1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
5 use parser::{FragmentKind, 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};
15 use crate::{subtree_source::SubtreeTokenSource, tt_iter::TtIter};
16 use crate::{ExpandError, TokenMap};
18 /// Convert the syntax node to a `TokenTree` with the censored nodes excluded (what macro
20 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
21 syntax_node_to_token_tree_censored(node, None)
24 /// Convert the syntax node to a `TokenTree` with the censored nodes excluded (what macro
26 pub fn syntax_node_to_token_tree_censored(
28 censor: Option<TextRange>,
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 fragment_kind: FragmentKind,
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_fragment(&mut token_source, &mut tree_sink, fragment_kind);
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 seperate 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(FragmentKind::Expr);
103 if expanded.err.is_some() {
107 res.push(match expanded.value {
109 Some(tt @ tt::TokenTree::Leaf(_)) => {
110 tt::Subtree { delimiter: None, token_trees: vec![tt] }
112 Some(tt::TokenTree::Subtree(tt)) => tt,
115 let mut fork = iter.clone();
116 if fork.expect_char(sep).is_err() {
122 if iter.peek_n(0).is_some() {
123 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
129 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
131 subtree: tt::Subtree,
133 open_range: TextRange,
136 let entry = StackEntry {
137 subtree: tt::Subtree { delimiter: None, ..Default::default() },
138 // never used (delimiter is `None`)
140 open_range: TextRange::empty(TextSize::of('.')),
142 let mut stack = vec![entry];
145 let entry = stack.last_mut().unwrap();
146 let result = &mut entry.subtree.token_trees;
147 let (token, range) = match conv.bump() {
152 let k: SyntaxKind = token.kind();
154 if let Some(tokens) = conv.convert_doc_comment(&token) {
155 result.extend(tokens);
160 result.push(if k.is_punct() && k != UNDERSCORE {
161 assert_eq!(range.len(), TextSize::of('.'));
163 if let Some(delim) = entry.subtree.delimiter {
164 let expected = match delim.kind {
165 tt::DelimiterKind::Parenthesis => T![')'],
166 tt::DelimiterKind::Brace => T!['}'],
167 tt::DelimiterKind::Bracket => T![']'],
171 let entry = stack.pop().unwrap();
172 conv.id_alloc().close_delim(entry.idx, Some(range));
173 stack.last_mut().unwrap().subtree.token_trees.push(entry.subtree.into());
178 let delim = match k {
179 T!['('] => Some(tt::DelimiterKind::Parenthesis),
180 T!['{'] => Some(tt::DelimiterKind::Brace),
181 T!['['] => Some(tt::DelimiterKind::Bracket),
185 if let Some(kind) = delim {
186 let mut subtree = tt::Subtree::default();
187 let (id, idx) = conv.id_alloc().open_delim(range);
188 subtree.delimiter = Some(tt::Delimiter { id, kind });
189 stack.push(StackEntry { subtree, idx, open_range: range });
192 let spacing = match conv.peek() {
194 if next.kind().is_trivia()
195 || next.kind() == T!['[']
196 || next.kind() == T!['{']
197 || next.kind() == T!['('] =>
201 Some(next) if next.kind().is_punct() && next.kind() != UNDERSCORE => {
204 _ => tt::Spacing::Alone,
206 let char = match token.to_char() {
209 panic!("Token from lexer must be single char: token = {:#?}", token);
212 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
215 macro_rules! make_leaf {
217 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text() }.into()
220 let leaf: tt::Leaf = match k {
221 T![true] | T![false] => make_leaf!(Ident),
222 IDENT => make_leaf!(Ident),
223 UNDERSCORE => make_leaf!(Ident),
224 k if k.is_keyword() => make_leaf!(Ident),
225 k if k.is_literal() => make_leaf!(Literal),
227 let char_unit = TextSize::of('\'');
228 let r = TextRange::at(range.start(), char_unit);
229 let apostrophe = tt::Leaf::from(tt::Punct {
231 spacing: tt::Spacing::Joint,
232 id: conv.id_alloc().alloc(r),
234 result.push(apostrophe.into());
236 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
237 let ident = tt::Leaf::from(tt::Ident {
238 text: SmolStr::new(&token.to_text()[1..]),
239 id: conv.id_alloc().alloc(r),
241 result.push(ident.into());
251 // If we get here, we've consumed all input tokens.
252 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
253 // Merge them so we're left with one.
