1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
3 use rustc_hash::{FxHashMap, FxHashSet};
5 ast::{self, make::tokens::doc_comment},
6 AstToken, Parse, PreorderWithTokens, SmolStr, SyntaxElement, SyntaxKind,
8 SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, WalkEvent, T,
10 use tt::buffer::{Cursor, TokenBuffer};
12 use crate::{to_parser_input::to_parser_input, tt_iter::TtIter, ExpandError, TokenMap};
14 /// Convert the syntax node to a `TokenTree` (what macro
16 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
17 syntax_node_to_token_tree_censored(node, &Default::default())
20 /// Convert the syntax node to a `TokenTree` (what macro will consume)
21 /// with the censored range excluded.
22 pub fn syntax_node_to_token_tree_censored(
24 censor: &FxHashSet<SyntaxNode>,
25 ) -> (tt::Subtree, TokenMap) {
26 let global_offset = node.text_range().start();
27 let mut c = Convertor::new(node, global_offset, censor);
28 let subtree = convert_tokens(&mut c);
29 c.id_alloc.map.shrink_to_fit();
30 (subtree, c.id_alloc.map)
33 // The following items are what `rustc` macro can be parsed into :
34 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
35 // * Expr(P<ast::Expr>) -> token_tree_to_expr
36 // * Pat(P<ast::Pat>) -> token_tree_to_pat
37 // * Ty(P<ast::Ty>) -> token_tree_to_ty
38 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
39 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
41 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
42 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
43 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
45 pub fn token_tree_to_syntax_node(
47 entry_point: parser::TopEntryPoint,
48 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
49 let buffer = match tt {
50 tt::Subtree { delimiter: None, token_trees } => {
51 TokenBuffer::from_tokens(token_trees.as_slice())
53 _ => TokenBuffer::from_subtree(tt),
55 let parser_input = to_parser_input(&buffer);
56 let parser_output = entry_point.parse(&parser_input);
57 let mut tree_sink = TtTreeSink::new(buffer.begin());
58 for event in parser_output.iter() {
60 parser::Step::Token { kind, n_input_tokens: n_raw_tokens } => {
61 tree_sink.token(kind, n_raw_tokens)
63 parser::Step::Enter { kind } => tree_sink.start_node(kind),
64 parser::Step::Exit => tree_sink.finish_node(),
65 parser::Step::Error { msg } => tree_sink.error(msg.to_string()),
68 if tree_sink.roots.len() != 1 {
69 return Err(ExpandError::ConversionError);
71 //FIXME: would be cool to report errors
72 let (parse, range_map) = tree_sink.finish();
73 Ok((parse, range_map))
76 /// Convert a string to a `TokenTree`
77 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
78 let lexed = parser::LexedStr::new(text);
79 if lexed.errors().next().is_some() {
83 let mut conv = RawConvertor {
86 id_alloc: TokenIdAlloc {
87 map: Default::default(),
88 global_offset: TextSize::default(),
93 let subtree = convert_tokens(&mut conv);
94 Some((subtree, conv.id_alloc.map))
97 /// Split token tree with separate expr: $($e:expr)SEP*
98 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
99 if tt.token_trees.is_empty() {
103 let mut iter = TtIter::new(tt);
104 let mut res = Vec::new();
106 while iter.peek_n(0).is_some() {
107 let expanded = iter.expect_fragment(parser::PrefixEntryPoint::Expr);
109 res.push(match expanded.value {
111 Some(tt @ tt::TokenTree::Leaf(_)) => {
112 tt::Subtree { delimiter: None, token_trees: vec![tt] }
114 Some(tt::TokenTree::Subtree(tt)) => tt,
117 let mut fork = iter.clone();
118 if fork.expect_char(sep).is_err() {
124 if iter.peek_n(0).is_some() {
125 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
131 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
133 subtree: tt::Subtree,
135 open_range: TextRange,
138 let entry = StackEntry {
139 subtree: tt::Subtree { delimiter: None, ..Default::default() },
140 // never used (delimiter is `None`)
142 open_range: TextRange::empty(TextSize::of('.')),
144 let mut stack = vec![entry];
147 let entry = stack.last_mut().unwrap();
148 let result = &mut entry.subtree.token_trees;
149 let (token, range) = match conv.bump() {
154 let k: SyntaxKind = token.kind(&conv);
156 if let Some(tokens) = conv.convert_doc_comment(&token) {
157 // FIXME: There has to be a better way to do this
158 // Add the comments token id to the converted doc string
159 let id = conv.id_alloc().alloc(range);
160 result.extend(tokens.into_iter().map(|mut tt| {
161 if let tt::TokenTree::Subtree(sub) = &mut tt {
