2 pub use Nonterminal::*;
10 use crate::tokenstream::TokenTree;
12 use syntax_pos::symbol::Symbol;
13 use syntax_pos::{self, Span, DUMMY_SP};
17 #[cfg(target_arch = "x86_64")]
18 use rustc_data_structures::static_assert_size;
19 use rustc_data_structures::sync::Lrc;
21 #[derive(Clone, PartialEq, RustcEncodable, RustcDecodable, Hash, Debug, Copy)]
35 /// A delimiter token.
36 #[derive(Clone, PartialEq, Eq, RustcEncodable, RustcDecodable, Hash, Debug, Copy)]
38 /// A round parenthesis (i.e., `(` or `)`).
40 /// A square bracket (i.e., `[` or `]`).
42 /// A curly brace (i.e., `{` or `}`).
44 /// An empty delimiter.
49 pub fn len(self) -> usize {
50 if self == NoDelim { 0 } else { 1 }
53 pub fn is_empty(self) -> bool {
58 #[derive(Clone, Copy, PartialEq, RustcEncodable, RustcDecodable, Debug)]
60 Bool, // AST only, must never appear in a `Token`
66 StrRaw(u16), // raw string delimited by `n` hash symbols
68 ByteStrRaw(u16), // raw byte string delimited by `n` hash symbols
73 #[derive(Clone, Copy, PartialEq, RustcEncodable, RustcDecodable, Debug)]
77 pub suffix: Option<Symbol>,
80 impl fmt::Display for Lit {
81 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
82 let Lit { kind, symbol, suffix } = *self;
84 Byte => write!(f, "b'{}'", symbol)?,
85 Char => write!(f, "'{}'", symbol)?,
86 Str => write!(f, "\"{}\"", symbol)?,
87 StrRaw(n) => write!(f, "r{delim}\"{string}\"{delim}",
88 delim="#".repeat(n as usize),
90 ByteStr => write!(f, "b\"{}\"", symbol)?,
91 ByteStrRaw(n) => write!(f, "br{delim}\"{string}\"{delim}",
92 delim="#".repeat(n as usize),
97 Err => write!(f, "{}", symbol)?,
100 if let Some(suffix) = suffix {
101 write!(f, "{}", suffix)?;
109 /// An English article for the literal token kind.
110 crate fn article(self) -> &'static str {
112 Integer | Err => "an",
117 crate fn descr(self) -> &'static str {
119 Bool => panic!("literal token contains `Lit::Bool`"),
122 Integer => "integer",
124 Str | StrRaw(..) => "string",
125 ByteStr | ByteStrRaw(..) => "byte string",
130 crate fn may_have_suffix(self) -> bool {
132 Integer | Float | Err => true,
139 pub fn new(kind: LitKind, symbol: Symbol, suffix: Option<Symbol>) -> Lit {
140 Lit { kind, symbol, suffix }
144 pub(crate) fn ident_can_begin_expr(name: ast::Name, span: Span, is_raw: bool) -> bool {
145 let ident_token = Token::new(Ident(name, is_raw), span);
147 !ident_token.is_reserved_ident() ||
148 ident_token.is_path_segment_keyword() ||
152 // FIXME: remove when `await!(..)` syntax is removed
153 // https://github.com/rust-lang/rust/issues/60610
176 fn ident_can_begin_type(name: ast::Name, span: Span, is_raw: bool) -> bool {
177 let ident_token = Token::new(Ident(name, is_raw), span);
179 !ident_token.is_reserved_ident() ||
180 ident_token.is_path_segment_keyword() ||
193 #[derive(Clone, PartialEq, RustcEncodable, RustcDecodable, Debug)]
195 /* Expression-operator symbols. */
210 /* Structural symbols */
226 /// Used by proc macros for representing lifetimes, not generated by lexer right now.
228 /// An opening delimiter (e.g., `{`).
229 OpenDelim(DelimToken),
230 /// A closing delimiter (e.g., `}`).
231 CloseDelim(DelimToken),
236 /* Name components */
237 Ident(ast::Name, /* is_raw */ bool),
240 Interpolated(Lrc<Nonterminal>),
242 // Can be expanded into several tokens.
244 DocComment(ast::Name),
246 // Junk. These carry no data because we don't really care about the data
247 // they *would* carry, and don't really want to allocate a new ident for
248 // them. Instead, users could extract that from the associated span.
255 /// A completely invalid token which should be skipped.
261 // `TokenKind` is used a lot. Make sure it doesn't unintentionally get bigger.
262 #[cfg(target_arch = "x86_64")]
263 static_assert_size!(TokenKind, 16);
265 #[derive(Clone, PartialEq, RustcEncodable, RustcDecodable, Debug)]
272 pub fn lit(kind: LitKind, symbol: Symbol, suffix: Option<Symbol>) -> TokenKind {
273 Literal(Lit::new(kind, symbol, suffix))
276 /// Returns tokens that are likely to be typed accidentally instead of the current token.
