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::{self, Lrc};
20 use rustc_macros::HashStable_Generic;
21 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
22 use rustc_span::{Span, DUMMY_SP};
23 use smallvec::{smallvec, SmallVec};
25 use std::{fmt, iter, mem};
27 /// When the main rust parser encounters a syntax-extension invocation, it
28 /// parses the arguments to the invocation as a token-tree. This is a very
29 /// loose structure, such that all sorts of different AST-fragments can
30 /// be passed to syntax extensions using a uniform type.
32 /// If the syntax extension is an MBE macro, it will attempt to match its
33 /// LHS token tree against the provided token tree, and if it finds a
34 /// match, will transcribe the RHS token tree, splicing in any captured
35 /// `macro_parser::matched_nonterminals` into the `SubstNt`s it finds.
37 /// The RHS of an MBE macro is the only place `SubstNt`s are substituted.
38 /// Nothing special happens to misnamed or misplaced `SubstNt`s.
39 #[derive(Debug, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
43 /// A delimited sequence of token trees
44 Delimited(DelimSpan, DelimToken, TokenStream),
47 #[derive(Copy, Clone)]
48 pub enum CanSynthesizeMissingTokens {
53 // Ensure all fields of `TokenTree` is `Send` and `Sync`.
54 #[cfg(parallel_compiler)]
58 DelimSpan: Send + Sync,
59 DelimToken: Send + Sync,
60 TokenStream: Send + Sync,
65 /// Checks if this TokenTree is equal to the other, regardless of span information.
66 pub fn eq_unspanned(&self, other: &TokenTree) -> bool {
68 (TokenTree::Token(token), TokenTree::Token(token2)) => token.kind == token2.kind,
69 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
70 delim == delim2 && tts.eq_unspanned(&tts2)
76 /// Retrieves the TokenTree's span.
77 pub fn span(&self) -> Span {
79 TokenTree::Token(token) => token.span,
80 TokenTree::Delimited(sp, ..) => sp.entire(),
84 /// Modify the `TokenTree`'s span in-place.
85 pub fn set_span(&mut self, span: Span) {
87 TokenTree::Token(token) => token.span = span,
88 TokenTree::Delimited(dspan, ..) => *dspan = DelimSpan::from_single(span),
92 pub fn joint(self) -> TokenStream {
93 TokenStream::new(vec![(self, Spacing::Joint)])
96 pub fn token(kind: TokenKind, span: Span) -> TokenTree {
97 TokenTree::Token(Token::new(kind, span))
100 /// Returns the opening delimiter as a token tree.
101 pub fn open_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
102 TokenTree::token(token::OpenDelim(delim), span.open)
105 /// Returns the closing delimiter as a token tree.
106 pub fn close_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
107 TokenTree::token(token::CloseDelim(delim), span.close)
110 pub fn uninterpolate(self) -> TokenTree {
112 TokenTree::Token(token) => TokenTree::Token(token.uninterpolate().into_owned()),
118 impl<CTX> HashStable<CTX> for TokenStream
120 CTX: crate::HashStableContext,
122 fn hash_stable(&self, hcx: &mut CTX, hasher: &mut StableHasher) {
123 for sub_tt in self.trees() {
124 sub_tt.hash_stable(hcx, hasher);
129 pub trait CreateTokenStream: sync::Send + sync::Sync {
130 fn add_trailing_semi(&self) -> Box<dyn CreateTokenStream>;
131 fn create_token_stream(&self) -> TokenStream;
134 impl CreateTokenStream for TokenStream {
135 fn add_trailing_semi(&self) -> Box<dyn CreateTokenStream> {
136 panic!("Cannot call `add_trailing_semi` on a `TokenStream`!");
138 fn create_token_stream(&self) -> TokenStream {
143 /// A lazy version of `TokenStream`, which defers creation
144 /// of an actual `TokenStream` until it is needed.
145 /// `Box` is here only to reduce the structure size.
147 pub struct LazyTokenStream(Lrc<Box<dyn CreateTokenStream>>);
149 impl LazyTokenStream {
150 pub fn new(inner: impl CreateTokenStream + 'static) -> LazyTokenStream {
151 LazyTokenStream(Lrc::new(Box::new(inner)))
154 /// Extends the captured stream by one token,
155 /// which must be a trailing semicolon. This
156 /// affects the `TokenStream` created by `make_tokenstream`.
157 pub fn add_trailing_semi(&self) -> LazyTokenStream {
158 LazyTokenStream(Lrc::new(self.0.add_trailing_semi()))
161 pub fn create_token_stream(&self) -> TokenStream {
162 self.0.create_token_stream()
166 impl fmt::Debug for LazyTokenStream {
167 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
168 fmt::Debug::fmt("LazyTokenStream", f)
172 impl<S: Encoder> Encodable<S> for LazyTokenStream {
173 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
174 // Used by AST json printing.
