src/libsyntax/parse/lexer/mod.rs

   1 use crate::ast::{self, Ident};
   2 use crate::parse::{token, ParseSess};
   3 use crate::symbol::Symbol;
   4
   5 use errors::{Applicability, FatalError, Diagnostic, DiagnosticBuilder};
   6 use syntax_pos::{BytePos, Pos, Span, NO_EXPANSION};
   7 use core::unicode::property::Pattern_White_Space;
   8
   9 use std::borrow::Cow;
  10 use std::char;
  11 use std::iter;
  12 use std::mem::replace;
  13 use rustc_data_structures::sync::Lrc;
  14 use log::debug;
  15
  16 pub mod comments;
  17 mod tokentrees;
  18 mod unicode_chars;
  19
  20 #[derive(Clone, Debug)]
  21 pub struct TokenAndSpan {
  22     pub tok: token::Token,
  23     pub sp: Span,
  24 }
  25
  26 impl Default for TokenAndSpan {
  27     fn default() -> Self {
  28         TokenAndSpan {
  29             tok: token::Whitespace,
  30             sp: syntax_pos::DUMMY_SP,
  31         }
  32     }
  33 }
  34
  35 #[derive(Clone, Debug)]
  36 pub struct UnmatchedBrace {
  37     pub expected_delim: token::DelimToken,
  38     pub found_delim: token::DelimToken,
  39     pub found_span: Span,
  40     pub unclosed_span: Option<Span>,
  41     pub candidate_span: Option<Span>,
  42 }
  43
  44 pub struct StringReader<'a> {
  45     crate sess: &'a ParseSess,
  46     /// The absolute offset within the source_map of the next character to read
  47     crate next_pos: BytePos,
  48     /// The absolute offset within the source_map of the current character
  49     crate pos: BytePos,
  50     /// The current character (which has been read from self.pos)
  51     crate ch: Option<char>,
  52     crate source_file: Lrc<syntax_pos::SourceFile>,
  53     /// Stop reading src at this index.
  54     crate end_src_index: usize,
  55     // cached:
  56     peek_tok: token::Token,
  57     peek_span: Span,
  58     peek_span_src_raw: Span,
  59     fatal_errs: Vec<DiagnosticBuilder<'a>>,
  60     // cache a direct reference to the source text, so that we don't have to
  61     // retrieve it via `self.source_file.src.as_ref().unwrap()` all the time.
  62     src: Lrc<String>,
  63     token: token::Token,
  64     span: Span,
  65     /// The raw source span which *does not* take `override_span` into account
  66     span_src_raw: Span,
  67     /// Stack of open delimiters and their spans. Used for error message.
  68     open_braces: Vec<(token::DelimToken, Span)>,
  69     crate unmatched_braces: Vec<UnmatchedBrace>,
  70     /// The type and spans for all braces
  71     ///
  72     /// Used only for error recovery when arriving to EOF with mismatched braces.
  73     matching_delim_spans: Vec<(token::DelimToken, Span, Span)>,
  74     crate override_span: Option<Span>,
  75     last_unclosed_found_span: Option<Span>,
  76 }
  77
  78 impl<'a> StringReader<'a> {
  79     fn mk_sp(&self, lo: BytePos, hi: BytePos) -> Span {
  80         self.mk_sp_and_raw(lo, hi).0
  81     }
  82
  83     fn mk_sp_and_raw(&self, lo: BytePos, hi: BytePos) -> (Span, Span) {
  84         let raw = Span::new(lo, hi, NO_EXPANSION);
  85         let real = self.override_span.unwrap_or(raw);
  86
  87         (real, raw)
  88     }
  89
  90     fn mk_ident(&self, string: &str) -> Ident {
  91         let mut ident = Ident::from_str(string);
  92         if let Some(span) = self.override_span {
  93             ident.span = span;
  94         }
  95
  96         ident
  97     }
  98
  99     fn unwrap_or_abort(&mut self, res: Result<TokenAndSpan, ()>) -> TokenAndSpan {
 100         match res {
 101             Ok(tok) => tok,
 102             Err(_) => {
 103                 self.emit_fatal_errors();
 104                 FatalError.raise();
 105             }
 106         }
 107     }
 108
 109     fn next_token(&mut self) -> TokenAndSpan where Self: Sized {
 110         let res = self.try_next_token();
 111         self.unwrap_or_abort(res)
 112     }
 113
 114     /// Returns the next token. EFFECT: advances the string_reader.
 115     pub fn try_next_token(&mut self) -> Result<TokenAndSpan, ()> {
 116         assert!(self.fatal_errs.is_empty());
 117         let ret_val = TokenAndSpan {
 118             tok: replace(&mut self.peek_tok, token::Whitespace),
 119             sp: self.peek_span,
 120         };
 121         self.advance_token()?;
 122         self.span_src_raw = self.peek_span_src_raw;
 123
 124         Ok(ret_val)
 125     }
 126
 127     /// Immutably extract string if found at current position with given delimiters
 128     fn peek_delimited(&self, from_ch: char, to_ch: char) -> Option<String> {
 129         let mut pos = self.pos;
 130         let mut idx = self.src_index(pos);
 131         let mut ch = char_at(&self.src, idx);
 132         if ch != from_ch {
 133             return None;
 134         }
 135         pos = pos + Pos::from_usize(ch.len_utf8());
 136         let start_pos = pos;
 137         idx = self.src_index(pos);
 138         while idx < self.end_src_index {
 139             ch = char_at(&self.src, idx);
 140             if ch == to_ch {
 141                 return Some(self.src[self.src_index(start_pos)..self.src_index(pos)].to_string());
 142             }
 143             pos = pos + Pos::from_usize(ch.len_utf8());
 144             idx = self.src_index(pos);
 145         }
 146         return None;
 147     }
 148
 149     fn try_real_token(&mut self) -> Result<TokenAndSpan, ()> {
 150         let mut t = self.try_next_token()?;
 151         loop {
 152             match t.tok {
 153                 token::Whitespace | token::Comment | token::Shebang(_) => {
 154                     t = self.try_next_token()?;
 155                 }
 156                 _ => break,
 157             }
 158         }
 159
 160         self.token = t.tok.clone();
 161         self.span = t.sp;
 162
 163         Ok(t)
 164     }
 165
 166     pub fn real_token(&mut self) -> TokenAndSpan {
 167         let res = self.try_real_token();
 168         self.unwrap_or_abort(res)
 169     }
 170
 171     #[inline]
 172     fn is_eof(&self) -> bool {
 173         self.ch.is_none()
 174     }
 175
 176     fn fail_unterminated_raw_string(&self, pos: BytePos, hash_count: u16) {
 177         let mut err = self.struct_span_fatal(pos, pos, "unterminated raw string");
 178         err.span_label(self.mk_sp(pos, pos), "unterminated raw string");
 179
 180         if hash_count > 0 {
 181             err.note(&format!("this raw string should be terminated with `\"{}`",
 182                               "#".repeat(hash_count as usize)));
 183         }
 184
 185         err.emit();
 186         FatalError.raise();
 187     }
 188
 189     fn fatal(&self, m: &str) -> FatalError {
 190         self.fatal_span(self.peek_span, m)
 191     }
 192
 193     crate fn emit_fatal_errors(&mut self) {
 194         for err in &mut self.fatal_errs {
 195             err.emit();
 196         }
 197
 198         self.fatal_errs.clear();
 199     }
 200
 201     pub fn buffer_fatal_errors(&mut self) -> Vec<Diagnostic> {
 202         let mut buffer = Vec::new();
 203
 204         for err in self.fatal_errs.drain(..) {
 205             err.buffer(&mut buffer);
 206         }
 207
 208         buffer
 209     }
 210
 211     pub fn peek(&self) -> TokenAndSpan {
 212         // FIXME(pcwalton): Bad copy!
 213         TokenAndSpan {
 214             tok: self.peek_tok.clone(),
 215             sp: self.peek_span,
 216         }
 217     }
 218
 219     /// For comments.rs, which hackily pokes into next_pos and ch
 220     fn new_raw(sess: &'a ParseSess,
 221                source_file: Lrc<syntax_pos::SourceFile>,
 222                override_span: Option<Span>) -> Self {
 223         let mut sr = StringReader::new_raw_internal(sess, source_file, override_span);
 224         sr.bump();
 225
 226         sr
 227     }
 228
 229     fn new_raw_internal(sess: &'a ParseSess, source_file: Lrc<syntax_pos::SourceFile>,
 230         override_span: Option<Span>) -> Self
 231     {
 232         if source_file.src.is_none() {
 233             sess.span_diagnostic.bug(&format!("Cannot lex source_file without source: {}",
 234                                               source_file.name));
 235         }
 236
 237         let src = (*source_file.src.as_ref().unwrap()).clone();
 238
 239         StringReader {
 240             sess,
 241             next_pos: source_file.start_pos,
 242             pos: source_file.start_pos,
 243             ch: Some('\n'),
 244             source_file,
 245             end_src_index: src.len(),
 246             // dummy values; not read
 247             peek_tok: token::Eof,
 248             peek_span: syntax_pos::DUMMY_SP,
 249             peek_span_src_raw: syntax_pos::DUMMY_SP,
 250             src,
 251             fatal_errs: Vec::new(),
 252             token: token::Eof,
 253             span: syntax_pos::DUMMY_SP,
 254             span_src_raw: syntax_pos::DUMMY_SP,
 255             open_braces: Vec::new(),
 256             unmatched_braces: Vec::new(),
 257             matching_delim_spans: Vec::new(),
 258             override_span,
 259             last_unclosed_found_span: None,
 260         }
 261     }
 262
 263     pub fn new(sess: &'a ParseSess,
 264                source_file: Lrc<syntax_pos::SourceFile>,
 265                override_span: Option<Span>) -> Self {
 266         let mut sr = StringReader::new_raw(sess, source_file, override_span);
 267         if sr.advance_token().is_err() {
 268             sr.emit_fatal_errors();
 269             FatalError.raise();
 270         }
 271
 272         sr
 273     }
 274
 275     pub fn new_or_buffered_errs(sess: &'a ParseSess,
 276                                 source_file: Lrc<syntax_pos::SourceFile>,
 277                                 override_span: Option<Span>) -> Result<Self, Vec<Diagnostic>> {
 278         let mut sr = StringReader::new_raw(sess, source_file, override_span);
 279         if sr.advance_token().is_err() {
 280             Err(sr.buffer_fatal_errors())
 281         } else {
 282             Ok(sr)
 283         }
 284     }
 285
 286     pub fn retokenize(sess: &'a ParseSess, mut span: Span) -> Self {
 287         let begin = sess.source_map().lookup_byte_offset(span.lo());
 288         let end = sess.source_map().lookup_byte_offset(span.hi());
 289
 290         // Make the range zero-length if the span is invalid.
