What is the => assignment in C# in a property signature
What you're looking at is an expression-bodied member not a lambda expression.
When the compiler encounters an expression-bodied property member, it essentially converts it to a getter like this:
public int MaxHealth
{
get
{
return Memory[Address].IsValid ? Memory[Address].Read<int>(Offs.Life.MaxHp) : 0;
}
}
(You can verify this for yourself by pumping the code into a tool called TryRoslyn.)
Expression-bodied members - like most C# 6 features - are just syntactic sugar. This means that they don’t provide functionality that couldn't otherwise be achieved through existing features. Instead, these new features allow a more expressive and succinct syntax to be used
As you can see, expression-bodied members have a handful of shortcuts that make property members more compact:
- There is no need to use a
returnstatement because the compiler can infer that you want to return the result of the expression - There is no need to create a statement block because the body is only one expression
- There is no need to use the
getkeyword because it is implied by the use of the expression-bodied member syntax.
I have made the final point bold because it is relevant to your actual question, which I will answer now.
The difference between...
// expression-bodied member property
public int MaxHealth => x ? y:z;
And...
// field with field initializer
public int MaxHealth = x ? y:z;
Is the same as the difference between...
public int MaxHealth
{
get
{
return x ? y:z;
}
}
And...
public int MaxHealth = x ? y:z;
Which - if you understand properties - should be obvious.
Just to be clear, though: the first listing is a property with a getter under the hood that will be called each time you access it. The second listing is is a field with a field initializer, whose expression is only evaluated once, when the type is instantiated.
This difference in syntax is actually quite subtle and can lead to a "gotcha" which is described by Bill Wagner in a post entitled "A C# 6 gotcha: Initialization vs. Expression Bodied Members".
While expression-bodied members are lambda expression-like, they are not lambda expressions. The fundamental difference is that a lambda expression results in either a delegate instance or an expression tree. Expression-bodied members are just a directive to the compiler to generate a property behind the scenes. The similarity (more or less) starts and end with the arrow (=>).
I'll also add that expression-bodied members are not limited to property members. They work on all these members:
- Properties
- Indexers
- Methods
- Operators
Added in C# 7.0
- Constructors
- Finalizers
However, they do not work on these members:
- Nested Types
- Events
- Fields
Ok... I made a comment that they were different but couldn't explain exactly how but now I know.
String Property { get; } = "value";
is not the same as
String Property => "value";
Here's the difference...
When you use the auto initializer the property creates the instance of value and uses that value persistently. In the above post there is a broken link to Bill Wagner, that explains this well, and I searched the correct link to understand it myself.
In my situation I had my property auto initialize a command in a ViewModel for a View. I changed the property to use expression bodied initializer and the command CanExecute stopped working.
Here's what it looked like and here's what was happening.
Command MyCommand { get; } = new Command(); //works
here's what I changed it to.
Command MyCommand => new Command(); //doesn't work properly
The difference here is when I use { get; } = I create and reference the SAME command in that property. When I use => I actually create a new command and return it every time the property is called. Therefore, I could never update the CanExecute on my command because I was always telling it to update a new reference of that command.
{ get; } = // same reference
=> // new reference
All that said, if you are just pointing to a backing field then it works fine. This only happens when the auto or expression body creates the return value.
This is a new feature of C# 6 called an expression bodied member that allows you to define a getter only property using a lambda like function.
While it is considered syntactic sugar for the following, they may not produce identical IL:
public int MaxHealth
{
get
{
return Memory[Address].IsValid
? Memory[Address].Read<int>(Offs.Life.MaxHp)
: 0;
}
}
It turns out that if you compile both versions of the above and compare the IL generated for each you'll see that they are NEARLY the same.
Here is the IL for the classic version in this answer when defined in a class named TestClass:
.property instance int32 MaxHealth()
{
.get instance int32 TestClass::get_MaxHealth()
}
.method public hidebysig specialname
instance int32 get_MaxHealth () cil managed
{
// Method begins at RVA 0x2458
// Code size 71 (0x47)
.maxstack 2
.locals init (
[0] int32
)
IL_0000: nop
IL_0001: ldarg.0
IL_0002: ldfld class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress> TestClass::Memory
IL_0007: ldarg.0
IL_0008: ldfld int64 TestClass::Address
IL_000d: callvirt instance !1 class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress>::get_Item(!0)
IL_0012: ldfld bool MemoryAddress::IsValid
IL_0017: brtrue.s IL_001c
IL_0019: ldc.i4.0
IL_001a: br.s IL_0042
IL_001c: ldarg.0
IL_001d: ldfld class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress> TestClass::Memory
IL_0022: ldarg.0
IL_0023: ldfld int64 TestClass::Address
IL_0028: callvirt instance !1 class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress>::get_Item(!0)
IL_002d: ldarg.0
IL_002e: ldfld class Offs TestClass::Offs
IL_0033: ldfld class Life Offs::Life
IL_0038: ldfld int64 Life::MaxHp
IL_003d: callvirt instance !!0 MemoryAddress::Read<int32>(int64)
IL_0042: stloc.0
IL_0043: br.s IL_0045
IL_0045: ldloc.0
IL_0046: ret
} // end of method TestClass::get_MaxHealth
And here is the IL for the expression bodied member version when defined in a class named TestClass:
.property instance int32 MaxHealth()
{
.get instance int32 TestClass::get_MaxHealth()
}
.method public hidebysig specialname
instance int32 get_MaxHealth () cil managed
{
// Method begins at RVA 0x2458
// Code size 66 (0x42)
.maxstack 2
IL_0000: ldarg.0
IL_0001: ldfld class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress> TestClass::Memory
IL_0006: ldarg.0
IL_0007: ldfld int64 TestClass::Address
IL_000c: callvirt instance !1 class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress>::get_Item(!0)
IL_0011: ldfld bool MemoryAddress::IsValid
IL_0016: brtrue.s IL_001b
IL_0018: ldc.i4.0
IL_0019: br.s IL_0041
IL_001b: ldarg.0
IL_001c: ldfld class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress> TestClass::Memory
IL_0021: ldarg.0
IL_0022: ldfld int64 TestClass::Address
IL_0027: callvirt instance !1 class [mscorlib]System.Collections.Generic.Dictionary`2<int64, class MemoryAddress>::get_Item(!0)
IL_002c: ldarg.0
IL_002d: ldfld class Offs TestClass::Offs
IL_0032: ldfld class Life Offs::Life
IL_0037: ldfld int64 Life::MaxHp
IL_003c: callvirt instance !!0 MemoryAddress::Read<int32>(int64)
IL_0041: ret
} // end of method TestClass::get_MaxHealth
See https://msdn.microsoft.com/en-us/magazine/dn802602.aspx for more information on this and other new features in C# 6.
See this post Difference between Property and Field in C# 3.0+ on the difference between a field and a property getter in C#.
Update:
Note that expression-bodied members were expanded to include properties, constructors, finalizers and indexers in C# 7.0.