Casting generic types to subclass - c#

I'm starting with generics types and I'm stuck with my project. Maybe I did not understand generics very well. Explanations have been inserted inline. Basically I need to implement the Do() method but I don't know how to resolve <T2>:
public abstract class MyGenericClass<T> { }
public class MyGenericClass<T, T2> : MyGenericClass<T>
{
public Expression<Func<T, T2>> expression;
public MyGenericClass(Expression<Func<T, T2>> expression)
{
this.expression = expression;
}
}
public class MyClass<T>
{
// I need to mantain a list of my generic class for later use.
// I don't know T2 at this point.
// So I Chose to use Inheritance as a workaround (MyGenericClass<T> and MyGenericClass<T, T2>).
// Maybe it's not a good solution but I counldn't find out other solution.
public List<MyGenericClass<T>> MyGenericList = new List<MyGenericClass<T>>();
// I receive the parametric argument T2 here as part of an Expresion.
// And I keep the expression in my list.
public MyGenericClass<T, T2> ReceivingMethod<T2>(Expression<Func<T, T2>> expression)
{
MyGenericClass<T, T2> genericImp = new MyGenericClass<T, T2>(expression);
MyGenericList.Add(genericImp);
return genericImp;
}
}
public class Client<T>
{
MyClass<T> class1;
// class1 has been created and his field MyGenericList has been populated.
// Then when I call Do()....
public void Do()
{
foreach (var item in class1.MyGenericList)
{
// I need something like this here,
// but it does not compile because I don't know T2 here.
// The caller of Do() method doesn't know T2.
MyGenericClass<T, T2> myGenericItem = (MyGenericClass<T, T2>)item;
var a = myGenericItem.expression;
}
}
}

You have to give Do() the T2 parameter somehow. So my solution is to create a method parameter of the same type. I also nested the types in order to make sure all of them refer to the same T.
I also renamed the parameters to be more descriptive
// T -> TArg
// T2 -> TResult
public abstract class MyBaseClass<TArg>
{
public class MyExpressionClass<TResult> : MyBaseClass<TArg>
{
public Expression<Func<TArg, TResult>> Expression { get; private set; }
public MyExpressionClass(Expression<Func<TArg, TResult>> expression)
{
this.Expression=expression;
}
}
public class MyCollectionClass
{
public List<MyBaseClass<TArg>> MyGenericList = new List<MyBaseClass<TArg>>();
public MyExpressionClass<TResult> ReceivingMethod<TResult>(Expression<Func<TArg, TResult>> expression)
{
var genericImp = new MyExpressionClass<TResult>(expression);
MyGenericList.Add(genericImp);
return genericImp;
}
}
public class Client
{
public MyCollectionClass List = new MyCollectionClass();
public void Do<TResult>()
{
foreach(var item in List.MyGenericList)
{
var expr = item as MyExpressionClass<TResult>;
if(expr!=null)
{
var a = expr.Expression;
Console.WriteLine(a);
}
}
}
}
}
class Program
{
static void Main(string[] args)
{
var client = new MyBaseClass<int>.Client();
// add conversion expressions
client.List.ReceivingMethod((i) => (i).ToString());
client.List.ReceivingMethod((i) => (2*i).ToString());
client.List.ReceivingMethod((i) => (3*i).ToString());
// The programmer has to manually enforce the `string` type
// below based on the results of the expressions above. There
// is no way to enforce consistency because `TResult` can be
// _any_ type.
client.Do<string>();
// Produces the following output
//
// i => i.ToString()
// i => (2*i).ToString()
// i => (3*i).ToString()
}
}

