1 //! Conversions between [`SyntaxNode`] and [`tt::TokenTree`].
3 use parser::{FragmentKind, ParseError, TreeSink};
4 use rustc_hash::FxHashMap;
6 ast::{self, make::tokens::doc_comment},
7 tokenize, AstToken, Parse, SmolStr, SyntaxKind,
9 SyntaxNode, SyntaxToken, SyntaxTreeBuilder, TextRange, TextSize, Token as RawToken, T,
11 use tt::buffer::{Cursor, TokenBuffer};
13 use crate::{subtree_source::SubtreeTokenSource, tt_iter::TtIter};
14 use crate::{ExpandError, TokenMap};
16 /// Convert the syntax tree (what user has written) to a `TokenTree` (what macro
18 pub fn ast_to_token_tree(ast: &impl ast::AstNode) -> (tt::Subtree, TokenMap) {
19 syntax_node_to_token_tree(ast.syntax())
22 /// Convert the syntax node to a `TokenTree` (what macro
24 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> (tt::Subtree, TokenMap) {
25 let global_offset = node.text_range().start();
26 let mut c = Convertor::new(node, global_offset);
28 c.id_alloc.map.shrink_to_fit();
29 (subtree, c.id_alloc.map)
32 // The following items are what `rustc` macro can be parsed into :
33 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
34 // * Expr(P<ast::Expr>) -> token_tree_to_expr
35 // * Pat(P<ast::Pat>) -> token_tree_to_pat
36 // * Ty(P<ast::Ty>) -> token_tree_to_ty
37 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
38 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
40 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
41 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
42 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
44 pub fn token_tree_to_syntax_node(
46 fragment_kind: FragmentKind,
47 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
48 let buffer = match tt {
49 tt::Subtree { delimiter: None, token_trees } => {
50 TokenBuffer::from_tokens(token_trees.as_slice())
52 _ => TokenBuffer::from_subtree(tt),
54 let mut token_source = SubtreeTokenSource::new(&buffer);
55 let mut tree_sink = TtTreeSink::new(buffer.begin());
56 parser::parse_fragment(&mut token_source, &mut tree_sink, fragment_kind);
57 if tree_sink.roots.len() != 1 {
58 return Err(ExpandError::ConversionError);
60 //FIXME: would be cool to report errors
61 let (parse, range_map) = tree_sink.finish();
62 Ok((parse, range_map))
65 /// Convert a string to a `TokenTree`
66 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
67 let (tokens, errors) = tokenize(text);
68 if !errors.is_empty() {
72 let mut conv = RawConvertor {
74 offset: TextSize::default(),
76 id_alloc: TokenIdAlloc {
77 map: Default::default(),
78 global_offset: TextSize::default(),
83 let subtree = conv.go();
84 Some((subtree, conv.id_alloc.map))
87 /// Split token tree with seperate expr: $($e:expr)SEP*
88 pub fn parse_exprs_with_sep(tt: &tt::Subtree, sep: char) -> Vec<tt::Subtree> {
89 if tt.token_trees.is_empty() {
93 let mut iter = TtIter::new(tt);
94 let mut res = Vec::new();
96 while iter.peek_n(0).is_some() {
97 let expanded = iter.expect_fragment(FragmentKind::Expr);
98 if expanded.err.is_some() {
102 res.push(match expanded.value {
104 Some(tt @ tt::TokenTree::Leaf(_)) => {
105 tt::Subtree { delimiter: None, token_trees: vec![tt] }
107 Some(tt::TokenTree::Subtree(tt)) => tt,
110 let mut fork = iter.clone();
111 if fork.expect_char(sep).is_err() {
117 if iter.peek_n(0).is_some() {
118 res.push(tt::Subtree { delimiter: None, token_trees: iter.into_iter().cloned().collect() });
124 /// Returns the textual content of a doc comment block as a quoted string
125 /// That is, strips leading `///` (or `/**`, etc)
126 /// and strips the ending `*/`
127 /// And then quote the string, which is needed to convert to `tt::Literal`
128 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
129 let prefix_len = comment.prefix().len();
130 let mut text = &comment.text()[prefix_len..];
132 // Remove ending "*/"
133 if comment.kind().shape == ast::CommentShape::Block {
