1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
3 use rustc_hash::{FxHashMap, FxHashSet};
4 use stdx::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, ExpandError, 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())
21 /// Convert the syntax node to a `TokenTree` (what macro will consume)
22 /// with the censored range excluded.
23 pub fn syntax_node_to_token_tree_censored(
25 censor: &FxHashSet<SyntaxNode>,
26 ) -> (tt::Subtree, TokenMap) {
27 let global_offset = node.text_range().start();
28 let mut c = Convertor::new(node, global_offset, censor);
29 let subtree = convert_tokens(&mut c);
30 c.id_alloc.map.shrink_to_fit();
31 (subtree, c.id_alloc.map)
34 // The following items are what `rustc` macro can be parsed into :
35 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
36 // * Expr(P<ast::Expr>) -> token_tree_to_expr
37 // * Pat(P<ast::Pat>) -> token_tree_to_pat
38 // * Ty(P<ast::Ty>) -> token_tree_to_ty
39 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
40 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
42 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
43 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
44 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
46 pub fn token_tree_to_syntax_node(
48 entry_point: parser::TopEntryPoint,
49 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
50 let buffer = match tt {
51 tt::Subtree { delimiter: None, token_trees } => {
52 TokenBuffer::from_tokens(token_trees.as_slice())
54 _ => TokenBuffer::from_subtree(tt),
56 let parser_input = to_parser_input(&buffer);
57 let parser_output = entry_point.parse(&parser_input);
58 let mut tree_sink = TtTreeSink::new(buffer.begin());
59 for event in parser_output.iter() {
61 parser::Step::Token { kind, n_input_tokens: n_raw_tokens } => {
62 tree_sink.token(kind, n_raw_tokens)
64 parser::Step::Enter { kind } => tree_sink.start_node(kind),
65 parser::Step::Exit => tree_sink.finish_node(),
66 parser::Step::Error { msg } => tree_sink.error(msg.to_string()),
69 let (parse, range_map) = tree_sink.finish();
70 Ok((parse, range_map))
73 /// Convert a string to a `TokenTree`
74 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
75 let lexed = parser::LexedStr::new(text);
76 if lexed.errors().next().is_some() {
80 let mut conv = RawConvertor {
83 id_alloc: TokenIdAlloc {
84 map: Default::default(),
85 global_offset: TextSize::default(),
90 let subtree = convert_tokens(&mut conv);
91 Some((subtree, conv.id_alloc.map))
94 /// Split token tree with separate expr: $($e:expr)SEP*
95 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
96 if tt.token_trees.is_empty() {
100 let mut iter = TtIter::new(tt);
101 let mut res = Vec::new();
103 while iter.peek_n(0).is_some() {
104 let expanded = iter.expect_fragment(parser::PrefixEntryPoint::Expr);
106 res.push(match expanded.value {
108 Some(tt @ tt::TokenTree::Leaf(_)) => {
109 tt::Subtree { delimiter: None, token_trees: vec![tt] }
111 Some(tt::TokenTree::Subtree(tt)) => tt,
114 let mut fork = iter.clone();
115 if fork.expect_char(sep).is_err() {
121 if iter.peek_n(0).is_some() {
122 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
128 fn convert_tokens<C: TokenConvertor>(conv: &mut C) -> tt::Subtree {
130 subtree: tt::Subtree,
132 open_range: TextRange,
135 let entry = StackEntry {
136 subtree: tt::Subtree { delimiter: None, ..Default::default() },
137 // never used (delimiter is `None`)
139 open_range: TextRange::empty(TextSize::of('.')),
141 let mut stack = NonEmptyVec::new(entry);
144 let StackEntry { subtree, .. } = stack.last_mut();
145 let result = &mut subtree.token_trees;
146 let (token, range) = match conv.bump() {
151 let kind = token.kind(&conv);
153 if let Some(tokens) = conv.convert_doc_comment(&token) {
154 // FIXME: There has to be a better way to do this
155 // Add the comments token id to the converted doc string
156 let id = conv.id_alloc().alloc(range);
157 result.extend(tokens.into_iter().map(|mut tt| {
158 if let tt::TokenTree::Subtree(sub) = &mut tt {
159 if let Some(tt::TokenTree::Leaf(tt::Leaf::Literal(lit))) =
160 sub.token_trees.get_mut(2)
170 let tt = if kind.is_punct() && kind != UNDERSCORE {
171 assert_eq!(range.len(), TextSize::of('.'));
173 if let Some(delim) = subtree.delimiter {
