1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
3 use rustc_hash::{FxHashMap, FxHashSet};
4 use stdx::{always, non_empty_vec::NonEmptyVec};
6 ast::{self, make::tokens::doc_comment},
7 AstToken, Parse, PreorderWithTokens, SmolStr, SyntaxElement, SyntaxKind,
9 SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, WalkEvent, T,
11 use tt::buffer::{Cursor, TokenBuffer};
13 use crate::{to_parser_input::to_parser_input, tt_iter::TtIter, TokenMap};
15 /// Convert the syntax node to a `TokenTree` (what macro
17 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
18 syntax_node_to_token_tree_censored(node, Default::default(), 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 replace: FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
27 append: FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
28 ) -> (tt::Subtree, TokenMap) {
29 let global_offset = node.text_range().start();
30 let mut c = Convertor::new(node, global_offset, replace, append);
31 let subtree = convert_tokens(&mut c);
32 c.id_alloc.map.shrink_to_fit();
33 always!(c.replace.is_empty());
34 always!(c.append.is_empty());
35 (subtree, c.id_alloc.map)
38 pub type SyntheticToken = (SyntaxKind, SmolStr);
40 // The following items are what `rustc` macro can be parsed into :
41 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
42 // * Expr(P<ast::Expr>) -> token_tree_to_expr
43 // * Pat(P<ast::Pat>) -> token_tree_to_pat
44 // * Ty(P<ast::Ty>) -> token_tree_to_ty
45 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
46 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
48 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
49 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
50 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
52 pub fn token_tree_to_syntax_node(
54 entry_point: parser::TopEntryPoint,
55 ) -> (Parse<SyntaxNode>, TokenMap) {
56 let buffer = match tt {
57 tt::Subtree { delimiter: None, token_trees } => {
58 TokenBuffer::from_tokens(token_trees.as_slice())
60 _ => TokenBuffer::from_subtree(tt),
62 let parser_input = to_parser_input(&buffer);
63 let parser_output = entry_point.parse(&parser_input);
64 let mut tree_sink = TtTreeSink::new(buffer.begin());
65 for event in parser_output.iter() {
67 parser::Step::Token { kind, n_input_tokens: n_raw_tokens } => {
68 tree_sink.token(kind, n_raw_tokens)
70 parser::Step::Enter { kind } => tree_sink.start_node(kind),
71 parser::Step::Exit => tree_sink.finish_node(),
72 parser::Step::Error { msg } => tree_sink.error(msg.to_string()),
75 let (parse, range_map) = tree_sink.finish();
79 /// Convert a string to a `TokenTree`
80 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
81 let lexed = parser::LexedStr::new(text);
82 if lexed.errors().next().is_some() {
86 let mut conv = RawConvertor {
89 id_alloc: TokenIdAlloc {
90 map: Default::default(),
91 global_offset: TextSize::default(),
96 let subtree = convert_tokens(&mut conv);
97 Some((subtree, conv.id_alloc.map))
100 /// Split token tree with separate expr: $($e:expr)SEP*
101 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
102 if tt.token_trees.is_empty() {
106 let mut iter = TtIter::new(tt);
107 let mut res = Vec::new();
109 while iter.peek_n(0).is_some() {
110 let expanded = iter.expect_fragment(parser::PrefixEntryPoint::Expr);
112 res.push(match expanded.value {
114 Some(tt @ tt::TokenTree::Leaf(_)) => {