254 while stack.len() > 1 {
255 let entry = stack.pop().unwrap();
256 let parent = stack.last_mut().unwrap();
258 conv.id_alloc().close_delim(entry.idx, None);
259 let leaf: tt::Leaf = tt::Punct {
260 id: conv.id_alloc().alloc(entry.open_range),
261 char: match entry.subtree.delimiter.unwrap().kind {
262 tt::DelimiterKind::Parenthesis => '(',
263 tt::DelimiterKind::Brace => '{',
264 tt::DelimiterKind::Bracket => '[',
266 spacing: tt::Spacing::Alone,
269 parent.subtree.token_trees.push(leaf.into());
270 parent.subtree.token_trees.extend(entry.subtree.token_trees);
273 let subtree = stack.pop().unwrap().subtree;
274 if subtree.token_trees.len() == 1 {
275 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
276 return first.clone();
282 /// Returns the textual content of a doc comment block as a quoted string
283 /// That is, strips leading `///` (or `/**`, etc)
284 /// and strips the ending `*/`
285 /// And then quote the string, which is needed to convert to `tt::Literal`
286 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
287 let prefix_len = comment.prefix().len();
288 let mut text = &comment.text()[prefix_len..];
290 // Remove ending "*/"
291 if comment.kind().shape == ast::CommentShape::Block {
292 text = &text[0..text.len() - 2];
296 // Note that `tt::Literal` expect an escaped string
297 let text = format!("\"{}\"", text.escape_debug());
301 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
302 let comment = ast::Comment::cast(token.clone())?;
303 let doc = comment.kind().doc?;
305 // Make `doc="\" Comments\""
306 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
309 let mut token_trees = vec![mk_punct('#')];
310 if let ast::CommentPlacement::Inner = doc {
311 token_trees.push(mk_punct('!'));
313 token_trees.push(tt::TokenTree::from(tt::Subtree {
314 delimiter: Some(tt::Delimiter {
315 kind: tt::DelimiterKind::Bracket,
316 id: tt::TokenId::unspecified(),
318 token_trees: meta_tkns,
321 return Some(token_trees);
324 fn mk_ident(s: &str) -> tt::TokenTree {
325 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
327 id: tt::TokenId::unspecified(),
331 fn mk_punct(c: char) -> tt::TokenTree {
332 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
334 spacing: tt::Spacing::Alone,
335 id: tt::TokenId::unspecified(),
339 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
340 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
342 tt::TokenTree::from(tt::Leaf::from(lit))
346 struct TokenIdAlloc {
348 global_offset: TextSize,
353 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
354 let relative_range = absolute_range - self.global_offset;
355 let token_id = tt::TokenId(self.next_id);
357 self.map.insert(token_id, relative_range);
361 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
362 let token_id = tt::TokenId(self.next_id);
364 let idx = self.map.insert_delim(
366 open_abs_range - self.global_offset,
367 open_abs_range - self.global_offset,
372 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
373 match close_abs_range {
375 self.map.remove_delim(idx);
378 self.map.update_close_delim(idx, close - self.global_offset);
384 /// A Raw Token (straightly from lexer) convertor
385 struct RawConvertor<'a> {
388 id_alloc: TokenIdAlloc,
389 inner: std::slice::Iter<'a, RawToken>,
392 trait SrcToken: std::fmt::Debug {
393 fn kind(&self) -> SyntaxKind;
395 fn to_char(&self) -> Option<char>;
397 fn to_text(&self) -> SmolStr;
400 trait TokenConvertor {
401 type Token: SrcToken;
403 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
405 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
407 fn peek(&self) -> Option<Self::Token>;
409 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
412 impl<'a> SrcToken for (&'a RawToken, &'a str) {
413 fn kind(&self) -> SyntaxKind {
417 fn to_char(&self) -> Option<char> {
418 self.1.chars().next()
421 fn to_text(&self) -> SmolStr {
426 impl RawConvertor<'_> {}
428 impl<'a> TokenConvertor for RawConvertor<'a> {
429 type Token = (&'a RawToken, &'a str);
431 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
432 convert_doc_comment(&doc_comment(token.1))
435 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
436 let token = self.inner.next()?;
437 let range = TextRange::at(self.offset, token.len);
438 self.offset += token.len;
440 Some(((token, &self.text[range]), range))
443 fn peek(&self) -> Option<Self::Token> {
444 let token = self.inner.as_slice().get(0);
447 let range = TextRange::at(self.offset, it.len);
448 (it, &self.text[range])
452 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
458 id_alloc: TokenIdAlloc,
459 current: Option<SyntaxToken>,
460 censor: Option<TextRange>,
462 punct_offset: Option<(SyntaxToken, TextSize)>,
466 fn new(node: &SyntaxNode, global_offset: TextSize, censor: Option<TextRange>) -> Convertor {
467 let first = node.first_token();
468 let current = match censor {
469 Some(censor) => iter::successors(first, |token| token.next_token())
470 .find(|token| !censor.contains_range(token.text_range())),
474 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
476 range: node.text_range(),
485 Ordinary(SyntaxToken),
486 Punch(SyntaxToken, TextSize),