174 let expected = match delim.kind {
175 tt::DelimiterKind::Parenthesis => T![')'],
176 tt::DelimiterKind::Brace => T!['}'],
177 tt::DelimiterKind::Bracket => T![']'],
180 if kind == expected {
181 if let Some(entry) = stack.pop() {
182 conv.id_alloc().close_delim(entry.idx, Some(range));
183 stack.last_mut().subtree.token_trees.push(entry.subtree.into());
189 let delim = match kind {
190 T!['('] => Some(tt::DelimiterKind::Parenthesis),
191 T!['{'] => Some(tt::DelimiterKind::Brace),
192 T!['['] => Some(tt::DelimiterKind::Bracket),
196 if let Some(kind) = delim {
197 let mut subtree = tt::Subtree::default();
198 let (id, idx) = conv.id_alloc().open_delim(range);
199 subtree.delimiter = Some(tt::Delimiter { id, kind });
200 stack.push(StackEntry { subtree, idx, open_range: range });
204 let spacing = match conv.peek().map(|next| next.kind(&conv)) {
211 && kind != UNDERSCORE =>
215 _ => tt::Spacing::Alone,
217 let char = match token.to_char(&conv) {
220 panic!("Token from lexer must be single char: token = {:#?}", token);
223 tt::Leaf::from(tt::Punct { char, spacing, id: conv.id_alloc().alloc(range) }).into()
225 macro_rules! make_leaf {
227 tt::$i { id: conv.id_alloc().alloc(range), text: token.to_text(conv) }.into()
230 let leaf: tt::Leaf = match kind {
231 T![true] | T![false] => make_leaf!(Ident),
232 IDENT => make_leaf!(Ident),
233 UNDERSCORE => make_leaf!(Ident),
234 k if k.is_keyword() => make_leaf!(Ident),
235 k if k.is_literal() => make_leaf!(Literal),
237 let char_unit = TextSize::of('\'');
238 let r = TextRange::at(range.start(), char_unit);
239 let apostrophe = tt::Leaf::from(tt::Punct {
241 spacing: tt::Spacing::Joint,
242 id: conv.id_alloc().alloc(r),
244 result.push(apostrophe.into());
246 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
247 let ident = tt::Leaf::from(tt::Ident {
248 text: SmolStr::new(&token.to_text(conv)[1..]),
249 id: conv.id_alloc().alloc(r),
251 result.push(ident.into());
262 // If we get here, we've consumed all input tokens.
263 // We might have more than one subtree in the stack, if the delimiters are improperly balanced.
264 // Merge them so we're left with one.
265 while let Some(entry) = stack.pop() {
266 let parent = stack.last_mut();
268 conv.id_alloc().close_delim(entry.idx, None);
269 let leaf: tt::Leaf = tt::Punct {
270 id: conv.id_alloc().alloc(entry.open_range),
271 char: match entry.subtree.delimiter.unwrap().kind {
272 tt::DelimiterKind::Parenthesis => '(',
273 tt::DelimiterKind::Brace => '{',
274 tt::DelimiterKind::Bracket => '[',
276 spacing: tt::Spacing::Alone,
279 parent.subtree.token_trees.push(leaf.into());
280 parent.subtree.token_trees.extend(entry.subtree.token_trees);
283 let subtree = stack.into_last().subtree;
284 if let [tt::TokenTree::Subtree(first)] = &*subtree.token_trees {
291 /// Returns the textual content of a doc comment block as a quoted string
292 /// That is, strips leading `///` (or `/**`, etc)
293 /// and strips the ending `*/`
294 /// And then quote the string, which is needed to convert to `tt::Literal`
295 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
296 let prefix_len = comment.prefix().len();
297 let mut text = &comment.text()[prefix_len..];
299 // Remove ending "*/"
300 if comment.kind().shape == ast::CommentShape::Block {
301 text = &text[0..text.len() - 2];
305 // Note that `tt::Literal` expect an escaped string
306 let text = format!("\"{}\"", text.escape_debug());
310 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
311 cov_mark::hit!(test_meta_doc_comments);
312 let comment = ast::Comment::cast(token.clone())?;
313 let doc = comment.kind().doc?;
315 // Make `doc="\" Comments\""
316 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
319 let mut token_trees = Vec::with_capacity(3);
320 token_trees.push(mk_punct('#'));
321 if let ast::CommentPlacement::Inner = doc {
322 token_trees.push(mk_punct('!'));
324 token_trees.push(tt::TokenTree::from(tt::Subtree {
325 delimiter: Some(tt::Delimiter {
326 kind: tt::DelimiterKind::Bracket,
327 id: tt::TokenId::unspecified(),
329 token_trees: meta_tkns,
332 return Some(token_trees);
335 fn mk_ident(s: &str) -> tt::TokenTree {
336 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
338 id: tt::TokenId::unspecified(),
342 fn mk_punct(c: char) -> tt::TokenTree {
