|
dev
newsgroups
|
|||||||||||||||||||||||
|
|||||||||||||||||||||||
Why does the generic constraint new() use Activator<T>.CreateInstance instead of new T()?instead of new T()? An example public class Foo<T> : where T : class, new() { T objectValue = new T(); public foo() {} } If you look at Lutz's reflector you will get something like T objectValue = Activator<T>.CreateInstance() in place of T objectValue = new T(); Using the Activator class is obviously slower and unintuitive to me, so why does the framework compile my code using Activator<T>.CreateInstance() instead of new T()? >Using the Activator class is obviously slower and unintuitive to me, new T usually compiles to the newobj IL instruction (when T is a>so why does the framework compile my code using >Activator<T>.CreateInstance() instead of new T()? reference type). But newobj doesn't support generic type arguments. Mattias -- Mattias Sjögren [C# MVP] mattias @ mvps.org http://www.msjogren.net/dotnet/ | http://www.dotnetinterop.com Please reply only to the newsgroup. Mattias Sjögren wrote:
> >Using the Activator class is obviously slower and unintuitive to me, Actually, the CLR does if you constrain the type argument to reference> >so why does the framework compile my code using > >Activator<T>.CreateInstance() instead of new T()? > > new T usually compiles to the newobj IL instruction (when T is a > reference type). But newobj doesn't support generic type arguments. types: ..method private hidebysig static void Foo<class .ctor T>() cil managed // ... newobj instance void !!T::.ctor() However, it will not pass PEVerify. -- Barry mjdun***@gmail.com wrote:
Show quote > Why does the generic constraint new() use Activator<T>.CreateInstance Probably because the compiler hasn't special-cased code generation based> instead of new T()? > > An example > > public class Foo<T> : where T : class, new() > { > T objectValue = new T(); > > public foo() {} > } > > If you look at Lutz's reflector you will get something like > T objectValue = Activator<T>.CreateInstance() in place of T > objectValue = new T(); > > Using the Activator class is obviously slower and unintuitive to me, > so why does the framework compile my code using > Activator<T>.CreateInstance() instead of new T()? on reference type constraint. It does special-case struct, in which case it produces initobj rather than calling CreateInstance(). Because of the two different ways of initializing reference types and value types (first by newobj with constructor argument, second by initobj), in the most general case of a new() constraint something like Activator<T>.CreateInstance() is needed. -- Barry You can use newobj when you already know the type.
In the case of generics, you are creating a new type (closed type) that newobj cannot do. <mjdun***@gmail.com> ha scritto nel messaggio Show quote news:1170703813.568619.137330@a34g2000cwb.googlegroups.com... > Why does the generic constraint new() use Activator<T>.CreateInstance > instead of new T()? > > An example > > public class Foo<T> : where T : class, new() > { > T objectValue = new T(); > > public foo() {} > } > > If you look at Lutz's reflector you will get something like > T objectValue = Activator<T>.CreateInstance() in place of T > objectValue = new T(); > > Using the Activator class is obviously slower and unintuitive to me, > so why does the framework compile my code using > Activator<T>.CreateInstance() instead of new T()? >  |
|||||||||||||||||||||||
Other interesting topics