1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
5 use parser::{ParseError, TreeSink};
6 use rustc_hash::FxHashMap;
8 ast::{self, make::tokens::doc_comment},
9 tokenize, AstToken, Parse, SmolStr, SyntaxKind,
11 SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, Token as RawToken, T,
13 use tt::buffer::{Cursor, TokenBuffer};
16 subtree_source::SubtreeTokenSource, tt_iter::TtIter, ExpandError, ParserEntryPoint, TokenMap,
19 /// Convert the syntax node to a `TokenTree` (what macro
21 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
22 syntax_node_to_token_tree_censored(node, None)
25 /// Convert the syntax node to a `TokenTree` (what macro will consume)
26 /// with the censored range excluded.
27 pub fn syntax_node_to_token_tree_censored(
29 censor: Option<TextRange>,
30 ) -> (tt::Subtree, TokenMap) {
31 let global_offset = node.text_range().start();
32 let mut c = Convertor::new(node, global_offset, censor);
33 let subtree = convert_tokens(&mut c);
34 c.id_alloc.map.shrink_to_fit();
35 (subtree, c.id_alloc.map)
38 // The following items are what `rustc` macro can be parsed into :
39 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
40 // * Expr(P<ast::Expr>) -> token_tree_to_expr
41 // * Pat(P<ast::Pat>) -> token_tree_to_pat
42 // * Ty(P<ast::Ty>) -> token_tree_to_ty
43 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
44 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
46 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
47 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
48 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
50 pub fn token_tree_to_syntax_node(
52 entry_point: ParserEntryPoint,
53 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
54 let buffer = match tt {
55 tt::Subtree { delimiter: None, token_trees } => {
56 TokenBuffer::from_tokens(token_trees.as_slice())
58 _ => TokenBuffer::from_subtree(tt),
60 let mut token_source = SubtreeTokenSource::new(&buffer);
61 let mut tree_sink = TtTreeSink::new(buffer.begin());
62 parser::parse(&mut token_source, &mut tree_sink, entry_point);
63 if tree_sink.roots.len() != 1 {
64 return Err(ExpandError::ConversionError);
66 //FIXME: would be cool to report errors
67 let (parse, range_map) = tree_sink.finish();
68 Ok((parse, range_map))
71 /// Convert a string to a `TokenTree`
72 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
73 let (tokens, errors) = tokenize(text);
74 if !errors.is_empty() {
78 let mut conv = RawConvertor {
80 offset: TextSize::default(),
82 id_alloc: TokenIdAlloc {
83 map: Default::default(),
84 global_offset: TextSize::default(),
89 let subtree = convert_tokens(&mut conv);
90 Some((subtree, conv.id_alloc.map))
93 /// Split token tree with seperate expr: $($e:expr)SEP*
94 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
95 if tt.token_trees.is_empty() {
99 let mut iter = TtIter::new(tt);
100 let mut res = Vec::new();
102 while iter.peek_n(0).is_some() {
103 let expanded = iter.expect_fragment(ParserEntryPoint::Expr);
104 if expanded.err.is_some() {
108 res.push(match expanded.value {
110 Some(tt @ tt::TokenTree::Leaf(_)) => {
111 tt::Subtree { delimiter: None, token_trees: vec![tt] }
113 Some(tt::TokenTree::Subtree(tt)) => tt,
116 let mut fork = iter.clone();
117 if fork.expect_char(sep).is_err() {
123 if iter.peek_n(0).is_some() {
124 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
130 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
132 subtree: tt::Subtree,
134 open_range: TextRange,
137 let entry = StackEntry {
138 subtree: tt::Subtree { delimiter: None, ..Default::default() },
139 // never used (delimiter is `None`)
141 open_range: TextRange::empty(TextSize::of('.')),
143 let mut stack = vec![entry];
146 let entry = stack.last_mut().unwrap();
147 let result = &mut entry.subtree.token_trees;
148 let (token, range) = match conv.bump() {
153 let k: SyntaxKind = token.kind();
155 if let Some(tokens) = conv.convert_doc_comment(&token) {
156 result.extend(tokens);
161 result.push(if k.is_punct() && k != UNDERSCORE {
162 assert_eq!(range.len(), TextSize::of('.'));
164 if let Some(delim) = entry.subtree.delimiter {
165 let expected = match delim.kind {
