diff options
Diffstat (limited to 'src/librustc_parse/parser/mod.rs')
| -rw-r--r-- | src/librustc_parse/parser/mod.rs | 1393 |
1 files changed, 1393 insertions, 0 deletions
diff --git a/src/librustc_parse/parser/mod.rs b/src/librustc_parse/parser/mod.rs new file mode 100644 index 00000000000..a491d91e20f --- /dev/null +++ b/src/librustc_parse/parser/mod.rs @@ -0,0 +1,1393 @@ +pub mod attr; +mod expr; +mod pat; +mod item; +mod module; +mod ty; +mod path; +pub use path::PathStyle; +mod stmt; +mod generics; +mod diagnostics; +use diagnostics::Error; + +use crate::{Directory, DirectoryOwnership}; +use crate::lexer::UnmatchedBrace; + +use syntax::ast::{ + self, Abi, DUMMY_NODE_ID, AttrStyle, Attribute, CrateSugar, Ident, + IsAsync, MacDelimiter, Mutability, StrStyle, Visibility, VisibilityKind, Unsafety, +}; + +use syntax::print::pprust; +use syntax::ptr::P; +use syntax::token::{self, Token, TokenKind, DelimToken}; +use syntax::tokenstream::{self, DelimSpan, TokenTree, TokenStream, TreeAndJoint}; +use syntax::sess::ParseSess; +use syntax::source_map::respan; +use syntax::struct_span_err; +use syntax::util::comments::{doc_comment_style, strip_doc_comment_decoration}; +use syntax_pos::symbol::{kw, sym, Symbol}; +use syntax_pos::{Span, BytePos, DUMMY_SP, FileName}; +use rustc_data_structures::thin_vec::ThinVec; +use errors::{PResult, Applicability, DiagnosticBuilder, DiagnosticId, FatalError}; +use log::debug; + +use std::borrow::Cow; +use std::{cmp, mem, slice}; +use std::path::PathBuf; + +bitflags::bitflags! { + struct Restrictions: u8 { + const STMT_EXPR = 1 << 0; + const NO_STRUCT_LITERAL = 1 << 1; + } +} + +#[derive(Clone, Copy, PartialEq, Debug)] +enum SemiColonMode { + Break, + Ignore, + Comma, +} + +#[derive(Clone, Copy, PartialEq, Debug)] +enum BlockMode { + Break, + Ignore, +} + +/// Like `maybe_whole_expr`, but for things other than expressions. +#[macro_export] +macro_rules! maybe_whole { + ($p:expr, $constructor:ident, |$x:ident| $e:expr) => { + if let token::Interpolated(nt) = &$p.token.kind { + if let token::$constructor(x) = &**nt { + let $x = x.clone(); + $p.bump(); + return Ok($e); + } + } + }; +} + +/// If the next tokens are ill-formed `$ty::` recover them as `<$ty>::`. +#[macro_export] +macro_rules! maybe_recover_from_interpolated_ty_qpath { + ($self: expr, $allow_qpath_recovery: expr) => { + if $allow_qpath_recovery && $self.look_ahead(1, |t| t == &token::ModSep) { + if let token::Interpolated(nt) = &$self.token.kind { + if let token::NtTy(ty) = &**nt { + let ty = ty.clone(); + $self.bump(); + return $self.maybe_recover_from_bad_qpath_stage_2($self.prev_span, ty); + } + } + } + } +} + +#[derive(Debug, Clone, Copy, PartialEq)] +enum PrevTokenKind { + DocComment, + Comma, + Plus, + Interpolated, + Eof, + Ident, + BitOr, + Other, +} + +// NOTE: `Ident`s are handled by `common.rs`. + +#[derive(Clone)] +pub struct Parser<'a> { + pub sess: &'a ParseSess, + /// The current normalized token. + /// "Normalized" means that some interpolated tokens + /// (`$i: ident` and `$l: lifetime` meta-variables) are replaced + /// with non-interpolated identifier and lifetime tokens they refer to. + /// Perhaps the normalized / non-normalized setup can be simplified somehow. + pub token: Token, + /// The span of the current non-normalized token. + meta_var_span: Option<Span>, + /// The span of the previous non-normalized token. + pub prev_span: Span, + /// The kind of the previous normalized token (in simplified form). + prev_token_kind: PrevTokenKind, + restrictions: Restrictions, + /// Used to determine the path to externally loaded source files. + pub(super) directory: Directory<'a>, + /// `true` to parse sub-modules in other files. + pub(super) recurse_into_file_modules: bool, + /// Name of the root module this parser originated from. If `None`, then the + /// name is not known. This does not change while the parser is descending + /// into modules, and sub-parsers have new values for this name. + pub root_module_name: Option<String>, + expected_tokens: Vec<TokenType>, + token_cursor: TokenCursor, + desugar_doc_comments: bool, + /// `true` we should configure out of line modules as we parse. + cfg_mods: bool, + /// This field is used to keep track of how many left angle brackets we have seen. This is + /// required in order to detect extra leading left angle brackets (`<` characters) and error + /// appropriately. + /// + /// See the comments in the `parse_path_segment` function for more details. + unmatched_angle_bracket_count: u32, + max_angle_bracket_count: u32, + /// A list of all unclosed delimiters found by the lexer. If an entry is used for error recovery + /// it gets removed from here. Every entry left at the end gets emitted as an independent + /// error. + pub(super) unclosed_delims: Vec<UnmatchedBrace>, + last_unexpected_token_span: Option<Span>, + pub last_type_ascription: Option<(Span, bool /* likely path typo */)>, + /// If present, this `Parser` is not parsing Rust code but rather a macro call. + subparser_name: Option<&'static str>, +} + +impl<'a> Drop for Parser<'a> { + fn drop(&mut self) { + emit_unclosed_delims(&mut