 291         if span.lo() > span.hi() || begin.sf.start_pos != end.sf.start_pos {
 292             span = span.shrink_to_lo();
 293         }
 294
 295         let mut sr = StringReader::new_raw_internal(sess, begin.sf, None);
 296
 297         // Seek the lexer to the right byte range.
 298         sr.next_pos = span.lo();
 299         sr.end_src_index = sr.src_index(span.hi());
 300
 301         sr.bump();
 302
 303         if sr.advance_token().is_err() {
 304             sr.emit_fatal_errors();
 305             FatalError.raise();
 306         }
 307
 308         sr
 309     }
 310
 311     #[inline]
 312     fn ch_is(&self, c: char) -> bool {
 313         self.ch == Some(c)
 314     }
 315
 316     /// Report a fatal lexical error with a given span.
 317     fn fatal_span(&self, sp: Span, m: &str) -> FatalError {
 318         self.sess.span_diagnostic.span_fatal(sp, m)
 319     }
 320
 321     /// Report a lexical error with a given span.
 322     fn err_span(&self, sp: Span, m: &str) {
 323         self.sess.span_diagnostic.struct_span_err(sp, m).emit();
 324     }
 325
 326
 327     /// Report a fatal error spanning [`from_pos`, `to_pos`).
 328     fn fatal_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str) -> FatalError {
 329         self.fatal_span(self.mk_sp(from_pos, to_pos), m)
 330     }
 331
 332     /// Report a lexical error spanning [`from_pos`, `to_pos`).
 333     fn err_span_(&self, from_pos: BytePos, to_pos: BytePos, m: &str) {
 334         self.err_span(self.mk_sp(from_pos, to_pos), m)
 335     }
 336
 337     /// Pushes a character to a message string for error reporting
 338     fn push_escaped_char_for_msg(m: &mut String, c: char) {
 339         match c {
 340             '\u{20}'..='\u{7e}' => {
 341                 // Don't escape \, ' or " for user-facing messages
 342                 m.push(c);
 343             }
 344             _ => {
 345                 m.extend(c.escape_default());
 346             }
 347         }
 348     }
 349
 350     /// Report a lexical error spanning [`from_pos`, `to_pos`), appending an
 351     /// escaped character to the error message
 352     fn fatal_span_char(&self, from_pos: BytePos, to_pos: BytePos, m: &str, c: char) -> FatalError {
 353         let mut m = m.to_string();
 354         m.push_str(": ");
 355         Self::push_escaped_char_for_msg(&mut m, c);
 356
 357         self.fatal_span_(from_pos, to_pos, &m[..])
 358     }
 359
 360     fn struct_span_fatal(&self, from_pos: BytePos, to_pos: BytePos, m: &str)
 361         -> DiagnosticBuilder<'a>
 362     {
 363         self.sess.span_diagnostic.struct_span_fatal(self.mk_sp(from_pos, to_pos), m)
 364     }
 365
 366     fn struct_fatal_span_char(&self, from_pos: BytePos, to_pos: BytePos, m: &str, c: char)
 367         -> DiagnosticBuilder<'a>
 368     {
 369         let mut m = m.to_string();
 370         m.push_str(": ");
 371         Self::push_escaped_char_for_msg(&mut m, c);
 372
 373         self.sess.span_diagnostic.struct_span_fatal(self.mk_sp(from_pos, to_pos), &m[..])
 374     }
 375
 376     /// Report a lexical error spanning [`from_pos`, `to_pos`), appending an
 377     /// escaped character to the error message
 378     fn err_span_char(&self, from_pos: BytePos, to_pos: BytePos, m: &str, c: char) {
 379         let mut m = m.to_string();
 380         m.push_str(": ");
 381         Self::push_escaped_char_for_msg(&mut m, c);
 382         self.err_span_(from_pos, to_pos, &m[..]);
 383     }
 384
 385     fn struct_err_span_char(&self, from_pos: BytePos, to_pos: BytePos, m: &str, c: char)
 386         -> DiagnosticBuilder<'a>
 387     {
 388         let mut m = m.to_string();
 389         m.push_str(": ");
 390         Self::push_escaped_char_for_msg(&mut m, c);
 391
 392         self.sess.span_diagnostic.struct_span_err(self.mk_sp(from_pos, to_pos), &m[..])
 393     }
 394
 395     /// Report a lexical error spanning [`from_pos`, `to_pos`), appending the
 396     /// offending string to the error message
 397     fn fatal_span_verbose(&self, from_pos: BytePos, to_pos: BytePos, mut m: String) -> FatalError {
 398         m.push_str(": ");
 399         m.push_str(&self.src[self.src_index(from_pos)..self.src_index(to_pos)]);
 400
 401         self.fatal_span_(from_pos, to_pos, &m[..])
 402     }
 403
 404     /// Advance peek_tok and peek_span to refer to the next token, and
 405     /// possibly update the interner.
 406     fn advance_token(&mut self) -> Result<(), ()> {
 407         match self.scan_whitespace_or_comment() {
 408             Some(comment) => {
 409                 self.peek_span_src_raw = comment.sp;
 410                 self.peek_span = comment.sp;
 411                 self.peek_tok = comment.tok;
 412             }
 413             None => {
 414                 if self.is_eof() {
 415                     self.peek_tok = token::Eof;
 416                     let (real, raw) = self.mk_sp_and_raw(
 417                         self.source_file.end_pos,
 418                         self.source_file.end_pos,
 419                     );
 420                     self.peek_span = real;
 421                     self.peek_span_src_raw = raw;
 422                 } else {
 423                     let start_bytepos = self.pos;
 424                     self.peek_tok = self.next_token_inner()?;
 425                     let (real, raw) = self.mk_sp_and_raw(start_bytepos, self.pos);
 426                     self.peek_span = real;
 427                     self.peek_span_src_raw = raw;
 428                 };
 429             }
 430         }
 431
 432         Ok(())
 433     }
 434
 435     #[inline]
 436     fn src_index(&self, pos: BytePos) -> usize {
 437         (pos - self.source_file.start_pos).to_usize()
 438     }
 439
 440     /// Calls `f` with a string slice of the source text spanning from `start`
 441     /// up to but excluding `self.pos`, meaning the slice does not include
 442     /// the character `self.ch`.
 443     fn with_str_from<T, F>(&self, start: BytePos, f: F) -> T
 444         where F: FnOnce(&str) -> T
 445     {
 446         self.with_str_from_to(start, self.pos, f)
 447     }
 448
 449     /// Creates a Name from a given offset to the current offset, each
 450     /// adjusted 1 towards each other (assumes that on either side there is a
 451     /// single-byte delimiter).
 452     fn name_from(&self, start: BytePos) -> ast::Name {
 453         debug!("taking an ident from {:?} to {:?}", start, self.pos);
 454         self.with_str_from(start, Symbol::intern)
 455     }
 456
 457     /// As name_from, with an explicit endpoint.
 458     fn name_from_to(&self, start: BytePos, end: BytePos) -> ast::Name {
 459         debug!("taking an ident from {:?} to {:?}", start, end);
 460         self.with_str_from_to(start, end, Symbol::intern)
 461     }
 462
 463     /// Calls `f` with a string slice of the source text spanning from `start`
 464     /// up to but excluding `end`.
 465     fn with_str_from_to<T, F>(&self, start: BytePos, end: BytePos, f: F) -> T
 466         where F: FnOnce(&str) -> T
 467     {
 468         f(&self.src[self.src_index(start)..self.src_index(end)])
 469     }
 470
 471     /// Converts CRLF to LF in the given string, raising an error on bare CR.
