1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
3 use rustc_hash::{FxHashMap, FxHashSet};
5 ast::{self, make::tokens::doc_comment},
6 AstToken, Parse, PreorderWithTokens, SmolStr, SyntaxElement, SyntaxKind,
8 SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, WalkEvent, T,
10 use tt::buffer::{Cursor, TokenBuffer};
13 to_parser_tokens::to_parser_tokens, tt_iter::TtIter, ExpandError, ParserEntryPoint, TokenMap,
16 /// Convert the syntax node to a `TokenTree` (what macro
18 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
19 syntax_node_to_token_tree_censored(node, &Default::default())
22 /// Convert the syntax node to a `TokenTree` (what macro will consume)
23 /// with the censored range excluded.
24 pub fn syntax_node_to_token_tree_censored(
26 censor: &FxHashSet<SyntaxNode>,
27 ) -> (tt::Subtree, TokenMap) {
28 let global_offset = node.text_range().start();
29 let mut c = Convertor::new(node, global_offset, censor);
30 let subtree = convert_tokens(&mut c);
31 c.id_alloc.map.shrink_to_fit();
32 (subtree, c.id_alloc.map)
35 // The following items are what `rustc` macro can be parsed into :
36 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
37 // * Expr(P<ast::Expr>) -> token_tree_to_expr
38 // * Pat(P<ast::Pat>) -> token_tree_to_pat
39 // * Ty(P<ast::Ty>) -> token_tree_to_ty
40 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
41 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
43 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
44 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
45 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
47 pub fn token_tree_to_syntax_node(
49 entry_point: ParserEntryPoint,
50 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
51 let buffer = match tt {
52 tt::Subtree { delimiter: None, token_trees } => {
53 TokenBuffer::from_tokens(token_trees.as_slice())
55 _ => TokenBuffer::from_subtree(tt),
57 let parser_tokens = to_parser_tokens(&buffer);
58 let tree_traversal = parser::parse(&parser_tokens, entry_point);
59 let mut tree_sink = TtTreeSink::new(buffer.begin());
60 for event in tree_traversal.iter() {
62 parser::TraversalStep::Token { kind, n_raw_tokens } => {
63 tree_sink.token(kind, n_raw_tokens)
65 parser::TraversalStep::EnterNode { kind } => tree_sink.start_node(kind),
66 parser::TraversalStep::LeaveNode => tree_sink.finish_node(),
67 parser::TraversalStep::Error { msg } => tree_sink.error(msg.to_string()),
70 if tree_sink.roots.len() != 1 {
71 return Err(ExpandError::ConversionError);
73 //FIXME: would be cool to report errors
74 let (parse, range_map) = tree_sink.finish();
75 Ok((parse, range_map))
78 /// Convert a string to a `TokenTree`
79 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
80 let lexed = parser::LexedStr::new(text);
81 if lexed.errors().next().is_some() {
85 let mut conv = RawConvertor {
88 id_alloc: TokenIdAlloc {
89 map: Default::default(),
90 global_offset: TextSize::default(),
95 let subtree = convert_tokens(&mut conv);
96 Some((subtree, conv.id_alloc.map))
99 /// Split token tree with separate expr: $($e:expr)SEP*
100 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
101 if tt.token_trees.is_empty() {
105 let mut iter = TtIter::new(tt);
106 let mut res = Vec::new();
108 while iter.peek_n(0).is_some() {
109 let expanded = iter.expect_fragment(ParserEntryPoint::Expr);
111 res.push(match expanded.value {
113 Some(tt @ tt::TokenTree::Leaf(_)) => {
114 tt::Subtree { delimiter: None, token_trees: vec![tt] }
116 Some(tt::TokenTree::Subtree(tt)) => tt,
119 let mut fork = iter.clone();
120 if fork.expect_char(sep).is_err() {
126 if iter.peek_n(0).is_some() {
127 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
133 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
135 subtree: tt::Subtree,
137 open_range: TextRange,
140 let entry = StackEntry {
141 subtree: tt::Subtree { delimiter: None, ..Default::default() },
142 // never used (delimiter is `None`)
144 open_range: TextRange::empty(TextSize::of('.')),
146 let mut stack = vec![entry];
149 let entry = stack.last_mut().unwrap();
150 let result = &mut entry.subtree.token_trees;
151 let (token, range) = match conv.bump() {
156 let k: SyntaxKind = token.kind(&conv);
158 if let Some(tokens) = conv.convert_doc_comment(&token) {
159 // FIXME: There has to be a better way to do this
160 // Add the comments token id to the converted doc string
161 let id = conv.id_alloc().alloc(range);
162 result.extend(tokens.into_iter().map(|mut tt| {
163 if let tt::TokenTree::Subtree(sub) = &mut tt {
