Fix issues surrounding type checking for Uint8Arrays and Buffers

[chjj]

Logic behind this:

1. b instanceof Uint8Array always returns true for Buffers created in the same context.
2. ArrayBuffer.isView(b) returns true for both Uint8Arrays and Buffers created in any context (at least in v8 -- hopefully other browsers too).
3. isInstance(b, Uint8Array) returns true for Uint8rrays created in any context but currently only returns true for Buffers created in the same context.

Number 3 has an inconsistency. This PR seeks to rectify that and also optimize several functions in doing so.

If we modify the code such that isInstance(b, Uint8Array) returns true for Buffers created in all contexts, we can drop a ton of the "buffer casting" for Uint8Arrays (which in most cases is actually entirely redundant and causes a slowdown -- for example, in Buffer.compare). I suspect a lot of this code is actually left over from the days when buffer.js still supported IE6.

Not only does this fix a few bugs (e.g. Buffer#equals and Buffer#copy not accepting Uint8Arrays as their target), but it can also optimize a few functions by avoiding the buffer casts.

In effect, we treat Buffers and Uint8Arrays (from any context) as though they were of the exact same class, which is what node.js does.

[dcousens]

LGTM

[dcousens]

What isn't necessary about this now?
(citation?)

[chjj]

Because ArrayBuffer.isView returns true for Uint8Array subclasses like Buffer. So this case will actually be handled earlier, in Buffer.from:

  if (ArrayBuffer.isView(value)) {
    return fromArrayView(value)
  }

I left the Buffer.isBuffer(obj) check in there because I was concerned that maybe non-v8 javascript engines' ArrayBuffer.isView does not return true for Uint8Array subclasses (from the same context or another).

Because ArrayByffer.isView is so magic in v8, I actually wanted to test it in FF and Safari to make sure they had the same behavior, but I didn't get around to it, so leaving the check in is just me being cautious.

[dcousens]

You should add comments like ^ into your code 💙

[chjj]

Yeah, it's something I need to work on.

[hedi-edelbloute]

Hi @chjj, your update looks well ! Thanks

Do you know if it could fix this issue #329 ? 🤔

[chjj]

@hedi-edelbloute, it shouldn't affect subarray at all. feross/buffer doesn't ever try to patch or change subarray, but if you are experiencing the issue described in #329, please post the affected javascript engine and version in that issue.

[hedi-edelbloute]

@chjj I'm going to provide infos on the issue then, thanks