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
|
//! Shortcuts that span lexer/parser abstraction.
//!
//! The way Rust works, parser doesn't necessary parse text, and you might
//! tokenize text without parsing it further. So, it makes sense to keep
//! abstract token parsing, and string tokenization as completely separate
//! layers.
//!
//! However, often you do pares text into syntax trees and the glue code for
//! that needs to live somewhere. Rather than putting it to lexer or parser, we
//! use a separate shortcuts module for that.
use std::mem;
use crate::{
LexedStr, Step,
SyntaxKind::{self, *},
};
#[derive(Debug)]
pub enum StrStep<'a> {
Token { kind: SyntaxKind, text: &'a str },
Enter { kind: SyntaxKind },
Exit,
Error { msg: &'a str, pos: usize },
}
impl<'a> LexedStr<'a> {
pub fn to_input(&self) -> crate::Input {
let mut res = crate::Input::default();
let mut was_joint = false;
for i in 0..self.len() {
let kind = self.kind(i);
if kind.is_trivia() {
was_joint = false
} else {
if kind == SyntaxKind::IDENT {
let token_text = self.text(i);
let contextual_kw = SyntaxKind::from_contextual_keyword(token_text)
.unwrap_or(SyntaxKind::IDENT);
res.push_ident(contextual_kw);
} else {
if was_joint {
res.was_joint();
}
res.push(kind);
}
was_joint = true;
}
}
res
}
/// NB: only valid to call with Output from Reparser/TopLevelEntry.
pub fn intersperse_trivia(
&self,
output: &crate::Output,
sink: &mut dyn FnMut(StrStep<'_>),
) -> bool {
let mut builder = Builder { lexed: self, pos: 0, state: State::PendingEnter, sink };
for event in output.iter() {
match event {
Step::Token { kind, n_input_tokens: n_raw_tokens } => {
builder.token(kind, n_raw_tokens)
}
Step::Enter { kind } => builder.enter(kind),
Step::Exit => builder.exit(),
Step::Error { msg } => {
let text_pos = builder.lexed.text_start(builder.pos);
(builder.sink)(StrStep::Error { msg, pos: text_pos });
}
}
}
match mem::replace(&mut builder.state, State::Normal) {
State::PendingExit => {
builder.eat_trivias();
(builder.sink)(StrStep::Exit);
}
State::PendingEnter | State::Normal => unreachable!(),
}
// is_eof?
builder.pos == builder.lexed.len()
}
}
struct Builder<'a, 'b> {
lexed: &'a LexedStr<'a>,
pos: usize,
state: State,
sink: &'b mut dyn FnMut(StrStep<'_>),
}
enum State {
PendingEnter,
Normal,
PendingExit,
}
impl Builder<'_, '_> {
fn token(&mut self, kind: SyntaxKind, n_tokens: u8) {
match mem::replace(&mut self.state, State::Normal) {
State::PendingEnter => unreachable!(),
State::PendingExit => (self.sink)(StrStep::Exit),
State::Normal => (),
}
self.eat_trivias();
self.do_token(kind, n_tokens as usize);
}
fn enter(&mut self, kind: SyntaxKind) {
match mem::replace(&mut self.state, State::Normal) {
State::PendingEnter => {
(self.sink)(StrStep::Enter { kind });
// No need to attach trivias to previous node: there is no
// previous node.
return;
}
State::PendingExit => (self.sink)(StrStep::Exit),
State::Normal => (),
}
let n_trivias =
(self.pos..self.lexed.len()).take_while(|&it| self.lexed.kind(it).is_trivia()).count();
let leading_trivias = self.pos..self.pos + n_trivias;
let n_attached_trivias = n_attached_trivias(
kind,
leading_trivias.rev().map(|it| (self.lexed.kind(it), self.lexed.text(it))),
);
self.eat_n_trivias(n_trivias - n_attached_trivias);
(self.sink)(StrStep::Enter { kind });
self.eat_n_trivias(n_attached_trivias);
}
fn exit(&mut self) {
match mem::replace(&mut self.state, State::PendingExit) {
State::PendingEnter => unreachable!(),
State::PendingExit => (self.sink)(StrStep::Exit),
State::Normal => (),
}
}
fn eat_trivias(&mut self) {
while self.pos < self.lexed.len() {
let kind = self.lexed.kind(self.pos);
if !kind.is_trivia() {
break;
}
self.do_token(kind, 1);
}
}
fn eat_n_trivias(&mut self, n: usize) {
for _ in 0..n {
let kind = self.lexed.kind(self.pos);
assert!(kind.is_trivia());
self.do_token(kind, 1);
}
}
fn do_token(&mut self, kind: SyntaxKind, n_tokens: usize) {
let text = &self.lexed.range_text(self.pos..self.pos + n_tokens);
self.pos += n_tokens;
(self.sink)(StrStep::Token { kind, text });
}
}
fn n_attached_trivias<'a>(
kind: SyntaxKind,
trivias: impl Iterator<Item = (SyntaxKind, &'a str)>,
) -> usize {
match kind {
CONST | ENUM | FN | IMPL | MACRO_CALL | MACRO_DEF | MACRO_RULES | MODULE | RECORD_FIELD
| STATIC | STRUCT | TRAIT | TUPLE_FIELD | TYPE_ALIAS | UNION | USE | VARIANT => {
let mut res = 0;
let mut trivias = trivias.enumerate().peekable();
while let Some((i, (kind, text))) = trivias.next() {
match kind {
WHITESPACE if text.contains("\n\n") => {
// we check whether the next token is a doc-comment
// and skip the whitespace in this case
if let Some((COMMENT, peek_text)) = trivias.peek().map(|(_, pair)| pair) {
if is_outer(peek_text) {
continue;
}
}
break;
}
COMMENT => {
if is_inner(text) {
break;
}
res = i + 1;
}
_ => (),
}
}
res
}
_ => 0,
}
}
fn is_outer(text: &str) -> bool {
if text.starts_with("////") || text.starts_with("/***") {
return false;
}
text.starts_with("///") || text.starts_with("/**")
}
fn is_inner(text: &str) -> bool {
text.starts_with("//!") || text.starts_with("/*!")
}
|
