Doughlang/src/ast.rs

//! The Abstract Syntax Tree defines an interface between the parser and type checker

use crate::span::Span;

pub mod macro_matcher;

pub mod visit;

pub mod fold;

mod display;

/// An annotation: extra data added on to important AST nodes.
pub trait Annotation: Clone + std::fmt::Display + std::fmt::Debug + PartialEq + Eq {}
impl<T: Clone + std::fmt::Debug + std::fmt::Display + PartialEq + Eq> Annotation for T {}

/// A value with an annotation.
#[derive(Clone, PartialEq, Eq)]
pub struct Anno<T: Annotation, A: Annotation = Span>(pub T, pub A);

/// Expressions: The beating heart of Dough.
///
/// A program in Doughlang is a single expression which, at compile time,
/// sets up the state in which a program will run. This expression binds types,
/// functions, and values to names which are exposed at runtime.
///
/// Whereas in the body of a function, `do` sequences are ordered, in the global
/// scope (or subsequent module scopes, which are children of the global module,)
/// `do` sequences are considered unordered, and subexpressions may be reordered
/// in whichever way the compiler sees fit. This is especially important when
/// performing import resolution, as imports typically depend on the order
/// in which names are bound.
#[derive(Clone, Debug, PartialEq, Eq)]
pub enum Expr<A: Annotation = Span> {
    /// Omitted by semicolon insertion-elision rules
    Omitted,
    /// An identifier
    Id(Path),
    /// An escaped token for macro binding
    MetId(String),
    /// A literal bool, string, char, or int
    Lit(Literal),
    /// use Use
    Use(Use),
    /// `(let | const | static) Pat::NoTopAlt (= expr (else expr)?)?` |
    /// `(fn | mod | impl) Pat::Fn Expr`
    Bind(Box<Bind<A>>),
    /// Expr { (Ident (: Expr)?),* }
    Make(Box<Make<A>>),
    /// Op Expr | Expr Op | Expr (Op Expr)+ | Op Expr Expr else Expr
    Op(Op, Vec<Anno<Self, A>>),
}

/// Doughlang's AST is partitioned by data representation, so it
/// considers any expression which is composed solely of keywords,
/// symbols, and other expressions as operator expressions.
///
/// This includes:
/// - Do-sequence expressions: `Expr ; Expr `
/// - Type-cast expressions `Expr as Expr`
/// - Binding-modifier expressions: `pub Expr`, `#[Expr] Expr`
/// - Block and Group expressions: `{Expr?}`, `(Expr?)`
/// - Control flow: `if`, `while`, `loop`, `match`, `break`, `return`
/// - Function calls `Expr (Expr,*)`
/// - Traditional binary and unary operators (add, sub, neg, assign)
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum Op {
    // -- true operators
    Do,    // Expr ; Expr
    As,    // Expr as Expr
    Macro, // macro { (Pat => Expr)* }
    Block, // { Expr }
    Array, // [ Expr,* ]
    ArRep, // [ Expr ; Expr ]
    Group, // ( Expr ,?)
    Tuple, // Expr (, Expr)*
    Meta,  // #[ Expr ]

    Try,   // Expr '?'
    Index, // Expr [ Expr,* ]
    Call,  // Expr ( Expr,* )

    Pub,      // pub Expr
    Loop,     // loop Expr
    Match,    // match Expr { <Let(Match, ..)>,* }
    If,       // if Expr Expr (else Expr)?
    While,    // while Expr Expr (else Expr)?
    Break,    // break Expr
    Return,   // return Expr
    Continue, // continue

    Dot, // Expr . Expr

    RangeEx,  // Expr? ..Expr
    RangeIn,  // Expr? ..=Expr
    Neg,      // -Expr
    Not,      // !Expr
    Identity, // !!Expr
    Refer,    // &Expr
    Deref,    // *Expr

    Mul, // Expr * Expr
    Div, // Expr / Expr
    Rem, // Expr % Expr

    Add, // Expr + Expr
    Sub, // Expr - Expr

    Shl, // Expr << Expr
    Shr, // Expr >> Expr

    And, // Expr & Expr
    Xor, // Expr ^ Expr
    Or,  // Expr | Expr

    Lt,  // Expr < Expr
    Leq, // Expr <= Expr
    Eq,  // Expr == Expr
    Neq, // Expr != Expr
    Geq, // Expr >= Expr
    Gt,  // Expr > Expr

    LogAnd, // Expr && Expr
    LogXor, // Expr ^^ Expr
    LogOr,  // Expr || Expr

    Set,    // Expr = Expr
    MulSet, // Expr *= Expr
    DivSet, // Expr /= Expr
    RemSet, // Expr %= Expr
    AddSet, // Expr += Expr
    SubSet, // Expr -= Expr
    ShlSet, // Expr <<= Expr
    ShrSet, // Expr >>= Expr
    AndSet, // Expr &= Expr
    XorSet, // Expr ^= Expr
    OrSet,  // Expr |= Expr
}

/// A qualified identifier
///
/// TODO: qualify identifier
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct Path {
    // TODO: Identifier interning
    pub parts: Vec<String>,
    // TODO: generic parameters
}

/// A literal value (boolean, character, integer, string)
#[derive(Clone, Debug, PartialEq, Eq)]
pub enum Literal {
    /// A boolean literal: true | false
    Bool(bool),
    /// A character literal: 'a', '\u{1f988}'
    Char(char),
    /// An integer literal: 0, 123, 0x10
    Int(u128, u32),
    /// A string literal:
    Str(String),
}

