ReferenceHandler.IgnoreCycles not handling duplicate strings properly

[jsheetzati]

Description

Using System.Text.Json's ReferenceHandler.IgnoreCycles is not working properly with string properties. It seems that System.Text.Json is improperly detecting strings with the same value as a cycle.

Configuration

System.Text.Json Nuget Version: 6.0.0-preview.3.21201.4

Other information

Here is some sample code that demonstrates the issue. In this case, parent.child.name will be serialized as null. No cycles here so I was expecting the name 'John' to be serialized for both parent.name and parent.child.name.

This same test works fine when not specifying ReferenceHandler and no cyclic exceptions are thrown.

[Fact]
public void Test()
{
	var parent = new Parent
	{
		Name = "John",
		Age = 60,
		Child = new Child
		{
			Name = "John",
			Age = 15
		}
	};

	var json = JsonSerializer.Serialize(parent, new JsonSerializerOptions
	{
		ReferenceHandler = ReferenceHandler.IgnoreCycles,
		WriteIndented = true
	});
}

public class Parent
{
	public string Name { get; set; }

	public int Age { get; set; }

	public Child Child { get; set; }
}

public class Child
{
	public string Name { get; set; }

	public int Age { get; set; }
}

Output without specifying ReferenceHandler

{
  "Name": "John",
  "Age": 60,
  "Child": {
    "Name": "John",
    "Age": 15
  }
}

Output when specifying ReferenceHandler.IgnoreCycles

{
  "Name": "John",
  "Age": 60,
  "Child": {
    "Name": null,
    "Age": 15
  }
}

[ghost]

Tagging subscribers to this area: @eiriktsarpalis, @layomia
See info in area-owners.md if you want to be subscribed.

Issue Details

---

Description

Using System.Text.Json's ReferenceHandler.IgnoreCycles is not working properly with string properties. It seems that System.Text.Json is improperly detecting strings with the same value as a cycle.

Configuration

System.Text.Json Nuget Version: 6.0.0-preview.3.21201.4

Other information

Here is some sample code that demonstrates the issue. In this case, parent.child.name will be serialized as null. No cycles here so I was expecting the name 'John' to be serialized for both parent.name and parent.child.name.

This same test works fine when not specifying ReferenceHandler and no cyclic exceptions are thrown.

[Fact]
public void Test()
{
	var parent = new Parent
	{
		Name = "John",
		Age = 60,
		Child = new Child
		{
			Name = "John",
			Age = 15
		}
	};

	var json = JsonSerializer.Serialize(parent, new JsonSerializerOptions
	{
		ReferenceHandler = ReferenceHandler.IgnoreCycles,
		WriteIndented = true
	});
}

public class Parent
{
	public string Name { get; set; }

	public int Age { get; set; }

	public Child Child { get; set; }
}

public class Child
{
	public string Name { get; set; }

	public int Age { get; set; }
}

Author:	jsheetzati
Assignees:	-
Labels:	area-System.Text.Json, untriaged
Milestone:	-

[GrabYourPitchforks]

JSON team - may want to investigate this behavior for other primitives that have reference equality, such as if a single instance of a boxed integer appears in multiple places.

[eiriktsarpalis]

Curiously this seem to only impact strings in my tests:

using System;
using System.Collections.Generic;
using System.Text.Json;
using System.Text.Json.Serialization;

void Test(object? value)
{
    var list = new List<object?> { value, value };
    var options = new JsonSerializerOptions { ReferenceHandler = ReferenceHandler.IgnoreCycles };
    var json = JsonSerializer.Serialize(list, options);
    Console.WriteLine(json);
}

Test(42); // [42,42]
Test(new object()); // [{},{}]
Test(new List<int>()); // [[],[]]
Test("John"); // ["John",null]

There seems to be something about strings in particular that makes the serializer erroneously treat all recurring references as cycles. cc @jozkee