 472     fn translate_crlf<'b>(&self, start: BytePos, s: &'b str, errmsg: &'b str) -> Cow<'b, str> {
 473         let mut chars = s.char_indices().peekable();
 474         while let Some((i, ch)) = chars.next() {
 475             if ch == '\r' {
 476                 if let Some((lf_idx, '\n')) = chars.peek() {
 477                     return translate_crlf_(self, start, s, *lf_idx, chars, errmsg).into();
 478                 }
 479                 let pos = start + BytePos(i as u32);
 480                 let end_pos = start + BytePos((i + ch.len_utf8()) as u32);
 481                 self.err_span_(pos, end_pos, errmsg);
 482             }
 483         }
 484         return s.into();
 485
 486         fn translate_crlf_(rdr: &StringReader<'_>,
 487                            start: BytePos,
 488                            s: &str,
 489                            mut j: usize,
 490                            mut chars: iter::Peekable<impl Iterator<Item = (usize, char)>>,
 491                            errmsg: &str)
 492                            -> String {
 493             let mut buf = String::with_capacity(s.len());
 494             // Skip first CR
 495             buf.push_str(&s[.. j - 1]);
 496             while let Some((i, ch)) = chars.next() {
 497                 if ch == '\r' {
 498                     if j < i {
 499                         buf.push_str(&s[j..i]);
 500                     }
 501                     let next = i + ch.len_utf8();
 502                     j = next;
 503                     if chars.peek().map(|(_, ch)| *ch) != Some('\n') {
 504                         let pos = start + BytePos(i as u32);
 505                         let end_pos = start + BytePos(next as u32);
 506                         rdr.err_span_(pos, end_pos, errmsg);
 507                     }
 508                 }
 509             }
 510             if j < s.len() {
 511                 buf.push_str(&s[j..]);
 512             }
 513             buf
 514         }
 515     }
 516
 517     /// Advance the StringReader by one character.
 518     crate fn bump(&mut self) {
 519         let next_src_index = self.src_index(self.next_pos);
 520         if next_src_index < self.end_src_index {
 521             let next_ch = char_at(&self.src, next_src_index);
 522             let next_ch_len = next_ch.len_utf8();
 523
 524             self.ch = Some(next_ch);
 525             self.pos = self.next_pos;
 526             self.next_pos = self.next_pos + Pos::from_usize(next_ch_len);
 527         } else {
 528             self.ch = None;
 529             self.pos = self.next_pos;
 530         }
 531     }
 532
 533     fn nextch(&self) -> Option<char> {
 534         let next_src_index = self.src_index(self.next_pos);
 535         if next_src_index < self.end_src_index {
 536             Some(char_at(&self.src, next_src_index))
 537         } else {
 538             None
 539         }
 540     }
 541
 542     #[inline]
 543     fn nextch_is(&self, c: char) -> bool {
 544         self.nextch() == Some(c)
 545     }
 546
 547     fn nextnextch(&self) -> Option<char> {
 548         let next_src_index = self.src_index(self.next_pos);
 549         if next_src_index < self.end_src_index {
 550             let next_next_src_index =
 551                 next_src_index + char_at(&self.src, next_src_index).len_utf8();
 552             if next_next_src_index < self.end_src_index {
 553                 return Some(char_at(&self.src, next_next_src_index));
 554             }
 555         }
 556         None
 557     }
 558
 559     #[inline]
 560     fn nextnextch_is(&self, c: char) -> bool {
 561         self.nextnextch() == Some(c)
 562     }
 563
 564     /// Eats <XID_start><XID_continue>*, if possible.
 565     fn scan_optional_raw_name(&mut self) -> Option<ast::Name> {
 566         if !ident_start(self.ch) {
 567             return None;
 568         }
 569
 570         let start = self.pos;
 571         self.bump();
 572
 573         while ident_continue(self.ch) {
 574             self.bump();
 575         }
 576
 577         self.with_str_from(start, |string| {
 578             if string == "_" {
 579                 self.sess.span_diagnostic
 580                     .struct_span_warn(self.mk_sp(start, self.pos),
 581                                       "underscore literal suffix is not allowed")
 582                     .warn("this was previously accepted by the compiler but is \
 583                           being phased out; it will become a hard error in \
 584                           a future release!")
 585                     .note("for more information, see issue #42326 \
 586                           <https://github.com/rust-lang/rust/issues/42326>")
 587                     .emit();
 588                 None
 589             } else {
 590                 Some(Symbol::intern(string))
 591             }
 592         })
 593     }
 594
 595     /// PRECONDITION: self.ch is not whitespace
 596     /// Eats any kind of comment.
 597     fn scan_comment(&mut self) -> Option<TokenAndSpan> {
 598         if let Some(c) = self.ch {
 599             if c.is_whitespace() {
 600                 let msg = "called consume_any_line_comment, but there was whitespace";
 601                 self.sess.span_diagnostic.span_err(self.mk_sp(self.pos, self.pos), msg);
 602             }
 603         }
 604
 605         if self.ch_is('/') {
 606             match self.nextch() {
 607                 Some('/') => {
 608                     self.bump();
 609                     self.bump();
 610
 611                     // line comments starting with "///" or "//!" are doc-comments
 612                     let doc_comment = (self.ch_is('/') && !self.nextch_is('/')) || self.ch_is('!');
 613                     let start_bpos = self.pos - BytePos(2);
 614
 615                     while !self.is_eof() {
 616                         match self.ch.unwrap() {
 617                             '\n' => break,
 618                             '\r' => {
 619                                 if self.nextch_is('\n') {
 620                                     // CRLF
 621                                     break;
 622                                 } else if doc_comment {
 623                                     self.err_span_(self.pos,
 624                                                    self.next_pos,
 625                                                    "bare CR not allowed in doc-comment");
 626                                 }
 627                             }
 628                             _ => (),
 629                         }
 630                         self.bump();
 631                     }
 632
 633                     let tok = if doc_comment {
 634                         self.with_str_from(start_bpos, |string| {
 635                             token::DocComment(Symbol::intern(string))
 636                         })
 637                     } else {
 638                         token::Comment
 639                     };
 640                     Some(TokenAndSpan { tok, sp: self.mk_sp(start_bpos, self.pos) })
 641                 }
 642                 Some('*') => {
 643                     self.bump();
 644                     self.bump();
 645                     self.scan_block_comment()
 646                 }
 647                 _ => None,
 648             }
 649         } else if self.ch_is('#') {
 650             if self.nextch_is('!') {
 651
 652                 // Parse an inner attribute.
 653                 if self.nextnextch_is('[') {
 654                     return None;
 655                 }
 656
 657                 let is_beginning_of_file = self.pos == self.source_file.start_pos;
 658                 if is_beginning_of_file {
 659                     debug!("Skipping a shebang");
 660                     let start = self.pos;
 661                     while !self.ch_is('\n') && !self.is_eof() {
 662                         self.bump();
 663                     }
 664                     return Some(TokenAndSpan {
 665                         tok: token::Shebang(self.name_from(start)),
 666                         sp: self.mk_sp(start, self.pos),
 667                     });
 668                 }
 669             }
 670             None
 671         } else {
 672             None
 673         }
 674     }
 675
 676     /// If there is whitespace, shebang, or a comment, scan it. Otherwise,
 677     /// return `None`.
 678     fn scan_whitespace_or_comment(&mut self) -> Option<TokenAndSpan> {
 679         match self.ch.unwrap_or('\0') {
 680             // # to handle shebang at start of file -- this is the entry point
 681             // for skipping over all "junk"
 682             '/' | '#' => {
 683                 let c = self.scan_comment();
 684                 debug!("scanning a comment {:?}", c);
 685                 c
 686             },
 687             c if is_pattern_whitespace(Some(c)) => {
 688                 let start_bpos = self.pos;
 689                 while is_pattern_whitespace(self.ch) {
 690                     self.bump();
 691                 }
 692                 let c = Some(TokenAndSpan {
 693                     tok: token::Whitespace,
 694                     sp: self.mk_sp(start_bpos, self.pos),
 695                 });
 696                 debug!("scanning whitespace: {:?}", c);
 697                 c
 698             }
 699             _ => None,
 700         }
 701     }
 702
 703     /// Might return a sugared-doc-attr
 704     fn scan_block_comment(&mut self) -> Option<TokenAndSpan> {
 705         // block comments starting with "/**" or "/*!" are doc-comments
 706         let is_doc_comment = self.ch_is('*') || self.ch_is('!');
 707         let start_bpos = self.pos - BytePos(2);
 708
 709         let mut level: isize = 1;
 710         let mut has_cr = false;
 711         while level > 0 {
 712             if self.is_eof() {
 713                 let msg = if is_doc_comment {
 714                     "unterminated block doc-comment"
 715                 } else {
 716                     "unterminated block comment"
 717                 };
 718                 let last_bpos = self.pos;
 719                 self.fatal_span_(start_bpos, last_bpos, msg).raise();
 720             }
 721             let n = self.ch.unwrap();
 722             match n {
 723                 '/' if self.nextch_is('*') => {
 724                     level += 1;
 725                     self.bump();
 726                 }
 727                 '*' if self.nextch_is('/') => {
 728                     level -= 1;
 729                     self.bump();
 730                 }
 731                 '\r' => {
 732                     has_cr = true;
 733                 }
 734                 _ => (),
 735             }
 736             self.bump();
 737         }
 738
 739         self.with_str_from(start_bpos, |string| {
 740             // but comments with only "*"s between two "/"s are not
 741             let tok = if is_block_doc_comment(string) {
 742                 let string = if has_cr {
 743                     self.translate_crlf(start_bpos,
 744                                         string,
 745                                         "bare CR not allowed in block doc-comment")
 746                 } else {
 747                     string.into()
 748                 };
 749                 token::DocComment(Symbol::intern(&string[..]))
 750             } else {
 751                 token::Comment
 752             };
 753
 754             Some(TokenAndSpan {
 755                 tok,
 756                 sp: self.mk_sp(start_bpos, self.pos),
 757             })
 758         })
 759     }
 760
 761     /// Scan through any digits (base `scan_radix`) or underscores,
 762     /// and return how many digits there were.
 763     ///
 764     /// `real_radix` represents the true radix of the number we're
 765     /// interested in, and errors will be emitted for any digits
 766     /// between `real_radix` and `scan_radix`.