490 fn token(&self) -> &SyntaxToken {
492 SynToken::Ordinary(it) => it,
493 SynToken::Punch(it, _) => it,
498 impl SrcToken for SynToken {
499 fn kind(&self) -> SyntaxKind {
502 fn to_char(&self) -> Option<char> {
504 SynToken::Ordinary(_) => None,
505 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
508 fn to_text(&self) -> SmolStr {
509 self.token().text().into()
513 impl TokenConvertor for Convertor {
514 type Token = SynToken;
515 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
516 convert_doc_comment(token.token())
519 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
520 if let Some((punct, offset)) = self.punct_offset.clone() {
521 if usize::from(offset) + 1 < punct.text().len() {
522 let offset = offset + TextSize::of('.');
523 let range = punct.text_range();
524 self.punct_offset = Some((punct.clone(), offset));
525 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
526 return Some((SynToken::Punch(punct, offset), range));
530 let curr = self.current.clone()?;
531 if !&self.range.contains_range(curr.text_range()) {
534 self.current = match self.censor {
535 Some(censor) => iter::successors(curr.next_token(), |token| token.next_token())
536 .find(|token| !censor.contains_range(token.text_range())),
537 None => curr.next_token(),
539 let token = if curr.kind().is_punct() {
540 let range = curr.text_range();
541 let range = TextRange::at(range.start(), TextSize::of('.'));
542 self.punct_offset = Some((curr.clone(), 0.into()));
543 (SynToken::Punch(curr, 0.into()), range)
545 self.punct_offset = None;
546 let range = curr.text_range();
547 (SynToken::Ordinary(curr), range)
553 fn peek(&self) -> Option<Self::Token> {
554 if let Some((punct, mut offset)) = self.punct_offset.clone() {
555 offset += TextSize::of('.');
556 if usize::from(offset) < punct.text().len() {
557 return Some(SynToken::Punch(punct, offset));
561 let curr = self.current.clone()?;
562 if !self.range.contains_range(curr.text_range()) {
566 let token = if curr.kind().is_punct() {
567 SynToken::Punch(curr, 0.into())
569 SynToken::Ordinary(curr)
574 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
579 struct TtTreeSink<'a> {
582 open_delims: FxHashMap<tt::TokenId, TextSize>,
584 inner: SyntaxTreeBuilder,
588 // Use for detect ill-form tree which is not single root
589 roots: smallvec::SmallVec<[usize; 1]>,
592 impl<'a> TtTreeSink<'a> {
593 fn new(cursor: Cursor<'a>) -> Self {
597 open_delims: FxHashMap::default(),
599 inner: SyntaxTreeBuilder::default(),
600 roots: smallvec::SmallVec::new(),
601 token_map: TokenMap::default(),
605 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
606 self.token_map.shrink_to_fit();
607 (self.inner.finish(), self.token_map)
611 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> &'static str {
612 let texts = match d {
613 Some(tt::DelimiterKind::Parenthesis) => "()",
614 Some(tt::DelimiterKind::Brace) => "{}",
615 Some(tt::DelimiterKind::Bracket) => "[]",
619 let idx = closing as usize;
620 &texts[idx..texts.len() - (1 - idx)]
623 impl<'a> TreeSink for TtTreeSink<'a> {
624 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
625 if kind == L_DOLLAR || kind == R_DOLLAR {
626 self.cursor = self.cursor.bump_subtree();
629 if kind == LIFETIME_IDENT {
633 let mut last = self.cursor;
634 for _ in 0..n_tokens {
635 let tmp_str: SmolStr;
636 if self.cursor.eof() {
640 let text: &str = match self.cursor.token_tree() {
641 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
642 // Mark the range if needed
643 let (text, id) = match leaf {
644 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
645 tt::Leaf::Punct(punct) => {
646 assert!(punct.char.is_ascii());
647 let char = &(punct.char as u8);
648 tmp_str = SmolStr::new_inline(
649 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
653 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
655 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
656 self.token_map.insert(id, range);
657 self.cursor = self.cursor.bump();
660 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
661 self.cursor = self.cursor.subtree().unwrap();
662 if let Some(id) = subtree.delimiter.map(|it| it.id) {
663 self.open_delims.insert(id, self.text_pos);
665 delim_to_str(subtree.delimiter_kind(), false)
668 if let Some(parent) = self.cursor.end() {
669 self.cursor = self.cursor.bump();
670 if let Some(id) = parent.delimiter.map(|it| it.id) {
671 if let Some(open_delim) = self.open_delims.get(&id) {
672 let open_range = TextRange::at(*open_delim, TextSize::of('('));
673 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
674 self.token_map.insert_delim(id, open_range, close_range);
677 delim_to_str(parent.delimiter_kind(), true)
684 self.text_pos += TextSize::of(text);
687 self.inner.token(kind, self.buf.as_str());
689 // Add whitespace between adjoint puncts
690 let next = last.bump();
692 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
693 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
694 ) = (last.token_tree(), next.token_tree())
696 // Note: We always assume the semi-colon would be the last token in
697 // other parts of RA such that we don't add whitespace here.
