Related
Given:
static TDest Gimme<TSource,TDest>(TSource source)
{
return default(TDest);
}
Why can't I do:
string dest = Gimme(5);
without getting the compiler error:
error CS0411: The type arguments for method 'Whatever.Gimme<TSource,TDest>(TSource)' cannot be inferred from the usage. Try specifying the type arguments explicitly.
The 5 can be inferred as int, but there's a restriction where the compiler won't/can't resolve the return type as a string. I've read in several places that this is by design but no real explanation. I read somewhere that this might change in C# 4, but it hasn't.
Anyone know why return types cannot be inferred from generic methods? Is this one of those questions where the answer's so obvious it's staring you in the face? I hope not!
The general principle here is that type information flows only "one way", from the inside to the outside of an expression. The example you give is extremely simple. Suppose we wanted to have type information flow "both ways" when doing type inference on a method R G<A, R>(A a), and consider some of the crazy scenarios that creates:
N(G(5))
Suppose there are ten different overloads of N, each with a different argument type. Should we make ten different inferences for R? If we did, should we somehow pick the "best" one?
double x = b ? G(5) : 123;
What should the return type of G be inferred to be? Int, because the other half of the conditional expression is int? Or double, because ultimately this thing is going to be assigned to double? Now perhaps you begin to see how this goes; if you're going to say that you reason from outside to inside, how far out do you go? There could be many steps along the way. See what happens when we start to combine these:
N(b ? G(5) : 123)
Now what do we do? We have ten overloads of N to choose from. Do we say that R is int? It could be int or any type that int is implicitly convertible to. But of those types, which ones are implicitly convertible to an argument type of N? Do we write ourselves a little prolog program and ask the prolog engine to solve what are all the possible return types that R could be in order to satisfy each of the possible overloads on N, and then somehow pick the best one?
(I'm not kidding; there are languages that essentially do write a little prolog program and then use a logic engine to work out what the types of everything are. F# for example, does way more complex type inference than C# does. Haskell's type system is actually Turing Complete; you can encode arbitrarily complex problems in the type system and ask the compiler to solve them. As we'll see later, the same is true of overload resolution in C# - you cannot encode the Halting Problem in the C# type system like you can in Haskell but you can encode NP-HARD problems into overload resolution problems.) (See below)
This is still a very simple expression. Suppose you had something like
N(N(b ? G(5) * G("hello") : 123));
Now we have to solve this problem multiple times for G, and possibly for N as well, and we have to solve them in combination. We have five overload resolution problems to solve and all of them, to be fair, should be considering both their arguments and their context type. If there are ten possibilities for N then there are potentially a hundred possibilities to consider for N(N(...)) and a thousand for N(N(N(...))) and very quickly you would have us solving problems that easily had billions of possible combinations and made the compiler very slow.
This is why we have the rule that type information only flows one way. It prevents these sorts of chicken and egg problems, where you are trying to both determine the outer type from the inner type, and determine the inner type from the outer type and cause a combinatorial explosion of possibilities.
Notice that type information does flow both ways for lambdas! If you say N(x=>x.Length) then sure enough, we consider all the possible overloads of N that have function or expression types in their arguments and try out all the possible types for x. And sure enough, there are situations in which you can easily make the compiler try out billions of possible combinations to find the unique combination that works. The type inference rules that make it possible to do that for generic methods are exceedingly complex and make even Jon Skeet nervous. This feature makes overload resolution NP-HARD.
Getting type information to flow both ways for lambdas so that generic overload resolution works correctly and efficiently took me about a year. It is such a complex feature that we only wanted to take it on if we absolutely positively would have an amazing return on that investment. Making LINQ work was worth it. But there is no corresponding feature like LINQ that justifies the immense expense of making this work in general.
UPDATE: It turns out that you can encode arbitrarily difficult problems in the C# type system. C# has nominal generic subtyping with generic contravariance, and it has been shown that you can build a Turing Machine out of generic type definitions and force the compiler to execute the machine, possibly going into infinite loops. At the time I wrote this answer the undecidability of such type systems was an open question. See https://stackoverflow.com/a/23968075/88656 for details.