166 tt::DelimiterKind::Parenthesis => T![')'],
167 tt::DelimiterKind::Brace => T!['}'],
168 tt::DelimiterKind::Bracket => T![']'],
172 let entry = stack.pop().unwrap();
173 conv.id_alloc().close_delim(entry.idx, Some(range));
174 stack.last_mut().unwrap().subtree.token_trees.push(entry.subtree.into());
179 let delim = match k {
180 T!['('] => Some(tt::DelimiterKind::Parenthesis),
181 T!['{'] => Some(tt::DelimiterKind::Brace),
182 T!['['] => Some(tt::DelimiterKind::Bracket),
186 if let Some(kind) = delim {
187 let mut subtree = tt::Subtree::default();
188 let (id, idx) = conv.id_alloc().open_delim(range);
189 subtree.delimiter = Some(tt::Delimiter { id, kind });
190 stack.push(StackEntry { subtree, idx, open_range: range });
193 let spacing = match conv.peek() {
195 if next.kind().is_trivia()
196 || next.kind() == T!['[']
197 || next.kind() == T!['{']
198 || next.kind() == T!['('] =>
202 Some(next) if next.kind().is_punct() && next.kind() != UNDERSCORE => {
205 _ => tt::Spacing::Alone,
207 let char = match token.to_char() {
210 panic!("Token from lexer must be single char: token = {:#?}", token);
213 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
216 macro_rules! make_leaf {
218 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text() }.into()
221 let leaf: tt::Leaf = match k {
222 T![true] | T![false] => make_leaf!(Ident),
223 IDENT => make_leaf!(Ident),
224 UNDERSCORE => make_leaf!(Ident),
225 k if k.is_keyword() => make_leaf!(Ident),
226 k if k.is_literal() => make_leaf!(Literal),
228 let char_unit = TextSize::of('\'');
229 let r = TextRange::at(range.start(), char_unit);
230 let apostrophe = tt::Leaf::from(tt::Punct {
232 spacing: tt::Spacing::Joint,
233 id: conv.id_alloc().alloc(r),
235 result.push(apostrophe.into());
237 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
238 let ident = tt::Leaf::from(tt::Ident {
239 text: SmolStr::new(&token.to_text()[1..]),
240 id: conv.id_alloc().alloc(r),
242 result.push(ident.into());
252 // If we get here, we've consumed all input tokens.
253 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
254 // Merge them so we're left with one.
255 while stack.len() > 1 {
256 let entry = stack.pop().unwrap();
257 let parent = stack.last_mut().unwrap();
259 conv.id_alloc().close_delim(entry.idx, None);
260 let leaf: tt::Leaf = tt::Punct {
261 id: conv.id_alloc().alloc(entry.open_range),
262 char: match entry.subtree.delimiter.unwrap().kind {
263 tt::DelimiterKind::Parenthesis => '(',
264 tt::DelimiterKind::Brace => '{',
265 tt::DelimiterKind::Bracket => '[',
267 spacing: tt::Spacing::Alone,
270 parent.subtree.token_trees.push(leaf.into());
271 parent.subtree.token_trees.extend(entry.subtree.token_trees);
274 let subtree = stack.pop().unwrap().subtree;
275 if subtree.token_trees.len() == 1 {
276 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
277 return first.clone();
283 /// Returns the textual content of a doc comment block as a quoted string
284 /// That is, strips leading `///` (or `/**`, etc)
285 /// and strips the ending `*/`
286 /// And then quote the string, which is needed to convert to `tt::Literal`
287 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
288 let prefix_len = comment.prefix().len();
289 let mut text = &comment.text()[prefix_len..];
291 // Remove ending "*/"
292 if comment.kind().shape == ast::CommentShape::Block {
293 text = &text[0..text.len() - 2];
297 // Note that `tt::Literal` expect an escaped string
298 let text = format!("\"{}\"", text.escape_debug());
302 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
303 let comment = ast::Comment::cast(token.clone())?;
304 let doc = comment.kind().doc?;
306 // Make `doc="\" Comments\""
307 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
310 let mut token_trees = vec![mk_punct('#')];
311 if let ast::CommentPlacement::Inner = doc {
312 token_trees.push(mk_punct('!'));
314 token_trees.push(tt::TokenTree::from(tt::Subtree {
315 delimiter: Some(tt::Delimiter {
316 kind: tt::DelimiterKind::Bracket,
317 id: tt::TokenId::unspecified(),
319 token_trees: meta_tkns,
322 return Some(token_trees);
325 fn mk_ident(s: &str) -> tt::TokenTree {
