We Keep Coding

Refactoring classes into multi-layered generic classes in C#

I have a problem with C# generics, and I'm not sure about the most elegant solution. I've been programming a while but am new to the C# ecosystem so don't know common terminology for searching.
I'm trying to refactor code to reduce existing copy-paste duplication of classes. It is easy to resolve with one level of generics, but I can't get it working with two.
A very simplified example is below. The core issue is that BaseProfile cannot use any implementation details relating to DetailsA or DetailsB as it does not know the type. So UpdateDetailsId() has to be duplicated in 2 derived classes, instead of having a single Profile class handle it. Keep in mind this is a toy example just to express the relationships. The real classes have tens of fields, but a common subset which we are using in the class in question, so even if DetailsA and DetailsB look identical assume we need both.
public abstract class BaseProfile<TypeOfPerson>
{
public TypeOfPerson Person { get; set; }
}
public class Profile1 : BaseProfile<PersonA>
{
public void UpdateDetailsId(int id)
{
this.Person.Details.Id = id;
}
}
public class Profile2 : BaseProfile<PersonB>
{
public void UpdateDetailsId(int id)
{
this.Person.Details.Id = id;
}
}
public class PersonA
{
public DetailsA Details { get; set; }
}
public class PersonB
{
public DetailsB Details { get; set; }
}
public class DetailsA
{
public int Id { get; set; }
}
public class DetailsB
{
public int Id { get; set; }
}
I can add interfaces as it is referring to all the same fields for each type. However, C# will not allow an interface to include another interface and automatically resolve it in the implementation, because the member has to exactly match i.e. I thought I could just add IDetails Details to the IPerson interface but the fields now need to be type IDetails instead of DetailsA which implements IDetails. If I do that then I lose compiler type safety and can put the wrong Details on the wrong Person.
I have had success doing a public/private field pair like below, but this only validates and throws at runtime when casting value to DetailsA. I'd prefer something safer but I don't know if this is the best option. The goal of this example is a single Profile class, handling multiple Person classes, each with their own Details type that has an int Id field.
public class PersonA : IPerson
{
public IDetails Details
{
get { return _details; }
set { _details = (DetailsA)value; }
}
private DetailsA _details { get; set; }
}

