1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
3 use parser::{ParseError, TreeSink};
4 use rustc_hash::{FxHashMap, FxHashSet};
6 ast::{self, make::tokens::doc_comment},
7 tokenize, AstToken, Parse, PreorderWithTokens, SmolStr, SyntaxElement, SyntaxKind,
9 SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, Token as RawToken, WalkEvent,
12 use tt::buffer::{Cursor, TokenBuffer};
15 subtree_source::SubtreeTokenSource, tt_iter::TtIter, ExpandError, ParserEntryPoint, TokenMap,
18 /// Convert the syntax node to a `TokenTree` (what macro
20 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
21 syntax_node_to_token_tree_censored(node, &Default::default())
24 /// Convert the syntax node to a `TokenTree` (what macro will consume)
25 /// with the censored range excluded.
26 pub fn syntax_node_to_token_tree_censored(
28 censor: &FxHashSet<SyntaxNode>,
29 ) -> (tt::Subtree, TokenMap) {
30 let global_offset = node.text_range().start();
31 let mut c = Convertor::new(node, global_offset, censor);
32 let subtree = convert_tokens(&mut c);
33 c.id_alloc.map.shrink_to_fit();
34 (subtree, c.id_alloc.map)
37 // The following items are what `rustc` macro can be parsed into :
38 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
39 // * Expr(P<ast::Expr>) -> token_tree_to_expr
40 // * Pat(P<ast::Pat>) -> token_tree_to_pat
41 // * Ty(P<ast::Ty>) -> token_tree_to_ty
42 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
43 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
45 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
46 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
47 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
49 pub fn token_tree_to_syntax_node(
51 entry_point: ParserEntryPoint,
52 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
53 let buffer = match tt {
54 tt::Subtree { delimiter: None, token_trees } => {
55 TokenBuffer::from_tokens(token_trees.as_slice())
57 _ => TokenBuffer::from_subtree(tt),
59 let mut token_source = SubtreeTokenSource::new(&buffer);
60 let mut tree_sink = TtTreeSink::new(buffer.begin());
61 parser::parse(&mut token_source, &mut tree_sink, entry_point);
62 if tree_sink.roots.len() != 1 {
63 return Err(ExpandError::ConversionError);
65 //FIXME: would be cool to report errors
66 let (parse, range_map) = tree_sink.finish();
67 Ok((parse, range_map))
70 /// Convert a string to a `TokenTree`
71 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
72 let (tokens, errors) = tokenize(text);
73 if !errors.is_empty() {
77 let mut conv = RawConvertor {
79 offset: TextSize::default(),
81 id_alloc: TokenIdAlloc {
82 map: Default::default(),
83 global_offset: TextSize::default(),
88 let subtree = convert_tokens(&mut conv);
89 Some((subtree, conv.id_alloc.map))
92 /// Split token tree with separate expr: $($e:expr)SEP*
93 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
94 if tt.token_trees.is_empty() {
98 let mut iter = TtIter::new(tt);
99 let mut res = Vec::new();
101 while iter.peek_n(0).is_some() {
102 let expanded = iter.expect_fragment(ParserEntryPoint::Expr);
104 res.push(match expanded.value {
106 Some(tt @ tt::TokenTree::Leaf(_)) => {
107 tt::Subtree { delimiter: None, token_trees: vec![tt] }
109 Some(tt::TokenTree::Subtree(tt)) => tt,
112 let mut fork = iter.clone();
113 if fork.expect_char(sep).is_err() {
119 if iter.peek_n(0).is_some() {
120 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
126 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
128 subtree: tt::Subtree,
130 open_range: TextRange,
133 let entry = StackEntry {
134 subtree: tt::Subtree { delimiter: None, ..Default::default() },
135 // never used (delimiter is `None`)
137 open_range: TextRange::empty(TextSize::of('.')),
139 let mut stack = vec![entry];
142 let entry = stack.last_mut().unwrap();
143 let result = &mut entry.subtree.token_trees;
144 let (token, range) = match conv.bump() {
149 let k: SyntaxKind = token.kind();
151 if let Some(tokens) = conv.convert_doc_comment(&token) {
152 // FIXME: There has to be a better way to do this
153 // Add the comments token id to the converted doc string
154 let id = conv.id_alloc().alloc(range);
155 result.extend(tokens.into_iter().map(|mut tt| {
156 if let tt::TokenTree::Subtree(sub) = &mut tt {
157 if let tt::TokenTree::Leaf(tt::Leaf::Literal(lit)) = &mut sub.token_trees[2]
168 result.push(if k.is_punct() && k != UNDERSCORE {