326 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
328 id: tt::TokenId::unspecified(),
332 fn mk_punct(c: char) -> tt::TokenTree {
333 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
335 spacing: tt::Spacing::Alone,
336 id: tt::TokenId::unspecified(),
340 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
341 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
343 tt::TokenTree::from(tt::Leaf::from(lit))
347 struct TokenIdAlloc {
349 global_offset: TextSize,
354 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
355 let relative_range = absolute_range - self.global_offset;
356 let token_id = tt::TokenId(self.next_id);
358 self.map.insert(token_id, relative_range);
362 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
363 let token_id = tt::TokenId(self.next_id);
365 let idx = self.map.insert_delim(
367 open_abs_range - self.global_offset,
368 open_abs_range - self.global_offset,
373 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
374 match close_abs_range {
376 self.map.remove_delim(idx);
379 self.map.update_close_delim(idx, close - self.global_offset);
385 /// A Raw Token (straightly from lexer) convertor
386 struct RawConvertor<'a> {
389 id_alloc: TokenIdAlloc,
390 inner: std::slice::Iter<'a, RawToken>,
393 trait SrcToken: std::fmt::Debug {
394 fn kind(&self) -> SyntaxKind;
396 fn to_char(&self) -> Option<char>;
398 fn to_text(&self) -> SmolStr;
401 trait TokenConvertor {
402 type Token: SrcToken;
404 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
406 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
408 fn peek(&self) -> Option<Self::Token>;
410 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
413 impl<'a> SrcToken for (&'a RawToken, &'a str) {
414 fn kind(&self) -> SyntaxKind {
418 fn to_char(&self) -> Option<char> {
419 self.1.chars().next()
422 fn to_text(&self) -> SmolStr {
427 impl RawConvertor<'_> {}
429 impl<'a> TokenConvertor for RawConvertor<'a> {
430 type Token = (&'a RawToken, &'a str);
432 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
433 convert_doc_comment(&doc_comment(token.1))
436 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
437 let token = self.inner.next()?;
438 let range = TextRange::at(self.offset, token.len);
439 self.offset += token.len;
441 Some(((token, &self.text[range]), range))
444 fn peek(&self) -> Option<Self::Token> {
445 let token = self.inner.as_slice().get(0);
448 let range = TextRange::at(self.offset, it.len);
449 (it, &self.text[range])
453 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
459 id_alloc: TokenIdAlloc,
460 current: Option<SyntaxToken>,
461 censor: Option<TextRange>,
463 punct_offset: Option<(SyntaxToken, TextSize)>,
467 fn new(node: &SyntaxNode, global_offset: TextSize, censor: Option<TextRange>) -> Convertor {
468 let first = node.first_token();
469 let current = match censor {
470 Some(censor) => iter::successors(first, |token| token.next_token())
471 .find(|token| !censor.contains_range(token.text_range())),
475 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
477 range: node.text_range(),
486 Ordinary(SyntaxToken),
487 Punch(SyntaxToken, TextSize),
491 fn token(&self) -> &SyntaxToken {
493 SynToken::Ordinary(it) => it,
494 SynToken::Punch(it, _) => it,
499 impl SrcToken for SynToken {
500 fn kind(&self) -> SyntaxKind {
503 fn to_char(&self) -> Option<char> {
505 SynToken::Ordinary(_) => None,
506 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
509 fn to_text(&self) -> SmolStr {
510 self.token().text().into()
514 impl TokenConvertor for Convertor {
515 type Token = SynToken;
516 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
517 convert_doc_comment(token.token())
520 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
521 if let Some((punct, offset)) = self.punct_offset.clone() {
522 if usize::from(offset) + 1 < punct.text().len() {
523 let offset = offset + TextSize::of('.');
524 let range = punct.text_range();
525 self.punct_offset = Some((punct.clone(), offset));
526 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
527 return Some((SynToken::Punch(punct, offset), range));
531 let curr = self.current.clone()?;