One way of achieving this is by defining the type relationship between PersonA to DetailsA in a generic way, and specify a second generic type on BaseProfile.
Profile1 : BaseProfile<PersonA, DetailsA>
Consider the following code (note that I'm using Net6, so I have all these nullable reference type operators):
public abstract class BaseProfile<TPerson, TDetails>
where TDetails : IDetails, new()
where TPerson : PersonDetails<TDetails>, new()
{
public TPerson? Person { get; set; } = new TPerson();
public virtual void UpdateDetailsId(int id)
{
Person!.Details!.Id = id;
}
}
public class Profile1 : BaseProfile<PersonA, DetailsA>
{
}
public class Profile2 : BaseProfile<PersonB, DetailsB>
{
}
public abstract class PersonDetails<TDetails>
where TDetails : IDetails, new()
{
public virtual TDetails? Details { get; set; } = new TDetails();
}
public class PersonA : PersonDetails<DetailsA>
{
}
public class PersonB : PersonDetails<DetailsB>
{
}
public interface IDetails
{
int Id { get; set; }
}
public class DetailsA : IDetails
{
public int Id { get; set; }
public string? FirstName { get; set; }
}
public class DetailsB : IDetails
{
public int Id { get; set; }
public string? LastName { get; set; }
}
Testing with the following snippet
var profile1 = new Profile1();
var profile2 = new Profile2();
profile1.UpdateDetailsId(10);
profile2.UpdateDetailsId(12);
Console.WriteLine(profile1.Person!.Details!.Id);
Console.WriteLine(profile2.Person!.Details!.Id);
Console.WriteLine();
Update:
Because you included explicit casting in your snippet for Details property getters and setter, I also want to show a pattern using a concrete type inheriting on these generic types -- then demonstrate implicit/explicit operator user-defined conversion patterns.
Add the following declarations:
public abstract class BaseProfile<TPerson>
where TPerson : PersonDetails<GenericDetails>, new()
{
public TPerson? Person { get; set; } = new TPerson();
public virtual void UpdateDetailsId(int id)
{
Person!.Details!.Id = id;
}
public static explicit operator Profile1(BaseProfile<TPerson> details)
{
var profile = new Profile1();
profile.Person!.Details = (GenericDetails)details.Person!.Details!;
return profile;
}
public static explicit operator Profile2(BaseProfile<TPerson> details)
{
var profile = new Profile2();
profile.Person!.Details = (GenericDetails)details.Person!.Details!;
return profile;
}
}
public class GenericProfile : BaseProfile<GenericPerson>
{
}
public abstract class GenericPersonDetails : PersonDetails<GenericDetails>
{
}
public class GenericPerson : GenericPersonDetails
{
}
public class GenericDetails : IDetails
{
public int Id { get; set; }
public static implicit operator DetailsA(GenericDetails details)
{
return new DetailsA() { Id = details.Id };
}
public static implicit operator DetailsB(GenericDetails details)
{
return new DetailsB() { Id = details.Id };
}
}
and, update the testing functional scope:
var profile1 = new Profile1();
var profile2 = new Profile2();
var genericProfile = new GenericProfile();
profile1.UpdateDetailsId(10);
profile2.UpdateDetailsId(12);
genericProfile.UpdateDetailsId(20);
Console.WriteLine(profile1.Person!.Details!.Id);
Console.WriteLine(profile1.Person!.Details!.FirstName ?? "No First Name");
Console.WriteLine(profile2.Person!.Details!.Id);
Console.WriteLine(profile2.Person!.Details!.LastName ?? "No Last Name");
Console.WriteLine(genericProfile.Person!.Details!.Id);
Console.WriteLine(((Profile1)genericProfile).Person!.Details!.FirstName ?? "No First Name");
Console.WriteLine(((Profile2)genericProfile).Person!.Details!.LastName ?? "No Last Name");
Console.WriteLine();

Using empty base class just for polymorphism purpose?

I have my ResponseDto which includes a simple string property named Answer.
public string Answer { get; set; }
Now, the requirement came such that I could either be getting an answer as a string, or as an array of int.
I decided to create two classes for this:
public class AnswerType {
public string Answer { get; set; }
}
public class OptionAnswerType {
public int[] AnswerOptionIds { get; set; }
}
I could serialize / deserialize it accordingly.
But to still keep a single response property type, I thought about creating an empty base class:
public class BaseAnswerType { }
public class AnswerType : BaseAnswerType {
public string Answer { get; set; }
}
public class OptionAnswerType : BaseAnswerType {
public Guid[] AnswerOptionIds { get; set; }
}
and change my property in ResponseDto to:
public BaseAnswerType Answer { get; set }
through which via run time, I would be returning either of the two classes.
Is this a bad approach? Any alternate would be greatly appreciated.

C# Counting properties of Class with child/nested objects

I have the following construction of classes, here simplified as child classes of a 'mother' class called DataClass, which also contains one simple method:
public class DataClass
{
public int num { get; set; }
public string code { get; set; }
public PartClass part { get; set; }
public MemberClass member { get; set; }
public int Count()
{
Type t = typeof(DataClass);
return typeof(DataClass).GetProperties().Length;
}
}
public class PartClass
{
public int seriesNum { get; set; }
public string seriesCode { get; set; }
}
public class MemberClass
{
public int versionNum { get; set; }
public SideClass side { get; set; }
}
public class SideClass
{
public string firstDetail { get; set; }
public string secondDetail { get; set; }
public bool include { get; set; }
}
The issue is, I want to refactor the method so that it can give me an accurate counting of all properties found, including the ones in nested or child classes. In the above example, it only counts properties of DataClass, while I wanted it to return 2 for DataClass + 2 for PartClass + 1 for MemberClass + 3 for SideClass, sums up to 8 properties you may set through DataClass.
Can someone help me with this?

