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 token(kind: TokenKind, span: Span) -> TokenTree {
93 TokenTree::Token(Token::new(kind, span))
96 /// Returns the opening delimiter as a token tree.
97 pub fn open_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
98 TokenTree::token(token::OpenDelim(delim), span.open)
101 /// Returns the closing delimiter as a token tree.
102 pub fn close_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
103 TokenTree::token(token::CloseDelim(delim), span.close)
106 pub fn uninterpolate(self) -> TokenTree {
108 TokenTree::Token(token) => TokenTree::Token(token.uninterpolate().into_owned()),
114 impl<CTX> HashStable<CTX> for TokenStream
116 CTX: crate::HashStableContext,
118 fn hash_stable(&self, hcx: &mut CTX, hasher: &mut StableHasher) {
119 for sub_tt in self.trees() {
120 sub_tt.hash_stable(hcx, hasher);
125 pub trait CreateTokenStream: sync::Send + sync::Sync {
126 fn create_token_stream(&self) -> TokenStream;
129 impl CreateTokenStream for TokenStream {
130 fn create_token_stream(&self) -> TokenStream {
135 /// A lazy version of [`TokenStream`], which defers creation
136 /// of an actual `TokenStream` until it is needed.
137 /// `Box` is here only to reduce the structure size.
139 pub struct LazyTokenStream(Lrc<Box<dyn CreateTokenStream>>);
141 impl LazyTokenStream {
142 pub fn new(inner: impl CreateTokenStream + 'static) -> LazyTokenStream {
143 LazyTokenStream(Lrc::new(Box::new(inner)))
146 pub fn create_token_stream(&self) -> TokenStream {
147 self.0.create_token_stream()
151 impl fmt::Debug for LazyTokenStream {
152 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
153 fmt::Debug::fmt("LazyTokenStream", f)
157 impl<S: Encoder> Encodable<S> for LazyTokenStream {
158 fn encode(&self, s: &mut S) -> Result<(), S::Error> {
159 // Used by AST json printing.
160 Encodable::encode(&self.create_token_stream(), s)
164 impl<D: Decoder> Decodable<D> for LazyTokenStream {
165 fn decode(_d: &mut D) -> Result<Self, D::Error> {
166 panic!("Attempted to decode LazyTokenStream");
170 impl<CTX> HashStable<CTX> for LazyTokenStream {
171 fn hash_stable(&self, _hcx: &mut CTX, _hasher: &mut StableHasher) {
172 panic!("Attempted to compute stable hash for LazyTokenStream");
176 /// A `TokenStream` is an abstract sequence of tokens, organized into [`TokenTree`]s.
178 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
179 /// instead of a representation of the abstract syntax tree.
180 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for
181 /// backwards compatability.
182 #[derive(Clone, Debug, Default, Encodable, Decodable)]
183 pub struct TokenStream(pub(crate) Lrc<Vec<TreeAndSpacing>>);
185 pub type TreeAndSpacing = (TokenTree, Spacing);
187 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
188 #[cfg(all(target_arch = "x86_64", target_pointer_width = "64"))]
189 rustc_data_structures::static_assert_size!(TokenStream, 8);
191 #[derive(Clone, Copy, Debug, PartialEq, Encodable, Decodable)]
198 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
199 /// separating the two arguments with a comma for diagnostic suggestions.
200 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
201 // Used to suggest if a user writes `foo!(a b);`
202 let mut suggestion = None;
203 let mut iter = self.0.iter().enumerate().peekable();
204 while let Some((pos, ts)) = iter.next() {
205 if let Some((_, next)) = iter.peek() {
206 let sp = match (&ts, &next) {
207 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
209 (TokenTree::Token(token_left), Spacing::Alone),
210 (TokenTree::Token(token_right), _),
211 ) if ((token_left.is_ident() && !token_left.is_reserved_ident())
212 || token_left.is_lit())
213 && ((token_right.is_ident() && !token_right.is_reserved_ident())
214 || token_right.is_lit()) =>
218 ((TokenTree::Delimited(sp, ..), Spacing::Alone), _) => sp.entire(),
221 let sp = sp.shrink_to_hi();
222 let comma = (TokenTree::token(token::Comma, sp), Spacing::Alone);
223 suggestion = Some((pos, comma, sp));
226 if let Some((pos, comma, sp)) = suggestion {
227 let mut new_stream = Vec::with_capacity(self.0.len() + 1);
228 let parts = self.0.split_at(pos + 1);
229 new_stream.extend_from_slice(parts.0);
230 new_stream.push(comma);
231 new_stream.extend_from_slice(parts.1);
232 return Some((TokenStream::new(new_stream), sp));
238 impl From<TokenTree> for TokenStream {
239 fn from(tree: TokenTree) -> TokenStream {
240 TokenStream::new(vec![(tree, Spacing::Alone)])
244 impl From<TokenTree> for TreeAndSpacing {
245 fn from(tree: TokenTree) -> TreeAndSpacing {
246 (tree, Spacing::Alone)
250 impl iter::FromIterator<TokenTree> for TokenStream {
251 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
252 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndSpacing>>())
256 impl Eq for TokenStream {}
258 impl PartialEq<TokenStream> for TokenStream {
259 fn eq(&self, other: &TokenStream) -> bool {
260 self.trees().eq(other.trees())
265 pub fn new(streams: Vec<TreeAndSpacing>) -> TokenStream {
266 TokenStream(Lrc::new(streams))
269 pub fn is_empty(&self) -> bool {
273 pub fn len(&self) -> usize {
277 pub fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
278 match streams.len() {
279 0 => TokenStream::default(),
280 1 => streams.pop().unwrap(),
282 // We are going to extend the first stream in `streams` with
283 // the elements from the subsequent streams. This requires
284 // using `make_mut()` on the first stream, and in practice this
285 // doesn't cause cloning 99.9% of the time.