 767     fn scan_digits(&mut self, real_radix: u32, scan_radix: u32) -> usize {
 768         assert!(real_radix <= scan_radix);
 769         let mut len = 0;
 770
 771         loop {
 772             let c = self.ch;
 773             if c == Some('_') {
 774                 debug!("skipping a _");
 775                 self.bump();
 776                 continue;
 777             }
 778             match c.and_then(|cc| cc.to_digit(scan_radix)) {
 779                 Some(_) => {
 780                     debug!("{:?} in scan_digits", c);
 781                     // check that the hypothetical digit is actually
 782                     // in range for the true radix
 783                     if c.unwrap().to_digit(real_radix).is_none() {
 784                         self.err_span_(self.pos,
 785                                        self.next_pos,
 786                                        &format!("invalid digit for a base {} literal", real_radix));
 787                     }
 788                     len += 1;
 789                     self.bump();
 790                 }
 791                 _ => return len,
 792             }
 793         }
 794     }
 795
 796     /// Lex a LIT_INTEGER or a LIT_FLOAT
 797     fn scan_number(&mut self, c: char) -> token::Lit {
 798         let mut base = 10;
 799         let start_bpos = self.pos;
 800         self.bump();
 801
 802         let num_digits = if c == '0' {
 803             match self.ch.unwrap_or('\0') {
 804                 'b' => {
 805                     self.bump();
 806                     base = 2;
 807                     self.scan_digits(2, 10)
 808                 }
 809                 'o' => {
 810                     self.bump();
 811                     base = 8;
 812                     self.scan_digits(8, 10)
 813                 }
 814                 'x' => {
 815                     self.bump();
 816                     base = 16;
 817                     self.scan_digits(16, 16)
 818                 }
 819                 '0'..='9' | '_' | '.' | 'e' | 'E' => {
 820                     self.scan_digits(10, 10) + 1
 821                 }
 822                 _ => {
 823                     // just a 0
 824                     return token::Integer(self.name_from(start_bpos));
 825                 }
 826             }
 827         } else if c.is_digit(10) {
 828             self.scan_digits(10, 10) + 1
 829         } else {
 830             0
 831         };
 832
 833         if num_digits == 0 {
 834             self.err_span_(start_bpos, self.pos, "no valid digits found for number");
 835
 836             return token::Integer(Symbol::intern("0"));
 837         }
 838
 839         // might be a float, but don't be greedy if this is actually an
 840         // integer literal followed by field/method access or a range pattern
 841         // (`0..2` and `12.foo()`)
 842         if self.ch_is('.') && !self.nextch_is('.') &&
 843            !ident_start(self.nextch()) {
 844             // might have stuff after the ., and if it does, it needs to start
 845             // with a number
 846             self.bump();
 847             if self.ch.unwrap_or('\0').is_digit(10) {
 848                 self.scan_digits(10, 10);
 849                 self.scan_float_exponent();
 850             }
 851             let pos = self.pos;
 852             self.check_float_base(start_bpos, pos, base);
 853
 854             token::Float(self.name_from(start_bpos))
 855         } else {
 856             // it might be a float if it has an exponent
 857             if self.ch_is('e') || self.ch_is('E') {
 858                 self.scan_float_exponent();
 859                 let pos = self.pos;
 860                 self.check_float_base(start_bpos, pos, base);
 861                 return token::Float(self.name_from(start_bpos));
 862             }
 863             // but we certainly have an integer!
 864             token::Integer(self.name_from(start_bpos))
 865         }
 866     }
 867
 868     /// Scan over `n_digits` hex digits, stopping at `delim`, reporting an
 869     /// error if too many or too few digits are encountered.
 870     fn scan_hex_digits(&mut self, n_digits: usize, delim: char, below_0x7f_only: bool) -> bool {
 871         debug!("scanning {} digits until {:?}", n_digits, delim);
 872         let start_bpos = self.pos;
 873         let mut accum_int = 0;
 874
 875         let mut valid = true;
 876         for _ in 0..n_digits {
 877             if self.is_eof() {
 878                 let last_bpos = self.pos;
 879                 self.fatal_span_(start_bpos,
 880                                  last_bpos,
 881                                  "unterminated numeric character escape").raise();
 882             }
 883             if self.ch_is(delim) {
 884                 let last_bpos = self.pos;
 885                 self.err_span_(start_bpos,
 886                                last_bpos,
 887                                "numeric character escape is too short");
 888                 valid = false;
 889                 break;
 890             }
 891             let c = self.ch.unwrap_or('\x00');
 892             accum_int *= 16;
 893             accum_int += c.to_digit(16).unwrap_or_else(|| {
 894                 self.err_span_char(self.pos,
 895                                    self.next_pos,
 896                                    "invalid character in numeric character escape",
 897                                    c);
 898
 899                 valid = false;
 900                 0
 901             });
 902             self.bump();
 903         }
 904
 905         if below_0x7f_only && accum_int >= 0x80 {
 906             self.err_span_(start_bpos,
 907                            self.pos,
 908                            "this form of character escape may only be used with characters in \
 909                             the range [\\x00-\\x7f]");
 910             valid = false;
 911         }
 912
 913         match char::from_u32(accum_int) {
 914             Some(_) => valid,
 915             None => {
 916                 let last_bpos = self.pos;
 917                 self.err_span_(start_bpos, last_bpos, "invalid numeric character escape");
 918                 false
 919             }
 920         }
 921     }
 922
 923     /// Scan for a single (possibly escaped) byte or char
 924     /// in a byte, (non-raw) byte string, char, or (non-raw) string literal.
 925     /// `start` is the position of `first_source_char`, which is already consumed.
 926     ///
 927     /// Returns `true` if there was a valid char/byte.
 928     fn scan_char_or_byte(&mut self,
 929                          start: BytePos,
 930                          first_source_char: char,
 931                          ascii_only: bool,
 932                          delim: char)
 933                          -> bool
 934     {
 935         match first_source_char {
 936             '\\' => {
 937                 // '\X' for some X must be a character constant:
 938                 let escaped = self.ch;
 939                 let escaped_pos = self.pos;
 940                 self.bump();
 941                 match escaped {
 942                     None => {}  // EOF here is an error that will be checked later.
 943                     Some(e) => {
 944                         return match e {
 945                             'n' | 'r' | 't' | '\\' | '\'' | '"' | '0' => true,
 946                             'x' => self.scan_byte_escape(delim, !ascii_only),
 947                             'u' => {
 948                                 let valid = if self.ch_is('{') {
 949                                     self.scan_unicode_escape(delim) && !ascii_only
 950                                 } else {
 951                                     let span = self.mk_sp(start, self.pos);
 952                                     let mut suggestion = "\\u{".to_owned();
 953                                     let msg = "incorrect unicode escape sequence";
 954                                     let mut err = self.sess.span_diagnostic.struct_span_err(
 955                                         span,
 956                                         msg,
 957                                     );
 958                                     let mut i = 0;
 959                                     while let (Some(ch), true) = (self.ch, i < 6) {
 960                                         if ch.is_digit(16) {
 961                                             suggestion.push(ch);
 962                                             self.bump();
 963                                             i += 1;
 964                                         } else {
 965                                             break;
 966                                         }
 967                                     }
 968                                     if i != 0 {
 969                                         suggestion.push('}');
 970                                         err.span_suggestion(
 971                                             self.mk_sp(start, self.pos),
 972                                             "format of unicode escape sequences uses braces",
 973                                             suggestion,
 974                                             Applicability::MaybeIncorrect,
 975                                         );
 976                                     } else {
 977                                         err.span_label(span, msg);
 978                                         err.help(
 979                                             "format of unicode escape sequences is `\\u{...}`",
 980                                         );
 981                                     }
 982                                     err.emit();
 983                                     false
 984                                 };
 985                                 if ascii_only {
 986                                     self.err_span_(start,
 987                                                    self.pos,
 988                                                    "unicode escape sequences cannot be used as a \
 989                                                     byte or in a byte string");
 990                                 }
 991                                 valid
 992
 993                             }
 994                             '\n' if delim == '"' => {
 995                                 self.consume_whitespace();
 996                                 true
 997                             }
 998                             '\r' if delim == '"' && self.ch_is('\n') => {
 999                                 self.consume_whitespace();
1000                                 true
1001                             }
1002                             c => {
1003                                 let pos = self.pos;
1004                                 let msg = if ascii_only {
1005                                     "unknown byte escape"
1006                                 } else {
1007                                     "unknown character escape"
1008                                 };
1009                                 let mut err = self.struct_err_span_char(escaped_pos, pos, msg, c);
1010                                 err.span_label(self.mk_sp(escaped_pos, pos), msg);
1011                                 if e == '\r' {
1012                                     err.help(
1013                                         "this is an isolated carriage return; consider checking \
1014                                          your editor and version control settings",
1015                                     );
1016                                 }
1017                                 if (e == '{' || e == '}') && !ascii_only {
1018                                     err.help(
1019                                         "if used in a formatting string, curly braces are escaped \
1020                                          with `{{` and `}}`",
1021                                     );
1022                                 }
1023                                 err.emit();
1024                                 false
1025                             }
1026                         }
1027                     }
1028                 }
1029             }
1030             '\t' | '\n' | '\r' | '\'' if delim == '\'' => {
1031                 let pos = self.pos;
1032                 self.err_span_char(start,
1033                                    pos,
1034                                    if ascii_only {
1035                                        "byte constant must be escaped"
1036                                    } else {
1037                                        "character constant must be escaped"
1038                                    },
1039                                    first_source_char);
1040                 return false;
1041             }
1042             '\r' => {
1043                 if self.ch_is('\n') {
1044                     self.bump();
1045                     return true;
1046                 } else {
1047                     self.err_span_(start,
1048                                    self.pos,
1049                                    "bare CR not allowed in string, use \\r instead");
1050                     return false;
1051                 }
1052             }
1053             _ => {
1054                 if ascii_only && first_source_char > '\x7F' {
1055                     let pos = self.pos;
1056                     self.err_span_(start,
1057                                    pos,
1058                                    "byte constant must be ASCII. Use a \\xHH escape for a \
1059                                     non-ASCII byte");
1060                     return false;
1061                 }
1062             }
1063         }
1064         true
1065     }
1066
1067     /// Scan over a `\u{...}` escape
1068     ///
1069     /// At this point, we have already seen the `\` and the `u`, the `{` is the current character.
