The Source Generator Toolkit
provides functionality to easily generate c# code/files using a fluent builder pattern - while initially designed to be used in conjunction with the .NET Roslyn Source Generator process, the functionality can be leveraged outside of process to generate c# source code.
The static SourceGenerator
class is the starting point for building up the source code. This process will generate a formatted string representation of the c# code:
var strCode = SourceGenerator.GenerateSource(gen =>
{
gen.WithFile("file1", file =>
{
file.WithNamespace("SampleNamespace", ns =>
{
ns.WithClass("SampleClass", cls => { });
});
});
});
The string output of the above being (the value of strCode
):
namespace SampleNamespace
{
[System.CodeDom.Compiler.GeneratedCode("SourceGeneratorToolkit", "0.0.0.1")]
class SampleClass
{
}
}
When used in conjunction with a Source Generator, the GenerateSource
extension method on the GeneratorExecutionContext
class can be leveraged.
The below example shows how to generate source code without any information from a SyntaxReceiver
- see further down on how the Source Generator Toolkit
can be used to generate code in conjunction with a ISyntaxReceiver
implementation.
public class SampleGenerator : ISourceGenerator
{
public void Execute(GeneratorExecutionContext context)
{
context.GenerateSource("file1", fileBuilder =>
{
fileBuilder.WithNamespace("SampleNamespace", nsBuilder =>
{
ns.WithClass("SampleClass", cls => { });
});
});
}
public void Initialize(GeneratorInitializationContext context)
{
// no ISyntaxReceiver implementation registered here
}
}
In the case of a Source Generator, an actual file named file1.cs
will be output as part of the generation process.
The output content of the file will be the same as in the previous example:
namespace SampleNamespace
{
[System.CodeDom.Compiler.GeneratedCode("SourceGeneratorToolkit", "0.0.0.1")]
class SampleClass
{
}
}
There is optional configuration which can be specified when generating the code using either of the above two methods (when calling the GenerateSource
method). If no configuration is specified, the default configuration is used.
Configuration Name | Description | Default Value |
---|---|---|
OutputGeneratedCodeAttribute | Flag to indicate if the System.CodeDom.Compiler.GeneratedCode attribute should be output with generated code. This attribute is used as an indicator to various tools that the code was auto generated |
true |
OutputDebuggerStepThroughAttribute | Flag to indicate if the System.Diagnostics.DebuggerStepThrough attribute should be output with generated code. When set to true, this attribute allows stepping into the generated code when debugging |
false |
In the unit test project of the source code, there are numerous examples of all the possible c# language components the library is capable of generating.
When using the .NET Roslyn Source Generator process, the actual generation of the source is only one step of the process - the other step is determining if the source should be generated in the first place. This qualification check is done in the OnVisitSyntaxNode
method of the ISyntaxReceiver
implementation.
The OnVisitSyntaxNode
method takes a SyntaxNode
as an argument (this is part of the normal Roslyn Source Generator process) - the Source Generator Toolkit
provides an extension method (NodeQualifiesWhen
) which accepts a qualification builder which is used to determine if the SyntaxNode qualifies to have source code generated.
The fluent builder pattern is again used to build up the qualification check for for the syntax:
class SampleClassSyntaxReceiver : ISyntaxReceiver
{
public List<SyntaxReceiverResult> Results { get; set; } = new List<SyntaxReceiverResult>();
public void OnVisitSyntaxNode(SyntaxNode syntaxNode)
{
syntaxNode.NodeQualifiesWhen(Results, node =>
{
node.IsClass(c => c
.WithName("SampleClass")
.IsNotStatic()
.IsPublic()
.Implements("ISerializable")
);
});
}
}
In the above example, if the qualification checks determines the node is:
- a class named
SampleClass
- which is public
- and not static
- and also implements
ISerializable
then the specific SyntaxNode
qualifies, and the Results list will be populated and passed to the Execute method of the generator.
A most complex, but less practical example:
syntaxNode.NodeQualifiesWhen(Results, node =>
{
node.IsClass(c => c
.WithName("SampleClass")
.IsNotStatic()
.IsNotPrivateProtected()
.IsPublic()
.Implements("ISerializable")
// the class must have the Obsolete attribute
.WithAttribute(a =>
{
a.WithName("Obsolete");
})
.WithMethod(m =>
{
// the class must have a method called "SampleMethod"
m.WithName("SampleMethod")
// which is async
.IsAsync()
// with the Obsolete attribute with a parameter in position 1 supplied
.WithAttribute(a =>
{
a.WithName("Obsolete")
.WithArgument(arg =>
{
arg.WithPosition(1);
});
})
// method must have a return type of Task
.WithReturnType(typeof(Task));
})
);
});
When generating code based on the output of the qualification process (OnVisitSyntaxNode
method in the ISyntaxReceiver
implementation, shown above), the Results
list is populated with the qualifying SyntaxNode(s)
, and passed to the Execute
method of the ISourceGenerator
implementation.