343 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
345 spacing: tt::Spacing::Alone,
346 id: tt::TokenId::unspecified(),
350 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
351 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
353 tt::TokenTree::from(tt::Leaf::from(lit))
357 struct TokenIdAlloc {
359 global_offset: TextSize,
364 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
365 let relative_range = absolute_range - self.global_offset;
366 let token_id = tt::TokenId(self.next_id);
368 self.map.insert(token_id, relative_range);
372 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
373 let token_id = tt::TokenId(self.next_id);
375 let idx = self.map.insert_delim(
377 open_abs_range - self.global_offset,
378 open_abs_range - self.global_offset,
383 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
384 match close_abs_range {
386 self.map.remove_delim(idx);
389 self.map.update_close_delim(idx, close - self.global_offset);
395 /// A raw token (straight from lexer) convertor
396 struct RawConvertor<'a> {
397 lexed: parser::LexedStr<'a>,
399 id_alloc: TokenIdAlloc,
402 trait SrcToken<Ctx>: std::fmt::Debug {
403 fn kind(&self, ctx: &Ctx) -> SyntaxKind;
405 fn to_char(&self, ctx: &Ctx) -> Option<char>;
407 fn to_text(&self, ctx: &Ctx) -> SmolStr;
410 trait TokenConvertor: Sized {
411 type Token: SrcToken<Self>;
413 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
415 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
417 fn peek(&self) -> Option<Self::Token>;
419 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
422 impl<'a> SrcToken<RawConvertor<'a>> for usize {
423 fn kind(&self, ctx: &RawConvertor<'a>) -> SyntaxKind {
424 ctx.lexed.kind(*self)
427 fn to_char(&self, ctx: &RawConvertor<'a>) -> Option<char> {
428 ctx.lexed.text(*self).chars().next()
431 fn to_text(&self, ctx: &RawConvertor<'_>) -> SmolStr {
432 ctx.lexed.text(*self).into()
436 impl<'a> TokenConvertor for RawConvertor<'a> {
439 fn convert_doc_comment(&self, &token: &usize) -> Option<Vec<tt::TokenTree>> {
440 let text = self.lexed.text(token);
441 convert_doc_comment(&doc_comment(text))
444 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
445 if self.pos == self.lexed.len() {
448 let token = self.pos;
450 let range = self.lexed.text_range(token);
451 let range = TextRange::new(range.start.try_into().unwrap(), range.end.try_into().unwrap());
456 fn peek(&self) -> Option<Self::Token> {
457 if self.pos == self.lexed.len() {
463 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
468 struct Convertor<'c> {
469 id_alloc: TokenIdAlloc,
470 current: Option<SyntaxToken>,
471 preorder: PreorderWithTokens,
472 censor: &'c FxHashSet<SyntaxNode>,
474 punct_offset: Option<(SyntaxToken, TextSize)>,
477 impl<'c> Convertor<'c> {
480 global_offset: TextSize,
481 censor: &'c FxHashSet<SyntaxNode>,
483 let range = node.text_range();
484 let mut preorder = node.preorder_with_tokens();
485 let first = Self::next_token(&mut preorder, censor);
487 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
497 preorder: &mut PreorderWithTokens,
498 censor: &FxHashSet<SyntaxNode>,
499 ) -> Option<SyntaxToken> {
500 while let Some(ev) = preorder.next() {
502 WalkEvent::Enter(ele) => ele,
506 SyntaxElement::Token(t) => return Some(t),
507 SyntaxElement::Node(node) if censor.contains(&node) => preorder.skip_subtree(),
508 SyntaxElement::Node(_) => (),
517 Ordinary(SyntaxToken),
518 Punch(SyntaxToken, TextSize),
522 fn token(&self) -> &SyntaxToken {
524 SynToken::Ordinary(it) | SynToken::Punch(it, _) => it,
529 impl<'a> SrcToken<Convertor<'a>> for SynToken {
530 fn kind(&self, _ctx: &Convertor<'a>) -> SyntaxKind {
533 fn to_char(&self, _ctx: &Convertor<'a>) -> Option<char> {
535 SynToken::Ordinary(_) => None,
536 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
539 fn to_text(&self, _ctx: &Convertor<'a>) -> SmolStr {
540 self.token().text().into()
544 impl TokenConvertor for Convertor<'_> {
545 type Token = SynToken;
546 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
547 convert_doc_comment(token.token())
550 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
551 if let Some((punct, offset)) = self.punct_offset.clone() {
552 if usize::from(offset) + 1 < punct.text().len() {