164 if let tt::TokenTree::Leaf(tt::Leaf::Literal(lit)) = &mut sub.token_trees[2]
175 result.push(if k.is_punct() && k != UNDERSCORE {
176 assert_eq!(range.len(), TextSize::of('.'));
178 if let Some(delim) = entry.subtree.delimiter {
179 let expected = match delim.kind {
180 tt::DelimiterKind::Parenthesis => T![')'],
181 tt::DelimiterKind::Brace => T!['}'],
182 tt::DelimiterKind::Bracket => T![']'],
186 let entry = stack.pop().unwrap();
187 conv.id_alloc().close_delim(entry.idx, Some(range));
188 stack.last_mut().unwrap().subtree.token_trees.push(entry.subtree.into());
193 let delim = match k {
194 T!['('] => Some(tt::DelimiterKind::Parenthesis),
195 T!['{'] => Some(tt::DelimiterKind::Brace),
196 T!['['] => Some(tt::DelimiterKind::Bracket),
200 if let Some(kind) = delim {
201 let mut subtree = tt::Subtree::default();
202 let (id, idx) = conv.id_alloc().open_delim(range);
203 subtree.delimiter = Some(tt::Delimiter { id, kind });
204 stack.push(StackEntry { subtree, idx, open_range: range });
207 let spacing = match conv.peek() {
209 if next.kind(&conv).is_trivia()
210 || next.kind(&conv) == T!['[']
211 || next.kind(&conv) == T!['{']
212 || next.kind(&conv) == T!['('] =>
216 Some(next) if next.kind(&conv).is_punct() && next.kind(&conv) != UNDERSCORE => {
219 _ => tt::Spacing::Alone,
221 let char = match token.to_char(&conv) {
224 panic!("Token from lexer must be single char: token = {:#?}", token);
227 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
230 macro_rules! make_leaf {
232 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text(conv) }.into()
235 let leaf: tt::Leaf = match k {
236 T![true] | T![false] => make_leaf!(Ident),
237 IDENT => make_leaf!(Ident),
238 UNDERSCORE => make_leaf!(Ident),
239 k if k.is_keyword() => make_leaf!(Ident),
240 k if k.is_literal() => make_leaf!(Literal),
242 let char_unit = TextSize::of('\'');
243 let r = TextRange::at(range.start(), char_unit);
244 let apostrophe = tt::Leaf::from(tt::Punct {
246 spacing: tt::Spacing::Joint,
247 id: conv.id_alloc().alloc(r),
249 result.push(apostrophe.into());
251 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
252 let ident = tt::Leaf::from(tt::Ident {
253 text: SmolStr::new(&token.to_text(conv)[1..]),
254 id: conv.id_alloc().alloc(r),
256 result.push(ident.into());
266 // If we get here, we've consumed all input tokens.
267 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
268 // Merge them so we're left with one.
269 while stack.len() > 1 {
270 let entry = stack.pop().unwrap();
271 let parent = stack.last_mut().unwrap();
273 conv.id_alloc().close_delim(entry.idx, None);
274 let leaf: tt::Leaf = tt::Punct {
275 id: conv.id_alloc().alloc(entry.open_range),
276 char: match entry.subtree.delimiter.unwrap().kind {
277 tt::DelimiterKind::Parenthesis => '(',
278 tt::DelimiterKind::Brace => '{',
279 tt::DelimiterKind::Bracket => '[',
281 spacing: tt::Spacing::Alone,
284 parent.subtree.token_trees.push(leaf.into());
285 parent.subtree.token_trees.extend(entry.subtree.token_trees);
288 let subtree = stack.pop().unwrap().subtree;
289 if subtree.token_trees.len() == 1 {
290 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
291 return first.clone();
297 /// Returns the textual content of a doc comment block as a quoted string
298 /// That is, strips leading `///` (or `/**`, etc)
299 /// and strips the ending `*/`
300 /// And then quote the string, which is needed to convert to `tt::Literal`
301 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
302 let prefix_len = comment.prefix().len();
303 let mut text = &comment.text()[prefix_len..];
305 // Remove ending "*/"
306 if comment.kind().shape == ast::CommentShape::Block {
307 text = &text[0..text.len() - 2];
311 // Note that `tt::Literal` expect an escaped string
312 let text = format!("\"{}\"", text.escape_debug());
316 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
317 cov_mark::hit!(test_meta_doc_comments);
318 let comment = ast::Comment::cast(token.clone())?;
319 let doc = comment.kind().doc?;
321 // Make `doc="\" Comments\""
322 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
325 let mut token_trees = vec![mk_punct('#')];
326 if let ast::CommentPlacement::Inner = doc {
327 token_trees.push(mk_punct('!'));
329 token_trees.push(tt::TokenTree::from(tt::Subtree {
330 delimiter: Some(tt::Delimiter {
331 kind: tt::DelimiterKind::Bracket,
332 id: tt::TokenId::unspecified(),
334 token_trees: meta_tkns,
337 return Some(token_trees);
340 fn mk_ident(s: &str) -> tt::TokenTree {