self.unclosed_delims, &self.sess); + } +} + +#[derive(Clone)] +struct TokenCursor { + frame: TokenCursorFrame, + stack: Vec<TokenCursorFrame>, +} + +#[derive(Clone)] +struct TokenCursorFrame { + delim: token::DelimToken, + span: DelimSpan, + open_delim: bool, + tree_cursor: tokenstream::Cursor, + close_delim: bool, + last_token: LastToken, +} + +/// This is used in `TokenCursorFrame` above to track tokens that are consumed +/// by the parser, and then that's transitively used to record the tokens that +/// each parse AST item is created with. +/// +/// Right now this has two states, either collecting tokens or not collecting +/// tokens. If we're collecting tokens we just save everything off into a local +/// `Vec`. This should eventually though likely save tokens from the original +/// token stream and just use slicing of token streams to avoid creation of a +/// whole new vector. +/// +/// The second state is where we're passively not recording tokens, but the last +/// token is still tracked for when we want to start recording tokens. This +/// "last token" means that when we start recording tokens we'll want to ensure +/// that this, the first token, is included in the output. +/// +/// You can find some more example usage of this in the `collect_tokens` method +/// on the parser. +#[derive(Clone)] +enum LastToken { + Collecting(Vec<TreeAndJoint>), + Was(Option<TreeAndJoint>), +} + +impl TokenCursorFrame { + fn new(span: DelimSpan, delim: DelimToken, tts: &TokenStream) -> Self { + TokenCursorFrame { + delim, + span, + open_delim: delim == token::NoDelim, + tree_cursor: tts.clone().into_trees(), + close_delim: delim == token::NoDelim, + last_token: LastToken::Was(None), + } + } +} + +impl TokenCursor { + fn next(&mut self) -> Token { + loop { + let tree = if !self.frame.open_delim { + self.frame.open_delim = true; + TokenTree::open_tt(self.frame.span, self.frame.delim) + } else if let Some(tree) = self.frame.tree_cursor.next() { + tree + } else if !self.frame.close_delim { + self.frame.close_delim = true; + TokenTree::close_tt(self.frame.span, self.frame.delim) + } else if let Some(frame) = self.stack.pop() { + self.frame = frame; + continue + } else { + return Token::new(token::Eof, DUMMY_SP); + }; + + match self.frame.last_token { + LastToken::Collecting(ref mut v) => v.push(tree.clone().into()), + LastToken::Was(ref mut t) => *t = Some(tree.clone().into()), + } + + match tree { + TokenTree::Token(token) => return token, + TokenTree::Delimited(sp, delim, tts) => { + let frame = TokenCursorFrame::new(sp, delim, &tts); + self.stack.push(mem::replace(&mut self.frame, frame)); + } + } + } + } + + fn next_desugared(&mut self) -> Token { + let (name, sp) = match self.next() { + Token { kind: token::DocComment(name), span } => (name, span), + tok => return tok, + }; + + let stripped = strip_doc_comment_decoration(&name.as_str()); + + // Searches for the occurrences of `"#*` and returns the minimum number of `#`s + // required to wrap the text. + let mut num_of_hashes = 0; + let mut count = 0; + for ch in stripped.chars() { + count = match ch { + '"' => 1, + '#' if count > 0 => count + 1, + _ => 0, + }; + num_of_hashes = cmp::max(num_of_hashes, count); + } + + let delim_span = DelimSpan::from_single(sp); + let body = TokenTree::Delimited( + delim_span, + token::Bracket, + [ + TokenTree::token(token::Ident(sym::doc, false), sp), + TokenTree::token(token::Eq, sp), + TokenTree::token(TokenKind::lit( + token::StrRaw(num_of_hashes), Symbol::intern(&stripped), None + ), sp), + ] + .iter().cloned().collect::<TokenStream>().into(), + ); + + self.stack.push(mem::replace(&mut self.frame, TokenCursorFrame::new( + delim_span, + token::NoDelim, + &if doc_comment_style(&name.as_str()) == AttrStyle::Inner { + [TokenTree::token(token::Pound, sp), TokenTree::token(token::Not, sp), body] + .iter().cloned().collect::<TokenStream>() + } else { + [TokenTree::token(token::Pound, sp), body] + .iter().cloned().collect::<TokenStream>() + }, + ))); + + self.next() + } +} + +#[derive(Clone, PartialEq)] +enum TokenType { + Token(TokenKind), + Keyword(Symbol), + Operator, + Lifetime, + Ident, + Path, + Type, + Const, +} + +impl TokenType { + fn to_string(&self) -> String { + match *self { + TokenType::Token(ref t) => format!("`{}`", pprust::token_kind_to_string(t)), + TokenType::Keyword(kw) => format!("`{}`", kw), + TokenType::Operator => "an operator".to_string(), + TokenType::Lifetime => "lifetime".to_string(), + TokenType::Ident => "identifier".to_string(), + TokenType::Path => "path".to_string(), + TokenType::Type => "type".to_string(), + TokenType::Const => "const".to_string(), + } + } +} + +#[derive(Copy, Clone, Debug)] +enum TokenExpectType { + Expect, + NoExpect, +} + +/// A sequence separator. +struct SeqSep { + /// The separator token. + sep: Option<TokenKind>, + /// `true` if a trailing separator is allowed. + trailing_sep_allowed: bool, +} + +impl SeqSep { + fn trailing_allowed(t: TokenKind) -> SeqSep { + SeqSep { + sep: Some(t), + trailing_sep_allowed: true, + } + } + + fn none() -> SeqSep { + SeqSep { + sep: None, + trailing_sep_allowed: false, + } + } +} + +impl<'a> Parser<'a> { + pub fn new( + sess: &'a