Xyn Compiler

Chapters

1. Overview
2. Variables

Overview

Xyn is an experimental, self-designed programming language and compiler project focused on clarity, control, and explicitness, while still being grounded in solid compiler theory and modern implementation techniques.

The project is developed entirely in Java, with the explicit goal of being understandable, extensible, and maintainable over time.

This repository documents the design decisions, internal architecture, and theoretical foundations of the Xyn compiler.

Xyn is not designed to compete with mainstream languages. It is a learning-driven and research-oriented project that prioritizes understanding how compilers really work.

Variables

Sections:

• Types in Xyn
• Example 0
• Nullability
• Static or Dynamic?

Types

• There are so many types in Xyn that can you use, from built-ins, libraries, and user-defined types.
  They have their own functionality and cost, so it's a good thing to use different types for different situations.

There are three categories about how types in Xyn can splits into.

Categories	Types	Description
Primitive	int, float, char, str, bool	basic types with fixed size
Collections	List<T>, Map<K, V>, Set<E>	type-safe data structure with dynamic size
User-defined	PersonClass, Player, Inventory	flexible data structure stored in heap

Example 0:

let age: int = 12;
// this is not allowed because `age` is not nullable
age = null;

dyn name = age;
name = null; // another error

Nullability

Type Inference and Static Typing

Xyn provides type inference for statement declarations such as variables, functions, and more.
Xyn also supports static typing, which offers its own advantages compared to type inference.

Benefits of Type Inference

• Shorter and less boilerplate declarations for long and repetitive types.
  Write what matters, not redundant information
  (e.g. SystemInformerInterface, List<Map<Str, Set>>, etc.).

• Improved local reasoning.
  Code becomes easier to read from top to bottom without mentally parsing explicit type declarations.
  Especially powerful for expressions, closures / lambdas, and temporary values.

• Safer and faster refactoring.
  Changing a single definition is easier when types are inferred, because the compiler automatically propagates type changes.

• Enables powerful abstractions.
  Type inference makes generics, pattern matching, and other advanced features much easier to write without explicitly specifying type parameters.

Examples

// before
let list: List<Integer?> = new List<Integer?>();

// after (note: `int` is lowered to `Integer` by the compiler)
let list = new List<int?>();


// before
let type: enumType = enumType.A;

// after
let type = enumType.A;


// before
func put(x: List<Integer>, y: int) -> List<Integer> {
  x.add(y);
  return x;
}

// after
func put(x: List<int>, y: int) {
  x.add(y);
  return x;
}

The examples above demonstrate how type inference works in Xyn. However, some developers prefer explicit static typing for clarity and intent. Xyn fully supports static typing, enhanced with additional syntactic sugar to keep code concise and expressive.

Benefits of Static Typing:

• Extra clarity and safety.
  These are especially important in large projects with multiple contributors, where developers cannot always infer each other’s intentions. Explicit types serve as a clear contract and documentation for shared code. Bugs and errors are caught earlier, before the program runs.

• With static typing, the compiler can detect invalid type usage at compile time.

• Simpler and faster runtime execution.
  Since types are fully known at compile time, the runtime can be simpler, more predictable, and more optimized without dynamic dispatch or other complex runtime mechanisms.

Examples

// before
let john: Person = new Person("John");

// after
let john: Person("John");


// before
let personJob: JobType = JobType.None;

// after
let personJob: JobType = None; // resolved to JobType.None at compile time


// before
let name = null; // unknown type

// after
let name: str? = null; // nullable string
name = "John";

---

Project Goals

• Design and implement a full compiler pipeline from scratch.

• Prioritize imperative semantics and explicit control flow.

• Keep abstractions minimal and understandable.

• Avoid unnecessary magic or hidden behavior.

• Learn and apply real-world compiler techniques (used in LLVM, JVM, and modern PLs)

---

Language Philosophy

Xyn is intentionally designed around these principles:

1. Imperative-first: You explicitly describe how things happen, not just what happens.

2. Predictable semantics: Language behavior should be obvious from reading the code.

3. Minimal core: The language starts small and grows deliberately.

4. Bootstrappable: Even fundamental concepts (like strings) are not assumed magically.

The language intentionally avoids overly abstract or purely functional paradigms, favoring transparency and control.

