JSON 序列化过程中的自引用循环问题与 EFCore 如何加载相关数据有关 (docs).加载集合时，相关实体可能会或可能不会自动填充，具体取决于这些对象之前是否已加载.这称为导航属性的自动修复.或者，它们可以通过 .Include().

The problem of self-referencing loops during JSON serialization is connected to how EFCore loads related data (docs). When you load a collection, related entities may or may not be automatically populated depending on whether or not those object have been previously loaded. It's called automatic fix-up of navigation properties. Alternatively, they may be eagerly loaded via .Include().

每当您获得要序列化的实体的自引用图时，有几种选择:

Whenever you get a self-referencing graph of entities to be serialized, there are several options:

• Newtonsoft.Json.ReferenceLoopHandling.Ignore (官方推荐).如果自引用发生在层次结构的深处，它可以工作，但仍然会导致序列化过多的数据.

• Newtonsoft.Json.ReferenceLoopHandling.Ignore (officially recommended). It works but still can result in serializing excessive data if self-reference occurs deep in the hierarchy.

[JsonIgnore] 导航属性上的属性.正如您所指出的，重新生成模型类时，属性会消失.因此，它们的使用可能会很不方便.

[JsonIgnore] attribute on navigation properties. As you noted, attributes disappear when model classes are regenerated. Thus, their use can be inconvenient.

(***选择) 预选属性子集:

var minimallyNecessarySet= _nwind.Products.Select(p => new {
    p.ProductID,
    p.ProductName,
    p.UnitPrice,
    p.CategoryID
});

return minimallyNecessarySet.ToList();

这种方法的优点是仅序列化所需的数据.它与 DevExtreme 的 DataSourceLoader 兼容:

This approach has the advantage of serializing only required data. It's compatible with DevExtreme's DataSourceLoader:

return DataSourceLoader.Load(minimallyNecessarySet, loadOptions);