]>
Commit | Line | Data |
---|---|---|
49aad941 FG |
1 | use std::ops::Range; |
2 | ||
9ffffee4 | 3 | use crate::errors; |
3c0e092e | 4 | use crate::lexer::unicode_chars::UNICODE_ARRAY; |
9ffffee4 | 5 | use crate::make_unclosed_delims_error; |
136023e0 | 6 | use rustc_ast::ast::{self, AttrStyle}; |
04454e1e | 7 | use rustc_ast::token::{self, CommentKind, Delimiter, Token, TokenKind}; |
2b03887a | 8 | use rustc_ast::tokenstream::TokenStream; |
3c0e092e | 9 | use rustc_ast::util::unicode::contains_text_flow_control_chars; |
9ffffee4 | 10 | use rustc_errors::{error_code, Applicability, Diagnostic, DiagnosticBuilder, StashKey}; |
49aad941 | 11 | use rustc_lexer::unescape::{self, EscapeError, Mode}; |
2b03887a | 12 | use rustc_lexer::Cursor; |
1b1a35ee | 13 | use rustc_lexer::{Base, DocStyle, RawStrError}; |
c295e0f8 | 14 | use rustc_session::lint::builtin::{ |
3c0e092e | 15 | RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX, TEXT_DIRECTION_CODEPOINT_IN_COMMENT, |
c295e0f8 | 16 | }; |
136023e0 | 17 | use rustc_session::lint::BuiltinLintDiagnostics; |
74b04a01 | 18 | use rustc_session::parse::ParseSess; |
dfeec247 | 19 | use rustc_span::symbol::{sym, Symbol}; |
136023e0 | 20 | use rustc_span::{edition::Edition, BytePos, Pos, Span}; |
1a4d82fc | 21 | |
9ffffee4 | 22 | mod diagnostics; |
32a655c1 | 23 | mod tokentrees; |
60c5eb7d | 24 | mod unescape_error_reporting; |
dfeec247 | 25 | mod unicode_chars; |
ba9703b0 | 26 | |
5869c6ff | 27 | use unescape_error_reporting::{emit_unescape_error, escaped_char}; |
1a4d82fc | 28 | |
064997fb FG |
29 | // This type is used a lot. Make sure it doesn't unintentionally get bigger. |
30 | // | |
31 | // This assertion is in this crate, rather than in `rustc_lexer`, because that | |
32 | // crate cannot depend on `rustc_data_structures`. | |
33 | #[cfg(all(target_arch = "x86_64", target_pointer_width = "64"))] | |
34 | rustc_data_structures::static_assert_size!(rustc_lexer::Token, 12); | |
35 | ||
9fa01778 | 36 | #[derive(Clone, Debug)] |
9ffffee4 | 37 | pub struct UnmatchedDelim { |
04454e1e FG |
38 | pub expected_delim: Delimiter, |
39 | pub found_delim: Option<Delimiter>, | |
9fa01778 XL |
40 | pub found_span: Span, |
41 | pub unclosed_span: Option<Span>, | |
42 | pub candidate_span: Option<Span>, | |
43 | } | |
44 | ||
923072b8 | 45 | pub(crate) fn parse_token_trees<'a>( |
1b1a35ee | 46 | sess: &'a ParseSess, |
2b03887a FG |
47 | mut src: &'a str, |
48 | mut start_pos: BytePos, | |
1b1a35ee | 49 | override_span: Option<Span>, |
9ffffee4 | 50 | ) -> Result<TokenStream, Vec<Diagnostic>> { |
2b03887a FG |
51 | // Skip `#!`, if present. |
52 | if let Some(shebang_len) = rustc_lexer::strip_shebang(src) { | |
53 | src = &src[shebang_len..]; | |
54 | start_pos = start_pos + BytePos::from_usize(shebang_len); | |
55 | } | |
56 | ||
57 | let cursor = Cursor::new(src); | |
9c376795 FG |
58 | let string_reader = StringReader { |
59 | sess, | |
60 | start_pos, | |
61 | pos: start_pos, | |
62 | src, | |
63 | cursor, | |
64 | override_span, | |
65 | nbsp_is_whitespace: false, | |
66 | }; | |
9ffffee4 FG |
67 | let (token_trees, unmatched_delims) = |
68 | tokentrees::TokenTreesReader::parse_all_token_trees(string_reader); | |
69 | match token_trees { | |
70 | Ok(stream) if unmatched_delims.is_empty() => Ok(stream), | |
71 | _ => { | |
49aad941 | 72 | // Return error if there are unmatched delimiters or unclosed delimiters. |
9ffffee4 FG |
73 | // We emit delimiter mismatch errors first, then emit the unclosing delimiter mismatch |
74 | // because the delimiter mismatch is more likely to be the root cause of error | |
75 | ||
76 | let mut buffer = Vec::with_capacity(1); | |
77 | // Not using `emit_unclosed_delims` to use `db.buffer` | |
78 | for unmatched in unmatched_delims { | |
79 | if let Some(err) = make_unclosed_delims_error(unmatched, &sess) { | |
80 | err.buffer(&mut buffer); | |
81 | } | |
82 | } | |
83 | if let Err(err) = token_trees { | |
84 | // Add unclosing delimiter error | |
85 | err.buffer(&mut buffer); | |
86 | } | |
87 | Err(buffer) | |
88 | } | |
89 | } | |
1b1a35ee XL |
90 | } |
91 | ||
92 | struct StringReader<'a> { | |
416331ca XL |
93 | sess: &'a ParseSess, |