You can introduce interface with Count() method
public interface ICountable
{
int Count();
}
And use this interface to mark all types, which properties are participating in Count() calculation.
You can see the generic abstract class to implement this interface below. Generic T parameter is type whose properties need to be calculated. You implement a calculation logic only once and inherit this class where needed. You also go through all of properties, implementing ICountable, to calculate them as well (some kind of recursion)
public abstract class Countable<T> : ICountable
{
public int Count()
{
Type t = typeof(T);
var properties = t.GetProperties();
var countable = properties.Select(p => p.PropertyType).Where(p => typeof(ICountable).IsAssignableFrom(p));
var sum = countable.Sum(c => c.GetProperties().Length);
return properties.Length + sum;
}
}
and inherit it in your classes
public class DataClass : Countable<DataClass>
{
...
}
public class PartClass : Countable<PartClass>
{
...
}
public class MemberClass : Countable<MemberClass>
{
...
}
public class SideClass : Countable<SideClass>
{
...
}
And this is for the test
var dataClass = new DataClass();
var count = dataClass.Count();
It returns 8 as expected

Generic Attributes in MVC models

I created the GenericAttribute.cs file in my Models
public class GenericAttributes<T>
{
public T Id { get; set; }
public bool IsActive { get; set; }
public DateTime CreatedDate { get; set; }
}
Now I want to add 'int id' field in my User Model
public class User
{
//here I want to add 'Id' field
public string UserId { get; set; }
public string password { get; set; }
public string UserType { get; set; }
public int EmployeeId { get; set; }
public virtual Employee employee { get; set; }
}
How should I do this? Please help

You can make GenericAttributes an interface so you can implement it where ever.
Such as;
public interface IGenericAttributes<T>
{
//properties
}
And use in your class declaration;
public class User : IGenericAttributes<int>
{
//properties
}
This will force your concrete type User to implement the properties of the interface.

You are getting some conflicting answers due to your naming convention. Any class of the form xxxAttribute is expected to be a subclass of the Attribute class. Attributes are metadata that you can attach to classes, fields, etc. Using reflection you can read these attributes, which is a powerful way to inform various APIs about how to interact with your custom classes - without inheritance or an interface.
If this sort of metadata is your intent, then Barr J's answer is correct. However, if your intent is for the GenericAttributes class to serve as a base class that you can inherit these properties from, then Tom Johnson is correct (although he did change GenericAttributes into an interface instead of a base class, but same result if all you have are properties like this). The latter is most likely what you are looking for.
I would suggest renaming GenericAttributes to something more descriptive, like BaseRecord or IRecord (as an interface), since User looks like data coming from or going to a database.
It would also be handy to have a non-generic version of the class/interface so that you can non-generically reference such records.
public class BaseRecord {
public Type IdType { get; }
private Object _id = null;
public Object Id {
get {
return _id;
}
set {
if(value != null) {
if(!IdType.IsAssignableFrom(value.GetType()))
throw new Exception("IdType mismatch");
}
_id = value;
}
}
public bool IsActive { get; set; }
public DateTime CreatedTime { get; set; }
public BaseRecord(Type idType)
{
if(idType == null) throw new ArgumentNullException("idType");
this.IdType = idType;
}
}
namespace Generic {
public class BaseRecord<T> : BaseRecord
{
new public T Id {
get { return (T)base.Id; }
set { base.Id = value; }
}
public BaseRecord() : base(typeof(T))
{
}
}
}
public class User : Generic.BaseRecord<int>
{}
public class OtherRecord : Generic.BaseRecord<string>
{}
// This inheritence scheme gives you the flexibility to non-generically reference record objects
// which can't be done if you only have generic base classes
BaseRecord r = new User();
r = new OtherRecord();
BaseRecord records[] = { new User(), new OtherRecord() };

