You cannot select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
582 lines
20 KiB
Rust
582 lines
20 KiB
Rust
pub(crate) mod tree;
|
|
|
|
use crate::error::Error;
|
|
use crate::lex::token;
|
|
use crate::lex::token::Token;
|
|
use crate::lex::Lexer;
|
|
|
|
use crate::ast::tree::Operator;
|
|
use std::fs;
|
|
use std::io;
|
|
|
|
#[derive(PartialEq)]
|
|
enum Scope {
|
|
File,
|
|
Target,
|
|
DepList,
|
|
SourceList,
|
|
}
|
|
|
|
struct Parser {
|
|
lexer: Lexer,
|
|
scope: Vec<Scope>,
|
|
filename: String,
|
|
}
|
|
|
|
pub fn parse(filename: String) -> io::Result<Result<tree::Node, Error>> {
|
|
let raw: String = fs::read_to_string(filename.clone())?;
|
|
let mut p = Parser::new(filename, raw);
|
|
Ok(p.parse_file())
|
|
}
|
|
|
|
/// All of the functions expect the leading token to not be consumed yet,
|
|
/// meaning you need to use `self.lexer.peek()` when determining what other
|
|
/// parsing function to call next. They consume every token up to and
|
|
/// *including* the terminating one (like a semicolon or closing brace).
|
|
///
|
|
/// Note: For now, the production rules for grammatical elements are more
|
|
/// like "freestyle guidelines" to help people understand the code.
|
|
/// In their final form, they will most likely be much more restrictive.
|
|
impl Parser {
|
|
pub fn new(filename: String, raw: String) -> Parser {
|
|
let lexer = Lexer::new(filename.clone(), raw);
|
|
Parser {
|
|
lexer,
|
|
scope: Vec::new(),
|
|
filename,
|
|
}
|
|
}
|
|
|
|
/// ```notrust
|
|
/// File
|
|
/// : Statement [ File ]
|
|
/// ```
|
|
pub fn parse_file(&mut self) -> Result<tree::Node, Error> {
|
|
let mut nodes = Vec::new();
|
|
self.scope.push(Scope::File);
|
|
|
|
while self.lexer.peek().is_some() {
|
|
nodes.push(self.parse_stmt()?);
|
|
}
|
|
|
|
self.scope.pop();
|
|
Ok(tree::Node::File {
|
|
name: self.filename.clone(),
|
|
content: nodes,
|
|
})
|
|
}
|
|
|
|
/// ```notrust
|
|
/// Statement
|
|
/// : DependStatement
|
|
/// | SetStatement
|
|
/// | SourceStatement
|
|
/// | TargetStatement
|
|
/// | TypeStatement
|
|
/// | ExpressionStatement
|
|
/// ```
|
|
fn parse_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
let token = self.lexer.peek_or_err()?;
|
|
match token.kind {
|
|
token::Kind::DependKeyword => self.parse_depend_stmt(),
|
|
token::Kind::IfKeyword => self.parse_if_stmt(),
|
|
token::Kind::SetKeyword => self.parse_set_stmt(),
|
|
token::Kind::SourceKeyword => self.parse_source_stmt(),
|
|
token::Kind::TargetKeyword => self.parse_target_stmt(),
|
|
token::Kind::TypeKeyword => self.parse_type_stmt(),
|
|
token::Kind::Ident => self.parse_expr_stmt(),
|
|
_ => self.syntax_error(format!("Unexpected token {}", token), &token),
|
|
}
|
|
}
|
|
|
|
/// ```notrust
|
|
/// BlockStatement
|
|
/// : "{" [ StatementList ] "}"
|
|
///
|
|
/// StatementList
|
|
/// : Statement [ StatementList ]
|
|
/// ```
|
|
fn parse_block_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
let mut nodes = Vec::new();
|
|
self.lexer.expect_kind(token::Kind::OBrace)?;
|
|
while let Some(result) = self.lexer.peek() {
|
|
match result?.kind {
|
|