341 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
343 id: tt::TokenId::unspecified(),
347 fn mk_punct(c: char) -> tt::TokenTree {
348 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
350 spacing: tt::Spacing::Alone,
351 id: tt::TokenId::unspecified(),
355 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
356 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
358 tt::TokenTree::from(tt::Leaf::from(lit))
362 struct TokenIdAlloc {
364 global_offset: TextSize,
369 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
370 let relative_range = absolute_range - self.global_offset;
371 let token_id = tt::TokenId(self.next_id);
373 self.map.insert(token_id, relative_range);
377 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
378 let token_id = tt::TokenId(self.next_id);
380 let idx = self.map.insert_delim(
382 open_abs_range - self.global_offset,
383 open_abs_range - self.global_offset,
388 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
389 match close_abs_range {
391 self.map.remove_delim(idx);
394 self.map.update_close_delim(idx, close - self.global_offset);
400 /// A Raw Token (straightly from lexer) convertor
401 struct RawConvertor<'a> {
402 lexed: parser::LexedStr<'a>,
404 id_alloc: TokenIdAlloc,
407 trait SrcToken<Ctx>: std::fmt::Debug {
408 fn kind(&self, ctx: &Ctx) -> SyntaxKind;
410 fn to_char(&self, ctx: &Ctx) -> Option<char>;
412 fn to_text(&self, ctx: &Ctx) -> SmolStr;
415 trait TokenConvertor: Sized {
416 type Token: SrcToken<Self>;
418 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
420 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
422 fn peek(&self) -> Option<Self::Token>;
424 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
427 impl<'a> SrcToken<RawConvertor<'a>> for usize {
428 fn kind(&self, ctx: &RawConvertor<'a>) -> SyntaxKind {
429 ctx.lexed.kind(*self)
432 fn to_char(&self, ctx: &RawConvertor<'a>) -> Option<char> {
433 ctx.lexed.text(*self).chars().next()
436 fn to_text(&self, ctx: &RawConvertor<'_>) -> SmolStr {
437 ctx.lexed.text(*self).into()
441 impl<'a> TokenConvertor for RawConvertor<'a> {
444 fn convert_doc_comment(&self, token: &usize) -> Option<Vec<tt::TokenTree>> {
445 let text = self.lexed.text(*token);
446 convert_doc_comment(&doc_comment(text))
449 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
450 if self.pos == self.lexed.len() {
453 let token = self.pos;
455 let range = self.lexed.text_range(token);
456 let range = TextRange::new(range.start.try_into().unwrap(), range.end.try_into().unwrap());
461 fn peek(&self) -> Option<Self::Token> {
462 if self.pos == self.lexed.len() {
468 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
473 struct Convertor<'c> {
474 id_alloc: TokenIdAlloc,
475 current: Option<SyntaxToken>,
476 preorder: PreorderWithTokens,
477 censor: &'c FxHashSet<SyntaxNode>,
479 punct_offset: Option<(SyntaxToken, TextSize)>,
482 impl<'c> Convertor<'c> {
485 global_offset: TextSize,
486 censor: &'c FxHashSet<SyntaxNode>,
488 let range = node.text_range();
489 let mut preorder = node.preorder_with_tokens();
490 let first = Self::next_token(&mut preorder, censor);
492 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
502 preorder: &mut PreorderWithTokens,
503 censor: &FxHashSet<SyntaxNode>,
504 ) -> Option<SyntaxToken> {
505 while let Some(ev) = preorder.next() {
507 WalkEvent::Enter(ele) => ele,
511 SyntaxElement::Token(t) => return Some(t),
512 SyntaxElement::Node(node) if censor.contains(&node) => preorder.skip_subtree(),
513 SyntaxElement::Node(_) => (),
522 Ordinary(SyntaxToken),
523 Punch(SyntaxToken, TextSize),
527 fn token(&self) -> &SyntaxToken {
529 SynToken::Ordinary(it) => it,
530 SynToken::Punch(it, _) => it,
535 impl<'a> SrcToken<Convertor<'a>> for SynToken {
536 fn kind(&self, _ctx: &Convertor<'a>) -> SyntaxKind {
539 fn to_char(&self, _ctx: &Convertor<'a>) -> Option<char> {
541 SynToken::Ordinary(_) => None,
542 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
545 fn to_text(&self, _ctx: &Convertor<'a>) -> SmolStr {
546 self.token().text().into()
550 impl TokenConvertor for Convertor<'_> {
551 type Token = SynToken;
552 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
553 convert_doc_comment(token.token())
556 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
557 if let Some((punct, offset)) = self.punct_offset.clone() {
558 if usize::from(offset) + 1 < punct.text().len() {
559 let offset = offset + TextSize::of('.');
560 let range = punct.text_range();
