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 syntax_pos::{Span, DUMMY_SP};
19 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
20 use rustc_macros::HashStable_Generic;
21 use rustc_data_structures::sync::Lrc;
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,
57 /// Checks if this TokenTree is equal to the other, regardless of span information.
58 pub fn eq_unspanned(&self, other: &TokenTree) -> bool {
60 (TokenTree::Token(token), TokenTree::Token(token2)) => token.kind == token2.kind,
61 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
62 delim == delim2 && tts.eq_unspanned(&tts2)
68 // See comments in `Nonterminal::to_tokenstream` for why we care about
69 // *probably* equal here rather than actual equality
71 // This is otherwise the same as `eq_unspanned`, only recursing with a
73 pub fn probably_equal_for_proc_macro(&self, other: &TokenTree) -> bool {
75 (TokenTree::Token(token), TokenTree::Token(token2)) => {
76 token.probably_equal_for_proc_macro(token2)
78 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
79 delim == delim2 && tts.probably_equal_for_proc_macro(&tts2)
85 /// Retrieves the TokenTree's span.
86 pub fn span(&self) -> Span {
88 TokenTree::Token(token) => token.span,
89 TokenTree::Delimited(sp, ..) => sp.entire(),
93 /// Modify the `TokenTree`'s span in-place.
94 pub fn set_span(&mut self, span: Span) {
96 TokenTree::Token(token) => token.span = span,
97 TokenTree::Delimited(dspan, ..) => *dspan = DelimSpan::from_single(span),
101 pub fn joint(self) -> TokenStream {
102 TokenStream::new(vec![(self, Joint)])
105 pub fn token(kind: TokenKind, span: Span) -> TokenTree {
106 TokenTree::Token(Token::new(kind, span))
109 /// Returns the opening delimiter as a token tree.
110 pub fn open_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
111 TokenTree::token(token::OpenDelim(delim), span.open)
114 /// Returns the closing delimiter as a token tree.
115 pub fn close_tt(span: DelimSpan, delim: DelimToken) -> TokenTree {
116 TokenTree::token(token::CloseDelim(delim), span.close)
120 impl<CTX> HashStable<CTX> for TokenStream
121 where CTX: crate::HashStableContext
123 fn hash_stable(&self, hcx: &mut CTX, hasher: &mut StableHasher) {
124 for sub_tt in self.trees() {
125 sub_tt.hash_stable(hcx, hasher);
130 /// A `TokenStream` is an abstract sequence of tokens, organized into `TokenTree`s.
132 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
133 /// instead of a representation of the abstract syntax tree.
134 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for back-compat.
135 #[derive(Clone, Debug, Default, RustcEncodable, RustcDecodable)]
136 pub struct TokenStream(pub Lrc<Vec<TreeAndJoint>>);
138 pub type TreeAndJoint = (TokenTree, IsJoint);
140 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
141 #[cfg(target_arch = "x86_64")]
142 rustc_data_structures::static_assert_size!(TokenStream, 8);
144 #[derive(Clone, Copy, Debug, PartialEq, RustcEncodable, RustcDecodable)]
153 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
154 /// separating the two arguments with a comma for diagnostic suggestions.