Solution 1 Inspired from #KMoussa comment. I've delegate the responsability to MyGenericClass using an abstract method. This seems a better design. All subclasses will implement this method DoTheWork(). And can be invoked from my Client.Do() method with only T param:
public abstract class MyGenericClass<T>
{
public abstract string DoTheWork();
}
public class MyGenericClass<T, T2> : MyGenericClass<T>
{
public override string DoTheWork()
{
// I can use expression here
}
private Expression<Func<T, T2>> expression { get; set; }
public MyGenericClass(Expression<Func<T, T2>> expression)
{
this.expression = expression;
}
}
public class MyClass<T>
{
public List<MyGenericClass<T>> MyGenericList = new List<MyGenericClass<T>>();
public void ReceivingMethod<T2>(Expression<Func<T, T2>> expression)
{
MyGenericClass<T, T2> genericImp = new MyGenericClass<T, T2>(expression);
}
}
public class Client<T>
{
MyClass<T> class1;
public void Do()
{
// I don't need to cast to MyGenericClass<T, T2>
foreach (MyGenericClass<T> myGenericItem in class1.MyGenericList)
{
string result = myGenericItem.DoTheWork();
}
}
}
Solution 2 Inspired from #ja72 and #juharr comments. Using reflection. First I save the type T2 on MyGenericClass using an abstract property. Then I can invoke a generic method MethodWithArgument using reflection so I can introduce the parameter for the casting:
public abstract class MyGenericClass<T>
{
public abstract Type type { get; set; }
}
public class MyGenericClass<T, T2> : MyGenericClass<T>
{
public Expression<Func<T, T2>> expression { get; set; }
public MyGenericClass(Expression<Func<T, T2>> expression)
{
type = typeof(T2); // I save the type of T2
this.expression = expression;
}
}
public class MyClass<T>
{
public List<MyGenericClass<T>> MyGenericList = new List<MyGenericClass<T>>();
public void ReceivingMethod<T2>(Expression<Func<T, T2>> expression)
{
MyGenericClass<T, T2> genericImp = new MyGenericClass<T, T2>(expression);
}
}
public class Client<T>
{
MyClass<T> class1;
public void Do()
{
foreach (MyGenericClass<T> myGenericItem in class1.MyGenericList)
{
MethodInfo method = GetType().GetMethod("MethodWithArgument");
MethodInfo generic = method.MakeGenericMethod(new Type[] { myGenericItem.type });
string g = (string)generic.Invoke(this, new object[] { myGenericItem });
}
}
// I introduce T2 in this method
public string MethodWithArgument<T2>(MyGenericClass<T> myClass)
{
// Now, the casting is valid
MyGenericClass<T, T2> mySubClass = (MyGenericClass<T, T2>)myClass;
var a = mySubClass.expression;
// I can work with expression here
}
}

Related

How to prevent Convert in generated expression for type parameter is interface

Here is the example code:
public interface A
{
int MyProperty { get; set; }
}
public class B : A
{
public int MyProperty { get; set; }
}
public class C<T> where T : A
{
public Expression<Func<T, bool>> Expression { get; } = a => a.MyProperty > 0;
}
class Program
{
static void Main(string[] args)
{
var cExpression = new C<B>();
Console.WriteLine(cExpression.Expression.ToString());
}
}
The output of the expression: a => (Convert(a, A).MyProperty > 0)
And I decompiled the code above and I got below:
public class C<T> where T : A
{
public Expression<Func<T, bool>> Expression { get; } = (T a) => ((A)a).MyProperty > 0;
}
As you can see, the compiler added a cast option ((A)a) to the expression, that is what I DO NOT want, so the question is how can I tell the compiler NOT to do that?
There is no way to "tell the compiler" not to do this, but you can construct the expression tree directly:
public class C<T> where T : A {
private static readonly MethodInfo s_propGetter = typeof(A).GetProperty(nameof(A.MyProperty)).GetMethod;
public Expression<Func<T, bool>> Expr { get; private set; }
public C() {
var param = Expression.Parameter(typeof(T), "a");
Expr = Expression.Lambda(
typeof(Func<T, bool>),
Expression.GreaterThan(Expression.Property(param, s_propGetter), Expression.Constant(0)),
new[] {param}
) as Expression<Func<T, bool>>;
}
}
Looks like the compiler is emitting the Convert because of the possibility of the generic type parameter to be struct (value type), even though the manual approach demonstrated by another answer proves that it is unnecessary.
But is is what it is. So the only way currently to tell compiler to not generate Convert (cast) is to put class constraint:
where T : class, A
Now the output of the original sample expression will be a => (a.MyProperty > 0)