115 tt::Subtree { delimiter: None, token_trees: vec![tt] }
117 Some(tt::TokenTree::Subtree(tt)) => tt,
120 let mut fork = iter.clone();
121 if fork.expect_char(sep).is_err() {
127 if iter.peek_n(0).is_some() {
128 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
134 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
136 subtree: tt::Subtree,
138 open_range: TextRange,
141 let entry = StackEntry {
142 subtree: tt::Subtree { delimiter: None, ..Default::default() },
143 // never used (delimiter is `None`)
145 open_range: TextRange::empty(TextSize::of('.')),
147 let mut stack = NonEmptyVec::new(entry);
150 let StackEntry { subtree, .. } = stack.last_mut();
151 let result = &mut subtree.token_trees;
152 let (token, range) = match conv.bump() {
157 let kind = token.kind(&conv);
159 if let Some(tokens) = conv.convert_doc_comment(&token) {
160 // FIXME: There has to be a better way to do this
161 // Add the comments token id to the converted doc string
162 let id = conv.id_alloc().alloc(range);
163 result.extend(tokens.into_iter().map(|mut tt| {
164 if let tt::TokenTree::Subtree(sub) = &mut tt {
165 if let Some(tt::TokenTree::Leaf(tt::Leaf::Literal(lit))) =
166 sub.token_trees.get_mut(2)
176 let tt = if kind.is_punct() && kind != UNDERSCORE {
177 assert_eq!(range.len(), TextSize::of('.'));
179 if let Some(delim) = subtree.delimiter {
180 let expected = match delim.kind {
181 tt::DelimiterKind::Parenthesis => T![')'],
182 tt::DelimiterKind::Brace => T!['}'],
183 tt::DelimiterKind::Bracket => T![']'],
186 if kind == expected {
187 if let Some(entry) = stack.pop() {
188 conv.id_alloc().close_delim(entry.idx, Some(range));
189 stack.last_mut().subtree.token_trees.push(entry.subtree.into());
195 let delim = match kind {
196 T!['('] => Some(tt::DelimiterKind::Parenthesis),
197 T!['{'] => Some(tt::DelimiterKind::Brace),
198 T!['['] => Some(tt::DelimiterKind::Bracket),
202 if let Some(kind) = delim {
203 let mut subtree = tt::Subtree::default();
204 let (id, idx) = conv.id_alloc().open_delim(range);
205 subtree.delimiter = Some(tt::Delimiter { id, kind });
206 stack.push(StackEntry { subtree, idx, open_range: range });
210 let spacing = match conv.peek().map(|next| next.kind(&conv)) {
217 && kind != UNDERSCORE =>
221 _ => tt::Spacing::Alone,
223 let char = match token.to_char(&conv) {
226 panic!("Token from lexer must be single char: token = {:#?}", token);
229 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
231 macro_rules! make_leaf {
233 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text(conv) }.into()
236 let leaf: tt::Leaf = match kind {
237 T![true] | T![false] => make_leaf!(Ident),
238 IDENT => make_leaf!(Ident),
239 UNDERSCORE => make_leaf!(Ident),
240 k if k.is_keyword() => make_leaf!(Ident),
241 k if k.is_literal() => make_leaf!(Literal),
243 let char_unit = TextSize::of('\'');
244 let r = TextRange::at(range.start(), char_unit);
245 let apostrophe = tt::Leaf::from(tt::Punct {
247 spacing: tt::Spacing::Joint,
248 id: conv.id_alloc().alloc(r),
250 result.push(apostrophe.into());
252 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
253 let ident = tt::Leaf::from(tt::Ident {
254 text: SmolStr::new(&token.to_text(conv)[1..]),
255 id: conv.id_alloc().alloc(r),
257 result.push(ident.into());
268 // If we get here, we've consumed all input tokens.
269 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
270 // Merge them so we're left with one.
