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::parse::token::{self, DelimToken, Token, TokenKind};
18 use syntax_pos::{BytePos, Span, DUMMY_SP};
19 #[cfg(target_arch = "x86_64")]
20 use rustc_data_structures::static_assert_size;
21 use rustc_data_structures::sync::Lrc;
22 use smallvec::{SmallVec, smallvec};
29 /// When the main rust parser encounters a syntax-extension invocation, it
30 /// parses the arguments to the invocation as a token-tree. This is a very
31 /// loose structure, such that all sorts of different AST-fragments can
32 /// be passed to syntax extensions using a uniform type.
34 /// If the syntax extension is an MBE macro, it will attempt to match its
35 /// LHS token tree against the provided token tree, and if it finds a
36 /// match, will transcribe the RHS token tree, splicing in any captured
37 /// `macro_parser::matched_nonterminals` into the `SubstNt`s it finds.
39 /// The RHS of an MBE macro is the only place `SubstNt`s are substituted.
40 /// Nothing special happens to misnamed or misplaced `SubstNt`s.
41 #[derive(Debug, Clone, PartialEq, RustcEncodable, RustcDecodable)]
45 /// A delimited sequence of token trees
46 Delimited(DelimSpan, DelimToken, TokenStream),
49 // Ensure all fields of `TokenTree` is `Send` and `Sync`.
50 #[cfg(parallel_compiler)]
54 DelimSpan: Send + Sync,
55 DelimToken: Send + Sync,
56 TokenStream: Send + Sync,
60 /// Checks if this TokenTree is equal to the other, regardless of span information.
61 pub fn eq_unspanned(&self, other: &TokenTree) -> bool {
63 (TokenTree::Token(token), TokenTree::Token(token2)) => token.kind == token2.kind,
64 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
65 delim == delim2 && tts.eq_unspanned(&tts2)
71 // See comments in `Nonterminal::to_tokenstream` for why we care about
72 // *probably* equal here rather than actual equality
74 // This is otherwise the same as `eq_unspanned`, only recursing with a
76 pub fn probably_equal_for_proc_macro(&self, other: &TokenTree) -> bool {
78 (TokenTree::Token(token), TokenTree::Token(token2)) => {
79 token.probably_equal_for_proc_macro(token2)
81 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
82 delim == delim2 && tts.probably_equal_for_proc_macro(&tts2)
88 /// Retrieves the TokenTree's span.
89 pub fn span(&self) -> Span {
91 TokenTree::Token(token) => token.span,
92 TokenTree::Delimited(sp, ..) => sp.entire(),
96 /// Modify the `TokenTree`'s span in-place.
97 pub fn set_span(&mut self, span: Span) {
99 TokenTree::Token(token) => token.span = span,
100 TokenTree::Delimited(dspan, ..) => *dspan = DelimSpan::from_single(span),
104 pub fn joint(self) -> TokenStream {
105 TokenStream::new(vec![(self, Joint)])
108 pub fn token(kind: TokenKind, span: Span) -> TokenTree {
109 TokenTree::Token(Token::new(kind, span))
112 /// Returns the opening delimiter as a token tree.
113 pub fn open_tt(span: Span, delim: DelimToken) -> TokenTree {
114 let open_span = if span.is_dummy() {
117 span.with_hi(span.lo() + BytePos(delim.len() as u32))
119 TokenTree::token(token::OpenDelim(delim), open_span)
122 /// Returns the closing delimiter as a token tree.
123 pub fn close_tt(span: Span, delim: DelimToken) -> TokenTree {
124 let close_span = if span.is_dummy() {
127 span.with_lo(span.hi() - BytePos(delim.len() as u32))
129 TokenTree::token(token::CloseDelim(delim), close_span)
133 /// A `TokenStream` is an abstract sequence of tokens, organized into `TokenTree`s.
135 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
136 /// instead of a representation of the abstract syntax tree.
137 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for back-compat.
138 #[derive(Clone, Debug, Default, RustcEncodable, RustcDecodable)]
139 pub struct TokenStream(pub Lrc<Vec<TreeAndJoint>>);
141 pub type TreeAndJoint = (TokenTree, IsJoint);
143 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
144 #[cfg(target_arch = "x86_64")]
145 static_assert_size!(TokenStream, 8);
147 #[derive(Clone, Copy, Debug, PartialEq, RustcEncodable, RustcDecodable)]
156 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
157 /// separating the two arguments with a comma for diagnostic suggestions.