162 if let tt::TokenTree::Leaf(tt::Leaf::Literal(lit)) = &mut sub.token_trees[2]
173 result.push(if k.is_punct() && k != UNDERSCORE {
174 assert_eq!(range.len(), TextSize::of('.'));
176 if let Some(delim) = entry.subtree.delimiter {
177 let expected = match delim.kind {
178 tt::DelimiterKind::Parenthesis => T![')'],
179 tt::DelimiterKind::Brace => T!['}'],
180 tt::DelimiterKind::Bracket => T![']'],
184 let entry = stack.pop().unwrap();
185 conv.id_alloc().close_delim(entry.idx, Some(range));
186 stack.last_mut().unwrap().subtree.token_trees.push(entry.subtree.into());
191 let delim = match k {
192 T!['('] => Some(tt::DelimiterKind::Parenthesis),
193 T!['{'] => Some(tt::DelimiterKind::Brace),
194 T!['['] => Some(tt::DelimiterKind::Bracket),
198 if let Some(kind) = delim {
199 let mut subtree = tt::Subtree::default();
200 let (id, idx) = conv.id_alloc().open_delim(range);
201 subtree.delimiter = Some(tt::Delimiter { id, kind });
202 stack.push(StackEntry { subtree, idx, open_range: range });
205 let spacing = match conv.peek() {
207 if next.kind(&conv).is_trivia()
208 || next.kind(&conv) == T!['[']
209 || next.kind(&conv) == T!['{']
210 || next.kind(&conv) == T!['('] =>
214 Some(next) if next.kind(&conv).is_punct() && next.kind(&conv) != UNDERSCORE => {
217 _ => tt::Spacing::Alone,
219 let char = match token.to_char(&conv) {
222 panic!("Token from lexer must be single char: token = {:#?}", token);
225 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
228 macro_rules! make_leaf {
230 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text(conv) }.into()
233 let leaf: tt::Leaf = match k {
234 T![true] | T![false] => make_leaf!(Ident),
235 IDENT => make_leaf!(Ident),
236 UNDERSCORE => make_leaf!(Ident),
237 k if k.is_keyword() => make_leaf!(Ident),
238 k if k.is_literal() => make_leaf!(Literal),
240 let char_unit = TextSize::of('\'');
241 let r = TextRange::at(range.start(), char_unit);
242 let apostrophe = tt::Leaf::from(tt::Punct {
244 spacing: tt::Spacing::Joint,
245 id: conv.id_alloc().alloc(r),
247 result.push(apostrophe.into());
249 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
250 let ident = tt::Leaf::from(tt::Ident {
251 text: SmolStr::new(&token.to_text(conv)[1..]),
252 id: conv.id_alloc().alloc(r),
254 result.push(ident.into());
264 // If we get here, we've consumed all input tokens.
265 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
266 // Merge them so we're left with one.
267 while stack.len() > 1 {
268 let entry = stack.pop().unwrap();
269 let parent = stack.last_mut().unwrap();
271 conv.id_alloc().close_delim(entry.idx, None);
272 let leaf: tt::Leaf = tt::Punct {
273 id: conv.id_alloc().alloc(entry.open_range),
274 char: match entry.subtree.delimiter.unwrap().kind {
275 tt::DelimiterKind::Parenthesis => '(',
276 tt::DelimiterKind::Brace => '{',
277 tt::DelimiterKind::Bracket => '[',
279 spacing: tt::Spacing::Alone,
282 parent.subtree.token_trees.push(leaf.into());
283 parent.subtree.token_trees.extend(entry.subtree.token_trees);
286 let subtree = stack.pop().unwrap().subtree;
287 if subtree.token_trees.len() == 1 {
288 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
289 return first.clone();
295 /// Returns the textual content of a doc comment block as a quoted string
296 /// That is, strips leading `///` (or `/**`, etc)
297 /// and strips the ending `*/`
298 /// And then quote the string, which is needed to convert to `tt::Literal`
299 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
300 let prefix_len = comment.prefix().len();
301 let mut text = &comment.text()[prefix_len..];
303 // Remove ending "*/"
304 if comment.kind().shape == ast::CommentShape::Block {
305 text = &text[0..text.len() - 2];
309 // Note that `tt::Literal` expect an escaped string
310 let text = format!("\"{}\"", text.escape_debug());
314 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
315 cov_mark::hit!(test_meta_doc_comments);
316 let comment = ast::Comment::cast(token.clone())?;
317 let doc = comment.kind().doc?;
319 // Make `doc="\" Comments\""
320 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
323 let mut token_trees = vec![mk_punct('#')];
324 if let ast::CommentPlacement::Inner = doc {
325 token_trees.push(mk_punct('!'));
327 token_trees.push(tt::TokenTree::from(tt::Subtree {
328 delimiter: Some(tt::Delimiter {
329 kind: tt::DelimiterKind::Bracket,
330 id: tt::TokenId::unspecified(),
332 token_trees: meta_tkns,
335 return Some(token_trees);
338 fn mk_ident(s: &str) -> tt::TokenTree {