271 while let Some(entry) = stack.pop() {
272 let parent = stack.last_mut();
274 conv.id_alloc().close_delim(entry.idx, None);
275 let leaf: tt::Leaf = tt::Punct {
276 id: conv.id_alloc().alloc(entry.open_range),
277 char: match entry.subtree.delimiter.unwrap().kind {
278 tt::DelimiterKind::Parenthesis => '(',
279 tt::DelimiterKind::Brace => '{',
280 tt::DelimiterKind::Bracket => '[',
282 spacing: tt::Spacing::Alone,
285 parent.subtree.token_trees.push(leaf.into());
286 parent.subtree.token_trees.extend(entry.subtree.token_trees);
289 let subtree = stack.into_last().subtree;
290 if let [tt::TokenTree::Subtree(first)] = &*subtree.token_trees {
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::with_capacity(3);
326 token_trees.push(mk_punct('#'));
327 if let ast::CommentPlacement::Inner = doc {
328 token_trees.push(mk_punct('!'));
330 token_trees.push(tt::TokenTree::from(tt::Subtree {
331 delimiter: Some(tt::Delimiter {
332 kind: tt::DelimiterKind::Bracket,
333 id: tt::TokenId::unspecified(),
335 token_trees: meta_tkns,
338 return Some(token_trees);
341 fn mk_ident(s: &str) -> tt::TokenTree {
342 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
344 id: tt::TokenId::unspecified(),
348 fn mk_punct(c: char) -> tt::TokenTree {
349 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
351 spacing: tt::Spacing::Alone,
352 id: tt::TokenId::unspecified(),
356 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
357 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
359 tt::TokenTree::from(tt::Leaf::from(lit))
363 struct TokenIdAlloc {
365 global_offset: TextSize,
370 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
371 let relative_range = absolute_range - self.global_offset;
372 let token_id = tt::TokenId(self.next_id);
374 self.map.insert(token_id, relative_range);
378 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
379 let token_id = tt::TokenId(self.next_id);
381 let idx = self.map.insert_delim(
383 open_abs_range - self.global_offset,
384 open_abs_range - self.global_offset,
389 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
390 match close_abs_range {
392 self.map.remove_delim(idx);
395 self.map.update_close_delim(idx, close - self.global_offset);
401 /// A raw token (straight from lexer) convertor
402 struct RawConvertor<'a> {
403 lexed: parser::LexedStr<'a>,
405 id_alloc: TokenIdAlloc,
408 trait SrcToken<Ctx>: std::fmt::Debug {
409 fn kind(&self, ctx: &Ctx) -> SyntaxKind;
411 fn to_char(&self, ctx: &Ctx) -> Option<char>;
413 fn to_text(&self, ctx: &Ctx) -> SmolStr;
416 trait TokenConvertor: Sized {
417 type Token: SrcToken<Self>;
419 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
421 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
423 fn peek(&self) -> Option<Self::Token>;
425 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
428 impl<'a> SrcToken<RawConvertor<'a>> for usize {
429 fn kind(&self, ctx: &RawConvertor<'a>) -> SyntaxKind {
430 ctx.lexed.kind(*self)
433 fn to_char(&self, ctx: &RawConvertor<'a>) -> Option<char> {
434 ctx.lexed.text(*self).chars().next()
437 fn to_text(&self, ctx: &RawConvertor<'_>) -> SmolStr {
438 ctx.lexed.text(*self).into()
442 impl<'a> TokenConvertor for RawConvertor<'a> {
445 fn convert_doc_comment(&self, &token: &usize) -> Option<Vec<tt::TokenTree>> {
446 let text = self.lexed.text(token);
447 convert_doc_comment(&doc_comment(text))
450 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
451 if self.pos == self.lexed.len() {
454 let token = self.pos;
456 let range = self.lexed.text_range(token);
457 let range = TextRange::new(range.start.try_into().unwrap(), range.end.try_into().unwrap());
462 fn peek(&self) -> Option<Self::Token> {
463 if self.pos == self.lexed.len() {
469 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
