When generating code, you may need to instantiate a variable or a parameter with some placeholder value (On my end, I needed to assign values of base class’s constructor parameters). The easiest solution will be using the default literal. Or if you want to be more traditional and support older C# versions, using 0 for value types (except structs) and null for reference types will give the same result. But these approaches have some issues with method overloading.
Let’s say there are multiple Foo methods, and our intention is calling method Foo(float f). Calling Foo(0) or Foo(default) will both fail. Even worse, Foo(default) might cause an ambiguity error.
void FooTest(){
Foo(default); // will give ambiguity error
Foo(0); // returns "int"
}
string Foo(int i) => "int";
string Foo(byte b) => "byte";
string Foo(float f) => "float";
string Foo(string s) => "string";
void FooTest(){
Foo(default); // returns "byte"
Foo(0); // returns "int"
}
string Foo(int i) => "int";
string Foo(byte b) => "byte";
string Foo(float f) => "float";
// string Foo(string s) => "string";
To overcome overloading ambiguity, we need to explicitly specify type of the value. Calling Foo((float)0) or Foo(default(float)) will work. (Also another way of calling is with the constructor, Foo(new float()) )
void FooTest(){
Foo(default(float)); // returns "float"
Foo((float)0); // returns "float"
}
string Foo(int i) => "int";
string Foo(byte b) => "byte";
string Foo(float f) => "float";
string Foo(string s) => "string";
Explicitly specifying types is totally fine, except it is not easy to read for the human eye. No one would prefer to see (float)0 or default(float) over 0f.
void FooTest(){
Foo(0f); // returns "float"
}
string Foo(int i) => "int";
string Foo(byte b) => "byte";
string Foo(float f) => "float";
string Foo(string s) => "string";
Float is an exception, along with some other types. Not all types have a unique identifier suffix like f. But since there are some, and these are among the most used types; I decided to write a helper method for generating default value in the user-friendly way. Searching Roslyn methods and digging Roslyn source code didn’t help much, so I had to write it myself.
Here is the code snippet, it should be portable, ready for usage. I preferred to use "" over null and string.empty for strings, Datetime.Now over new DateTime(), IntPtr.Zero over new IntPtr().
public static ExpressionSyntax TypeToDefaultValue(ITypeSymbol typeSymbol, bool isExplicit = false)
{
var syntax = TypeToDefaultValueAux(typeSymbol, isExplicit, out bool cannotBeExplicit);
if (isExplicit && cannotBeExplicit)
{
syntax = TypeToDefaultValueAux(typeSymbol, false, out _);
}
return syntax;
}
static ExpressionSyntax TypeToDefaultValueAux(ITypeSymbol typeSymbol, bool isExplicit, out bool cannotBeExplicit)
{
// https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/keywords/default-values-table
// https://docs.microsoft.com/tr-tr/dotnet/csharp/language-reference/keywords/value-types-table
cannotBeExplicit = false;
if (typeSymbol.IsAnonymousType && isExplicit)
{
// type cannot be expressed since it is anonymous, typeSymbol.ToDisplayString(format) wont work
isExplicit = false;
}
switch (typeSymbol.SpecialType)
{
case SpecialType.System_Enum:
return SyntaxFactory.ParseExpression("(System.Enum)null")
.WithAdditionalAnnotations(Simplifier.Annotation);
case SpecialType.System_ValueType:
return SyntaxFactory.ParseExpression("(System.ValueType)null")
.WithAdditionalAnnotations(Simplifier.Annotation);
case SpecialType.System_Boolean:
return SyntaxFactory.ParseExpression("false");
case SpecialType.System_Char:
return SyntaxFactory.ParseExpression("'\0'");
case SpecialType.System_SByte:
return isExplicit ?
SyntaxFactory.ParseExpression("(sbyte)0") :
SyntaxFactory.ParseExpression("0");
case SpecialType.System_Byte:
return isExplicit ?
SyntaxFactory.ParseExpression("(byte)0") :
SyntaxFactory.ParseExpression("0");
case SpecialType.System_Int16:
return isExplicit ?
SyntaxFactory.ParseExpression("(short)0") :
SyntaxFactory.ParseExpression("0");
case SpecialType.System_UInt16:
return isExplicit ?
SyntaxFactory.ParseExpression("(ushort)0") :
SyntaxFactory.ParseExpression("0");
case SpecialType.System_Int32:
return isExplicit ?