561 self.punct_offset = Some((punct.clone(), offset));
562 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
563 return Some((SynToken::Punch(punct, offset), range));
567 let curr = self.current.clone()?;
568 if !&self.range.contains_range(curr.text_range()) {
571 self.current = Self::next_token(&mut self.preorder, self.censor);
572 let token = if curr.kind().is_punct() {
573 let range = curr.text_range();
574 let range = TextRange::at(range.start(), TextSize::of('.'));
575 self.punct_offset = Some((curr.clone(), 0.into()));
576 (SynToken::Punch(curr, 0.into()), range)
578 self.punct_offset = None;
579 let range = curr.text_range();
580 (SynToken::Ordinary(curr), range)
586 fn peek(&self) -> Option<Self::Token> {
587 if let Some((punct, mut offset)) = self.punct_offset.clone() {
588 offset += TextSize::of('.');
589 if usize::from(offset) < punct.text().len() {
590 return Some(SynToken::Punch(punct, offset));
594 let curr = self.current.clone()?;
595 if !self.range.contains_range(curr.text_range()) {
599 let token = if curr.kind().is_punct() {
600 SynToken::Punch(curr, 0.into())
602 SynToken::Ordinary(curr)
607 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
612 struct TtTreeSink<'a> {
615 open_delims: FxHashMap<tt::TokenId, TextSize>,
617 inner: SyntaxTreeBuilder,
621 // Use for detect ill-form tree which is not single root
622 roots: smallvec::SmallVec<[usize; 1]>,
625 impl<'a> TtTreeSink<'a> {
626 fn new(cursor: Cursor<'a>) -> Self {
630 open_delims: FxHashMap::default(),
632 inner: SyntaxTreeBuilder::default(),
633 roots: smallvec::SmallVec::new(),
634 token_map: TokenMap::default(),
638 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
639 self.token_map.shrink_to_fit();
640 (self.inner.finish(), self.token_map)
644 fn delim_to_str(d: tt::DelimiterKind, closing: bool) -> &'static str {
645 let texts = match d {
646 tt::DelimiterKind::Parenthesis => "()",
647 tt::DelimiterKind::Brace => "{}",
648 tt::DelimiterKind::Bracket => "[]",
651 let idx = closing as usize;
652 &texts[idx..texts.len() - (1 - idx)]
655 impl<'a> TtTreeSink<'a> {
656 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
657 if kind == LIFETIME_IDENT {
661 let mut last = self.cursor;
662 for _ in 0..n_tokens {
663 let tmp_str: SmolStr;
664 if self.cursor.eof() {
668 let text: &str = loop {
669 break match self.cursor.token_tree() {
670 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
671 // Mark the range if needed
672 let (text, id) = match leaf {
673 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
674 tt::Leaf::Punct(punct) => {
675 assert!(punct.char.is_ascii());
676 let char = &(punct.char as u8);
677 tmp_str = SmolStr::new_inline(
678 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
682 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
684 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
685 self.token_map.insert(id, range);
686 self.cursor = self.cursor.bump();
689 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
690 self.cursor = self.cursor.subtree().unwrap();
691 match subtree.delimiter {
693 self.open_delims.insert(d.id, self.text_pos);
694 delim_to_str(d.kind, false)
700 let parent = self.cursor.end().unwrap();
701 self.cursor = self.cursor.bump();
702 match parent.delimiter {
704 if let Some(open_delim) = self.open_delims.get(&d.id) {
705 let open_range = TextRange::at(*open_delim, TextSize::of('('));
707 TextRange::at(self.text_pos, TextSize::of('('));
708 self.token_map.insert_delim(d.id, open_range, close_range);
710 delim_to_str(d.kind, true)
718 self.text_pos += TextSize::of(text);
721 self.inner.token(kind, self.buf.as_str());
723 // Add whitespace between adjoint puncts
724 let next = last.bump();
726 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
727 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
728 ) = (last.token_tree(), next.token_tree())
730 // Note: We always assume the semi-colon would be the last token in
731 // other parts of RA such that we don't add whitespace here.
732 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
733 self.inner.token(WHITESPACE, " ");
734 self.text_pos += TextSize::of(' ');
739 fn start_node(&mut self, kind: SyntaxKind) {
740 self.inner.start_node(kind);
742 match self.roots.last_mut() {
743 None | Some(0) => self.roots.push(1),
744 Some(ref mut n) => **n += 1,
748 fn finish_node(&mut self) {
749 self.inner.finish_node();
750 *self.roots.last_mut().unwrap() -= 1;
753 fn error(&mut self, error: String) {
754 self.inner.error(error, self.text_pos)