169 assert_eq!(range.len(), TextSize::of('.'));
171 if let Some(delim) = entry.subtree.delimiter {
172 let expected = match delim.kind {
173 tt::DelimiterKind::Parenthesis => T![')'],
174 tt::DelimiterKind::Brace => T!['}'],
175 tt::DelimiterKind::Bracket => T![']'],
179 let entry = stack.pop().unwrap();
180 conv.id_alloc().close_delim(entry.idx, Some(range));
181 stack.last_mut().unwrap().subtree.token_trees.push(entry.subtree.into());
186 let delim = match k {
187 T!['('] => Some(tt::DelimiterKind::Parenthesis),
188 T!['{'] => Some(tt::DelimiterKind::Brace),
189 T!['['] => Some(tt::DelimiterKind::Bracket),
193 if let Some(kind) = delim {
194 let mut subtree = tt::Subtree::default();
195 let (id, idx) = conv.id_alloc().open_delim(range);
196 subtree.delimiter = Some(tt::Delimiter { id, kind });
197 stack.push(StackEntry { subtree, idx, open_range: range });
200 let spacing = match conv.peek() {
202 if next.kind().is_trivia()
203 || next.kind() == T!['[']
204 || next.kind() == T!['{']
205 || next.kind() == T!['('] =>
209 Some(next) if next.kind().is_punct() && next.kind() != UNDERSCORE => {
212 _ => tt::Spacing::Alone,
214 let char = match token.to_char() {
217 panic!("Token from lexer must be single char: token = {:#?}", token);
220 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
223 macro_rules! make_leaf {
225 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text() }.into()
228 let leaf: tt::Leaf = match k {
229 T![true] | T![false] => make_leaf!(Ident),
230 IDENT => make_leaf!(Ident),
231 UNDERSCORE => make_leaf!(Ident),
232 k if k.is_keyword() => make_leaf!(Ident),
233 k if k.is_literal() => make_leaf!(Literal),
235 let char_unit = TextSize::of('\'');
236 let r = TextRange::at(range.start(), char_unit);
237 let apostrophe = tt::Leaf::from(tt::Punct {
239 spacing: tt::Spacing::Joint,
240 id: conv.id_alloc().alloc(r),
242 result.push(apostrophe.into());
244 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
245 let ident = tt::Leaf::from(tt::Ident {
246 text: SmolStr::new(&token.to_text()[1..]),
247 id: conv.id_alloc().alloc(r),
249 result.push(ident.into());
259 // If we get here, we've consumed all input tokens.
260 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
261 // Merge them so we're left with one.
262 while stack.len() > 1 {
263 let entry = stack.pop().unwrap();
264 let parent = stack.last_mut().unwrap();
266 conv.id_alloc().close_delim(entry.idx, None);
267 let leaf: tt::Leaf = tt::Punct {
268 id: conv.id_alloc().alloc(entry.open_range),
269 char: match entry.subtree.delimiter.unwrap().kind {
270 tt::DelimiterKind::Parenthesis => '(',
271 tt::DelimiterKind::Brace => '{',
272 tt::DelimiterKind::Bracket => '[',
274 spacing: tt::Spacing::Alone,
277 parent.subtree.token_trees.push(leaf.into());
278 parent.subtree.token_trees.extend(entry.subtree.token_trees);
281 let subtree = stack.pop().unwrap().subtree;
282 if subtree.token_trees.len() == 1 {
283 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
284 return first.clone();
290 /// Returns the textual content of a doc comment block as a quoted string
291 /// That is, strips leading `///` (or `/**`, etc)
292 /// and strips the ending `*/`
293 /// And then quote the string, which is needed to convert to `tt::Literal`
294 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
295 let prefix_len = comment.prefix().len();
296 let mut text = &comment.text()[prefix_len..];
298 // Remove ending "*/"
299 if comment.kind().shape == ast::CommentShape::Block {
300 text = &text[0..text.len() - 2];
304 // Note that `tt::Literal` expect an escaped string
305 let text = format!("\"{}\"", text.escape_debug());
309 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
310 let comment = ast::Comment::cast(token.clone())?;
311 let doc = comment.kind().doc?;
313 // Make `doc="\" Comments\""
314 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
317 let mut token_trees = vec![mk_punct('#')];
318 if let ast::CommentPlacement::Inner = doc {
319 token_trees.push(mk_punct('!'));
321 token_trees.push(tt::TokenTree::from(tt::Subtree {
322 delimiter: Some(tt::Delimiter {
323 kind: tt::DelimiterKind::Bracket,
324 id: tt::TokenId::unspecified(),
326 token_trees: meta_tkns,
329 return Some(token_trees);
332 fn mk_ident(s: &str) -> tt::TokenTree {
333 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
335 id: tt::TokenId::unspecified(),
339 fn mk_punct(c: char) -> tt::TokenTree {
