Specify path to .wasm file in v2.0.0-beta.5

[ryan-codingintrigue]

Problem.
In the latest version of tesseract.js-core@v2.0.0-beta.5 the .wasm file is embedded within the file as a Base64 Data URI.

Any workers which have an appropriately strict connect-src Content-Security-Policy header in place will fail to load.

In addition to the above, it might be nice to be able to cache the .wasm separately from the worker .js

Describe the solution you'd like
It would be great if we could optionally specify a path to the .wasm file for the worker to fetch from. If that path isn't specified, then the worker may download from Base64 as a fallback

[jeromewu]

@ryan-codingintrigue

Is there any side effect if we choose to provide only non-embedded version? I am not quiet sure about how to do both with fallback capability.

[ryan-codingintrigue]

It does mean that you must specify some form of wasmPath to locate the file which is unfortunately another parameter.

I think the usual thing to do is, if a path is provided use that, otherwise fallback to the current directory + wasm filename.

if(options.wasmPath) {
    fetch(options.wasmPath)
} else {
    fetch("tesseract.wasm");
}

If you wanted to avoid that, you could provide two distinct core bundles which can be used for corePath if desired:

• tesseract-core.js - Default. Contains fetch("data:application/octet-stream...)
• tesseract.js-core-external.js - Optional. Contains fetch(options.wasmPath)

[jeromewu]

I have upgrade tesseract-core.js to 2.0.0-beta.8 which contains 3 versions of tesseract-core:

• ASM version: https://unpkg.com/tesseract.js-core@2.0.0-beta.8/tesseract-core.asm.js
• WASM version (embedded): https://unpkg.com/tesseract.js-core@2.0.0-beta.8/tesseract-core.wasm.js
• WASM version:
  • javascript: https://unpkg.com/tesseract.js-core@2.0.0-beta.8/tesseract-core.js
  • wasm: https://unpkg.com/tesseract.js-core@2.0.0-beta.8/tesseract-core.wasm

Right now ASM version and WASM version (embedded) and it works fine, but I actually cannot get WASM version working. I leave it there for now, and feel free to send PR 😃

BTW, you can try it using code below (in browser) with tesseract.js@2.0.0-alpha.4:

const { TesseractWorker } = Tesseract;
const worker = new TesseractWorker({
  corePath: 'https://unpkg.com/tesseract.js-core@2.0.0-beta.8/tesseract-core.js',
});

worker
  .recognize(image)
  .then(result => console.log(result.text));

[WebReflection]

FWIW, this is how I managed to make it work via an extension:

const worker = createWorker({
  errorHandler: console.error,
  workerBlobURL: false,
  workerPath: extension.getURL("js/worker.js"),
  corePath: extension.getURL("js/tesseract-core.wasm.js"),
  langPath: extension.getURL("tessdata")
});

At least in Firefox, this doesn't result into loading anything embedded, and the js/tesseract-core.wasm.js path is retrieved without issues.

[Balearica]

Closing the issue as the existing solution (embedded js/wasm) seems to work fine. Feel free to reopen if there's a compelling reason for pursuing this further.