MRubyCS.Compiler 0.61.3

mruby/cs

MRubyCS is a pure C# mruby virtual machine designed for seamless integration with C# game engines. It combines high Ruby-level compatibility with the performance and extensibility of modern C#.

Easily embed Ruby into Unity or .NET—empowering users to script game logic while keeping your core engine in C#.

Why MRubyCS?

• Zero native dependencies — runs anywhere Unity/.NET runs. No per-platform native builds to maintain.
• High performance — leverages .NET JIT, GC, and modern C# optimizations with minimal overhead.
• Ruby compatible — all opcodes implemented; passes mruby's official test suite
  • Syntax, Literals, Lang, Methods, Module, Exception, ...
  • Supported Types: Array, Class, Fiber, Float, Hash, Integer, Module, Nil, Proc, Random, Range, Symbol, String, Time
• Fiber & async/await — suspend Ruby execution and await C# async methods without blocking threads.
• Prism-based compiler — uses mruby-compiler2, the next-generation mruby compiler built on Prism (the official CRuby parser), for more accurate and modern Ruby syntax support.

Performance

In the .NET JIT environment, execution speeds are equal to or faster than the original native mruby.

<img width="594" height="389" alt="ss 2026-03-04 22 11 01" src="https://github.com/user-attachments/assets/00cd3644-e460-4b21-a41e-661d484fe30c" />

The above results were obtained on macOS with Apple M4 over 10 iterations.

Please refer to the following for the benchmark code.

Limitations

• As of mruby 3.3, almost all bundled classes/methods are supported.
  • Support for extensions split into mrbgems remains limited.
  • Some methods/specs added in 3.4 are not yet covered.
• However, basic private/protected visibility is already supported.

Table of Contents

• Installation
  • NuGet
  • Unity
• Basic Usage
  • MRubyState Lifecycle
  • MRubyValue
    • Symbol/String
    • Array/Hash
    • Embedded custom C# data into MRubyValue
  • Define ruby class/module/method by C#
    • Error handling & validation in C# methods
    • Constants
  • Call ruby method from C# side
    • Send with block / keyword arguments
    • Type conversion & introspection
    • Instance variables / class variables
    • Clone / Dup / Freeze
• Compiling Ruby source code
  • MRubyCS.Compiler.Cli (dotnet tool)
  • MRubyCS.Compiler (library)
• Fiber (Coroutine)
• MRubyCS.Serializer

Installation

NuGet

Package	Description	Latest version
MRubyCS	Main package. A mruby vm implementation.
MRubyCS.Compiler	Compile ruby source code utility. (Native binding)
MRubyCS.Compiler.Cli	dotnet tool for compiling Ruby source to bytecode
MRubyCS.Serializer	Converting Ruby and C# Objects Between Each Other

Unity

1. Install NuGetForUnity.
2. Install following packages via NugetForUnity
   • Utf8StringInterpolation
   • MRubyCS
   • (Optional) MRubyCS.Serializer
3. (Optional) To install utilities for generating mrb bytecode, refer to the Compiling Ruby source code section.

Basic Usage

MRubyState Lifecycle

MRubyState is the central object that holds the entire Ruby VM — symbol table, built-in classes, call stack, and all runtime state.

// Create a new VM instance.
var mrb = MRubyState.Create();

// Or, with additional configuration.
var mrb = MRubyState.Create(state =>
{
    state.DefineMethod(state.KernelModule, state.Intern("puts"u8), (s, self) =>
    {
        Console.WriteLine(s.GetArgumentAt(0));
        return MRubyValue.Nil;
    });
});

• No Dispose needed — MRubyState is fully managed by .NET GC. No native resources to release.
• Not thread-safe — each MRubyState instance must be used from a single thread. For multi-threaded scenarios, create a separate instance per thread.

MRubyState exposes all built-in Ruby classes as properties, which are used for class definitions, type checks, and method definitions:

Option A: Pre-compile bytecode

Pre-compiling Ruby source to .mrb bytecode keeps the native compiler out of your production deployment. You can use either the CLI tool or the C# API.

