3 //! `TokenStream`s represent syntactic objects before they are converted into ASTs.
4 //! A `TokenStream` is, roughly speaking, a sequence 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};
19 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
20 use rustc_data_structures::sync::{self, Lrc};
21 use rustc_macros::HashStable_Generic;
22 use rustc_serialize::{Decodable, Decoder, Encodable, Encoder};
23 use rustc_span::{Span, DUMMY_SP};
24 use smallvec::{smallvec, SmallVec};
26 use std::{fmt, iter, mem};
28 /// When the main Rust parser encounters a syntax-extension invocation, it
29 /// parses the arguments to the invocation as a token tree. This is a very
30 /// loose structure, such that all sorts of different AST fragments can
31 /// be passed to syntax extensions using a uniform type.
33 /// If the syntax extension is an MBE macro, it will attempt to match its
34 /// LHS token tree against the provided token tree, and if it finds a
35 /// match, will transcribe the RHS token tree, splicing in any captured
36 /// `macro_parser::matched_nonterminals` into the `SubstNt`s it finds.
38 /// The RHS of an MBE macro is the only place `SubstNt`s are substituted.
39 /// Nothing special happens to misnamed or misplaced `SubstNt`s.
40 #[derive(Debug, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
44 /// A delimited sequence of token trees.
45 Delimited(DelimSpan, DelimToken, TokenStream),
48 #[derive(Copy, Clone)]
49 pub enum CanSynthesizeMissingTokens {
54 // Ensure all fields of `TokenTree` is `Send` and `Sync`.
55 #[cfg(parallel_compiler)]
59 DelimSpan: Send + Sync,
60 DelimToken: Send + Sync,
61 TokenStream: Send + Sync,
66 /// Checks if this `TokenTree` is equal to the other, regardless of span information.
67 pub fn eq_unspanned(&self, other: &TokenTree) -> bool {
69 (TokenTree::Token(token), TokenTree::Token(token2)) => token.kind == token2.kind,
70 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
71 delim == delim2 && tts.eq_unspanned(&tts2)
77 /// Retrieves the `TokenTree`'s span.
78 pub fn span(&self) -> Span {
80 TokenTree::Token(token) => token.span,
81 TokenTree::Delimited(sp, ..) => sp.entire(),
85 /// Modify the `TokenTree`'s span in-place.
86 pub fn set_span(&mut self, span: Span) {
88 TokenTree::Token(token) => token.span = span,
89 TokenTree::Delimited(dspan, ..) => *dspan = DelimSpan::from_single(span),
93 pub fn token(kind: TokenKind, span: Span) -> TokenTree {
94 TokenTree::Token(Token::new(kind, span))
97 /// Returns the opening delimiter as a token tree.
98 pub fn open_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
99 TokenTree::token(token::OpenDelim(delim), span.open)
102 /// Returns the closing delimiter as a token tree.
103 pub fn close_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
104 TokenTree::token(token::CloseDelim(delim), span.close)
107 pub fn uninterpolate(self) -> TokenTree {
109 TokenTree::Token(token) => TokenTree::Token(token.uninterpolate().into_owned()),
115 impl<CTX> HashStable<CTX> for TokenStream
117 CTX: crate::HashStableContext,
119 fn hash_stable(&self, hcx: &mut CTX, hasher: &mut StableHasher) {
120 for sub_tt in self.trees() {
121 sub_tt.hash_stable(hcx, hasher);
126 pub trait CreateTokenStream: sync::Send + sync::Sync {
127 fn create_token_stream(&self) -> AttrAnnotatedTokenStream;
130 impl CreateTokenStream for AttrAnnotatedTokenStream {
131 fn create_token_stream(&self) -> AttrAnnotatedTokenStream {
136 /// A lazy version of [`TokenStream`], which defers creation
137 /// of an actual `TokenStream` until it is needed.
138 /// `Box` is here only to reduce the structure size.
140 pub struct LazyTokenStream(Lrc<Box<dyn CreateTokenStream>>);
142 impl LazyTokenStream {
143 pub fn new(inner: impl CreateTokenStream + 'static) -> LazyTokenStream {
144 LazyTokenStream(Lrc::new(Box::new(inner)))
147 pub fn create_token_stream(&self) -> AttrAnnotatedTokenStream {
148 self.0.create_token_stream()
152 impl fmt::Debug for LazyTokenStream {
153 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
154 write!(f, "LazyTokenStream({:?})", self.create_token_stream())
158 impl<S: Encoder> Encodable<S> for LazyTokenStream {
159 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
160 // Used by AST json printing.