ParseSess, + tokens: TokenStream, + directory: Option<Directory<'a>>, + recurse_into_file_modules: bool, + desugar_doc_comments: bool, + subparser_name: Option<&'static str>, + ) -> Self { + let mut parser = Parser { + sess, + token: Token::dummy(), + prev_span: DUMMY_SP, + meta_var_span: None, + prev_token_kind: PrevTokenKind::Other, + restrictions: Restrictions::empty(), + recurse_into_file_modules, + directory: Directory { + path: Cow::from(PathBuf::new()), + ownership: DirectoryOwnership::Owned { relative: None } + }, + root_module_name: None, + expected_tokens: Vec::new(), + token_cursor: TokenCursor { + frame: TokenCursorFrame::new( + DelimSpan::dummy(), + token::NoDelim, + &tokens.into(), + ), + stack: Vec::new(), + }, + desugar_doc_comments, + cfg_mods: true, + unmatched_angle_bracket_count: 0, + max_angle_bracket_count: 0, + unclosed_delims: Vec::new(), + last_unexpected_token_span: None, + last_type_ascription: None, + subparser_name, + }; + + parser.token = parser.next_tok(); + + if let Some(directory) = directory { + parser.directory = directory; + } else if !parser.token.span.is_dummy() { + if let Some(FileName::Real(path)) = + &sess.source_map().lookup_char_pos(parser.token.span.lo()).file.unmapped_path { + if let Some(directory_path) = path.parent() { + parser.directory.path = Cow::from(directory_path.to_path_buf()); + } + } + } + + parser.process_potential_macro_variable(); + parser + } + + fn next_tok(&mut self) -> Token { + let mut next = if self.desugar_doc_comments { + self.token_cursor.next_desugared() + } else { + self.token_cursor.next() + }; + if next.span.is_dummy() { + // Tweak the location for better diagnostics, but keep syntactic context intact. + next.span = self.prev_span.with_ctxt(next.span.ctxt()); + } + next + } + + /// Converts the current token to a string using `self`'s reader. + pub fn this_token_to_string(&self) -> String { + pprust::token_to_string(&self.token) + } + + fn token_descr(&self) -> Option<&'static str> { + Some(match &self.token.kind { + _ if self.token.is_special_ident() => "reserved identifier", + _ if self.token.is_used_keyword() => "keyword", + _ if self.token.is_unused_keyword() => "reserved keyword", + token::DocComment(..) => "doc comment", + _ => return None, + }) + } + + pub(super) fn this_token_descr(&self) -> String { + if let Some(prefix) = self.token_descr() { + format!("{} `{}`", prefix, self.this_token_to_string()) + } else { + format!("`{}`", self.this_token_to_string()) + } + } + + crate fn unexpected<T>(&mut self) -> PResult<'a, T> { + match self.expect_one_of(&[], &[]) { + Err(e) => Err(e), + Ok(_) => unreachable!(), + } + } + + /// Expects and consumes the token `t`. Signals an error if the next token is not `t`. + pub fn expect(&mut self, t: &TokenKind) -> PResult<'a, bool /* recovered */> { + if self.expected_tokens.is_empty() { + if self.token == *t { + self.bump(); + Ok(false) + } else { + self.unexpected_try_recover(t) + } + } else { + self.expect_one_of(slice::from_ref(t), &[]) + } + } + + /// Expect next token to be edible or inedible token. If edible, + /// then consume it; if inedible, then return without consuming + /// anything. Signal a fatal error if next token is unexpected. + pub fn expect_one_of( + &mut self, + edible: &[TokenKind], + inedible: &[TokenKind], + ) -> PResult<'a, bool /* recovered */> { + if edible.contains(&self.token.kind) { + self.bump(); + Ok(false) + } else if inedible.contains(&self.token.kind) { + // leave it in the input + Ok(false) + } else if self.last_unexpected_token_span == Some(self.token.span) { + FatalError.raise(); + } else { + self.expected_one_of_not_found(edible, inedible) + } + } + + fn parse_ident(&mut self) -> PResult<'a, ast::Ident> { + self.parse_ident_common(true) + } + + fn parse_ident_common(&mut self, recover: bool) -> PResult<'a, ast::Ident> { + match self.token.kind { + token::Ident(name, _) => { + if self.token.is_reserved_ident() { + let mut err = self.expected_ident_found(); + if recover { + err.emit(); + } else { + return Err(err); + } + } + let span = self.token.span; + self.bump(); + Ok(Ident::new(name, span)) + } + _ => { + Err(if self.prev_token_kind == PrevTokenKind::DocComment { + self.span_fatal_err(self.prev_span, Error::UselessDocComment) + } else { + self.expected_ident_found() + }) + } + } + } + + /// Checks if the next token is `tok`, and returns `true` if so. + /// + /// This method will automatically add `tok` to `expected_tokens` if `tok` is not + /// encountered. + fn check(&mut self, tok: &TokenKind) -> bool { + let is_present = self.token == *tok; + if !is_present { self.expected_tokens.push(TokenType::Token(tok.clone())); } + is_present + } + + /// Consumes a token 'tok' if it exists. Returns whether the given token was present. + pub fn eat(&mut self, tok: &TokenKind) -> bool { + let is_present = self.check(tok); + if is_present { self.bump() } + is_present + } + + /// If the next token is the given keyword, returns `true` without eating it. + /// An expectation is also added for diagnostics purposes. + fn check_keyword(&mut self, kw: Symbol) -> bool { + self.expected_tokens.push(TokenType::Keyword(kw)); + self.token.is_keyword(kw) + } + + /// If