339 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
341 id: tt::TokenId::unspecified(),
345 fn mk_punct(c: char) -> tt::TokenTree {
346 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
348 spacing: tt::Spacing::Alone,
349 id: tt::TokenId::unspecified(),
353 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
354 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
356 tt::TokenTree::from(tt::Leaf::from(lit))
360 struct TokenIdAlloc {
362 global_offset: TextSize,
367 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
368 let relative_range = absolute_range - self.global_offset;
369 let token_id = tt::TokenId(self.next_id);
371 self.map.insert(token_id, relative_range);
375 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
376 let token_id = tt::TokenId(self.next_id);
378 let idx = self.map.insert_delim(
380 open_abs_range - self.global_offset,
381 open_abs_range - self.global_offset,
386 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
387 match close_abs_range {
389 self.map.remove_delim(idx);
392 self.map.update_close_delim(idx, close - self.global_offset);
398 /// A Raw Token (straightly from lexer) convertor
399 struct RawConvertor<'a> {
400 lexed: parser::LexedStr<'a>,
402 id_alloc: TokenIdAlloc,
405 trait SrcToken<Ctx>: std::fmt::Debug {
406 fn kind(&self, ctx: &Ctx) -> SyntaxKind;
408 fn to_char(&self, ctx: &Ctx) -> Option<char>;
410 fn to_text(&self, ctx: &Ctx) -> SmolStr;
413 trait TokenConvertor: Sized {
414 type Token: SrcToken<Self>;
416 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
418 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
420 fn peek(&self) -> Option<Self::Token>;
422 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
425 impl<'a> SrcToken<RawConvertor<'a>> for usize {
426 fn kind(&self, ctx: &RawConvertor<'a>) -> SyntaxKind {
427 ctx.lexed.kind(*self)
430 fn to_char(&self, ctx: &RawConvertor<'a>) -> Option<char> {
431 ctx.lexed.text(*self).chars().next()
434 fn to_text(&self, ctx: &RawConvertor<'_>) -> SmolStr {
435 ctx.lexed.text(*self).into()
439 impl<'a> TokenConvertor for RawConvertor<'a> {
442 fn convert_doc_comment(&self, token: &usize) -> Option<Vec<tt::TokenTree>> {
443 let text = self.lexed.text(*token);
444 convert_doc_comment(&doc_comment(text))
447 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
448 if self.pos == self.lexed.len() {
451 let token = self.pos;
453 let range = self.lexed.text_range(token);
454 let range = TextRange::new(range.start.try_into().unwrap(), range.end.try_into().unwrap());
459 fn peek(&self) -> Option<Self::Token> {
460 if self.pos == self.lexed.len() {
466 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
471 struct Convertor<'c> {
472 id_alloc: TokenIdAlloc,
473 current: Option<SyntaxToken>,
474 preorder: PreorderWithTokens,
475 censor: &'c FxHashSet<SyntaxNode>,
477 punct_offset: Option<(SyntaxToken, TextSize)>,
480 impl<'c> Convertor<'c> {
483 global_offset: TextSize,
484 censor: &'c FxHashSet<SyntaxNode>,
486 let range = node.text_range();
487 let mut preorder = node.preorder_with_tokens();
488 let first = Self::next_token(&mut preorder, censor);
490 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
500 preorder: &mut PreorderWithTokens,
501 censor: &FxHashSet<SyntaxNode>,
502 ) -> Option<SyntaxToken> {
503 while let Some(ev) = preorder.next() {
505 WalkEvent::Enter(ele) => ele,
509 SyntaxElement::Token(t) => return Some(t),
510 SyntaxElement::Node(node) if censor.contains(&node) => preorder.skip_subtree(),
511 SyntaxElement::Node(_) => (),
520 Ordinary(SyntaxToken),
521 Punch(SyntaxToken, TextSize),
525 fn token(&self) -> &SyntaxToken {
527 SynToken::Ordinary(it) => it,
528 SynToken::Punch(it, _) => it,
533 impl<'a> SrcToken<Convertor<'a>> for SynToken {
534 fn kind(&self, _ctx: &Convertor<'a>) -> SyntaxKind {
537 fn to_char(&self, _ctx: &Convertor<'a>) -> Option<char> {
539 SynToken::Ordinary(_) => None,
540 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
543 fn to_text(&self, _ctx: &Convertor<'a>) -> SmolStr {
544 self.token().text().into()
548 impl TokenConvertor for Convertor<'_> {
549 type Token = SynToken;
550 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
551 convert_doc_comment(token.token())
554 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
555 if let Some((punct, offset)) = self.punct_offset.clone() {
556 if usize::from(offset) + 1 < punct.text().len() {
557 let offset = offset + TextSize::of('.');
558 let range = punct.text_range();