134 text = &text[0..text.len() - 2];
138 // Note that `tt::Literal` expect an escaped string
139 let text = format!("\"{}\"", text.escape_debug());
143 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
144 let comment = ast::Comment::cast(token.clone())?;
145 let doc = comment.kind().doc?;
147 // Make `doc="\" Comments\""
148 let meta_tkns = vec![mk_ident("doc"), mk_punct('='), mk_doc_literal(&comment)];
151 let mut token_trees = vec![mk_punct('#')];
152 if let ast::CommentPlacement::Inner = doc {
153 token_trees.push(mk_punct('!'));
155 token_trees.push(tt::TokenTree::from(tt::Subtree {
156 delimiter: Some(tt::Delimiter {
157 kind: tt::DelimiterKind::Bracket,
158 id: tt::TokenId::unspecified(),
160 token_trees: meta_tkns,
163 return Some(token_trees);
166 fn mk_ident(s: &str) -> tt::TokenTree {
167 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
169 id: tt::TokenId::unspecified(),
173 fn mk_punct(c: char) -> tt::TokenTree {
174 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
176 spacing: tt::Spacing::Alone,
177 id: tt::TokenId::unspecified(),
181 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
182 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
184 tt::TokenTree::from(tt::Leaf::from(lit))
188 struct TokenIdAlloc {
190 global_offset: TextSize,
195 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
196 let relative_range = absolute_range - self.global_offset;
197 let token_id = tt::TokenId(self.next_id);
199 self.map.insert(token_id, relative_range);
203 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
204 let token_id = tt::TokenId(self.next_id);
206 let idx = self.map.insert_delim(
208 open_abs_range - self.global_offset,
209 open_abs_range - self.global_offset,
214 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
215 match close_abs_range {
217 self.map.remove_delim(idx);
220 self.map.update_close_delim(idx, close - self.global_offset);
226 /// A Raw Token (straightly from lexer) convertor
227 struct RawConvertor<'a> {
230 id_alloc: TokenIdAlloc,
231 inner: std::slice::Iter<'a, RawToken>,
234 trait SrcToken: std::fmt::Debug {
235 fn kind(&self) -> SyntaxKind;
237 fn to_char(&self) -> Option<char>;
239 fn to_text(&self) -> SmolStr;
242 trait TokenConvertor {
243 type Token: SrcToken;
245 fn go(&mut self) -> tt::Subtree {
246 let mut subtree = tt::Subtree { delimiter: None, ..Default::default() };
247 while self.peek().is_some() {
248 self.collect_leaf(&mut subtree.token_trees);
250 if subtree.token_trees.len() == 1 {
251 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
252 return first.clone();
258 fn collect_leaf(&mut self, result: &mut Vec<tt::TokenTree>) {
259 let (token, range) = match self.bump() {
264 let k: SyntaxKind = token.kind();
266 if let Some(tokens) = self.convert_doc_comment(&token) {
267 result.extend(tokens);
272 result.push(if k.is_punct() && k != UNDERSCORE {
273 assert_eq!(range.len(), TextSize::of('.'));
274 let delim = match k {
275 T!['('] => Some((tt::DelimiterKind::Parenthesis, T![')'])),
276 T!['{'] => Some((tt::DelimiterKind::Brace, T!['}'])),
277 T!['['] => Some((tt::DelimiterKind::Bracket, T![']'])),
281 if let Some((kind, closed)) = delim {
282 let mut subtree = tt::Subtree::default();
283 let (id, idx) = self.id_alloc().open_delim(range);
284 subtree.delimiter = Some(tt::Delimiter { id, kind });
286 while self.peek().map(|it| it.kind() != closed).unwrap_or(false) {
287 self.collect_leaf(&mut subtree.token_trees);
289 let last_range = match self.bump() {
291 // For error resilience, we insert an char punct for the opening delim here
292 self.id_alloc().close_delim(idx, None);
293 let leaf: tt::Leaf = tt::Punct {
294 id: self.id_alloc().alloc(range),
295 char: token.to_char().unwrap(),
296 spacing: tt::Spacing::Alone,
299 result.push(leaf.into());
300 result.extend(subtree.token_trees);
