crafting-interpreters/rust/rox/src/parser.rs

use thiserror::Error;
use tracing::error;

use crate::{
    expression::Expression,
    statement::Statement,
    token::{
        self, Token,
        TokenType::{self, *},
    },
};

#[derive(Error, Debug)]
pub enum ParserError {
    #[error("empty token stream")]
    NoTokens,
    #[error("[line {0}] expected expression")]
    ExpressionExpected(usize),
    #[error("[line {0}] expected ')' after expression.")]
    ParenAfterExpression(usize),
    #[error("[Out of bounds access at index {0}.")]
    OutOfBoundsAccess(usize),
    #[error("[line {0}] literal expected.")]
    LiteralExpected(usize),
    #[error("[line {0}] expected ';' after value.")]
    SemicolonAfterValueExpected(usize),
    #[error("[line {0}] expected ';' after expression.")]
    SemicolonAfterExpressionExpected(usize),
    #[error("[line {0}] expected variable name.")]
    VariableNameExpected(usize),
    #[error("[line {0}] invalid assignment target.")]
    InvalidAssignmentTarget(usize),
    #[error("[line {0}] expected '}}' after block.")]
    RightBraceAfterBlockExpected(usize),
}

/// Parse the Lox language.
#[derive(Debug, Clone)]
struct Parser {
    current: usize,
    current_token: Token,
    tokens: Vec<Token>,
}

impl Parser {
    /// Create a new parser instance, fail if the tokens vector is empty.
    fn new(tokens: Vec<Token>) -> Result<Self, ParserError> {
        let current_token = tokens.first().ok_or(ParserError::NoTokens)?.clone();

        Ok(Self {
            current: 0,
            current_token,
            tokens,
        })
    }

    /// Parse all tokens to a list of statements for execution.
    fn run(&mut self) -> Result<Vec<Statement>, ParserError> {
        let mut statements = Vec::new();

        while !self.is_at_end() {
            match self.declaration() {
                Ok(x) => statements.push(x),
                Err(e) => {
                    error!("{e}");
                    self.synchronize();
                }
            }
        }

        Ok(statements)
    }

    /// Check if any of the provided types match the type of the current token.
    ///
    /// If so, advance the current token.
    fn matches(&mut self, types: &[TokenType]) -> bool {
        let matches = types.iter().any(|x| self.check(x));
        matches.then(|| self.advance());
        matches
    }

    /// Return true if the current token type matches the match_type, false otherwise.
    fn check(&self, match_type: &TokenType) -> bool {
        self.current_token.token_type == *match_type
    }

    /// Advance the current token if we have not hit Eof yet.
    ///
    /// Return the token before the advancement.
    fn advance(&mut self) -> Result<&Token, ParserError> {
        if !self.is_at_end() {
            self.current += 1;
            self.current_token = self
                .tokens
                .get(self.current)
                .ok_or(ParserError::OutOfBoundsAccess(self.current))?
                .clone();
        }

        self.previous()
    }

    /// Return true if the current token is Eof, false otherwise.
    fn is_at_end(&self) -> bool {
        self.current_token.token_type == Eof
    }

    /// Return the token before the current one or an error if there is none.
    fn previous(&self) -> Result<&Token, ParserError> {
        self.tokens
            .get(self.current - 1)
            .ok_or_else(|| ParserError::OutOfBoundsAccess(self.current - 1))
    }

    /// Consume the current token if its token type matches the provided token_type and advance the
    /// current token. Otherwise return None..
    fn consume(&mut self, token_type: &TokenType) -> Option<&Token> {
        if self.check(token_type) {
            self.advance().ok()
        } else {
            None
        }
    }

    /// Parse a binary expression using the next_precedence function and operators to match.
    fn binary_expr(
        &mut self,
        next_precedence: impl Fn(&mut Self) -> Result<Expression, ParserError>,
        operators: &[TokenType],
    ) -> Result<Expression, ParserError> {
        let mut expr = next_precedence(self)?;

        while self.matches(operators) {
            let operator = self.previous()?.clone();
            let right = next_precedence(self)?;
            expr = Expression::Binary {
                left: Box::new(expr.clone()),
                operator,
                right: Box::new(right),
            };
        }
        Ok(expr)
    }

    /// expression -> equality ;
    fn expression(&mut self) -> Result<Expression, ParserError> {
        self.assignment()
    }

    fn declaration(&mut self) -> Result<Statement, ParserError> {
        if self.matches(&[Var]) {
            self.var_declaration()
        } else {
            self.statement()
        }
    }

    fn statement(&mut self) -> Result<Statement, ParserError> {
        if self.matches(&[Print]) {
            self.print_statement()
        } else if self.matches(&[LeftBrace]) {
            Ok(Statement::Block(self.block()?))
        } else {
            self.expression_statement()
        }
    }

    fn print_statement(&mut self) -> Result<Statement, ParserError> {
        let value = self.expression()?;
        let line = self.current_token.line;
        self.consume(&Semicolon)
            .ok_or(ParserError::SemicolonAfterValueExpected(line))?;

