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 separate 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);
105 res.push(match expanded.value {
107 Some(tt @ tt::TokenTree::Leaf(_)) => {
108 tt::Subtree { delimiter: None, token_trees: vec![tt] }
110 Some(tt::TokenTree::Subtree(tt)) => tt,
113 let mut fork = iter.clone();
114 if fork.expect_char(sep).is_err() {
120 if iter.peek_n(0).is_some() {
121 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
127 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
129 subtree: tt::Subtree,
131 open_range: TextRange,
134 let entry = StackEntry {
135 subtree: tt::Subtree { delimiter: None, ..Default::default() },
136 // never used (delimiter is `None`)
138 open_range: TextRange::empty(TextSize::of('.')),
140 let mut stack = vec![entry];
143 let entry = stack.last_mut().unwrap();
144 let result = &mut entry.subtree.token_trees;
145 let (token, range) = match conv.bump() {
150 let k: SyntaxKind = token.kind();
152 if let Some(tokens) = conv.convert_doc_comment(&token) {
153 result.extend(tokens);
158 result.push(if k.is_punct() && k != UNDERSCORE {
159 assert_eq!(range.len(), TextSize::of('.'));
161 if let Some(delim) = entry.subtree.delimiter {
162 let expected = match delim.kind {
163 tt::DelimiterKind::Parenthesis => T![')'],
164 tt::DelimiterKind::Brace => T!['}'],
165 tt::DelimiterKind::Bracket => T![']'],
169 let entry = stack.pop().unwrap();
170 conv.id_alloc().close_delim(entry.idx, Some(range));
171 stack.last_mut().unwrap().subtree.token_trees.push(entry.subtree.into());
176 let delim = match k {
177 T!['('] => Some(tt::DelimiterKind::Parenthesis),
178 T!['{'] => Some(tt::DelimiterKind::Brace),
179 T!['['] => Some(tt::DelimiterKind::Bracket),
183 if let Some(kind) = delim {
184 let mut subtree = tt::Subtree::default();
185 let (id, idx) = conv.id_alloc().open_delim(range);
186 subtree.delimiter = Some(tt::Delimiter { id, kind });
187 stack.push(StackEntry { subtree, idx, open_range: range });
190 let spacing = match conv.peek() {
192 if next.kind().is_trivia()
193 || next.kind() == T!['[']
194 || next.kind() == T!['{']
195 || next.kind() == T!['('] =>
199 Some(next) if next.kind().is_punct() && next.kind() != UNDERSCORE => {
202 _ => tt::Spacing::Alone,
204 let char = match token.to_char() {
207 panic!("Token from lexer must be single char: token = {:#?}", token);
210 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
213 macro_rules! make_leaf {
215 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text() }.into()
218 let leaf: tt::Leaf = match k {
219 T![true] | T![false] => make_leaf!(Ident),
220 IDENT => make_leaf!(Ident),
221 UNDERSCORE => make_leaf!(Ident),
222 k if k.is_keyword() => make_leaf!(Ident),
223 k if k.is_literal() => make_leaf!(Literal),
225 let char_unit = TextSize::of('\'');
226 let r = TextRange::at(range.start(), char_unit);
227 let apostrophe = tt::Leaf::from(tt::Punct {
229 spacing: tt::Spacing::Joint,
230 id: conv.id_alloc().alloc(r),
232 result.push(apostrophe.into());
234 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
235 let ident = tt::Leaf::from(tt::Ident {
236 text: SmolStr::new(&token.to_text()[1..]),
237 id: conv.id_alloc().alloc(r),
239 result.push(ident.into());
249 // If we get here, we've consumed all input tokens.
250 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
251 // Merge them so we're left with one.