Moq Extended Setup With Single shared Return

What I want to do is to define one or more Setup(s) to one Return call on multiple method calls of the mocked object; hence avoiding multiple single calls to do Setup().Return()s.
The compiler provides an error when attempting the following which demonstrates the goal, so this is not an appropriate way to achieve that goal.
var mPlatform = new Mock<IPlatformCommunicator>();
mPlatform.Setup(mp => mp.PreStart(It.IsAny<Action<IStatus>>()))
.Setup(mp => mp.Start(It.IsAny<Action<IStatus>>()))
...
.Returns(mockStatusIndeterminate.Object as IStatus);
Is there way to define multiple method calls in Setup to economize the total lines of code?
This is not an option too:
.Setup(mp => mp.PreStart(It.IsAny<Action<IStatus>>()) || mp.Start(It.IsAny<Action<IStatus>>()))
As #Nkosi mentions, not really. But there are options that may work depending on your usage.
SetReturnsDefault
void Main()
{
var fooMock = new Mock<IFoo>();
fooMock.SetReturnsDefault<string>("This is a mocked value");
var foo = fooMock.Object;
Console.WriteLine($"foo.Bar(): {foo.Bar()}");
Console.WriteLine($"foo.Baz(): {foo.Baz()}");
}
public interface IFoo
{
string Bar();
string Baz();
}
Any method that returns a string will return whatever you specify.
The default value provider is similar but across the board.
void Main()
{
var fooMock = new Mock<IFoo>();
fooMock.DefaultValueProvider = new MyDefaultValueProvider();
var foo = fooMock.Object;
Console.WriteLine($"foo.Bar(): {foo.Bar()}");
Console.WriteLine($"foo.Baz(): {foo.Baz()}");
}
public interface IFoo
{
string Bar();
string Baz();
}
public class MyDefaultValueProvider : DefaultValueProvider
{
protected override object GetDefaultValue(Type type, Mock mock)
{
return "This is my default value";
}
}
Create your own extension; the following is a working mvp
void Main()
{
var fooMock = new Mock<IFoo>();
fooMock.Setup(new Expression<Func<IFoo, string>>[] { x => x.Bar(), x => x.Baz() }, "This is a mocked value");
var foo = fooMock.Object;
Console.WriteLine($"foo.Bar(): {foo.Bar()}");
Console.WriteLine($"foo.Baz(): {foo.Baz()}");
}
public interface IFoo
{
string Bar();
string Baz();
}
public static class MoqExtensions
{
public static Mock<T> Setup<T, U>(this Mock<T> self, Expression<Func<T, U>>[] setups, U returns)
where T : class
{
foreach (var setup in setups)
{
self.Setup(setup).Returns(returns);
}
return self;
}
}
All of the above produce the following result
You could create your own extension method on Mock<T> that does what you want:
public static class MoqExt {
public static void SetupReturnOnAll<T, TResult>(
this Mock<T> mock,
TResult returnValue,
params Expression<Func<T, TResult>>[] expressions)
where T: class {
foreach (var expr in expressions)
mock.Setup(expr).Returns(returnValue);
}
}
Usage looks like this:
mPlatform
.SetupReturnOnAll(
mockStatusIndeterminate.Object as IStatus,
mp => mp.PreStart(It.IsAny<Action<IStatus>>()),
mp => mp.Start(It.IsAny<Action<IStatus>>()));
With some extra effort to could improve the interface to this:
mPlatform
.SetupAll(
mp => mp.PreStart(It.IsAny<Action<IStatus>>()),
mp => mp.Start(It.IsAny<Action<IStatus>>()))
.Return(mockStatusIndeterminate.Object as IStatus);
For that you'll need an extra class though:
public class MultiSetup<T, TResult>
where T: class {
public Mock<T> Mock { get; }
public Expression<Func<T, TResult>>[] Expressions { get; }
public MultiSetup(Mock<T> mock, Expression<Func<T, TResult>>[] expressions)
=> (Mock, Expressions) = (mock, expressions);
public void Return(TResult returnValue) {
foreach (var expr in Expressions)
Mock.Setup(expr).Returns(returnValue);
}
}
And you'd create it with this extension:
public static class MoqExt {
public static MultiSetup<T, TResult> SetupAll<T, TResult>(
this Mock<T> mock,
params Expression<Func<T, TResult>>[] expressions)
where T : class
=> new MultiSetup<T, TResult>(mock, expressions);
}