To access the id for GenericAttributes class, you'll have to cast User object as base class type.
namespace SampleApp
{
class SampleProgram
{
static void Main(string[] args)
{
User User = new User() { Id = 1 };
var genericAttribute = (User as GenericAttributes<int>);
genericAttribute.Id = 2;
var genericAttributeId = genericAttribute.Id;
var classId = User.Id;
}
}
public class GenericAttributes<T>
{
public T Id { get; set; }
}
public class User : GenericAttributes<int>
{
public new int Id { get; set; }
}
}

C# - Passing different type of objects through the same method

Original Question
So I have this 3 objects...
public class obj1
{
public int Id { get; set; }
public string Name { get; set; }
}
public class obj2
{
public int AccNum { get; set; }
public string Name { get; set; }
}
public class obj3
{
public string Email { get; set; }
public string Phone { get; set; }
}
... and one method that is supposed to receive one of them, after evaluating the object type the program should decide which function to call.
I've tried with generics but it doesn't work as I expected. So far this is what I've got...
public class NotificationHelper: INotificationHelper
{
public bool SendNotification<TNotInfo>(TNotInfo obj) where TNotInfo : class
{
if (contract.GetType() == typeof (obj1))
{
var sender = new SendSMS();
return sender.Send(obj);
}
if (contract.GetType() == typeof(obj2))
{
var sender = new SendPush();
return sender.Send(obj);
}
else
{
var sender = new SendEmail();
return sender.Send(obj);
}
}
}
but I get the error "Cannot convert from TNotInfo to Models.obj1". Is there any way to overcome this issue? Or I have to change my logic?
Appreciate any help, thanks in advance.
*Edit
using System;
namespace EmailNotifications
{
public interface IEmailNotification
{
void SendEmailNotification();
}
public class EmailNotificationA : IEmailNotification
{
public void SendEmailNotification(Contract1 a)
{
Console.WriteLine($"Sending EmailNotificationA ({a})");
}
}
public class EmailNotificationB : IEmailNotification
{
public void SendEmailNotification(Contract2 b)
{
Console.WriteLine($"Sending EmailNotificationB ({b})");
}
}
public class EmailNotificationC : IEmailNotification
{
public void SendEmailNotification(Contrac3 c)
{
Console.WriteLine($"Sending EmailNotificationC ({c})");
}
}
public class EmailNotificationService
{
private readonly IEmailNotification _emailNotification;
public EmailNotificationService(IEmailNotification emailNotification)
{
this._emailNotification = emailNotification;
}
public void ServiceHelper()
{
_emailNotification.SendEmailNotification();
}
}
}
Above solution is what I was trying to achieve, applying strategy design pattern. But I couldn't manage to make my interface method receive different objects, this is required because each notification has is own implementation. As visible at the none working example above, I have 3 different implementation of the same method all of them receiving different objects. Any idea of how to make this logic work?