559 self.punct_offset = Some((punct.clone(), offset));
560 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
561 return Some((SynToken::Punch(punct, offset), range));
565 let curr = self.current.clone()?;
566 if !&self.range.contains_range(curr.text_range()) {
569 self.current = Self::next_token(&mut self.preorder, self.censor);
570 let token = if curr.kind().is_punct() {
571 let range = curr.text_range();
572 let range = TextRange::at(range.start(), TextSize::of('.'));
573 self.punct_offset = Some((curr.clone(), 0.into()));
574 (SynToken::Punch(curr, 0.into()), range)
576 self.punct_offset = None;
577 let range = curr.text_range();
578 (SynToken::Ordinary(curr), range)
584 fn peek(&self) -> Option<Self::Token> {
585 if let Some((punct, mut offset)) = self.punct_offset.clone() {
586 offset += TextSize::of('.');
587 if usize::from(offset) < punct.text().len() {
588 return Some(SynToken::Punch(punct, offset));
592 let curr = self.current.clone()?;
593 if !self.range.contains_range(curr.text_range()) {
597 let token = if curr.kind().is_punct() {
598 SynToken::Punch(curr, 0.into())
600 SynToken::Ordinary(curr)
605 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
610 struct TtTreeSink<'a> {
613 open_delims: FxHashMap<tt::TokenId, TextSize>,
615 inner: SyntaxTreeBuilder,
619 // Use for detect ill-form tree which is not single root
620 roots: smallvec::SmallVec<[usize; 1]>,
623 impl<'a> TtTreeSink<'a> {
624 fn new(cursor: Cursor<'a>) -> Self {
628 open_delims: FxHashMap::default(),
630 inner: SyntaxTreeBuilder::default(),
631 roots: smallvec::SmallVec::new(),
632 token_map: TokenMap::default(),
636 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
637 self.token_map.shrink_to_fit();
638 (self.inner.finish(), self.token_map)
642 fn delim_to_str(d: tt::DelimiterKind, closing: bool) -> &'static str {
643 let texts = match d {
644 tt::DelimiterKind::Parenthesis => "()",
645 tt::DelimiterKind::Brace => "{}",
646 tt::DelimiterKind::Bracket => "[]",
649 let idx = closing as usize;
650 &texts[idx..texts.len() - (1 - idx)]
653 impl<'a> TtTreeSink<'a> {
654 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
655 if kind == LIFETIME_IDENT {
659 let mut last = self.cursor;
660 for _ in 0..n_tokens {
661 let tmp_str: SmolStr;
662 if self.cursor.eof() {
666 let text: &str = loop {
667 break match self.cursor.token_tree() {
668 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
669 // Mark the range if needed
670 let (text, id) = match leaf {
671 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
672 tt::Leaf::Punct(punct) => {
673 assert!(punct.char.is_ascii());
674 let char = &(punct.char as u8);
675 tmp_str = SmolStr::new_inline(
676 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
680 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
682 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
683 self.token_map.insert(id, range);
684 self.cursor = self.cursor.bump();
687 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
688 self.cursor = self.cursor.subtree().unwrap();
689 match subtree.delimiter {
691 self.open_delims.insert(d.id, self.text_pos);
692 delim_to_str(d.kind, false)
698 let parent = self.cursor.end().unwrap();
699 self.cursor = self.cursor.bump();
700 match parent.delimiter {
702 if let Some(open_delim) = self.open_delims.get(&d.id) {
703 let open_range = TextRange::at(*open_delim, TextSize::of('('));
705 TextRange::at(self.text_pos, TextSize::of('('));
706 self.token_map.insert_delim(d.id, open_range, close_range);
708 delim_to_str(d.kind, true)
716 self.text_pos += TextSize::of(text);
719 self.inner.token(kind, self.buf.as_str());
721 // Add whitespace between adjoint puncts
722 let next = last.bump();
724 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
725 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
726 ) = (last.token_tree(), next.token_tree())
728 // Note: We always assume the semi-colon would be the last token in
729 // other parts of RA such that we don't add whitespace here.
730 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
731 self.inner.token(WHITESPACE, " ");
732 self.text_pos += TextSize::of(' ');
737 fn start_node(&mut self, kind: SyntaxKind) {
738 self.inner.start_node(kind);
740 match self.roots.last_mut() {
741 None | Some(0) => self.roots.push(1),
742 Some(ref mut n) => **n += 1,
746 fn finish_node(&mut self) {
747 self.inner.finish_node();
748 *self.roots.last_mut().unwrap() -= 1;
751 fn error(&mut self, error: String) {
752 self.inner.error(error, self.text_pos)