532 if !&self.range.contains_range(curr.text_range()) {
535 self.current = match self.censor {
536 Some(censor) => iter::successors(curr.next_token(), |token| token.next_token())
537 .find(|token| !censor.contains_range(token.text_range())),
538 None => curr.next_token(),
540 let token = if curr.kind().is_punct() {
541 let range = curr.text_range();
542 let range = TextRange::at(range.start(), TextSize::of('.'));
543 self.punct_offset = Some((curr.clone(), 0.into()));
544 (SynToken::Punch(curr, 0.into()), range)
546 self.punct_offset = None;
547 let range = curr.text_range();
548 (SynToken::Ordinary(curr), range)
554 fn peek(&self) -> Option<Self::Token> {
555 if let Some((punct, mut offset)) = self.punct_offset.clone() {
556 offset += TextSize::of('.');
557 if usize::from(offset) < punct.text().len() {
558 return Some(SynToken::Punch(punct, offset));
562 let curr = self.current.clone()?;
563 if !self.range.contains_range(curr.text_range()) {
567 let token = if curr.kind().is_punct() {
568 SynToken::Punch(curr, 0.into())
570 SynToken::Ordinary(curr)
575 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
580 struct TtTreeSink<'a> {
583 open_delims: FxHashMap<tt::TokenId, TextSize>,
585 inner: SyntaxTreeBuilder,
589 // Use for detect ill-form tree which is not single root
590 roots: smallvec::SmallVec<[usize; 1]>,
593 impl<'a> TtTreeSink<'a> {
594 fn new(cursor: Cursor<'a>) -> Self {
598 open_delims: FxHashMap::default(),
600 inner: SyntaxTreeBuilder::default(),
601 roots: smallvec::SmallVec::new(),
602 token_map: TokenMap::default(),
606 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
607 self.token_map.shrink_to_fit();
608 (self.inner.finish(), self.token_map)
612 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> &'static str {
613 let texts = match d {
614 Some(tt::DelimiterKind::Parenthesis) => "()",
615 Some(tt::DelimiterKind::Brace) => "{}",
616 Some(tt::DelimiterKind::Bracket) => "[]",
620 let idx = closing as usize;
621 &texts[idx..texts.len() - (1 - idx)]
624 impl<'a> TreeSink for TtTreeSink<'a> {
625 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
626 if kind == L_DOLLAR || kind == R_DOLLAR {
627 self.cursor = self.cursor.bump_subtree();
630 if kind == LIFETIME_IDENT {
634 let mut last = self.cursor;
635 for _ in 0..n_tokens {
636 let tmp_str: SmolStr;
637 if self.cursor.eof() {
641 let text: &str = match self.cursor.token_tree() {
642 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
643 // Mark the range if needed
644 let (text, id) = match leaf {
645 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
646 tt::Leaf::Punct(punct) => {
647 assert!(punct.char.is_ascii());
648 let char = &(punct.char as u8);
649 tmp_str = SmolStr::new_inline(
650 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
654 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
656 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
657 self.token_map.insert(id, range);
658 self.cursor = self.cursor.bump();
661 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
662 self.cursor = self.cursor.subtree().unwrap();
663 if let Some(id) = subtree.delimiter.map(|it| it.id) {
664 self.open_delims.insert(id, self.text_pos);
666 delim_to_str(subtree.delimiter_kind(), false)
669 if let Some(parent) = self.cursor.end() {
670 self.cursor = self.cursor.bump();
671 if let Some(id) = parent.delimiter.map(|it| it.id) {
672 if let Some(open_delim) = self.open_delims.get(&id) {
673 let open_range = TextRange::at(*open_delim, TextSize::of('('));
674 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
675 self.token_map.insert_delim(id, open_range, close_range);
678 delim_to_str(parent.delimiter_kind(), true)
685 self.text_pos += TextSize::of(text);
688 self.inner.token(kind, self.buf.as_str());
690 // Add whitespace between adjoint puncts
691 let next = last.bump();
693 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
694 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
695 ) = (last.token_tree(), next.token_tree())
697 // Note: We always assume the semi-colon would be the last token in
698 // other parts of RA such that we don't add whitespace here.
