16336579506

Committed 17 Jul 2025 04:57AM UTC coverage: 75.6% (+0.2%) from 75.352%

Build # 16336579506

Build Type

push

github

Committed by

vcfxb

Commit Message

chore: clippy

Run Details

0 of 1 new or added line in 1 file covered. (0.0%)

32 existing lines in 5 files now uncovered.

976 of 1291 relevant lines covered (75.6%)

35.0 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

81.63

/wright/src/parser.rs

//! This parser module is responsible for turning the stream of [Token]s from the [Lexer] into a tree of [AST] nodes.
//!
//! [AST]: crate::ast
//! [Token]: crate::lexer::token::Token

use error::{ParserError, ParserErrorKind};

use super::lexer::Lexer;
use crate::{
    lexer::token::{Token, TokenTy},
    source_tracking::fragment::Fragment,
};
use std::collections::VecDeque;

mod decl;
pub mod error;
mod identifier;
mod literal;
mod path;
mod ty;

/// The [Parser] struct wraps a [Lexer] and adds lookahead and functions that are useful for parsing.
#[derive(Debug)]
pub struct Parser {
    lexer: Lexer,
    lookahead: VecDeque<Token>,
}

impl Parser {
    /// Construct a new parser around a given [Lexer].
    pub fn new(lexer: Lexer) -> Self {
        Parser {
            lexer,
            lookahead: VecDeque::new(),
        }
    }

    /// Get the [Lexer] that's wrapped.
    pub fn lexer(&self) -> &Lexer {
        &self.lexer
    }

    /// Lookahead `k` [Token]s.
    ///
    /// If `k == 0` then this is effectively peeking at the next [Token] from the wrapped [Lexer].
    pub fn lookahead(&mut self, k: usize) -> Option<&Token> {
        while self.lookahead.len() <= k {
            self.lookahead.push_back(self.lexer.next_token()?);
        }

        // SAFETY: It's certain that if this function reaches this line, this access is infallible.
        Some(unsafe { self.lookahead.get(k).unwrap_unchecked() })
    }

    /// Peek at the next token from the [Lexer] (cached in the lookahead queue if peeked before).
    pub fn peek(&mut self) -> Option<&Token> {
        self.lookahead(0)
    }

    /// Peek the [Fragment] of the next [Token].
    pub fn peek_fragment(&mut self) -> Option<&Fragment> {
        self.peek().map(|token| &token.fragment)
    }

    /// Peek the [TokenTy] of the next [Token].
    pub fn peek_variant(&mut self) -> Option<TokenTy> {
        self.peek().map(|token| token.variant)
    }

    /// Similar to [Parser::lookahead] but instead returns a slice of `n` [Token]s, starting with the next [Token].
    ///
    /// Returns [None] if `n` is greater than the number of remaining [Token]s for this [Parser].
    pub fn lookahead_window(&mut self, n: usize) -> Option<&[Token]> {
        // Pull tokens from lexer
        while self.lookahead.len() < n {
            self.lookahead.push_back(self.lexer.next_token()?);
        }

        // Use make contiguous here to get a unified/single slice.
        Some(&self.lookahead.make_contiguous()[..n])
    }

    /// Peek the next token that's not whitespace.
    pub fn peek_next_not_whitespace(&mut self) -> Option<&Token> {
        // There's no way to do this in safe rust, despite the memory accesses being fine,
        // so we do it unsafely here

        for i in 0.. {
            let peek = self.lookahead(i)?;

            if peek.variant != TokenTy::Whitespace {
                // This bit prevents the rust compiler from thinking we're breaking
                // lifetime/aliasing rules by mutating the internal state in the next
                // iteration of the loop while still holding a reference to the peeked token.
                unsafe {
                    let const_ref = &raw const *peek;
                    let upcast = &*const_ref;
                    return Some(upcast);
                }
            }
        }

        // Safety: For large enough values of `i`, self.lookahead eventually has to return `None`
        unsafe { std::hint::unreachable_unchecked() }
    }

    /// Get the number of remaining bytes on this parser. This is potentially useful for checking
    /// if a parser has advanced between two calls (or checking if a parser has reached end of input).
    pub fn bytes_remaining(&self) -> usize {
        let bytes_remaining_in_lookahead_buffer = self
            .lookahead
            .iter()
            .map(|t| t.fragment.len())
            .sum::<usize>();

        let bytes_remaining_in_lexer = self.lexer.bytes_remaining();

        bytes_remaining_in_lexer + bytes_remaining_in_lookahead_buffer
    }

    /// Get the next [Token] from this [Parser]. This may be a token that's already been peeked.
    ///
    /// Skips any non-document comments encountered via the lexer implementation.
    ///
    /// Return an error if a [Token] with [TokenTy::Unknown] is encountered.
    pub fn next_token(&mut self) -> Result<Option<Token>, ParserError> {
        let token = self
            .lookahead
            .pop_front()
            .or_else(|| self.lexer.next_token());