You have to do:
string dest = Gimme<int, string>(5);
You need to specify what your types are in the call to the generic method. How could it know that you wanted a string in the output?
System.String is a bad example because it's a sealed class, but say it wasn't. How could the compiler know that you didn't want one of its subclasses instead if you didn't specify the type in the call?
Take this example:
System.Windows.Forms.Control dest = Gimme(5);
How would the compiler know what control to actually make? You'd need to specify it like so:
System.Windows.Forms.Control dest = Gimme<int, System.Windows.Forms.Button>(5);
Calling Gimme(5) ignoring the return value is a legal statement how would the compiler know which type to return?
I use this technique when I need to do something like that:
static void Gimme<T>(out T myVariable)
{
myVariable = default(T);
}
and use it like this:
Gimme(out int myVariable);
Print(myVariable); //myVariable is already declared and usable.
Note that inline declaration of out variables is available since C# 7.0
This was a design decision I guess. I also find it useful while programming in Java.
Unlike Java, C# seems to evolve towards a functional programming language, and you can get type inference the other way round, so you can have:
var dest = Gimme<int, string>(5);
which will infer the type of dest. I guess mixing this and the java style inference could prove to be fairly difficult to implement.
If a function is supposed to return one of a small number of types, you could have it return a class with defined widening conversions to those types. I don't think it's possible to do that in a generic way, since the widening ctype operator doesn't accept a generic type parameter.
public class ReturnString : IReq<string>
{
}
public class ReturnInt : IReq<int>
{
}
public interface IReq<T>
{
}
public class Handler
{
public T MakeRequest<T>(IReq<T> requestObject)
{
return default(T);
}
}
var handler = new Handler();
string stringResponse = handler.MakeRequest(new ReturnString());
int intResponse = handler.MakeRequest(new ReturnInt());
I have a static method:
public class Example
{
//for demonstration purposes - just returns default(T)
public static T Foo<T>() { return default(T); }
}
And I need to be able to invoke it using a Type parameter calls to which could be numerous, so my standard pattern is to create a thread-safe cache of delegates (using ConcurrentDictionary in .Net 4) which dynamically invoke the Foo<T> method with the correct T. Without the caching, though, the code is this:
static object LateFoo(Type t)
{
//creates the delegate and invokes it in one go
return (Func<object>)Delegate.CreateDelegate(
typeof(Func<object>),
typeof(Example).GetMethod("Foo", BindingFlags.Public | BindingFlags.Static).
MakeGenericMethod(t))();
}
This is not the first time I've had to do this - and in the past I have use Expression trees to build and compile a proxy to invoke the target method - to ensure that return type conversion and boxing from int -> object (for example) is handled correctly.
Update - example of Expression code that works
static object LateFoo(Type t)
{
var method = typeof(Example)
.GetMethod("Foo", BindingFlags.Public | BindingFlags.Static)
.MakeGenericMethod(t);
//in practise I cache the delegate, invoking it freshly built or from the cache
return Expression.Lambda<Func<IField, object>>(Expression.Convert(
Expression.Call(method), typeof(object))).Compile()();
}
What's slightly amusing is that I learned early on with expressions that an explicit Convert was required and accepted it - and in lieu of the answers here it does now make sense why the .Net framework doesn't automatically stick the equivalent in.
End update
However, this time I thought I'd just use Delegate.CreateDelegate as it makes great play of the fact that (from MSDN):
Similarly, the return type of a delegate is compatible with the return type of a method if the return type of the method is more restrictive than the return type of the delegate, because this guarantees that the return value of the method can be cast safely to the return type of the delegate.
Now - if I pass typeof(string) to LateFoo method, everything is fine.
If, however, I pass typeof(int) I get an ArgumentException on the CreateDelegate call, message: Error binding to target method. There is no inner exception or further information.
So it would seem that, for method binding purposes, object is not considered more restrictive than int. Obviously, this must be to do with boxing being a different operation than a simple type conversion and value types not being treated as covariant to object in the .Net framework; despite the actual type relationship at runtime.