A-1. CLI tool

dotnet tool install -g MRubyCS.Compiler.Cli
mruby-compiler fibonacci.rb -o fibonacci.mrb

A-2. C# API

using MRubyCS;
using MRubyCS.Compiler;

var mrb = MRubyState.Create();
var compiler = MRubyCompiler.Create(mrb);

var source = """
    def fibonacci(n)
      return n if n <= 1
      fibonacci(n - 1) + fibonacci(n - 2)
    end

    fibonacci 10
    """u8;

// Compile and save as .mrb file
using var compilation = compiler.Compile(source);
File.WriteAllBytes("fibonacci.mrb", compilation.AsBytecode());

Execute pre-compiled bytecode

using MRubyCS;

var mrb = MRubyState.Create();
var bytecode = File.ReadAllBytes("fibonacci.mrb");
var result = mrb.LoadBytecode(bytecode);

result.IntegerValue //=> 55

Option B: Using Compiler library (runtime compile)

dotnet add package MRubyCS
dotnet add package MRubyCS.Compiler

using MRubyCS;
using MRubyCS.Compiler;

var mrb = MRubyState.Create();
var compiler = MRubyCompiler.Create(mrb);

var result = compiler.LoadSourceCode("""
    def fibonacci(n)
      return n if n <= 1
      fibonacci(n - 1) + fibonacci(n - 2)
    end

    fibonacci 10
    """u8);

result.IntegerValue //=> 55

Irep

You can also parse bytecode in advance. The result is called Irep in mruby terminology. Pre-parsing is useful when you want to execute the same bytecode multiple times without re-parsing overhead.

Irep irep = mrb.ParseBytecode(bytecode);
mrb.Execute(irep);

Irep can be executed as is, or converted to Proc, Fiber before use. For details on Fiber, refer to the Fiber section.

MRubyValue

Above result is MRubyValue. This represents a Ruby value.

value.IsNil //=> true if `nil`
value.IsInteger //=> true if integer
value.IsFloat //=> true if float
value.IsSymbol //=> true if Symbol
value.IsObject //=> true if any allocated object type

value.VType //=> get known ruby-type as C# enum.

value.IntegerValue //=> get as C# Int64
value.FloatValue //=> get as C# float
value.SymbolValue //=> get as `Symbol`

value.As<RString>() //=> get as internal String representation
value.As<RArray>() //=> get as internal Array representation
value.As<RHash>() //=> get as internal Hash representation

// pattern matching
if (value.Object is RString str)
{
    // ...
}

switch (value)
{
    case { IsInteger: true }:
        // ...
        break;
    case { Object: RString str }:
        // ...
        break;
}

var intValue = new MRubyValue(100); // create int value
var floatValue = new MRubyValue(1.234f); // create float value
var objValue = new MRubyValue(str); // create allocated ruby object value

// Implicit conversions — no `new MRubyValue(...)` needed
MRubyValue a = 42;         // int
MRubyValue b = 3.14;       // double
MRubyValue c = true;       // bool
MRubyValue d = sym;         // Symbol
MRubyValue e = rstring;     // RObject (RString, RArray, etc.)

// Static constants
MRubyValue.Nil   // Ruby nil
MRubyValue.True  // Ruby true
MRubyValue.False // Ruby false

// Boolean / truthiness
value.BoolValue //=> C# bool
value.Truthy    //=> true unless nil or false (Ruby semantics)
value.Falsy     //=> true if nil or false

Symbol/String

The string representation within mruby is utf8. Therefore, to generate a ruby string from C#, Utf8StringInterpolation is used internally.

// Create string literal.
var str1 = mrb.NewString("HOGE HOGE"u8); // use u8 literal (C# 11 or newer)
var str2 = mrb.NewString($"FOO BAR"); // use string interpolation

var x = 123;
var str3 = mrb.NewString($"x={x}");

// wrap MRubyValue..
var strValue = new MRubyValue(str1);

There is a concept in mruby similar to String called Symbol. Like String, it is created using utf8 strings, but internally it is a uint integer. Symbols are usually used for method IDs and class IDs.

To create a symbol from C#, use Intern.

// symbol literal
var sym1 = mrb.Intern("sym");

// create symbol from string interporation
var x = 123;
var sym2 = mrb.Intern($"sym{x}");

// symbol to utf8 bytes
mrb.NameOf(sym1); //=> "sym"u8
mrb.NameOf(sym2); //=> "sym123"u8

// create symbol from string
var sym2 = mrb.AsSymbol(mrb.NewString($"hoge"));

RString also provides methods for in-place manipulation and direct UTF-8 byte access:

var str = mrb.NewString(“hello”u8);

// UTF-8 byte access
ReadOnlySpan<byte> bytes = str.AsSpan(); // raw UTF-8 bytes

// In-place modification
str.Concat(“ world”u8);  // append bytes
str.Upcase();             // “HELLO WORLD”
str.Downcase();           // “hello world”
str.Capitalize();         // “Hello world”
str.Chomp();              // remove trailing newline
str.Chop();               // remove last character

Array/Hash

RArray and RHash are the internal representations of Ruby's Array and Hash.