277 /// Enables better error recovery when the wrong token is found.
278 crate fn similar_tokens(&self) -> Option<Vec<TokenKind>> {
280 Comma => Some(vec![Dot, Lt, Semi]),
281 Semi => Some(vec![Colon, Comma]),
288 pub fn new(kind: TokenKind, span: Span) -> Self {
292 /// Some token that will be thrown away later.
293 crate fn dummy() -> Self {
294 Token::new(TokenKind::Whitespace, DUMMY_SP)
297 /// Recovers a `Token` from an `ast::Ident`. This creates a raw identifier if necessary.
298 pub fn from_ast_ident(ident: ast::Ident) -> Self {
299 Token::new(Ident(ident.name, ident.is_raw_guess()), ident.span)
302 /// Return this token by value and leave a dummy token in its place.
303 pub fn take(&mut self) -> Self {
304 mem::replace(self, Token::dummy())
307 crate fn is_op(&self) -> bool {
309 OpenDelim(..) | CloseDelim(..) | Literal(..) | DocComment(..) |
310 Ident(..) | Lifetime(..) | Interpolated(..) |
311 Whitespace | Comment | Shebang(..) | Eof => false,
316 crate fn is_like_plus(&self) -> bool {
318 BinOp(Plus) | BinOpEq(Plus) => true,
323 /// Returns `true` if the token can appear at the start of an expression.
324 pub fn can_begin_expr(&self) -> bool {
326 Ident(name, is_raw) =>
327 ident_can_begin_expr(name, self.span, is_raw), // value name or keyword
328 OpenDelim(..) | // tuple, array or block
329 Literal(..) | // literal
330 Not | // operator not
331 BinOp(Minus) | // unary minus
332 BinOp(Star) | // dereference
333 BinOp(Or) | OrOr | // closure
334 BinOp(And) | // reference
335 AndAnd | // double reference
336 // DotDotDot is no longer supported, but we need some way to display the error
337 DotDot | DotDotDot | DotDotEq | // range notation
338 Lt | BinOp(Shl) | // associated path
339 ModSep | // global path
340 Lifetime(..) | // labeled loop
341 Pound => true, // expression attributes
342 Interpolated(ref nt) => match **nt {
348 NtLifetime(..) => true,
355 /// Returns `true` if the token can appear at the start of a type.
356 pub fn can_begin_type(&self) -> bool {
358 Ident(name, is_raw) =>
359 ident_can_begin_type(name, self.span, is_raw), // type name or keyword
360 OpenDelim(Paren) | // tuple
361 OpenDelim(Bracket) | // array
363 BinOp(Star) | // raw pointer
364 BinOp(And) | // reference
365 AndAnd | // double reference
366 Question | // maybe bound in trait object
367 Lifetime(..) | // lifetime bound in trait object
368 Lt | BinOp(Shl) | // associated path
369 ModSep => true, // global path
370 Interpolated(ref nt) => match **nt {
371 NtIdent(..) | NtTy(..) | NtPath(..) | NtLifetime(..) => true,
378 /// Returns `true` if the token can appear at the start of a const param.
379 crate fn can_begin_const_arg(&self) -> bool {
381 OpenDelim(Brace) => true,
382 Interpolated(ref nt) => match **nt {
385 NtLiteral(..) => true,
388 _ => self.can_begin_literal_or_bool(),
392 /// Returns `true` if the token can appear at the start of a generic bound.
393 crate fn can_begin_bound(&self) -> bool {
394 self.is_path_start() || self.is_lifetime() || self.is_keyword(kw::For) ||
395 self == &Question || self == &OpenDelim(Paren)
398 /// Returns `true` if the token is any literal
399 pub fn is_lit(&self) -> bool {
406 crate fn expect_lit(&self) -> Lit {
409 _ => panic!("`expect_lit` called on non-literal"),
413 /// Returns `true` if the token is any literal, a minus (which can prefix a literal,
414 /// for example a '-42', or one of the boolean idents).
415 pub fn can_begin_literal_or_bool(&self) -> bool {
417 Literal(..) | BinOp(Minus) => true,
418 Ident(name, false) if name.is_bool_lit() => true,
419 Interpolated(ref nt) => match **nt {
420 NtLiteral(..) => true,
427 /// Returns an identifier if this token is an identifier.
428 pub fn ident(&self) -> Option<(ast::Ident, /* is_raw */ bool)> {
430 Ident(name, is_raw) => Some((ast::Ident::new(name, self.span), is_raw)),
431 Interpolated(ref nt) => match **nt {
432 NtIdent(ident, is_raw) => Some((ident, is_raw)),
439 /// Returns a lifetime identifier if this token is a lifetime.