305 self.id_alloc().close_delim(idx, Some(last_range));
308 let spacing = match self.peek() {
310 if next.kind().is_trivia()
311 || next.kind() == T!['[']
312 || next.kind() == T!['{']
313 || next.kind() == T!['('] =>
317 Some(next) if next.kind().is_punct() && next.kind() != UNDERSCORE => {
320 _ => tt::Spacing::Alone,
322 let char = match token.to_char() {
325 panic!("Token from lexer must be single char: token = {:#?}", token);
328 tt::Leaf::from(tt::Punct { char, spacing, id: self.id_alloc().alloc(range) }).into()
331 macro_rules! make_leaf {
333 tt::$i { id: self.id_alloc().alloc(range), text: token.to_text() }.into()
336 let leaf: tt::Leaf = match k {
337 T![true] | T![false] => make_leaf!(Ident),
338 IDENT => make_leaf!(Ident),
339 UNDERSCORE => make_leaf!(Ident),
340 k if k.is_keyword() => make_leaf!(Ident),
341 k if k.is_literal() => make_leaf!(Literal),
343 let char_unit = TextSize::of('\'');
344 let r = TextRange::at(range.start(), char_unit);
345 let apostrophe = tt::Leaf::from(tt::Punct {
347 spacing: tt::Spacing::Joint,
348 id: self.id_alloc().alloc(r),
350 result.push(apostrophe.into());
352 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
353 let ident = tt::Leaf::from(tt::Ident {
354 text: SmolStr::new(&token.to_text()[1..]),
355 id: self.id_alloc().alloc(r),
357 result.push(ident.into());
367 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
369 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
371 fn peek(&self) -> Option<Self::Token>;
373 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
376 impl<'a> SrcToken for (&'a RawToken, &'a str) {
377 fn kind(&self) -> SyntaxKind {
381 fn to_char(&self) -> Option<char> {
382 self.1.chars().next()
385 fn to_text(&self) -> SmolStr {
390 impl RawConvertor<'_> {}
392 impl<'a> TokenConvertor for RawConvertor<'a> {
393 type Token = (&'a RawToken, &'a str);
395 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
396 convert_doc_comment(&doc_comment(token.1))
399 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
400 let token = self.inner.next()?;
401 let range = TextRange::at(self.offset, token.len);
402 self.offset += token.len;
404 Some(((token, &self.text[range]), range))
407 fn peek(&self) -> Option<Self::Token> {
408 let token = self.inner.as_slice().get(0);
411 let range = TextRange::at(self.offset, it.len);
412 (it, &self.text[range])
416 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
422 id_alloc: TokenIdAlloc,
423 current: Option<SyntaxToken>,
425 punct_offset: Option<(SyntaxToken, TextSize)>,
429 fn new(node: &SyntaxNode, global_offset: TextSize) -> Convertor {
431 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
432 current: node.first_token(),
433 range: node.text_range(),
441 Ordinary(SyntaxToken),
442 Punch(SyntaxToken, TextSize),
446 fn token(&self) -> &SyntaxToken {
448 SynToken::Ordinary(it) => it,
449 SynToken::Punch(it, _) => it,
454 impl SrcToken for SynToken {
455 fn kind(&self) -> SyntaxKind {
458 fn to_char(&self) -> Option<char> {
460 SynToken::Ordinary(_) => None,
461 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
464 fn to_text(&self) -> SmolStr {
465 self.token().text().into()
469 impl TokenConvertor for Convertor {
470 type Token = SynToken;
471 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
472 convert_doc_comment(token.token())
475 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
476 if let Some((punct, offset)) = self.punct_offset.clone() {
477 if usize::from(offset) + 1 < punct.text().len() {
478 let offset = offset + TextSize::of('.');
479 let range = punct.text_range();
480 self.punct_offset = Some((punct.clone(), offset));
481 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
482 return Some((SynToken::Punch(punct, offset), range));
486 let curr = self.current.clone()?;
487 if !&self.range.contains_range(curr.text_range()) {
490 self.current = curr.next_token();