161 Encodable::encode(&self.create_token_stream(), s)
165 impl<D: Decoder> Decodable<D> for LazyTokenStream {
166 fn decode(_d: &mut D) -> Self {
167 panic!("Attempted to decode LazyTokenStream");
171 impl<CTX> HashStable<CTX> for LazyTokenStream {
172 fn hash_stable(&self, _hcx: &mut CTX, _hasher: &mut StableHasher) {
173 panic!("Attempted to compute stable hash for LazyTokenStream");
177 /// A `AttrAnnotatedTokenStream` is similar to a `TokenStream`, but with extra
178 /// information about the tokens for attribute targets. This is used
179 /// during expansion to perform early cfg-expansion, and to process attributes
180 /// during proc-macro invocations.
181 #[derive(Clone, Debug, Default, Encodable, Decodable)]
182 pub struct AttrAnnotatedTokenStream(pub Lrc<Vec<(AttrAnnotatedTokenTree, Spacing)>>);
184 /// Like `TokenTree`, but for `AttrAnnotatedTokenStream`
185 #[derive(Clone, Debug, Encodable, Decodable)]
186 pub enum AttrAnnotatedTokenTree {
188 Delimited(DelimSpan, DelimToken, AttrAnnotatedTokenStream),
189 /// Stores the attributes for an attribute target,
190 /// along with the tokens for that attribute target.
191 /// See `AttributesData` for more information
192 Attributes(AttributesData),
195 impl AttrAnnotatedTokenStream {
196 pub fn new(tokens: Vec<(AttrAnnotatedTokenTree, Spacing)>) -> AttrAnnotatedTokenStream {
197 AttrAnnotatedTokenStream(Lrc::new(tokens))
200 /// Converts this `AttrAnnotatedTokenStream` to a plain `TokenStream
201 /// During conversion, `AttrAnnotatedTokenTree::Attributes` get 'flattened'
202 /// back to a `TokenStream` of the form `outer_attr attr_target`.
203 /// If there are inner attributes, they are inserted into the proper
204 /// place in the attribute target tokens.
205 pub fn to_tokenstream(&self) -> TokenStream {
206 let trees: Vec<_> = self
209 .flat_map(|tree| match &tree.0 {
210 AttrAnnotatedTokenTree::Token(inner) => {
211 smallvec![(TokenTree::Token(inner.clone()), tree.1)].into_iter()
213 AttrAnnotatedTokenTree::Delimited(span, delim, stream) => smallvec![(
214 TokenTree::Delimited(*span, *delim, stream.to_tokenstream()),
218 AttrAnnotatedTokenTree::Attributes(data) => {
219 let mut outer_attrs = Vec::new();
220 let mut inner_attrs = Vec::new();
221 for attr in &data.attrs {
223 crate::AttrStyle::Outer => {
224 outer_attrs.push(attr);
226 crate::AttrStyle::Inner => {
227 inner_attrs.push(attr);
232 let mut target_tokens: Vec<_> = data
234 .create_token_stream()
240 if !inner_attrs.is_empty() {
241 let mut found = false;
242 // Check the last two trees (to account for a trailing semi)
243 for (tree, _) in target_tokens.iter_mut().rev().take(2) {
244 if let TokenTree::Delimited(span, delim, delim_tokens) = tree {
245 // Inner attributes are only supported on extern blocks, functions, impls,
246 // and modules. All of these have their inner attributes placed at
247 // the beginning of the rightmost outermost braced group:
248 // e.g. fn foo() { #![my_attr} }
250 // Therefore, we can insert them back into the right location
251 // without needing to do any extra position tracking.
253 // Note: Outline modules are an exception - they can
254 // have attributes like `#![my_attr]` at the start of a file.
255 // Support for custom attributes in this position is not
256 // properly implemented - we always synthesize fake tokens,
257 // so we never reach this code.
259 let mut builder = TokenStreamBuilder::new();
260 for inner_attr in inner_attrs {
261 builder.push(inner_attr.tokens().to_tokenstream());
263 builder.push(delim_tokens.clone());
264 *tree = TokenTree::Delimited(*span, *delim, builder.build());
272 "Failed to find trailing delimited group in: {:?}",
276 let mut flat: SmallVec<[_; 1]> = SmallVec::new();
277 for attr in outer_attrs {
278 // FIXME: Make this more efficient
279 flat.extend(attr.tokens().to_tokenstream().0.clone().iter().cloned());
281 flat.extend(target_tokens);
286 TokenStream::new(trees)
290 /// Stores the tokens for an attribute target, along
291 /// with its attributes.
293 /// This is constructed during parsing when we need to capture
296 /// For example, `#[cfg(FALSE)] struct Foo {}` would
297 /// have an `attrs` field containing the `#[cfg(FALSE)]` attr,
298 /// and a `tokens` field storing the (unparsed) tokens `struct Foo {}`
299 #[derive(Clone, Debug, Encodable, Decodable)]
300 pub struct AttributesData {
301 /// Attributes, both outer and inner.