175 Encodable::encode(&self.create_token_stream(), s)
179 impl<D: Decoder> Decodable<D> for LazyTokenStream {
180 fn decode(_d: &mut D) -> Result<Self, D::Error> {
181 panic!("Attempted to decode LazyTokenStream");
185 impl<CTX> HashStable<CTX> for LazyTokenStream {
186 fn hash_stable(&self, _hcx: &mut CTX, _hasher: &mut StableHasher) {
187 panic!("Attempted to compute stable hash for LazyTokenStream");
191 /// A `TokenStream` is an abstract sequence of tokens, organized into `TokenTree`s.
193 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
194 /// instead of a representation of the abstract syntax tree.
195 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for back-compat.
196 #[derive(Clone, Debug, Default, Encodable, Decodable)]
197 pub struct TokenStream(pub(crate) Lrc<Vec<TreeAndSpacing>>);
199 pub type TreeAndSpacing = (TokenTree, Spacing);
201 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
202 #[cfg(target_arch = "x86_64")]
203 rustc_data_structures::static_assert_size!(TokenStream, 8);
205 #[derive(Clone, Copy, Debug, PartialEq, Encodable, Decodable)]
212 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
213 /// separating the two arguments with a comma for diagnostic suggestions.
214 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
215 // Used to suggest if a user writes `foo!(a b);`
216 let mut suggestion = None;
217 let mut iter = self.0.iter().enumerate().peekable();
218 while let Some((pos, ts)) = iter.next() {
219 if let Some((_, next)) = iter.peek() {
220 let sp = match (&ts, &next) {
221 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
223 (TokenTree::Token(token_left), Spacing::Alone),
224 (TokenTree::Token(token_right), _),
225 ) if ((token_left.is_ident() && !token_left.is_reserved_ident())
226 || token_left.is_lit())
227 && ((token_right.is_ident() && !token_right.is_reserved_ident())
228 || token_right.is_lit()) =>
232 ((TokenTree::Delimited(sp, ..), Spacing::Alone), _) => sp.entire(),
235 let sp = sp.shrink_to_hi();
236 let comma = (TokenTree::token(token::Comma, sp), Spacing::Alone);
237 suggestion = Some((pos, comma, sp));
240 if let Some((pos, comma, sp)) = suggestion {
241 let mut new_stream = Vec::with_capacity(self.0.len() + 1);
242 let parts = self.0.split_at(pos + 1);
243 new_stream.extend_from_slice(parts.0);
244 new_stream.push(comma);
245 new_stream.extend_from_slice(parts.1);
246 return Some((TokenStream::new(new_stream), sp));
252 impl From<TokenTree> for TokenStream {
253 fn from(tree: TokenTree) -> TokenStream {
254 TokenStream::new(vec![(tree, Spacing::Alone)])
258 impl From<TokenTree> for TreeAndSpacing {
259 fn from(tree: TokenTree) -> TreeAndSpacing {
260 (tree, Spacing::Alone)
264 impl iter::FromIterator<TokenTree> for TokenStream {
265 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
266 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndSpacing>>())
270 impl Eq for TokenStream {}
272 impl PartialEq<TokenStream> for TokenStream {
273 fn eq(&self, other: &TokenStream) -> bool {
274 self.trees().eq(other.trees())
279 pub fn new(streams: Vec<TreeAndSpacing>) -> TokenStream {
280 TokenStream(Lrc::new(streams))
283 pub fn is_empty(&self) -> bool {
287 pub fn len(&self) -> usize {
291 pub fn span(&self) -> Option<Span> {
294 [(tt, _)] => Some(tt.span()),
295 [(tt_start, _), .., (tt_end, _)] => Some(tt_start.span().to(tt_end.span())),
299 pub fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
300 match streams.len() {
301 0 => TokenStream::default(),
302 1 => streams.pop().unwrap(),
304 // We are going to extend the first stream in `streams` with
305 // the elements from the subsequent streams. This requires
306 // using `make_mut()` on the first stream, and in practice this
307 // doesn't cause cloning 99.9% of the time.
309 // One very common use case is when `streams` has two elements,
310 // where the first stream has any number of elements within
311 // (often 1, but sometimes many more) and the second stream has
312 // a single element within.
314 // Determine how much the first stream will be extended.
315 // Needed to avoid quadratic blow up from on-the-fly
316 // reallocations (#57735).
317 let num_appends = streams.iter().skip(1).map(|ts| ts.len()).sum();
319 // Get the first stream. If it's `None`, create an empty
321 let mut iter = streams.drain(..);
322 let mut first_stream_lrc = iter.next().unwrap().0;
324 // Append the elements to the first stream, after reserving
326 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
327 first_vec_mut.reserve(num_appends);
329 first_vec_mut.extend(stream.0.iter().cloned());
332 // Create the final `TokenStream`.