252 while stack.len() > 1 {
253 let entry = stack.pop().unwrap();
254 let parent = stack.last_mut().unwrap();
256 conv.id_alloc().close_delim(entry.idx, None);
257 let leaf: tt::Leaf = tt::Punct {
258 id: conv.id_alloc().alloc(entry.open_range),
259 char: match entry.subtree.delimiter.unwrap().kind {
260 tt::DelimiterKind::Parenthesis => '(',
261 tt::DelimiterKind::Brace => '{',
262 tt::DelimiterKind::Bracket => '[',
264 spacing: tt::Spacing::Alone,
267 parent.subtree.token_trees.push(leaf.into());
268 parent.subtree.token_trees.extend(entry.subtree.token_trees);
271 let subtree = stack.pop().unwrap().subtree;
272 if subtree.token_trees.len() == 1 {
273 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
274 return first.clone();
280 /// Returns the textual content of a doc comment block as a quoted string
281 /// That is, strips leading `///` (or `/**`, etc)
282 /// and strips the ending `*/`
283 /// And then quote the string, which is needed to convert to `tt::Literal`
284 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
285 let prefix_len = comment.prefix().len();
286 let mut text = &comment.text()[prefix_len..];
288 // Remove ending "*/"
289 if comment.kind().shape == ast::CommentShape::Block {
290 text = &text[0..text.len() - 2];
294 // Note that `tt::Literal` expect an escaped string
295 let text = format!("\"{}\"", text.escape_debug());
299 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
300 let comment = ast::Comment::cast(token.clone())?;
301 let doc = comment.kind().doc?;
303 // Make `doc="\" Comments\""
304 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
307 let mut token_trees = vec![mk_punct('#')];
308 if let ast::CommentPlacement::Inner = doc {
309 token_trees.push(mk_punct('!'));
311 token_trees.push(tt::TokenTree::from(tt::Subtree {
312 delimiter: Some(tt::Delimiter {
313 kind: tt::DelimiterKind::Bracket,
314 id: tt::TokenId::unspecified(),
316 token_trees: meta_tkns,
319 return Some(token_trees);
322 fn mk_ident(s: &str) -> tt::TokenTree {
323 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
325 id: tt::TokenId::unspecified(),
329 fn mk_punct(c: char) -> tt::TokenTree {
330 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
332 spacing: tt::Spacing::Alone,
333 id: tt::TokenId::unspecified(),
337 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
338 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
340 tt::TokenTree::from(tt::Leaf::from(lit))
344 struct TokenIdAlloc {
346 global_offset: TextSize,
351 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
352 let relative_range = absolute_range - self.global_offset;
353 let token_id = tt::TokenId(self.next_id);
355 self.map.insert(token_id, relative_range);
359 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
360 let token_id = tt::TokenId(self.next_id);
362 let idx = self.map.insert_delim(
364 open_abs_range - self.global_offset,
365 open_abs_range - self.global_offset,
370 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
371 match close_abs_range {
373 self.map.remove_delim(idx);
376 self.map.update_close_delim(idx, close - self.global_offset);
382 /// A Raw Token (straightly from lexer) convertor
383 struct RawConvertor<'a> {
386 id_alloc: TokenIdAlloc,
387 inner: std::slice::Iter<'a, RawToken>,
390 trait SrcToken: std::fmt::Debug {
391 fn kind(&self) -> SyntaxKind;
393 fn to_char(&self) -> Option<char>;
395 fn to_text(&self) -> SmolStr;
398 trait TokenConvertor {
399 type Token: SrcToken;
401 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
403 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
405 fn peek(&self) -> Option<Self::Token>;
407 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
410 impl<'a> SrcToken for (&'a RawToken, &'a str) {
411 fn kind(&self) -> SyntaxKind {
415 fn to_char(&self) -> Option<char> {
416 self.1.chars().next()
419 fn to_text(&self) -> SmolStr {
424 impl RawConvertor<'_> {}
426 impl<'a> TokenConvertor for RawConvertor<'a> {
427 type Token = (&'a RawToken, &'a str);
429 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
430 convert_doc_comment(&doc_comment(token.1))
433 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
434 let token = self.inner.next()?;
435 let range = TextRange::at(self.offset, token.len);
436 self.offset += token.len;
438 Some(((token, &self.text[range]), range))
441 fn peek(&self) -> Option<Self::Token> {
442 let token = self.inner.as_slice().get(0);
445 let range = TextRange::at(self.offset, it.len);
446 (it, &self.text[range])
450 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
456 id_alloc: TokenIdAlloc,
457 current: Option<SyntaxToken>,
458 censor: Option<TextRange>,
460 punct_offset: Option<(SyntaxToken, TextSize)>,
464 fn new(node: &SyntaxNode, global_offset: TextSize, censor: Option<TextRange>) -> Convertor {
465 let first = node.first_token();
466 let current = match censor {
467 Some(censor) => iter::successors(first, |token| token.next_token())
468 .find(|token| !censor.contains_range(token.text_range())),
472 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
474 range: node.text_range(),
483 Ordinary(SyntaxToken),
484 Punch(SyntaxToken, TextSize),