158 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
159 // Used to suggest if a user writes `foo!(a b);`
160 let mut suggestion = None;
161 let mut iter = self.0.iter().enumerate().peekable();
162 while let Some((pos, ts)) = iter.next() {
163 if let Some((_, next)) = iter.peek() {
164 let sp = match (&ts, &next) {
165 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
166 ((TokenTree::Token(token_left), NonJoint),
167 (TokenTree::Token(token_right), _))
168 if ((token_left.is_ident() && !token_left.is_reserved_ident())
169 || token_left.is_lit()) &&
170 ((token_right.is_ident() && !token_right.is_reserved_ident())
171 || token_right.is_lit()) => token_left.span,
172 ((TokenTree::Delimited(sp, ..), NonJoint), _) => sp.entire(),
175 let sp = sp.shrink_to_hi();
176 let comma = (TokenTree::token(token::Comma, sp), NonJoint);
177 suggestion = Some((pos, comma, sp));
180 if let Some((pos, comma, sp)) = suggestion {
181 let mut new_stream = vec![];
182 let parts = self.0.split_at(pos + 1);
183 new_stream.extend_from_slice(parts.0);
184 new_stream.push(comma);
185 new_stream.extend_from_slice(parts.1);
186 return Some((TokenStream::new(new_stream), sp));
192 impl From<TokenTree> for TokenStream {
193 fn from(tree: TokenTree) -> TokenStream {
194 TokenStream::new(vec![(tree, NonJoint)])
198 impl From<TokenTree> for TreeAndJoint {
199 fn from(tree: TokenTree) -> TreeAndJoint {
204 impl iter::FromIterator<TokenTree> for TokenStream {
205 fn from_iter<I: IntoIterator<Item = TokenTree>>(iter: I) -> Self {
206 TokenStream::new(iter.into_iter().map(Into::into).collect::<Vec<TreeAndJoint>>())
210 impl Eq for TokenStream {}
212 impl PartialEq<TokenStream> for TokenStream {
213 fn eq(&self, other: &TokenStream) -> bool {
214 self.trees().eq(other.trees())
219 pub fn new(streams: Vec<TreeAndJoint>) -> TokenStream {
220 TokenStream(Lrc::new(streams))
223 pub fn is_empty(&self) -> bool {
227 pub fn len(&self) -> usize {
231 pub(crate) fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
232 match streams.len() {
233 0 => TokenStream::default(),
234 1 => streams.pop().unwrap(),
236 // We are going to extend the first stream in `streams` with
237 // the elements from the subsequent streams. This requires
238 // using `make_mut()` on the first stream, and in practice this
239 // doesn't cause cloning 99.9% of the time.
241 // One very common use case is when `streams` has two elements,
242 // where the first stream has any number of elements within
243 // (often 1, but sometimes many more) and the second stream has
244 // a single element within.
246 // Determine how much the first stream will be extended.
247 // Needed to avoid quadratic blow up from on-the-fly
248 // reallocations (#57735).
249 let num_appends = streams.iter()
254 // Get the first stream. If it's `None`, create an empty
256 let mut iter = streams.drain();
257 let mut first_stream_lrc = iter.next().unwrap().0;
259 // Append the elements to the first stream, after reserving
261 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
262 first_vec_mut.reserve(num_appends);
264 first_vec_mut.extend(stream.0.iter().cloned());
267 // Create the final `TokenStream`.
268 TokenStream(first_stream_lrc)
273 pub fn trees(&self) -> Cursor {
274 self.clone().into_trees()
277 pub fn into_trees(self) -> Cursor {
281 /// Compares two `TokenStream`s, checking equality without regarding span information.
282 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
283 let mut t1 = self.trees();
284 let mut t2 = other.trees();
285 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
286 if !t1.eq_unspanned(&t2) {
290 t1.next().is_none() && t2.next().is_none()
293 // See comments in `Nonterminal::to_tokenstream` for why we care about
294 // *probably* equal here rather than actual equality
296 // This is otherwise the same as `eq_unspanned`, only recursing with a
298 pub fn probably_equal_for_proc_macro(&self, other: &TokenStream) -> bool {
299 // When checking for `probably_eq`, we ignore certain tokens that aren't
300 // preserved in the AST. Because they are not preserved, the pretty
301 // printer arbitrarily adds or removes them when printing as token
302 // streams, making a comparison between a token stream generated from an
303 // AST and a token stream which was parsed into an AST more reliable.
