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 rustc_serialize::{Decoder, Decodable, Encoder, Encodable};
23 use smallvec::{SmallVec, smallvec};
30 /// When the main rust parser encounters a syntax-extension invocation, it
31 /// parses the arguments to the invocation as a token-tree. This is a very
32 /// loose structure, such that all sorts of different AST-fragments can
33 /// be passed to syntax extensions using a uniform type.
35 /// If the syntax extension is an MBE macro, it will attempt to match its
36 /// LHS token tree against the provided token tree, and if it finds a
37 /// match, will transcribe the RHS token tree, splicing in any captured
38 /// `macro_parser::matched_nonterminals` into the `SubstNt`s it finds.
40 /// The RHS of an MBE macro is the only place `SubstNt`s are substituted.
41 /// Nothing special happens to misnamed or misplaced `SubstNt`s.
42 #[derive(Debug, Clone, PartialEq, RustcEncodable, RustcDecodable)]
46 /// A delimited sequence of token trees
47 Delimited(DelimSpan, DelimToken, TokenStream),
50 // Ensure all fields of `TokenTree` is `Send` and `Sync`.
51 #[cfg(parallel_compiler)]
55 DelimSpan: Send + Sync,
56 DelimToken: Send + Sync,
57 TokenStream: Send + Sync,
61 /// Checks if this TokenTree is equal to the other, regardless of span information.
62 pub fn eq_unspanned(&self, other: &TokenTree) -> bool {
64 (TokenTree::Token(token), TokenTree::Token(token2)) => token.kind == token2.kind,
65 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
66 delim == delim2 && tts.eq_unspanned(&tts2)
72 // See comments in `Nonterminal::to_tokenstream` for why we care about
73 // *probably* equal here rather than actual equality
75 // This is otherwise the same as `eq_unspanned`, only recursing with a
77 pub fn probably_equal_for_proc_macro(&self, other: &TokenTree) -> bool {
79 (TokenTree::Token(token), TokenTree::Token(token2)) => {
80 token.probably_equal_for_proc_macro(token2)
82 (TokenTree::Delimited(_, delim, tts), TokenTree::Delimited(_, delim2, tts2)) => {
83 delim == delim2 && tts.probably_equal_for_proc_macro(&tts2)
89 /// Retrieves the TokenTree's span.
90 pub fn span(&self) -> Span {
92 TokenTree::Token(token) => token.span,
93 TokenTree::Delimited(sp, ..) => sp.entire(),
97 /// Modify the `TokenTree`'s span in-place.
98 pub fn set_span(&mut self, span: Span) {
100 TokenTree::Token(token) => token.span = span,
101 TokenTree::Delimited(dspan, ..) => *dspan = DelimSpan::from_single(span),
105 pub fn joint(self) -> TokenStream {
106 TokenStream::new(vec![(self, Joint)])
109 pub fn token(kind: TokenKind, span: Span) -> TokenTree {
110 TokenTree::Token(Token::new(kind, span))
113 /// Returns the opening delimiter as a token tree.
114 pub fn open_tt(span: Span, delim: DelimToken) -> TokenTree {
115 let open_span = if span.is_dummy() {
118 span.with_hi(span.lo() + BytePos(delim.len() as u32))
120 TokenTree::token(token::OpenDelim(delim), open_span)
123 /// Returns the closing delimiter as a token tree.
124 pub fn close_tt(span: Span, delim: DelimToken) -> TokenTree {
125 let close_span = if span.is_dummy() {
128 span.with_lo(span.hi() - BytePos(delim.len() as u32))
130 TokenTree::token(token::CloseDelim(delim), close_span)
134 /// A `TokenStream` is an abstract sequence of tokens, organized into `TokenTree`s.
136 /// The goal is for procedural macros to work with `TokenStream`s and `TokenTree`s
137 /// instead of a representation of the abstract syntax tree.
138 /// Today's `TokenTree`s can still contain AST via `token::Interpolated` for back-compat.
139 #[derive(Clone, Debug, Default)]
140 pub struct TokenStream(pub Lrc<Vec<TreeAndJoint>>);
142 pub type TreeAndJoint = (TokenTree, IsJoint);
144 // `TokenStream` is used a lot. Make sure it doesn't unintentionally get bigger.
145 #[cfg(target_arch = "x86_64")]
146 static_assert_size!(TokenStream, 8);
148 #[derive(Clone, Copy, Debug, PartialEq)]
157 /// Given a `TokenStream` with a `Stream` of only two arguments, return a new `TokenStream`
158 /// separating the two arguments with a comma for diagnostic suggestions.
