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};
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
196 /// backwards compatability.
197 #[derive(Clone, Debug, Default, Encodable, Decodable)]
198 pub struct TokenStream(pub(crate) Lrc<Vec<TreeAndSpacing>>);
200 pub type TreeAndSpacing = (TokenTree, Spacing);
202 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
203 #[cfg(target_arch = "x86_64")]
204 rustc_data_structures::static_assert_size!(TokenStream, 8);
206 #[derive(Clone, Copy, Debug, PartialEq, Encodable, Decodable)]
213 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
214 /// separating the two arguments with a comma for diagnostic suggestions.
215 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
216 // Used to suggest if a user writes `foo!(a b);`
217 let mut suggestion = None;
218 let mut iter = self.0.iter().enumerate().peekable();
219 while let Some((pos, ts)) = iter.next() {
220 if let Some((_, next)) = iter.peek() {
221 let sp = match (&ts, &next) {
222 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
224 (TokenTree::Token(token_left), Spacing::Alone),
225 (TokenTree::Token(token_right), _),
226 ) if ((token_left.is_ident() && !token_left.is_reserved_ident())
227 || token_left.is_lit())
228 && ((token_right.is_ident() && !token_right.is_reserved_ident())
229 || token_right.is_lit()) =>
233 ((TokenTree::Delimited(sp, ..), Spacing::Alone), _) => sp.entire(),
236 let sp = sp.shrink_to_hi();
237 let comma = (TokenTree::token(token::Comma, sp), Spacing::Alone);
238 suggestion = Some((pos, comma, sp));
241 if let Some((pos, comma, sp)) = suggestion {
242 let mut new_stream = Vec::with_capacity(self.0.len() + 1);
243 let parts = self.0.split_at(pos + 1);
244 new_stream.extend_from_slice(parts.0);
245 new_stream.push(comma);
246 new_stream.extend_from_slice(parts.1);
247 return Some((TokenStream::new(new_stream), sp));
253 impl From<TokenTree> for TokenStream {
254 fn from(tree: TokenTree) -> TokenStream {
255 TokenStream::new(vec![(tree, Spacing::Alone)])
259 impl From<TokenTree> for TreeAndSpacing {
260 fn from(tree: TokenTree) -> TreeAndSpacing {
261 (tree, Spacing::Alone)
265 impl iter::FromIterator<TokenTree> for TokenStream {
266 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
267 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndSpacing>>())
271 impl Eq for TokenStream {}
273 impl PartialEq<TokenStream> for TokenStream {
274 fn eq(&self, other: &TokenStream) -> bool {
275 self.trees().eq(other.trees())
280 pub fn new(streams: Vec<TreeAndSpacing>) -> TokenStream {
281 TokenStream(Lrc::new(streams))
284 pub fn is_empty(&self) -> bool {
288 pub fn len(&self) -> usize {
292 pub fn span(&self) -> Option<Span> {
295 [(tt, _)] => Some(tt.span()),
296 [(tt_start, _), .., (tt_end, _)] => Some(tt_start.span().to(tt_end.span())),
300 pub fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
301 match streams.len() {
302 0 => TokenStream::default(),
303 1 => streams.pop().unwrap(),
305 // We are going to extend the first stream in `streams` with
306 // the elements from the subsequent streams. This requires
307 // using `make_mut()` on the first stream, and in practice this
308 // doesn't cause cloning 99.9% of the time.
310 // One very common use case is when `streams` has two elements,
311 // where the first stream has any number of elements within
312 // (often 1, but sometimes many more) and the second stream has
313 // a single element within.
315 // Determine how much the first stream will be extended.
316 // Needed to avoid quadratic blow up from on-the-fly
317 // reallocations (#57735).
318 let num_appends = streams.iter().skip(1).map(|ts| ts.len()).sum();
320 // Get the first stream. If it's `None`, create an empty
322 let mut iter = streams.drain(..);
323 let mut first_stream_lrc = iter.next().unwrap().0;
325 // Append the elements to the first stream, after reserving
327 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
328 first_vec_mut.reserve(num_appends);
330 first_vec_mut.extend(stream.0.iter().cloned());
333 // Create the final `TokenStream`.
334 TokenStream(first_stream_lrc)
339 pub fn trees_ref(&self) -> CursorRef<'_> {
343 pub fn trees(&self) -> Cursor {
344 self.clone().into_trees()
347 pub fn into_trees(self) -> Cursor {
351 /// Compares two `TokenStream`s, checking equality without regarding span information.