155 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
156 // Used to suggest if a user writes `foo!(a b);`
157 let mut suggestion = None;
158 let mut iter = self.0.iter().enumerate().peekable();
159 while let Some((pos, ts)) = iter.next() {
160 if let Some((_, next)) = iter.peek() {
161 let sp = match (&ts, &next) {
162 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
163 ((TokenTree::Token(token_left), NonJoint),
164 (TokenTree::Token(token_right), _))
165 if ((token_left.is_ident() && !token_left.is_reserved_ident())
166 || token_left.is_lit()) &&
167 ((token_right.is_ident() && !token_right.is_reserved_ident())
168 || token_right.is_lit()) => token_left.span,
169 ((TokenTree::Delimited(sp, ..), NonJoint), _) => sp.entire(),
172 let sp = sp.shrink_to_hi();
173 let comma = (TokenTree::token(token::Comma, sp), NonJoint);
174 suggestion = Some((pos, comma, sp));
177 if let Some((pos, comma, sp)) = suggestion {
178 let mut new_stream = vec![];
179 let parts = self.0.split_at(pos + 1);
180 new_stream.extend_from_slice(parts.0);
181 new_stream.push(comma);
182 new_stream.extend_from_slice(parts.1);
183 return Some((TokenStream::new(new_stream), sp));
189 impl From<TokenTree> for TokenStream {
190 fn from(tree: TokenTree) -> TokenStream {
191 TokenStream::new(vec![(tree, NonJoint)])
195 impl From<TokenTree> for TreeAndJoint {
196 fn from(tree: TokenTree) -> TreeAndJoint {
201 impl iter::FromIterator<TokenTree> for TokenStream {
202 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
203 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndJoint>>())
207 impl Eq for TokenStream {}
209 impl PartialEq<TokenStream> for TokenStream {
210 fn eq(&self, other: &TokenStream) -> bool {
211 self.trees().eq(other.trees())
216 pub fn new(streams: Vec<TreeAndJoint>) -> TokenStream {
217 TokenStream(Lrc::new(streams))
220 pub fn is_empty(&self) -> bool {
224 pub fn len(&self) -> usize {
228 pub fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
229 match streams.len() {
230 0 => TokenStream::default(),
231 1 => streams.pop().unwrap(),
233 // We are going to extend the first stream in `streams` with
234 // the elements from the subsequent streams. This requires
235 // using `make_mut()` on the first stream, and in practice this
236 // doesn't cause cloning 99.9% of the time.
238 // One very common use case is when `streams` has two elements,
239 // where the first stream has any number of elements within
240 // (often 1, but sometimes many more) and the second stream has
241 // a single element within.
243 // Determine how much the first stream will be extended.
244 // Needed to avoid quadratic blow up from on-the-fly
245 // reallocations (#57735).
246 let num_appends = streams.iter()
251 // Get the first stream. If it's `None`, create an empty
253 let mut iter = streams.drain(..);
254 let mut first_stream_lrc = iter.next().unwrap().0;
256 // Append the elements to the first stream, after reserving
258 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
259 first_vec_mut.reserve(num_appends);
261 first_vec_mut.extend(stream.0.iter().cloned());
264 // Create the final `TokenStream`.
265 TokenStream(first_stream_lrc)
270 pub fn trees(&self) -> Cursor {
271 self.clone().into_trees()
274 pub fn into_trees(self) -> Cursor {
278 /// Compares two `TokenStream`s, checking equality without regarding span information.
279 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
280 let mut t1 = self.trees();
281 let mut t2 = other.trees();
282 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
283 if !t1.eq_unspanned(&t2) {
287 t1.next().is_none() && t2.next().is_none()
290 // See comments in `Nonterminal::to_tokenstream` for why we care about
291 // *probably* equal here rather than actual equality
293 // This is otherwise the same as `eq_unspanned`, only recursing with a
295 pub fn probably_equal_for_proc_macro(&self, other: &TokenStream) -> bool {
296 // When checking for `probably_eq`, we ignore certain tokens that aren't
297 // preserved in the AST. Because they are not preserved, the pretty
298 // printer arbitrarily adds or removes them when printing as token
299 // streams, making a comparison between a token stream generated from an
300 // AST and a token stream which was parsed into an AST more reliable.
301 fn semantic_tree(tree: &TokenTree) -> bool {
302 if let TokenTree::Token(token) = tree {
304 // The pretty printer tends to add trailing commas to
305 // everything, and in particular, after struct fields.
307 // The pretty printer emits `NoDelim` as whitespace.
308 | token::OpenDelim(DelimToken::NoDelim)
309 | token::CloseDelim(DelimToken::NoDelim)
310 // The pretty printer collapses many semicolons into one.
