]>
Commit | Line | Data |
---|---|---|
064997fb FG |
1 | //! Shortcuts that span lexer/parser abstraction. |
2 | //! | |
3 | //! The way Rust works, parser doesn't necessary parse text, and you might | |
4 | //! tokenize text without parsing it further. So, it makes sense to keep | |
5 | //! abstract token parsing, and string tokenization as completely separate | |
6 | //! layers. | |
7 | //! | |
8 | //! However, often you do pares text into syntax trees and the glue code for | |
9 | //! that needs to live somewhere. Rather than putting it to lexer or parser, we | |
10 | //! use a separate shortcuts module for that. | |
11 | ||
12 | use std::mem; | |
13 | ||
14 | use crate::{ | |
15 | LexedStr, Step, | |
16 | SyntaxKind::{self, *}, | |
17 | }; | |
18 | ||
19 | #[derive(Debug)] | |
20 | pub enum StrStep<'a> { | |
21 | Token { kind: SyntaxKind, text: &'a str }, | |
22 | Enter { kind: SyntaxKind }, | |
23 | Exit, | |
24 | Error { msg: &'a str, pos: usize }, | |
25 | } | |
26 | ||
27 | impl<'a> LexedStr<'a> { | |
28 | pub fn to_input(&self) -> crate::Input { | |
29 | let mut res = crate::Input::default(); | |
30 | let mut was_joint = false; | |
31 | for i in 0..self.len() { | |
32 | let kind = self.kind(i); | |
33 | if kind.is_trivia() { | |
34 | was_joint = false | |
35 | } else { | |
36 | if kind == SyntaxKind::IDENT { | |
37 | let token_text = self.text(i); | |
38 | let contextual_kw = SyntaxKind::from_contextual_keyword(token_text) | |
39 | .unwrap_or(SyntaxKind::IDENT); | |
40 | res.push_ident(contextual_kw); | |
41 | } else { | |
42 | if was_joint { | |
43 | res.was_joint(); | |
44 | } | |
45 | res.push(kind); | |
46 | } | |
47 | was_joint = true; | |
48 | } | |
49 | } | |
50 | res | |
51 | } | |
52 | ||
53 | /// NB: only valid to call with Output from Reparser/TopLevelEntry. | |
54 | pub fn intersperse_trivia( | |
55 | &self, | |
56 | output: &crate::Output, | |
57 | sink: &mut dyn FnMut(StrStep<'_>), | |
58 | ) -> bool { | |
59 | let mut builder = Builder { lexed: self, pos: 0, state: State::PendingEnter, sink }; | |
60 | ||
61 | for event in output.iter() { | |
62 | match event { | |
63 | Step::Token { kind, n_input_tokens: n_raw_tokens } => { | |
64 | builder.token(kind, n_raw_tokens) | |
65 | } | |
66 | Step::Enter { kind } => builder.enter(kind), | |
67 | Step::Exit => builder.exit(), | |
68 | Step::Error { msg } => { | |
69 | let text_pos = builder.lexed.text_start(builder.pos); | |
70 | (builder.sink)(StrStep::Error { msg, pos: text_pos }); | |
71 | } | |
72 | } | |
73 | } | |
74 | ||
75 | match mem::replace(&mut builder.state, State::Normal) { | |
76 | State::PendingExit => { | |
77 | builder.eat_trivias(); | |
78 | (builder.sink)(StrStep::Exit); | |
79 | } | |
80 | State::PendingEnter | State::Normal => unreachable!(), | |
81 | } | |
82 | ||
f25598a0 FG |
83 | // is_eof? |
84 | builder.pos == builder.lexed.len() | |
064997fb FG |
85 | } |
86 | } | |
87 | ||
88 | struct Builder<'a, 'b> { | |
89 | lexed: &'a LexedStr<'a>, | |
90 | pos: usize, | |
91 | state: State, | |
92 | sink: &'b mut dyn FnMut(StrStep<'_>), | |
93 | } | |
94 | ||
95 | enum State { | |
96 | PendingEnter, | |
97 | Normal, | |
98 | PendingExit, | |
99 | } | |
100 | ||
101 | impl Builder<'_, '_> { | |
102 | fn token(&mut self, kind: SyntaxKind, n_tokens: u8) { | |
103 | match mem::replace(&mut self.state, State::Normal) { | |
104 | State::PendingEnter => unreachable!(), | |
105 | State::PendingExit => (self.sink)(StrStep::Exit), | |
106 | State::Normal => (), | |
107 | } | |
108 | self.eat_trivias(); | |