340 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
342 spacing: tt::Spacing::Alone,
343 id: tt::TokenId::unspecified(),
347 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
348 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
350 tt::TokenTree::from(tt::Leaf::from(lit))
354 struct TokenIdAlloc {
356 global_offset: TextSize,
361 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
362 let relative_range = absolute_range - self.global_offset;
363 let token_id = tt::TokenId(self.next_id);
365 self.map.insert(token_id, relative_range);
369 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
370 let token_id = tt::TokenId(self.next_id);
372 let idx = self.map.insert_delim(
374 open_abs_range - self.global_offset,
375 open_abs_range - self.global_offset,
380 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
381 match close_abs_range {
383 self.map.remove_delim(idx);
386 self.map.update_close_delim(idx, close - self.global_offset);
392 /// A Raw Token (straightly from lexer) convertor
393 struct RawConvertor<'a> {
396 id_alloc: TokenIdAlloc,
397 inner: std::slice::Iter<'a, RawToken>,
400 trait SrcToken: std::fmt::Debug {
401 fn kind(&self) -> SyntaxKind;
403 fn to_char(&self) -> Option<char>;
405 fn to_text(&self) -> SmolStr;
408 trait TokenConvertor {
409 type Token: SrcToken;
411 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
413 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
415 fn peek(&self) -> Option<Self::Token>;
417 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
420 impl<'a> SrcToken for (&'a RawToken, &'a str) {
421 fn kind(&self) -> SyntaxKind {
425 fn to_char(&self) -> Option<char> {
426 self.1.chars().next()
429 fn to_text(&self) -> SmolStr {
434 impl<'a> TokenConvertor for RawConvertor<'a> {
435 type Token = (&'a RawToken, &'a str);
437 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
438 convert_doc_comment(&doc_comment(token.1))
441 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
442 let token = self.inner.next()?;
443 let range = TextRange::at(self.offset, token.len);
444 self.offset += token.len;
446 Some(((token, &self.text[range]), range))
449 fn peek(&self) -> Option<Self::Token> {
450 let token = self.inner.as_slice().get(0);
453 let range = TextRange::at(self.offset, it.len);
454 (it, &self.text[range])
458 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
463 struct Convertor<'c> {
464 id_alloc: TokenIdAlloc,
465 current: Option<SyntaxToken>,
466 preorder: PreorderWithTokens,
467 censor: &'c FxHashSet<SyntaxNode>,
469 punct_offset: Option<(SyntaxToken, TextSize)>,
472 impl<'c> Convertor<'c> {
475 global_offset: TextSize,
476 censor: &'c FxHashSet<SyntaxNode>,
478 let range = node.text_range();
479 let mut preorder = node.preorder_with_tokens();
480 let first = Self::next_token(&mut preorder, censor);
482 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
492 preorder: &mut PreorderWithTokens,
493 censor: &FxHashSet<SyntaxNode>,
494 ) -> Option<SyntaxToken> {
495 while let Some(ev) = preorder.next() {
497 WalkEvent::Enter(ele) => ele,
501 SyntaxElement::Token(t) => return Some(t),
502 SyntaxElement::Node(node) if censor.contains(&node) => preorder.skip_subtree(),
503 SyntaxElement::Node(_) => (),
512 Ordinary(SyntaxToken),
513 Punch(SyntaxToken, TextSize),
517 fn token(&self) -> &SyntaxToken {
519 SynToken::Ordinary(it) => it,
520 SynToken::Punch(it, _) => it,
525 impl SrcToken for SynToken {
526 fn kind(&self) -> SyntaxKind {
529 fn to_char(&self) -> Option<char> {
531 SynToken::Ordinary(_) => None,
532 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
535 fn to_text(&self) -> SmolStr {
536 self.token().text().into()
540 impl TokenConvertor for Convertor<'_> {
541 type Token = SynToken;
542 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
543 convert_doc_comment(token.token())
546 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
547 if let Some((punct, offset)) = self.punct_offset.clone() {
548 if usize::from(offset) + 1 < punct.text().len() {
549 let offset = offset + TextSize::of('.');
550 let range = punct.text_range();
551 self.punct_offset = Some((punct.clone(), offset));
552 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
553 return Some((SynToken::Punch(punct, offset), range));
557 let curr = self.current.clone()?;
558 if !&self.range.contains_range(curr.text_range()) {
561 self.current = Self::next_token(&mut self.preorder, self.censor);