492 let token = if curr.kind().is_punct() {
493 let range = curr.text_range();
494 let range = TextRange::at(range.start(), TextSize::of('.'));
495 self.punct_offset = Some((curr.clone(), 0.into()));
496 (SynToken::Punch(curr, 0.into()), range)
498 self.punct_offset = None;
499 let range = curr.text_range();
500 (SynToken::Ordinary(curr), range)
506 fn peek(&self) -> Option<Self::Token> {
507 if let Some((punct, mut offset)) = self.punct_offset.clone() {
508 offset += TextSize::of('.');
509 if usize::from(offset) < punct.text().len() {
510 return Some(SynToken::Punch(punct, offset));
514 let curr = self.current.clone()?;
515 if !self.range.contains_range(curr.text_range()) {
519 let token = if curr.kind().is_punct() {
520 SynToken::Punch(curr, 0.into())
522 SynToken::Ordinary(curr)
527 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
532 struct TtTreeSink<'a> {
535 open_delims: FxHashMap<tt::TokenId, TextSize>,
537 inner: SyntaxTreeBuilder,
541 // Use for detect ill-form tree which is not single root
542 roots: smallvec::SmallVec<[usize; 1]>,
545 impl<'a> TtTreeSink<'a> {
546 fn new(cursor: Cursor<'a>) -> Self {
550 open_delims: FxHashMap::default(),
552 inner: SyntaxTreeBuilder::default(),
553 roots: smallvec::SmallVec::new(),
554 token_map: TokenMap::default(),
558 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
559 self.token_map.shrink_to_fit();
560 (self.inner.finish(), self.token_map)
564 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> &'static str {
565 let texts = match d {
566 Some(tt::DelimiterKind::Parenthesis) => "()",
567 Some(tt::DelimiterKind::Brace) => "{}",
568 Some(tt::DelimiterKind::Bracket) => "[]",
572 let idx = closing as usize;
573 &texts[idx..texts.len() - (1 - idx)]
576 impl<'a> TreeSink for TtTreeSink<'a> {
577 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
578 if kind == L_DOLLAR || kind == R_DOLLAR {
579 self.cursor = self.cursor.bump_subtree();
582 if kind == LIFETIME_IDENT {
586 let mut last = self.cursor;
587 for _ in 0..n_tokens {
588 let tmp_str: SmolStr;
589 if self.cursor.eof() {
593 let text: &str = match self.cursor.token_tree() {
594 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
595 // Mark the range if needed
596 let (text, id) = match leaf {
597 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
598 tt::Leaf::Punct(punct) => {
599 assert!(punct.char.is_ascii());
600 let char = &(punct.char as u8);
601 tmp_str = SmolStr::new_inline(
602 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
606 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
608 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
609 self.token_map.insert(id, range);
610 self.cursor = self.cursor.bump();
613 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
614 self.cursor = self.cursor.subtree().unwrap();
615 if let Some(id) = subtree.delimiter.map(|it| it.id) {
616 self.open_delims.insert(id, self.text_pos);
618 delim_to_str(subtree.delimiter_kind(), false)
621 if let Some(parent) = self.cursor.end() {
622 self.cursor = self.cursor.bump();
623 if let Some(id) = parent.delimiter.map(|it| it.id) {
624 if let Some(open_delim) = self.open_delims.get(&id) {
625 let open_range = TextRange::at(*open_delim, TextSize::of('('));
626 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
627 self.token_map.insert_delim(id, open_range, close_range);
630 delim_to_str(parent.delimiter_kind(), true)
637 self.text_pos += TextSize::of(text);
640 self.inner.token(kind, self.buf.as_str());
642 // Add whitespace between adjoint puncts
643 let next = last.bump();
645 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
646 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
647 ) = (last.token_tree(), next.token_tree())
649 // Note: We always assume the semi-colon would be the last token in
650 // other parts of RA such that we don't add whitespace here.