352 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
353 let mut t1 = self.trees();
354 let mut t2 = other.trees();
355 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
356 if !t1.eq_unspanned(&t2) {
360 t1.next().is_none() && t2.next().is_none()
363 pub fn map_enumerated<F: FnMut(usize, &TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
364 TokenStream(Lrc::new(
368 .map(|(i, (tree, is_joint))| (f(i, tree), *is_joint))
374 // 99.5%+ of the time we have 1 or 2 elements in this vector.
376 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
378 impl TokenStreamBuilder {
379 pub fn new() -> TokenStreamBuilder {
380 TokenStreamBuilder(SmallVec::new())
383 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
384 let mut stream = stream.into();
386 // If `self` is not empty and the last tree within the last stream is a
387 // token tree marked with `Joint`...
388 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
389 if let Some((TokenTree::Token(last_token), Spacing::Joint)) = last_stream_lrc.last() {
390 // ...and `stream` is not empty and the first tree within it is
392 let TokenStream(ref mut stream_lrc) = stream;
393 if let Some((TokenTree::Token(token), spacing)) = stream_lrc.first() {
394 // ...and the two tokens can be glued together...
395 if let Some(glued_tok) = last_token.glue(&token) {
396 // ...then do so, by overwriting the last token
397 // tree in `self` and removing the first token tree
398 // from `stream`. This requires using `make_mut()`
399 // on the last stream in `self` and on `stream`,
400 // and in practice this doesn't cause cloning 99.9%
403 // Overwrite the last token tree with the merged
405 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
406 *last_vec_mut.last_mut().unwrap() = (TokenTree::Token(glued_tok), *spacing);
408 // Remove the first token tree from `stream`. (This
409 // is almost always the only tree in `stream`.)
410 let stream_vec_mut = Lrc::make_mut(stream_lrc);
411 stream_vec_mut.remove(0);
413 // Don't push `stream` if it's empty -- that could
414 // block subsequent token gluing, by getting
415 // between two token trees that should be glued
417 if !stream.is_empty() {
428 pub fn build(self) -> TokenStream {
429 TokenStream::from_streams(self.0)
433 /// By-reference iterator over a [`TokenStream`].
435 pub struct CursorRef<'t> {
436 stream: &'t TokenStream,
440 impl<'t> CursorRef<'t> {
441 fn new(stream: &TokenStream) -> CursorRef<'_> {
442 CursorRef { stream, index: 0 }
445 fn next_with_spacing(&mut self) -> Option<&'t TreeAndSpacing> {
446 self.stream.0.get(self.index).map(|tree| {
453 impl<'t> Iterator for CursorRef<'t> {
454 type Item = &'t TokenTree;
456 fn next(&mut self) -> Option<&'t TokenTree> {
457 self.next_with_spacing().map(|(tree, _)| tree)
461 /// Owning by-value iterator over a [`TokenStream`].
462 // FIXME: Many uses of this can be replaced with by-reference iterator to avoid clones.
465 pub stream: TokenStream,
469 impl Iterator for Cursor {
470 type Item = TokenTree;
472 fn next(&mut self) -> Option<TokenTree> {
473 self.next_with_spacing().map(|(tree, _)| tree)
478 fn new(stream: TokenStream) -> Self {
479 Cursor { stream, index: 0 }
482 pub fn next_with_spacing(&mut self) -> Option<TreeAndSpacing> {
483 if self.index < self.stream.len() {
485 Some(self.stream.0[self.index - 1].clone())
491 pub fn append(&mut self, new_stream: TokenStream) {
492 if new_stream.is_empty() {
495 let index = self.index;
496 let stream = mem::take(&mut self.stream);
497 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
501 pub fn look_ahead(&self, n: usize) -> Option<&TokenTree> {
502 self.stream.0[self.index..].get(n).map(|(tree, _)| tree)
506 #[derive(Debug, Copy, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
507 pub struct DelimSpan {
513 pub fn from_single(sp: Span) -> Self {
514 DelimSpan { open: sp, close: sp }
517 pub fn from_pair(open: Span, close: Span) -> Self {
518 DelimSpan { open, close }
521 pub fn dummy() -> Self {
522 Self::from_single(DUMMY_SP)
525 pub fn entire(self) -> Span {
526 self.open.with_hi(self.close.hi())