94 | /// Initial position, read-only. | |
95 | start_pos: BytePos, | |
96 | /// The absolute offset within the source_map of the current character. | |
f9f354fc | 97 | pos: BytePos, |
416331ca | 98 | /// Source text to tokenize. |
1b1a35ee | 99 | src: &'a str, |
2b03887a FG |
100 | /// Cursor for getting lexer tokens. |
101 | cursor: Cursor<'a>, | |
48663c56 | 102 | override_span: Option<Span>, |
9c376795 FG |
103 | /// When a "unknown start of token: \u{a0}" has already been emitted earlier |
104 | /// in this file, it's safe to treat further occurrences of the non-breaking | |
105 | /// space character as whitespace. | |
106 | nbsp_is_whitespace: bool, | |
cc61c64b XL |
107 | } |
108 | ||
32a655c1 | 109 | impl<'a> StringReader<'a> { |
416331ca | 110 | fn mk_sp(&self, lo: BytePos, hi: BytePos) -> Span { |
e1599b0c | 111 | self.override_span.unwrap_or_else(|| Span::with_root_ctxt(lo, hi)) |
8bb4bdeb XL |
112 | } |
113 | ||
2b03887a FG |
114 | /// Returns the next token, paired with a bool indicating if the token was |
115 | /// preceded by whitespace. | |
116 | fn next_token(&mut self) -> (Token, bool) { | |
117 | let mut preceded_by_whitespace = false; | |
9c376795 | 118 | let mut swallow_next_invalid = 0; |
1b1a35ee XL |
119 | // Skip trivial (whitespace & comments) tokens |
120 | loop { | |
2b03887a | 121 | let token = self.cursor.advance_token(); |
1b1a35ee | 122 | let start = self.pos; |
064997fb | 123 | self.pos = self.pos + BytePos(token.len); |
416331ca | 124 | |
1b1a35ee | 125 | debug!("next_token: {:?}({:?})", token.kind, self.str_from(start)); |
416331ca | 126 | |
2b03887a FG |
127 | // Now "cook" the token, converting the simple `rustc_lexer::TokenKind` enum into a |
128 | // rich `rustc_ast::TokenKind`. This turns strings into interned symbols and runs | |
129 | // additional validation. | |
130 | let kind = match token.kind { | |
131 | rustc_lexer::TokenKind::LineComment { doc_style } => { | |
132 | // Skip non-doc comments | |
133 | let Some(doc_style) = doc_style else { | |
134 | self.lint_unicode_text_flow(start); | |
135 | preceded_by_whitespace = true; | |
136 | continue; | |
137 | }; | |
138 | ||
139 | // Opening delimiter of the length 3 is not included into the symbol. | |
140 | let content_start = start + BytePos(3); | |
141 | let content = self.str_from(content_start); | |
142 | self.cook_doc_comment(content_start, content, CommentKind::Line, doc_style) | |
143 | } | |
144 | rustc_lexer::TokenKind::BlockComment { doc_style, terminated } => { | |
145 | if !terminated { | |
146 | self.report_unterminated_block_comment(start, doc_style); | |
147 | } | |
148 | ||
149 | // Skip non-doc comments | |
150 | let Some(doc_style) = doc_style else { | |
151 | self.lint_unicode_text_flow(start); | |
152 | preceded_by_whitespace = true; | |
153 | continue; | |
154 | }; | |
155 | ||
156 | // Opening delimiter of the length 3 and closing delimiter of the length 2 | |
157 | // are not included into the symbol. | |
158 | let content_start = start + BytePos(3); | |
159 | let content_end = self.pos - BytePos(if terminated { 2 } else { 0 }); | |
160 | let content = self.str_from_to(content_start, content_end); | |
161 | self.cook_doc_comment(content_start, content, CommentKind::Block, doc_style) | |
162 | } | |
163 | rustc_lexer::TokenKind::Whitespace => { | |
164 | preceded_by_whitespace = true; | |
165 | continue; | |
166 | } | |
167 | rustc_lexer::TokenKind::Ident => { | |
168 | let sym = nfc_normalize(self.str_from(start)); | |
1b1a35ee | 169 | let span = self.mk_sp(start, self.pos); |
2b03887a FG |
170 | self.sess.symbol_gallery.insert(sym, span); |
171 | token::Ident(sym, false) | |
1b1a35ee | 172 | } |
2b03887a FG |
173 | rustc_lexer::TokenKind::RawIdent => { |
174 | let sym = nfc_normalize(self.str_from(start + BytePos(2))); | |
175 | let span = self.mk_sp(start, self.pos); | |
176 | self.sess.symbol_gallery.insert(sym, span); | |
177 | if !sym.can_be_raw() { | |
9ffffee4 | 178 | self.sess.emit_err(errors::CannotBeRawIdent { span, ident: sym }); |
2b03887a | 179 | } |
353b0b11 | 180 | self.sess.raw_identifier_spans.push(span); |
2b03887a FG |
181 | token::Ident(sym, true) |
182 | } | |
183 | rustc_lexer::TokenKind::UnknownPrefix => { | |
