treat node buffer as uint8array

[jimmywarting]

Don't do any special handling for ppl using node:Buffer treat it as any normal uint8array instead, so that ppl can pass in Uint8Array instead.

I suggest that you change the following code:

ExifReader/src/exif-reader.js

Lines 179 to 183 in 27d92fa

	if (isNodeBuffer(data)) {
	// File data read in Node can share the underlying buffer with other
	// data. Therefore it's safest to get a new one to avoid weird bugs.
	data = (new Uint8Array(data)).buffer;
	}

to instead do this:

if (ArrayBuffer.isView(data)) {
  data = new Uint8Array(data.buffer, data.byteOffset, data.byteLength).slice().buffer
}

i think any appearance of the keyword Buffer should be omited, why? cdn delivery networks have a hard time figuring out if node:Buffer should be be bundled into the code and therefore include it anyway just to be safe,
look for instance in: https://cdn.jsdelivr.net/npm/exifreader@4.31.1/dist/exif-reader.js/+esm and search for allocUnsafe and you will notice that there is a node:buffer polyfill being imported into your project even if this can work without NodeJS specific features.

just by doing typeof Buffer === undefined means that cdn will include it (b/c it implies that you try to access globalThis.Buffer)

related:

• Introducing esm.run - A New-Age CDN for JavaScript modules jsdelivr/jsdelivr#18263 (comment)
• https://sindresorhus.com/blog/goodbye-nodejs-buffer
• Start moving to Uint8Array in new APIs? nodejs/node#41588

[jimmywarting]

there are a bunch of things i don't think that you should include at all... such as loading images from fs, using node:http.get()
it should rather be the responsibility of the developer who depends on this lib to do it. a image can come from many different destinations, and it's not your job to support every possible method.

related:
https://github.com/cross-js/cross-js?tab=readme-ov-file#dont-use-fs
https://web.archive.org/web/20190328210220/https://www.codeguru.com/csharp/csharp/cs_misc/designtechniques/understanding-onion-architecture.html

[jimmywarting]

i would rather wish you suggest to ppl how they themself how they can get a hold of a blob by a simple demonstration in your readme:

import { openAsBlob } from 'node:fs'
import ExifReader from 'exifreader'

const blob = await openAsBlob(path)
const tags = await ExifReader.load(blob)

[mattiasw]

Hi!

I will look into Buffer removal.

Regarding the other things, it started out that way but users wanted to be able to pass in URLs and filenames. One reason was probably that several (maybe most at the time) of the other similar packages did/do support this. That does of course not mean it has to be supported here too but after a while it was hard to resist.

[jimmywarting]

maybe you can at least replace requireNodeGet with the new built in fetch api.

[mattiasw]

I will do that for the next major version bump. Which I really should start working on soon, but I can't say when I'll have time.