488 fn token(&self) -> &SyntaxToken {
490 SynToken::Ordinary(it) => it,
491 SynToken::Punch(it, _) => it,
496 impl SrcToken for SynToken {
497 fn kind(&self) -> SyntaxKind {
500 fn to_char(&self) -> Option<char> {
502 SynToken::Ordinary(_) => None,
503 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
506 fn to_text(&self) -> SmolStr {
507 self.token().text().into()
511 impl TokenConvertor for Convertor {
512 type Token = SynToken;
513 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
514 convert_doc_comment(token.token())
517 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
518 if let Some((punct, offset)) = self.punct_offset.clone() {
519 if usize::from(offset) + 1 < punct.text().len() {
520 let offset = offset + TextSize::of('.');
521 let range = punct.text_range();
522 self.punct_offset = Some((punct.clone(), offset));
523 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
524 return Some((SynToken::Punch(punct, offset), range));
528 let curr = self.current.clone()?;
529 if !&self.range.contains_range(curr.text_range()) {
532 self.current = match self.censor {
533 Some(censor) => iter::successors(curr.next_token(), |token| token.next_token())
534 .find(|token| !censor.contains_range(token.text_range())),
535 None => curr.next_token(),
537 let token = if curr.kind().is_punct() {
538 let range = curr.text_range();
539 let range = TextRange::at(range.start(), TextSize::of('.'));
540 self.punct_offset = Some((curr.clone(), 0.into()));
541 (SynToken::Punch(curr, 0.into()), range)
543 self.punct_offset = None;
544 let range = curr.text_range();
545 (SynToken::Ordinary(curr), range)
551 fn peek(&self) -> Option<Self::Token> {
552 if let Some((punct, mut offset)) = self.punct_offset.clone() {
553 offset += TextSize::of('.');
554 if usize::from(offset) < punct.text().len() {
555 return Some(SynToken::Punch(punct, offset));
559 let curr = self.current.clone()?;
560 if !self.range.contains_range(curr.text_range()) {
564 let token = if curr.kind().is_punct() {
565 SynToken::Punch(curr, 0.into())
567 SynToken::Ordinary(curr)
572 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
577 struct TtTreeSink<'a> {
580 open_delims: FxHashMap<tt::TokenId, TextSize>,
582 inner: SyntaxTreeBuilder,
586 // Use for detect ill-form tree which is not single root
587 roots: smallvec::SmallVec<[usize; 1]>,
590 impl<'a> TtTreeSink<'a> {
591 fn new(cursor: Cursor<'a>) -> Self {
595 open_delims: FxHashMap::default(),
597 inner: SyntaxTreeBuilder::default(),
598 roots: smallvec::SmallVec::new(),
599 token_map: TokenMap::default(),
603 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
604 self.token_map.shrink_to_fit();
605 (self.inner.finish(), self.token_map)
609 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> &'static str {
610 let texts = match d {
611 Some(tt::DelimiterKind::Parenthesis) => "()",
612 Some(tt::DelimiterKind::Brace) => "{}",
613 Some(tt::DelimiterKind::Bracket) => "[]",
617 let idx = closing as usize;
618 &texts[idx..texts.len() - (1 - idx)]
621 impl<'a> TreeSink for TtTreeSink<'a> {
622 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
623 if kind == L_DOLLAR || kind == R_DOLLAR {
624 self.cursor = self.cursor.bump_subtree();
627 if kind == LIFETIME_IDENT {
631 let mut last = self.cursor;
632 for _ in 0..n_tokens {
633 let tmp_str: SmolStr;
634 if self.cursor.eof() {
638 let text: &str = match self.cursor.token_tree() {
639 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
640 // Mark the range if needed
641 let (text, id) = match leaf {
642 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
643 tt::Leaf::Punct(punct) => {
644 assert!(punct.char.is_ascii());
645 let char = &(punct.char as u8);
646 tmp_str = SmolStr::new_inline(
647 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
651 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
653 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
654 self.token_map.insert(id, range);
655 self.cursor = self.cursor.bump();
658 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
659 self.cursor = self.cursor.subtree().unwrap();
660 if let Some(id) = subtree.delimiter.map(|it| it.id) {
661 self.open_delims.insert(id, self.text_pos);
663 delim_to_str(subtree.delimiter_kind(), false)
666 if let Some(parent) = self.cursor.end() {
667 self.cursor = self.cursor.bump();
668 if let Some(id) = parent.delimiter.map(|it| it.id) {
669 if let Some(open_delim) = self.open_delims.get(&id) {
670 let open_range = TextRange::at(*open_delim, TextSize::of('('));
671 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
672 self.token_map.insert_delim(id, open_range, close_range);
675 delim_to_str(parent.delimiter_kind(), true)
682 self.text_pos += TextSize::of(text);
685 self.inner.token(kind, self.buf.as_str());
687 // Add whitespace between adjoint puncts
688 let next = last.bump();
690 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
691 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
692 ) = (last.token_tree(), next.token_tree())
694 // Note: We always assume the semi-colon would be the last token in
695 // other parts of RA such that we don't add whitespace here.