the next token is the given keyword, eats it and returns `true`. + /// Otherwise, returns `false`. An expectation is also added for diagnostics purposes. + fn eat_keyword(&mut self, kw: Symbol) -> bool { + if self.check_keyword(kw) { + self.bump(); + true + } else { + false + } + } + + fn eat_keyword_noexpect(&mut self, kw: Symbol) -> bool { + if self.token.is_keyword(kw) { + self.bump(); + true + } else { + false + } + } + + /// If the given word is not a keyword, signals an error. + /// If the next token is not the given word, signals an error. + /// Otherwise, eats it. + fn expect_keyword(&mut self, kw: Symbol) -> PResult<'a, ()> { + if !self.eat_keyword(kw) { + self.unexpected() + } else { + Ok(()) + } + } + + fn check_or_expected(&mut self, ok: bool, typ: TokenType) -> bool { + if ok { + true + } else { + self.expected_tokens.push(typ); + false + } + } + + fn check_ident(&mut self) -> bool { + self.check_or_expected(self.token.is_ident(), TokenType::Ident) + } + + fn check_path(&mut self) -> bool { + self.check_or_expected(self.token.is_path_start(), TokenType::Path) + } + + fn check_type(&mut self) -> bool { + self.check_or_expected(self.token.can_begin_type(), TokenType::Type) + } + + fn check_const_arg(&mut self) -> bool { + self.check_or_expected(self.token.can_begin_const_arg(), TokenType::Const) + } + + /// Checks to see if the next token is either `+` or `+=`. + /// Otherwise returns `false`. + fn check_plus(&mut self) -> bool { + self.check_or_expected( + self.token.is_like_plus(), + TokenType::Token(token::BinOp(token::Plus)), + ) + } + + /// Expects and consumes a `+`. if `+=` is seen, replaces it with a `=` + /// and continues. If a `+` is not seen, returns `false`. + /// + /// This is used when token-splitting `+=` into `+`. + /// See issue #47856 for an example of when this may occur. + fn eat_plus(&mut self) -> bool { + self.expected_tokens.push(TokenType::Token(token::BinOp(token::Plus))); + match self.token.kind { + token::BinOp(token::Plus) => { + self.bump(); + true + } + token::BinOpEq(token::Plus) => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + self.bump_with(token::Eq, span); + true + } + _ => false, + } + } + + /// Expects and consumes an `&`. If `&&` is seen, replaces it with a single + /// `&` and continues. If an `&` is not seen, signals an error. + fn expect_and(&mut self) -> PResult<'a, ()> { + self.expected_tokens.push(TokenType::Token(token::BinOp(token::And))); + match self.token.kind { + token::BinOp(token::And) => { + self.bump(); + Ok(()) + } + token::AndAnd => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + Ok(self.bump_with(token::BinOp(token::And), span)) + } + _ => self.unexpected() + } + } + + /// Expects and consumes an `|`. If `||` is seen, replaces it with a single + /// `|` and continues. If an `|` is not seen, signals an error. + fn expect_or(&mut self) -> PResult<'a, ()> { + self.expected_tokens.push(TokenType::Token(token::BinOp(token::Or))); + match self.token.kind { + token::BinOp(token::Or) => { + self.bump(); + Ok(()) + } + token::OrOr => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + Ok(self.bump_with(token::BinOp(token::Or), span)) + } + _ => self.unexpected() + } + } + + /// Attempts to consume a `<`. If `<<` is seen, replaces it with a single + /// `<` and continue. If `<-` is seen, replaces it with a single `<` + /// and continue. If a `<` is not seen, returns false. + /// + /// This is meant to be used when parsing generics on a path to get the + /// starting token. + fn eat_lt(&mut self) -> bool { + self.expected_tokens.push(TokenType::Token(token::Lt)); + let ate = match self.token.kind { + token::Lt => { + self.bump(); + true + } + token::BinOp(token::Shl) => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + self.bump_with(token::Lt, span); + true + } + token::LArrow => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + self.bump_with(token::BinOp(token::Minus), span); + true + } + _ => false, + }; + + if ate { + // See doc comment for `unmatched_angle_bracket_count`. + self.unmatched_angle_bracket_count += 1; + self.max_angle_bracket_count += 1; + debug!("eat_lt: (increment) count={:?}", self.unmatched_angle_bracket_count); + } + + ate + } + + fn expect_lt(&mut self) -> PResult<'a, ()> { + if !self.eat_lt() { + self.unexpected() + } else { + Ok(()) + } + } + + /// Expects and consumes a single `>` token. if a `>>` is seen, replaces it + /// with a single `>` and continues. If a `>` is not seen, signals an error. + fn expect_gt(&mut self) -> PResult<'a, ()> { + self.expected_tokens.push(TokenType::Token(token::Gt)); + let ate = match self.token.kind { + token::Gt => { + self.bump(); + Some(()) + } + token::BinOp(token::Shr) => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + Some(self.bump_with(token::Gt, span)) + } + token::BinOpEq(token::Shr) => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + Some(self.bump_with(token::Ge, span)) + } + token::Ge => { + let span = self.token.span.with_lo(self.token.span.lo() + BytePos(1)); + Some(self.bump_with(token::Eq, span)) + } + _ => None, + }; + + match ate { + Some(_) => { + // See doc comment for `unmatched_angle_bracket_count`. + if self.unmatched_angle_bracket_count > 0 { + self.unmatched_angle_bracket_count -= 1; + debug!