/// A compound import declaration
#[derive(Clone, Debug, PartialEq, Eq)]
pub enum Use {
    /// "*"
    Glob,
    /// Identifier
    Name(String),
    /// Identifier :: Use
    Path(String, Box<Use>),
    /// { Use, * }
    Tree(Vec<Use>),
}

/// A pattern binding
/// ```ignore
/// let    Pat (= Expr (else Expr)?)?
/// const  Pat (= Expr (else Expr)?)?
/// static Pat (= Expr (else Expr)?)?
/// type   Pat (= Expr)?
/// struct Pat
/// enum   Pat
/// fn     Pat Expr
/// mod    Pat Expr
/// impl   Pat Expr
/// Pat => Expr // in match
/// ```
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct Bind<A: Annotation = Span>(
    pub BindOp,
    pub Vec<Path>,
    pub Pat,
    pub Vec<Anno<Expr<A>, A>>,
);

#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum BindOp {
    /// A `let Pat (= Expr (else Expr)?)?` binding
    Let,
    /// A `const Pat = Expr` binding
    Const,
    /// A `static Pat = Expr` binding
    Static,
    /// A type-alias binding
    Type,
    /// A `fn Pat Expr` binding
    Fn,
    /// A `mod Pat Expr` binding
    Mod,
    /// An `impl Pat Expr` binding
    Impl,
    /// A struct definition
    Struct,
    /// An enum definition
    Enum,
    /// A `Pat => Expr` binding
    Match,
}

/// A make (constructor) expression
/// ```ignore
/// Expr { (Ident (: Expr)?),* }
/// ```
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct Make<A: Annotation = Span>(pub Anno<Expr<A>, A>, pub Vec<MakeArm<A>>);

/// A single "arm" of a make expression
/// ```text
/// Identifier (':' Expr)?
/// ```
#[derive(Clone, Debug, PartialEq, Eq)]
pub struct MakeArm<A: Annotation = Span>(pub String, pub Option<Anno<Expr<A>, A>>);

/// Binding patterns for each kind of matchable value.
///
/// This covers both patterns in Match expressions, and type annotations.
#[derive(Clone, Debug, PartialEq, Eq)]
pub enum Pat {
    /// Matches anything without binding
    Ignore,
    /// Matches nothing, ever
    Never,
    /// Matches nothing; used for macro substitution
    MetId(String),
    /// Matches anything, and binds it to a name
    Name(String),
    /// Matches against a named const value
    Path(Path),
    /// Matches a Struct Expression `Ident { Pat }`
    NamedStruct(Path, Box<Pat>),
    /// Matches a Tuple Struct Expression `Ident ( Pat )`
    NamedTuple(Path, Box<Pat>),
    /// Matches a literal value by equality comparison
    Lit(Literal),
    /// Matches a compound pattern
    Op(PatOp, Vec<Pat>),
}

/// Operators on lists of patterns
#[derive(Clone, Copy, Debug, PartialEq, Eq)]
pub enum PatOp {
    /// Changes the visibility mode to "public"
    Pub,
    /// Changes the binding mode to "mutable"
    Mut,
    /// Matches the dereference of a pointer (`&pat`)
    Ref,
    /// Matches the dereference of a raw pointer (`*pat`)
    Ptr,
    /// Matches a partial decomposition (`..rest`) or upper-bounded range (`..100`)
    Rest,
    /// Matches an exclusive bounded range (`0..100`)
    RangeEx,
    /// Matches an inclusive bounded range (`0..=100`)
    RangeIn,
    /// Matches the elements of a tuple
    Tuple,
    /// Matches the elements of a slice or array
    Slice,
    /// Matches a constant-size slice with repeating elements
    Arrep,
    /// Matches a type annotation or struct member
    Typed,
    /// Matches a function signature
    Fn,
    /// Matches one of a list of alternatives
    Alt,
}

impl<T: Annotation, A: Annotation> std::fmt::Debug for Anno<T, A> {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        <A as std::fmt::Debug>::fmt(&self.1, f)?;
        f.write_str(": ")?;
        <T as std::fmt::Debug>::fmt(&self.0, f)
    }
}
impl<A: Annotation> Default for Expr<A> {
    fn default() -> Self {
        Self::Op(Op::Tuple, vec![])
    }
}

impl<A: Annotation> Expr<A> {
    pub const fn anno(self, annotation: A) -> Anno<Expr<A>, A> {
        Anno(self, annotation)
    }

    pub fn and_do(self, annotation: A, other: Anno<Expr<A>, A>) -> Self {
        let Self::Op(Op::Do, mut exprs) = self else {
            return Self::Op(Op::Do, vec![self.anno(annotation), other]);
        };
        let Anno(Self::Op(Op::Do, mut other), _) = other else {
            exprs.push(other);
            return Self::Op(Op::Do, exprs);
        };
        exprs.append(&mut other);
        Self::Op(Op::Do, exprs)
    }

    pub const fn is_place(&self) -> bool {
        matches!(
            self,
            Self::Id(_) | Self::Op(Op::Index | Op::Dot | Op::Deref, _)
        )
    }

    #[allow(clippy::type_complexity)]
    pub const fn as_slice(&self) -> Option<(Op, &[Anno<Expr<A>, A>])> {
        match self {
            Expr::Op(op, args) => Some((*op, args.as_slice())),
            _ => None,
        }
    }
}

impl From<&str> for Path {
    fn from(value: &str) -> Self {
        Self { parts: vec![value.to_owned()] }
    }
}