Write main file

README.md

@@ -34,7 +34,7 @@ def fibPrompt
   = do {
     print "Enter a number: ";
     num <- read <$> getLine;
-    sequence_ (print . fibonacci <$> [0 .. num]);
+    sequence_ (print . fibonacci <$> (0 .. num));
   };
 
 // Program entry point.

axc/src/ast2ir.rs

@@ -0,0 +1,10 @@
+//! Conversion of AST to intermediate representation.
+
+use crate::ir::IR;
+use crate::SyntaxTree;
+
+/// Compiles an abstract syntax tree into intermediate representation; this assumes the code already
+/// type-checks, and emits unoptimized IR.
+pub fn ast2ir(_: SyntaxTree) -> IR {
+    todo!()
+}

axc/src/backends/c.rs

@@ -0,0 +1,9 @@
+//! Backend for direct generation of C code.
+
+use crate::ir::IR;
+
+/// Generates C code with the same semantics as the intermediate representation. Returns the C
+/// source code as a string.
+pub fn generate_c(_ir: &IR) -> String {
+    todo!()
+}

axc/src/backends/mod.rs

@@ -0,0 +1,3 @@
+//! Different back-ends for code generation.
+
+pub mod c;

axc/src/ir.rs

@@ -0,0 +1,5 @@
+//! Intermediate code representation.
+
+/// The IR tree representing the whole program.
+#[derive(Debug)]
+pub struct IR;

axc/src/lib.rs

@@ -2,6 +2,14 @@
 //!
 //! AlexScript is a based programming language, for based people.
 
+#![deny(missing_docs)]
+
+pub mod ast2ir;
+pub mod ir;
+pub mod parser;
+pub mod typeck;
+pub mod backends;
+
 use num_bigint::BigUint;
 
 /// A concrete syntax tree. This represents the full content of an AlexScript program, including all
@@ -88,46 +96,70 @@ pub struct TypeConstructor {
 /// Expressions.
 pub enum Expr {
     /// Unary operators, e.g., `-5`.
-    UnaryOp { kind: String, val: Box<Expr> },
+    UnaryOp {
+        /// The text of the operator.
+        kind: String,
+
+        /// The value being operated upon.
+        val: Box<Expr>,
+    },
 
     /// Binary operators, e.g., `5 + 5`.
     BinaryOp {
+        /// The text of the operator.
         kind: String,
+
+        /// The left side of the operator.
         left: Box<Expr>,
+
+        /// The right side of the operator.
         right: Box<Expr>,
     },
 
     /// Function application, e.g., `sin x`.
     Application {
+        /// The function being applied. For curried functions with multiple arguments (e.g., `atan2
+        /// y x`), this is another expression of type `Application`.
         func: Box<Expr>,
+
+        /// The argument to which the function is being applied.
         argument: Box<Expr>,
     },
 
     /// Defining of temporary variables, e.g., `let x = 5 in x + x`.
-    Let { left: Pattern, right: Box<Expr> },
+    Let {
+        /// The pattern being bound.
+        left: Pattern,
+
+        /// The variable the pattern is matching.
+        right: Box<Expr>,
+
+        /// The expression the pattern is being substituted into.
+        into: Box<Expr>,
+    },
 
     /// Matching of multiple cases, e.g., `match x { 5 => 'a', 6 => 'b' }`.
     Match {
+        /// The expression being matched upon.
         matcher: Box<Expr>,
+
+        /// The possible cases of the `match` expression.
         cases: Vec<(Pattern, Expr)>,
     },
 
-    /// Syntax sugar for matching on booleans, e.g., `if foo then bar else baz`.
+    /// Record initialization, e.g., `{ pointer: xyz, length: 12 }`.
-    If {
+    Record {
-        subject: Box<Expr>,
+        /// The elements of the record.
-        iftrue: Box<Expr>,
-        iffalse: Box<Expr>,
-    },
-
-    /// Struct initialization, e.g., `Vector { pointer: xyz, length: 12 }`.
-    StructInit {
-        name: String,
         elements: Vec<(String, Expr)>,
     },
 
     /// Anonymous functions, e.g., `fn x -> x + 1`.
     Lambda {
+        /// Arguments to the lambda; multiple of these are equivalent to stacking lambdas by
+        /// currying.
         arguments: Vec<Pattern>,
+
+        /// The result of the lambda.
         result: Box<Expr>,
     },
 
@@ -135,11 +167,20 @@ pub enum Expr {
     VariableReference(Vec<String>),
 
     /// Dot subscripts, e.g., `foo.bar`.
-    DotSubscript { value: Box<Expr>, subscript: String },
+    DotSubscript {
+        /// The left side of the subscript.
+        value: Box<Expr>,
+
+        /// The right side of the subscript; this is only allowed to be a single word.
+        subscript: String,
+    },
 
     /// Bracket subscripts, e.g., `foo[bar]`.
     BracketSubscript {
+        /// The left side of the subscript.
         value: Box<Expr>,
+
+        /// The right side of the subscript.
         subscript: Box<Expr>,
     },
 