The C# compiler seems to agree with this (just shortest way I can model the error, ignore what the code would do):
public static int Foo()
{
Func<object> f = new Func<object>(Foo);
return 0;
}
Does not compile because the Foo method 'has the wrong return type' - given the CreateDelegate problem, C# is simply following .Net's lead.
It seems to me that .Net is inconsistent in it's treatment of covariance - either a value type is an object or it's not; & if it's not it should not expose object as a base (despite how much more difficult it would make our lives). Since it does expose object as a base (or is it only the language that does that?), then according to logic a value type should be covariant to object (or whichever way around you're supposed to say it) making this delegate bind correctly. If that covariance can only be achieved via a boxing operation; then the framework should take care of that.
I dare say the answer here will be that CreateDelegate doesn't say that it will treat a box operation in covariance because it only uses the word 'cast'. I also expect there are whole treatises on the wider subject of value types and object covariance, and I'm shouting about a long-defunct and settled subject. I think there's something I either don't understand or have missed, though - so please enlighten!
If this is unanswerable - I'm happy to delete.
You can only convert a delegate in this way if the parameters and return value can be converted using a representation conserving conversion.
Reference types can only be converted to other reference types in this way
Integral values can be converted to other integer values of the same size (int, uint, and enums of the same size are compatible)
A few more relevant blog articles:
This dichotomy motivates yet another classification scheme for conversions (†). We can divide conversions into representation-preserving conversions (B to D) and representation-changing conversions (T to U). (‡) We can think of representation-preserving conversions on reference types as those conversions which preserve the identity of the object. When you cast a B to a D, you’re not doing anything to the existing object; you’re merely verifying that it is actually the type you say it is, and moving on. The identity of the object and the bits which represent the reference stay the same. But when you cast an int to a double, the resulting bits are very different.
This is why covariant and contravariant conversions of interface and delegate types require that all varying type arguments be of reference types. To ensure that a variant reference conversion is always identity-preserving, all of the conversions involving type arguments must also be identity-preserving. The easiest way to ensure that all the non-trivial conversions on type arguments are identity-preserving is to restrict them to be reference conversions.
http://blogs.msdn.com/b/ericlippert/archive/2009/03/19/representation-and-identity.aspx
"but how can a value type, like int, which is 32 bits of memory, no more, no less, possibly inherit from object? An object laid out in memory is way bigger than 32 bits; it's got a sync block and a virtual function table and all kinds of stuff in there." Apparently lots of people think that inheritance has something to do with how a value is laid out in memory. But how a value is laid out in memory is an implementation detail, not a contractual obligation of the inheritance relationship! When we say that int inherits from object, what we mean is that if object has a member -- say, ToString -- then int has that member as well.
http://ericlippert.com/2011/09/19/inheritance-and-representation/
It seems to me that .Net is inconsistent in it's treatment of covariance - either a value type is an object or it's not; if it's not it should not expose object as a base
It depends on what the meaning of "is" is, as President Clinton famously said.
For the purposes of covariance, int is not object because int is not assignment compatible with object. A variable of type object expects a particular bit pattern with a particular meaning to be stored in it. A variable of type int expects a particular bit pattern with a particular meaning, but a different meaning than the meaning of a variable of object type.
However, for the purposes of inheritance, an int is an object because every member of object is also a member of int. If you want to invoke a method of object -- ToString, say -- on int, you are guaranteed that you can do so, because an int is a kind of object, and an object has ToString.
It is unfortunate, I agree, that the truth value of "an int is an object" varies depending on whether you mean "is assignment-compatible with" or "is a kind of".
If that covariance can only be achieved via a boxing operation; then the framework should take care of that.
OK. Where? Where should the boxing operation go? Someone, somewhere has to generate a hunk of IL that has a boxing instruction. Are you suggesting that when the framework sees:
Func<int> f1 = ()=>1;
Func<object> f2 = f1;
then the framework should automatically pretend that you said:
Func<object> f2 = ()=>(object)f1();
and thereby generate the boxing instruction?