SyntaxFactory.ParseExpression("(int)0") :
SyntaxFactory.ParseExpression("0");
case SpecialType.System_UInt32:
return isExplicit ?
SyntaxFactory.ParseExpression("(uint)0") :
SyntaxFactory.ParseExpression("0u");
case SpecialType.System_Int64:
return isExplicit ?
SyntaxFactory.ParseExpression("(long)0") :
SyntaxFactory.ParseExpression("0L");
case SpecialType.System_UInt64:
return SyntaxFactory.ParseExpression("0ul");
case SpecialType.System_Decimal:
return SyntaxFactory.ParseExpression("0m");
case SpecialType.System_Single:
return SyntaxFactory.ParseExpression("0f");
case SpecialType.System_Double:
return SyntaxFactory.ParseExpression("0d");
case SpecialType.System_String:
return SyntaxFactory.ParseExpression("\"\"");
//return SyntaxFactory.ParseExpression("string.Empty");
//return SyntaxFactory.ParseExpression("null");
case SpecialType.System_IntPtr:
return SyntaxFactory.ParseExpression("System.IntPtr.Zero")
.WithAdditionalAnnotations(Simplifier.Annotation);
case SpecialType.System_UIntPtr:
return SyntaxFactory.ParseExpression("System.UIntPtr.Zero")
.WithAdditionalAnnotations(Simplifier.Annotation);
case SpecialType.System_Nullable_T:
break;
case SpecialType.System_DateTime:
return SyntaxFactory.ParseExpression("System.DateTime.Now")// warning, this is not the default value, 0
.WithAdditionalAnnotations(Simplifier.Annotation);
default:
break;
}
var isStruct = false;
switch (typeSymbol.TypeKind)
{
case TypeKind.Enum:
var defaultEnumField = typeSymbol.GetMembers().FirstOrDefault(
m => m is IFieldSymbol mf && mf.HasConstantValue && Convert.ToInt64(mf.ConstantValue) == 0);
if (defaultEnumField != null)
return SyntaxFactory.ParseExpression(getDisplayString(typeSymbol, out cannotBeExplicit) + "." + defaultEnumField.Name)
.WithAdditionalAnnotations(Simplifier.Annotation);
else
return SyntaxFactory.ParseExpression("(" + getDisplayString(typeSymbol, out cannotBeExplicit) + ")0")
.WithAdditionalAnnotations(Simplifier.Annotation);
case TypeKind.Error:
// throw new Exception();// can be uncommented when debugging
break;
case TypeKind.Struct:
isStruct = true;
break;
case TypeKind.TypeParameter:
//var typeParameterSymbol = typeSymbol as ITypeParameterSymbol;
if (typeSymbol.IsReferenceType)
break;
else if (typeSymbol.IsValueType)
{
isStruct = true;
break;
}
else
return SyntaxFactory.ParseExpression("default(" + getDisplayString(typeSymbol, out cannotBeExplicit) + ")")
.WithAdditionalAnnotations(Simplifier.Annotation);
default:
break;
}
if (isStruct)
{
if (typeSymbol.IsTupleType)
{
var named = typeSymbol as INamedTypeSymbol;
return SyntaxFactory.ParseExpression("(" +
String.Join(", ", named.TupleElements.Select(te => TypeToDefaultValue(te.Type, isExplicit)))
+ ")");
}
else if (typeSymbol.IsNullable())
{
// do nothing
}
else
{
return SyntaxFactory.ParseExpression("new " + getDisplayString(typeSymbol, out cannotBeExplicit) + "()")
.WithAdditionalAnnotations(Simplifier.Annotation);
}
}
return isExplicit ?
SyntaxFactory.ParseExpression("(" + getDisplayString(typeSymbol, out cannotBeExplicit) + ")null")
.WithAdditionalAnnotations(Simplifier.Annotation) :
SyntaxFactory.ParseExpression("null");
string getDisplayString(ITypeSymbol ts, out bool cantBeExplicit)
{
cantBeExplicit = false;
var format = SymbolDisplayFormat.FullyQualifiedFormat
.WithGlobalNamespaceStyle(SymbolDisplayGlobalNamespaceStyle.OmittedAsContaining);
var display = typeSymbol.ToDisplayString(format);
// if an inner type is anonymous, disable explicit
if (isExplicit && display.Contains("<anonymous type"))
cantBeExplicit = true;
return display;
}
}
Related useful links:
https://docs.microsoft.com/en-us/dotnet/csharp/language-reference/builtin-types/default-values