@@ -154,8 +195,11 @@ pub enum Type {
 
     /// `List Int`
     Application {
-        // TODO: is this right?
+        /// The function being applied. This must be a generic type.
-        function: Box<Expr>,
+        function: Box<Type>,
+
+        /// The expression given as an argument to the type. This can be any expression, to allow
+        /// const generics; in most cases, though, it should be just a normal type.
         expression: Box<Expr>,
     },
 
@@ -174,9 +218,11 @@ pub enum Pattern {
 
     /// `a: String`
     TypeAnnotated {
+        /// The pattern being annotated.
         pat: Box<Pattern>,
-        // Note that types are expressions, to simplify parsing.
+
-        typ: Box<Expr>,
+        /// The type that `pat` is being asserted to have.
+        typ: Box<Type>,
     },
 
     /// `Foo`

axc/src/main.rs

@@ -1,3 +1,276 @@
-fn main() {
+//! AlexScript CLI.
-    println!("Hello, world!");
+
+use std::{error::Error, fmt::Display, fs::File, io::Write, process::exit, str::FromStr};
+
+use axc::{
+    ast2ir::ast2ir,
+    backends,
+    parser::{parser, ParserError},
+    typeck::typeck,
+};
+use clap::Parser;
+
+/// Optimization levels.
+#[derive(Debug)]
+enum Optimization {
+    Numeric(usize),
+    Size,
+    Debugging,
+    Speed,
+}
+
+impl Display for Optimization {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(
+            f,
+            "{}",
+            match self {
+                Optimization::Numeric(n) => format!("{}", n),
+                Optimization::Size => "z".to_string(),
+                Optimization::Debugging => "g".to_string(),
+                Optimization::Speed => "fast".to_string(),
+            }
+        )
+    }
+}
+
+impl FromStr for Optimization {
+    type Err = String;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        use Optimization::*;
+
+        let levels = [
+            (
+                "0",
+                "Disable all optimizations, for best results when debugging.",
+                Numeric(0),
+            ),
+            (
+                "1",
+                "Enable basic optimizations, without significantly affecting build times.",
+                Numeric(1),
+            ),
+            (
+                "2",
+                "Enable standard optimizations for release builds.",
+                Numeric(2),
+            ),
+            (
+                "3",
+                "Optimize for speed at the expense of size.",
+                Numeric(3),
+            ),
+            (
+                "z",
+                "Aggresively optimize for size rather than speed.",
+                Size,
+            ),
+            (
+                "g",
+                "Enable some optimizations that do not interfere with debugging.",
+                Debugging,
+            ),
+            (
+                "fast",
+                "Optimize for speed at the expense of strict standards compliance.",
+                Speed,
+            ),
+        ];
+
+        if s == "list" {
+            eprintln!(
+                "{}",
+                levels
+                    .iter()
+                    .map(|(name, description, _)| format!("\n    -O{name} : {description}"))
+                    .fold("optimization levels:".to_string(), |s, line| s + &line)
+            );
+            exit(1)
+        } else {
+            levels
+                .into_iter()
+                .find(|(name, _, _)| name == &s)
+                .map(|(_, _, opt)| opt)
+                .ok_or_else(|| "use -Olist to list".to_string())
+        }
+    }
+}
+
+/// Targets for compiling to.
+#[derive(Debug)]
+enum Target {
+    // -------------------- C generation --------------------
+
+    // Highest priority codegen, since it allows us to compile to the vast majority of possible
+    // targets.
+    /// Directly generate C source code.
+    CSource,
+
+    /// Pass generated C code into the system C compiler, and emit an assembly code file.
+    Assembly,
+
+    /// Pass Assembly code into the system assembler, and emit an object file.
+    ObjectFile,
+
+    /// Link generated object files using the system linker, and generate an executable file.
+    Executable,
+
+    /// Link generated object files using the system linker, and generate a shared object file.
+    SharedObject,
+
+    // -------------------- GPU generation --------------------
+
+    // Medium-priority codegen, since GPUs typically can't be portably programmed in C and GPU
+    // programming is a minor goal of the language.
+    /// Directly generate SPIR-V code.
+    Spirv,
+
+    // TODO: add more GPU targets derived from spirv.
+
+    // -------------------- WebAssembly generation --------------------
+
+    // Very low-priority codegen, since efficient WebAssembly can be generated from C.
+    /// Pass generated C code into the WebAssembly compiler, and generate WebAssembly text format
+    /// code. At some point this might get replaced with direct WAT generation.
+    Wat,
+
+    /// Pass generated WAT code into a WebAssembly compiler, and generate WebAssembly binary format
+    /// code.
+    Wasm,
+
+    // -------------------- Other language generation --------------------
+    /// Directly generate Lua code.
+    // Medium-priority codegen, since Lua is the only type of code that runs in e.g. plugins and
+    // scripts in some software, and it is difficult to generate efficient Lua without direct
+    // support from the compiler.
+    Lua,
+
+    /// Directly generate Python code.
+    // Low-priority codegen; the same situation as Lua, but situations that require Python code with
+    // no alternatives are much less common than situations that require Lua code with no
+    // alternatives.
+    Python,
+
+    /// Directly generate Go code.
+    // Extremely low-priority codegen; almost no valid use cases.
+    Go,
+
+    /// Directly generate Ada code.
+    // Currently zero-priority codegen; no valid use cases whatsoever as far as I (Alex) can
+    // determine.
+    Ada,
+}
+
+impl Display for Target {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(
+            f,
+            "{}",
+            match self {
+                Target::CSource => "c",
+                Target::Assembly => "asm",
+                Target::ObjectFile => "obj",
+                Target::Executable => "exe",
+                Target::SharedObject => "so",
+                Target::Spirv => "spirv",
+                Target::Wat => "wat",
+                Target::Wasm => "wasm",
+                Target::Lua => "lua",
+                Target::Python => "py",
+                Target::Go => "go",
+                Target::Ada => "ada",
+            }
+        )
+    }
+}
+
+impl FromStr for Target {
+    type Err = String;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        use Target::*;
+
+        let targets = [
+            ("c", "C source code.", CSource),
+            ("asm", "Assembly code.", Assembly),
+            ("obj", "Object file.", ObjectFile),
+            ("exe", "Executable binary file.", Executable),
+            ("so", "Shared object file.", SharedObject),
+            ("spirv", "Spir-V source code.", Spirv),
+            ("wat", "WebAssembly text format code.", Wat),
+            ("wasm", "WebAssembly binary format file.", Wasm),
+            ("lua", "Lua source code.", Lua),
+            ("python", "Python source code.", Python),
+            ("go", "Go source code.", Go),
+            ("ada", "Ada source code.", Ada),
+        ];
+
+        if s == "list" {
+            eprintln!(
+                "{}",
+                targets
+                    .iter()
+                    .map(|(name, description, _)| format!("\n    -t{name} : {description}"))
+                    .fold("targets:".to_string(), |s, line| s + &line)
+            );
+            exit(1)
+        } else {
+            targets
+                .into_iter()
+                .find(|(name, _, _)| name == &s)
+                .map(|(_, _, target)| target)
+                .ok_or_else(|| "use -tlist to list".to_string())
+        }
+    }
+}
+
+/// The AlexScript compiler.
+#[derive(Parser, Debug)]
+#[clap(version = "0.1.0")]
+struct Args {
+    /// Output file to generate.
+    #[clap(short = 'o')]
+    output: Option<String>,
+
+    /// Optimization level. Use `-Olist` to list possible optimization levels.
+    #[clap(short = 'O', default_value_t = Optimization::Numeric(0))]
+    optimization: Optimization,
+
+    /// Generate debug information in resultant code, if possible.
+    #[clap(short = 'g')]
+    debug_syms: bool,
+
+    /// Target to generate code for. Use `-tlist` to list possible targets.
+    #[clap(short = 't', default_value_t = Target::Executable)]
+    target: Target,
+
+    /// The source file to compile.
+    source_file: String,
+}
+
+fn main() {
+    fn main_e() -> Result<(), Box<dyn Error>> {
+        let args = Args::parse();
+
+        let source = std::fs::read_to_string(&args.source_file)?;
+        let ast = chumsky::Parser::parse(&parser(), source).map_err(ParserError)?;
+        typeck(&ast)?;
+
+        let ir = ast2ir(ast);
+        let c = backends::c::generate_c(&ir);
+
+        let mut out_file = File::create("out.c")?;
+        write!(out_file, "{}", c)?;
+
+        Ok(())
+    }
+
+    match main_e() {
+        Ok(()) => (),
+        Err(e) => {
+            eprintln!("axc fatal error: {}", e);
+            exit(1);
+        }
+    }
 }