699 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
700 self.inner.token(WHITESPACE, " ");
701 self.text_pos += TextSize::of(' ');
706 fn start_node(&mut self, kind: SyntaxKind) {
707 self.inner.start_node(kind);
709 match self.roots.last_mut() {
710 None | Some(0) => self.roots.push(1),
711 Some(ref mut n) => **n += 1,
715 fn finish_node(&mut self) {
716 self.inner.finish_node();
717 *self.roots.last_mut().unwrap() -= 1;
720 fn error(&mut self, error: ParseError) {
721 self.inner.error(error, self.text_pos)
728 use crate::tests::parse_macro;
729 use parser::TokenSource;
731 ast::{make, AstNode},
734 use test_utils::assert_eq_text;
737 fn convert_tt_token_source() {
738 let expansion = parse_macro(
740 macro_rules! literals {
752 .expand_tt("literals!(foo);");
753 let tts = &[expansion.into()];
754 let buffer = tt::buffer::TokenBuffer::from_tokens(tts);
755 let mut tt_src = SubtreeTokenSource::new(&buffer);
756 let mut tokens = vec![];
757 while tt_src.current().kind != EOF {
758 tokens.push((tt_src.current().kind, tt_src.text()));
763 // [let] [a] [=] ['c'] [;]
764 assert_eq!(tokens[2 + 3].1, "'c'");
765 assert_eq!(tokens[2 + 3].0, CHAR);
766 // [let] [c] [=] [1000] [;]
767 assert_eq!(tokens[2 + 5 + 3].1, "1000");
768 assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
769 // [let] [f] [=] [12E+99_f64] [;]
770 assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
771 assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
773 // [let] [s] [=] ["rust1"] [;]
774 assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
775 assert_eq!(tokens[2 + 15 + 3].0, STRING);
779 fn stmts_token_trees_to_expr_is_err() {
780 let expansion = parse_macro(
792 .expand_tt("stmts!();");
793 assert!(token_tree_to_syntax_node(&expansion, ParserEntryPoint::Expr).is_err());
797 fn test_token_tree_last_child_is_white_space() {
798 let source_file = ast::SourceFile::parse("f!{}").ok().unwrap();
799 let macro_call = source_file.syntax().descendants().find_map(ast::MacroCall::cast).unwrap();
800 let token_tree = macro_call.token_tree().unwrap();
802 // Token Tree now is :
808 let token_tree = token_tree.clone_for_update();
809 ted::append_child(token_tree.syntax(), make::tokens::single_space());
810 let token_tree = token_tree.clone_subtree();
811 // Token Tree now is :
817 let tt = syntax_node_to_token_tree(token_tree.syntax()).0;
818 assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
822 fn test_token_tree_multi_char_punct() {
823 let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
824 let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
825 let tt = syntax_node_to_token_tree(struct_def.syntax()).0;
826 token_tree_to_syntax_node(&tt, ParserEntryPoint::Item).unwrap();
830 fn test_missing_closing_delim() {
831 let source_file = ast::SourceFile::parse("m!(x").tree();
832 let node = source_file.syntax().descendants().find_map(ast::TokenTree::cast).unwrap();
833 let tt = syntax_node_to_token_tree(node.syntax()).0.to_string();
834 assert_eq_text!(&*tt, "( x");