("expect_gt: (decrement) count={:?}", self.unmatched_angle_bracket_count); + } + + Ok(()) + }, + None => self.unexpected(), + } + } + + /// Parses a sequence, including the closing delimiter. The function + /// `f` must consume tokens until reaching the next separator or + /// closing bracket. + fn parse_seq_to_end<T>( + &mut self, + ket: &TokenKind, + sep: SeqSep, + f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, + ) -> PResult<'a, Vec<T>> { + let (val, _, recovered) = self.parse_seq_to_before_end(ket, sep, f)?; + if !recovered { + self.bump(); + } + Ok(val) + } + + /// Parses a sequence, not including the closing delimiter. The function + /// `f` must consume tokens until reaching the next separator or + /// closing bracket. + fn parse_seq_to_before_end<T>( + &mut self, + ket: &TokenKind, + sep: SeqSep, + f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, + ) -> PResult<'a, (Vec<T>, bool, bool)> { + self.parse_seq_to_before_tokens(&[ket], sep, TokenExpectType::Expect, f) + } + + fn expect_any_with_type(&mut self, kets: &[&TokenKind], expect: TokenExpectType) -> bool { + kets.iter().any(|k| { + match expect { + TokenExpectType::Expect => self.check(k), + TokenExpectType::NoExpect => self.token == **k, + } + }) + } + + fn parse_seq_to_before_tokens<T>( + &mut self, + kets: &[&TokenKind], + sep: SeqSep, + expect: TokenExpectType, + mut f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, + ) -> PResult<'a, (Vec<T>, bool /* trailing */, bool /* recovered */)> { + let mut first = true; + let mut recovered = false; + let mut trailing = false; + let mut v = vec![]; + while !self.expect_any_with_type(kets, expect) { + if let token::CloseDelim(..) | token::Eof = self.token.kind { + break + } + if let Some(ref t) = sep.sep { + if first { + first = false; + } else { + match self.expect(t) { + Ok(false) => {} + Ok(true) => { + recovered = true; + break; + } + Err(mut e) => { + // Attempt to keep parsing if it was a similar separator. + if let Some(ref tokens) = t.similar_tokens() { + if tokens.contains(&self.token.kind) { + self.bump(); + } + } + e.emit(); + // Attempt to keep parsing if it was an omitted separator. + match f(self) { + Ok(t) => { + v.push(t); + continue; + }, + Err(mut e) => { + e.cancel(); + break; + } + } + } + } + } + } + if sep.trailing_sep_allowed && self.expect_any_with_type(kets, expect) { + trailing = true; + break; + } + + let t = f(self)?; + v.push(t); + } + + Ok((v, trailing, recovered)) + } + + /// Parses a sequence, including the closing delimiter. The function + /// `f` must consume tokens until reaching the next separator or + /// closing bracket. + fn parse_unspanned_seq<T>( + &mut self, + bra: &TokenKind, + ket: &TokenKind, + sep: SeqSep, + f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, + ) -> PResult<'a, (Vec<T>, bool)> { + self.expect(bra)?; + let (result, trailing, recovered) = self.parse_seq_to_before_end(ket, sep, f)?; + if !recovered { + self.eat(ket); + } + Ok((result, trailing)) + } + + fn parse_delim_comma_seq<T>( + &mut self, + delim: DelimToken, + f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, + ) -> PResult<'a, (Vec<T>, bool)> { + self.parse_unspanned_seq( + &token::OpenDelim(delim), + &token::CloseDelim(delim), + SeqSep::trailing_allowed(token::Comma), + f, + ) + } + + fn parse_paren_comma_seq<T>( + &mut self, + f: impl FnMut(&mut Parser<'a>) -> PResult<'a, T>, + ) -> PResult<'a, (Vec<T>, bool)> { + self.parse_delim_comma_seq(token::Paren, f) + } + + /// Advance the parser by one token. + pub fn bump(&mut self) { + if self.prev_token_kind == PrevTokenKind::Eof { + // Bumping after EOF is a bad sign, usually an infinite loop. + self.bug("attempted to bump the parser past EOF (may be stuck in a loop)"); + } + + self.prev_span = self.meta_var_span.take().unwrap_or(self.token.span); + + // Record last token kind for possible error recovery. + self.prev_token_kind = match self.token.kind { + token::DocComment(..) => PrevTokenKind::DocComment, + token::Comma => PrevTokenKind::Comma, + token::BinOp(token::Plus) => PrevTokenKind::Plus, + token::BinOp(token::Or) => PrevTokenKind::BitOr, + token::Interpolated(..) => PrevTokenKind::Interpolated, + token::Eof => PrevTokenKind::Eof, + token::Ident(..) => PrevTokenKind::Ident, + _ => PrevTokenKind::Other, + }; + + self.token = self.next_tok(); + self.expected_tokens.clear(); + // Check after each token. + self.process_potential_macro_variable(); + } + + /// Advances the parser using provided token as a next one. Use this when + /// consuming a part of a token. For example a single `<` from `<<`. + fn bump_with(&mut self, next: TokenKind, span: Span) { + self.prev_span = self.token.span.with_hi(span.lo()); + // It would be incorrect to record the kind of the current token, but + // fortunately for tokens currently using `bump_with`, the + // `prev_token_kind` will be of no use anyway. + self.prev_token_kind = PrevTokenKind::Other; + self.token = Token::new(next, span); + self.expected_tokens.clear(); + } + + /// Look-ahead `dist` tokens of `self.token` and get access to that token there. + /// When `dist == 0` then the current token is looked at. + pub fn look_ahead<R>(&self, dist: usize, looker: impl FnOnce(&Token) -> R) -> R { + if dist == 0 { + return looker(&self.token); + } + + let frame = &self.token_cursor.frame; + looker(&match frame.tree_cursor.look_ahead(dist - 1) { + Some(tree) => match tree { + TokenTree::Token(token) => token, + TokenTree::Delimited(dspan, delim, _) => + Token::new(token::OpenDelim(delim), dspan.open), + } + None => Token::new(token::CloseDelim(frame.delim), frame.span.close) + }) + } + + /// Returns whether any of the given keywords are `dist` tokens ahead of the current one. + fn is_keyword_ahead(&self, dist: usize, kws: &[Symbol]) -> bool { + self.look_ahead(dist, |t| kws.iter().any(|&kw| t.is_keyword(kw))) + } + + /// Parses asyncness: `async` or nothing. + fn parse_asyncness(&mut self) -> IsAsync { + if self.eat_keyword(kw::Async) { + IsAsync::Async { + closure_id: DUMMY_NODE_ID, + return_impl_trait_id: DUMMY_NODE_ID, + } + } else { + IsAsync::NotAsync + } + } + + /// Parses unsafety: `unsafe` or nothing. + fn parse_unsafety(&mut self) -> Unsafety { + if self.eat_keyword(kw::Unsafe) { + Unsafety::Unsafe + } else { + Unsafety::Normal + } + } + + /// Parses mutability (`mut` or nothing). + fn parse_mutability(&mut self) -> Mutability { + if self.eat_keyword(kw::Mut) { + Mutability::Mutable + } else { + Mutability::Immutable + } + } + + /// Possibly parses mutability (`const` or `mut`). + fn parse_const_or_mut(&mut self) -> Option<Mutability> { + if self.eat_keyword(kw::Mut) { + Some(Mutability::Mutable) + } else if self.eat_keyword(kw::Const) { + Some(Mutability::Immutable) + } else { + None + } + } + + fn parse_field_name(&mut self) -> PResult<'a, Ident> { + if let token::Literal(token::Lit { kind: token::Integer, symbol, suffix }) = + self.token.kind { + self.expect_no_suffix(self.token.span, "a tuple index", suffix); + self.bump(); + Ok(Ident::new(symbol, self.prev_span)) + } else { + self.parse_ident_common(false) + } + } + + fn expect_delimited_token_tree(&mut self) -> PResult<'a, (MacDelimiter, TokenStream)> { + let delim = match self.token.kind { + token::OpenDelim(delim) => delim, + _ => { + let msg = "expected open delimiter"; + let mut err = self.fatal(msg); + err.span_label(self.token.span, msg); + return Err(err) + } + }; + let tts = match self.parse_token_tree() { + TokenTree::Delimited(_, _, tts) => tts, + _ => unreachable!(), + }; + let delim = match delim { + token::Paren => MacDelimiter::Parenthesis, + token::Bracket => MacDelimiter::Bracket, + token::Brace => MacDelimiter::Brace, + token::NoDelim => self.bug("unexpected no delimiter"), + }; + Ok((delim, tts.into())) + } + + fn parse_or_use_outer_attributes( + &mut self, + already_parsed_attrs: Option<ThinVec<Attribute>>, + ) -> PResult<'a, ThinVec<Attribute>> { + if let Some(attrs) = already_parsed_attrs { + Ok(attrs) + } else { + self.parse_outer_attributes().map(|a| a.into()) + } + } + + pub fn process_potential_macro_variable(&mut self) { + self.token = match self.token.kind { + token::Dollar if self.token.span.from_expansion() && + self.look_ahead(1, |t| t.is_ident()) => { + self.bump(); + let name = match self.token.kind { + token::Ident(name, _) => name, + _ => unreachable!() + }; + let span = self.prev_span.to(self.token.span); + self.diagnostic() + .struct_span_fatal(span, &format!("unknown macro variable `{}`", name)) + .span_label(span, "unknown macro variable") + .emit(); + self.bump(); + return + } + token::Interpolated(ref nt) => { + self.meta_var_span = Some(self.token.span); + // Interpolated identifier and lifetime tokens are replaced with usual identifier + // and lifetime tokens, so the former are never encountered during normal parsing. + match **nt { + token::NtIdent(ident, is_raw) => + Token::new(token::Ident(ident.name, is_raw), ident.span), + token::NtLifetime(ident) => + Token::new(token::Lifetime(ident.name), ident.span), + _ => return, + } + } + _ => return, + }; + } + + /// Parses a single token tree from the input. + pub fn parse_token_tree(&mut self) -> TokenTree { + match self.token.kind { + token::OpenDelim(..) => { + let frame = mem::replace(&mut self.token_cursor.frame, + self.token_cursor.stack.pop().unwrap()); + self.token.span = frame.span.entire(); + self.bump(); + TokenTree::Delimited( + frame.span, + frame.delim, + frame.tree_cursor.stream.into(), + ) + }, + token::CloseDelim(_) | token::Eof => unreachable!