overhaul operator precedence
arithmetic operators now have the same precedence. if you want them to bind tighter, you remove the spaces around them: - 2 + 2 * 2 = 8 - 2 + 2*2 = 6
This commit is contained in:
parent
09f2292e62
commit
b52c1b26c9
4 changed files with 144 additions and 43 deletions
|
@ -7,7 +7,7 @@ use crate::{
|
|||
ast::{Ast, NodeAllocError, NodeId, NodeKind},
|
||||
diagnostic::Diagnostic,
|
||||
source::Span,
|
||||
token::{Lexis, TokenKind, TokenKindSet},
|
||||
token::{Lexis, Spaces, TokenKind, TokenKindSet},
|
||||
};
|
||||
|
||||
#[derive(Debug, Clone, Copy)]
|
||||
|
@ -132,6 +132,11 @@ impl<'a> Parser<'a> {
|
|||
self.tokens.kind(self.position)
|
||||
}
|
||||
|
||||
#[track_caller]
|
||||
fn peek_with_spaces(&self) -> (TokenKind, Spaces) {
|
||||
(self.peek(), self.tokens.spaces(self.position))
|
||||
}
|
||||
|
||||
fn span(&self) -> Span {
|
||||
self.tokens.span(self.position)
|
||||
}
|
||||
|
@ -298,33 +303,61 @@ impl fmt::Display for IntoAstError {
|
|||
|
||||
impl Error for IntoAstError {}
|
||||
|
||||
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
|
||||
enum Tighter {
|
||||
Left,
|
||||
Right,
|
||||
}
|
||||
|
||||
fn tighter(left: TokenKind, right: TokenKind) -> Tighter {
|
||||
fn tightness(kind: TokenKind) -> Option<usize> {
|
||||
match kind {
|
||||
TokenKind::Equal | TokenKind::Colon => Some(0),
|
||||
fn tighter(left: (TokenKind, Spaces), right: (TokenKind, Spaces)) -> Tighter {
|
||||
#[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
|
||||
enum Spacing {
|
||||
Loose,
|
||||
Call,
|
||||
Tight,
|
||||
}
|
||||
|
||||
fn tightness((kind, spaces): (TokenKind, Spaces)) -> Option<(Spacing, usize)> {
|
||||
let spacing = match kind {
|
||||
// There are a few types of operators which are independent of tightness.
|
||||
|
||||
// For : and =, it does not matter if they're spelled one way or the other, because
|
||||
// there is only one way to use them (at the beginning of the expression).
|
||||
TokenKind::Colon | TokenKind::Equal => Spacing::Loose,
|
||||
|
||||
// For calls, there is a special intermediate level, such that they can sit between
|
||||
// loose operators and tight operators.
|
||||
_ if PREFIX_TOKENS.contains(kind) => Spacing::Call,
|
||||
|
||||
// For everything else, the usual rules apply.
|
||||
_ => match spaces.pair() {
|
||||
(false, false) => Spacing::Tight,
|
||||
(true, true) => Spacing::Loose,
|
||||
_ => return None, // not a valid infix operator
|
||||
},
|
||||
};
|
||||
let index = match kind {
|
||||
TokenKind::Equal | TokenKind::Colon => 0,
|
||||
// 1: reserved for `and` and `or`
|
||||
TokenKind::EqualEqual
|
||||
| TokenKind::NotEqual
|
||||
| TokenKind::Less
|
||||
| TokenKind::LessEqual
|
||||
| TokenKind::Greater
|
||||
| TokenKind::GreaterEqual => Some(1),
|
||||
TokenKind::Plus | TokenKind::Minus => Some(2),
|
||||
TokenKind::Star | TokenKind::Slash => Some(3),
|
||||
_ if PREFIX_TOKENS.contains(kind) => Some(4),
|
||||
_ => None,
|
||||
}
|
||||
| TokenKind::GreaterEqual => 2,
|
||||
TokenKind::Plus | TokenKind::Minus | TokenKind::Star | TokenKind::Slash => 3,
|
||||
// 4: reserve for `.`
|
||||
_ if PREFIX_TOKENS.contains(kind) => 5,
|
||||
_ => return None, // not an infix operator
|
||||
};
|
||||
Some((spacing, index))
|
||||
}
|
||||
|
||||
let Some(right_tightness) = tightness(right) else {
|
||||
return Tighter::Left;
|
||||
};
|
||||
let Some(left_tightness) = tightness(left) else {
|
||||
assert!(left == TokenKind::Eof);
|
||||
assert!(left.0 == TokenKind::Eof);
|
||||
return Tighter::Right;
|
||||
};
|
||||
|
||||
|
@ -335,12 +368,13 @@ fn tighter(left: TokenKind, right: TokenKind) -> Tighter {
|
|||
}
|
||||
}
|
||||
|
||||
fn precedence_parse(p: &mut Parser, left: TokenKind) {
|
||||
fn precedence_parse(p: &mut Parser, left: (TokenKind, Spaces)) {
|
||||
let mut lhs = prefix(p);
|
||||
|
||||
loop {
|
||||
let right = p.peek();
|
||||
match tighter(left, right) {
|
||||
let right = p.peek_with_spaces();
|
||||
let tighter = tighter(left, right);
|
||||
match tighter {
|
||||
Tighter::Left => break,
|
||||
Tighter::Right => {
|
||||
let o = p.open_before(lhs);
|
||||
|
@ -536,14 +570,12 @@ fn if_expr(p: &mut Parser) -> Closed {
|
|||
p.close(o, NodeKind::If)
|
||||
}
|
||||
|
||||
// NOTE: This must be synchronised with the match expression in prefix().