159 pub fn add_comma(&self) -> Option<(TokenStream, Span)> {
160 // Used to suggest if a user writes `foo!(a b);`
161 let mut suggestion = None;
162 let mut iter = self.0.iter().enumerate().peekable();
163 while let Some((pos, ts)) = iter.next() {
164 if let Some((_, next)) = iter.peek() {
165 let sp = match (&ts, &next) {
166 (_, (TokenTree::Token(Token { kind: token::Comma, .. }), _)) => continue,
167 ((TokenTree::Token(token_left), NonJoint),
168 (TokenTree::Token(token_right), _))
169 if ((token_left.is_ident() && !token_left.is_reserved_ident())
170 || token_left.is_lit()) &&
171 ((token_right.is_ident() && !token_right.is_reserved_ident())
172 || token_right.is_lit()) => token_left.span,
173 ((TokenTree::Delimited(sp, ..), NonJoint), _) => sp.entire(),
176 let sp = sp.shrink_to_hi();
177 let comma = (TokenTree::token(token::Comma, sp), NonJoint);
178 suggestion = Some((pos, comma, sp));
181 if let Some((pos, comma, sp)) = suggestion {
182 let mut new_stream = vec![];
183 let parts = self.0.split_at(pos + 1);
184 new_stream.extend_from_slice(parts.0);
185 new_stream.push(comma);
186 new_stream.extend_from_slice(parts.1);
187 return Some((TokenStream::new(new_stream), sp));
193 impl From<TokenTree> for TokenStream {
194 fn from(tree: TokenTree) -> TokenStream {
195 TokenStream::new(vec![(tree, NonJoint)])
199 impl From<TokenTree> for TreeAndJoint {
200 fn from(tree: TokenTree) -> TreeAndJoint {
205 impl<T: Into<TokenStream>> iter::FromIterator<T> for TokenStream {
206 fn from_iter<I: IntoIterator<Item = T>>(iter: I) -> Self {
207 TokenStream::from_streams(iter.into_iter().map(Into::into).collect::<SmallVec<_>>())
211 impl Eq for TokenStream {}
213 impl PartialEq<TokenStream> for TokenStream {
214 fn eq(&self, other: &TokenStream) -> bool {
215 self.trees().eq(other.trees())
220 pub fn new(streams: Vec<TreeAndJoint>) -> TokenStream {
221 TokenStream(Lrc::new(streams))
224 pub fn is_empty(&self) -> bool {
228 pub fn len(&self) -> usize {
232 pub(crate) fn from_streams(mut streams: SmallVec<[TokenStream; 2]>) -> TokenStream {
233 match streams.len() {
234 0 => TokenStream::default(),
235 1 => streams.pop().unwrap(),
237 // We are going to extend the first stream in `streams` with
238 // the elements from the subsequent streams. This requires
239 // using `make_mut()` on the first stream, and in practice this
240 // doesn't cause cloning 99.9% of the time.
242 // One very common use case is when `streams` has two elements,
243 // where the first stream has any number of elements within
244 // (often 1, but sometimes many more) and the second stream has
245 // a single element within.
247 // Determine how much the first stream will be extended.
248 // Needed to avoid quadratic blow up from on-the-fly
249 // reallocations (#57735).
250 let num_appends = streams.iter()
255 // Get the first stream. If it's `None`, create an empty
257 let mut iter = streams.drain();
258 let mut first_stream_lrc = iter.next().unwrap().0;
260 // Append the elements to the first stream, after reserving
262 let first_vec_mut = Lrc::make_mut(&mut first_stream_lrc);
263 first_vec_mut.reserve(num_appends);
265 first_vec_mut.extend(stream.0.iter().cloned());
268 // Create the final `TokenStream`.
269 TokenStream(first_stream_lrc)
274 pub fn append_to_tree_and_joint_vec(self, vec: &mut Vec<TreeAndJoint>) {
275 vec.extend(self.0.iter().cloned());
278 pub fn trees(&self) -> Cursor {
279 self.clone().into_trees()
282 pub fn into_trees(self) -> Cursor {
286 /// Compares two `TokenStream`s, checking equality without regarding span information.
287 pub fn eq_unspanned(&self, other: &TokenStream) -> bool {
288 let mut t1 = self.trees();
289 let mut t2 = other.trees();
290 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
291 if !t1.eq_unspanned(&t2) {
295 t1.next().is_none() && t2.next().is_none()
298 // See comments in `Nonterminal::to_tokenstream` for why we care about
299 // *probably* equal here rather than actual equality
301 // This is otherwise the same as `eq_unspanned`, only recursing with a
303 pub fn probably_equal_for_proc_macro(&self, other: &TokenStream) -> bool {
304 // When checking for `probably_eq`, we ignore certain tokens that aren't
305 // preserved in the AST. Because they are not preserved, the pretty
306 // printer arbitrarily adds or removes them when printing as token
307 // streams, making a comparison between a token stream generated from an
308 // AST and a token stream which was parsed into an AST more reliable.
309 fn semantic_tree(tree: &TokenTree) -> bool {
310 if let TokenTree::Token(token) = tree {
312 // The pretty printer tends to add trailing commas to
313 // everything, and in particular, after struct fields.
315 // The pretty printer emits `NoDelim` as whitespace.
316 | token::OpenDelim(DelimToken::NoDelim)
317 | token::CloseDelim(DelimToken::NoDelim)
318 // The pretty printer collapses many semicolons into one.