304 fn semantic_tree(tree: &TokenTree) -> bool {
305 if let TokenTree::Token(token) = tree {
307 // The pretty printer tends to add trailing commas to
308 // everything, and in particular, after struct fields.
310 // The pretty printer emits `NoDelim` as whitespace.
311 | token::OpenDelim(DelimToken::NoDelim)
312 | token::CloseDelim(DelimToken::NoDelim)
313 // The pretty printer collapses many semicolons into one.
315 // The pretty printer collapses whitespace arbitrarily and can
316 // introduce whitespace from `NoDelim`.
318 // The pretty printer can turn `$crate` into `::crate_name`
319 | token::ModSep = token.kind {
326 let mut t1 = self.trees().filter(semantic_tree);
327 let mut t2 = other.trees().filter(semantic_tree);
328 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
329 if !t1.probably_equal_for_proc_macro(&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.clone()), *is_joint))
346 pub fn map<F: FnMut(TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
347 TokenStream(Lrc::new(
350 .map(|(tree, is_joint)| (f(tree.clone()), *is_joint))
356 // 99.5%+ of the time we have 1 or 2 elements in this vector.
358 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
360 impl TokenStreamBuilder {
361 pub fn new() -> TokenStreamBuilder {
362 TokenStreamBuilder(SmallVec::new())
365 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
366 let mut stream = stream.into();
368 // If `self` is not empty and the last tree within the last stream is a
369 // token tree marked with `Joint`...
370 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
371 if let Some((TokenTree::Token(last_token), Joint)) = last_stream_lrc.last() {
373 // ...and `stream` is not empty and the first tree within it is
375 let TokenStream(ref mut stream_lrc) = stream;
376 if let Some((TokenTree::Token(token), is_joint)) = stream_lrc.first() {
378 // ...and the two tokens can be glued together...
379 if let Some(glued_tok) = last_token.glue(&token) {
381 // ...then do so, by overwriting the last token
382 // tree in `self` and removing the first token tree
383 // from `stream`. This requires using `make_mut()`
384 // on the last stream in `self` and on `stream`,
385 // and in practice this doesn't cause cloning 99.9%
388 // Overwrite the last token tree with the merged
390 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
391 *last_vec_mut.last_mut().unwrap() =
392 (TokenTree::Token(glued_tok), *is_joint);
394 // Remove the first token tree from `stream`. (This
395 // is almost always the only tree in `stream`.)
396 let stream_vec_mut = Lrc::make_mut(stream_lrc);
397 stream_vec_mut.remove(0);
399 // Don't push `stream` if it's empty -- that could
400 // block subsequent token gluing, by getting
401 // between two token trees that should be glued
403 if !stream.is_empty() {
414 pub fn build(self) -> TokenStream {
415 TokenStream::from_streams(self.0)
421 pub stream: TokenStream,
425 impl Iterator for Cursor {
426 type Item = TokenTree;
428 fn next(&mut self) -> Option<TokenTree> {
429 self.next_with_joint().map(|(tree, _)| tree)
434 fn new(stream: TokenStream) -> Self {
435 Cursor { stream, index: 0 }
438 pub fn next_with_joint(&mut self) -> Option<TreeAndJoint> {
439 if self.index < self.stream.len() {
441 Some(self.stream.0[self.index - 1].clone())
447 pub fn append(&mut self, new_stream: TokenStream) {
448 if new_stream.is_empty() {
451 let index = self.index;
452 let stream = mem::take(&mut self.stream);
453 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
457 pub fn look_ahead(&self, n: usize) -> Option<TokenTree> {
458 self.stream.0[self.index ..].get(n).map(|(tree, _)| tree.clone())
462 #[derive(Debug, Copy, Clone, PartialEq, RustcEncodable, RustcDecodable)]
463 pub struct DelimSpan {
469 pub fn from_single(sp: Span) -> Self {
476 pub fn from_pair(open: Span, close: Span) -> Self {
477 DelimSpan { open, close }
480 pub fn dummy() -> Self {
481 Self::from_single(DUMMY_SP)
484 pub fn entire(self) -> Span {
485 self.open.with_hi(self.close.hi())