ll-lang User Guide

A practical guide for programmers (human or LLM) who want to write .lll programs, compile them with lllc, and target F#, TypeScript, Python, Java, C#, with optional experimental LLVM IR output.

ll-lang is a statically typed functional language with Hindley-Milner inference, algebraic data types, traits, semantic tags, and unit algebra.

Contents

1. Installation — build the compiler, run hello world
2. Syntax — modules, bindings, lambdas, if, match, ADTs, traits
3. Types and inference — H-M, polymorphism, annotations
4. Tags and units — tags, phantom types, unit algebra
5. Traits — declarations, impls, constrained generics, dispatch
6. Modules — module headers, imports, project mode (lll.toml)
7. Error codes — compact diagnostics (E001..E008, E020, E024..E031)
8. CLI — build/check/run/new/install/self commands (+ experimental reverse)
9. MCP server — lllc mcp tools for LLM clients
10. Compilation targets — --target fs|ts|py|java|cs|llvm
11. Java target details — Java backend mapping details
12. LLM prompting — short guide for ll-lang prompting

Status

Compiler phases 1 to 10 are complete. Current stable surface includes:

• End-to-end pipeline (lexer -> parser -> elaborator -> HM infer -> codegen)
• Multi-file projects with lll.toml and target selection
• Self-hosted stdlib/compiler bootstrap flow
• Canonical default operator fixities from Std.Operators
• MCP server for structured LLM tooling
• Targets: F#, TypeScript, Python, Java, C#

Current caveats:

• LLVM backend is intentionally subset-based compared to the primary F# backend.
• lllc reverse is available as experimental tooling and is outside 1.0 compatibility guarantees.
• See Release Contract 1.0.