// Create array
var array = mrb.NewArray(3); // with capacity
var array2 = mrb.NewArray(new MRubyValue(1), new MRubyValue(2), new MRubyValue(3));

// Access elements (supports negative indices)
var first = array2[0];   //=> 1
var last  = array2[-1];  //=> 3

// Add elements
array.Push(new MRubyValue(100));
array.Push(new MRubyValue(200));

// Get length
array.Length //=> 2

// Iterate over elements
foreach (var item in array)
{
    Console.WriteLine(item.IntegerValue);
}

// Pop / Shift
if (array.TryPop(out var popped)) { /* ... */ }
var shifted = array.Shift(); // remove and return first element

// Extract RArray from MRubyValue
var value = mrb.LoadBytecode(bytecode); // returns MRubyValue
var arr = value.As<RArray>();

// Create hash
var hash = mrb.NewHash();

// Set values (key can be any MRubyValue — Symbol, String, Integer, etc.)
hash[new MRubyValue(mrb.Intern("name"u8))] = new MRubyValue(mrb.NewString("Alice"u8));
hash[new MRubyValue(mrb.Intern("age"u8))]  = new MRubyValue(30);

// Get values
var name = hash[new MRubyValue(mrb.Intern("name"u8))];

// Check existence
hash.ContainsKey(new MRubyValue(mrb.Intern("name"u8))); //=> true
hash.TryGetValue(new MRubyValue(mrb.Intern("age"u8)), out var age); //=> true, age = 30

// Get length
hash.Length //=> 2

// Iterate over key-value pairs
foreach (var kv in hash)
{
    // kv.Key, kv.Value are MRubyValue
}

// Delete
hash.TryDelete(new MRubyValue(mrb.Intern("age"u8)), out var deleted);

// Extract RHash from MRubyValue
var hashValue = mrb.LoadBytecode(bytecode);
var h = hashValue.As<RHash>();

Embedded custom C# data into MRubyValue

You can stuff any C# object into an MRubyValue via RData. The RData.Data property accepts any object and can be freely get/set from C#.

This is useful when calling C# functionality from Ruby methods defined in C#.

class YourCustomClass
{
    public string Value { get; set; }
}

var csharpInstance = new YourCustomClass { Value = "abcde" };

var mrb = MRubyState.Create();

var data = new RData(csharpInstance);
state.SetConst(state.Intern("MYDATA"u8), state.ObjectClass, data);

// Use custom data from ruby
mrb.DefineMethod(mrb.ObjectClass, mrb.Intern("from_csharp_data"), (_, self) =>
{
    var dataValue = mrb.GetConst(state.Intern("MYDATA"u8), mrb.ObjectClass);
    var csharpInstance = dataValue.As<RData>().Data as YourCustomClass;
    // ...
});

Embedded custom C# data with ruby class

// Instances of classes that specify `MRubyVType.CSharpData` have `self` as RData.
var yourClass = DefineClass(Intern("MyCustomClass"u8), ObjectClass, MRubyVType.CSharpData);

// Define custom `initialize` with C# data
mrb.DefineMethod(yourClass, "initialize", (s, self) =>
{
    if (self.Object is RData x)
    {
        x.Data = new YourCustomClass { Value = "abcde" };
    }
    return self;
});

// Use custom C# data
mrb.DefineMethod(yourClass, "foo_method", (s, self) =>
{
    if (self.Object is RData { Data: YourCustomClass csharpInstance })
    {
        // Use C# data..
        csharpInstance.Value = "fghij";
    }
    // ...
});

Define ruby class/module/method by C#

// Define class
var classA = mrb.DefineClass(Intern("A"), c =>
{
    // Method definition that takes a required argument.
    c.DefineMethod(Intern("plus100"), (_, self) =>
    {
        var arg0 = mrb.GetArgumentAsIntegerAt(0); // get first argument (index:0)
        return arg0 + 100;
    });

    // Method definition that takes a block argument.
    c.DefineMethod(mrb.Intern("method2"), (_, self) =>
    {
        var arg0 = mrb.GetArgumentAt(0);
        var blockArg = mrb.GetBlockArgument();
        if (!blockArg.IsNil)
        {
            // Execute `Proc#call`
            mrb.Send(blockArg, mrb.Intern("call"), arg0);
        }
    });

