1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
//! The character scanner: turns pattern text into [`Token`]s.
use super::token::{Token, TokenKind};
use crate::error::{Result, Span};
/// The lexer for regex patterns.
pub struct Lexer<'a> {
/// The source pattern.
pub(super) src: &'a str,
/// Iterator over characters.
chars: std::iter::Peekable<std::str::CharIndices<'a>>,
/// Current position in bytes.
pub(super) pos: usize,
/// Nesting depth of character classes (0 = outside any class). Tracked as a
/// depth (not a bool) so nested classes like `[a[b]c]` lex correctly.
class_depth: usize,
/// Extended mode (`x`): unescaped ASCII whitespace and `#` comments outside
/// a character class are not part of the pattern.
extended: bool,
/// Inside a `\Q…\E` span, where every character is a literal.
quoted: bool,
}
impl<'a> Lexer<'a> {
/// Creates a new lexer.
pub fn new(src: &'a str) -> Self {
Self {
src,
chars: src.char_indices().peekable(),
pos: 0,
class_depth: 0,
extended: false,
quoted: false,
}
}
/// Returns the source pattern.
pub fn source(&self) -> &str {
self.src
}
/// Peeks at the next character without consuming it.
pub(super) fn peek_char(&mut self) -> Option<char> {
self.chars.peek().map(|(_, c)| *c)
}
/// Consumes the next character.
pub(super) fn next_char(&mut self) -> Option<(usize, char)> {
let result = self.chars.next();
if let Some((pos, c)) = result {
self.pos = pos + c.len_utf8();
}
result
}
/// Enter (`true`) or exit (`false`) a character class, maintaining the
/// nesting depth so nested classes lex in class mode until fully closed.
pub fn set_in_class(&mut self, in_class: bool) {
if in_class {
self.class_depth += 1;
} else {
self.class_depth = self.class_depth.saturating_sub(1);
}
}
/// Turns extended (`x`) mode on or off for subsequent tokens.
///
/// The parser drives this as it enters and leaves `(?x)` / `(?x:…)` scopes;
/// the lexer needs it because whitespace and comments are removed before a
/// token is ever produced.
pub fn set_extended(&mut self, extended: bool) {
self.extended = extended;
}
/// Returns the next token.
pub fn next_token(&mut self) -> Result<Token> {
loop {
// Inside `\Q…\E` every character stands for itself: metacharacters,
// whitespace under extended mode, and `]` inside a class alike. Only
// `\E` ends the span, and an unterminated `\Q` quotes to the end of
// the pattern.
if self.quoted {
return match self.next_char() {
None => Ok(Token {
kind: TokenKind::Eof,
span: Span::point(self.pos),
}),
Some((_, '\\')) if self.peek_char() == Some('E') => {
self.next_char();
self.quoted = false;
continue;
}
Some((start, c)) => Ok(Token {
kind: TokenKind::Literal(c),
span: Span::new(start, self.pos),
}),
};
}
if self.extended && self.class_depth == 0 {
self.skip_extended_trivia();
}
let (start, c) = match self.next_char() {
Some(pair) => pair,
None => {
return Ok(Token {
kind: TokenKind::Eof,
span: Span::point(self.pos),
});
}
};
// `\Q` opens a quoted span. A stray `\E` is a no-op, as in PCRE, so
// that `\Q…\E` can be spliced in unconditionally. `\\Q` is a literal
// backslash followed by `Q` and must not reach here — it doesn't,
// because the peek only fires when `\` is the escape's own opener.
if c == '\\' {
match self.peek_char() {
Some('Q') => {
self.next_char();
self.quoted = true;
continue;
}
Some('E') => {
self.next_char();
continue;
}
_ => {}
}
}
let kind = if self.class_depth > 0 {
self.lex_class_char(c, start)?
} else {
self.lex_char(c, start)?
};
return Ok(Token {
kind,
span: Span::new(start, self.pos),
});
}
}
/// Consumes whitespace and `#` comments that extended mode strips from the
/// pattern. Only reached outside a character class, where whitespace stays
/// literal in every engine that implements this mode.
fn skip_extended_trivia(&mut self) {
loop {
match self.peek_char() {
Some(c) if c.is_ascii_whitespace() => {
self.next_char();
}
Some('#') => {
while let Some((_, c)) = self.next_char() {
if c == '\n' {
break;
}
}
}
_ => return,
}
}
}
/// Lexes a character outside a character class.
fn lex_char(&mut self, c: char, start: usize) -> Result<TokenKind> {
match c {
'.' => Ok(TokenKind::Dot),
'*' => Ok(TokenKind::Star),
'+' => Ok(TokenKind::Plus),
'?' => Ok(TokenKind::Question),
'|' => Ok(TokenKind::Pipe),
'^' => Ok(TokenKind::Caret),
'$' => Ok(TokenKind::Dollar),
'(' => Ok(TokenKind::OpenParen),
')' => Ok(TokenKind::CloseParen),
'[' => Ok(TokenKind::OpenBracket),
']' => Ok(TokenKind::CloseBracket),
'{' => Ok(TokenKind::OpenBrace),
'}' => Ok(TokenKind::CloseBrace),
':' => Ok(TokenKind::Colon),
'<' => Ok(TokenKind::LessThan),
'>' => Ok(TokenKind::GreaterThan),
'=' => Ok(TokenKind::Equals),
'!' => Ok(TokenKind::Exclamation),
',' => Ok(TokenKind::Comma),
'\\' => self.lex_escape(start),
c if c.is_ascii_digit() => Ok(TokenKind::Digit(c.to_digit(10).unwrap())),
c => Ok(TokenKind::Literal(c)),
}
}
/// Lexes a character inside a character class.
fn lex_class_char(&mut self, c: char, start: usize) -> Result<TokenKind> {
match c {
']' => Ok(TokenKind::CloseBracket),
// `[` opens a nested class (set union), matching the Rust `regex`
// crate / HuggingFace tokenizers — e.g. `[a[b-c]]` or `[^(\s|[.,!])]`.
'[' => Ok(TokenKind::OpenBracket),
'-' => Ok(TokenKind::Hyphen),
'^' => Ok(TokenKind::Caret),
'\\' => self.lex_escape(start),
c => Ok(TokenKind::Literal(c)),
}
}
/// Reads the remainder of an identifier (for named groups).
///
/// The caller has already consumed the first character, so an empty result
/// is the ordinary single-character-name case, not a failure.
pub fn read_ident_rest(&mut self) -> String {
let mut ident = String::new();
while let Some(c) = self.peek_char() {
if c.is_alphanumeric() || c == '_' {
self.next_char();
ident.push(c);
} else {
break;
}
}
ident
}
/// Reads a number (for repetition bounds).
pub fn read_number(&mut self) -> Result<Option<u32>> {
let mut n: u32 = 0;
let mut has_digit = false;
while let Some(c) = self.peek_char() {
if c.is_ascii_digit() {
self.next_char();
has_digit = true;
n = n.saturating_mul(10).saturating_add(c.to_digit(10).unwrap());
} else {
break;
}
}
Ok(if has_digit { Some(n) } else { None })
}
}