553 let offset = offset + TextSize::of('.');
554 let range = punct.text_range();
555 self.punct_offset = Some((punct.clone(), offset));
556 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
557 return Some((SynToken::Punch(punct, offset), range));
561 let curr = self.current.clone()?;
562 if !&self.range.contains_range(curr.text_range()) {
565 self.current = Self::next_token(&mut self.preorder, self.censor);
566 let token = if curr.kind().is_punct() {
567 self.punct_offset = Some((curr.clone(), 0.into()));
568 let range = curr.text_range();
569 let range = TextRange::at(range.start(), TextSize::of('.'));
570 (SynToken::Punch(curr, 0.into()), range)
572 self.punct_offset = None;
573 let range = curr.text_range();
574 (SynToken::Ordinary(curr), range)
580 fn peek(&self) -> Option<Self::Token> {
581 if let Some((punct, mut offset)) = self.punct_offset.clone() {
582 offset += TextSize::of('.');
583 if usize::from(offset) < punct.text().len() {
584 return Some(SynToken::Punch(punct, offset));
588 let curr = self.current.clone()?;
589 if !self.range.contains_range(curr.text_range()) {
593 let token = if curr.kind().is_punct() {
594 SynToken::Punch(curr, 0.into())
596 SynToken::Ordinary(curr)
601 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
606 struct TtTreeSink<'a> {
609 open_delims: FxHashMap<tt::TokenId, TextSize>,
611 inner: SyntaxTreeBuilder,
615 impl<'a> TtTreeSink<'a> {
616 fn new(cursor: Cursor<'a>) -> Self {
620 open_delims: FxHashMap::default(),
622 inner: SyntaxTreeBuilder::default(),
623 token_map: TokenMap::default(),
627 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
628 self.token_map.shrink_to_fit();
629 (self.inner.finish(), self.token_map)
633 fn delim_to_str(d: tt::DelimiterKind, closing: bool) -> &'static str {
634 let texts = match d {
635 tt::DelimiterKind::Parenthesis => "()",
636 tt::DelimiterKind::Brace => "{}",
637 tt::DelimiterKind::Bracket => "[]",
640 let idx = closing as usize;
641 &texts[idx..texts.len() - (1 - idx)]
644 impl<'a> TtTreeSink<'a> {
645 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
646 if kind == LIFETIME_IDENT {
650 let mut last = self.cursor;
651 for _ in 0..n_tokens {
653 if self.cursor.eof() {
657 let text: &str = loop {
658 break match self.cursor.token_tree() {
659 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
660 // Mark the range if needed
661 let (text, id) = match leaf {
662 tt::Leaf::Ident(ident) => (ident.text.as_str(), ident.id),
663 tt::Leaf::Punct(punct) => {
664 assert!(punct.char.is_ascii());
665 tmp = punct.char as u8;
666 (std::str::from_utf8(std::slice::from_ref(&tmp)).unwrap(), punct.id)
668 tt::Leaf::Literal(lit) => (lit.text.as_str(), lit.id),
670 let range = TextRange::at(self.text_pos, TextSize::of(text));
671 self.token_map.insert(id, range);
672 self.cursor = self.cursor.bump();
675 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
676 self.cursor = self.cursor.subtree().unwrap();
677 match subtree.delimiter {
679 self.open_delims.insert(d.id, self.text_pos);
680 delim_to_str(d.kind, false)
686 let parent = self.cursor.end().unwrap();
687 self.cursor = self.cursor.bump();
688 match parent.delimiter {
690 if let Some(open_delim) = self.open_delims.get(&d.id) {
691 let open_range = TextRange::at(*open_delim, TextSize::of('('));
693 TextRange::at(self.text_pos, TextSize::of('('));
694 self.token_map.insert_delim(d.id, open_range, close_range);
696 delim_to_str(d.kind, true)
704 self.text_pos += TextSize::of(text);
707 self.inner.token(kind, self.buf.as_str());
709 // Add whitespace between adjoint puncts
710 let next = last.bump();
712 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
713 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
714 ) = (last.token_tree(), next.token_tree())
716 // Note: We always assume the semi-colon would be the last token in
717 // other parts of RA such that we don't add whitespace here.
718 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
719 self.inner.token(WHITESPACE, " ");
720 self.text_pos += TextSize::of(' ');
725 fn start_node(&mut self, kind: SyntaxKind) {
726 self.inner.start_node(kind);
729 fn finish_node(&mut self) {
730 self.inner.finish_node();
733 fn error(&mut self, error: String) {
734 self.inner.error(error, self.text_pos)