(), + _ => { + let token = self.token.take(); + self.bump(); + TokenTree::Token(token) + } + } + } + + /// Parses a stream of tokens into a list of `TokenTree`s, up to EOF. + pub fn parse_all_token_trees(&mut self) -> PResult<'a, Vec<TokenTree>> { + let mut tts = Vec::new(); + while self.token != token::Eof { + tts.push(self.parse_token_tree()); + } + Ok(tts) + } + + pub fn parse_tokens(&mut self) -> TokenStream { + let mut result = Vec::new(); + loop { + match self.token.kind { + token::Eof | token::CloseDelim(..) => break, + _ => result.push(self.parse_token_tree().into()), + } + } + TokenStream::new(result) + } + + /// Evaluates the closure with restrictions in place. + /// + /// Afters the closure is evaluated, restrictions are reset. + fn with_res<T>(&mut self, res: Restrictions, f: impl FnOnce(&mut Self) -> T) -> T { + let old = self.restrictions; + self.restrictions = res; + let res = f(self); + self.restrictions = old; + res + } + + fn is_crate_vis(&self) -> bool { + self.token.is_keyword(kw::Crate) && self.look_ahead(1, |t| t != &token::ModSep) + } + + /// Parses `pub`, `pub(crate)` and `pub(in path)` plus shortcuts `crate` for `pub(crate)`, + /// `pub(self)` for `pub(in self)` and `pub(super)` for `pub(in super)`. + /// If the following element can't be a tuple (i.e., it's a function definition), then + /// it's not a tuple struct field), and the contents within the parentheses isn't valid, + /// so emit a proper diagnostic. + pub fn parse_visibility(&mut self, can_take_tuple: bool) -> PResult<'a, Visibility> { + maybe_whole!(self, NtVis, |x| x); + + self.expected_tokens.push(TokenType::Keyword(kw::Crate)); + if self.is_crate_vis() { + self.bump(); // `crate` + self.sess.gated_spans.gate(sym::crate_visibility_modifier, self.prev_span); + return Ok(respan(self.prev_span, VisibilityKind::Crate(CrateSugar::JustCrate))); + } + + if !self.eat_keyword(kw::Pub) { + // We need a span for our `Spanned<VisibilityKind>`, but there's inherently no + // keyword to grab a span from for inherited visibility; an empty span at the + // beginning of the current token would seem to be the "Schelling span". + return Ok(respan(self.token.span.shrink_to_lo(), VisibilityKind::Inherited)) + } + let lo = self.prev_span; + + if self.check(&token::OpenDelim(token::Paren)) { + // We don't `self.bump()` the `(` yet because this might be a struct definition where + // `()` or a tuple might be allowed. For example, `struct Struct(pub (), pub (usize));`. + // Because of this, we only `bump` the `(` if we're assured it is appropriate to do so + // by the following tokens. + if self.is_keyword_ahead(1, &[kw::Crate]) + && self.look_ahead(2, |t| t != &token::ModSep) // account for `pub(crate::foo)` + { + // Parse `pub(crate)`. + self.bump(); // `(` + self.bump(); // `crate` + self.expect(&token::CloseDelim(token::Paren))?; // `)` + let vis = VisibilityKind::Crate(CrateSugar::PubCrate); + return Ok(respan(lo.to(self.prev_span), vis)); + } else if self.is_keyword_ahead(1, &[kw::In]) { + // Parse `pub(in path)`. + self.bump(); // `(` + self.bump(); // `in` + let path = self.parse_path(PathStyle::Mod)?; // `path` + self.expect(&token::CloseDelim(token::Paren))?; // `)` + let vis = VisibilityKind::Restricted { + path: P(path), + id: ast::DUMMY_NODE_ID, + }; + return Ok(respan(lo.to(self.prev_span), vis)); + } else if self.look_ahead(2, |t| t == &token::CloseDelim(token::Paren)) + && self.is_keyword_ahead(1, &[kw::Super, kw::SelfLower]) + { + // Parse `pub(self)` or `pub(super)`. + self.bump(); // `(` + let path = self.parse_path(PathStyle::Mod)?; // `super`/`self` + self.expect(&token::CloseDelim(token::Paren))?; // `)` + let vis = VisibilityKind::Restricted { + path: P(path), + id: ast::DUMMY_NODE_ID, + }; + return Ok(respan(lo.to(self.prev_span), vis)); + } else if !can_take_tuple { // Provide this diagnostic if this is not a tuple struct. + self.recover_incorrect_vis_restriction()?; + // Emit diagnostic, but continue with public visibility. + } + } + + Ok(respan(lo, VisibilityKind::Public)) + } + + /// Recovery for e.g. `pub(something) fn ...` or `struct X { pub(something) y: Z }` + fn recover_incorrect_vis_restriction(&mut self) -> PResult<'a, ()> { + self.bump(); // `(` + let path = self.parse_path(PathStyle::Mod)?; + self.expect(&token::CloseDelim(token::Paren))?; // `)` + + let msg = "incorrect visibility restriction"; + let suggestion = r##"some possible visibility restrictions are: +`pub(crate)`: visible only on the current crate +`pub(super)`: visible only in the current module's parent +`pub(in path::to::module)`: visible only on the specified path"##; + + let path_str = pprust::path_to_string(&path); + + struct_span_err!(self.sess.span_diagnostic, path.span, E0704, "{}", msg) + .help(suggestion) + .span_suggestion( + path.span, + &format!("make this visible only to module `{}` with `in`", path_str), + format!("in {}", path_str), + Applicability::MachineApplicable, + ) + .emit(); + + Ok(()) + } + + /// Parses `extern string_literal?`. + /// If `extern` is not found, the Rust ABI is used. + /// If `extern` is found and a `string_literal` does not follow, the C ABI is used. + fn parse_extern_abi(&mut self) -> PResult<'a, Abi> { + Ok(if self.eat_keyword(kw::Extern) { + self.parse_opt_abi()? + } else { + Abi::default() + }) + } + + /// Parses a string literal as an ABI spec. + /// If one is not found, the "C" ABI is used. + fn parse_opt_abi(&mut self) -> PResult<'a, Abi> { + let span = if self.token.can_begin_literal_or_bool() { + let ast::Lit { span, kind, .. } = self.parse_lit()?; + match kind { + ast::LitKind::Str(symbol, _) => return Ok(Abi::new(symbol, span)), + ast::LitKind::Err(_) => {} + _ => { + self.struct_span_err(span, "non-string ABI literal") + .span_suggestion( + span, + "specify the ABI with a