Lack of documentation for encoding fallbacks

[ulrj]

Description

Hi

neither me not copilot found any documentation why the following program (in particular the GetBytes part) does what it does and if there are other character replacements. I think there should be a full table of character replacements applied somewhere documented or easily retrievable in code.

using System;
using System.Text;

namespace EncodingFun1
{
    class Program
    {
        static void Main(string[] args)
        {
            Console.WriteLine(Encoding.Latin1.GetString(Encoding.Latin1.GetBytes("© Š š Ž ž Œ œ Ÿ €")));
            Console.ReadKey();
        }
    }
}

Copilot asked me to report this here. So I did.

Best Regards

Marc

Reproduction Steps

Encoding.Latin1.GetBytes("Š")[0]; // returns 83 and not 63. This is fine, but needs documentation.

Expected behavior

documented behavior

Actual behavior

undocumented behavior

Regression?

no.

Known Workarounds

guessing.

Configuration

No response

Other information

No response

[tarekgh]

This is documented in the EncoderFallback page. It has the following:

Best-fit fallback, which maps valid Unicode characters that cannot be encoded to an approximate equivalent. For example, a best-fit fallback handler for the `ASCIIEncoding` class might map Æ (U+00C6) to AE (U+0041 + U+0045). 
A best-fit fallback handler might also be implemented to transliterate one alphabet (such as Cyrillic) to another (such as Latin or Roman). The .NET Framework does not provide any public best-fit fallback implementations.

@ulrj If you want to clarify more the docs, feel free to submit a PR with your suggested contents.

[ulrj]

Hi @tarekgh ,

thanks for the feedback. However: to me

The .NET Framework does not provide any public best-fit fallback implementations.

sounds wrong. There seems to be a best fit fallback (see above). If I knew where the one source of truth is in that regard I would not need documentation. I can work with a code pointer either.

[tarekgh]

@ulrj

sounds wrong.

I'm not sure I fully understand what you mean. When we say, "The .NET Framework does not provide any public best-fit fallback implementations," it means there's no public API for best-fit fallback—but it is used internally by the encoding classes. In this respect, the behavior between .NET Framework and .NET should be the same.
That said, I agree the documentation could use some clarification.