3 //! `TokenStream`s represent syntactic objects before they are converted into ASTs.
4 //! A `TokenStream` is, roughly speaking, a sequence (eg stream) of `TokenTree`s,
5 //! which are themselves a single `Token` or a `Delimited` subsequence of tokens.
9 //! `TokenStream`s are persistent data structures constructed as ropes with reference
10 //! counted-children. In general, this means that calling an operation on a `TokenStream`
11 //! (such as `slice`) produces an entirely new `TokenStream` from the borrowed reference to
12 //! the original. This essentially coerces `TokenStream`s into 'views' of their subparts,
13 //! and a borrowed `TokenStream` is sufficient to build an owned `TokenStream` without taking
14 //! ownership of the original.
16 use crate::token::{self, DelimToken, Token, TokenKind};
18 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
19 use rustc_data_structures::sync::Lrc;
20 use rustc_macros::HashStable_Generic;
21 use rustc_span::{Span, DUMMY_SP};
22 use smallvec::{smallvec, SmallVec};
26 /// When the main rust parser encounters a syntax-extension invocation, it
27 /// parses the arguments to the invocation as a token-tree. This is a very
28 /// loose structure, such that all sorts of different AST-fragments can
29 /// be passed to syntax extensions using a uniform type.
31 /// If the syntax extension is an MBE macro, it will attempt to match its
32 /// LHS token tree against the provided token tree, and if it finds a
33 /// match, will transcribe the RHS token tree, splicing in any captured
34 /// `macro_parser::matched_nonterminals` into the `SubstNt`s it finds.
36 /// The RHS of an MBE macro is the only place `SubstNt`s are substituted.
37 /// Nothing special happens to misnamed or misplaced `SubstNt`s.
38 #[derive(Debug, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
42 /// A delimited sequence of token trees
43 Delimited(DelimSpan, DelimToken, TokenStream),
46 // Ensure all fields of `TokenTree` is `Send` and `Sync`.
47 #[cfg(parallel_compiler)]
51 DelimSpan: Send + Sync,
52 DelimToken: Send + Sync,
53 TokenStream: Send + Sync,
58 /// Checks if this TokenTree is equal to the other, regardless of span information.
59 pub fn eq_unspanned(&self, other: &TokenTree) -> bool {
61 (TokenTree::Token(token), TokenTree::Token(token2)) => token.kind == token2.kind,
62 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
63 delim == delim2 && tts.eq_unspanned(&tts2)
69 /// Retrieves the TokenTree's span.
70 pub fn span(&self) -> Span {
72 TokenTree::Token(token) => token.span,
73 TokenTree::Delimited(sp, ..) => sp.entire(),
77 /// Modify the `TokenTree`'s span in-place.
78 pub fn set_span(&mut self, span: Span) {
80 TokenTree::Token(token) => token.span = span,
81 TokenTree::Delimited(dspan, ..) => *dspan = DelimSpan::from_single(span),
85 pub fn joint(self) -> TokenStream {
86 TokenStream::new(vec![(self, Joint)])
89 pub fn token(kind: TokenKind, span: Span) -> TokenTree {
90 TokenTree::Token(Token::new(kind, span))
93 /// Returns the opening delimiter as a token tree.
94 pub fn open_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
95 TokenTree::token(token::OpenDelim(delim), span.open)
98 /// Returns the closing delimiter as a token tree.
99 pub fn close_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
100 TokenTree::token(token::CloseDelim(delim), span.close)
103 pub fn uninterpolate(self) -> TokenTree {
105 TokenTree::Token(token) => TokenTree::Token(token.uninterpolate().into_owned()),
111 impl<CTX> HashStable<CTX> for TokenStream
113 CTX: crate::HashStableContext,
115 fn hash_stable(&self, hcx: &mut CTX, hasher: &mut StableHasher) {
116 for sub_tt in self.trees() {
117 sub_tt.hash_stable(hcx, hasher);
122 /// A `TokenStream` is an abstract sequence of tokens, organized into `TokenTree`s.