    // Other complex arguments...
    c.DefineMethod(mrb.Intern("method3"), (_, self) =>
    {
        var keywordArg = mrb.GetKeywordArgument(mrb.Intern("foo"))
        Console.WriteLine($"foo: {keywordArg}");

        // argument type checking
        mrb.EnsureValueType(keywordArg, MRubyVType.Integer);

        var restArguments = mrb.GetRestArgumentsAfter(0);
        for (var i = 0; i < restArguments.Length; i++)
        {
            Console.WriteLine($"rest arg({i}: {restArguments[i]})");
        }
    });

    // class method
    c.DefineClassMethod(Intern("classmethod1"), (_, self) =>
    {
        return mrb.NewString($"hoge fuga");
    });
});

// Monkey patching
classA.DefineMethod(mrb.Intern("additional_method1"u8), (_, self) => { /* ... */ });

// Define module
var moduleA = mrb.DefineModule(mrb.Intern("ModuleA"));
mrb.DefineMethod(moduleA, mrb.Intern("additional_method2"), (_, self) => new MRubyValue(123));

mrb.IncludeModule(classA, moduleA);

As a result of the definition, the following Ruby code can now be executed.

a = A.new
a.plus100(123) #=> 223

a.method2(1) { |a| a } #=> 1

a.additional_method2 #=> 123

A.classmethod1 #=> "hoge fuga"

Error handling & validation in C# methods

Inside C#-defined methods, you can raise Ruby exceptions and validate arguments:

mrb.DefineMethod(myClass, mrb.Intern("safe_divide"u8), (s, self) =>
{
    s.EnsureArgumentCount(2, 2); // require exactly 2 arguments

    var a = s.GetArgumentAsIntegerAt(0);
    var b = s.GetArgumentAsIntegerAt(1);

    if (b == 0)
    {
        s.Raise(s.StandardErrorClass, "division by zero"u8);
    }
    return new MRubyValue(a / b);
});

// Available validation helpers
mrb.EnsureArgumentCount(min, max);               // check argument count
mrb.EnsureValueType(value, MRubyVType.Integer);  // check value type
mrb.EnsureBlockGiven(block);                     // check block is provided
mrb.EnsureNotFrozen(value);                      // check object is not frozen

// Raise Ruby exceptions
mrb.Raise(mrb.StandardErrorClass, "message"u8);
mrb.Raise(mrb.ExceptionClass, mrb.NewString($"detail: {info}"));

To catch Ruby exceptions raised during execution on the C# side:

try
{
    mrb.Send(obj, mrb.Intern("may_raise"u8));
}
catch (MRubyRaiseException ex)
{
    Console.WriteLine($"Ruby exception: {ex.Message}");
}

Constants

// Define a constant under Object (global)
mrb.DefineConst(mrb.Intern("MAX_SIZE"u8), new MRubyValue(1024));

// Define a constant under a specific class/module
mrb.DefineConst(myClass, mrb.Intern("VERSION"u8), new MRubyValue(mrb.NewString("1.0"u8)));

// Check if a constant exists
mrb.ConstDefinedAt(mrb.Intern("MAX_SIZE"u8));                         //=> true
mrb.ConstDefinedAt(mrb.Intern("VERSION"u8), myClass);                 //=> true
mrb.ConstDefinedAt(mrb.Intern("VERSION"u8), myClass, recursive: true); // search ancestors

// Safe lookup
if (mrb.TryGetConst(mrb.Intern("MAX_SIZE"u8), out var constValue))
{
    // use constValue...
}

Call ruby method from C# side

class A
  def self.foo = @@foo

  def self.foo=(x)
    @@foo = x
  end
end

class B
  attr_accessor :bar
end
@b = B.new

module M
  class C
    def self.foo = 999
  end
end

// get class instance
var classA = mrb.GetConst(mrb.Intern("A"), mrb.ObjectClass);

// call class method
mrb.Send(classA, mrb.Intern("foo="), new MRubyValue(123));
mrb.Send(classA, mrb.Intern("foo")); //=> 123

// call global-scope method — use TopSelf as the receiver
mrb.Send(mrb.TopSelf, mrb.Intern("puts"u8), mrb.NewString("hello"u8));