302 /// These are stored in the original order that they were parsed in.
304 /// The underlying tokens for the attribute target that `attrs`
306 pub tokens: LazyTokenStream,
309 /// A `TokenStream` is an abstract sequence of tokens, organized into [`TokenTree`]s.
311 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
312 /// instead of a representation of the abstract syntax tree.
313 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for
314 /// backwards compatibility.
315 #[derive(Clone, Debug, Default, Encodable, Decodable)]
316 pub struct TokenStream(pub(crate) Lrc<Vec<TreeAndSpacing>>);
318 pub type TreeAndSpacing = (TokenTree, Spacing);
320 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
321 #[cfg(all(target_arch = "x86_64", target_pointer_width = "64"))]
322 rustc_data_structures::static_assert_size!(TokenStream, 8);
324 #[derive(Clone, Copy, Debug, PartialEq, Encodable, Decodable)]
331 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
332 /// separating the two arguments with a comma for diagnostic suggestions.
333 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
334 // Used to suggest if a user writes `foo!(a b);`
335 let mut suggestion = None;
336 let mut iter = self.0.iter().enumerate().peekable();
337 while let Some((pos, ts)) = iter.next() {
338 if let Some((_, next)) = iter.peek() {
339 let sp = match (&ts, &next) {
340 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
342 (TokenTree::Token(token_left), Spacing::Alone),
343 (TokenTree::Token(token_right), _),
344 ) if ((token_left.is_ident() && !token_left.is_reserved_ident())
345 || token_left.is_lit())
346 && ((token_right.is_ident() && !token_right.is_reserved_ident())
347 || token_right.is_lit()) =>
351 ((TokenTree::Delimited(sp, ..), Spacing::Alone), _) => sp.entire(),
354 let sp = sp.shrink_to_hi();
355 let comma = (TokenTree::token(token::Comma, sp), Spacing::Alone);
356 suggestion = Some((pos, comma, sp));
359 if let Some((pos, comma, sp)) = suggestion {
360 let mut new_stream = Vec::with_capacity(self.0.len() + 1);
361 let parts = self.0.split_at(pos + 1);
362 new_stream.extend_from_slice(parts.0);
363 new_stream.push(comma);
364 new_stream.extend_from_slice(parts.1);
365 return Some((TokenStream::new(new_stream), sp));
371 impl From<(AttrAnnotatedTokenTree, Spacing)> for AttrAnnotatedTokenStream {
372 fn from((tree, spacing): (AttrAnnotatedTokenTree, Spacing)) -> AttrAnnotatedTokenStream {
373 AttrAnnotatedTokenStream::new(vec![(tree, spacing)])
377 impl From<TokenTree> for TokenStream {
378 fn from(tree: TokenTree) -> TokenStream {
379 TokenStream::new(vec![(tree, Spacing::Alone)])
383 impl From<TokenTree> for TreeAndSpacing {
384 fn from(tree: TokenTree) -> TreeAndSpacing {
385 (tree, Spacing::Alone)
389 impl iter::FromIterator<TokenTree> for TokenStream {
390 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
391 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndSpacing>>())
395 impl Eq for TokenStream {}
397 impl PartialEq<TokenStream> for TokenStream {
398 fn eq(&self, other: &TokenStream) -> bool {
399 self.trees().eq(other.trees())
404 pub fn new(streams: Vec<TreeAndSpacing>) -> TokenStream {
405 TokenStream(Lrc::new(streams))
408 pub fn is_empty(&self) -> bool {
412 pub fn len(&self) -> usize {
416 pub fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
417 match streams.len() {
418 0 => TokenStream::default(),
419 1 => streams.pop().unwrap(),
421 // We are going to extend the first stream in `streams` with
422 // the elements from the subsequent streams. This requires
423 // using `make_mut()` on the first stream, and in practice this
424 // doesn't cause cloning 99.9% of the time.
426 // One very common use case is when `streams` has two elements,
427 // where the first stream has any number of elements within
428 // (often 1, but sometimes many more) and the second stream has
429 // a single element within.
431 // Determine how much the first stream will be extended.
432 // Needed to avoid quadratic blow up from on-the-fly
433 // reallocations (#57735).
434 let num_appends = streams.iter().skip(1).map(|ts| ts.len()).sum();
436 // Get the first stream. If it's `None`, create an empty
438 let mut iter = streams.drain(..);
439 let mut first_stream_lrc = iter.next().unwrap().0;
441 // Append the elements to the first stream, after reserving
443 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
444 first_vec_mut.reserve(num_appends);
446 first_vec_mut.extend(stream.0.iter().cloned());
449 // Create the final `TokenStream`.