That's a reasonable feature, but what are the consequences? Func<int> and Func<object> are reference types. If you do f2 = f1 on reference types like this, do you not expect that f2 and f1 have reference identity? Would it not be exceedingly strange for this test case to fail?
f2 = f1;
Debug.Assert(object.ReferenceEquals(f1, f2));
Because if the framework implemented that feature, it would.
Similarly, if you said:
f1 = MyMethod;
f2 = f1;
and you asked the two delegates whether they referred to the same method or not, would it not be exceedingly weird if they referred to different methods?
I think that would be weird. However, the VB designers do not. If you try to pull shenanigans like that in VB, the compiler will not stop you. The VB code generator will generate non-reference-equal delegates for you that refer to different methods. Try it!
Moral of the story: maybe C# is not the language for you. Maybe you prefer a language like VB, where the language is designed to take a "make a guess about what the user probably meant and just make it work" attitude. That's not the attitude of the C# designers. We are more "tell the user when something looks suspiciously wrong and let them figure out how they want to fix it" kind of people.
Even though I think #CodeInChaos is absolutely right, I can't help pointing this Eric Lippert's blog post out. In reply to the last comment to his post (at the very bottom of the page) Eric explains the rationale for such behaviour, and I think this is exactly what you're interested in.
UPDATE: As #Sheepy pointed out Microsoft moved old MSDN blogs into archive and removed all comments. Luckily, the Wayback Machine preserved the blog post in its original form.
This question gives me curiosity... When you want to define a type you must say GetType(Type) ex.: GetType(string), but ain't String a type itself?
Why do you need to use GetType in those situations? And, if the reason is because it is expecting a Type 'Type'... why isn't the conversion implicit... I mean, all the data is there...
What you're doing is getting a reference to the meta-data of the type ... it might be a little more obvious if you look at the C# version of the API, which is typeof(string) ... which returns a Type object with information about the string type.
You would generally do this when using reflection or other meta-programming techniques
string is type, int is type and Type is type and they are not the same. but about why there is no implicit conversion it's not recommended by MSDN:
By eliminating unnecessary casts,
implicit conversions can improve
source code readability. However,
because implicit conversions can occur
without the programmer's specifying
them, care must be taken to prevent
unpleasant surprises. In general,
implicit conversion operators should
never throw exceptions and never lose
information so that they can be used
safely without the programmer's
awareness. If a conversion operator
cannot meet those criteria, it should
be marked explicit.
Take attention to :
never lose information so that they
can be used safely without the
programmer's awareness
When you want to define a type you must say GetType(Type) ex.: GetType(string)...
That's not true. Every time you do any of the following
class MyClass
{
///...
}
class MyChildClass : MyClass
{
}
struct MyStruct
{
///...
}
you're defining a new type.
if the reason is because it is expecting a Type 'Type'... why isn't the conversion implicit... I mean, all the data is there...
One reason for this is polymorphism. For instance, if we were allowed to do the following:
MyChildClass x;
....GetType(x)
GetType(x) could return MyChildClass, MyClass, or Object, since x is really an instance of all of those types.
It's also worth noting that Type is itself a class (ie, it inherits from Object), so you can inherit from it. Although I'm not sure why you'd want to do this other than overriding the default reflection behavior (for instance, to hide the internals from prying eyes).
GetType(string) will return the same information. Look at it like you would a constant. The only other way to get the Type object that represents a string would be to instantiate the string object and call o.GetType(). Also, this is not possible for interfaces and abstract types.
If you want to know the runtime type of a variable, call the .GetType() method off of it, as the runtime type may not be the same as the declared type of the variable.
An example would be:
XNamespace ns = "my namespace"
Why not?:
XNamespace ns = new XNamespace ( "my namespace" )
What's the idea behind using implicit/explicit convertions instead of constructors? Convenience?
Is there a guideline for this?
Convenience?
More or less, yes. Consider the case for when you’ve got a number-like object (say, a Complex) on which you do calculations. Clearly, writing code such as:
Complex result = c1 * new Complex(2) + new Complex(32);
is very annoying and hard to read. Implicit conversions help here (an alternative would be operator overloads in this example, but that would lead to lots of similar overloads).
Is there a guideline for this?