---

Implementation Language

Xyn made from purely all Java ecosystem and tooling.

Java was chosen because:

1. Strong tooling and debuggability.

2. Explicit memory and object model.

3. Easier long-term maintenance.

4. Better alignment with imperative reasoning.

---

Compiler Architecture Overview

The Xyn compiler follows a traditional but carefully designed pipeline:

Source Code → Lexer (Lexical Analysis) → Parser (Syntax Analysis) → AST (Abstract Syntax Tree) → Type Checker (Semantic Analysis) → IR (Intermediate Representation with Three-Address Code) → SSA Transformation → Optimizations → Backend / Code Generation (planned).

Each stage is intentionally isolated and inspectable.

---

Memory Management Strategy

Fast allocation, cheap deallocation (bulk-free)

Especially suited for:

• AST nodes

• IR instructions

• Temporary compiler data

This mirrors techniques used in real compilers (LLVM, GCC).

---

Error Reporting & Diagnostics

A major focus of Xyn is high-quality diagnostics:

• Colored error output.
  Error message colored white, Error location with red, and others with green.

• Exact source highlighting.
  Xyn's Error Engine use span location from every Error class and then display it to the screen.

• Multiple severity levels.
  Xyn's errors divided line by line, start from one and so on orderly.

• Clear messages specified to aimed at humans, not machines.

The goal is for error messages to teach, not merely report failure.

---

What are done for now:

Features:

[version 0.01]

• Basic variable declaration.
• Variable declaration supports static typing and type inference.
• Supports three types (i.e, String, Integer, Float)

Technical:

[version 0.01]

• Lexer, supports assignment operator (e.g., ident, equal sign, let keyword, etc.), three values (i.e, int, float, string), and operation operators.
• Parser, supports basic declarations and error recovery with maximum lookahead are 5.
• Semantic, have a type checking task that check expression type and compare it with it's declared type (if declared).
• HIR (High IR) generator, can generate basic declarations IR based on three address code, do type conversion, and do optimization to the generated IR.
• HIR Pass (subset of IR generator), do optimization on the generated IR for multiple pass.
• Error Engine, stores all error. When reporting, it separates all the errors line by line orderly from least to greatest, and display the errors with pretty output.

---

Planned & Ongoing Work

Disclaimer: Feature or technical stuff maybe appear not as the expected version.

Feature:

[for version 0.05]

• Dynamic Variable. [ ]
• Simple control flow statement (e.g., if and else statement). [ ]

Technical:

[for version 0.02]

• Completing the HIR Pass and makes the optimization works. [X]
• Clean the HIR code and make it more modular. [X]

[for version 0.03]

• Make a LIR (Low IR) Generator that generate VM-ready instruction from the HIR. [X]

[for version 0.04]

• Make a LIR storer that will store the LIR code into a .xir file [X]
• Make a VM that used the LIR and execute it. [X]

[for version 0.045]

• Cleaner code. [ ]
• Modular code. [ ]
• Faster runtime. [ ]
• Making comments to some parts of the code, so it can be more readable. [ ]
• review all of it, and planning for the next feature. [ ]

[for version 0.046]

• revise and rewrite the README file. [ ]

---

Inspirations & References

Xyn is inspired by:

• LLVM IR & SSA design.

• JVM and bytecode-based architectures.

• ML-family compiler theory (conceptually).

• Classic compiler literature (Dragon Book–style pipelines).

• And some features from other programming languages.

While inspired by many systems, Xyn deliberately avoids copying any single one.

---

Contribution & Collaboration

This project is open to critique, discussion, and collaboration.

Design criticism is welcome.

Alternative approaches are encouraged.

Discussions about trade-offs are valued.

If you are interested in:

• Compiler design.

• Programming language theory.

• IR / SSA / optimizations.

Feel free to open issues, propose changes, or start discussions.

---

Status

Xyn is an active experimental project.

Expect:

• Breaking changes

• Refactors

• Design evolution

The primary goal is deep understanding, not short-term stability.

---

Xyn exists to explore how programming languages are built — deliberately, transparently, and thoughtfully.