450 TokenStream(first_stream_lrc)
455 pub fn trees(&self) -> Cursor {
456 self.clone().into_trees()
459 pub fn into_trees(self) -> Cursor {
463 /// Compares two `TokenStream`s, checking equality without regarding span information.
464 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
465 let mut t1 = self.trees();
466 let mut t2 = other.trees();
467 for (t1, t2) in iter::zip(&mut t1, &mut t2) {
468 if !t1.eq_unspanned(&t2) {
472 t1.next().is_none() && t2.next().is_none()
475 pub fn map_enumerated<F: FnMut(usize, &TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
476 TokenStream(Lrc::new(
480 .map(|(i, (tree, is_joint))| (f(i, tree), *is_joint))
486 // 99.5%+ of the time we have 1 or 2 elements in this vector.
488 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
490 impl TokenStreamBuilder {
491 pub fn new() -> TokenStreamBuilder {
492 TokenStreamBuilder(SmallVec::new())
495 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
496 let mut stream = stream.into();
498 // If `self` is not empty and the last tree within the last stream is a
499 // token tree marked with `Joint`...
500 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut()
501 && let Some((TokenTree::Token(last_token), Spacing::Joint)) = last_stream_lrc.last()
502 // ...and `stream` is not empty and the first tree within it is
504 && let TokenStream(ref mut stream_lrc) = stream
505 && let Some((TokenTree::Token(token), spacing)) = stream_lrc.first()
506 // ...and the two tokens can be glued together...
507 && let Some(glued_tok) = last_token.glue(&token)
509 // ...then do so, by overwriting the last token
510 // tree in `self` and removing the first token tree
511 // from `stream`. This requires using `make_mut()`
512 // on the last stream in `self` and on `stream`,
513 // and in practice this doesn't cause cloning 99.9%
516 // Overwrite the last token tree with the merged
518 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
519 *last_vec_mut.last_mut().unwrap() = (TokenTree::Token(glued_tok), *spacing);
521 // Remove the first token tree from `stream`. (This
522 // is almost always the only tree in `stream`.)
523 let stream_vec_mut = Lrc::make_mut(stream_lrc);
524 stream_vec_mut.remove(0);
526 // Don't push `stream` if it's empty -- that could
527 // block subsequent token gluing, by getting
528 // between two token trees that should be glued
530 if !stream.is_empty() {
538 pub fn build(self) -> TokenStream {
539 TokenStream::from_streams(self.0)
543 /// By-reference iterator over a [`TokenStream`].
545 pub struct CursorRef<'t> {
546 stream: &'t TokenStream,
550 impl<'t> CursorRef<'t> {
551 fn next_with_spacing(&mut self) -> Option<&'t TreeAndSpacing> {
552 self.stream.0.get(self.index).map(|tree| {
559 impl<'t> Iterator for CursorRef<'t> {
560 type Item = &'t TokenTree;
562 fn next(&mut self) -> Option<&'t TokenTree> {
563 self.next_with_spacing().map(|(tree, _)| tree)
567 /// Owning by-value iterator over a [`TokenStream`].
568 // FIXME: Many uses of this can be replaced with by-reference iterator to avoid clones.
571 pub stream: TokenStream,
575 impl Iterator for Cursor {
576 type Item = TokenTree;
578 fn next(&mut self) -> Option<TokenTree> {
579 self.next_with_spacing().map(|(tree, _)| tree)
584 fn new(stream: TokenStream) -> Self {
585 Cursor { stream, index: 0 }
588 pub fn next_with_spacing(&mut self) -> Option<TreeAndSpacing> {
589 if self.index < self.stream.len() {
591 Some(self.stream.0[self.index - 1].clone())
597 pub fn index(&self) -> usize {
601 pub fn append(&mut self, new_stream: TokenStream) {
602 if new_stream.is_empty() {
605 let index = self.index;
606 let stream = mem::take(&mut self.stream);
607 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
611 pub fn look_ahead(&self, n: usize) -> Option<&TokenTree> {
612 self.stream.0[self.index..].get(n).map(|(tree, _)| tree)
616 #[derive(Debug, Copy, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
617 pub struct DelimSpan {
623 pub fn from_single(sp: Span) -> Self {
624 DelimSpan { open: sp, close: sp }
627 pub fn from_pair(open: Span, close: Span) -> Self {
628 DelimSpan { open, close }
631 pub fn dummy() -> Self {
632 Self::from_single(DUMMY_SP)
635 pub fn entire(self) -> Span {
636 self.open.with_hi(self.close.hi())