|
||||
const PREFIX_TOKENS: TokenKindSet = TokenKindSet::new(&[
|
||||
TokenKind::Ident,
|
||||
TokenKind::Tag,
|
||||
TokenKind::Number,
|
||||
TokenKind::Color,
|
||||
// NOTE: This is ambiguous in function calls.
|
||||
// In that case, the infix operator takes precedence (because the `match` arms for the infix op
|
||||
// come first.)
|
||||
TokenKind::Minus,
|
||||
TokenKind::Not,
|
||||
TokenKind::LParen,
|
||||
|
@ -582,8 +614,8 @@ fn prefix(p: &mut Parser) -> Closed {
|
|||
}
|
||||
}
|
||||
|
||||
fn infix(p: &mut Parser, op: TokenKind) -> NodeKind {
|
||||
match op {
|
||||
fn infix(p: &mut Parser, op: (TokenKind, Spaces)) -> NodeKind {
|
||||
match op.0 {
|
||||
TokenKind::Plus
|
||||
| TokenKind::Minus
|
||||
| TokenKind::Star
|
||||
|
@ -598,13 +630,13 @@ fn infix(p: &mut Parser, op: TokenKind) -> NodeKind {
|
|||
|
||||
TokenKind::Equal => infix_let(p, op),
|
||||
|
||||
_ if PREFIX_TOKENS.contains(op) => infix_call(p),
|
||||
_ if PREFIX_TOKENS.contains(op.0) => infix_call(p, op),
|
||||
|
||||
_ => panic!("unhandled infix operator {op:?}"),
|
||||
}
|
||||
}
|
||||
|
||||
fn infix_binary(p: &mut Parser, op: TokenKind) -> NodeKind {
|
||||
fn infix_binary(p: &mut Parser, op: (TokenKind, Spaces)) -> NodeKind {
|
||||
let o = p.open();
|
||||
p.advance();
|
||||
p.close(o, NodeKind::Op);
|
||||
|
@ -617,15 +649,16 @@ fn infix_binary(p: &mut Parser, op: TokenKind) -> NodeKind {
|
|||
NodeKind::Binary
|
||||
}
|
||||
|
||||
fn infix_call(p: &mut Parser) -> NodeKind {
|
||||
fn infix_call(p: &mut Parser, mut arg: (TokenKind, Spaces)) -> NodeKind {
|
||||
while PREFIX_TOKENS.contains(p.peek()) {
|
||||
prefix(p);
|
||||
precedence_parse(p, arg);
|
||||
arg = p.peek_with_spaces();
|
||||
}
|
||||
|
||||
NodeKind::Call
|
||||
}
|
||||
|
||||
fn infix_let(p: &mut Parser, op: TokenKind) -> NodeKind {
|
||||
fn infix_let(p: &mut Parser, op: (TokenKind, Spaces)) -> NodeKind {
|
||||
p.advance();
|
||||
|
||||
if p.peek() == TokenKind::Newline {
|
||||
|
@ -651,7 +684,7 @@ fn infix_let(p: &mut Parser, op: TokenKind) -> NodeKind {
|
|||
}
|
||||
|
||||
pub fn expr(p: &mut Parser) {
|
||||
precedence_parse(p, TokenKind::Eof)
|
||||
precedence_parse(p, (TokenKind::Eof, Spaces::new(true, false)))
|
||||
}
|
||||
|
||||
pub fn toplevel(p: &mut Parser) {
|
||||
|
|
Loading…
Add table
Add a link
Reference in a new issue