Provide as few implicit conversions as possible, since they may hide problems. Implicit conversion reduce explicitness by the same amount by which they increase terseness. Sometimes this is good, but sometimes not.
I find it best to restrict implicit conversions to very similar types, such as the number-like objects in my example above: an int essentially is-a Complex (from a mathematical standpoint; even if it’s not modelled via inheritance), hence an implicit conversion makes sense.
In VB, an implicit conversion is called “Widening” (as opposed to Narrowing, which is explicit) and this describes it well: no information is lost in the course of the conversion.
Furthermore, an operator is essentially a builder function, and has (some of) the usual advantages of a builder function over a constructor: namely, it can re-use cached values instead of always creating new instances.
Consider my Complex example. We may want to cache values for often-used Complex numbers:
Class Complex {
// Rest of implementation.
private static Complex[] cache = new[] {
new Complex(-1), new Complex(0), new Complex(1) };
public implicit operator Complex(int value) {
if (value >= -1 && value <= 1)
return cache[value];
else
return new Complex(value);
}
}
Of course, whether this micro-optimization is effective is another question.
One of the reasons behind using implicit conversion with such simple types as XName is, I believe, convenience in calling methods.
For example, you can write
var info = root.Elements ("user").Element ("info").Value;
Simplicity at extracting data is what LINQ is all about, and if we had to write
var info = root.Elements (new XName ("user")).Element (new XName ("info")).Value;
even for simplest queries, would LINQ be totally worth it for complex ones?
Another important issue here is that XNames are atomized.
See MSDN:
XName objects are guaranteed to be atomized; that is, if two XName objects have exactly the same namespace and exactly the same local name, they will share the same instance. The equality and comparison operators are also provided explicitly for this purpose.
Among other benefits, this feature allows for faster execution of queries. When filtering on the name of elements or attributes, the comparisons expressed in predicates use identity comparison, not value comparison. It is much faster to determine that two references actually refer to the same object than to compare two strings.
You can't provide atomization in constructor, but defining a conversion allows you to pick corresponding object from the pool and return it as if it were a new instance.
The use of implicit / explicit conversions is issue of convenience and one that many programming guidelines suggest you avoid in favor of explicit ConvertToXXX methods.
One of the problems is tha the use of implicit / explicit conversions further overloads functions of the casting operator. It gives it the dual purpose of
Viewing the same object through a different type / interface in the object's hierarchy
Coverting the object to a new type altogether
Unfortunately C# already does the latter in other areas (with primitives and boxing).
If two classes should be convertible to one another, but they do not share an interface of a base class that allows this behavior automatically, you would use the conversions. Implicit conversions should never have a possibility of data loss; they are often considered "widening" conversions. For example, converting an int to a long is a widening conversion, and there is no problem inherent in the conversion being implicit. Explicit conversions may involve the possibility of data loss; a long may or may not be convertible to an int, depending on its value.
One trick I have used with implicit conversions is to convert classes in different namespaces to each other when I did not have another reasonable option. For example, one WCF service returns an AuthenticationToken object that I need to pass to a WCF service in a different namespace. Both have this AuthenticationToken object, and constant conversion would have been a pain. My solution involved using public static implicit operator in a partial class to add the functionality to convert each way.
Personally, I use the conversions when I know that the rhs may be converted into a static member of a class (like saying color = "red" to imply color = Colors.Red)
I use the new operator when I intend to actually create a new instance.
Section 6.1 Implicit conversions defines an identity conversion thusly:
An identity conversion converts from any type to the same type. This conversion exists such that an entity that already has a required type can be said to be convertible to that type.
Now, what is the purpose of sentences such as these?
(In §6.1.6 Implicit reference conversions)
The implicit reference conversions are:
[...]
From any reference-type to a reference-type T if it has an implicit identity or reference conversion to a reference-type T0 and T0 has an identity conversion to T.
and:
(In §6.1.7 Boxing conversions)
A value type has a boxing conversion to an interface type I if it has a boxing conversion to an interface type I0 and I0 has an identity conversion to I.
Initially they seem redundant (tautologous). But they must be there for a purpose, so why are they there?