axc/src/parser.rs

@@ -0,0 +1,28 @@
+//! AlexScript parser.
+
+use std::{error::Error, fmt::Display};
+
+use chumsky::{
+    prelude::{filter, Simple},
+    Parser,
+};
+
+/// Adapter to make `chumsky`'s parser errors usable as standard Rust errors.
+#[derive(Debug)]
+pub struct ParserError(pub Vec<Simple<char>>);
+
+impl Display for ParserError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        for e in &self.0 {
+            write!(f, "{}", e)?;
+        }
+        Ok(())
+    }
+}
+
+impl Error for ParserError {}
+
+/// Parser for AlexScript code.
+pub fn parser() -> impl Parser<char, crate::SyntaxTree, Error = Simple<char>> {
+    filter(|c: &char| c.is_numeric()).map(|_| todo!())
+}

axc/src/typeck.rs

@@ -0,0 +1,22 @@
+//! Type checker.
+
+use std::{error::Error, fmt::Display};
+
+use crate::SyntaxTree;
+
+/// A compile-time type error from the user's source code.
+#[derive(Debug)]
+pub struct TypeError;
+
+impl Display for TypeError {
+    fn fmt(&self, _f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        todo!()
+    }
+}
+
+impl Error for TypeError {}
+
+/// Type-checks the syntax tree.
+pub fn typeck(_: &SyntaxTree) -> Result<(), TypeError> {
+    todo!()
+}