Covariance in generic parameter and convention based on parameter type

I'm really struggling to create interface/convention based rules for FluentValidator. It has following class
abstract class AbstractValidator<T>
{
IRuleBuilderInitial<T, TProperty> RuleFor<TProperty>(Expression<Func<T, TProperty>> expression)
...
}
public interface IWithPropertyA
{
string PropertyA{get;set;}
}
public interface IWithPropertyB
{
string PropertyB{get;set;}
}
public class Handler1Data: IWithPropertyA
{
public string PropertyA {get;set;}
}
public class Handler2Data: IWithPropertyA, IWithPropertyB
{
public string PropertyA {get;set;}
public string PropertyB {get;set;}
}
public class Handler1 : AbstractValidator<Handler1Data> {}
public class Handler2 : AbstractValidator<Handler2Data> {}
I'm trying to create extension method which will basically check if generic argument implements specific interface and then adds rule to it:
public static void ValidateAll<T>(this AbstractValidator<T> validator)
{
(validator as AbstractValidator<IWithPropertyA>)?.RuleFor(x => x.PropertyA).NotEmpty().WithMessage("PropertyA Missing");
(validator as AbstractValidator<IWithPropertyB>)?.RuleFor(x => x.PropertyB).NotEmpty().WithMessage("PropertyB Missing");
}
The problem here is obviously AbstractValidator is not covariant so validator is not castable to neither AbstractValidator<PropertyA> nor AbstractValidator<PropertyB>. I've tried to create my own Base Validator like below and then create extension method based on that but I can't.
public interface IMyValidator<in T>
{
void AddMyRule<TProperty>(Expression<Func<T, TProperty>> expression) //it doesn't work because Expression<Func<T,Property> cannont be covariant
}
public abstract class MyBaseValidator<T>: AbstractValidator<T> ,IMyValidator<T>
{
void AddMyRule<TProperty>(Expression<Func<T, TProperty>> expression)
}
Method will be called in each Handler like this:
public class Handler1 : AbstractValidator<Handler1Data> {
this.ValidateAll();
}
Didn't you mean to write:
(validator as AbstractValidator<IWithPropertyA>)
?.RuleFor(x => x.PropertyA)
.NotEmpty()
.WithMessage("Property1 Missing");
Because Property1 is not defined anywhere.
I found one workaround which avoids using Expression which is obviously the issue here. The downfall of it is that we lose property name and we have to configure message manually.
public interface IMyValidator<out T>
{
void AddMyRule<TProperty>(Func<T, TProperty> expression, string message);
}
public abstract class MyBaseValidator<T> : AbstractValidator<T>, IMyValidator<T>
{
public void AddMyRule<TProperty>(Func<T, TProperty> expression, string message)
{
var exp = FuncToExpression(expression);
RuleFor(exp).NotEmpty().WithMessage(message);
}
private static Expression<Func<T, P>> FuncToExpression<T, P>(Func<T, P> f) => x => f(x);
}
public static class Ext
{
public static void ValidateAll<T>(this AbstractValidator<T> validator)
{
(validator as IMyValidator<IWithPropertyA>)?.AddMyRule(x => x.PropertyA, "PropA Cant be empty");
(validator as IMyValidator<IWithPropertyB>)?.AddMyRule(x => x.PropertyB, "PropB Cant be empty");
}
}
public class Handler1 : MyBaseValidator<Handler1Data>
{
public Handler1()
{
this.ValidateAll();
}
}
public class Handler2 : MyBaseValidator<Handler2Data> { }