token::Kind::CBrace => {
|
|
self.lexer.next();
|
|
break;
|
|
}
|
|
_ => nodes.push(self.parse_stmt()?),
|
|
}
|
|
}
|
|
Ok(tree::Node::Block(nodes))
|
|
}
|
|
|
|
/// ```notrust
|
|
/// TargetStatement
|
|
/// : "target" Expression BlockStatement
|
|
/// ```
|
|
fn parse_target_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
self.assert_scope(Scope::File)?;
|
|
self.assert_scope_not(Scope::Target)?;
|
|
self.scope.push(Scope::Target);
|
|
|
|
self.lexer.expect_kind(token::Kind::TargetKeyword)?;
|
|
let name_token = self.lexer.expect_kind(token::Kind::Ident)?;
|
|
|
|
let children = self.parse_block_stmt()?;
|
|
|
|
self.scope.pop();
|
|
Ok(tree::Node::Target {
|
|
name: Box::new(tree::Node::Ident(name_token.raw)),
|
|
content: Box::new(children),
|
|
})
|
|
}
|
|
|
|
/// ```notrust
|
|
/// DependStatement
|
|
/// : "depend" Expression ";"
|
|
/// ```
|
|
fn parse_depend_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
self.assert_scope(Scope::Target)?;
|
|
self.scope.push(Scope::DepList);
|
|
self.lexer.expect_kind(token::Kind::DependKeyword)?;
|
|
let rvalue = self.parse_expr(&[token::Kind::Semi])?;
|
|
self.scope.pop();
|
|
Ok(tree::Node::DepList(Box::new(rvalue)))
|
|
}
|
|
|
|
/// ```notrust
|
|
/// IfStatement
|
|
/// : "if" "(" Expression ")" BlockStatement [ "else" BlockStatement ]
|
|
/// ```
|
|
fn parse_if_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
self.lexer.expect_kind(token::Kind::IfKeyword)?;
|
|
self.lexer.expect_kind(token::Kind::OParen)?;
|
|
let condition = self.parse_expr(&[token::Kind::CParen])?;
|
|
let then_block = self.parse_block_stmt()?;
|
|
let mut else_block = None;
|
|
if let Some(Ok(token)) = self.lexer.peek() {
|
|
if token.kind == token::Kind::ElseKeyword {
|
|
self.lexer.next();
|
|
else_block = Some(Box::new(self.parse_block_stmt()?));
|
|
}
|
|
}
|
|
Ok(tree::Node::IfStmt {
|
|
condition: Box::new(condition),
|
|
then_block: Box::new(then_block),
|
|
else_block,
|
|
})
|
|
}
|
|
|
|
/// ```notrust
|
|
/// SetStatement
|
|
/// : "set" AssignmentExpression ";"
|
|
/// ```
|
|
fn parse_set_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
self.assert_scope(Scope::File)?;
|
|
self.lexer.expect_kind(token::Kind::SetKeyword)?;
|
|
let expr = self.parse_expr(&[token::Kind::Semi])?;
|
|
match expr {
|
|
tree::Node::BinaryExpr { op, lhs, rhs } => {
|
|
if op == Operator::Eq {
|
|
Ok(tree::Node::SetExpr {
|
|
name: lhs,
|
|
val: rhs,
|
|
})
|
|
} else {
|
|
self.syntax_error(format!("Invalid operator"), self.lexer.current().unwrap())
|
|
}
|
|
}
|
|
_ => self.syntax_error(
|
|
format!("Expected an assignment"),
|
|
self.lexer.current().unwrap(),
|
|
),
|
|
}
|
|
}
|
|
|
|
/// ```notrust
|
|
/// TypeStatement
|
|
/// : "type" Expression ";"
|
|
/// ```
|
|
fn parse_type_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
self.assert_scope(Scope::Target)?;
|
|
self.lexer.expect_kind(token::Kind::TypeKeyword)?;
|
|
let expr = self.parse_expr(&[token::Kind::Semi])?;
|
|
Ok(tree::Node::TypeExpr(Box::new(expr)))
|
|
}
|
|
|
|
/// ```notrust
|
|
/// SourceStatement
|
|
/// : "source" Expression ";"
|
|
/// ```
|
|