312 // The pretty printer collapses whitespace arbitrarily and can
313 // introduce whitespace from `NoDelim`.
315 // The pretty printer can turn `$crate` into `::crate_name`
316 | token::ModSep = token.kind {
323 let mut t1 = self.trees().filter(semantic_tree);
324 let mut t2 = other.trees().filter(semantic_tree);
325 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
326 if !t1.probably_equal_for_proc_macro(&t2) {
330 t1.next().is_none() && t2.next().is_none()
333 pub fn map_enumerated<F: FnMut(usize, TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
334 TokenStream(Lrc::new(
338 .map(|(i, (tree, is_joint))| (f(i, tree.clone()), *is_joint))
343 pub fn map<F: FnMut(TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
344 TokenStream(Lrc::new(
347 .map(|(tree, is_joint)| (f(tree.clone()), *is_joint))
353 // 99.5%+ of the time we have 1 or 2 elements in this vector.
355 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
357 impl TokenStreamBuilder {
358 pub fn new() -> TokenStreamBuilder {
359 TokenStreamBuilder(SmallVec::new())
362 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
363 let mut stream = stream.into();
365 // If `self` is not empty and the last tree within the last stream is a
366 // token tree marked with `Joint`...
367 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
368 if let Some((TokenTree::Token(last_token), Joint)) = last_stream_lrc.last() {
370 // ...and `stream` is not empty and the first tree within it is
372 let TokenStream(ref mut stream_lrc) = stream;
373 if let Some((TokenTree::Token(token), is_joint)) = stream_lrc.first() {
375 // ...and the two tokens can be glued together...
376 if let Some(glued_tok) = last_token.glue(&token) {
378 // ...then do so, by overwriting the last token
379 // tree in `self` and removing the first token tree
380 // from `stream`. This requires using `make_mut()`
381 // on the last stream in `self` and on `stream`,
382 // and in practice this doesn't cause cloning 99.9%
385 // Overwrite the last token tree with the merged
387 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
388 *last_vec_mut.last_mut().unwrap() =
389 (TokenTree::Token(glued_tok), *is_joint);
391 // Remove the first token tree from `stream`. (This
392 // is almost always the only tree in `stream`.)
393 let stream_vec_mut = Lrc::make_mut(stream_lrc);
394 stream_vec_mut.remove(0);
396 // Don't push `stream` if it's empty -- that could
397 // block subsequent token gluing, by getting
398 // between two token trees that should be glued
400 if !stream.is_empty() {
411 pub fn build(self) -> TokenStream {
412 TokenStream::from_streams(self.0)
418 pub stream: TokenStream,
422 impl Iterator for Cursor {
423 type Item = TokenTree;
425 fn next(&mut self) -> Option<TokenTree> {
426 self.next_with_joint().map(|(tree, _)| tree)
431 fn new(stream: TokenStream) -> Self {
432 Cursor { stream, index: 0 }
435 pub fn next_with_joint(&mut self) -> Option<TreeAndJoint> {
436 if self.index < self.stream.len() {
438 Some(self.stream.0[self.index - 1].clone())
444 pub fn append(&mut self, new_stream: TokenStream) {
445 if new_stream.is_empty() {
448 let index = self.index;
449 let stream = mem::take(&mut self.stream);
450 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
454 pub fn look_ahead(&self, n: usize) -> Option<TokenTree> {
455 self.stream.0[self.index ..].get(n).map(|(tree, _)| tree.clone())
459 #[derive(Debug, Copy, Clone, PartialEq, RustcEncodable, RustcDecodable, HashStable_Generic)]
460 pub struct DelimSpan {
466 pub fn from_single(sp: Span) -> Self {
473 pub fn from_pair(open: Span, close: Span) -> Self {
474 DelimSpan { open, close }
477 pub fn dummy() -> Self {
478 Self::from_single(DUMMY_SP)
481 pub fn entire(self) -> Span {
482 self.open.with_hi(self.close.hi())