        // Check for unknown tokens, which should always convert to an error.
        if let Some(ref t) = token
            && t.variant == TokenTy::Unknown
        {
            // Clone here is avoidable but this code path is super unlikely and
            // probably optimized heavily.
            Err(ParserErrorKind::EncounteredUnknownToken.at(t.fragment.clone()))
        } else {
            Ok(token)
        }
    }

    /// Advance this [Parser] by `n` [Token]s. If this [Parser] runs out of [Token]s, panic.
    ///
    /// Panics
    /// - If `n` is greater than the number of remaining tokens.
    pub fn advance(&mut self, n: usize) {
        // Add tokens to the lookahead buffer until we have enough to split off.
        while self.lookahead.len() < n {
            let token = self
                .lexer
                .next_token()
                .expect("advance: `n` <= number of remaining tokens");

            self.lookahead.push_back(token);
        }

        // Split them off.
        self.lookahead = self.lookahead.split_off(n);
    }

    /// Peek the [Fragment] of the next [Token] and clone it or return a clone of the
    /// remainder [Fragment] of the internal [Lexer]
    /// (which will be empty, since there wasn't a [Token] to peek).
    ///
    /// This is likely only useful for error reporting -- a clone of a potentially empty fragment is
    /// rarely ever useful otherwise.
    pub fn peek_fragment_or_rest_cloned(&mut self) -> Fragment {
        match self.peek() {
            Some(Token { fragment, .. }) => fragment.clone(),
            None => {
                let rest = self.lexer.remaining.clone();

                // Assert that we're making the right assumptions about the remaining fragment.
                // These are (unidiomatically) done using debug_assert -- perhaps that changes eventually
                // however it should be fine for now, since this can only produce logic bugs (never memory or
                // concurrency bugs).
                debug_assert!(rest.is_valid());
                debug_assert!(rest.is_empty());
                debug_assert!(rest.is_empty_at_end_of_source());

                rest
            }
        }
    }

    /// Get the next [Token] from this parser if its [Token::variant] is the given `token_ty`.
    pub fn next_if_is(&mut self, token_ty: TokenTy) -> Option<Token> {
        // Peeking successfully first means that the lookahead vec will never be empty here.
        (self.peek_variant()? == token_ty)
            // SAFETY: We just peeked a token to check its variant so this unwrap is always ok.
            .then(|| unsafe { self.lookahead.pop_front().unwrap_unchecked() })
    }

    /// Peek at the next [Token]s of this [Parser] and determine if the [Token::variant]s match this
    /// sequence of [TokenTy]s.
    pub fn matches(&mut self, seq: &[TokenTy]) -> bool {
        // Use the rare let-else to ensure there are at minimum, the given number of tokens remaining.
        let Some(lookahead_window) = self.lookahead_window(seq.len()) else {
            return false;
        };

        // Use a zipped iterator to compare all the token variants.
        lookahead_window
            .iter()
            .zip(seq)
            .all(|(token, matches)| token.variant == *matches)
    }

    /// Consume & remove all whitespace tokens from the front of the parser.
    pub fn consume_optional_whitespace(&mut self) {
        // Iterate until the next token is not whitespace.
        while self.next_if_is(TokenTy::Whitespace).is_some() {}
    }

    /// Require a whitespace from the [Parser]. Do not advance if the next [Token] is not a whitespace.
    pub fn consume_at_least_one_whitespace(&mut self) -> Result<(), ParserError> {
        if self.next_if_is(TokenTy::Whitespace).is_some() {
            self.consume_optional_whitespace();
            Ok(())
        } else {
            Err(ParserErrorKind::ExpectedWhitespace.at(self.peek_fragment_or_rest_cloned()))
        }
    }
}

