83 lines
3.7 KiB
Markdown
83 lines
3.7 KiB
Markdown
# Compiler Roadmap
|
|
|
|
A Rust-flavored, C-targeting language - built pipeline-first.
|
|
|
|
**Implementation language:** Rust
|
|
**Code generation target:** Native object files (`.o`) via Cranelift JIT/AOT
|
|
|
|
## Phase 1 - Lexer
|
|
|
|
- [x] Define token enum (int literal, bool literal, ident, keywords, operators, punctuation)
|
|
- [x] Implement character-by-character scanner loop
|
|
- [x] Handle whitespace & single-line comments (`//`)
|
|
- [x] Produce source spans (file, line, col) on every token
|
|
- [x] Unit-test: known inputs → expected token streams
|
|
|
|
## Phase 2 - Parser
|
|
|
|
- [x] Write grammar for the base subset
|
|
- `fn` declarations, `return`, int/bool literals
|
|
- Arithmetic (`+`, `-`, `*`, `/`)
|
|
- [x] Implement recursive-descent parser
|
|
- [x] Attach source spans to every AST node
|
|
- [x] Emit structured parse errors with span info
|
|
- [x] Unit-test: parse valid snippets, expect correct AST shapes
|
|
|
|
## Phase 3 - Semantic Analysis
|
|
|
|
- [x] Implement scope-aware symbol table (environment)
|
|
- [x] Name resolution pass - resolve all `Ident` nodes to their declarations
|
|
- [x] Hindley-Milner type inference (unification, type variables, occurs check)
|
|
- [x] Integer literal sizing and unary minus type promotion logic
|
|
- [x] Translate untyped AST directly into a fully-typed AST (Typed AST)
|
|
- [x] Validate function return types match declared signature
|
|
- [x] Error on use-before-declaration, undeclared symbols, and type mismatches
|
|
- [x] Unit-test: HM unification, type mappings, and ill-typed program diagnostics
|
|
|
|
## Phase 4 - Code Generation via Cranelift
|
|
|
|
- [x] Integrate `cranelift-codegen`, `cranelift-frontend`, and `cranelift-object`
|
|
- [x] Implement CLI with `clap` (`--emit-ir` flag, input/output files)
|
|
- [x] Map Typed AST types (`Ty`) to Cranelift IR types
|
|
- [x] Lower functions, parameters, and variable definitions to Cranelift IR
|
|
- [x] Codegen for arithmetic, unary operations, and `return` statements
|
|
- [x] Run built-in optimization passes (constant folding, e-graphs, DCE)
|
|
- [x] Output System V AMD64 ABI-compliant `.o` machine code files
|
|
- [x] End-to-end test: compile a simple `fn` → link via `gcc` → run → correct exit code
|
|
|
|
## Phase 5 - The Ray Tracer Milestone (Current)
|
|
|
|
To successfully write a simple ray tracer, we need continuous math, data structures, and I/O. The following path establishes these prerequisites:
|
|
|
|
- [ ] **Floating-Point Support:** Add `f32`/`f64` types, decimal literals, and Cranelift lowering for `fadd`, `fmul`, etc.
|
|
- [ ] **FFI & Interop:** Implement `extern fn` declarations to bind C standard library functions (like `putchar` or `printf`) for `.ppm` image output.
|
|
- [ ] **Type Casting:** Add the `as` operator to convert floating-point color bounds `[0.0, 1.0]` into integer byte formats `[0, 255]`.
|
|
- [ ] **Pointers:** Add pointer types (`*T`), address-of (`&`), and dereference (`*`) operators.
|
|
- [ ] **Structs:** Add `struct` definitions, initializers, and field access (`ray.origin.x`) to represent 3D vectors, rays, and spheres.
|
|
- [ ] **Arrays:** Add fixed-size arrays (`[T; N]`) or heap allocations for the scene and framebuffers.
|
|
|
|
## Planned Features (Backlog)
|
|
|
|
### Control flow & Variables
|
|
- [x] booleans and comparision operators
|
|
- [x] `if` / `else` branching
|
|
- [x] `while` loops, `break`, `continue`
|
|
- [x] `let` bindings and variable assignments
|
|
|
|
### Types & memory
|
|
- [ ] Opaque pointers (`*void` / `*opaque`)
|
|
- [ ] Raw pointer arithmetic
|
|
- [ ] Slices (`&[T]` / `[]T`)
|
|
|
|
### Composite types
|
|
- [ ] Enums (C-style tagged unions)
|
|
- [ ] Pattern matching (`match` / `switch`)
|
|
|
|
### Strings & interop
|
|
- [ ] String literals
|
|
- [ ] Variadic functions (for `printf` interop)
|
|
|
|
### Tooling & backend
|
|
- [ ] Proper register allocator
|
|
- [ ] Debug info (DWARF)
|
|
- [ ] Standard library bootstrap (`print`, `malloc` wrapper) |