Constructor can't accept input type Expression<Func<T, TResult>> in C#

I want to create the class Foo as follows.
The Foo constructor input expression has to get some member of Type T to do some work.
class Foo<T>
{
public Foo<TResult>(Expression<Func<T, TResult>> selector)
{
List<string> memberNames = typeof(TR).GetProperties().Select(p => p.Name).ToList();
....//some work on memberNames
}
}
and create Foo<T> instance by this code:
Foo<ClassA> foo = new Foo<ClassA>(u=>new{u.Property1, u.Property2});
However, this did not work and gave the following error:
Constructor can't have Generic type TResult.
How can I fix this?
Edit:
After struggling to understand the expression I found an answer and have written it as an answer.
You can move the logic outside of consrtuctor to separate method. Instead of using constructor to create Foo - you can use static method, like this:
class Foo<T> {
private Foo() {
}
private void Init<TResult>(Expression<Func<T, TResult>> selector) {
List<string> memberNames = typeof(TResult).GetProperties().Select(p => p.Name).ToList();
//some work on memberNames
}
public static Foo<T> Create<TResult>(Expression<Func<T, TResult>> selector) {
var foo = new Foo<T>();
foo.Init(selector);
return foo;
}
}
And use it like this:
Foo<ClassA> foo = Foo<ClassA>.Create(u=>new{u.Property1, u.Property2});
I solve that as follow
class ClassA
{
public int MyProperty { get; set; }
public int MyProperty2 { get; set; }
}
class Foo<T>
{
public Foo(Expression<Func<T, Object>> selector)
{
var props = ((NewExpression)selector.Body).Members.Select(p => p.Name).ToList();
}
}
Foo<ClassA> foo = new Foo<ClassA>(u => new { u.MyProperty, u.MyProperty2 });