184 | self.report_unknown_prefix(start); | |
185 | let sym = nfc_normalize(self.str_from(start)); | |
186 | let span = self.mk_sp(start, self.pos); | |
187 | self.sess.symbol_gallery.insert(sym, span); | |
188 | token::Ident(sym, false) | |
189 | } | |
190 | rustc_lexer::TokenKind::InvalidIdent | |
191 | // Do not recover an identifier with emoji if the codepoint is a confusable | |
192 | // with a recoverable substitution token, like `➖`. | |
193 | if !UNICODE_ARRAY | |
194 | .iter() | |
195 | .any(|&(c, _, _)| { | |
196 | let sym = self.str_from(start); | |
197 | sym.chars().count() == 1 && c == sym.chars().next().unwrap() | |
198 | }) => | |
199 | { | |
200 | let sym = nfc_normalize(self.str_from(start)); | |
201 | let span = self.mk_sp(start, self.pos); | |
202 | self.sess.bad_unicode_identifiers.borrow_mut().entry(sym).or_default() | |
203 | .push(span); | |
204 | token::Ident(sym, false) | |
205 | } | |
206 | rustc_lexer::TokenKind::Literal { kind, suffix_start } => { | |
207 | let suffix_start = start + BytePos(suffix_start); | |
208 | let (kind, symbol) = self.cook_lexer_literal(start, suffix_start, kind); | |
49aad941 FG |
209 | if let token::LitKind::CStr | token::LitKind::CStrRaw(_) = kind { |
210 | self.sess.gated_spans.gate(sym::c_str_literals, self.mk_sp(start, self.pos)); | |
211 | } | |
2b03887a FG |
212 | let suffix = if suffix_start < self.pos { |
213 | let string = self.str_from(suffix_start); | |
214 | if string == "_" { | |
215 | self.sess | |
216 | .span_diagnostic | |
49aad941 | 217 | .emit_err(errors::UnderscoreLiteralSuffix { span: self.mk_sp(suffix_start, self.pos) }); |
2b03887a FG |
218 | None |
219 | } else { | |
220 | Some(Symbol::intern(string)) | |
221 | } | |
222 | } else { | |
223 | None | |
224 | }; | |
225 | token::Literal(token::Lit { kind, symbol, suffix }) | |
226 | } | |
227 | rustc_lexer::TokenKind::Lifetime { starts_with_number } => { | |
228 | // Include the leading `'` in the real identifier, for macro | |
229 | // expansion purposes. See #12512 for the gory details of why | |
230 | // this is necessary. | |
231 | let lifetime_name = self.str_from(start); | |
232 | if starts_with_number { | |
233 | let span = self.mk_sp(start, self.pos); | |
234 | let mut diag = self.sess.struct_err("lifetimes cannot start with a number"); | |
235 | diag.set_span(span); | |
236 | diag.stash(span, StashKey::LifetimeIsChar); | |
237 | } | |
238 | let ident = Symbol::intern(lifetime_name); | |
239 | token::Lifetime(ident) | |
240 | } | |
241 | rustc_lexer::TokenKind::Semi => token::Semi, | |
242 | rustc_lexer::TokenKind::Comma => token::Comma, | |
243 | rustc_lexer::TokenKind::Dot => token::Dot, | |
244 | rustc_lexer::TokenKind::OpenParen => token::OpenDelim(Delimiter::Parenthesis), | |
245 | rustc_lexer::TokenKind::CloseParen => token::CloseDelim(Delimiter::Parenthesis), | |
246 | rustc_lexer::TokenKind::OpenBrace => token::OpenDelim(Delimiter::Brace), | |
247 | rustc_lexer::TokenKind::CloseBrace => token::CloseDelim(Delimiter::Brace), | |
248 | rustc_lexer::TokenKind::OpenBracket => token::OpenDelim(Delimiter::Bracket), | |
249 | rustc_lexer::TokenKind::CloseBracket => token::CloseDelim(Delimiter::Bracket), | |
250 | rustc_lexer::TokenKind::At => token::At, | |
251 | rustc_lexer::TokenKind::Pound => token::Pound, | |
252 | rustc_lexer::TokenKind::Tilde => token::Tilde, | |
253 | rustc_lexer::TokenKind::Question => token::Question, | |
254 | rustc_lexer::TokenKind::Colon => token::Colon, | |
255 | rustc_lexer::TokenKind::Dollar => token::Dollar, | |
256 | rustc_lexer::TokenKind::Eq => token::Eq, | |
257 | rustc_lexer::TokenKind::Bang => token::Not, | |
258 | rustc_lexer::TokenKind::Lt => token::Lt, | |
259 | rustc_lexer::TokenKind::Gt => token::Gt, | |
260 | rustc_lexer::TokenKind::Minus => token::BinOp(token::Minus), | |
261 | rustc_lexer::TokenKind::And => token::BinOp(token::And), | |
262 | rustc_lexer::TokenKind::Or => token::BinOp(token::Or), | |
263 | rustc_lexer::TokenKind::Plus => token::BinOp(token::Plus), | |
264 | rustc_lexer::TokenKind::Star => token::BinOp(token::Star), | |
265 | rustc_lexer::TokenKind::Slash => token::BinOp(token::Slash), | |