1070     /// We will read a hex number (with `_` separators), with 1 to 6 actual digits,
1071     /// and pass over the `}`.
1072     fn scan_unicode_escape(&mut self, delim: char) -> bool {
1073         self.bump(); // past the {
1074         let start_bpos = self.pos;
1075         let mut valid = true;
1076
1077         if let Some('_') = self.ch {
1078             // disallow leading `_`
1079             self.err_span_(self.pos,
1080                            self.next_pos,
1081                            "invalid start of unicode escape");
1082             valid = false;
1083         }
1084
1085         let count = self.scan_digits(16, 16);
1086
1087         if count > 6 {
1088             self.err_span_(start_bpos,
1089                            self.pos,
1090                            "overlong unicode escape (must have at most 6 hex digits)");
1091             valid = false;
1092         }
1093
1094         loop {
1095             match self.ch {
1096                 Some('}') => {
1097                     if valid && count == 0 {
1098                         self.err_span_(start_bpos,
1099                                        self.pos,
1100                                        "empty unicode escape (must have at least 1 hex digit)");
1101                         valid = false;
1102                     }
1103                     self.bump(); // past the ending `}`
1104                     break;
1105                 },
1106                 Some(c) => {
1107                     if c == delim {
1108                         self.err_span_(self.pos,
1109                                        self.pos,
1110                                        "unterminated unicode escape (needed a `}`)");
1111                         valid = false;
1112                         break;
1113                     } else if valid {
1114                         self.err_span_char(start_bpos,
1115                                            self.pos,
1116                                            "invalid character in unicode escape",
1117                                            c);
1118                         valid = false;
1119                     }
1120                 },
1121                 None => {
1122                     self.fatal_span_(start_bpos,
1123                                      self.pos,
1124                                      "unterminated unicode escape (found EOF)").raise();
1125                 }
1126             }
1127             self.bump();
1128         }
1129
1130         valid
1131     }
1132
1133     /// Scan over a float exponent.
1134     fn scan_float_exponent(&mut self) {
1135         if self.ch_is('e') || self.ch_is('E') {
1136             self.bump();
1137
1138             if self.ch_is('-') || self.ch_is('+') {
1139                 self.bump();
1140             }
1141
1142             if self.scan_digits(10, 10) == 0 {
1143                 let mut err = self.struct_span_fatal(
1144                     self.pos, self.next_pos,
1145                     "expected at least one digit in exponent"
1146                 );
1147                 if let Some(ch) = self.ch {
1148                     // check for e.g., Unicode minus '−' (Issue #49746)
1149                     if unicode_chars::check_for_substitution(self, ch, &mut err) {
1150                         self.bump();
1151                         self.scan_digits(10, 10);
1152                     }
1153                 }
1154                 err.emit();
1155             }
1156         }
1157     }
1158
1159     /// Checks that a base is valid for a floating literal, emitting a nice
1160     /// error if it isn't.
1161     fn check_float_base(&mut self, start_bpos: BytePos, last_bpos: BytePos, base: usize) {
1162         match base {
1163             16 => {
1164                 self.err_span_(start_bpos,
1165                                last_bpos,
1166                                "hexadecimal float literal is not supported")
1167             }
1168             8 => {
1169                 self.err_span_(start_bpos,
1170                                last_bpos,
1171                                "octal float literal is not supported")
1172             }
1173             2 => {
1174                 self.err_span_(start_bpos,
1175                                last_bpos,
1176                                "binary float literal is not supported")
1177             }
1178             _ => (),
1179         }
1180     }
1181
1182     fn binop(&mut self, op: token::BinOpToken) -> token::Token {
1183         self.bump();
1184         if self.ch_is('=') {
1185             self.bump();
1186             token::BinOpEq(op)
1187         } else {
1188             token::BinOp(op)
1189         }
1190     }
1191
1192     /// Returns the next token from the string, advances the input past that
1193     /// token, and updates the interner
1194     fn next_token_inner(&mut self) -> Result<token::Token, ()> {
1195         let c = self.ch;
1196
1197         if ident_start(c) {
1198             let (is_ident_start, is_raw_ident) =
1199                 match (c.unwrap(), self.nextch(), self.nextnextch()) {
1200                     // r# followed by an identifier starter is a raw identifier.
1201                     // This is an exception to the r# case below.
1202                     ('r', Some('#'), x) if ident_start(x) => (true, true),
1203                     // r as in r" or r#" is part of a raw string literal.
1204                     // b as in b' is part of a byte literal.
1205                     // They are not identifiers, and are handled further down.
1206                     ('r', Some('"'), _) |
1207                     ('r', Some('#'), _) |
1208                     ('b', Some('"'), _) |
1209                     ('b', Some('\''), _) |
1210                     ('b', Some('r'), Some('"')) |
1211                     ('b', Some('r'), Some('#')) => (false, false),
1212                     _ => (true, false),
1213                 };
1214
1215             if is_ident_start {
1216                 let raw_start = self.pos;
1217                 if is_raw_ident {
1218                     // Consume the 'r#' characters.
1219                     self.bump();
1220                     self.bump();
1221                 }
1222
1223                 let start = self.pos;
1224                 self.bump();
1225
1226                 while ident_continue(self.ch) {
1227                     self.bump();
1228                 }
1229
1230                 return Ok(self.with_str_from(start, |string| {
1231                     // FIXME: perform NFKC normalization here. (Issue #2253)
1232                     let ident = self.mk_ident(string);
1233
1234                     if is_raw_ident {
1235                         let span = self.mk_sp(raw_start, self.pos);
1236                         if !ident.can_be_raw() {
1237                             self.err_span(span, &format!("`{}` cannot be a raw identifier", ident));
1238                         }
1239                         self.sess.raw_identifier_spans.borrow_mut().push(span);
1240                     }
1241
1242                     token::Ident(ident, is_raw_ident)
1243                 }));
1244             }
1245         }
1246
1247         if is_dec_digit(c) {
1248             let num = self.scan_number(c.unwrap());
1249             let suffix = self.scan_optional_raw_name();
1250             debug!("next_token_inner: scanned number {:?}, {:?}", num, suffix);
1251             return Ok(token::Literal(num, suffix));
1252         }
1253
1254         match c.expect("next_token_inner called at EOF") {
1255             // One-byte tokens.
1256             ';' => {
1257                 self.bump();
1258                 Ok(token::Semi)
1259             }
1260             ',' => {
1261                 self.bump();
1262                 Ok(token::Comma)
1263             }
1264             '.' => {
1265                 self.bump();
1266                 if self.ch_is('.') {
1267                     self.bump();
1268                     if self.ch_is('.') {
1269                         self.bump();
1270                         Ok(token::DotDotDot)
1271                     } else if self.ch_is('=') {
1272                         self.bump();
1273                         Ok(token::DotDotEq)
1274                     } else {
1275                         Ok(token::DotDot)
1276                     }
1277                 } else {
1278                     Ok(token::Dot)
1279                 }
1280             }
1281             '(' => {
1282                 self.bump();
1283                 Ok(token::OpenDelim(token::Paren))
1284             }
1285             ')' => {
1286                 self.bump();
1287                 Ok(token::CloseDelim(token::Paren))
1288             }
1289             '{' => {
1290                 self.bump();
1291                 Ok(token::OpenDelim(token::Brace))
1292             }
1293             '}' => {
1294                 self.bump();
1295                 Ok(token::CloseDelim(token::Brace))
1296             }
1297             '[' => {
1298                 self.bump();
1299                 Ok(token::OpenDelim(token::Bracket))
1300             }
1301             ']' => {
1302                 self.bump();
1303                 Ok(token::CloseDelim(token::Bracket))
1304             }
1305             '@' => {
1306                 self.bump();
1307                 Ok(token::At)
1308             }
1309             '#' => {
1310                 self.bump();
1311                 Ok(token::Pound)
1312             }
1313             '~' => {
1314                 self.bump();
1315                 Ok(token::Tilde)
1316             }
1317             '?' => {
1318                 self.bump();
1319                 Ok(token::Question)
1320             }
1321             ':' => {
1322                 self.bump();
1323                 if self.ch_is(':') {
1324                     self.bump();
1325                     Ok(token::ModSep)
1326                 } else {
1327                     Ok(token::Colon)
1328                 }
1329             }
1330
1331             '$' => {
1332                 self.bump();
1333                 Ok(token::Dollar)
1334             }
1335
1336             // Multi-byte tokens.