Using the same ISyntaxReceiver
implementation as above:
class SampleClassSyntaxReceiver : ISyntaxReceiver
{
public List<SyntaxReceiverResult> Results { get; set; } = new List<SyntaxReceiverResult>();
public void OnVisitSyntaxNode(SyntaxNode syntaxNode)
{
syntaxNode.NodeQualifiesWhen(Results, node =>
{
node.IsClass(c => c
.WithName("SampleClass")
.IsNotStatic()
.IsPublic()
.Implements("ISerializable")
);
});
}
}
If a qualifying node is found, the Results
property is populated with the SyntaxNode
.
Below is a sample of a ISourceGenerator
which used the Results output from the OnVisitSyntaxNode
method to generate source code:
[Generator]
public class PartialMethodGenerator : ISourceGenerator
{
public void Initialize(GeneratorInitializationContext context)
{
// Register our custom syntax receiver
context.RegisterForSyntaxNotifications(() => new PartialClassSyntaxReceiver());
}
public void Execute(GeneratorExecutionContext context)
{
if (context.SyntaxReceiver == null)
{
return;
}
PartialClassSyntaxReceiver syntaxReceiver = (PartialClassSyntaxReceiver)context.SyntaxReceiver;
if (syntaxReceiver != null && syntaxReceiver.Results != null && syntaxReceiver.Results.Any())
{
foreach (SyntaxReceiverResult result in syntaxReceiver.Results)
{
ClassDeclarationSyntax cls = result.Node.AsClass();
context.GenerateSource($"{cls.GetNamespace()}_file", fileBuilder =>
{
fileBuilder.WithNamespace($"{cls.GetNamespace()}", nsBuilder =>
{
nsBuilder.WithClass($"{cls.GetName()}_generated", clsBuilder =>
{
clsBuilder.AsPublic();
clsBuilder.WithMethod("Hello", "void", mthBuilder =>
{
mthBuilder.AsPublic()
.WithBody(@"Console.WriteLine($""Generator says: Hello"");");
});
});
});
});
}
}
}
}
- The
Initialize
is used to register the customISyntaxReceiver
implementation containing the qualification rules - this is part of the Roslyn source generation processes - The
GeneratorExecutionContext
parameter passed to theExecute
method contains a ISyntaxReceiver implementation property -PartialClassSyntaxReceiver
in this example, which contains the Results property with the qualifying SyntaxNode. A number of checks are performed to ensure the SyntaxReceiver is not null, and that the Results property on it is not null. - The code then iterates over each
SyntaxReceiverResult
in the Results property - effectively iterating through each qualifying node - The
AsClass
extension method (part of theSource Generator Toolkit
) will convert the generic SyntaxNode to the specific syntax type ('ClassDeclarationSyntax' in this example) - The
GenerateSource
extension method (again, part of theSource Generator Toolkit
) then allows for the building up of the required source code as described above. However, now, instead of explicitly supplying the values for the code (the file name, namespace and class name in this example), the provided extension methods are used to extract the values from the qualifying syntax node. - In this example, the
GetNamespace
andGetName
extension methods onClassDeclarationSyntax
are used to get the relevent details from the syntax to populate the generated source code
The Source Generator Toolkit
allows for custom qualification checks using the WithQualifyingCheck
method:
syntaxNode.NodeQualifiesWhen(Results, node =>
{
node.WithQualifyingCheck(customNode =>
{
// completely un-useful check
return customNode.ChildNodes().Count() == 10;
});
});
Here instead of checking if the node is a class or attribute for example, the qualification check is to see if the node contains 10 child nodes (a not very useful check)
The initial version of the library was written with the most common syntax type checks, extension methods - there is still a lot to add depending on requirements:
- SyntaxNode
AsAttribute
extension method - Additional extension methods to be used on ClassDeclarationSyntax and AttributeDeclarationSyntax to be leverage when doing code generation in a source generator
- Ability to determine qualification with generics and generic types
- Ability to determine qualification based on a code comment
Feel free to log a request or bug if you have a specific requirement and I'll try to implement asap.