266 | rustc_lexer::TokenKind::Caret => token::BinOp(token::Caret), | |
267 | rustc_lexer::TokenKind::Percent => token::BinOp(token::Percent), | |
268 | ||
269 | rustc_lexer::TokenKind::Unknown | rustc_lexer::TokenKind::InvalidIdent => { | |
9c376795 FG |
270 | // Don't emit diagnostics for sequences of the same invalid token |
271 | if swallow_next_invalid > 0 { | |
272 | swallow_next_invalid -= 1; | |
273 | continue; | |
274 | } | |
275 | let mut it = self.str_from_to_end(start).chars(); | |
276 | let c = it.next().unwrap(); | |
277 | if c == '\u{00a0}' { | |
278 | // If an error has already been reported on non-breaking | |
279 | // space characters earlier in the file, treat all | |
280 | // subsequent occurrences as whitespace. | |
281 | if self.nbsp_is_whitespace { | |
282 | preceded_by_whitespace = true; | |
283 | continue; | |
284 | } | |
285 | self.nbsp_is_whitespace = true; | |
286 | } | |
287 | let repeats = it.take_while(|c1| *c1 == c).count(); | |
2b03887a FG |
288 | // FIXME: the lexer could be used to turn the ASCII version of unicode |
289 | // homoglyphs, instead of keeping a table in `check_for_substitution`into the | |
290 | // token. Ideally, this should be inside `rustc_lexer`. However, we should | |
291 | // first remove compound tokens like `<<` from `rustc_lexer`, and then add | |
292 | // fancier error recovery to it, as there will be less overall work to do this | |
293 | // way. | |
9ffffee4 FG |
294 | let (token, sugg) = unicode_chars::check_for_substitution(self, start, c, repeats+1); |
295 | self.sess.emit_err(errors::UnknownTokenStart { | |
296 | span: self.mk_sp(start, self.pos + Pos::from_usize(repeats * c.len_utf8())), | |
297 | escaped: escaped_char(c), | |
298 | sugg, | |
299 | null: if c == '\x00' {Some(errors::UnknownTokenNull)} else {None}, | |
300 | repeat: if repeats > 0 { | |
301 | swallow_next_invalid = repeats; | |
302 | Some(errors::UnknownTokenRepeat { repeats }) | |
303 | } else {None} | |
304 | }); | |
305 | ||
2b03887a FG |
306 | if let Some(token) = token { |
307 | token | |
308 | } else { | |
309 | preceded_by_whitespace = true; | |
310 | continue; | |
311 | } | |
312 | } | |
313 | rustc_lexer::TokenKind::Eof => token::Eof, | |
314 | }; | |
315 | let span = self.mk_sp(start, self.pos); | |
316 | return (Token::new(kind, span), preceded_by_whitespace); | |
1b1a35ee | 317 | } |
1a4d82fc JJ |
318 | } |
319 | ||
dfeec247 XL |
320 | fn struct_fatal_span_char( |
321 | &self, | |
322 | from_pos: BytePos, | |
323 | to_pos: BytePos, | |
324 | m: &str, | |
325 | c: char, | |
5e7ed085 | 326 | ) -> DiagnosticBuilder<'a, !> { |
5869c6ff XL |
327 | self.sess |
328 | .span_diagnostic | |
49aad941 | 329 | .struct_span_fatal(self.mk_sp(from_pos, to_pos), format!("{}: {}", m, escaped_char(c))) |
9cc50fc6 | 330 | } |
1a4d82fc | 331 | |
c295e0f8 XL |
332 | /// Detect usages of Unicode codepoints changing the direction of the text on screen and loudly |
333 | /// complain about it. | |
334 | fn lint_unicode_text_flow(&self, start: BytePos) { | |
335 | // Opening delimiter of the length 2 is not included into the comment text. | |
336 | let content_start = start + BytePos(2); | |
337 | let content = self.str_from(content_start); | |
3c0e092e XL |
338 | if contains_text_flow_control_chars(content) { |
339 | let span = self.mk_sp(start, self.pos); | |
c295e0f8 XL |
340 | self.sess.buffer_lint_with_diagnostic( |
341 | &TEXT_DIRECTION_CODEPOINT_IN_COMMENT, | |
342 | span, | |
343 | ast::CRATE_NODE_ID, | |
344 | "unicode codepoint changing visible direction of text present in comment", | |
345 | BuiltinLintDiagnostics::UnicodeTextFlow(span, content.to_string()), | |
346 | ); | |
347 | } | |
348 | } | |
349 | ||
3dfed10e XL |
350 | fn cook_doc_comment( |
351 | &self, | |
352 | content_start: BytePos, | |
353 | content: &str, | |
354 | comment_kind: CommentKind, | |
355 | doc_style: DocStyle, | |
356 | ) -> TokenKind { | |
357 | if content.contains('\r') { | |
358 | for (idx, _) in content.char_indices().filter(|&(_, c)| c == '\r') { | |
9ffffee4 | 359 | let span = self.mk_sp( |
3dfed10e XL |
360 | content_start + BytePos(idx as u32), |
361 | content_start + BytePos(idx as u32 + 1), | |
3dfed10e | 362 | ); |