Can you give an example of two types T1, T2 such that T1 would not be implicitly convertible to T2 if it weren’t for the above-quoted paragraphs?
Update on 22-Sep-2010:
I doubt anybody is going to read this besides Timwi. Even so, I wanted to make a few edits to this answer in light of the fact that a new answer has now been accepted and the debate still continues (at least in my perhaps imaginary world) on whether or not the quoted excerpts of the spec are technically redundant. I am not adding much, but it's too substantial to fit in a comment. The bulk of the update can be found under the heading "Conversion involving the dynamic type" below.
Update on 19-Sep-2010:
In your comment:
[T]his doesn’t make sense.
Damn, Timwi, you say that a lot. But all right, then; you've put me on the defensive, so here goes!
Disclaimer: I have definitely not examined the spec as closely as you have. Based on some of your recent questions it seems like you've been looking into it quite a bit lately. This is naturally going to make you more familiar with a lot of details than most users on SO; so this, like most answers you're likely to receive from anyone other than Eric Lippert, may not satisfy you.
Different premises
Firstly, the premise of your question is that if the statements highlighted are redundant, then they serve no purpose. My answer's premise is that redundant statements are not necessarily without purpose if they clarify something that isn't obvious to everyone. These are contradictory premises. And if we can't agree on premises, we can't have a straightforward logical argument. I was simply asking you to rethink your premise.
Your response, however, was to reiterate your premise: "If the sentences are truly redundant, then they only confuse the reader and don't clarify anything."
(I like how you set yourself up as the representative for all readers of the spec there, by the way.)
I can't blame you for holding this position, exactly. I mean, it does seem obvious. And I didn't give any concrete examples in my original answer. So below I will try to include some concrete examples. But first, let me take a step back and offer my take on why this weird identity conversion concept exists in the spec in the first place.
The purpose of the identity conversion definition
Upon first glance, this definition seems rather superfluous; isn't it just saying that an instance of any type T is convertible to ... well, to T? Yes, it is. But I hypothesize* that the purpose of this definition is to provide the spec with the proper vocabulary to utilize the concept of type identity in the context of discussing conversions.
This allows for statements about conversions which are essentially transitive in nature. The first point you quoted from the spec as an example of a tautological statement falls into this category. It says that if an implicit conversion is defined for some type (I'll call it K) to another type T0 and T0 has an identity conversion to T, then K is implicitly convertible to T. By the definition of identity conversion given above, "has an identity conversion to" really means "is the same type as." So the statement is redundant.
But again: the identity conversion definition exists in the first place to equip the spec with a formal language for describing conversions without having to say things like "if T0 and T are really the same type."
OK, time for concrete examples.
Where the existence of an implicit conversion might not be obvious to some developers
Note: A much better example has been provided by Eric Lippert in his answer to the question. I leave these first two examples as only minor reinforcements of my point. I have also added a third example that concretizes the identity conversion that exists between object and dynamic as pointed out in Eric's answer.
Transitive reference conversion
Let's say you have two types, M and N, and you've got an implicit conversion defined like this:
public static implicit operator M(N n);
Then you can write code like this:
N n = new N();
M m = n;
Now let's say you've got a file with this using statement up top:
using K = M;
And then you have, later in the file:
N n = new N();
K k = n;
OK, before I proceed, I realize that this is obvious to you and me. But my answer is, and has been from the beginning, that it might not be obvious to everyone, and therefore specifying it--while redundant--still has a purpose.
That purpose is: to make clear to anyone scratching his or her head, looking at that code, it is legal. An implicit conversion exists from N to M, and an identity conversion exists from M to K (i.e., M and K are the same type); so an implicit conversion exists from N to K. It isn't just logical (though it may be logical); it's right there in the spec. Otherwise one might mistakenly believe that something like the following would be necessary:
K k = (M)n;
Clearly, it isn't.