124 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
125 /// instead of a representation of the abstract syntax tree.
126 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for back-compat.
127 #[derive(Clone, Debug, Default, Encodable, Decodable)]
128 pub struct TokenStream(pub Lrc<Vec<TreeAndJoint>>);
130 pub type TreeAndJoint = (TokenTree, IsJoint);
132 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
133 #[cfg(target_arch = "x86_64")]
134 rustc_data_structures::static_assert_size!(TokenStream, 8);
136 #[derive(Clone, Copy, Debug, PartialEq, Encodable, Decodable)]
145 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
146 /// separating the two arguments with a comma for diagnostic suggestions.
147 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
148 // Used to suggest if a user writes `foo!(a b);`
149 let mut suggestion = None;
150 let mut iter = self.0.iter().enumerate().peekable();
151 while let Some((pos, ts)) = iter.next() {
152 if let Some((_, next)) = iter.peek() {
153 let sp = match (&ts, &next) {
154 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
156 (TokenTree::Token(token_left), NonJoint),
157 (TokenTree::Token(token_right), _),
158 ) if ((token_left.is_ident() && !token_left.is_reserved_ident())
159 || token_left.is_lit())
160 && ((token_right.is_ident() && !token_right.is_reserved_ident())
161 || token_right.is_lit()) =>
165 ((TokenTree::Delimited(sp, ..), NonJoint), _) => sp.entire(),
168 let sp = sp.shrink_to_hi();
169 let comma = (TokenTree::token(token::Comma, sp), NonJoint);
170 suggestion = Some((pos, comma, sp));
173 if let Some((pos, comma, sp)) = suggestion {
174 let mut new_stream = vec![];
175 let parts = self.0.split_at(pos + 1);
176 new_stream.extend_from_slice(parts.0);
177 new_stream.push(comma);
178 new_stream.extend_from_slice(parts.1);
179 return Some((TokenStream::new(new_stream), sp));
185 impl From<TokenTree> for TokenStream {
186 fn from(tree: TokenTree) -> TokenStream {
187 TokenStream::new(vec![(tree, NonJoint)])
191 impl From<TokenTree> for TreeAndJoint {
192 fn from(tree: TokenTree) -> TreeAndJoint {
197 impl iter::FromIterator<TokenTree> for TokenStream {
198 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
199 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndJoint>>())
203 impl Eq for TokenStream {}
205 impl PartialEq<TokenStream> for TokenStream {
206 fn eq(&self, other: &TokenStream) -> bool {
207 self.trees().eq(other.trees())
212 pub fn new(streams: Vec<TreeAndJoint>) -> TokenStream {
213 TokenStream(Lrc::new(streams))
216 pub fn is_empty(&self) -> bool {
220 pub fn len(&self) -> usize {
224 pub fn span(&self) -> Option<Span> {
227 [(tt, _)] => Some(tt.span()),
228 [(tt_start, _), .., (tt_end, _)] => Some(tt_start.span().to(tt_end.span())),
232 pub fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
233 match streams.len() {
234 0 => TokenStream::default(),
235 1 => streams.pop().unwrap(),
237 // We are going to extend the first stream in `streams` with
238 // the elements from the subsequent streams. This requires
239 // using `make_mut()` on the first stream, and in practice this
240 // doesn't cause cloning 99.9% of the time.
242 // One very common use case is when `streams` has two elements,
243 // where the first stream has any number of elements within
244 // (often 1, but sometimes many more) and the second stream has
245 // a single element within.
247 // Determine how much the first stream will be extended.
248 // Needed to avoid quadratic blow up from on-the-fly
249 // reallocations (#57735).
250 let num_appends = streams.iter().skip(1).map(|ts| ts.len()).sum();
252 // Get the first stream. If it's `None`, create an empty
254 let mut iter = streams.drain(..);
255 let mut first_stream_lrc = iter.next().unwrap().0;
257 // Append the elements to the first stream, after reserving
259 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
260 first_vec_mut.reserve(num_appends);
262 first_vec_mut.extend(stream.0.iter().cloned());
265 // Create the final `TokenStream`.
