1 //! FIXME: write short doc here
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;
14 use crate::ExpandError;
16 #[derive(Debug, PartialEq, Eq, Clone, Copy)]
17 pub enum TokenTextRange {
19 Delimiter(TextRange, TextRange),
23 pub fn by_kind(self, kind: SyntaxKind) -> Option<TextRange> {
25 TokenTextRange::Token(it) => Some(it),
26 TokenTextRange::Delimiter(open, close) => match kind {
27 T!['{'] | T!['('] | T!['['] => Some(open),
28 T!['}'] | T![')'] | T![']'] => Some(close),
35 /// Maps `tt::TokenId` to the relative range of the original token.
36 #[derive(Debug, PartialEq, Eq, Clone, Default)]
38 /// Maps `tt::TokenId` to the *relative* source range.
39 entries: Vec<(tt::TokenId, TokenTextRange)>,
42 /// Convert the syntax tree (what user has written) to a `TokenTree` (what macro
44 pub fn ast_to_token_tree(ast: &impl ast::AstNode) -> Option<(tt::Subtree, TokenMap)> {
45 syntax_node_to_token_tree(ast.syntax())
48 /// Convert the syntax node to a `TokenTree` (what macro
50 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> Option<(tt::Subtree, TokenMap)> {
51 let global_offset = node.text_range().start();
52 let mut c = Convertor::new(node, global_offset);
53 let subtree = c.go()?;
54 Some((subtree, c.id_alloc.map))
57 // The following items are what `rustc` macro can be parsed into :
58 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
59 // * Expr(P<ast::Expr>) -> token_tree_to_expr
60 // * Pat(P<ast::Pat>) -> token_tree_to_pat
61 // * Ty(P<ast::Ty>) -> token_tree_to_ty
62 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
63 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
65 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
66 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
67 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
69 pub fn token_tree_to_syntax_node(
71 fragment_kind: FragmentKind,
72 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
73 let buffer = match tt {
74 tt::Subtree { delimiter: None, token_trees } => {
75 TokenBuffer::from_tokens(token_trees.as_slice())
77 _ => TokenBuffer::from_subtree(tt),
79 let mut token_source = SubtreeTokenSource::new(&buffer);
80 let mut tree_sink = TtTreeSink::new(buffer.begin());
81 parser::parse_fragment(&mut token_source, &mut tree_sink, fragment_kind);
82 if tree_sink.roots.len() != 1 {
83 return Err(ExpandError::ConversionError);
85 //FIXME: would be cool to report errors
86 let (parse, range_map) = tree_sink.finish();
87 Ok((parse, range_map))
90 /// Convert a string to a `TokenTree`
91 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
92 let (tokens, errors) = tokenize(text);
93 if !errors.is_empty() {
97 let mut conv = RawConvertor {
99 offset: TextSize::default(),
100 inner: tokens.iter(),
101 id_alloc: TokenIdAlloc {
102 map: Default::default(),
103 global_offset: TextSize::default(),
108 let subtree = conv.go()?;
109 Some((subtree, conv.id_alloc.map))
113 pub fn token_by_range(&self, relative_range: TextRange) -> Option<tt::TokenId> {
114 let &(token_id, _) = self.entries.iter().find(|(_, range)| match range {
115 TokenTextRange::Token(it) => *it == relative_range,
116 TokenTextRange::Delimiter(open, close) => {
117 *open == relative_range || *close == relative_range
123 pub fn range_by_token(&self, token_id: tt::TokenId) -> Option<TokenTextRange> {
124 let &(_, range) = self.entries.iter().find(|(tid, _)| *tid == token_id)?;
128 fn insert(&mut self, token_id: tt::TokenId, relative_range: TextRange) {
129 self.entries.push((token_id, TokenTextRange::Token(relative_range)));
134 token_id: tt::TokenId,
135 open_relative_range: TextRange,
136 close_relative_range: TextRange,
138 let res = self.entries.len();
140 .push((token_id, TokenTextRange::Delimiter(open_relative_range, close_relative_range)));
144 fn update_close_delim(&mut self, idx: usize, close_relative_range: TextRange) {
145 let (_, token_text_range) = &mut self.entries[idx];
146 if let TokenTextRange::Delimiter(dim, _) = token_text_range {
147 *token_text_range = TokenTextRange::Delimiter(*dim, close_relative_range);
151 fn remove_delim(&mut self, idx: usize) {
152 // FIXME: This could be accidently quadratic
153 self.entries.remove(idx);
157 /// Returns the textual content of a doc comment block as a quoted string
158 /// That is, strips leading `///` (or `/**`, etc)
159 /// and strips the ending `*/`
160 /// And then quote the string, which is needed to convert to `tt::Literal`
161 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
162 let prefix_len = comment.prefix().len();
163 let mut text = &comment.text()[prefix_len..];
165 // Remove ending "*/"
166 if comment.kind().shape == ast::CommentShape::Block {
167 text = &text[0..text.len() - 2];
171 // Note that `tt::Literal` expect an escaped string
172 let text = format!("{:?}", text.escape_default().to_string());
176 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
177 let comment = ast::Comment::cast(token.clone())?;
178 let doc = comment.kind().doc?;
180 // Make `doc="\" Comments\""
181 let mut meta_tkns = Vec::new();
182 meta_tkns.push(mk_ident("doc"));
183 meta_tkns.push(mk_punct('='));
184 meta_tkns.push(mk_doc_literal(&comment));