698 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
699 self.inner.token(WHITESPACE, " ");
700 self.text_pos += TextSize::of(' ');
705 fn start_node(&mut self, kind: SyntaxKind) {
706 self.inner.start_node(kind);
708 match self.roots.last_mut() {
709 None | Some(0) => self.roots.push(1),
710 Some(ref mut n) => **n += 1,
714 fn finish_node(&mut self) {
715 self.inner.finish_node();
716 *self.roots.last_mut().unwrap() -= 1;
719 fn error(&mut self, error: ParseError) {
720 self.inner.error(error, self.text_pos)
727 use crate::tests::parse_macro;
728 use parser::TokenSource;
730 ast::{make, AstNode},
733 use test_utils::assert_eq_text;
736 fn convert_tt_token_source() {
737 let expansion = parse_macro(
739 macro_rules! literals {
751 .expand_tt("literals!(foo);");
752 let tts = &[expansion.into()];
753 let buffer = tt::buffer::TokenBuffer::from_tokens(tts);
754 let mut tt_src = SubtreeTokenSource::new(&buffer);
755 let mut tokens = vec![];
756 while tt_src.current().kind != EOF {
757 tokens.push((tt_src.current().kind, tt_src.text()));
762 // [let] [a] [=] ['c'] [;]
763 assert_eq!(tokens[2 + 3].1, "'c'");
764 assert_eq!(tokens[2 + 3].0, CHAR);
765 // [let] [c] [=] [1000] [;]
766 assert_eq!(tokens[2 + 5 + 3].1, "1000");
767 assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
768 // [let] [f] [=] [12E+99_f64] [;]
769 assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
770 assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
772 // [let] [s] [=] ["rust1"] [;]
773 assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
774 assert_eq!(tokens[2 + 15 + 3].0, STRING);
778 fn stmts_token_trees_to_expr_is_err() {
779 let expansion = parse_macro(
791 .expand_tt("stmts!();");
792 assert!(token_tree_to_syntax_node(&expansion, FragmentKind::Expr).is_err());
796 fn test_token_tree_last_child_is_white_space() {
797 let source_file = ast::SourceFile::parse("f!{}").ok().unwrap();
798 let macro_call = source_file.syntax().descendants().find_map(ast::MacroCall::cast).unwrap();
799 let token_tree = macro_call.token_tree().unwrap();
801 // Token Tree now is :
807 let token_tree = token_tree.clone_for_update();
808 ted::append_child(token_tree.syntax(), make::tokens::single_space());
809 let token_tree = token_tree.clone_subtree();
810 // Token Tree now is :
816 let tt = syntax_node_to_token_tree(token_tree.syntax()).0;
817 assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
821 fn test_token_tree_multi_char_punct() {
822 let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
823 let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
824 let tt = syntax_node_to_token_tree(struct_def.syntax()).0;
825 token_tree_to_syntax_node(&tt, FragmentKind::Item).unwrap();
829 fn test_missing_closing_delim() {
830 let source_file = ast::SourceFile::parse("m!(x").tree();
831 let node = source_file.syntax().descendants().find_map(ast::TokenTree::cast).unwrap();
832 let tt = syntax_node_to_token_tree(node.syntax()).0.to_string();
833 assert_eq_text!(&*tt, "( x");