562 let token = if curr.kind().is_punct() {
563 let range = curr.text_range();
564 let range = TextRange::at(range.start(), TextSize::of('.'));
565 self.punct_offset = Some((curr.clone(), 0.into()));
566 (SynToken::Punch(curr, 0.into()), range)
568 self.punct_offset = None;
569 let range = curr.text_range();
570 (SynToken::Ordinary(curr), range)
576 fn peek(&self) -> Option<Self::Token> {
577 if let Some((punct, mut offset)) = self.punct_offset.clone() {
578 offset += TextSize::of('.');
579 if usize::from(offset) < punct.text().len() {
580 return Some(SynToken::Punch(punct, offset));
584 let curr = self.current.clone()?;
585 if !self.range.contains_range(curr.text_range()) {
589 let token = if curr.kind().is_punct() {
590 SynToken::Punch(curr, 0.into())
592 SynToken::Ordinary(curr)
597 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
602 struct TtTreeSink<'a> {
605 open_delims: FxHashMap<tt::TokenId, TextSize>,
607 inner: SyntaxTreeBuilder,
611 // Use for detect ill-form tree which is not single root
612 roots: smallvec::SmallVec<[usize; 1]>,
615 impl<'a> TtTreeSink<'a> {
616 fn new(cursor: Cursor<'a>) -> Self {
620 open_delims: FxHashMap::default(),
622 inner: SyntaxTreeBuilder::default(),
623 roots: smallvec::SmallVec::new(),
624 token_map: TokenMap::default(),
628 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
629 self.token_map.shrink_to_fit();
630 (self.inner.finish(), self.token_map)
634 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> &'static str {
635 let texts = match d {
636 Some(tt::DelimiterKind::Parenthesis) => "()",
637 Some(tt::DelimiterKind::Brace) => "{}",
638 Some(tt::DelimiterKind::Bracket) => "[]",
642 let idx = closing as usize;
643 &texts[idx..texts.len() - (1 - idx)]
646 impl<'a> TreeSink for TtTreeSink<'a> {
647 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
648 if kind == L_DOLLAR || kind == R_DOLLAR {
649 self.cursor = self.cursor.bump_subtree();
652 if kind == LIFETIME_IDENT {
656 let mut last = self.cursor;
657 for _ in 0..n_tokens {
658 let tmp_str: SmolStr;
659 if self.cursor.eof() {
663 let text: &str = match self.cursor.token_tree() {
664 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
665 // Mark the range if needed
666 let (text, id) = match leaf {
667 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
668 tt::Leaf::Punct(punct) => {
669 assert!(punct.char.is_ascii());
670 let char = &(punct.char as u8);
671 tmp_str = SmolStr::new_inline(
672 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
676 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
678 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
679 self.token_map.insert(id, range);
680 self.cursor = self.cursor.bump();
683 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
684 self.cursor = self.cursor.subtree().unwrap();
685 if let Some(id) = subtree.delimiter.map(|it| it.id) {
686 self.open_delims.insert(id, self.text_pos);
688 delim_to_str(subtree.delimiter_kind(), false)
691 if let Some(parent) = self.cursor.end() {
692 self.cursor = self.cursor.bump();
693 if let Some(id) = parent.delimiter.map(|it| it.id) {
694 if let Some(open_delim) = self.open_delims.get(&id) {
695 let open_range = TextRange::at(*open_delim, TextSize::of('('));
696 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
697 self.token_map.insert_delim(id, open_range, close_range);
700 delim_to_str(parent.delimiter_kind(), true)
707 self.text_pos += TextSize::of(text);
710 self.inner.token(kind, self.buf.as_str());
712 // Add whitespace between adjoint puncts
713 let next = last.bump();
715 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
716 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
717 ) = (last.token_tree(), next.token_tree())
719 // Note: We always assume the semi-colon would be the last token in
720 // other parts of RA such that we don't add whitespace here.
721 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
722 self.inner.token(WHITESPACE, " ");
723 self.text_pos += TextSize::of(' ');
728 fn start_node(&mut self, kind: SyntaxKind) {
729 self.inner.start_node(kind);
731 match self.roots.last_mut() {
732 None | Some(0) => self.roots.push(1),
733 Some(ref mut n) => **n += 1,
737 fn finish_node(&mut self) {
738 self.inner.finish_node();
739 *self.roots.last_mut().unwrap() -= 1;
742 fn error(&mut self, error: ParseError) {
743 self.inner.error(error, self.text_pos)