[API Proposal]: Consider support for IEquatable<Uri> on System.Uri

[dahlia]

Edited by @MihaZupan on 2024-02-16

Background and motivation

For a long time, the System.Uri type has provided equality comparison operations via the object.Equals(object?) method and the operator ==() & operator !=() operators. Therefore, there is no reason not to further implement the IEquatable<Uri> interface.

API Proposal

namespace System;

-public partial class Uri : ISpanFormattable, ISerializable
+public partial class Uri : ISpanFormattable, ISerializable, IEquatable<Uri>
{
    // Existing
    public override bool Equals(object? comparand);

+   public bool Equals(Uri? other);
}

API Usage

IEquatable<Uri> a = new Uri("https://example.com/a");
IEquatable<Uri> b = new Uri("https://example.com/b");
Console.WriteLine("Equals? {0}", a.Equals(b));

Alternative Designs

No response

Risks

No response

[ghost]

Tagging subscribers to this area: @dotnet/ncl
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Background and motivation

For a long time, the System.Uri type has provided equality comparison operations via the object.Equals(object?) method and the operator ==() & operator !=() operators. Therefore, there is no reason not to further implement the IEquatable<Uri> interface.

API Proposal

namespace System;

public class Uri : IEquatable<Uri>, ...
{
    bool IEquatable<Uri>.Equals(Uri other);

    ...
}

API Usage

IEquatable<Uri> a = new Uri("https://example.com/a");
IEquatable<Uri> b = new Uri("https://example.com/b");
Console.WriteLine("Equals? {0}", a.Equals(b));

Alternative Designs

No response

Risks

No response

Author:	dahlia
Assignees:	-
Labels:	api-suggestion, area-System.Net, untriaged
Milestone:	-

[teo-tsirpanis]

I don't see a benefit here; IEquatable<T> is particularly useful in value types to avoid boxing.

[RenderMichael]

This would allow Uris to participate in certain generic methods like MemoryExtensions.SequenceEqual.

Api Usage

bool AreEqual(Uri[] left, Uri[] right)
{
    return left.AsSpan().SequenceEqual(right.AsSpan());
}

[tfenise]

MemoryExtensions.SequenceEqual no longer needs IEquatable<T> with .NET 6. https://learn.microsoft.com/en-us/dotnet/api/system.memoryextensions.sequenceequal?view=net-8.0#system-memoryextensions-sequenceequal-1(system-readonlyspan((-0))-system-readonlyspan((-0))-system-collections-generic-iequalitycomparer((-0)))

[dahlia]

If implementing IEquatable<T> adds no benefit to reference types, why does System.String implement IEquatable<String>? I'm just curious.

[huoyaoyuan]

Dictionary requires IEquatable<T> for default value equality. For cases where an equality comparer is optional, implementing IEquatable<T> can change the behavior of fallback comparer.

[MichalPetryka]

Dictionary requires IEquatable<T> for default value equality.

Isn't overriding the default Equals enough for that?

[MihaZupan]

This seems reasonable to me. The main benefits I see are:

• being able to call methods that are limited to IEquatable (e.g. spanOfUris.Contains(uri))
• having an overload of Equals on Uri that takes a strongly typed Uri instead of just object
  • The implementation shouldn't be explicit, we do want it to show up in IntelliSense for Uri

namespace System;

-public partial class Uri : ISpanFormattable, ISerializable
+public partial class Uri : ISpanFormattable, ISerializable, IEquatable<Uri>
{
    // Existing
    public override bool Equals(object? comparand);

+   public bool Equals(Uri? other);
}

@dotnet/ncl any thoughts against marking this one as ready for review?

[stephentoub]

Seems fine.

[bartonjs]

Video

• Looks good as proposed
• Added a [NotNullWhen(true)].

namespace System;

-public partial class Uri : ISpanFormattable, ISerializable
+public partial class Uri : ISpanFormattable, ISerializable, IEquatable<Uri>
{
    // Existing
    public override bool Equals(object? comparand);

+   public bool Equals([NotNullWhen(true)] Uri? other);
}

[fubar-coder]

Might be a dumb question, but what is the behavior for internationalized domain names?

In other words: Are these two equal?

• xn--test--test-u5a0tyc.de
• test-äöü-test.de

[huoyaoyuan]

The behavior would be the same with existing operator == and Equals(object) overload. It's not related to this particular API proposal.

The Uri class doesn't accept IDN names at all. Passing IDN names in constructor will throw UriFormatException.

[MihaZupan]

As mentioned, the new API would exactly match the behavior of the existing one.

In other words: Are these two equal?

The docs suggest they should, but that's currently not the case. That's most likely a bug we should track separately though.