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, RustcEncodable, RustcDecodable, 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 // See comments in `Nonterminal::to_tokenstream` for why we care about
70 // *probably* equal here rather than actual equality
72 // This is otherwise the same as `eq_unspanned`, only recursing with a
74 pub fn probably_equal_for_proc_macro(&self, other: &TokenTree) -> bool {
76 (TokenTree::Token(token), TokenTree::Token(token2)) => {
77 token.probably_equal_for_proc_macro(token2)
79 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
80 delim == delim2 && tts.probably_equal_for_proc_macro(&tts2)
86 /// Retrieves the TokenTree's span.
87 pub fn span(&self) -> Span {
89 TokenTree::Token(token) => token.span,
90 TokenTree::Delimited(sp, ..) => sp.entire(),
94 /// Modify the `TokenTree`'s span in-place.
95 pub fn set_span(&mut self, span: Span) {
97 TokenTree::Token(token) => token.span = span,
98 TokenTree::Delimited(dspan, ..) => *dspan = DelimSpan::from_single(span),
102 pub fn joint(self) -> TokenStream {
103 TokenStream::new(vec![(self, Joint)])
106 pub fn token(kind: TokenKind, span: Span) -> TokenTree {
107 TokenTree::Token(Token::new(kind, span))
110 /// Returns the opening delimiter as a token tree.
111 pub fn open_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
112 TokenTree::token(token::OpenDelim(delim), span.open)
115 /// Returns the closing delimiter as a token tree.
116 pub fn close_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
117 TokenTree::token(token::CloseDelim(delim), span.close)
120 pub fn uninterpolate(self) -> TokenTree {
122 TokenTree::Token(token) => TokenTree::Token(token.uninterpolate().into_owned()),
128 impl<CTX> HashStable<CTX> for TokenStream
130 CTX: crate::HashStableContext,
132 fn hash_stable(&self, hcx: &mut CTX, hasher: &mut StableHasher) {
133 for sub_tt in self.trees() {
134 sub_tt.hash_stable(hcx, hasher);
139 /// A `TokenStream` is an abstract sequence of tokens, organized into `TokenTree`s.
141 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
142 /// instead of a representation of the abstract syntax tree.
143 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for back-compat.
144 #[derive(Clone, Debug, Default, RustcEncodable, RustcDecodable)]
145 pub struct TokenStream(pub Lrc<Vec<TreeAndJoint>>);
147 pub type TreeAndJoint = (TokenTree, IsJoint);
149 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
150 #[cfg(target_arch = "x86_64")]
151 rustc_data_structures::static_assert_size!(TokenStream, 8);
153 #[derive(Clone, Copy, Debug, PartialEq, RustcEncodable, RustcDecodable)]
162 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
163 /// separating the two arguments with a comma for diagnostic suggestions.