fn parse_source_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
self.assert_scope(Scope::Target)?;
|
|
self.lexer.expect_kind(token::Kind::SourceKeyword)?;
|
|
self.scope.push(Scope::SourceList);
|
|
let source = self.parse_expr(&[token::Kind::Semi])?;
|
|
self.scope.pop();
|
|
Ok(tree::Node::SourceList(Box::new(source)))
|
|
}
|
|
|
|
/// ```notrust
|
|
/// ExpressionStatement
|
|
/// : Expression ";"
|
|
/// ```
|
|
fn parse_expr_stmt(&mut self) -> Result<tree::Node, Error> {
|
|
self.parse_expr(&[token::Kind::Semi])
|
|
}
|
|
|
|
/// ```notrust
|
|
/// Expression
|
|
/// : AssignmentExpression
|
|
/// | BinaryExpression
|
|
/// | UnaryExpression
|
|
/// | PrimaryExpression
|
|
/// ```
|
|
fn parse_expr(&mut self, terminators: &[token::Kind]) -> Result<tree::Node, Error> {
|
|
self.assert_scope(Scope::File)?;
|
|
let expr = if let Some(result) = self.lexer.peek() {
|
|
let token = result?;
|
|
if !token.kind.is_start_of_expr() {
|
|
self.syntax_error(String::from("Expected an expression"), &token)
|
|
} else {
|
|
self.parse_assignment_expr_or_higher(terminators)
|
|
}
|
|
} else {
|
|
self.syntax_error(
|
|
String::from("Unexpected EOF"),
|
|
&self.lexer.current().unwrap(),
|
|
)
|
|
};
|
|
expr
|
|
}
|
|
|
|
/// Parse an assignment expression.
|
|
/// This is no different to parsing any other binary expression, with the
|
|
/// difference being that assignment operators are right associative.
|
|
/// Therefore, we need a separate function for this. Other than that,
|
|
/// this method is no different to `parse_binary_expr_or_higher()` (it even
|
|
/// returns the same kind of tree node).
|
|
///
|
|
/// ```notrust
|
|
/// AssignmentExpression
|
|
/// : PrimaryExpression AssignmentOperator Expression
|
|
///
|
|
/// AssignmentOperator
|
|
/// : "=" | "+=" | "-=" | "*=" | "/=" | "%="
|
|
/// | "&=" | "|=" | "^=" | ">>=" | "<<="
|
|
/// ```
|
|
fn parse_assignment_expr_or_higher(
|
|
&mut self,
|
|
terminators: &[token::Kind],
|
|
) -> Result<tree::Node, Error> {
|
|
// we speculate on this being an assignment expression, so we need to
|
|
// be able to undo our work in case this speculation doesn't hold true
|
|
// so parse_binary_expr_or_higher() can do its thing
|
|
let bookmark = self.lexer.save();
|
|
|
|
let lhs = self.parse_primary_expr()?;
|
|
if let Some(Ok(token)) = self.lexer.peek() {
|
|
if token.kind.is_assignment_op() {
|
|
let op_token = self.lexer.require_next()?;
|
|
let op = Operator::from_token(&op_token)?;
|
|
let rhs = self.parse_binary_expr_or_higher(terminators)?;
|
|
return Ok(tree::Node::BinaryExpr {
|
|
op,
|
|
lhs: Box::new(lhs),
|
|
rhs: Box::new(rhs),
|
|
});
|
|
} else if token.kind.binary_op_precedence().is_some() {
|
|
// shoot, this wasn't an assignment, all of our work was useless
|
|
self.lexer.restore(bookmark);
|
|
return self.parse_binary_expr_or_higher(terminators);
|
|
} else {
|
|
self.lexer.expect_kinds(terminators)?;
|
|
}
|
|
}
|
|
Ok(lhs)
|
|
}
|
|
|
|
/// Binary expressions are generally left associative (except for assignments,
|
|
/// which are handled separately in `parse_assignment_expr_or_higher()`).