475 id_alloc: TokenIdAlloc,
476 current: Option<SyntaxToken>,
477 current_synthetic: Vec<SyntheticToken>,
478 preorder: PreorderWithTokens,
479 replace: FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
480 append: FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
482 punct_offset: Option<(SyntaxToken, TextSize)>,
488 global_offset: TextSize,
489 mut replace: FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
490 mut append: FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
492 let range = node.text_range();
493 let mut preorder = node.preorder_with_tokens();
494 let (first, synthetic) = Self::next_token(&mut preorder, &mut replace, &mut append);
496 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
498 current_synthetic: synthetic,
508 preorder: &mut PreorderWithTokens,
509 replace: &mut FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
510 append: &mut FxHashMap<SyntaxNode, Vec<SyntheticToken>>,
511 ) -> (Option<SyntaxToken>, Vec<SyntheticToken>) {
512 while let Some(ev) = preorder.next() {
514 WalkEvent::Enter(ele) => ele,
515 WalkEvent::Leave(SyntaxElement::Node(node)) => {
516 if let Some(mut v) = append.remove(&node) {
517 eprintln!("after {:?}, appending {:?}", node, v);
528 SyntaxElement::Token(t) => return (Some(t), Vec::new()),
529 SyntaxElement::Node(node) => {
530 if let Some(mut v) = replace.remove(&node) {
531 preorder.skip_subtree();
532 eprintln!("replacing {:?} by {:?}", node, v);
547 Ordinary(SyntaxToken),
548 // FIXME is this supposed to be `Punct`?
549 Punch(SyntaxToken, TextSize),
550 Synthetic(SyntheticToken),
554 fn token(&self) -> Option<&SyntaxToken> {
556 SynToken::Ordinary(it) | SynToken::Punch(it, _) => Some(it),
557 SynToken::Synthetic(_) => None,
562 impl SrcToken<Convertor> for SynToken {
563 fn kind(&self, _ctx: &Convertor) -> SyntaxKind {
565 SynToken::Ordinary(token) => token.kind(),
566 SynToken::Punch(token, _) => token.kind(),
567 SynToken::Synthetic((kind, _)) => *kind,
570 fn to_char(&self, _ctx: &Convertor) -> Option<char> {
572 SynToken::Ordinary(_) => None,
573 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
574 SynToken::Synthetic(_) => None,
577 fn to_text(&self, _ctx: &Convertor) -> SmolStr {
579 SynToken::Ordinary(token) => token.text().into(),
580 SynToken::Punch(token, _) => token.text().into(),
581 SynToken::Synthetic((_, text)) => text.clone(),
586 impl TokenConvertor for Convertor {
587 type Token = SynToken;
588 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
589 convert_doc_comment(token.token()?)
592 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
593 if let Some((punct, offset)) = self.punct_offset.clone() {
594 if usize::from(offset) + 1 < punct.text().len() {
595 let offset = offset + TextSize::of('.');
596 let range = punct.text_range();
597 self.punct_offset = Some((punct.clone(), offset));
598 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
599 return Some((SynToken::Punch(punct, offset), range));
603 if let Some(synth_token) = self.current_synthetic.pop() {
604 if self.current_synthetic.is_empty() {
605 let (new_current, new_synth) =
606 Self::next_token(&mut self.preorder, &mut self.replace, &mut self.append);
607 self.current = new_current;
608 self.current_synthetic = new_synth;
611 return Some((SynToken::Synthetic(synth_token), self.range));
614 let curr = self.current.clone()?;
615 if !&self.range.contains_range(curr.text_range()) {
618 let (new_current, new_synth) =
619 Self::next_token(&mut self.preorder, &mut self.replace, &mut self.append);
620 self.current = new_current;
621 self.current_synthetic = new_synth;
622 let token = if curr.kind().is_punct() {
623 self.punct_offset = Some((curr.clone(), 0.into()));
624 let range = curr.text_range();
625 let range = TextRange::at(range.start(), TextSize::of('.'));