333 TokenStream(first_stream_lrc)
338 pub fn trees_ref(&self) -> CursorRef<'_> {
342 pub fn trees(&self) -> Cursor {
343 self.clone().into_trees()
346 pub fn into_trees(self) -> Cursor {
350 /// Compares two `TokenStream`s, checking equality without regarding span information.
351 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
352 let mut t1 = self.trees();
353 let mut t2 = other.trees();
354 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
355 if !t1.eq_unspanned(&t2) {
359 t1.next().is_none() && t2.next().is_none()
362 pub fn map_enumerated<F: FnMut(usize, &TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
363 TokenStream(Lrc::new(
367 .map(|(i, (tree, is_joint))| (f(i, tree), *is_joint))
373 // 99.5%+ of the time we have 1 or 2 elements in this vector.
375 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
377 impl TokenStreamBuilder {
378 pub fn new() -> TokenStreamBuilder {
379 TokenStreamBuilder(SmallVec::new())
382 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
383 let mut stream = stream.into();
385 // If `self` is not empty and the last tree within the last stream is a
386 // token tree marked with `Joint`...
387 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
388 if let Some((TokenTree::Token(last_token), Spacing::Joint)) = last_stream_lrc.last() {
389 // ...and `stream` is not empty and the first tree within it is
391 let TokenStream(ref mut stream_lrc) = stream;
392 if let Some((TokenTree::Token(token), spacing)) = stream_lrc.first() {
393 // ...and the two tokens can be glued together...
394 if let Some(glued_tok) = last_token.glue(&token) {
395 // ...then do so, by overwriting the last token
396 // tree in `self` and removing the first token tree
397 // from `stream`. This requires using `make_mut()`
398 // on the last stream in `self` and on `stream`,
399 // and in practice this doesn't cause cloning 99.9%
402 // Overwrite the last token tree with the merged
404 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
405 *last_vec_mut.last_mut().unwrap() = (TokenTree::Token(glued_tok), *spacing);
407 // Remove the first token tree from `stream`. (This
408 // is almost always the only tree in `stream`.)
409 let stream_vec_mut = Lrc::make_mut(stream_lrc);
410 stream_vec_mut.remove(0);
412 // Don't push `stream` if it's empty -- that could
413 // block subsequent token gluing, by getting
414 // between two token trees that should be glued
416 if !stream.is_empty() {
427 pub fn build(self) -> TokenStream {
428 TokenStream::from_streams(self.0)
432 /// By-reference iterator over a `TokenStream`.
434 pub struct CursorRef<'t> {
435 stream: &'t TokenStream,
439 impl<'t> CursorRef<'t> {
440 fn new(stream: &TokenStream) -> CursorRef<'_> {
441 CursorRef { stream, index: 0 }
444 fn next_with_spacing(&mut self) -> Option<&'t TreeAndSpacing> {
445 self.stream.0.get(self.index).map(|tree| {
452 impl<'t> Iterator for CursorRef<'t> {
453 type Item = &'t TokenTree;
455 fn next(&mut self) -> Option<&'t TokenTree> {
456 self.next_with_spacing().map(|(tree, _)| tree)
460 /// Owning by-value iterator over a `TokenStream`.
461 /// FIXME: Many uses of this can be replaced with by-reference iterator to avoid clones.
464 pub stream: TokenStream,
468 impl Iterator for Cursor {
469 type Item = TokenTree;
471 fn next(&mut self) -> Option<TokenTree> {
472 self.next_with_spacing().map(|(tree, _)| tree)
477 fn new(stream: TokenStream) -> Self {
478 Cursor { stream, index: 0 }
481 pub fn next_with_spacing(&mut self) -> Option<TreeAndSpacing> {
482 if self.index < self.stream.len() {
484 Some(self.stream.0[self.index - 1].clone())
490 pub fn append(&mut self, new_stream: TokenStream) {
491 if new_stream.is_empty() {
494 let index = self.index;
495 let stream = mem::take(&mut self.stream);
496 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
500 pub fn look_ahead(&self, n: usize) -> Option<&TokenTree> {
501 self.stream.0[self.index..].get(n).map(|(tree, _)| tree)
505 #[derive(Debug, Copy, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
506 pub struct DelimSpan {
512 pub fn from_single(sp: Span) -> Self {
513 DelimSpan { open: sp, close: sp }
516 pub fn from_pair(open: Span, close: Span) -> Self {
517 DelimSpan { open, close }
520 pub fn dummy() -> Self {
521 Self::from_single(DUMMY_SP)
524 pub fn entire(self) -> Span {
525 self.open.with_hi(self.close.hi())