9ffffee4 FG |
363 | let block = matches!(comment_kind, CommentKind::Block); |
364 | self.sess.emit_err(errors::CrDocComment { span, block }); | |
3dfed10e XL |
365 | } |
366 | } | |
367 | ||
368 | let attr_style = match doc_style { | |
369 | DocStyle::Outer => AttrStyle::Outer, | |
370 | DocStyle::Inner => AttrStyle::Inner, | |
371 | }; | |
372 | ||
373 | token::DocComment(comment_kind, attr_style, Symbol::intern(content)) | |
374 | } | |
375 | ||
416331ca XL |
376 | fn cook_lexer_literal( |
377 | &self, | |
378 | start: BytePos, | |
487cf647 | 379 | end: BytePos, |
dfeec247 | 380 | kind: rustc_lexer::LiteralKind, |
416331ca | 381 | ) -> (token::LitKind, Symbol) { |
487cf647 | 382 | match kind { |
416331ca XL |
383 | rustc_lexer::LiteralKind::Char { terminated } => { |
384 | if !terminated { | |
17df50a5 | 385 | self.sess.span_diagnostic.span_fatal_with_code( |
487cf647 | 386 | self.mk_sp(start, end), |
17df50a5 XL |
387 | "unterminated character literal", |
388 | error_code!(E0762), | |
389 | ) | |
416331ca | 390 | } |
487cf647 | 391 | self.cook_quoted(token::Char, Mode::Char, start, end, 1, 1) // ' ' |
dfeec247 | 392 | } |
416331ca XL |
393 | rustc_lexer::LiteralKind::Byte { terminated } => { |
394 | if !terminated { | |
17df50a5 | 395 | self.sess.span_diagnostic.span_fatal_with_code( |
487cf647 | 396 | self.mk_sp(start + BytePos(1), end), |
17df50a5 XL |
397 | "unterminated byte constant", |
398 | error_code!(E0763), | |
399 | ) | |
416331ca | 400 | } |
487cf647 | 401 | self.cook_quoted(token::Byte, Mode::Byte, start, end, 2, 1) // b' ' |
dfeec247 | 402 | } |
416331ca XL |
403 | rustc_lexer::LiteralKind::Str { terminated } => { |
404 | if !terminated { | |
17df50a5 | 405 | self.sess.span_diagnostic.span_fatal_with_code( |
487cf647 | 406 | self.mk_sp(start, end), |
17df50a5 XL |
407 | "unterminated double quote string", |
408 | error_code!(E0765), | |
409 | ) | |
416331ca | 410 | } |
487cf647 | 411 | self.cook_quoted(token::Str, Mode::Str, start, end, 1, 1) // " " |
416331ca XL |
412 | } |
413 | rustc_lexer::LiteralKind::ByteStr { terminated } => { | |
414 | if !terminated { | |
17df50a5 | 415 | self.sess.span_diagnostic.span_fatal_with_code( |
487cf647 | 416 | self.mk_sp(start + BytePos(1), end), |
17df50a5 XL |
417 | "unterminated double quote byte string", |
418 | error_code!(E0766), | |
419 | ) | |
416331ca | 420 | } |
487cf647 | 421 | self.cook_quoted(token::ByteStr, Mode::ByteStr, start, end, 2, 1) // b" " |
416331ca | 422 | } |
49aad941 FG |
423 | rustc_lexer::LiteralKind::CStr { terminated } => { |
424 | if !terminated { | |
425 | self.sess.span_diagnostic.span_fatal_with_code( | |
426 | self.mk_sp(start + BytePos(1), end), | |
427 | "unterminated C string", | |
428 | error_code!(E0767), | |
429 | ) | |
430 | } | |
431 | self.cook_c_string(token::CStr, Mode::CStr, start, end, 2, 1) // c" " | |
432 | } | |
064997fb FG |
433 | rustc_lexer::LiteralKind::RawStr { n_hashes } => { |
434 | if let Some(n_hashes) = n_hashes { | |
435 | let n = u32::from(n_hashes); | |
487cf647 FG |
436 | let kind = token::StrRaw(n_hashes); |
437 | self.cook_quoted(kind, Mode::RawStr, start, end, 2 + n, 1 + n) // r##" "## | |
064997fb FG |
438 | } else { |
439 | self.report_raw_str_error(start, 1); | |
440 | } | |
416331ca | 441 | } |
064997fb FG |
442 | rustc_lexer::LiteralKind::RawByteStr { n_hashes } => { |
443 | if let Some(n_hashes) = n_hashes { | |
444 | let n = u32::from(n_hashes); | |
487cf647 FG |
445 | let kind = token::ByteStrRaw(n_hashes); |
446 | self.cook_quoted(kind, Mode::RawByteStr, start, end, 3 + n, 1 + n) // br##" "## | |
064997fb FG |
447 | } else { |
448 | self.report_raw_str_error(start, 2); | |
449 | } | |
416331ca | 450 | } |
49aad941 FG |
451 | rustc_lexer::LiteralKind::RawCStr { n_hashes } => { |
452 | if let Some(n_hashes) = n_hashes { | |
453 | let n = u32::from(n_hashes); | |
454 | let kind = token::CStrRaw(n_hashes); | |
455 | self.cook_c_string(kind, Mode::RawCStr, start, end, 3 + n, 1 + n) // cr##" "## | |
456 | } else { | |
457 | self.report_raw_str_error(start, 2); | |
458 | } | |
459 | } | |
416331ca | 460 | rustc_lexer::LiteralKind::Int { base, empty_int } => { |
487cf647 | 461 | if empty_int { |