440 pub fn lifetime(&self) -> Option<ast::Ident> {
442 Lifetime(name) => Some(ast::Ident::new(name, self.span)),
443 Interpolated(ref nt) => match **nt {
444 NtLifetime(ident) => Some(ident),
451 /// Returns `true` if the token is an identifier.
452 pub fn is_ident(&self) -> bool {
453 self.ident().is_some()
456 /// Returns `true` if the token is a lifetime.
457 crate fn is_lifetime(&self) -> bool {
458 self.lifetime().is_some()
461 /// Returns `true` if the token is a identifier whose name is the given
463 crate fn is_ident_named(&self, name: Symbol) -> bool {
464 self.ident().map_or(false, |(ident, _)| ident.name == name)
467 /// Returns `true` if the token is an interpolated path.
468 fn is_path(&self) -> bool {
469 if let Interpolated(ref nt) = self.kind {
470 if let NtPath(..) = **nt {
477 /// Would `maybe_whole_expr` in `parser.rs` return `Ok(..)`?
478 /// That is, is this a pre-parsed expression dropped into the token stream
479 /// (which happens while parsing the result of macro expansion)?
480 crate fn is_whole_expr(&self) -> bool {
481 if let Interpolated(ref nt) = self.kind {
482 if let NtExpr(_) | NtLiteral(_) | NtPath(_) | NtIdent(..) | NtBlock(_) = **nt {
490 /// Returns `true` if the token is either the `mut` or `const` keyword.
491 crate fn is_mutability(&self) -> bool {
492 self.is_keyword(kw::Mut) ||
493 self.is_keyword(kw::Const)
496 crate fn is_qpath_start(&self) -> bool {
497 self == &Lt || self == &BinOp(Shl)
500 crate fn is_path_start(&self) -> bool {
501 self == &ModSep || self.is_qpath_start() || self.is_path() ||
502 self.is_path_segment_keyword() || self.is_ident() && !self.is_reserved_ident()
505 /// Returns `true` if the token is a given keyword, `kw`.
506 pub fn is_keyword(&self, kw: Symbol) -> bool {
507 self.is_non_raw_ident_where(|id| id.name == kw)
510 crate fn is_path_segment_keyword(&self) -> bool {
511 self.is_non_raw_ident_where(ast::Ident::is_path_segment_keyword)
514 // Returns true for reserved identifiers used internally for elided lifetimes,
515 // unnamed method parameters, crate root module, error recovery etc.
516 crate fn is_special_ident(&self) -> bool {
517 self.is_non_raw_ident_where(ast::Ident::is_special)
520 /// Returns `true` if the token is a keyword used in the language.
521 crate fn is_used_keyword(&self) -> bool {
522 self.is_non_raw_ident_where(ast::Ident::is_used_keyword)
525 /// Returns `true` if the token is a keyword reserved for possible future use.
526 crate fn is_unused_keyword(&self) -> bool {
527 self.is_non_raw_ident_where(ast::Ident::is_unused_keyword)
530 /// Returns `true` if the token is either a special identifier or a keyword.
531 pub fn is_reserved_ident(&self) -> bool {
532 self.is_non_raw_ident_where(ast::Ident::is_reserved)
535 /// Returns `true` if the token is the identifier `true` or `false`.
536 crate fn is_bool_lit(&self) -> bool {
537 self.is_non_raw_ident_where(|id| id.name.is_bool_lit())
540 /// Returns `true` if the token is a non-raw identifier for which `pred` holds.
