefined (reading '0')). However, I believe this is a separate issue given where in the code it occurs, so have opened a separate issue to track it: #26.

I will troubleshoot and fix that error, and update #26 accordingly. However, as noted above, I doubt this is directly related to the original error message, so let me know if you can reliably replicate the original error message with a minimal example.

[bennyzen]

Yes, you're right. While trying and re-trying, I was causing confusion and mixing different errors. I'll track the other error over there #26

Thank you for your effort. I really appreciate.

[Balearica]

@bennyzen #26 has been patched in v0.5.1, so if you update to the latest version, any errors should be specific to the original issue.