1	//! This parser module is responsible for turning the stream of [Token]s from the [Lexer] into a tree of [AST] nodes.
2	//!
3	//! [AST]: crate::ast
4	//! [Token]: crate::lexer::token::Token
5
6	use error::{ParserError, ParserErrorKind};
7
8	use super::lexer::Lexer;
9	use crate::{
10	lexer::token::{Token, TokenTy},
11	source_tracking::fragment::Fragment,
12	};
13	use std::collections::VecDeque;
14
15	mod decl;
16	pub mod error;
17	mod identifier;
18	mod literal;
19	mod path;
20	mod ty;
21
22	/// The [Parser] struct wraps a [Lexer] and adds lookahead and functions that are useful for parsing.
23	#[derive(Debug)]
24	pub struct Parser {
25	lexer: Lexer,
26	lookahead: VecDeque<Token>,
27	}
28
29	impl Parser {
30	/// Construct a new parser around a given [Lexer].
31	pub fn new(lexer: Lexer) -> Self {	28✔
32	Parser {	28✔
33	lexer,	28✔
34	lookahead: VecDeque::new(),	28✔
35	}	28✔
36	}	28✔
37
38	/// Get the [Lexer] that's wrapped.
39	pub fn lexer(&self) -> &Lexer {	1✔
40	&self.lexer	1✔
41	}	1✔
42
43	/// Lookahead `k` [Token]s.
44	///
45	/// If `k == 0` then this is effectively peeking at the next [Token] from the wrapped [Lexer].
46	pub fn lookahead(&mut self, k: usize) -> Option<&Token> {	124✔
47	while self.lookahead.len() <= k {	172✔
48	self.lookahead.push_back(self.lexer.next_token()?);	56✔
49	}
50
51	// SAFETY: It's certain that if this function reaches this line, this access is infallible.
52	Some(unsafe { self.lookahead.get(k).unwrap_unchecked() })	116✔
53	}	124✔
54
55	/// Peek at the next token from the [Lexer] (cached in the lookahead queue if peeked before).
56	pub fn peek(&mut self) -> Option<&Token> {	118✔
57	self.lookahead(0)	118✔
58	}	118✔
59
60	/// Peek the [Fragment] of the next [Token].
UNCOV 61	pub fn peek_fragment(&mut self) -> Option<&Fragment> {	×
UNCOV 62	self.peek().map(\|token\| &token.fragment)	×
UNCOV 63	}	×
64
65	/// Peek the [TokenTy] of the next [Token].
66	pub fn peek_variant(&mut self) -> Option<TokenTy> {	101✔
67	self.peek().map(\|token\| token.variant)	101✔
68	}	101✔
69
70	/// Similar to [Parser::lookahead] but instead returns a slice of `n` [Token]s, starting with the next [Token].
71	///
72	/// Returns [None] if `n` is greater than the number of remaining [Token]s for this [Parser].
73	pub fn lookahead_window(&mut self, n: usize) -> Option<&[Token]> {	63✔
74	// Pull tokens from lexer
75	while self.lookahead.len() < n {	102✔
76	self.lookahead.push_back(self.lexer.next_token()?);	68✔
77	}
78
79	// Use make contiguous here to get a unified/single slice.
80	Some(&self.lookahead.make_contiguous()[..n])	34✔
81	}	63✔
82
83	/// Peek the next token that's not whitespace.
84	pub fn peek_next_not_whitespace(&mut self) -> Option<&Token> {	6✔
85	// There's no way to do this in safe rust, despite the memory accesses being fine,
86	// so we do it unsafely here
87
88	for i in 0.. {	6✔
89	let peek = self.lookahead(i)?;	6✔
90
UNCOV 91	if peek.variant != TokenTy::Whitespace {	×
92	// This bit prevents the rust compiler from thinking we're breaking
93	// lifetime/aliasing rules by mutating the internal state in the next
94	// iteration of the loop while still holding a reference to the peeked token.
95	unsafe {
UNCOV 96	let const_ref = &raw const *peek;	×
UNCOV 97	let upcast = &*const_ref;	×
UNCOV 98	return Some(upcast);	×
99	}
UNCOV 100	}	×
101	}
102
103	// Safety: For large enough values of `i`, self.lookahead eventually has to return `None`
104	unsafe { std::hint::unreachable_unchecked() }	×
105	}	6✔
106
107	/// Get the number of remaining bytes on this parser. This is potentially useful for checking
108	/// if a parser has advanced between two calls (or checking if a parser has reached end of input).
109	pub fn bytes_remaining(&self) -> usize {	10✔
110	let bytes_remaining_in_lookahead_buffer = self	10✔
111	.lookahead	10✔
112	.iter()	10✔
113	.map(\|t\| t.fragment.len())	10✔
114	.sum::<usize>();	10✔
115
116	let bytes_remaining_in_lexer = self.lexer.bytes_remaining();	10✔
117
118	bytes_remaining_in_lexer + bytes_remaining_in_lookahead_buffer	10✔
119	}	10✔
120
121	/// Get the next [Token] from this [Parser]. This may be a token that's already been peeked.
122	///
123	/// Skips any non-document comments encountered via the lexer implementation.