696 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
697 self.inner.token(WHITESPACE, " ");
698 self.text_pos += TextSize::of(' ');
703 fn start_node(&mut self, kind: SyntaxKind) {
704 self.inner.start_node(kind);
706 match self.roots.last_mut() {
707 None | Some(0) => self.roots.push(1),
708 Some(ref mut n) => **n += 1,
712 fn finish_node(&mut self) {
713 self.inner.finish_node();
714 *self.roots.last_mut().unwrap() -= 1;
717 fn error(&mut self, error: ParseError) {
718 self.inner.error(error, self.text_pos)
725 use crate::tests::parse_macro;
726 use parser::TokenSource;
728 ast::{make, AstNode},
731 use test_utils::assert_eq_text;
734 fn convert_tt_token_source() {
735 let expansion = parse_macro(
737 macro_rules! literals {
749 .expand_tt("literals!(foo);");
750 let tts = &[expansion.into()];
751 let buffer = tt::buffer::TokenBuffer::from_tokens(tts);
752 let mut tt_src = SubtreeTokenSource::new(&buffer);
753 let mut tokens = vec![];
754 while tt_src.current().kind != EOF {
755 tokens.push((tt_src.current().kind, tt_src.text()));
760 // [let] [a] [=] ['c'] [;]
761 assert_eq!(tokens[2 + 3].1, "'c'");
762 assert_eq!(tokens[2 + 3].0, CHAR);
763 // [let] [c] [=] [1000] [;]
764 assert_eq!(tokens[2 + 5 + 3].1, "1000");
765 assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
766 // [let] [f] [=] [12E+99_f64] [;]
767 assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
768 assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
770 // [let] [s] [=] ["rust1"] [;]
771 assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
772 assert_eq!(tokens[2 + 15 + 3].0, STRING);
776 fn stmts_token_trees_to_expr_is_err() {
777 let expansion = parse_macro(
789 .expand_tt("stmts!();");
790 assert!(token_tree_to_syntax_node(&expansion, ParserEntryPoint::Expr).is_err());
794 fn test_token_tree_last_child_is_white_space() {
795 let source_file = ast::SourceFile::parse("f!{}").ok().unwrap();
796 let macro_call = source_file.syntax().descendants().find_map(ast::MacroCall::cast).unwrap();
797 let token_tree = macro_call.token_tree().unwrap();
799 // Token Tree now is :
805 let token_tree = token_tree.clone_for_update();
806 ted::append_child(token_tree.syntax(), make::tokens::single_space());
807 let token_tree = token_tree.clone_subtree();
808 // Token Tree now is :
814 let tt = syntax_node_to_token_tree(token_tree.syntax()).0;
815 assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
819 fn test_token_tree_multi_char_punct() {
820 let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
821 let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
822 let tt = syntax_node_to_token_tree(struct_def.syntax()).0;
823 token_tree_to_syntax_node(&tt, ParserEntryPoint::Item).unwrap();
827 fn test_missing_closing_delim() {
828 let source_file = ast::SourceFile::parse("m!(x").tree();
829 let node = source_file.syntax().descendants().find_map(ast::TokenTree::cast).unwrap();
830 let tt = syntax_node_to_token_tree(node.syntax()).0.to_string();
831 assert_eq_text!(&*tt, "( x");