164 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
165 // Used to suggest if a user writes `foo!(a b);`
166 let mut suggestion = None;
167 let mut iter = self.0.iter().enumerate().peekable();
168 while let Some((pos, ts)) = iter.next() {
169 if let Some((_, next)) = iter.peek() {
170 let sp = match (&ts, &next) {
171 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
173 (TokenTree::Token(token_left), NonJoint),
174 (TokenTree::Token(token_right), _),
175 ) if ((token_left.is_ident() && !token_left.is_reserved_ident())
176 || token_left.is_lit())
177 && ((token_right.is_ident() && !token_right.is_reserved_ident())
178 || token_right.is_lit()) =>
182 ((TokenTree::Delimited(sp, ..), NonJoint), _) => sp.entire(),
185 let sp = sp.shrink_to_hi();
186 let comma = (TokenTree::token(token::Comma, sp), NonJoint);
187 suggestion = Some((pos, comma, sp));
190 if let Some((pos, comma, sp)) = suggestion {
191 let mut new_stream = vec![];
192 let parts = self.0.split_at(pos + 1);
193 new_stream.extend_from_slice(parts.0);
194 new_stream.push(comma);
195 new_stream.extend_from_slice(parts.1);
196 return Some((TokenStream::new(new_stream), sp));
202 impl From<TokenTree> for TokenStream {
203 fn from(tree: TokenTree) -> TokenStream {
204 TokenStream::new(vec![(tree, NonJoint)])
208 impl From<TokenTree> for TreeAndJoint {
209 fn from(tree: TokenTree) -> TreeAndJoint {
214 impl iter::FromIterator<TokenTree> for TokenStream {
215 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
216 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndJoint>>())
220 impl Eq for TokenStream {}
222 impl PartialEq<TokenStream> for TokenStream {
223 fn eq(&self, other: &TokenStream) -> bool {
224 self.trees().eq(other.trees())
229 pub fn new(streams: Vec<TreeAndJoint>) -> TokenStream {
230 TokenStream(Lrc::new(streams))
233 pub fn is_empty(&self) -> bool {
237 pub fn len(&self) -> usize {
241 pub fn span(&self) -> Option<Span> {
244 [(tt, _)] => Some(tt.span()),
245 [(tt_start, _), .., (tt_end, _)] => Some(tt_start.span().to(tt_end.span())),
249 pub fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
250 match streams.len() {
251 0 => TokenStream::default(),
252 1 => streams.pop().unwrap(),
254 // We are going to extend the first stream in `streams` with
255 // the elements from the subsequent streams. This requires
256 // using `make_mut()` on the first stream, and in practice this
257 // doesn't cause cloning 99.9% of the time.
259 // One very common use case is when `streams` has two elements,
260 // where the first stream has any number of elements within
261 // (often 1, but sometimes many more) and the second stream has
262 // a single element within.
264 // Determine how much the first stream will be extended.
265 // Needed to avoid quadratic blow up from on-the-fly
266 // reallocations (#57735).
267 let num_appends = streams.iter().skip(1).map(|ts| ts.len()).sum();
269 // Get the first stream. If it's `None`, create an empty
271 let mut iter = streams.drain(..);
272 let mut first_stream_lrc = iter.next().unwrap().0;
274 // Append the elements to the first stream, after reserving
276 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
277 first_vec_mut.reserve(num_appends);
279 first_vec_mut.extend(stream.0.iter().cloned());
282 // Create the final `TokenStream`.
283 TokenStream(first_stream_lrc)
288 pub fn trees(&self) -> Cursor {
289 self.clone().into_trees()
292 pub fn into_trees(self) -> Cursor {
296 /// Compares two `TokenStream`s, checking equality without regarding span information.
297 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
298 let mut t1 = self.trees();
299 let mut t2 = other.trees();
300 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
301 if !t1.eq_unspanned(&t2) {
305 t1.next().is_none() && t2.next().is_none()
308 // See comments in `Nonterminal::to_tokenstream` for why we care about
309 // *probably* equal here rather than actual equality
311 // This is otherwise the same as `eq_unspanned`, only recursing with a
313 pub fn probably_equal_for_proc_macro(&self, other: &TokenStream) -> bool {
314 // When checking for `probably_eq`, we ignore certain tokens that aren't
315 // preserved in the AST. Because they are not preserved, the pretty
316 // printer arbitrarily adds or removes them when printing as token
317 // streams, making a comparison between a token stream generated from an
318 // AST and a token stream which was parsed into an AST more reliable.
319 fn semantic_tree(tree: &TokenTree) -> bool {
320 if let TokenTree::Token(token) = tree {
322 // The pretty printer tends to add trailing commas to
323 // everything, and in particular, after struct fields.
325 // The pretty printer emits `NoDelim` as whitespace.
326 | token::OpenDelim(DelimToken::NoDelim)
327 | token::CloseDelim(DelimToken::NoDelim)
328 // The pretty printer collapses many semicolons into one.
330 // The pretty printer collapses whitespace arbitrarily and can
331 // introduce whitespace from `NoDelim`.