320 // The pretty printer collapses whitespace arbitrarily and can
321 // introduce whitespace from `NoDelim`.
323 // The pretty printer can turn `$crate` into `::crate_name`
324 | token::ModSep = token.kind {
331 let mut t1 = self.trees().filter(semantic_tree);
332 let mut t2 = other.trees().filter(semantic_tree);
333 for (t1, t2) in t1.by_ref().zip(t2.by_ref()) {
334 if !t1.probably_equal_for_proc_macro(&t2) {
338 t1.next().is_none() && t2.next().is_none()
341 pub fn map_enumerated<F: FnMut(usize, TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
342 TokenStream(Lrc::new(
346 .map(|(i, (tree, is_joint))| (f(i, tree.clone()), *is_joint))
351 pub fn map<F: FnMut(TokenTree) -> TokenTree>(self, mut f: F) -> TokenStream {
352 TokenStream(Lrc::new(
355 .map(|(tree, is_joint)| (f(tree.clone()), *is_joint))
361 // 99.5%+ of the time we have 1 or 2 elements in this vector.
363 pub struct TokenStreamBuilder(SmallVec<[TokenStream; 2]>);
365 impl TokenStreamBuilder {
366 pub fn new() -> TokenStreamBuilder {
367 TokenStreamBuilder(SmallVec::new())
370 pub fn push<T: Into<TokenStream>>(&mut self, stream: T) {
371 let mut stream = stream.into();
373 // If `self` is not empty and the last tree within the last stream is a
374 // token tree marked with `Joint`...
375 if let Some(TokenStream(ref mut last_stream_lrc)) = self.0.last_mut() {
376 if let Some((TokenTree::Token(last_token), Joint)) = last_stream_lrc.last() {
378 // ...and `stream` is not empty and the first tree within it is
380 let TokenStream(ref mut stream_lrc) = stream;
381 if let Some((TokenTree::Token(token), is_joint)) = stream_lrc.first() {
383 // ...and the two tokens can be glued together...
384 if let Some(glued_tok) = last_token.glue(&token) {
386 // ...then do so, by overwriting the last token
387 // tree in `self` and removing the first token tree
388 // from `stream`. This requires using `make_mut()`
389 // on the last stream in `self` and on `stream`,
390 // and in practice this doesn't cause cloning 99.9%
393 // Overwrite the last token tree with the merged
395 let last_vec_mut = Lrc::make_mut(last_stream_lrc);
396 *last_vec_mut.last_mut().unwrap() =
397 (TokenTree::Token(glued_tok), *is_joint);
399 // Remove the first token tree from `stream`. (This
400 // is almost always the only tree in `stream`.)
401 let stream_vec_mut = Lrc::make_mut(stream_lrc);
402 stream_vec_mut.remove(0);
404 // Don't push `stream` if it's empty -- that could
405 // block subsequent token gluing, by getting
406 // between two token trees that should be glued
408 if !stream.is_empty() {
419 pub fn build(self) -> TokenStream {
420 TokenStream::from_streams(self.0)
426 pub stream: TokenStream,
430 impl Iterator for Cursor {
431 type Item = TokenTree;
433 fn next(&mut self) -> Option<TokenTree> {
434 self.next_with_joint().map(|(tree, _)| tree)
439 fn new(stream: TokenStream) -> Self {
440 Cursor { stream, index: 0 }
443 pub fn next_with_joint(&mut self) -> Option<TreeAndJoint> {
444 if self.index < self.stream.len() {
446 Some(self.stream.0[self.index - 1].clone())
452 pub fn append(&mut self, new_stream: TokenStream) {
453 if new_stream.is_empty() {
456 let index = self.index;
457 let stream = mem::take(&mut self.stream);
458 *self = TokenStream::from_streams(smallvec![stream, new_stream]).into_trees();
462 pub fn look_ahead(&self, n: usize) -> Option<TokenTree> {
463 self.stream.0[self.index ..].get(n).map(|(tree, _)| tree.clone())
467 impl Encodable for TokenStream {
468 fn encode<E: Encoder>(&self, encoder: &mut E) -> Result<(), E::Error> {
469 self.trees().collect::<Vec<_>>().encode(encoder)
473 impl Decodable for TokenStream {
474 fn decode<D: Decoder>(decoder: &mut D) -> Result<TokenStream, D::Error> {
475 Vec::<TokenTree>::decode(decoder).map(|vec| vec.into_iter().collect())
479 #[derive(Debug, Copy, Clone, PartialEq, RustcEncodable, RustcDecodable)]
480 pub struct DelimSpan {
486 pub fn from_single(sp: Span) -> Self {
493 pub fn from_pair(open: Span, close: Span) -> Self {
494 DelimSpan { open, close }
497 pub fn dummy() -> Self {
498 Self::from_single(DUMMY_SP)
501 pub fn entire(self) -> Span {
502 self.open.with_hi(self.close.hi())