rustc_lexer/cursor.rs
1use std::str::Chars;
2
3/// Peekable iterator over a char sequence.
4///
5/// Next characters can be peeked via `first` method,
6/// and position can be shifted forward via `bump` method.
7pub struct Cursor<'a> {
8 len_remaining: usize,
9 /// Iterator over chars. Slightly faster than a &str.
10 chars: Chars<'a>,
11 #[cfg(debug_assertions)]
12 prev: char,
13}
14
15pub(crate) const EOF_CHAR: char = '\0';
16
17impl<'a> Cursor<'a> {
18 pub fn new(input: &'a str) -> Cursor<'a> {
19 Cursor {
20 len_remaining: input.len(),
21 chars: input.chars(),
22 #[cfg(debug_assertions)]
23 prev: EOF_CHAR,
24 }
25 }
26
27 pub fn as_str(&self) -> &'a str {
28 self.chars.as_str()
29 }
30
31 /// Returns the last eaten symbol (or `'\0'` in release builds).
32 /// (For debug assertions only.)
33 pub(crate) fn prev(&self) -> char {
34 #[cfg(debug_assertions)]
35 {
36 self.prev
37 }
38
39 #[cfg(not(debug_assertions))]
40 {
41 EOF_CHAR
42 }
43 }
44
45 /// Peeks the next symbol from the input stream without consuming it.
46 /// If requested position doesn't exist, `EOF_CHAR` is returned.
47 /// However, getting `EOF_CHAR` doesn't always mean actual end of file,
48 /// it should be checked with `is_eof` method.
49 pub fn first(&self) -> char {
50 // `.next()` optimizes better than `.nth(0)`
51 self.chars.clone().next().unwrap_or(EOF_CHAR)
52 }
53
54 /// Peeks the second symbol from the input stream without consuming it.
55 pub(crate) fn second(&self) -> char {
56 // `.next()` optimizes better than `.nth(1)`
57 let mut iter = self.chars.clone();
58 iter.next();
59 iter.next().unwrap_or(EOF_CHAR)
60 }
61
62 /// Peeks the third symbol from the input stream without consuming it.
63 pub fn third(&self) -> char {
64 // `.next()` optimizes better than `.nth(1)`
65 let mut iter = self.chars.clone();
66 iter.next();
67 iter.next();
68 iter.next().unwrap_or(EOF_CHAR)
69 }
70
71 /// Checks if there is nothing more to consume.
72 pub(crate) fn is_eof(&self) -> bool {
73 self.chars.as_str().is_empty()
74 }
75
76 /// Returns amount of already consumed symbols.
77 pub(crate) fn pos_within_token(&self) -> u32 {
78 (self.len_remaining - self.chars.as_str().len()) as u32
79 }
80
81 /// Resets the number of bytes consumed to 0.
82 pub(crate) fn reset_pos_within_token(&mut self) {
83 self.len_remaining = self.chars.as_str().len();
84 }
85
86 /// Moves to the next character.
87 pub(crate) fn bump(&mut self) -> Option<char> {
88 let c = self.chars.next()?;
89
90 #[cfg(debug_assertions)]
91 {
92 self.prev = c;
93 }
94
95 Some(c)
96 }
97
98 /// Eats symbols while predicate returns true or until the end of file is reached.
99 pub(crate) fn eat_while(&mut self, mut predicate: impl FnMut(char) -> bool) {
100 // It was tried making optimized version of this for eg. line comments, but
101 // LLVM can inline all of this and compile it down to fast iteration over bytes.
102 while predicate(self.first()) && !self.is_eof() {
103 self.bump();
104 }
105 }
106
107 pub(crate) fn eat_until(&mut self, byte: u8) {
108 self.chars = match memchr::memchr(byte, self.as_str().as_bytes()) {
109 Some(index) => self.as_str()[index..].chars(),
110 None => "".chars(),
111 }
112 }
113}