// get instance variable from top
var instanceB = mrb.GetInstanceVariable(mrb.TopSelf, mrb.Intern("@b"));
mrb.Send(instanceB, mrb.Intern("bar="), 456);
mrb.Send(instanceB, mrb.Intern("bar")); //=> 456

// find class instance on the hierarchy
var classC = mrb.Send(mrb.ObjectClass, mrb.Intern("const_get"), mrb.NewString($"M::C"));

Send with block / keyword arguments

// Send with a block (RProc)
var proc = mrb.CreateProc(irep);
mrb.Send(obj, mrb.Intern("each"u8), proc);

// Send with keyword arguments
mrb.Send(
    obj,
    mrb.Intern("configure"u8),
    args: [],
    kargs: [new(mrb.Intern("verbose"u8), MRubyValue.True)],
    block: null);

// This does NOT compile in Unity:
// mrb.Send(klass, sym, arg0, arg1);

// Use an explicit array:
mrb.Send(klass, sym, new MRubyValue[] { arg0, arg1 });

Type conversion & introspection

// Convert values (calls Ruby's to_i / to_f / to_sym internally)
long   i = mrb.AsInteger(value);
double f = mrb.AsFloat(value);
Symbol s = mrb.AsSymbol(value);

// Convert to string (Ruby's to_s / inspect)
RString str     = mrb.Stringify(value);  // to_s
RString inspect = mrb.Inspect(value);    // inspect

// Class introspection
RClass  klass = mrb.ClassOf(value);
RString name  = mrb.ClassNameOf(value);

// Type checking (Ruby's instance_of? / kind_of?)
mrb.InstanceOf(value, mrb.StringClass);  //=> true if exact class
mrb.KindOf(value, mrb.ObjectClass);      //=> true if class or ancestor

// Equality and comparison (calls Ruby's == / <=>)
mrb.ValueEquals(a, b);   //=> true/false
mrb.ValueCompare(a, b);  //=> -1, 0, 1

// Check if method exists (Ruby's respond_to?)
mrb.RespondTo(value, mrb.Intern("to_s"u8)); //=> true

Instance variables / class variables

// Instance variables
mrb.SetInstanceVariable(obj, mrb.Intern("@name"u8), mrb.NewString("Alice"u8));
var name = mrb.GetInstanceVariable(obj, mrb.Intern("@name"u8));
mrb.RemoveInstanceVariable(obj, mrb.Intern("@name"u8));

// Class variables
mrb.SetClassVariable(myClass, mrb.Intern("@@count"u8), new MRubyValue(0));
var count = mrb.GetClassVariable(myClass, mrb.Intern("@@count"u8));

Clone / Dup / Freeze

// Clone (deep copy with singleton class)
var cloned = mrb.CloneObject(value);

// Dup (shallow copy)
var duped = mrb.DupObject(value);

// Freeze an object (RObject level)
var str = mrb.NewString("immutable"u8);
str.MarkAsFrozen();
str.IsFrozen //=> true

Compiling Ruby source code

mruby has the following architecture, and allows the compiler and runtime to be separated.

By distributing only precompiled bytecode, you can optimize the installation on the application.

graph TB
    subgraph host["host machine"]
        A[source code<br/>.rb files]
        C[byte-code<br/>.mrb files]

        A -->|compile| C
    end

    C -->|deploy/install| E

    subgraph application["application"]
        D[mruby VM]
        E[byte-code<br>.mrb files]

        E -->|execute bytecode| D
    end

By the way, MRubyCS only includes the mruby virtual machine. Therefore it is necessary to convert it to .mrb bytecode before executing the .rb source.

MRubyCS.Compiler.Cli (dotnet tool)

The easiest way to compile Ruby source files is using the mruby-compiler dotnet tool.

# Install globally
$ dotnet tool install -g MRubyCS.Compiler.Cli
$ mruby-compiler input.rb -o output.mrb

# Or, install locally
$ dotnet tool install MRubyCS.Compiler.Cli
$ dotnet mruby-compiler input.rb -o output.mrb

# Dump bytecode in human-readable format
$ mruby-compiler input.rb --dump

# Generate C# code with embedded bytecode
$ mruby-compiler input.rb -o Bytecode.cs --format csharp --csharp-namespace MyApp

Options

mrbc (original mruby compiler)

Alternatively, you can use the original mruby project's compiler.

$ git clone git@github.com:mruby/mruby.git
$ cd mruby
$ rake
$ ./build/host/bin/mrbc -o output.mrb input.rb

MRubyCS.Compiler (library)

To simplify compilation from C#, we provide the MRubyCS.Compiler package, which is a thin wrapper of the C# API for the native compiler.

dotnet add package MRubyCS.Compiler

Unity

Open the Package Manager window by selecting Window > Package Manager, then click on [+] > Add package from git URL and enter the following URL:

https://github.com/hadashiA/MRubyCS.git?path=src/MRubyCS.Unity/Assets/MRubyCS.Compiler.Unity#0.50.3

For manual compilation, refer to the following.

Usage

using MRubyCS.Compiler;

var source = """
def f(a)
  1 * a
end

f 100
"""u8;

var mrb = MRubyState.Create();
var compiler = MRubyCompiler.Create(mrb);

// Compile source code (returns CompilationResult)
using var compilation = compiler.Compile(source);

// Convert to irep (internal executable representation)
var irep = compilation.ToIrep();

// irep can be used later..
var result = mrb.Execute(irep); // => 100

// Or, get bytecode (mruby calls this format "Rite")
// bytecode can be saved to a file or any other storage
File.WriteAllBytes("compiled.mrb", compilation.AsBytecode());

// Can be used later from file
mrb.LoadBytecode(File.ReadAllBytes("compiled.mrb")); //=> 100

// or, you can evaluate source code directly
result = compiler.LoadSourceCode("f(100)"u8);
result = compiler.LoadSourceCode("f(100)");

Unity AssetImporter

In Unity, if you install this extension, importing a .rb text file will generate .mrb bytecode as a subasset.

For example, importing the text file hoge.rb into a project will result in the following.

This subasset is a TextAsset. To specify it in the inspector.

Or, to extract in C#, do the following:

var mrb = MRubyState.Create();

var bytecodeAsset = (TextAsset)AssetDatabase.LoadAllAssetsAtPath("Assets/hoge.rb")
       .First(x => x.name.EndsWith(".mrb"));
mrb.LoadBytecode(bytecodeAsset.GetData<byte>().AsSpan());

To read a subasset in Addressables, you would do the following.

Addressables.LoadAssetAsync<TextAsset>("Assets/hoge.rb[hoge.mrb]")

Alternatively, you can generate the .mrb bytecode yourself within your project.

Fiber (Coroutine)

MRubyCS supports Ruby Fibers, which are lightweight concurrency primitives that allow you to pause and resume code execution. In addition to standard Ruby Fiber features, MRubyCS provides seamless integration with C#'s async/await pattern.

Basic Fiber Usage

using MRubyCS;
using MRubyCS.Compiler;

// Create state and compiler
var mrb = MRubyState.Create();
var compiler = MRubyCompiler.Create(mrb);

// Define a fiber that yields values
var code = """
    Fiber.new do |x|
      Fiber.yield(x * 2)
      Fiber.yield(x * 3)
      x * 4
    end
    """u8;

// Load the Ruby code as a Fiber
var irep = compiler.Compile(code);
var fiber = mrb.Execute(irep).As<RFiber>();

// Resume the fiber with initial value
var result1 = fiber.Resume(new MRubyValue(10));  // => 20

var result2 = fiber.Resume(new MRubyValue(10));  // => 30

var result3 = fiber.Resume(new MRubyValue(10));  // => 40 (final return value)

// Check if fiber is still alive
fiber.IsAlive  // => false

If you want to execute arbitrary code snippets as fibers, do the following.

var code = """
  x = 1
  y = 2
  Fiber.yield (x + y) * 100
  Fiber.yield (x + y) * 200
"""u8;

var fiber = compiler.LoadSourceCodeAsFiber(code);

// `LoadSourceCodeAsFiber` is same as:
// using var compilation = compiler.Compile(code);
// var proc = mrb.CreateProc(compilation.ToIrep());
// var fiber = mrb.CreateFiber(proc);

fiber.Resume(); //=> 300
fiber.Resume(); //=> 600

Async/Await Integration

MRubyCS provides unique C# async integration features for working with Fibers:

// Wait for fiber to terminate
var code = """
    Fiber.new do |x|
      Fiber.yield
      Fiber.yield
      "done"
    end
    """u8;

var irep = compiler.Compile(code);
var fiber = mrb.Execute(irep).As<RFiber>();

// Start async wait before resuming
var terminateTask = fiber.WaitForTerminateAsync();

// Resume the fiber multiple times
fiber.Resume();
fiber.Resume();
fiber.Resume();

// Wait for completion
await terminateTask;
Console.WriteLine("Fiber has terminated");

You can consume fiber results as async enumerable:

var code = """
    Fiber.new do |x|
      3.times do |i|
        Fiber.yield(x * (i + 1))
      end
    end
    """u8;

var irep = compiler.Compile(code);
var fiber = mrb.Execute(irep).As<RFiber>();

// Process each yielded value asynchronously
await foreach (var value in fiber.AsAsyncEnumerable())
{
    Console.WriteLine($"Yielded: {value.IntegerValue}");
}

MRubyCS supports multiple consumers waiting for fiber results simultaneously:

var irep = compiler.Compile(code);
var fiber = mrb.Execute(irep).As<RFiber>();

// Create multiple consumers
var consumer1 = Task.Run(async () =>
{
    while (fiber.IsAlive)
    {
        var result = await fiber.WaitForResumeAsync();
        Console.WriteLine($"Consumer 1 received: {result}");
    }
});

var consumer2 = Task.Run(async () =>
{
    while (fiber.IsAlive)
    {
        var result = await fiber.WaitForResumeAsync();
        Console.WriteLine($"Consumer 2 received: {result}");
    }
});

// Resume fiber and both consumers will receive the results
fiber.Resume(new MRubyValue(10));
fiber.Resume(new MRubyValue(20));
fiber.Resume(new MRubyValue(30));

await Task.WhenAll(consumer1, consumer2);

Error Handling in Fibers

Exceptions raised within fibers are properly propagated:

var code = """
    Fiber.new do |x|
      Fiber.yield(x)
      raise "Something went wrong"
    end
    """u8;

var irep = compiler.Compile(code);
var fiber = mrb.Execute(irep).As<RFiber>();

// First resume succeeds
var result1 = fiber.Resume(new MRubyValue(10));  // => 10

// Second resume will throw
try
{
    fiber.Resume();
}
catch (MRubyRaiseException ex)
{
    Console.WriteLine($"Ruby exception: {ex.Message}");
}

// Async wait will also propagate the exception
var waitTask = fiber.WaitForResumeAsync();
try
{
    fiber.Resume();
    await waitTask;
}
catch (MRubyRaiseException ex)
{
    Console.WriteLine($"Async exception: {ex.Message}");
}

yield/resume from C#

It is possible to resume/yield from a method defined in C#.

mrb.DefineMethod(mrb.FiberClass, mrb.Intern("resume_by_csharp"u8), (state, self) =>
{
    return self.As<RFiber>().Resume();
});

 fiber = Fiber.new do
   3.times do
     Fiber.yield
   end
 end

 fiber.resume_by_csharp

MRubyCS.Serializer

Using the MRuby.Serializer package enables conversion between MRubyValue and C# objects.

// Deserialize (MRubyValue -> C#)

MRubyValue result1 = mrb.LoadSourceCode("111 + 222");
MRubyValueSerializer.Deserialize<int>(result1, mrb); //=> 333

MRubyValue result2 = mrb.LoadSourceCode("'hoge'.upcase");
MRubyValueSerializer.Deserialize<string>(result2, mrb); //=> "HOGE"

// Serialize (C# -> MRubyValue)

var intArray = new int[] { 111, 222, 333 };

MRubyValue value = MRubyValueSerializer.Serialize(intArray, mrb);

var mrubyArray = value.As<RArray>();
mrubyArray[0] //=> 111
mrubyArray[1] //=> 222
mrubyArray[2] //=> 333

MRubyValue mrubyStringValue = MRubyValueSerializer.Serialize("hoge fuga", mrb);

// Use the serialized value...
mrb.Send(mrubyStringValue, mrb.Intern("upcase")); //=> MRubyValue("UPCASE")

Builtin Supported types

The following C# types and MRubyValue type conversions are supported natively:

Unity-specific types

By introducing the following packages, serialization of Unity-specific types will also be supported.

Open the Package Manager window by selecting Window > Package Manager, then click on [+] > Add package from git URL and enter the following URL:

https://github.com/hadashiA/MRubyCS.git?path=src/MRubyCS.Unity/Assets/MRubyCS.Serializer.Unity#0.18.1

Naming Convention

• C# property/field names are converted to underscore style in Ruby
  • e.g) FooBar ↔ foo_bar
• C# enum values are converted to underscore-style symbols in Ruby
  • e.g) EnumType.FooBar ↔ :foo_bar

[MRubyObject] attribute

Marking with [MRubyObject] enables bidirectional conversion between custom C# types and MRubyValue.

• Converts C# type properties/fields into Ruby world Hash key/value pairs.
• class, struct, and record are all supported.
• A partial declaration is required.
• Members that meet the following conditions are converted from mruby:
  • public fields or properties, or fields or properties with the [MRubyMember] attribute.
  • And have a setter (private is acceptable).

[MRubyObject]
partial struct SerializeExample
{
    // this is serializable members
    public string Id { get; private set; }
    public int X { get; init; }
    public int FooBar;

    [MRubyMember]
    public int Z;

    // ignore members
    [MRubyIgnore]
    public float Foo;
}

// Deserialize (MRubyValue -> C#)

var value = mrb.LoadSourceCode("{ id: 'aiueo', x: 1234, foo_bar: 4567, z: 8901 }");

SerializeExample deserialized = MRubyValueSerializer.Deserialize<SerializeExample>(value, mrb);
deserialized.Id     //=> "aiueo"
deserialized.X      //=> 1234
deserialized.FooBar //=> 4567
deserialized.Z      //=> 8901

// Serialize (C# -> MRubyValue)
var value = MRubyValueSerializer.Serialize(new SerializeExample { Id = "aiueo", X = 1234, FooBar = 4567 });

var props = value.As<RHash>();
props[mrb.Intern("id")] //=> "aiueo"
props[mrb.Intern("x")] //=> 1234
props[mrb.Intern("foo_bar")] //=> 4567

The list of properties specified by mruby is assigned to the C# member names that match the key names.

Note:

• The names on the ruby side are converted to CamelCase.
  • Example: ruby's foo_bar maps to C#'s FooBar.
• The values of C# enums are serialized as Ruby symbols.
  • Example: Season.Summer becomes Ruby's :summer.

You can change the member name specified from Ruby by using [MRubyMember("alias name")].

[MRubyObject]
partial class Foo
{
    [MRubyMember("alias_y")]
    public int Y;
}

Also, you can receive data from Ruby via any constructor by using the [MRubyConstructor] attribute.

[MRubyObject]
partial class Foo
{
    public int X { get; }

    [MRubyConstructor]
    public Foo(int x)
    {
        X = x;
    }
}

Dynamic serialization

Specifying a dynamic type parameter allows conversion to C# Array/Dictionary and primitive types.

var array = mrb.NewArray();
array.Push(123);

var result = MRubyValueSerializer.Deserialize<dynamic>(array, mrb);

((object[])result).Length //=> 1
((object[])result)[0] //=> 123

Custom Formatter

You can also customize the conversion of any C# type to an MRubyValue.

 // custom type example
struct Vector3
{
    public int X;
    public int Y;
    public int Z;
}

// Implement `IMRubyValueFormatter`
class CustomVector3Formatter : IMRubyValueFormatter<Vector3>
{
    public static readonly CustomVector3Formatter Instance = new();

    public MRubyValue Serialize(Vector3 value, MRubyState mrb, MRubyValueSerializerOptions options)
    {
        var array = mrb.NewArray();
        array.Push(value.X);
        array.Push(value.Y);
        array.Push(value.Z);
        return array;
    }
    public Vector3 Deserialize(MRubyValue value, MRubyState mrb, MRubyValueSerializerOptions options)
    {
        // validation
        MRubySerializationException.ThrowIfTypeMismatch(value, MRubyVType.Array);
        MRubySerializationException.ThrowIfNotEnoughArrayLength(value, 3);

        var array = value.As<RArray>();
        return new Vector3
        {
            X = array[0].IntegerValue,
            Y = array[1].IntegerValue,
            Z = array[2].IntegerValue,
        }
    }
}

To set a custom formatter, specify options as an argument to MRubyValueSerializer.

Specify the enumeration of Formatter and Formatter's Resolver instances. StandardResolver supports the default behavior, so specify this along with additional formatters.

// Create a new formatter resolver.
var resolver = CompositeResolver.Create(
    [CustomVector3Formatter.Instance],
    [StandardResolver.Instance]
    );

var options = new MRubyValueSerializerOptions
{
    Resolver = resolver,
};

var value = mrb.LoadSourceCode("[111, 222, 333]");
Vector3 deserialized = MRubyValueSerializer.Deserialize<Vector3>(value, mrb, options);
deserialized.X //=> 111
deserialized.Y //=> 222
deserialized.Z //=> 333

LICENSE

MIT

Contact

@hadashiA