Factory class using generics but without base class

Ive got what I think may be an unusual problem (Ive searched around a lot for an answer, but I dont think Ive found one).
I have messages that are read from a queue and depending on the message type contains a payload that needs to be deserialized into a concrete c# class. This needs to eventually be concrete (I cant use generics the whole way) because Im using Expression Trees to perform Evaluations on the classes that arrive from the queue.
The base class looks like this:
public abstract class BaseRuleMessage<T>
{
public abstract Func<T, bool> CompileRule(Rule r, T msg);
public T Deserialize(ClientEventQueueMessage message)
{
return JsonConvert.DeserializeObject<T>(message.Payload);
}
public BaseRuleMessage()
{
RulesCompleted = new List<int>();
}
public IEnumerable<Rule> FilterRules(RuleGroup ruleGroup)
{
return ruleGroup.Rules.Where(item =>
!RulesCompleted.Any(r => r.Equals(item.Id)));
}
I implement the base class like this:
public class UiTransactionUpdate : BaseRuleMessage<UiTransactionUpdate>
{
public override Func<UiTransactionUpdate, bool> CompileRule(Rule r, UiTransactionUpdate msg)
{
var expression = Expression.Parameter(typeof(UiTransactionUpdate));
Expression expr = BuildExpr(r, expression, msg);
return Expression.Lambda<Func<UiTransactionUpdate, bool>>(expr, expression).Compile();
}
public Guid TransactionId { get; set; }
public Guid GroupId { get; set; }
public decimal StatusValue { get; set; }
I then do something like this to call:
switch (message.MessageType)
{
case "UI_UPDATE":
{
message.Payload = RemoveNullGroupIdReference(jsonPayload, message.Payload);
var deserializedMessage = new UiTransactionUpdate().Deserialize(message);
deserializedMessage.RulesCompleted = deserializedMessage.RulesCompleted ?? new List<int>();
foreach (var rule in deserializedMessage.FilterRules(ruleGroup))
{
What I really want to know is how can I create a factory (or can I?) to be able to define the implementation of the base class in such a way that I can return a concrete class to use for my expression tree evaluations without having to repeat all the calling code for each type.
I avoided using dynamic but this meant that I had pass the object around as an object. I prefer not to use dynamic but in this case, casting objects at run-time may not be any better.
I also had to change the code so that instead of returning a Func<T, bool>, there is a method that would execute the Func. This was to avoid referring to the generic class. I'm not sure if you actually need the Func in your actual implementation.
I had to create a new base class that wasn't generically typed.
// Horrible name, do change it to something more appropriate
public abstract class BaseBaseRuleMessage
{
public IList<int> RulesCompleted { get; set; }
public IEnumerable<Rule> FilterRules(RuleGroup ruleGroup)
{
return ruleGroup.Rules.Where(item =>
!RulesCompleted.Any(r => r.Equals(item.Id)));
}
public BaseBaseRuleMessage DeserializeToBaseBaseRuleMessage(ClientEventQueueMessage message)
{
return (BaseBaseRuleMessage) DeserializeToType(message);
}
protected abstract object DeserializeToType(ClientEventQueueMessage message);
public abstract bool ExecuteRule(Rule rule, object msg);
}
Updated the BaseRuleMessage to derive from BaseBaseRuleMessage (and moved some properties to the base class.
public abstract class BaseRuleMessage<T> : BaseBaseRuleMessage
where T : BaseRuleMessage<T>
{
public abstract Func<T, bool> CompileRule(Rule r, T msg);
protected override object DeserializeToType(ClientEventQueueMessage message)
{
return JsonConvert.DeserializeObject(message.Payload, typeof(T));
}
protected BaseRuleMessage()
{
RulesCompleted = new List<int>();
}
public override bool ExecuteRule(Rule rule, object msg)
{
var message = (T) msg;
if (message == null)
{
throw new InvalidOperationException();
}
return CompileRule(rule, message).Invoke(message);
}
}
The concrete class is basically the same. I've implemented my own BuildExpr to make sure the code can compile.
public class UiTransactionUpdate : BaseRuleMessage<UiTransactionUpdate>
{
public override Func<UiTransactionUpdate, bool> CompileRule(Rule r, UiTransactionUpdate msg)
{
var expression = Expression.Parameter(typeof(UiTransactionUpdate));
Expression expr = BuildExpr(r, expression, msg);
return Expression.Lambda<Func<UiTransactionUpdate, bool>>(expr, expression).Compile();
}
public Guid TransactionId { get; set; }
public Guid GroupId { get; set; }
public decimal StatusValue { get; set; }
private Expression BuildExpr(Rule rule, ParameterExpression parameterExpression, UiTransactionUpdate message)
{
var transactionIdProperty = Expression.Property(parameterExpression, "TransactionId");
var value = Expression.Constant(rule.TransactionId);
return Expression.Equal(transactionIdProperty, value);
}
}
To use it:
var messageTypeToTypeMap = new Dictionary<string, Func<BaseBaseRuleMessage>>
{
{"UI_UPDATE", () => new UiTransactionUpdate()}
};
var factoryFunc = messageTypeToTypeMap[message.MessageType];
message.Payload = RemoveNullGroupIdReference(jsonPayload, message.Payload);
var ruleMessage = factoryFunc.Invoke();
var deserializedMessage = ruleMessage.DeserializeToBaseBaseRuleMessage(message);
deserializedMessage.RulesCompleted = deserializedMessage.RulesCompleted ?? new List<int>();
foreach (var rule in deserializedMessage.FilterRules(ruleGroup))
{
var isTrue = deserializedMessage.ExecuteRule(rule, deserializedMessage);
}

Categories