287 // One very common use case is when `streams` has two elements,
288 // where the first stream has any number of elements within
289 // (often 1, but sometimes many more) and the second stream has
290 // a single element within.
292 // Determine how much the first stream will be extended.
293 // Needed to avoid quadratic blow up from on-the-fly
294 // reallocations (#57735).
295 let num_appends = streams.iter().skip(1).map(|ts| ts.len()).sum();
297 // Get the first stream. If it's `None`, create an empty
299 let mut iter = streams.drain(..);
300 let mut first_stream_lrc = iter.next().unwrap().0;
302 // Append the elements to the first stream, after reserving
304 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
305 first_vec_mut.reserve(num_appends);
307 first_vec_mut.extend(stream.0.iter().cloned());
310 // Create the final `TokenStream`.
311 TokenStream(first_stream_lrc)
316 pub fn trees(&self) -> Cursor {
317 self.clone().into_trees()
320 pub fn into_trees(self) -> Cursor {
324 /// Compares two `TokenStream`s, checking equality without regarding span information.
325 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
326 let mut t1 = self.trees();
327 let mut t2 = other.trees();
328 for (t1, t2) in iter::zip(&mut t1, &mut t2) {
329 if !t1.eq_unspanned(&t2) {
333 t1.next().is_none() && t2.next().is_none()
336 pub fn map_enumerated<F: FnMut(usize, &TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
337 TokenStream(Lrc::new(
341 .map(|(i, (tree, is_joint))| (f(i, tree), *is_joint))
347 // 99.5%+ of the time we have 1 or 2 elements in this vector.
349 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
351 impl TokenStreamBuilder {
352 pub fn new() -> TokenStreamBuilder {
353 TokenStreamBuilder(SmallVec::new())
356 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
357 let mut stream = stream.into();
359 // If `self` is not empty and the last tree within the last stream is a
360 // token tree marked with `Joint`...
361 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
362 if let Some((TokenTree::Token(last_token), Spacing::Joint)) = last_stream_lrc.last() {
363 // ...and `stream` is not empty and the first tree within it is
365 let TokenStream(ref mut stream_lrc) = stream;
366 if let Some((TokenTree::Token(token), spacing)) = stream_lrc.first() {
367 // ...and the two tokens can be glued together...
368 if let Some(glued_tok) = last_token.glue(&token) {
369 // ...then do so, by overwriting the last token
370 // tree in `self` and removing the first token tree
371 // from `stream`. This requires using `make_mut()`
372 // on the last stream in `self` and on `stream`,
373 // and in practice this doesn't cause cloning 99.9%
376 // Overwrite the last token tree with the merged
378 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
379 *last_vec_mut.last_mut().unwrap() = (TokenTree::Token(glued_tok), *spacing);
381 // Remove the first token tree from `stream`. (This
382 // is almost always the only tree in `stream`.)
383 let stream_vec_mut = Lrc::make_mut(stream_lrc);
384 stream_vec_mut.remove(0);
386 // Don't push `stream` if it's empty -- that could
387 // block subsequent token gluing, by getting
388 // between two token trees that should be glued
390 if !stream.is_empty() {
401 pub fn build(self) -> TokenStream {
402 TokenStream::from_streams(self.0)
406 /// By-reference iterator over a [`TokenStream`].
408 pub struct CursorRef<'t> {
409 stream: &'t TokenStream,
413 impl<'t> CursorRef<'t> {
414 fn next_with_spacing(&mut self) -> Option<&'t TreeAndSpacing> {
415 self.stream.0.get(self.index).map(|tree| {
422 impl<'t> Iterator for CursorRef<'t> {
423 type Item = &'t TokenTree;
425 fn next(&mut self) -> Option<&'t TokenTree> {
426 self.next_with_spacing().map(|(tree, _)| tree)
430 /// Owning by-value iterator over a [`TokenStream`].
431 // FIXME: Many uses of this can be replaced with by-reference iterator to avoid clones.
434 pub stream: TokenStream,
438 impl Iterator for Cursor {
439 type Item = TokenTree;
441 fn next(&mut self) -> Option<TokenTree> {
442 self.next_with_spacing().map(|(tree, _)| tree)
447 fn new(stream: TokenStream) -> Self {
448 Cursor { stream, index: 0 }
451 pub fn next_with_spacing(&mut self) -> Option<TreeAndSpacing> {
452 if self.index < self.stream.len() {
454 Some(self.stream.0[self.index - 1].clone())
460 pub fn append(&mut self, new_stream: TokenStream) {
461 if new_stream.is_empty() {
464 let index = self.index;
465 let stream = mem::take(&mut self.stream);
466 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
470 pub fn look_ahead(&self, n: usize) -> Option<&TokenTree> {
471 self.stream.0[self.index..].get(n).map(|(tree, _)| tree)
475 #[derive(Debug, Copy, Clone, PartialEq, Encodable, Decodable, HashStable_Generic)]
476 pub struct DelimSpan {
482 pub fn from_single(sp: Span) -> Self {
483 DelimSpan { open: sp, close: sp }
486 pub fn from_pair(open: Span, close: Span) -> Self {
487 DelimSpan { open, close }
490 pub fn dummy() -> Self {
491 Self::from_single(DUMMY_SP)
494 pub fn entire(self) -> Span {
495 self.open.with_hi(self.close.hi())