9ffffee4 FG |
462 | let span = self.mk_sp(start, end); |
463 | self.sess.emit_err(errors::NoDigitsLiteral { span }); | |
416331ca XL |
464 | (token::Integer, sym::integer(0)) |
465 | } else { | |
487cf647 FG |
466 | if matches!(base, Base::Binary | Base::Octal) { |
467 | let base = base as u32; | |
468 | let s = self.str_from_to(start + BytePos(2), end); | |
469 | for (idx, c) in s.char_indices() { | |
9ffffee4 FG |
470 | let span = self.mk_sp( |
471 | start + BytePos::from_usize(2 + idx), | |
472 | start + BytePos::from_usize(2 + idx + c.len_utf8()), | |
473 | ); | |
487cf647 | 474 | if c != '_' && c.to_digit(base).is_none() { |
9ffffee4 | 475 | self.sess.emit_err(errors::InvalidDigitLiteral { span, base }); |
487cf647 FG |
476 | } |
477 | } | |
478 | } | |
479 | (token::Integer, self.symbol_from_to(start, end)) | |
480 | } | |
dfeec247 | 481 | } |
416331ca XL |
482 | rustc_lexer::LiteralKind::Float { base, empty_exponent } => { |
483 | if empty_exponent { | |
9ffffee4 FG |
484 | let span = self.mk_sp(start, self.pos); |
485 | self.sess.emit_err(errors::EmptyExponentFloat { span }); | |
416331ca | 486 | } |
9ffffee4 FG |
487 | let base = match base { |
488 | Base::Hexadecimal => Some("hexadecimal"), | |
489 | Base::Octal => Some("octal"), | |
490 | Base::Binary => Some("binary"), | |
491 | _ => None, | |
492 | }; | |
493 | if let Some(base) = base { | |
494 | let span = self.mk_sp(start, end); | |
495 | self.sess.emit_err(errors::FloatLiteralUnsupportedBase { span, base }); | |
416331ca | 496 | } |
487cf647 | 497 | (token::Float, self.symbol_from_to(start, end)) |
dfeec247 | 498 | } |
487cf647 | 499 | } |
f9f354fc XL |
500 | } |
501 | ||
94b46f34 XL |
502 | #[inline] |
503 | fn src_index(&self, pos: BytePos) -> usize { | |
416331ca | 504 | (pos - self.start_pos).to_usize() |
1a4d82fc JJ |
505 | } |
506 | ||
dc9dc135 XL |
507 | /// Slice of the source text from `start` up to but excluding `self.pos`, |
508 | /// meaning the slice does not include the character `self.ch`. | |
9c376795 | 509 | fn str_from(&self, start: BytePos) -> &'a str { |
dc9dc135 | 510 | self.str_from_to(start, self.pos) |
1a4d82fc JJ |
511 | } |
512 | ||
dc9dc135 XL |
513 | /// As symbol_from, with an explicit endpoint. |
514 | fn symbol_from_to(&self, start: BytePos, end: BytePos) -> Symbol { | |
1a4d82fc | 515 | debug!("taking an ident from {:?} to {:?}", start, end); |
dc9dc135 | 516 | Symbol::intern(self.str_from_to(start, end)) |
1a4d82fc JJ |
517 | } |
518 | ||
dc9dc135 | 519 | /// Slice of the source text spanning from `start` up to but excluding `end`. |
9c376795 | 520 | fn str_from_to(&self, start: BytePos, end: BytePos) -> &'a str { |
dc9dc135 | 521 | &self.src[self.src_index(start)..self.src_index(end)] |
1a4d82fc JJ |
522 | } |
523 | ||
9c376795 FG |
524 | /// Slice of the source text spanning from `start` until the end |
525 | fn str_from_to_end(&self, start: BytePos) -> &'a str { | |
526 | &self.src[self.src_index(start)..] | |
527 | } | |
528 | ||
064997fb FG |
529 | fn report_raw_str_error(&self, start: BytePos, prefix_len: u32) -> ! { |
530 | match rustc_lexer::validate_raw_str(self.str_from(start), prefix_len) { | |
531 | Err(RawStrError::InvalidStarter { bad_char }) => { | |
f035d41b XL |
532 | self.report_non_started_raw_string(start, bad_char) |
533 | } | |
064997fb | 534 | Err(RawStrError::NoTerminator { expected, found, possible_terminator_offset }) => self |
f035d41b | 535 | .report_unterminated_raw_string(start, expected, possible_terminator_offset, found), |
064997fb | 536 | Err(RawStrError::TooManyDelimiters { found }) => { |
f035d41b | 537 | self.report_too_many_hashes(start, found) |
ba9703b0 | 538 | } |
064997fb | 539 | Ok(()) => panic!("no error found for supposedly invalid raw string literal"), |
ba9703b0 XL |
540 | } |
541 | } | |
542 | ||
f035d41b | 543 | fn report_non_started_raw_string(&self, start: BytePos, bad_char: char) -> ! { |
dfeec247 XL |
544 | self.struct_fatal_span_char( |
545 | start, | |
546 | self.pos, | |
f035d41b | 547 | "found invalid character; only `#` is allowed in raw string delimitation", |
dfeec247 XL |
548 | bad_char, |
549 | ) | |
5e7ed085 | 550 | .emit() |
1a4d82fc JJ |
551 | } |