|
|
/// However, things get a little more tricky when taking the fact that there
|
|
/// are 9 different levels of precedence into account.
|
|
///
|
|
/// ```notrust
|
|
/// BinaryExpression
|
|
/// : Expression BinaryOperator Expression
|
|
///
|
|
/// BinaryOperator
|
|
/// : "||" | "&&" | "==" | "!=" | "<" | "<=" | ">" | ">="
|
|
/// : "|" | "^" | "&" | "<<" | ">>" | "+" | "-" | "*" | "/" | "%"
|
|
/// ```
|
|
fn parse_binary_expr_or_higher(
|
|
&mut self,
|
|
terminators: &[token::Kind],
|
|
) -> Result<tree::Node, Error> {
|
|
let mut expr = self.parse_unary_expr_or_higher()?;
|
|
|
|
while let Some(Ok(token)) = self.lexer.peek() {
|
|
if terminators.contains(&token.kind) {
|
|
self.lexer.next();
|
|
break;
|
|
}
|
|
|
|
let op = Operator::from_token(&token)?;
|
|
self.lexer.next();
|
|
let precedence = token.kind.binary_op_precedence().unwrap();
|
|
expr = tree::Node::BinaryExpr {
|
|
op,
|
|
lhs: Box::new(expr),
|
|
rhs: Box::new(self.parse_binary_rhs(precedence, terminators)?),
|
|
};
|
|
}
|
|
|
|
Ok(expr)
|
|
}
|
|
|
|
/// This is for parsing the right-hand side of a binary expression.
|
|
/// If the expression is followed by another operator with higher precedence, we need to
|
|
/// consume that entire subexpression and return it to the caller. This is best described
|
|
/// by the following two examples: The left one would be the result of `1 + 2 - 3`, and
|
|
/// the right one is `1 + 2 * 3` (note how the plus operator moves to the top of the tree
|
|
/// in the right example due to the multiplication operator's higher precedence).
|
|
///
|
|
/// ```notrust
|
|
/// - +
|
|
/// / \ / \
|
|
/// + 3 1 *
|
|
/// / \ / \
|
|
/// 1 2 2 3
|
|
/// ```
|
|
///
|
|
/// `parse_binary_expr_or_higher()` parses only left associatively through iteration.
|
|
/// It always calls this method to try and parse any chained binary expressions of higher
|
|
/// precedence. In the simplest case, this method will only read one unary expression
|
|
/// or higher and immediately return (if the following binary operator has equal or lower
|
|
/// precedence). In other cases, it invokes one recursion per increase in precedence.