124	///
125	/// Return an error if a [Token] with [TokenTy::Unknown] is encountered.
126	pub fn next_token(&mut self) -> Result<Option<Token>, ParserError> {	5✔
127	let token = self	5✔
128	.lookahead	5✔
129	.pop_front()	5✔
130	.or_else(\|\| self.lexer.next_token());	5✔
131
132	// Check for unknown tokens, which should always convert to an error.
133	if let Some(ref t) = token	5✔
134	&& t.variant == TokenTy::Unknown	5✔
135	{
136	// Clone here is avoidable but this code path is super unlikely and
137	// probably optimized heavily.
UNCOV 138	Err(ParserErrorKind::EncounteredUnknownToken.at(t.fragment.clone()))	×
139	} else {
140	Ok(token)	5✔
141	}
142	}	5✔
143
144	/// Advance this [Parser] by `n` [Token]s. If this [Parser] runs out of [Token]s, panic.
145	///
146	/// Panics
147	/// - If `n` is greater than the number of remaining tokens.
148	pub fn advance(&mut self, n: usize) {	9✔
149	// Add tokens to the lookahead buffer until we have enough to split off.
150	while self.lookahead.len() < n {	9✔
151	let token = self	×
UNCOV 152	.lexer	×
UNCOV 153	.next_token()	×
UNCOV 154	.expect("advance: `n` <= number of remaining tokens");	×
UNCOV 155		×
UNCOV 156	self.lookahead.push_back(token);	×
UNCOV 157	}	×
158
159	// Split them off.
160	self.lookahead = self.lookahead.split_off(n);	9✔
161	}	9✔
162
163	/// Peek the [Fragment] of the next [Token] and clone it or return a clone of the
164	/// remainder [Fragment] of the internal [Lexer]
165	/// (which will be empty, since there wasn't a [Token] to peek).
166	///
167	/// This is likely only useful for error reporting -- a clone of a potentially empty fragment is
168	/// rarely ever useful otherwise.
169	pub fn peek_fragment_or_rest_cloned(&mut self) -> Fragment {	17✔
170	match self.peek() {	17✔
171	Some(Token { fragment, .. }) => fragment.clone(),	16✔
172	None => {
173	let rest = self.lexer.remaining.clone();	1✔
174
175	// Assert that we're making the right assumptions about the remaining fragment.
176	// These are (unidiomatically) done using debug_assert -- perhaps that changes eventually
177	// however it should be fine for now, since this can only produce logic bugs (never memory or
178	// concurrency bugs).
179	debug_assert!(rest.is_valid());	1✔
180	debug_assert!(rest.is_empty());	1✔
181	debug_assert!(rest.is_empty_at_end_of_source());	1✔
182
183	rest	1✔
184	}
185	}
186	}	17✔
187
188	/// Get the next [Token] from this parser if its [Token::variant] is the given `token_ty`.
189	pub fn next_if_is(&mut self, token_ty: TokenTy) -> Option<Token> {	94✔
190	// Peeking successfully first means that the lookahead vec will never be empty here.
191	(self.peek_variant()? == token_ty)	94✔
192	// SAFETY: We just peeked a token to check its variant so this unwrap is always ok.
193	.then(\|\| unsafe { self.lookahead.pop_front().unwrap_unchecked() })	93✔
194	}	94✔
195
196	/// Peek at the next [Token]s of this [Parser] and determine if the [Token::variant]s match this
197	/// sequence of [TokenTy]s.
198	pub fn matches(&mut self, seq: &[TokenTy]) -> bool {	63✔
199	// Use the rare let-else to ensure there are at minimum, the given number of tokens remaining.
200	let Some(lookahead_window) = self.lookahead_window(seq.len()) else {	63✔
201	return false;	29✔
202	};
203
204	// Use a zipped iterator to compare all the token variants.
205	lookahead_window	34✔
206	.iter()	34✔
207	.zip(seq)	34✔
208	.all(\|(token, matches)\| token.variant == *matches)	59✔
209	}	63✔
210
211	/// Consume & remove all whitespace tokens from the front of the parser.
212	pub fn consume_optional_whitespace(&mut self) {	23✔
213	// Iterate until the next token is not whitespace.
214	while self.next_if_is(TokenTy::Whitespace).is_some() {}	25✔
215	}	23✔
216
217	/// Require a whitespace from the [Parser]. Do not advance if the next [Token] is not a whitespace.
218	pub fn consume_at_least_one_whitespace(&mut self) -> Result<(), ParserError> {	8✔
219	if self.next_if_is(TokenTy::Whitespace).is_some() {	8✔
220	self.consume_optional_whitespace();	8✔
221	Ok(())	8✔
222	} else {
UNCOV 223	Err(ParserErrorKind::ExpectedWhitespace.at(self.peek_fragment_or_rest_cloned()))	×
224	}
225	}	8✔
226	}

vcfxb / wright-lang / 16336579506

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous