API proposal: expose System.Text.Encoding.Latin1 getter publicly

[GrabYourPitchforks]

The ISO-8859-1 encoding is used somewhat commonly in web scenarios since it's the encoding normally specified for headers. We should consider having a fast-access property for this encoding just like we have for Encoding.UTF8.

The workaround for now is to use Encoding.GetEncoding("iso-8859-1", ...), which works but is not convenient or optimized.

API Proposal

namespace System.Text
{
    public class Encoding
    {
        /* NEW static propery getter */
        public static Encoding Latin1 { get; }
    }
}

The return value of this would be equivalent to calling Encoding.GetEncoding("iso-8859-1"). The byte to char conversion is a naive widening operation. The char to byte conversion is a naive narrowing operation, with any char values above 0x007F being best-fit mapped to an ASCII byte. If a best-fit mapping is unavailable then the character is converted to (byte)'?'.

If the developer wishes to change the fallback behavior or to enable best fit mapping, they can call Encoding.GetEncoding("iso-8859-1", ...) and pass custom fallback behaviors as needed for the newly-created instance.

This proposal does not intend on making the existing internal Latin1Encoding class public.

/cc @davidfowl

[tarekgh]

We already have it and called Encoding.Latin1

runtime/src/libraries/System.Private.CoreLib/src/System/Text/Encoding.cs

Line 495 in 1767c68

private static Encoding Latin1 => Latin1Encoding.s_default;

[stephentoub]

We already have it and called Encoding.Latin1

I think @GrabYourPitchforks is proposing it be made public.

[tarekgh]

Ah, somehow I had the impression that Encoding.Latin1 is already public :-) so it make sense to expose it (maybe with different name).

[GrabYourPitchforks]

I'd be fine with keeping the property name Latin1. The actual concrete type doesn't need to be made public IMO.

[tarekgh]

I agree we don't have to expose Latin1Encoding class. For the naming, we need to choose the name which people more familiar with:

MIME / IANA	ISO-8859-1
Alias(es)	iso-ir-100, csISOLatin1, latin1, l1, IBM819, CP819

if we can have the name indicate ISO-8859-1 that would be better I think.

[scalablecory]

I think this API would be fine.

Outside of the web (iirc HTML5 actually considers the two identical), Windows-1252 is in my experience even more common than ISO-8859-1.

[GrabYourPitchforks]

I updated the proposal to perform best-fit mapping during the narrowing conversion. This provides compatibility with the existing Latin1Encoding default behaviors.

[terrajobst]

• Looks good.
• We won't be exposing the Latin1Encoding type, because it's not necessary

namespace System.Text
{
    public partial class Encoding
    {
        public static Encoding Latin1 => Encoding.GetEncoding("iso-8859-1");
    }
}