109 | self.do_token(kind, n_tokens as usize); | |
110 | } | |
111 | ||
112 | fn enter(&mut self, kind: SyntaxKind) { | |
113 | match mem::replace(&mut self.state, State::Normal) { | |
114 | State::PendingEnter => { | |
115 | (self.sink)(StrStep::Enter { kind }); | |
116 | // No need to attach trivias to previous node: there is no | |
117 | // previous node. | |
118 | return; | |
119 | } | |
120 | State::PendingExit => (self.sink)(StrStep::Exit), | |
121 | State::Normal => (), | |
122 | } | |
123 | ||
124 | let n_trivias = | |
125 | (self.pos..self.lexed.len()).take_while(|&it| self.lexed.kind(it).is_trivia()).count(); | |
126 | let leading_trivias = self.pos..self.pos + n_trivias; | |
127 | let n_attached_trivias = n_attached_trivias( | |
128 | kind, | |
129 | leading_trivias.rev().map(|it| (self.lexed.kind(it), self.lexed.text(it))), | |
130 | ); | |
131 | self.eat_n_trivias(n_trivias - n_attached_trivias); | |
132 | (self.sink)(StrStep::Enter { kind }); | |
133 | self.eat_n_trivias(n_attached_trivias); | |
134 | } | |
135 | ||
136 | fn exit(&mut self) { | |
137 | match mem::replace(&mut self.state, State::PendingExit) { | |
138 | State::PendingEnter => unreachable!(), | |
139 | State::PendingExit => (self.sink)(StrStep::Exit), | |
140 | State::Normal => (), | |
141 | } | |
142 | } | |
143 | ||
144 | fn eat_trivias(&mut self) { | |
145 | while self.pos < self.lexed.len() { | |
146 | let kind = self.lexed.kind(self.pos); | |
147 | if !kind.is_trivia() { | |
148 | break; | |
149 | } | |
150 | self.do_token(kind, 1); | |
151 | } | |
152 | } | |
153 | ||
154 | fn eat_n_trivias(&mut self, n: usize) { | |
155 | for _ in 0..n { | |
156 | let kind = self.lexed.kind(self.pos); | |
157 | assert!(kind.is_trivia()); | |
158 | self.do_token(kind, 1); | |
159 | } | |
160 | } | |
161 | ||
162 | fn do_token(&mut self, kind: SyntaxKind, n_tokens: usize) { | |
163 | let text = &self.lexed.range_text(self.pos..self.pos + n_tokens); | |
164 | self.pos += n_tokens; | |
165 | (self.sink)(StrStep::Token { kind, text }); | |
166 | } | |
167 | } | |
168 | ||
169 | fn n_attached_trivias<'a>( | |
170 | kind: SyntaxKind, | |
171 | trivias: impl Iterator<Item = (SyntaxKind, &'a str)>, | |
172 | ) -> usize { | |
173 | match kind { | |
174 | CONST | ENUM | FN | IMPL | MACRO_CALL | MACRO_DEF | MACRO_RULES | MODULE | RECORD_FIELD | |
175 | | STATIC | STRUCT | TRAIT | TUPLE_FIELD | TYPE_ALIAS | UNION | USE | VARIANT => { | |
176 | let mut res = 0; | |
177 | let mut trivias = trivias.enumerate().peekable(); | |
178 | ||
179 | while let Some((i, (kind, text))) = trivias.next() { | |
180 | match kind { | |
181 | WHITESPACE if text.contains("\n\n") => { | |
182 | // we check whether the next token is a doc-comment | |
183 | // and skip the whitespace in this case | |
184 | if let Some((COMMENT, peek_text)) = trivias.peek().map(|(_, pair)| pair) { | |
185 | if is_outer(peek_text) { | |
186 | continue; | |
187 | } | |
188 | } | |
189 | break; | |
190 | } | |
191 | COMMENT => { | |
192 | if is_inner(text) { | |
193 | break; | |
194 | } | |
195 | res = i + 1; | |
196 | } | |
197 | _ => (), | |
198 | } | |
199 | } | |
200 | res | |
201 | } | |
202 | _ => 0, | |
203 | } | |
204 | } | |
205 | ||
206 | fn is_outer(text: &str) -> bool { | |
207 | if text.starts_with("////") || text.starts_with("/***") { | |
208 | return false; | |
209 | } | |
210 | text.starts_with("///") || text.starts_with("/**") | |
211 | } | |
212 | ||
213 | fn is_inner(text: &str) -> bool { | |
214 | text.starts_with("//!") || text.starts_with("/*!") | |
215 | } |