552 | ||
ba9703b0 XL |
553 | fn report_unterminated_raw_string( |
554 | &self, | |
555 | start: BytePos, | |
064997fb FG |
556 | n_hashes: u32, |
557 | possible_offset: Option<u32>, | |
558 | found_terminators: u32, | |
ba9703b0 | 559 | ) -> ! { |
74b04a01 XL |
560 | let mut err = self.sess.span_diagnostic.struct_span_fatal_with_code( |
561 | self.mk_sp(start, start), | |
562 | "unterminated raw string", | |
563 | error_code!(E0748), | |
564 | ); | |
ba9703b0 | 565 | |
dfeec247 | 566 | err.span_label(self.mk_sp(start, start), "unterminated raw string"); |
1a4d82fc | 567 | |
416331ca | 568 | if n_hashes > 0 { |
49aad941 | 569 | err.note(format!( |
dfeec247 | 570 | "this raw string should be terminated with `\"{}`", |
064997fb | 571 | "#".repeat(n_hashes as usize) |
dfeec247 | 572 | )); |
48663c56 | 573 | } |
1a4d82fc | 574 | |
ba9703b0 | 575 | if let Some(possible_offset) = possible_offset { |
353b0b11 FG |
576 | let lo = start + BytePos(possible_offset); |
577 | let hi = lo + BytePos(found_terminators); | |
ba9703b0 XL |
578 | let span = self.mk_sp(lo, hi); |
579 | err.span_suggestion( | |
580 | span, | |
581 | "consider terminating the string here", | |
064997fb | 582 | "#".repeat(n_hashes as usize), |
ba9703b0 XL |
583 | Applicability::MaybeIncorrect, |
584 | ); | |
585 | } | |
586 | ||
5e7ed085 | 587 | err.emit() |
1a4d82fc JJ |
588 | } |
589 | ||
04454e1e FG |
590 | fn report_unterminated_block_comment(&self, start: BytePos, doc_style: Option<DocStyle>) { |
591 | let msg = match doc_style { | |
592 | Some(_) => "unterminated block doc-comment", | |
593 | None => "unterminated block comment", | |
594 | }; | |
595 | let last_bpos = self.pos; | |
596 | let mut err = self.sess.span_diagnostic.struct_span_fatal_with_code( | |
597 | self.mk_sp(start, last_bpos), | |
598 | msg, | |
599 | error_code!(E0758), | |
600 | ); | |
601 | let mut nested_block_comment_open_idxs = vec![]; | |
602 | let mut last_nested_block_comment_idxs = None; | |
603 | let mut content_chars = self.str_from(start).char_indices().peekable(); | |
604 | ||
605 | while let Some((idx, current_char)) = content_chars.next() { | |
606 | match content_chars.peek() { | |
607 | Some((_, '*')) if current_char == '/' => { | |
608 | nested_block_comment_open_idxs.push(idx); | |
609 | } | |
610 | Some((_, '/')) if current_char == '*' => { | |
611 | last_nested_block_comment_idxs = | |
612 | nested_block_comment_open_idxs.pop().map(|open_idx| (open_idx, idx)); | |
613 | } | |
614 | _ => {} | |
615 | }; | |
616 | } | |
617 | ||
618 | if let Some((nested_open_idx, nested_close_idx)) = last_nested_block_comment_idxs { | |
619 | err.span_label(self.mk_sp(start, start + BytePos(2)), msg) | |
620 | .span_label( | |
621 | self.mk_sp( | |
622 | start + BytePos(nested_open_idx as u32), | |
623 | start + BytePos(nested_open_idx as u32 + 2), | |
624 | ), | |
625 | "...as last nested comment starts here, maybe you want to close this instead?", | |
626 | ) | |
627 | .span_label( | |
628 | self.mk_sp( | |
629 | start + BytePos(nested_close_idx as u32), | |
630 | start + BytePos(nested_close_idx as u32 + 2), | |
631 | ), | |
632 | "...and last nested comment terminates here.", | |
633 | ); | |
634 | } | |
635 | ||
636 | err.emit(); | |
637 | } | |
638 | ||
136023e0 XL |
639 | // RFC 3101 introduced the idea of (reserved) prefixes. As of Rust 2021, |
640 | // using a (unknown) prefix is an error. In earlier editions, however, they | |
641 | // only result in a (allowed by default) lint, and are treated as regular | |
642 | // identifier tokens. | |
643 | fn report_unknown_prefix(&self, start: BytePos) { | |
644 | let prefix_span = self.mk_sp(start, self.pos); | |
9ffffee4 | 645 | let prefix = self.str_from_to(start, self.pos); |
136023e0 XL |
646 | |
647 | let expn_data = prefix_span.ctxt().outer_expn_data(); | |
648 | ||
649 | if expn_data.edition >= Edition::Edition2021 { | |
650 | // In Rust 2021, this is a hard error. | |
9ffffee4 FG |
651 | let sugg = if prefix == "rb" { |
652 | Some(errors::UnknownPrefixSugg::UseBr(prefix_span)) | |
94222f64 | 653 | } else if expn_data.is_root() { |
9ffffee4 FG |
654 | Some(errors::UnknownPrefixSugg::Whitespace(prefix_span.shrink_to_hi())) |