651 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
652 self.inner.token(WHITESPACE, " ");
653 self.text_pos += TextSize::of(' ');
658 fn start_node(&mut self, kind: SyntaxKind) {
659 self.inner.start_node(kind);
661 match self.roots.last_mut() {
662 None | Some(0) => self.roots.push(1),
663 Some(ref mut n) => **n += 1,
667 fn finish_node(&mut self) {
668 self.inner.finish_node();
669 *self.roots.last_mut().unwrap() -= 1;
672 fn error(&mut self, error: ParseError) {
673 self.inner.error(error, self.text_pos)
680 use crate::tests::parse_macro;
681 use parser::TokenSource;
683 algo::{insert_children, InsertPosition},
688 fn convert_tt_token_source() {
689 let expansion = parse_macro(
691 macro_rules! literals {
703 .expand_tt("literals!(foo);");
704 let tts = &[expansion.into()];
705 let buffer = tt::buffer::TokenBuffer::from_tokens(tts);
706 let mut tt_src = SubtreeTokenSource::new(&buffer);
707 let mut tokens = vec![];
708 while tt_src.current().kind != EOF {
709 tokens.push((tt_src.current().kind, tt_src.text()));
714 // [let] [a] [=] ['c'] [;]
715 assert_eq!(tokens[2 + 3].1, "'c'");
716 assert_eq!(tokens[2 + 3].0, CHAR);
717 // [let] [c] [=] [1000] [;]
718 assert_eq!(tokens[2 + 5 + 3].1, "1000");
719 assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
720 // [let] [f] [=] [12E+99_f64] [;]
721 assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
722 assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
724 // [let] [s] [=] ["rust1"] [;]
725 assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
726 assert_eq!(tokens[2 + 15 + 3].0, STRING);
730 fn stmts_token_trees_to_expr_is_err() {
731 let expansion = parse_macro(
743 .expand_tt("stmts!();");
744 assert!(token_tree_to_syntax_node(&expansion, FragmentKind::Expr).is_err());
748 fn test_token_tree_last_child_is_white_space() {
749 let source_file = ast::SourceFile::parse("f!({} );").ok().unwrap();
750 let macro_call = source_file.syntax().descendants().find_map(ast::MacroCall::cast).unwrap();
751 let token_tree = macro_call.token_tree().unwrap();
753 // Token Tree now is :
763 token_tree.syntax().descendants_with_tokens().find(|it| it.kind() == T!['}']).unwrap();
764 let space = token_tree
766 .descendants_with_tokens()
767 .find(|it| it.kind() == SyntaxKind::WHITESPACE)
770 // reorder th white space, such that the white is inside the inner token-tree.
771 let token_tree = insert_children(
772 &rbrace.parent().unwrap(),
773 InsertPosition::Last,
774 std::iter::once(space),
777 // Token Tree now is :
782 let token_tree = ast::TokenTree::cast(token_tree).unwrap();
783 let tt = ast_to_token_tree(&token_tree).0;
785 assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
789 fn test_token_tree_multi_char_punct() {
790 let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
791 let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
792 let tt = ast_to_token_tree(&struct_def).0;
793 token_tree_to_syntax_node(&tt, FragmentKind::Item).unwrap();