1337             '=' => {
1338                 self.bump();
1339                 if self.ch_is('=') {
1340                     self.bump();
1341                     Ok(token::EqEq)
1342                 } else if self.ch_is('>') {
1343                     self.bump();
1344                     Ok(token::FatArrow)
1345                 } else {
1346                     Ok(token::Eq)
1347                 }
1348             }
1349             '!' => {
1350                 self.bump();
1351                 if self.ch_is('=') {
1352                     self.bump();
1353                     Ok(token::Ne)
1354                 } else {
1355                     Ok(token::Not)
1356                 }
1357             }
1358             '<' => {
1359                 self.bump();
1360                 match self.ch.unwrap_or('\x00') {
1361                     '=' => {
1362                         self.bump();
1363                         Ok(token::Le)
1364                     }
1365                     '<' => {
1366                         Ok(self.binop(token::Shl))
1367                     }
1368                     '-' => {
1369                         self.bump();
1370                         Ok(token::LArrow)
1371                     }
1372                     _ => {
1373                         Ok(token::Lt)
1374                     }
1375                 }
1376             }
1377             '>' => {
1378                 self.bump();
1379                 match self.ch.unwrap_or('\x00') {
1380                     '=' => {
1381                         self.bump();
1382                         Ok(token::Ge)
1383                     }
1384                     '>' => {
1385                         Ok(self.binop(token::Shr))
1386                     }
1387                     _ => {
1388                         Ok(token::Gt)
1389                     }
1390                 }
1391             }
1392             '\'' => {
1393                 // Either a character constant 'a' OR a lifetime name 'abc
1394                 let start_with_quote = self.pos;
1395                 self.bump();
1396                 let start = self.pos;
1397
1398                 // the eof will be picked up by the final `'` check below
1399                 let c2 = self.ch.unwrap_or('\x00');
1400                 self.bump();
1401
1402                 // If the character is an ident start not followed by another single
1403                 // quote, then this is a lifetime name:
1404                 if (ident_start(Some(c2)) || c2.is_numeric()) && !self.ch_is('\'') {
1405                     while ident_continue(self.ch) {
1406                         self.bump();
1407                     }
1408                     // lifetimes shouldn't end with a single quote
1409                     // if we find one, then this is an invalid character literal
1410                     if self.ch_is('\'') {
1411                         self.err_span_(
1412                             start_with_quote,
1413                             self.next_pos,
1414                             "character literal may only contain one codepoint");
1415                         self.bump();
1416                         return Ok(token::Literal(token::Err(Symbol::intern("??")), None))
1417
1418                     }
1419
1420                     // Include the leading `'` in the real identifier, for macro
1421                     // expansion purposes. See #12512 for the gory details of why
1422                     // this is necessary.
1423                     let ident = self.with_str_from(start, |lifetime_name| {
1424                         self.mk_ident(&format!("'{}", lifetime_name))
1425                     });
1426
1427                     if c2.is_numeric() {
1428                         // this is a recovered lifetime written `'1`, error but accept it
1429                         self.err_span_(
1430                             start_with_quote,
1431                             self.pos,
1432                             "lifetimes cannot start with a number",
1433                         );
1434                     }
1435
1436                     return Ok(token::Lifetime(ident));
1437                 }
1438
1439                 let valid = self.scan_char_or_byte(start, c2, /* ascii_only */ false, '\'');
1440
1441                 if !self.ch_is('\'') {
1442                     let pos = self.pos;
1443
1444                     loop {
1445                         self.bump();
1446                         if self.ch_is('\'') {
1447                             let start = self.src_index(start);
1448                             let end = self.src_index(self.pos);
1449                             self.bump();
1450                             let span = self.mk_sp(start_with_quote, self.pos);
1451                             self.sess.span_diagnostic
1452                                 .struct_span_err(span,
1453                                                  "character literal may only contain one codepoint")
1454                                 .span_suggestion(
1455                                     span,
1456                                     "if you meant to write a `str` literal, use double quotes",
1457                                     format!("\"{}\"", &self.src[start..end]),
1458                                     Applicability::MachineApplicable
1459                                 ).emit();
1460                             return Ok(token::Literal(token::Err(Symbol::intern("??")), None))
1461                         }
1462                         if self.ch_is('\n') || self.is_eof() || self.ch_is('/') {
1463                             // Only attempt to infer single line string literals. If we encounter
1464                             // a slash, bail out in order to avoid nonsensical suggestion when
1465                             // involving comments.
1466                             break;
1467                         }
1468                     }
1469
1470                     self.fatal_span_verbose(start_with_quote, pos,
1471                         String::from("character literal may only contain one codepoint")).raise();
1472                 }
1473
1474                 let id = if valid {
1475                     self.name_from(start)
1476                 } else {
1477                     Symbol::intern("0")
1478                 };
1479
1480                 self.bump(); // advance ch past token
1481                 let suffix = self.scan_optional_raw_name();
1482
1483                 Ok(token::Literal(token::Char(id), suffix))
1484             }
1485             'b' => {
1486                 self.bump();
1487                 let lit = match self.ch {
1488                     Some('\'') => self.scan_byte(),
1489                     Some('"') => self.scan_byte_string(),
1490                     Some('r') => self.scan_raw_byte_string(),
1491                     _ => unreachable!(),  // Should have been a token::Ident above.
1492                 };
1493                 let suffix = self.scan_optional_raw_name();
1494
1495                 Ok(token::Literal(lit, suffix))
1496             }
1497             '"' => {
1498                 let start_bpos = self.pos;
1499                 let mut valid = true;
1500                 self.bump();
1501
1502                 while !self.ch_is('"') {
1503                     if self.is_eof() {
1504                         let last_bpos = self.pos;
1505                         self.fatal_span_(start_bpos,
1506                                          last_bpos,
1507                                          "unterminated double quote string").raise();
1508                     }
1509
1510                     let ch_start = self.pos;
1511                     let ch = self.ch.unwrap();
1512                     self.bump();
1513                     valid &= self.scan_char_or_byte(ch_start, ch, /* ascii_only */ false, '"');
1514                 }
1515                 // adjust for the ASCII " at the start of the literal
1516                 let id = if valid {
1517                     self.name_from(start_bpos + BytePos(1))
1518                 } else {
1519                     Symbol::intern("??")
1520                 };
1521                 self.bump();
1522                 let suffix = self.scan_optional_raw_name();
1523
1524                 Ok(token::Literal(token::Str_(id), suffix))
1525             }
1526             'r' => {
1527                 let start_bpos = self.pos;
1528                 self.bump();
1529                 let mut hash_count: u16 = 0;
1530                 while self.ch_is('#') {
1531                     if hash_count == 65535 {
1532                         let bpos = self.next_pos;
1533                         self.fatal_span_(start_bpos,
1534                                          bpos,
1535                                          "too many `#` symbols: raw strings may be \
1536                                          delimited by up to 65535 `#` symbols").raise();
1537                     }
1538                     self.bump();
1539                     hash_count += 1;
1540                 }
1541
1542                 if self.is_eof() {
1543                     self.fail_unterminated_raw_string(start_bpos, hash_count);
1544                 } else if !self.ch_is('"') {
1545                     let last_bpos = self.pos;
1546                     let curr_char = self.ch.unwrap();
1547                     self.fatal_span_char(start_bpos,
1548                                          last_bpos,
1549                                          "found invalid character; only `#` is allowed \
1550                                          in raw string delimitation",
1551                                          curr_char).raise();
1552                 }
1553                 self.bump();
1554                 let content_start_bpos = self.pos;
1555                 let mut content_end_bpos;
1556                 let mut valid = true;
1557                 'outer: loop {
1558                     if self.is_eof() {
1559                         self.fail_unterminated_raw_string(start_bpos, hash_count);
1560                     }
1561                     // if self.ch_is('"') {
1562                     // content_end_bpos = self.pos;
1563                     // for _ in 0..hash_count {
1564                     // self.bump();
1565                     // if !self.ch_is('#') {
1566                     // continue 'outer;
1567                     let c = self.ch.unwrap();
1568                     match c {
1569                         '"' => {
1570                             content_end_bpos = self.pos;
1571                             for _ in 0..hash_count {
1572                                 self.bump();
1573                                 if !self.ch_is('#') {
1574                                     continue 'outer;
1575                                 }
1576                             }
1577                             break;
1578                         }
1579                         '\r' => {
1580                             if !self.nextch_is('\n') {
1581                                 let last_bpos = self.pos;
1582                                 self.err_span_(start_bpos,
1583                                                last_bpos,
1584                                                "bare CR not allowed in raw string, use \\r \
1585                                                 instead");
1586                                 valid = false;
1587                             }
1588                         }
1589                         _ => (),
1590                     }
1591                     self.bump();
1592                 }
1593
1594                 self.bump();
1595                 let id = if valid {
1596                     self.name_from_to(content_start_bpos, content_end_bpos)