655 | } else { | |
656 | None | |
657 | }; | |
658 | self.sess.emit_err(errors::UnknownPrefix { span: prefix_span, prefix, sugg }); | |
136023e0 XL |
659 | } else { |
660 | // Before Rust 2021, only emit a lint for migration. | |
661 | self.sess.buffer_lint_with_diagnostic( | |
662 | &RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX, | |
663 | prefix_span, | |
664 | ast::CRATE_NODE_ID, | |
49aad941 | 665 | format!("prefix `{prefix}` is unknown"), |
136023e0 XL |
666 | BuiltinLintDiagnostics::ReservedPrefix(prefix_span), |
667 | ); | |
668 | } | |
669 | } | |
670 | ||
9ffffee4 FG |
671 | fn report_too_many_hashes(&self, start: BytePos, num: u32) -> ! { |
672 | self.sess.emit_fatal(errors::TooManyHashes { span: self.mk_sp(start, self.pos), num }); | |
1a4d82fc | 673 | } |
48663c56 | 674 | |
49aad941 | 675 | fn cook_common( |
5869c6ff | 676 | &self, |
487cf647 | 677 | kind: token::LitKind, |
5869c6ff | 678 | mode: Mode, |
487cf647 FG |
679 | start: BytePos, |
680 | end: BytePos, | |
5869c6ff XL |
681 | prefix_len: u32, |
682 | postfix_len: u32, | |
49aad941 | 683 | unescape: fn(&str, Mode, &mut dyn FnMut(Range<usize>, Result<(), EscapeError>)), |
487cf647 FG |
684 | ) -> (token::LitKind, Symbol) { |
685 | let mut has_fatal_err = false; | |
686 | let content_start = start + BytePos(prefix_len); | |
687 | let content_end = end - BytePos(postfix_len); | |
f9f354fc | 688 | let lit_content = self.str_from_to(content_start, content_end); |
49aad941 | 689 | unescape(lit_content, mode, &mut |range, result| { |
f9f354fc XL |
690 | // Here we only check for errors. The actual unescaping is done later. |
691 | if let Err(err) = result { | |
487cf647 | 692 | let span_with_quotes = self.mk_sp(start, end); |
fc512014 XL |
693 | let (start, end) = (range.start as u32, range.end as u32); |
694 | let lo = content_start + BytePos(start); | |
695 | let hi = lo + BytePos(end - start); | |
696 | let span = self.mk_sp(lo, hi); | |
487cf647 FG |
697 | if err.is_fatal() { |
698 | has_fatal_err = true; | |
699 | } | |
dc9dc135 XL |
700 | emit_unescape_error( |
701 | &self.sess.span_diagnostic, | |
f9f354fc XL |
702 | lit_content, |
703 | span_with_quotes, | |
fc512014 | 704 | span, |
f9f354fc | 705 | mode, |
dc9dc135 XL |
706 | range, |
707 | err, | |
f9f354fc | 708 | ); |
dc9dc135 | 709 | } |
f9f354fc | 710 | }); |
1a4d82fc | 711 | |
487cf647 FG |
712 | // We normally exclude the quotes for the symbol, but for errors we |
713 | // include it because it results in clearer error messages. | |
714 | if !has_fatal_err { | |
715 | (kind, Symbol::intern(lit_content)) | |
716 | } else { | |
717 | (token::Err, self.symbol_from_to(start, end)) | |
416331ca XL |
718 | } |
719 | } | |
49aad941 FG |
720 | |
721 | fn cook_quoted( | |
722 | &self, | |
723 | kind: token::LitKind, | |
724 | mode: Mode, | |
725 | start: BytePos, | |
726 | end: BytePos, | |
727 | prefix_len: u32, | |
728 | postfix_len: u32, | |
729 | ) -> (token::LitKind, Symbol) { | |
730 | self.cook_common(kind, mode, start, end, prefix_len, postfix_len, |src, mode, callback| { | |
731 | unescape::unescape_literal(src, mode, &mut |span, result| { | |
732 | callback(span, result.map(drop)) | |
733 | }) | |
734 | }) | |
735 | } | |
736 | ||
737 | fn cook_c_string( | |
738 | &self, | |
739 | kind: token::LitKind, | |
740 | mode: Mode, | |
741 | start: BytePos, | |
742 | end: BytePos, | |
743 | prefix_len: u32, | |
744 | postfix_len: u32, | |
745 | ) -> (token::LitKind, Symbol) { | |
746 | self.cook_common(kind, mode, start, end, prefix_len, postfix_len, |src, mode, callback| { | |
747 | unescape::unescape_c_string(src, mode, &mut |span, result| { | |
748 | callback(span, result.map(drop)) | |
749 | }) | |
750 | }) | |
751 | } | |
9cc50fc6 | 752 | } |
dfeec247 XL |
753 | |
754 | pub fn nfc_normalize(string: &str) -> Symbol { | |
755 | use unicode_normalization::{is_nfc_quick, IsNormalized, UnicodeNormalization}; | |
756 | match is_nfc_quick(string.chars()) { | |
757 | IsNormalized::Yes => Symbol::intern(string), | |
758 | _ => { | |
759 | let normalized_str: String = string.chars().nfc().collect(); | |
760 | Symbol::intern(&normalized_str) | |
761 | } | |
762 | } | |
763 | } |