Transitive boxing conversion
Or take the type int. An int can be boxed as an IComparable<int>, right? So this is legal:
int i = 10;
IComparable<int> x = i;
Now consider this:
int i = 10;
IComparable<System.Int32> x = i;
Again, yes, it may be obvious to you, me, and 90% of all developers who might ever come across it. But for that slim minority who don't see it right away: a boxing conversion exists from int to IComparable<int>, and an identity conversion exists from IComparable<int> to IComparable<System.Int32> (i.e., IComparable<int> and IComparable<System.Int32> are the same type); so a boxing conversion exists from int to IComparable<System.Int32>.
Conversion involving the dynamic type
I'm going to borrow from my reference conversion example above and just tweak it slightly to illustrate the identity relation between object and dynamic in version 4.0 of the spec.
Let's say we have the types M<T> and N, and have defined somewhere the following implicit conversion:
public static implicit operator M<object>(N n);
Then the following is legal:
N n = new N();
M<dynamic> m = n;
Clearly, the above is far less obvious than the two previous examples. But here's the million-dollar question: would the above still be legal even if the excerpts from the spec quoted in the question did not exist? (I'm going to call these excerpts Q for brevity.) If the answer is yes, then Q is in fact redundant. If no, then it is not.
I believe the answer is yes.
Consider the definition of identity conversion, defined in section 6.1.1 (I am including the entire section here as it is quite short):
An identity conversion converts from any type to the same type. This conversion exists such that an entity that already has a required type can be said to be convertible to that type.
Because object and dynamic are considered equivalent there is an identity conversion between object and dynamic, and between constructed types that are the same when replacing all occurences of dynamic with object. [emphasis mine]
(This last part is also included in section 4.7, which defines the dynamic type.)
Now let's look at the code again. In particular I'm interested in this one line:
M<dynamic> m = n;
The legality of this statement (disregarding Q -- remember, the issue being discussed is the hypothetical legality of the above statement if Q did not exist), since M<T> and N are custom types, depends on the existence of a user-defined implicit conversion between N and M<dynamic>.
There exists an implicit conversion from N to M<object>. By the section of the spec quoted above, there is an identity conversion between M<object> and M<dynamic>. By the definition of identity conversion, M<object> and M<dynamic> are the same type.
So, just as in the first two (more obvious) examples, I believe it is true that an implicit conversion exists from N to M<dynamic> even without taking Q into account, just as it is true that an implicit conversion exists from N to K in the first example and that a boxing conversion exists from int to IComparable<System.Int32> in the second example.
Without Q, this is much less obvious (hence Q's existence); but that does not make it false (i.e., Q is not necessary for this behavior to be defined). It just makes it less obvious.
Conclusion
I said in my original answer that this is the "obvious" explanation, because it seemed to me you were barking up the wrong tree. You initially posed this challenge:
Can you give an example of two types T1, T2 such that T1 would not be implicitly convertible to T2 if it weren’t for the above-quoted paragraphs?
No one's going to meet this challenge, Timwi, because it's impossible. Take the first excerpt about reference conversions. It is saying that a type K is implicitly convertible to a type T if it is implicitly convertible to T0 and T0 is the same as T. Deconstruct this, put it back together, and you're left with an obvious tautology: K is implicitly convertible to T if it's implicitly convertible to T. Does this introduce any new implicit conversions? Of course not.
So maybe Ben Voigt's comment was correct; maybe these points that you're asking about would've been better placed in footnotes, rather than in the body of the text. In any case, it's clear to me that they are redundant, and so to start with the premise they cannot be redundant, or else they wouldn't be there is to embark on a fool's errand. Be willing to accept that a redundant statement may still shed some light on a concept that may not be obvious to everyone, and it will become easier to accept these statements for what they are.
Redundant? Yes. Tautologous? Yes. Pointless? In my opinion, no.
*Obviously, I did not have any part in writing the C# language specification. If I did, this answer would be a lot more authoritative. As it is, it simply represents one guy's feeble attempt to make sense of a rather complex document.
Original answer
I think you're (perhaps intentionally) overlooking the most obvious answer here.
Consider these two sentences in your question:
(1) Initially they seem redundant
(tautologous). (2) But they must be there
for a purpose, so why are they there?