626 (SynToken::Punch(curr, 0.into()), range)
628 self.punct_offset = None;
629 let range = curr.text_range();
630 (SynToken::Ordinary(curr), range)
636 fn peek(&self) -> Option<Self::Token> {
637 if let Some((punct, mut offset)) = self.punct_offset.clone() {
638 offset += TextSize::of('.');
639 if usize::from(offset) < punct.text().len() {
640 return Some(SynToken::Punch(punct, offset));
644 if let Some(synth_token) = self.current_synthetic.last() {
646 return Some(SynToken::Synthetic(synth_token.clone()));
649 let curr = self.current.clone()?;
650 if !self.range.contains_range(curr.text_range()) {
654 let token = if curr.kind().is_punct() {
655 SynToken::Punch(curr, 0.into())
657 SynToken::Ordinary(curr)
662 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
667 struct TtTreeSink<'a> {
670 open_delims: FxHashMap<tt::TokenId, TextSize>,
672 inner: SyntaxTreeBuilder,
676 impl<'a> TtTreeSink<'a> {
677 fn new(cursor: Cursor<'a>) -> Self {
681 open_delims: FxHashMap::default(),
683 inner: SyntaxTreeBuilder::default(),
684 token_map: TokenMap::default(),
688 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
689 self.token_map.shrink_to_fit();
690 (self.inner.finish(), self.token_map)
694 fn delim_to_str(d: tt::DelimiterKind, closing: bool) -> &'static str {
695 let texts = match d {
696 tt::DelimiterKind::Parenthesis => "()",
697 tt::DelimiterKind::Brace => "{}",
698 tt::DelimiterKind::Bracket => "[]",
701 let idx = closing as usize;
702 &texts[idx..texts.len() - (1 - idx)]
705 impl<'a> TtTreeSink<'a> {
706 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
707 if kind == LIFETIME_IDENT {
711 let mut last = self.cursor;
712 for _ in 0..n_tokens {
714 if self.cursor.eof() {
718 let text: &str = loop {
719 break match self.cursor.token_tree() {
720 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
721 // Mark the range if needed
722 let (text, id) = match leaf {
723 tt::Leaf::Ident(ident) => (ident.text.as_str(), ident.id),
724 tt::Leaf::Punct(punct) => {
725 assert!(punct.char.is_ascii());
726 tmp = punct.char as u8;
727 (std::str::from_utf8(std::slice::from_ref(&tmp)).unwrap(), punct.id)
729 tt::Leaf::Literal(lit) => (lit.text.as_str(), lit.id),
731 let range = TextRange::at(self.text_pos, TextSize::of(text));
732 self.token_map.insert(id, range);
733 self.cursor = self.cursor.bump();
736 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
737 self.cursor = self.cursor.subtree().unwrap();
738 match subtree.delimiter {
740 self.open_delims.insert(d.id, self.text_pos);
741 delim_to_str(d.kind, false)
747 let parent = self.cursor.end().unwrap();
748 self.cursor = self.cursor.bump();
749 match parent.delimiter {
751 if let Some(open_delim) = self.open_delims.get(&d.id) {
752 let open_range = TextRange::at(*open_delim, TextSize::of('('));
754 TextRange::at(self.text_pos, TextSize::of('('));
755 self.token_map.insert_delim(d.id, open_range, close_range);
757 delim_to_str(d.kind, true)
765 self.text_pos += TextSize::of(text);
768 self.inner.token(kind, self.buf.as_str());
770 // Add whitespace between adjoint puncts
771 let next = last.bump();
773 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
774 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
775 ) = (last.token_tree(), next.token_tree())
777 // Note: We always assume the semi-colon would be the last token in
778 // other parts of RA such that we don't add whitespace here.
779 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
780 self.inner.token(WHITESPACE, " ");
781 self.text_pos += TextSize::of(' ');
786 fn start_node(&mut self, kind: SyntaxKind) {
787 self.inner.start_node(kind);
790 fn finish_node(&mut self) {
791 self.inner.finish_node();
794 fn error(&mut self, error: String) {
795 self.inner.error(error, self.text_pos)