1597                 } else {
1598                     Symbol::intern("??")
1599                 };
1600                 let suffix = self.scan_optional_raw_name();
1601
1602                 Ok(token::Literal(token::StrRaw(id, hash_count), suffix))
1603             }
1604             '-' => {
1605                 if self.nextch_is('>') {
1606                     self.bump();
1607                     self.bump();
1608                     Ok(token::RArrow)
1609                 } else {
1610                     Ok(self.binop(token::Minus))
1611                 }
1612             }
1613             '&' => {
1614                 if self.nextch_is('&') {
1615                     self.bump();
1616                     self.bump();
1617                     Ok(token::AndAnd)
1618                 } else {
1619                     Ok(self.binop(token::And))
1620                 }
1621             }
1622             '|' => {
1623                 match self.nextch() {
1624                     Some('|') => {
1625                         self.bump();
1626                         self.bump();
1627                         Ok(token::OrOr)
1628                     }
1629                     _ => {
1630                         Ok(self.binop(token::Or))
1631                     }
1632                 }
1633             }
1634             '+' => {
1635                 Ok(self.binop(token::Plus))
1636             }
1637             '*' => {
1638                 Ok(self.binop(token::Star))
1639             }
1640             '/' => {
1641                 Ok(self.binop(token::Slash))
1642             }
1643             '^' => {
1644                 Ok(self.binop(token::Caret))
1645             }
1646             '%' => {
1647                 Ok(self.binop(token::Percent))
1648             }
1649             c => {
1650                 let last_bpos = self.pos;
1651                 let bpos = self.next_pos;
1652                 let mut err = self.struct_fatal_span_char(last_bpos,
1653                                                           bpos,
1654                                                           "unknown start of token",
1655                                                           c);
1656                 unicode_chars::check_for_substitution(self, c, &mut err);
1657                 self.fatal_errs.push(err);
1658
1659                 Err(())
1660             }
1661         }
1662     }
1663
1664     fn consume_whitespace(&mut self) {
1665         while is_pattern_whitespace(self.ch) && !self.is_eof() {
1666             self.bump();
1667         }
1668     }
1669
1670     fn read_to_eol(&mut self) -> String {
1671         let mut val = String::new();
1672         while !self.ch_is('\n') && !self.is_eof() {
1673             val.push(self.ch.unwrap());
1674             self.bump();
1675         }
1676
1677         if self.ch_is('\n') {
1678             self.bump();
1679         }
1680
1681         val
1682     }
1683
1684     fn read_one_line_comment(&mut self) -> String {
1685         let val = self.read_to_eol();
1686         assert!((val.as_bytes()[0] == b'/' && val.as_bytes()[1] == b'/') ||
1687                 (val.as_bytes()[0] == b'#' && val.as_bytes()[1] == b'!'));
1688         val
1689     }
1690
1691     fn consume_non_eol_whitespace(&mut self) {
1692         while is_pattern_whitespace(self.ch) && !self.ch_is('\n') && !self.is_eof() {
1693             self.bump();
1694         }
1695     }
1696
1697     fn peeking_at_comment(&self) -> bool {
1698         (self.ch_is('/') && self.nextch_is('/')) || (self.ch_is('/') && self.nextch_is('*')) ||
1699         // consider shebangs comments, but not inner attributes
1700         (self.ch_is('#') && self.nextch_is('!') && !self.nextnextch_is('['))
1701     }
1702
1703     fn scan_byte(&mut self) -> token::Lit {
1704         self.bump();
1705         let start = self.pos;
1706
1707         // the eof will be picked up by the final `'` check below
1708         let c2 = self.ch.unwrap_or('\x00');
1709         self.bump();
1710
1711         let valid = self.scan_char_or_byte(start,
1712                                            c2,
1713                                            // ascii_only =
1714                                            true,
1715                                            '\'');
1716         if !self.ch_is('\'') {
1717             // Byte offsetting here is okay because the
1718             // character before position `start` are an
1719             // ascii single quote and ascii 'b'.
1720             let pos = self.pos;
1721             self.fatal_span_verbose(start - BytePos(2),
1722                                     pos,
1723                                     "unterminated byte constant".to_string()).raise();
1724         }
1725
1726         let id = if valid {
1727             self.name_from(start)
1728         } else {
1729             Symbol::intern("?")
1730         };
1731         self.bump(); // advance ch past token
1732
1733         token::Byte(id)
1734     }
1735
1736     #[inline]
1737     fn scan_byte_escape(&mut self, delim: char, below_0x7f_only: bool) -> bool {
1738         self.scan_hex_digits(2, delim, below_0x7f_only)
1739     }
1740
1741     fn scan_byte_string(&mut self) -> token::Lit {
1742         self.bump();
1743         let start = self.pos;
1744         let mut valid = true;
1745
1746         while !self.ch_is('"') {
1747             if self.is_eof() {
1748                 let pos = self.pos;
1749                 self.fatal_span_(start, pos, "unterminated double quote byte string").raise();
1750             }
1751
1752             let ch_start = self.pos;
1753             let ch = self.ch.unwrap();
1754             self.bump();
1755             valid &= self.scan_char_or_byte(ch_start,
1756                                             ch,
1757                                             // ascii_only =
1758                                             true,
1759                                             '"');
1760         }
1761
1762         let id = if valid {
1763             self.name_from(start)
1764         } else {
1765             Symbol::intern("??")
1766         };
1767         self.bump();
1768
1769         token::ByteStr(id)
1770     }
1771
1772     fn scan_raw_byte_string(&mut self) -> token::Lit {
1773         let start_bpos = self.pos;
1774         self.bump();
1775         let mut hash_count = 0;
1776         while self.ch_is('#') {
1777             if hash_count == 65535 {
1778                 let bpos = self.next_pos;
1779                 self.fatal_span_(start_bpos,
1780                                  bpos,
1781                                  "too many `#` symbols: raw byte strings may be \
1782                                  delimited by up to 65535 `#` symbols").raise();
1783             }
1784             self.bump();
1785             hash_count += 1;
1786         }
1787
1788         if self.is_eof() {
1789             self.fail_unterminated_raw_string(start_bpos, hash_count);
1790         } else if !self.ch_is('"') {
1791             let pos = self.pos;
1792             let ch = self.ch.unwrap();
1793             self.fatal_span_char(start_bpos,
1794                                         pos,
1795                                         "found invalid character; only `#` is allowed in raw \
1796                                          string delimitation",
1797                                         ch).raise();
1798         }
1799         self.bump();
1800         let content_start_bpos = self.pos;
1801         let mut content_end_bpos;
1802         'outer: loop {
1803             match self.ch {
1804                 None => {
1805                     self.fail_unterminated_raw_string(start_bpos, hash_count);
1806                 }
1807                 Some('"') => {
1808                     content_end_bpos = self.pos;
1809                     for _ in 0..hash_count {
1810                         self.bump();
1811                         if !self.ch_is('#') {
1812                             continue 'outer;
1813                         }
1814                     }
1815                     break;
1816                 }
1817                 Some(c) => {
1818                     if c > '\x7F' {
1819                         let pos = self.pos;
1820                         self.err_span_char(pos, pos, "raw byte string must be ASCII", c);
1821                     }
1822                 }
1823             }
1824             self.bump();
1825         }
1826
1827         self.bump();
1828
1829         token::ByteStrRaw(self.name_from_to(content_start_bpos, content_end_bpos), hash_count)
1830     }
1831 }
1832
1833 // This tests the character for the unicode property 'PATTERN_WHITE_SPACE' which
1834 // is guaranteed to be forward compatible. http://unicode.org/reports/tr31/#R3
1835 #[inline]
1836 crate fn is_pattern_whitespace(c: Option<char>) -> bool {
1837     c.map_or(false, Pattern_White_Space)
1838 }
1839
1840 #[inline]
1841 fn in_range(c: Option<char>, lo: char, hi: char) -> bool {
1842     c.map_or(false, |c| lo <= c && c <= hi)
1843 }
1844
1845 #[inline]
1846 fn is_dec_digit(c: Option<char>) -> bool {
1847     in_range(c, '0', '9')
1848 }
1849
1850 fn is_doc_comment(s: &str) -> bool {
1851     let res = (s.starts_with("///") && *s.as_bytes().get(3).unwrap_or(&b' ') != b'/') ||
1852               s.starts_with("//!");
1853     debug!("is {:?} a doc comment? {}", s, res);
1854     res
1855 }
1856
1857 fn is_block_doc_comment(s: &str) -> bool {
1858     // Prevent `/**/` from being parsed as a doc comment
1859     let res = ((s.starts_with("/**") && *s.as_bytes().get(3).unwrap_or(&b' ') != b'*') ||
1860                s.starts_with("/*!")) && s.len() >= 5;
1861     debug!("is {:?} a doc comment? {}", s, res);
1862     res
1863 }
1864
1865 /// Determine whether `c` is a valid start for an ident.