To me, the implication of these two sentences together is that a tautologous statement serves no purpose. But just because a statement follows logically from established premises, that does not make it obvious to everyone. In other words even if (1) is true, the answer to (2) may simply be: to make the described behavior clear to anyone reading the spec.
Now you might argue that even if something is not obvious, it still does not belong in a specification if it is providing a redundant definition. To this potential objection, I can only say: be realistic. It's not really practical (in my opinion) to comb through a document stripping out all statements which are simply stating facts that could have been deduced from prior statements.
If this were a common practice, I think you'd find a lot of literature out there -- not just specs, but research papers, articles, textbooks, etc. -- would be a lot shorter, denser, and more difficult to understand.
So: yes, perhaps they are redundant. But that does not negate their purpose.
Section 4.7 of the specification notes that there is an identity conversion from Foo<dynamic> to Foo<object> and vice versa. The portion of the spec you quoted is written to ensure that this case is handled. That is, if there is an implicit reference conversion from T to C<object, object> then there is also an implicit reference conversion to C<object, dynamic>, C<dynamic, object> and C<dynamic, dynamic>.
One might reasonably point out that (1) the intention of these phrases is unobvious - hence your question - and confusing, and (2) that the section on identity conversions ought to cross-reference the section on dynamic conversions, and (3) phrases like this in the spec make it difficult for an implementor of the specification to clearly translate the spec language into an implementation. How is one to know if any such type exists? The spec need not specify exact algorithms, but it would be nice if it gave more guidance.
The spec is, sadly, not a perfect document.
An identity conversion converts from
any type to the same type. This
conversion exists such that an entity
that already has a required type can
be said to be convertible to that
type.
This says that in C#-land, 1==1; a Spade is a Spade. This is the basis of assigning an object reference to a variable of the same type; if a variable typed T1 and one typed T2 are in reality both Spades, it is possible to assign one to the other without having to explicitly cast one as a Spade. Imagine a C# variant where assignments had to look like this:
Spade mySpade = new Spade();
Spade mySpade2;
mySpade2 = (Spade)mySpade; //explicit cast required
Also, an "identity" in mathematics states that an expression that evaluates to a result given a set of inputs is equivalent to another expression that produces the same result given the same inputs. In programming, this means that an expression or function that evaluates to an instance of a type is equivalent to that type, without explicit conversion. If that didn't hold, the following code would be required:
public int myMethod() { /*Do something*/ }
...
int myInt = (int)myMethod(); //required even though myMethod() evals to an int.
...
int myInt = (int)(1 + 2); //required even though 1, 2, and 1+2 eval to an int.
The second rule basically says that a value type can be assigned to a member variable on a class if, in part, the member variable (a boxed type by definition, as its container is a reference type) is declared to be the same type. If this rule were not specified, theoretically, a version of C# could exist in which pure value types would have to be explicitly converted to their reference analog in order to be stored as a member variable on a class. Imagine, for example, a version of C# in which the blue keyword types (int, float, decimal) and the light blue class names (Int32, Float, Decimal) referred to two very different, only-explicitly-convertible types, and int, float, decimal etc. were not legal as member variable types because they were not reference types:
public class MyClass
{
Int32 MyBoxedValueType; //using "int" not legal
}
...
MyClass myClass = new MyClass();
int theInt = 2;
myClass.MyBoxedValueType = (Int32)theInt; //explicit cast required
I know it sounds silly, but at some level, these things must be known, and in computers, you have to specify EVERYTHING. Read the USA Hockey rulebook for ice hockey sometime; the very first rule in the book is that the game shall be played on an ice surface. It's one of the ultimate "well duhs", but also a fundamental truth of the game that must be understood in order for any other rule to make sense.
May it is such that the code guarantees pass-through when called like Convert.ChangeType(client, typeof(Client)) regardless if IConvertible is implemented.
Look into the source of ChangeType from mscorlib with Reflector and notice the conditions at which value is returned as-is.
Remember a = operator is not a conversion, just a reference set. So code like Client client_2 = client_1 does not perform any implicit conversions. If an identity implicit conversion is declared then error CS0555 is issued.
I guess the spec says let the C# compiler handle such cases, and thus dot not manually try to define identity conversions.