string literal", + "\"C\"".to_string(), + Applicability::MaybeIncorrect, + ) + .emit(); + } + } + span + } else { + self.prev_span + }; + Ok(Abi::new(sym::C, span)) + } + + /// We are parsing `async fn`. If we are on Rust 2015, emit an error. + fn ban_async_in_2015(&self, async_span: Span) { + if async_span.rust_2015() { + self.diagnostic() + .struct_span_err_with_code( + async_span, + "`async fn` is not permitted in the 2015 edition", + DiagnosticId::Error("E0670".into()) + ) + .emit(); + } + } + + fn collect_tokens<R>( + &mut self, + f: impl FnOnce(&mut Self) -> PResult<'a, R>, + ) -> PResult<'a, (R, TokenStream)> { + // Record all tokens we parse when parsing this item. + let mut tokens = Vec::new(); + let prev_collecting = match self.token_cursor.frame.last_token { + LastToken::Collecting(ref mut list) => { + Some(mem::take(list)) + } + LastToken::Was(ref mut last) => { + tokens.extend(last.take()); + None + } + }; + self.token_cursor.frame.last_token = LastToken::Collecting(tokens); + let prev = self.token_cursor.stack.len(); + let ret = f(self); + let last_token = if self.token_cursor.stack.len() == prev { + &mut self.token_cursor.frame.last_token + } else if self.token_cursor.stack.get(prev).is_none() { + // This can happen due to a bad interaction of two unrelated recovery mechanisms with + // mismatched delimiters *and* recovery lookahead on the likely typo `pub ident(` + // (#62881). + return Ok((ret?, TokenStream::default())); + } else { + &mut self.token_cursor.stack[prev].last_token + }; + + // Pull out the tokens that we've collected from the call to `f` above. + let mut collected_tokens = match *last_token { + LastToken::Collecting(ref mut v) => mem::take(v), + LastToken::Was(ref was) => { + let msg = format!("our vector went away? - found Was({:?})", was); + debug!("collect_tokens: {}", msg); + self.sess.span_diagnostic.delay_span_bug(self.token.span, &msg); + // This can happen due to a bad interaction of two unrelated recovery mechanisms + // with mismatched delimiters *and* recovery lookahead on the likely typo + // `pub ident(` (#62895, different but similar to the case above). + return Ok((ret?, TokenStream::default())); + } + }; + + // If we're not at EOF our current token wasn't actually consumed by + // `f`, but it'll still be in our list that we pulled out. In that case + // put it back. + let extra_token = if self.token != token::Eof { + collected_tokens.pop() + } else { + None + }; + + // If we were previously collecting tokens, then this was a recursive + // call. In that case we need to record all the tokens we collected in + // our parent list as well. To do that we push a clone of our stream + // onto the previous list. + match prev_collecting { + Some(mut list) => { + list.extend(collected_tokens.iter().cloned()); + list.extend(extra_token); + *last_token = LastToken::Collecting(list); + } + None => { + *last_token = LastToken::Was(extra_token); + } + } + + Ok((ret?, TokenStream::new(collected_tokens))) + } + + /// `::{` or `::*` + fn is_import_coupler(&mut self) -> bool { + self.check(&token::ModSep) && + self.look_ahead(1, |t| *t == token::OpenDelim(token::Brace) || + *t == token::BinOp(token::Star)) + } + + fn parse_optional_str(&mut self) -> Option<(Symbol, ast::StrStyle, Option<ast::Name>)> { + let ret = match self.token.kind { + token::Literal(token::Lit { kind: token::Str, symbol, suffix }) => + (symbol, ast::StrStyle::Cooked, suffix), + token::Literal(token::Lit { kind: token::StrRaw(n), symbol, suffix }) => + (symbol, ast::StrStyle::Raw(n), suffix), + _ => return None + }; + self.bump(); + Some(ret) + } + + pub fn parse_str(&mut self) -> PResult<'a, (Symbol, StrStyle)> { + match self.parse_optional_str() { + Some((s, style, suf)) => { + let sp = self.prev_span; + self.expect_no_suffix(sp, "a string literal", suf); + Ok((s, style)) + } + _ => { + let msg = "expected string literal"; + let mut err = self.fatal(msg); + err.span_label(self.token.span, msg); + Err(err) + } + } + } +} + +crate fn make_unclosed_delims_error( + unmatched: UnmatchedBrace, + sess: &ParseSess, +) -> Option<DiagnosticBuilder<'_>> { + // `None` here means an `Eof` was found. We already emit those errors elsewhere, we add them to + // `unmatched_braces` only for error recovery in the `Parser`. + let found_delim = unmatched.found_delim?; + let mut err = sess.span_diagnostic.struct_span_err(unmatched.found_span, &format!( + "incorrect close delimiter: `{}`", + pprust::token_kind_to_string(&token::CloseDelim(found_delim)), + )); + err.span_label(unmatched.found_span, "incorrect close delimiter"); + if let Some(sp) = unmatched.candidate_span { + err.span_label(sp, "close delimiter possibly meant for this"); + } + if let Some(sp) = unmatched.unclosed_span { + err.span_label(sp, "un-closed delimiter"); + } + Some(err) +} + +pub fn emit_unclosed_delims(unclosed_delims: &mut Vec<UnmatchedBrace>, sess: &ParseSess) { + *sess.reached_eof.borrow_mut() |= unclosed_delims.iter() + .any(|unmatched_delim| unmatched_delim.found_delim.is_none()); + for unmatched in unclosed_delims.drain(..) { + make_unclosed_delims_error(unmatched, sess).map(|mut e| e.emit()); + } +} |