541 fn is_non_raw_ident_where(&self, pred: impl FnOnce(ast::Ident) -> bool) -> bool {
543 Some((id, false)) => pred(id),
548 crate fn glue(&self, joint: &Token) -> Option<Token> {
549 let kind = match self.kind {
550 Eq => match joint.kind {
555 Lt => match joint.kind {
559 BinOp(Minus) => LArrow,
562 Gt => match joint.kind {
568 Not => match joint.kind {
572 BinOp(op) => match joint.kind {
574 BinOp(And) if op == And => AndAnd,
575 BinOp(Or) if op == Or => OrOr,
576 Gt if op == Minus => RArrow,
579 Dot => match joint.kind {
584 DotDot => match joint.kind {
589 Colon => match joint.kind {
593 SingleQuote => match joint.kind {
594 Ident(name, false) => Lifetime(Symbol::intern(&format!("'{}", name))),
598 Le | EqEq | Ne | Ge | AndAnd | OrOr | Tilde | BinOpEq(..) | At | DotDotDot |
599 DotDotEq | Comma | Semi | ModSep | RArrow | LArrow | FatArrow | Pound | Dollar |
600 Question | OpenDelim(..) | CloseDelim(..) |
601 Literal(..) | Ident(..) | Lifetime(..) | Interpolated(..) | DocComment(..) |
602 Whitespace | Comment | Shebang(..) | Unknown(..) | Eof => return None,
605 Some(Token::new(kind, self.span.to(joint.span)))
608 // See comments in `Nonterminal::to_tokenstream` for why we care about
609 // *probably* equal here rather than actual equality
610 crate fn probably_equal_for_proc_macro(&self, other: &Token) -> bool {
611 if mem::discriminant(&self.kind) != mem::discriminant(&other.kind) {
614 match (&self.kind, &other.kind) {
629 (&DotDotDot, &DotDotDot) |
630 (&DotDotEq, &DotDotEq) |
637 (&FatArrow, &FatArrow) |
640 (&Question, &Question) |
641 (&Whitespace, &Whitespace) |
642 (&Comment, &Comment) |
643 (&Eof, &Eof) => true,
645 (&BinOp(a), &BinOp(b)) |
646 (&BinOpEq(a), &BinOpEq(b)) => a == b,
648 (&OpenDelim(a), &OpenDelim(b)) |
649 (&CloseDelim(a), &CloseDelim(b)) => a == b,
651 (&DocComment(a), &DocComment(b)) |
652 (&Shebang(a), &Shebang(b)) => a == b,
654 (&Literal(a), &Literal(b)) => a == b,
656 (&Lifetime(a), &Lifetime(b)) => a == b,
657 (&Ident(a, b), &Ident(c, d)) => b == d && (a == c ||
658 a == kw::DollarCrate ||
659 c == kw::DollarCrate),
661 (&Interpolated(_), &Interpolated(_)) => false,
663 _ => panic!("forgot to add a token?"),
668 impl PartialEq<TokenKind> for Token {
669 fn eq(&self, rhs: &TokenKind) -> bool {
674 #[derive(Clone, RustcEncodable, RustcDecodable)]
675 /// For interpolation during macro expansion.
676 pub enum Nonterminal {
677 NtItem(P<ast::Item>),
678 NtBlock(P<ast::Block>),
681 NtExpr(P<ast::Expr>),
683 NtIdent(ast::Ident, /* is_raw */ bool),
684 NtLifetime(ast::Ident),
685 NtLiteral(P<ast::Expr>),
686 /// Stuff inside brackets for attributes
687 NtMeta(ast::AttrItem),
689 NtVis(ast::Visibility),
691 // Used only for passing items to proc macro attributes (they are not
692 // strictly necessary for that, `Annotatable` can be converted into
693 // tokens directly, but doing that naively regresses pretty-printing).
694 NtTraitItem(ast::TraitItem),
695 NtImplItem(ast::ImplItem),
696 NtForeignItem(ast::ForeignItem),
699 impl PartialEq for Nonterminal {
700 fn eq(&self, rhs: &Self) -> bool {
702 (NtIdent(ident_lhs, is_raw_lhs), NtIdent(ident_rhs, is_raw_rhs)) =>
703 ident_lhs == ident_rhs && is_raw_lhs == is_raw_rhs,
704 (NtLifetime(ident_lhs), NtLifetime(ident_rhs)) => ident_lhs == ident_rhs,
705 (NtTT(tt_lhs), NtTT(tt_rhs)) => tt_lhs == tt_rhs,
706 // FIXME: Assume that all "complex" nonterminal are not equal, we can't compare them
707 // correctly based on data from AST. This will prevent them from matching each other
708 // in macros. The comparison will become possible only when each nonterminal has an
709 // attached token stream from which it was parsed.
715 impl fmt::Debug for Nonterminal {
716 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
718 NtItem(..) => f.pad("NtItem(..)"),
719 NtBlock(..) => f.pad("NtBlock(..)"),
720 NtStmt(..) => f.pad("NtStmt(..)"),
721 NtPat(..) => f.pad("NtPat(..)"),
722 NtExpr(..) => f.pad("NtExpr(..)"),
723 NtTy(..) => f.pad("NtTy(..)"),
724 NtIdent(..) => f.pad("NtIdent(..)"),
725 NtLiteral(..) => f.pad("NtLiteral(..)"),
726 NtMeta(..) => f.pad("NtMeta(..)"),
727 NtPath(..) => f.pad("NtPath(..)"),
728 NtTT(..) => f.pad("NtTT(..)"),
729 NtImplItem(..) => f.pad("NtImplItem(..)"),
730 NtTraitItem(..) => f.pad("NtTraitItem(..)"),
731 NtForeignItem(..) => f.pad("NtForeignItem(..)"),
732 NtVis(..) => f.pad("NtVis(..)"),
733 NtLifetime(..) => f.pad("NtLifetime(..)"),