1866 fn ident_start(c: Option<char>) -> bool {
1867     let c = match c {
1868         Some(c) => c,
1869         None => return false,
1870     };
1871
1872     (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || c == '_' || (c > '\x7f' && c.is_xid_start())
1873 }
1874
1875 fn ident_continue(c: Option<char>) -> bool {
1876     let c = match c {
1877         Some(c) => c,
1878         None => return false,
1879     };
1880
1881     (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') || (c >= '0' && c <= '9') || c == '_' ||
1882     (c > '\x7f' && c.is_xid_continue())
1883 }
1884
1885 #[inline]
1886 fn char_at(s: &str, byte: usize) -> char {
1887     s[byte..].chars().next().unwrap()
1888 }
1889
1890 #[cfg(test)]
1891 mod tests {
1892     use super::*;
1893
1894     use crate::ast::{Ident, CrateConfig};
1895     use crate::symbol::Symbol;
1896     use crate::source_map::{SourceMap, FilePathMapping};
1897     use crate::feature_gate::UnstableFeatures;
1898     use crate::parse::token;
1899     use crate::diagnostics::plugin::ErrorMap;
1900     use crate::with_globals;
1901     use std::io;
1902     use std::path::PathBuf;
1903     use syntax_pos::{BytePos, Span, NO_EXPANSION};
1904     use rustc_data_structures::fx::FxHashSet;
1905     use rustc_data_structures::sync::Lock;
1906
1907     fn mk_sess(sm: Lrc<SourceMap>) -> ParseSess {
1908         let emitter = errors::emitter::EmitterWriter::new(Box::new(io::sink()),
1909                                                           Some(sm.clone()),
1910                                                           false,
1911                                                           false,
1912                                                           false);
1913         ParseSess {
1914             span_diagnostic: errors::Handler::with_emitter(true, None, Box::new(emitter)),
1915             unstable_features: UnstableFeatures::from_environment(),
1916             config: CrateConfig::default(),
1917             included_mod_stack: Lock::new(Vec::new()),
1918             source_map: sm,
1919             missing_fragment_specifiers: Lock::new(FxHashSet::default()),
1920             raw_identifier_spans: Lock::new(Vec::new()),
1921             registered_diagnostics: Lock::new(ErrorMap::new()),
1922             buffered_lints: Lock::new(vec![]),
1923         }
1924     }
1925
1926     // open a string reader for the given string
1927     fn setup<'a>(sm: &SourceMap,
1928                  sess: &'a ParseSess,
1929                  teststr: String)
1930                  -> StringReader<'a> {
1931         let sf = sm.new_source_file(PathBuf::from(teststr.clone()).into(), teststr);
1932         StringReader::new(sess, sf, None)
1933     }
1934
1935     #[test]
1936     fn t1() {
1937         with_globals(|| {
1938             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
1939             let sh = mk_sess(sm.clone());
1940             let mut string_reader = setup(&sm,
1941                                         &sh,
1942                                         "/* my source file */ fn main() { println!(\"zebra\"); }\n"
1943                                             .to_string());
1944             let id = Ident::from_str("fn");
1945             assert_eq!(string_reader.next_token().tok, token::Comment);
1946             assert_eq!(string_reader.next_token().tok, token::Whitespace);
1947             let tok1 = string_reader.next_token();
1948             let tok2 = TokenAndSpan {
1949                 tok: token::Ident(id, false),
1950                 sp: Span::new(BytePos(21), BytePos(23), NO_EXPANSION),
1951             };
1952             assert_eq!(tok1.tok, tok2.tok);
1953             assert_eq!(tok1.sp, tok2.sp);
1954             assert_eq!(string_reader.next_token().tok, token::Whitespace);
1955             // the 'main' id is already read:
1956             assert_eq!(string_reader.pos.clone(), BytePos(28));
1957             // read another token:
1958             let tok3 = string_reader.next_token();
1959             let tok4 = TokenAndSpan {
1960                 tok: mk_ident("main"),
1961                 sp: Span::new(BytePos(24), BytePos(28), NO_EXPANSION),
1962             };
1963             assert_eq!(tok3.tok, tok4.tok);
1964             assert_eq!(tok3.sp, tok4.sp);
1965             // the lparen is already read:
1966             assert_eq!(string_reader.pos.clone(), BytePos(29))
1967         })
1968     }
1969
1970     // check that the given reader produces the desired stream
1971     // of tokens (stop checking after exhausting the expected vec)
1972     fn check_tokenization(mut string_reader: StringReader<'_>, expected: Vec<token::Token>) {
1973         for expected_tok in &expected {
1974             assert_eq!(&string_reader.next_token().tok, expected_tok);
1975         }
1976     }
1977
1978     // make the identifier by looking up the string in the interner
1979     fn mk_ident(id: &str) -> token::Token {
1980         token::Token::from_ast_ident(Ident::from_str(id))
1981     }
1982
1983     #[test]
1984     fn doublecolonparsing() {
1985         with_globals(|| {
1986             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
1987             let sh = mk_sess(sm.clone());
1988             check_tokenization(setup(&sm, &sh, "a b".to_string()),
1989                             vec![mk_ident("a"), token::Whitespace, mk_ident("b")]);
1990         })
1991     }
1992
1993     #[test]
1994     fn dcparsing_2() {
1995         with_globals(|| {
1996             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
1997             let sh = mk_sess(sm.clone());
1998             check_tokenization(setup(&sm, &sh, "a::b".to_string()),
1999                             vec![mk_ident("a"), token::ModSep, mk_ident("b")]);
2000         })
2001     }
2002
2003     #[test]
2004     fn dcparsing_3() {
2005         with_globals(|| {
2006             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2007             let sh = mk_sess(sm.clone());
2008             check_tokenization(setup(&sm, &sh, "a ::b".to_string()),
2009                             vec![mk_ident("a"), token::Whitespace, token::ModSep, mk_ident("b")]);
2010         })
2011     }
2012
2013     #[test]
2014     fn dcparsing_4() {
2015         with_globals(|| {
2016             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2017             let sh = mk_sess(sm.clone());
2018             check_tokenization(setup(&sm, &sh, "a:: b".to_string()),
2019                             vec![mk_ident("a"), token::ModSep, token::Whitespace, mk_ident("b")]);
2020         })
2021     }
2022
2023     #[test]
2024     fn character_a() {
2025         with_globals(|| {
2026             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2027             let sh = mk_sess(sm.clone());
2028             assert_eq!(setup(&sm, &sh, "'a'".to_string()).next_token().tok,
2029                     token::Literal(token::Char(Symbol::intern("a")), None));
2030         })
2031     }
2032
2033     #[test]
2034     fn character_space() {
2035         with_globals(|| {
2036             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2037             let sh = mk_sess(sm.clone());
2038             assert_eq!(setup(&sm, &sh, "' '".to_string()).next_token().tok,
2039                     token::Literal(token::Char(Symbol::intern(" ")), None));
2040         })
2041     }
2042
2043     #[test]
2044     fn character_escaped() {
2045         with_globals(|| {
2046             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2047             let sh = mk_sess(sm.clone());
2048             assert_eq!(setup(&sm, &sh, "'\\n'".to_string()).next_token().tok,
2049                     token::Literal(token::Char(Symbol::intern("\\n")), None));
2050         })
2051     }
2052
2053     #[test]
2054     fn lifetime_name() {
2055         with_globals(|| {
2056             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2057             let sh = mk_sess(sm.clone());
2058             assert_eq!(setup(&sm, &sh, "'abc".to_string()).next_token().tok,
2059                     token::Lifetime(Ident::from_str("'abc")));
2060         })
2061     }
2062
2063     #[test]
2064     fn raw_string() {
2065         with_globals(|| {
2066             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2067             let sh = mk_sess(sm.clone());
2068             assert_eq!(setup(&sm, &sh, "r###\"\"#a\\b\x00c\"\"###".to_string())
2069                         .next_token()
2070                         .tok,
2071                     token::Literal(token::StrRaw(Symbol::intern("\"#a\\b\x00c\""), 3), None));
2072         })
2073     }
2074
2075     #[test]
2076     fn literal_suffixes() {
2077         with_globals(|| {
2078             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2079             let sh = mk_sess(sm.clone());
2080             macro_rules! test {
2081                 ($input: expr, $tok_type: ident, $tok_contents: expr) => {{
2082                     assert_eq!(setup(&sm, &sh, format!("{}suffix", $input)).next_token().tok,
2083                             token::Literal(token::$tok_type(Symbol::intern($tok_contents)),
2084                                             Some(Symbol::intern("suffix"))));
2085                     // with a whitespace separator:
2086                     assert_eq!(setup(&sm, &sh, format!("{} suffix", $input)).next_token().tok,
2087                             token::Literal(token::$tok_type(Symbol::intern($tok_contents)),
2088                                             None));
2089                 }}
2090             }
2091
2092             test!("'a'", Char, "a");
2093             test!("b'a'", Byte, "a");
2094             test!("\"a\"", Str_, "a");
2095             test!("b\"a\"", ByteStr, "a");
2096             test!("1234", Integer, "1234");
2097             test!("0b101", Integer, "0b101");
2098             test!("0xABC", Integer, "0xABC");
2099             test!("1.0", Float, "1.0");
2100             test!("1.0e10", Float, "1.0e10");
2101
2102             assert_eq!(setup(&sm, &sh, "2us".to_string()).next_token().tok,
2103                     token::Literal(token::Integer(Symbol::intern("2")),
2104                                     Some(Symbol::intern("us"))));
2105             assert_eq!(setup(&sm, &sh, "r###\"raw\"###suffix".to_string()).next_token().tok,
2106                     token::Literal(token::StrRaw(Symbol::intern("raw"), 3),
2107                                     Some(Symbol::intern("suffix"))));
2108             assert_eq!(setup(&sm, &sh, "br###\"raw\"###suffix".to_string()).next_token().tok,
2109                     token::Literal(token::ByteStrRaw(Symbol::intern("raw"), 3),
2110                                     Some(Symbol::intern("suffix"))));
2111         })
2112     }
2113
2114     #[test]
2115     fn line_doc_comments() {
2116         assert!(is_doc_comment("///"));
2117         assert!(is_doc_comment("/// blah"));
2118         assert!(!is_doc_comment("////"));
2119     }
2120
2121     #[test]
2122     fn nested_block_comments() {
2123         with_globals(|| {
2124             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2125             let sh = mk_sess(sm.clone());
2126             let mut lexer = setup(&sm, &sh, "/* /* */ */'a'".to_string());
2127             match lexer.next_token().tok {
2128                 token::Comment => {}
2129                 _ => panic!("expected a comment!"),
2130             }
2131             assert_eq!(lexer.next_token().tok,
2132                     token::Literal(token::Char(Symbol::intern("a")), None));
2133         })
2134     }
2135
2136     #[test]
2137     fn crlf_comments() {
2138         with_globals(|| {
2139             let sm = Lrc::new(SourceMap::new(FilePathMapping::empty()));
2140             let sh = mk_sess(sm.clone());
2141             let mut lexer = setup(&sm, &sh, "// test\r\n/// test\r\n".to_string());
2142             let comment = lexer.next_token();
2143             assert_eq!(comment.tok, token::Comment);
2144             assert_eq!((comment.sp.lo(), comment.sp.hi()), (BytePos(0), BytePos(7)));
2145             assert_eq!(lexer.next_token().tok, token::Whitespace);
2146             assert_eq!(lexer.next_token().tok,
2147                     token::DocComment(Symbol::intern("/// test")));
2148         })
2149     }
2150 }