|
|
fn parse_binary_rhs(
|
|
&mut self,
|
|
precedence: u32,
|
|
terminators: &[token::Kind],
|
|
) -> Result<tree::Node, Error> {
|
|
let mut lhs = self.parse_unary_expr_or_higher()?;
|
|
|
|
while let Some(Ok(token)) = self.lexer.peek() {
|
|
if let Some(new_precedence) = token.kind.binary_op_precedence() {
|
|
if new_precedence > precedence {
|
|
let op = Operator::from_token(&token)?;
|
|
self.lexer.next();
|
|
lhs = tree::Node::BinaryExpr {
|
|
op,
|
|
lhs: Box::new(lhs),
|
|
rhs: Box::new(self.parse_binary_rhs(new_precedence, terminators)?),
|
|
};
|
|
} else {
|
|
break;
|
|
}
|
|
} else {
|
|
break;
|
|
}
|
|
}
|
|
|
|
Ok(lhs)
|
|
}
|
|
|
|
/// ```notrust
|
|
/// UnaryExpression
|
|
/// : UnaryOperator Expression
|
|
///
|
|
/// UnaryOperator
|
|
/// : "!" | "-"
|
|
/// ```
|
|
fn parse_unary_expr_or_higher(&mut self) -> Result<tree::Node, Error> {
|
|
if let Some(result) = self.lexer.peek() {
|
|
let token = result?;
|
|
if token.kind == token::Kind::Bang || token.kind == token::Kind::Minus {
|
|
self.lexer.next(); // consume unary operator token
|
|
let op = Operator::from_token(&token)?;
|
|
let expr = self.parse_primary_expr()?;
|
|
return Ok(tree::Node::UnaryExpr {
|
|
op,
|
|
node: Box::new(expr),
|
|
});
|
|
}
|
|
}
|
|
self.parse_primary_expr()
|
|
}
|
|
|
|
/// ```notrust
|
|
/// PrimaryExpression
|
|
/// : "(" Expression ")"
|
|
/// | ArrayExpression
|
|
/// | CallExpression
|
|
/// | Identifier
|
|
/// | StringLiteral
|
|
/// | IntLiteral
|
|
/// | ArrayLiteral
|
|
/// | BoolLiteral
|
|
///
|
|
/// ArrayExpression
|
|
/// : PrimaryExpression "[" Expression "]"
|
|
///
|
|
/// CallExpression
|
|
/// : PrimaryExpression "(" [ ParameterList ] ")"
|
|
///
|
|
/// ParameterList
|
|
/// : Expression [ "," ]
|
|
/// | Expression "," ParameterList
|
|
/// ```
|
|
fn parse_primary_expr(&mut self) -> Result<tree::Node, Error> {
|
|
let token = self.lexer.require_next()?;
|
|
match token.kind {
|
|
token::Kind::OParen => {
|
|
let expr = self.parse_binary_expr_or_higher(&[token::Kind::CParen])?;
|
|
self.parse_primary_expr_rest(expr)
|
|
}
|
|
token::Kind::Ident => {
|
|
let ident = tree::Node::Ident(String::from(token.raw));
|
|
self.parse_primary_expr_rest(ident)
|
|
}
|
|
token::Kind::IntLiteral => {
|
|
let raw = token.raw;
|
|
let num = match raw.chars().nth(1) {
|
|
Some('x') => i128::from_str_radix(&raw[2..], 16),
|
|
Some('o') => i128::from_str_radix(&raw[2..], 8),
|
|
Some('b') => i128::from_str_radix(&raw[2..], 2),
|
|
_ => raw.parse(),
|
|
}
|
|
.unwrap();
|
|
Ok(tree::Node::Int(num))
|
|
}
|
|
token::Kind::StringLiteral => Ok(tree::Node::String(token.raw)),
|
|
token::Kind::TrueKeyword => Ok(tree::Node::Bool(true)),
|
|
token::Kind::FalseKeyword => Ok(tree::Node::Bool(false)),
|
|
token::Kind::OBracket => {
|
|
let elements =
|
|
self.parse_delimited_list(token::Kind::Comma, token::Kind::CBracket, true)?;
|
|
Ok(tree::Node::Array(elements))
|
|
}
|
|
_ => self.syntax_error(format!("Unexpected token {}", token.kind), &token),
|
|
}
|
|
}
|
|
|
|
/// Parse an optional appendix to a primary expression, i.e. an array access
|
|
/// or function call. This can also be chained, for example when dealing
|
|
/// with a matrix or a function returning another function like this:
|
|
///
|
|
/// ```notrust
|
|
/// matrix[y][x]
|
|
/// array_of_functions[index](params)
|
|
/// function_returning_an_array(params)[index]
|
|
/// ```
|
|
fn parse_primary_expr_rest(&mut self, start: tree::Node) -> Result<tree::Node, Error> {
|
|
if let Some(Ok(token)) = self.lexer.peek() {
|
|
match token.kind {
|
|
token::Kind::OParen => {
|
|
// function call
|
|
self.lexer.next();
|
|
let params =
|
|
self.parse_delimited_list(token::Kind::Comma, token::Kind::CParen, false)?;
|
|
self.parse_primary_expr_rest(tree::Node::CallExpr {
|
|
func: Box::new(start),
|
|
params,
|
|
})
|
|
}
|
|
token::Kind::OBracket => {
|
|
// array index
|
|
self.lexer.next();
|
|
let index = self.parse_expr(&[token::Kind::CBracket])?;
|
|
self.parse_primary_expr_rest(tree::Node::ArrayExpr {
|
|
array: Box::new(start),
|
|
index: Box::new(index),
|
|
})
|
|
}
|
|
_ => Ok(start),
|
|
}
|
|
} else {
|
|
Ok(start)
|
|
}
|
|
}
|
|
|
|
/// Parse a terminated, delimited list of expressions. This is used for
|
|
/// parameter lists in function calls and elements in array literals.
|
|
fn parse_delimited_list(
|
|
&mut self,
|
|
delimiter: token::Kind,
|
|
terminator: token::Kind,
|
|
allow_trailing_delimiter: bool,
|
|
) -> Result<Vec<tree::Node>, Error> {
|
|
let mut list = Vec::new();
|
|
|
|
// In the simplest case, we immediately see the terminator.
|
|
// That means we are already finished and return an empty list.
|
|
if let Some(Ok(token)) = self.lexer.peek() {
|
|
if token.kind == terminator {
|
|
self.lexer.next();
|
|
return Ok(list);
|
|
}
|
|
}
|
|
|
|
// now we know the list must contain at least one item
|
|
while self.lexer.peek().is_some() {
|
|
list.push(self.parse_expr(&[delimiter, terminator])?);
|
|
|
|
let current = self.lexer.current().unwrap();
|
|
if current.kind == terminator {
|
|
// this is the end of the list, we are finished
|
|
break;
|
|
} else if current.kind == delimiter {
|
|
// depending on whether trailing delimiters are allowed,
|
|
// this might still be the end of the list
|
|
if let Some(Ok(token)) = self.lexer.peek() {
|
|
if token.kind == terminator && allow_trailing_delimiter {
|
|
// so we saw a trailing delimiter followed by the
|
|
// terminator *and* trailing delimiters are allowed;
|
|
// this means we are finished here
|
|
self.lexer.next();
|
|
break;
|
|
}
|
|
}
|
|
} else {
|
|
// this should never happen since parse_expr() always returns
|
|
// with the current token kind being one of the ones specified
|
|
// (otherwise it would return a syntax error, in which case we
|
|
// wouldn't even reach this entire if block in the first place)
|
|
panic!("parse_expr() ended with an illegal token");
|
|
}
|
|
}
|
|
Ok(list)
|
|
}
|
|
|
|
/// Ensure that the `scope` stack contains a certain scope.
|
|
fn assert_scope(&self, scope: Scope) -> Result<(), Error> {
|
|
if self.scope.contains(&scope) {
|
|
Ok(())
|
|
} else {
|
|
let token = self.lexer.current().unwrap();
|
|
self.syntax_error(
|
|
format!("Token {} cannot be used in this context", token),
|
|
token,
|
|
)
|
|
}
|
|
}
|
|
|
|
/// Ensure that the `scope` stack does not contain a certain scope.
|
|
fn assert_scope_not(&self, scope: Scope) -> Result<(), Error> {
|
|
if self.scope.contains(&scope) {
|
|
let token = self.lexer.current().unwrap();
|
|
self.syntax_error(
|
|
format!("Token {} cannot be used in this context", token),
|
|
token,
|
|
)
|
|
} else {
|
|
Ok(())
|
|
}
|
|
}
|
|
|
|
fn syntax_error<T>(&self, msg: String, token: &Token) -> Result<T, Error> {
|
|
Err(Error::syntax_error(token.pos.clone(), msg))
|
|
}
|
|
}
|