333 // The pretty printer can turn `$crate` into `::crate_name`
334 | token::ModSep = token.kind {
341 let mut t1 = self.trees().filter(semantic_tree);
342 let mut t2 = other.trees().filter(semantic_tree);
343 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
344 if !t1.probably_equal_for_proc_macro(&t2) {
348 t1.next().is_none() && t2.next().is_none()
351 pub fn map_enumerated<F: FnMut(usize, TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
352 TokenStream(Lrc::new(
356 .map(|(i, (tree, is_joint))| (f(i, tree.clone()), *is_joint))
361 pub fn map<F: FnMut(TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
362 TokenStream(Lrc::new(
363 self.0.iter().map(|(tree, is_joint)| (f(tree.clone()), *is_joint)).collect(),
368 // 99.5%+ of the time we have 1 or 2 elements in this vector.
370 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
372 impl TokenStreamBuilder {
373 pub fn new() -> TokenStreamBuilder {
374 TokenStreamBuilder(SmallVec::new())
377 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
378 let mut stream = stream.into();
380 // If `self` is not empty and the last tree within the last stream is a
381 // token tree marked with `Joint`...
382 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
383 if let Some((TokenTree::Token(last_token), Joint)) = last_stream_lrc.last() {
384 // ...and `stream` is not empty and the first tree within it is
386 let TokenStream(ref mut stream_lrc) = stream;
387 if let Some((TokenTree::Token(token), is_joint)) = stream_lrc.first() {
388 // ...and the two tokens can be glued together...
389 if let Some(glued_tok) = last_token.glue(&token) {
390 // ...then do so, by overwriting the last token
391 // tree in `self` and removing the first token tree
392 // from `stream`. This requires using `make_mut()`
393 // on the last stream in `self` and on `stream`,
394 // and in practice this doesn't cause cloning 99.9%
397 // Overwrite the last token tree with the merged
399 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
400 *last_vec_mut.last_mut().unwrap() =
401 (TokenTree::Token(glued_tok), *is_joint);
403 // Remove the first token tree from `stream`. (This
404 // is almost always the only tree in `stream`.)
405 let stream_vec_mut = Lrc::make_mut(stream_lrc);
406 stream_vec_mut.remove(0);
408 // Don't push `stream` if it's empty -- that could
409 // block subsequent token gluing, by getting
410 // between two token trees that should be glued
412 if !stream.is_empty() {
423 pub fn build(self) -> TokenStream {
424 TokenStream::from_streams(self.0)
430 pub stream: TokenStream,
434 impl Iterator for Cursor {
435 type Item = TokenTree;
437 fn next(&mut self) -> Option<TokenTree> {
438 self.next_with_joint().map(|(tree, _)| tree)
443 fn new(stream: TokenStream) -> Self {
444 Cursor { stream, index: 0 }
447 pub fn next_with_joint(&mut self) -> Option<TreeAndJoint> {
448 if self.index < self.stream.len() {
450 Some(self.stream.0[self.index - 1].clone())
456 pub fn append(&mut self, new_stream: TokenStream) {
457 if new_stream.is_empty() {
460 let index = self.index;
461 let stream = mem::take(&mut self.stream);
462 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
466 pub fn look_ahead(&self, n: usize) -> Option<TokenTree> {
467 self.stream.0[self.index..].get(n).map(|(tree, _)| tree.clone())
471 #[derive(Debug, Copy, Clone, PartialEq, RustcEncodable, RustcDecodable, HashStable_Generic)]
472 pub struct DelimSpan {
478 pub fn from_single(sp: Span) -> Self {
479 DelimSpan { open: sp, close: sp }
482 pub fn from_pair(open: Span, close: Span) -> Self {
483 DelimSpan { open, close }
486 pub fn dummy() -> Self {
487 Self::from_single(DUMMY_SP)
490 pub fn entire(self) -> Span {
491 self.open.with_hi(self.close.hi())