266 TokenStream(first_stream_lrc)
271 pub fn trees(&self) -> Cursor {
272 self.clone().into_trees()
275 pub fn into_trees(self) -> Cursor {
279 /// Compares two `TokenStream`s, checking equality without regarding span information.
280 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
281 let mut t1 = self.trees();
282 let mut t2 = other.trees();
283 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
284 if !t1.eq_unspanned(&t2) {
288 t1.next().is_none() && t2.next().is_none()
291 pub fn map_enumerated<F: FnMut(usize, TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
292 TokenStream(Lrc::new(
296 .map(|(i, (tree, is_joint))| (f(i, tree.clone()), *is_joint))
301 pub fn map<F: FnMut(TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
302 TokenStream(Lrc::new(
303 self.0.iter().map(|(tree, is_joint)| (f(tree.clone()), *is_joint)).collect(),
308 // 99.5%+ of the time we have 1 or 2 elements in this vector.
310 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
312 impl TokenStreamBuilder {
313 pub fn new() -> TokenStreamBuilder {
314 TokenStreamBuilder(SmallVec::new())
317 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
318 let mut stream = stream.into();
320 // If `self` is not empty and the last tree within the last stream is a
321 // token tree marked with `Joint`...
322 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
323 if let Some((TokenTree::Token(last_token), Joint)) = last_stream_lrc.last() {
324 // ...and `stream` is not empty and the first tree within it is
326 let TokenStream(ref mut stream_lrc) = stream;
327 if let Some((TokenTree::Token(token), is_joint)) = stream_lrc.first() {
328 // ...and the two tokens can be glued together...
329 if let Some(glued_tok) = last_token.glue(&token) {
330 // ...then do so, by overwriting the last token
331 // tree in `self` and removing the first token tree
332 // from `stream`. This requires using `make_mut()`
333 // on the last stream in `self` and on `stream`,
334 // and in practice this doesn't cause cloning 99.9%
337 // Overwrite the last token tree with the merged
339 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
340 *last_vec_mut.last_mut().unwrap() =
341 (TokenTree::Token(glued_tok), *is_joint);
343 // Remove the first token tree from `stream`. (This
344 // is almost always the only tree in `stream`.)
345 let stream_vec_mut = Lrc::make_mut(stream_lrc);
346 stream_vec_mut.remove(0);
348 // Don't push `stream` if it's empty -- that could
349 // block subsequent token gluing, by getting
350 // between two token trees that should be glued
352 if !stream.is_empty() {
363 pub fn build(self) -> TokenStream {
364 TokenStream::from_streams(self.0)
370 pub stream: TokenStream,
374 impl Iterator for Cursor {
375 type Item = TokenTree;
377 fn next(&mut self) -> Option<TokenTree> {
378 self.next_with_joint().map(|(tree, _)| tree)
383 fn new(stream: TokenStream) -> Self {
384 Cursor { stream, index: 0 }
387 pub fn next_with_joint(&mut self) -> Option<TreeAndJoint> {
388 if self.index < self.stream.len() {
390 Some(self.stream.0[self.index - 1].clone())
396 pub fn append(&mut self, new_stream: TokenStream) {
397 if new_stream.is_empty() {
400 let index = self.index;
401 let stream = mem::take(&mut self.stream);
402 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
406 pub fn look_ahead(&self, n: usize) -> Option<TokenTree> {
407 self.stream.0[self.index..].get(n).map(|(tree, _)| tree.clone())
411 #[derive(Debug, Copy, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
412 pub struct DelimSpan {
418 pub fn from_single(sp: Span) -> Self {
419 DelimSpan { open: sp, close: sp }
422 pub fn from_pair(open: Span, close: Span) -> Self {
423 DelimSpan { open, close }
426 pub fn dummy() -> Self {
427 Self::from_single(DUMMY_SP)
430 pub fn entire(self) -> Span {
431 self.open.with_hi(self.close.hi())