        Ok(Statement::Print(value))
    }

    fn var_declaration(&mut self) -> Result<Statement, ParserError> {
        let line = self.current_token.line;
        let name = self
            .consume(&Identifier)
            .ok_or(ParserError::VariableNameExpected(line))?
            .clone();

        let initializer = if self.matches(&[Equal]) {
            Some(self.expression()?)
        } else {
            None
        };

        self.consume(&Semicolon)
            .ok_or(ParserError::SemicolonAfterExpressionExpected(line))?;

        Ok(Statement::Var {
            name,
            initializer: Box::new(initializer),
        })
    }

    fn expression_statement(&mut self) -> Result<Statement, ParserError> {
        let expr = self.expression()?;
        let line = self.current_token.line;
        self.consume(&Semicolon)
            .ok_or(ParserError::SemicolonAfterExpressionExpected(line))?;

        Ok(Statement::Expression(expr))
    }

    fn block(&mut self) -> Result<Vec<Statement>, ParserError> {
        let mut statements = Vec::new();

        while !self.check(&RightBrace) && !self.is_at_end() {
            statements.push(self.declaration()?);
        }

        let line = self.previous()?.line;
        self.consume(&RightBrace)
            .ok_or(ParserError::RightBraceAfterBlockExpected(line))?;

        Ok(statements)
    }

    fn assignment(&mut self) -> Result<Expression, ParserError> {
        let expr = self.equality()?;

        if self.matches(&[Equal]) {
            let equals = self.previous()?.clone();
            let value = self.assignment()?;

            if let Expression::Variable { name } = expr {
                Ok(Expression::Assign {
                    name,
                    value: Box::new(value),
                })
            } else {
                Err(ParserError::InvalidAssignmentTarget(equals.line))
            }
        } else {
            Ok(expr)
        }
    }

    /// equality -> comparison ( ( "!=" | "==" ) comparison )* ;
    fn equality(&mut self) -> Result<Expression, ParserError> {
        self.binary_expr(Self::comparison, &[BangEqual, EqualEqual])
    }

    /// comparison -> term ( ( ">" | ">=" | "<" | "<=" ) term )* ;
    fn comparison(&mut self) -> Result<Expression, ParserError> {
        self.binary_expr(Self::term, &[Greater, GreaterEqual, Less, LessEqual])
    }

    /// term -> factor ( ( "-" | "+" ) factor )* ;
    fn term(&mut self) -> Result<Expression, ParserError> {
        self.binary_expr(Self::factor, &[Minus, Plus])
    }

    /// factor -> unary ( ( "/" | "*" ) unary )* ;
    fn factor(&mut self) -> Result<Expression, ParserError> {
        self.binary_expr(Self::unary, &[Slash, Star])
    }

    /// unary -> ( "!" | "-" ) unary | primary ;
    fn unary(&mut self) -> Result<Expression, ParserError> {
        if self.matches(&[Bang, Minus]) {
            let operator = self.previous()?.clone();
            let right = self.unary()?;

            Ok(Expression::Unary {
                operator,
                right: Box::new(right),
            })
        } else {
            self.primary()
        }
    }

    /// primary -> NUMBER | STRING | "true" | "false" | "nil" | "(" expression ")" ;
    fn primary(&mut self) -> Result<Expression, ParserError> {
        if self.matches(&[False]) {
            Ok(Expression::Literal {
                value: token::Literal::Boolean(false),
            })
        } else if self.matches(&[True]) {
            Ok(Expression::Literal {
                value: token::Literal::Boolean(true),
            })
        } else if self.matches(&[Nil]) {
            Ok(Expression::Literal {
                value: token::Literal::Nil,
            })
        } else if self.matches(&[Identifier]) {
            let prev = self.previous()?.clone();
            Ok(Expression::Variable { name: prev })
        } else if self.matches(&[Number, String]) {
            let prev = self.previous()?;
            let value = prev
                .literal
                .clone()
                .ok_or(ParserError::LiteralExpected(prev.line))?;

            Ok(Expression::Literal { value })
        } else if self.matches(&[LeftParen]) {
            let expr = self.expression()?;
            let line = self.current_token.line;
            self.consume(&RightParen)
                .ok_or(ParserError::ParenAfterExpression(line))?;

            Ok(Expression::Grouping {
                expression: Box::new(expr),
            })
        } else {
            let prev = self.previous()?;
            Err(ParserError::ExpressionExpected(prev.line))
        }
    }

    fn synchronize(&mut self) {
        let _ = self.advance();
        while !self.is_at_end()
            && self.previous().unwrap().token_type != Semicolon
            && !&[Class, Fun, Var, For, If, While, Print, Return]
                .contains(&self.current_token.token_type)
        {
            let _ = self.advance();
        }
    }
}

/// Try to parse the provided tokens into an AST.
pub fn ast(tokens: Vec<Token>) -> Result<Vec<Statement>, ParserError> {
    let mut parser = Parser::new(tokens)?;
    parser.run()
}