Add warning about encodings not supported by Python

[chardet]

Add warning about encodings not supported by Python

[dan-blanchard]

I'm also considering whether or not we should even detect these legacy encodings by default.

[sigmavirus24]

If we want to deprecate those, we should save that removal until a major version bump and communicate it well.

[dan-blanchard]

Yeah. And I don't think we should remove them entirely. We should just not check for them by default.

[sigmavirus24]

So in that case, I'd like to propose a different API than what you might be thinking about for these issues.

I think a Configuration object that works like this might be good:

from chardet import config

cfg = config.Configuration().enable(
    'EUC-TW'
    ).prerefine(
    'GB2312', 'Shift_JIS', 'EUC-TW',
    )

Then this would make the selections immutable and this could be passed like so:

import chardet

chardet.detect(input_bytes, config=cfg)

[dan-blanchard]

Could you layout the advantages of that over just adding kwargs to chardet.detect?

From my perspective I could see a configuration object being really handy if you were making loads of calls to chardet.detect at various points in your code, but my intuition is that most people are just calling it once (or repeatedly in a loop).

[sigmavirus24]

So, for one, I've grown to hate kwargs. Since we're supporting Python 2, it means we cannot make them keyword-only arguments which means that people will call them positionally. This makes it hard to deprecate features as we need to. Configuration objects as described above mean that we can make a method raise a DeprecationWarning, and removing it is as easy as making the method a no-op.

[dan-blanchard]

That's a really good point about deprecation. I also hate people calling keyword arguments positionally, so I think you've got the right approach there.