This is the kind of thing that interfaces were designed to do. First, define a common interface:
public interface INotifier
{
bool Notify();
}
Second, implement it in your objX classes:
public class obj1 : INotifier
{
public int Id { get; set; }
public string Name { get; set; }
public bool Notify()
{
var sender = new SendSMS();
return sender.Send(this);
}
}
public class obj2 : INotifier
{
public int AccNum { get; set; }
public string Name { get; set; }
public bool Notify()
{
var sender = new SendPush();
return sender.Send(this);
}
}
public class obj3 : INotifier
{
public string Email { get; set; }
public string Phone { get; set; }
public bool Notify()
{
var sender = new SendEmail();
return sender.Send(this);
}
}
And finally, change your notification method to accept the interface type as the parameter:
public class NotificationHelper : INotificationHelper
{
public bool SendNotification(INotifier obj)
{
return obj.Notify();
}
}
Edit (2019):
I'm revisiting this answer as it seems to be getting a fair amount of visibility. OP has probably long since moved on, but for others that may stumble upon this answer, here's another solution.
I still believe that interfaces are the way to go. However, the interface suggested above is extremely generic and ultimately not terribly useful. It also runs into some DRY violations because, as Fabio said in a comment, if two objX classes implement notifications in the same way, this approach forces you to duplicate the code between them.
Instead of one global interface, instead have interfaces for each specific notification task, i.e. ISMSNotification, IPushNotification, IEmailNotification. You can then use the mixin pattern to give each interface instance a default implementation of the send method:
interface ISmsNotifier
{
int SmsId { get; }
string SmsName { get; }
}
static class ISmsNotifierExtensions
{
public static bool NotifySms(this ISmsNotifier obj)
{
var sender = new SendSMS();
return sender.Send(obj);
}
}
// ---------------------------------------------
interface IPushNotifier
{
int PushAccNum { get; }
string PushName { get; }
}
static class IPushNotifierExtensions
{
public static bool NotifyPush(this IPushNotifier obj)
{
var sender = new SendEmail();
return sender.Send(obj);
}
}
// ---------------------------------------------
interface IEmailNotifier
{
string EmailAddress { get; }
string EmailPhone { get; }
}
static class IEmailNotifierExtensions
{
public static bool NotifyEmail(this IEmailNotifier obj)
{
var sender = new SendEmail();
return sender.Send(obj);
}
}
You can then implement it in the objX classes like so:
public class obj1 : INotifier, ISmsNotifier
{
public int SmsId { get; set; }
public string SmsName { get; set; }
public bool Notify() => this.NotifySms();
}
public class obj2 : INotifier, IPushNotifier
{
public int PushAccNum { get; set; }
public string PushName { get; set; }
public bool Notify() => this.NotifyPush();
}
public class obj3 : INotifier, IEmailNotifier
{
public string EmailAddress { get; set; }
public string EmailPhone { get; set; }
public bool Notify() => this.NotifyEmail();
}
Notice that using this approach it's easy to not only support objects which use identical notification systems, you can also support objects with multiple notification systems:
public class obj4 : INotifier, IEmailNotifier, IPushNotifier
{
public int PushAccNum { get; set; }
public string PushName { get; set; }
public string EmailAddress { get; set; }
public string EmailPhone { get; set; }
public bool Notify() => this.NotifyEmail() && this.NotifyPush();
}
You might notice that this approach makes NotificationHelper obsolete since it's no longer necessary to pass the objects through a processing step to determine which notification system to process the object through. That is true, and maybe rightfully so, since the objects should be fully capable of deciding that for themselves (depending on your mentality approaching this problem). However, NotificationHelper may still have its uses, such as if you wanted to preprocess the information that's getting sent to the notification services, or if you wanted a common point of entry to help with mocking and testing.
C# 8 Note:
A proposed feature of C# 8 is the ability to give interfaces a default implementation of methods within the interface definition itself. When (if) that happens, you don't need to use the mixin pattern anymore and can directly define the default method implementations in the interfaces. The feature hasn't yet been finalized, but it might look something like this:
interface ISmsNotifier
{
int SmsId { get; }
string SmsName { get; }
public bool NotifySms()
{
var sender = new SendSMS();
return sender.Send(this);
}
}
// ---------------------------------------------
interface IPushNotifier
{
int PushAccNum { get; }
string PushName { get; }
public bool NotifyPush()
{
var sender = new SendEmail();
return sender.Send(this);
}
}
// ---------------------------------------------
interface IEmailNotifier
{
string EmailAddress { get; }
string EmailPhone { get; }
public bool NotifyEmail()
{
var sender = new SendEmail();
return sender.Send(this);
}
}

Another approach will be overload methods.
Because you have different logic based on the given type. And types have nothing in common (interface/abstract class).
public class NotificationHelper
{
public bool SendNotification(obj1 obj)
{
var sender = new SendSMS();
return sender.Send(obj);
}
public bool SendNotification(obj2 obj)
{
var sender = new SendPush();
return sender.Send(obj);
}
public bool SendNotification(obj3 obj)
{
var sender = new SendEmail();
return sender.Send(obj);
}
}
Then using will be clear enough
var someObject = GetObjectFromSomeWhere();
var isSuccessful = SendNotification(someObject);

I would suggest creating a parent class from which these 3 inherit
public class ParentType { }
public class Obj1 : ParentType { ... }
The method would then just request the parent type, such as:
public bool SendNotification(ParentType obj) { ... }