187 let mut token_trees = Vec::new();
188 token_trees.push(mk_punct('#'));
189 if let ast::CommentPlacement::Inner = doc {
190 token_trees.push(mk_punct('!'));
192 token_trees.push(tt::TokenTree::from(tt::Subtree {
193 delimiter: Some(tt::Delimiter {
194 kind: tt::DelimiterKind::Bracket,
195 id: tt::TokenId::unspecified(),
197 token_trees: meta_tkns,
200 return Some(token_trees);
203 fn mk_ident(s: &str) -> tt::TokenTree {
204 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
206 id: tt::TokenId::unspecified(),
210 fn mk_punct(c: char) -> tt::TokenTree {
211 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
213 spacing: tt::Spacing::Alone,
214 id: tt::TokenId::unspecified(),
218 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
219 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
221 tt::TokenTree::from(tt::Leaf::from(lit))
225 struct TokenIdAlloc {
227 global_offset: TextSize,
232 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
233 let relative_range = absolute_range - self.global_offset;
234 let token_id = tt::TokenId(self.next_id);
236 self.map.insert(token_id, relative_range);
240 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
241 let token_id = tt::TokenId(self.next_id);
243 let idx = self.map.insert_delim(
245 open_abs_range - self.global_offset,
246 open_abs_range - self.global_offset,
251 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
252 match close_abs_range {
254 self.map.remove_delim(idx);
257 self.map.update_close_delim(idx, close - self.global_offset);
263 /// A Raw Token (straightly from lexer) convertor
264 struct RawConvertor<'a> {
267 id_alloc: TokenIdAlloc,
268 inner: std::slice::Iter<'a, RawToken>,
271 trait SrcToken: std::fmt::Debug {
272 fn kind(&self) -> SyntaxKind;
274 fn to_char(&self) -> Option<char>;
276 fn to_text(&self) -> SmolStr;
279 trait TokenConvertor {
280 type Token: SrcToken;
282 fn go(&mut self) -> Option<tt::Subtree> {
283 let mut subtree = tt::Subtree::default();
284 subtree.delimiter = None;
285 while self.peek().is_some() {
286 self.collect_leaf(&mut subtree.token_trees);
288 if subtree.token_trees.is_empty() {
291 if subtree.token_trees.len() == 1 {
292 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
293 return Some(first.clone());
299 fn collect_leaf(&mut self, result: &mut Vec<tt::TokenTree>) {
300 let (token, range) = match self.bump() {
305 let k: SyntaxKind = token.kind();
307 if let Some(tokens) = self.convert_doc_comment(&token) {
308 result.extend(tokens);
313 result.push(if k.is_punct() {
314 assert_eq!(range.len(), TextSize::of('.'));
315 let delim = match k {
316 T!['('] => Some((tt::DelimiterKind::Parenthesis, T![')'])),
317 T!['{'] => Some((tt::DelimiterKind::Brace, T!['}'])),
318 T!['['] => Some((tt::DelimiterKind::Bracket, T![']'])),
322 if let Some((kind, closed)) = delim {
323 let mut subtree = tt::Subtree::default();
324 let (id, idx) = self.id_alloc().open_delim(range);
325 subtree.delimiter = Some(tt::Delimiter { kind, id });
327 while self.peek().map(|it| it.kind() != closed).unwrap_or(false) {
328 self.collect_leaf(&mut subtree.token_trees);
330 let last_range = match self.bump() {
332 // For error resilience, we insert an char punct for the opening delim here
333 self.id_alloc().close_delim(idx, None);
334 let leaf: tt::Leaf = tt::Punct {
335 id: self.id_alloc().alloc(range),
336 char: token.to_char().unwrap(),
337 spacing: tt::Spacing::Alone,
340 result.push(leaf.into());
341 result.extend(subtree.token_trees);
346 self.id_alloc().close_delim(idx, Some(last_range));
349 let spacing = match self.peek() {
351 if next.kind().is_trivia()
352 || next.kind() == T!['[']
353 || next.kind() == T!['{']
354 || next.kind() == T!['('] =>
358 Some(next) if next.kind().is_punct() => tt::Spacing::Joint,
359 _ => tt::Spacing::Alone,
361 let char = match token.to_char() {
364 panic!("Token from lexer must be single char: token = {:#?}", token);
367 tt::Leaf::from(tt::Punct { char, spacing, id: self.id_alloc().alloc(range) }).into()
370 macro_rules! make_leaf {
372 tt::$i { id: self.id_alloc().alloc(range), text: token.to_text() }.into()
375 let leaf: tt::Leaf = match k {
376 T![true] | T![false] => make_leaf!(Ident),
377 IDENT => make_leaf!(Ident),
378 k if k.is_keyword() => make_leaf!(Ident),
379 k if k.is_literal() => make_leaf!(Literal),
381 let char_unit = TextSize::of('\'');
382 let r = TextRange::at(range.start(), char_unit);
383 let apostrophe = tt::Leaf::from(tt::Punct {
385 spacing: tt::Spacing::Joint,
386 id: self.id_alloc().alloc(r),
388 result.push(apostrophe.into());
390 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
391 let ident = tt::Leaf::from(tt::Ident {
392 text: SmolStr::new(&token.to_text()[1..]),
393 id: self.id_alloc().alloc(r),
395 result.push(ident.into());
405 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
407 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
409 fn peek(&self) -> Option<Self::Token>;
411 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
414 impl<'a> SrcToken for (&'a RawToken, &'a str) {
415 fn kind(&self) -> SyntaxKind {
419 fn to_char(&self) -> Option<char> {
420 self.1.chars().next()
423 fn to_text(&self) -> SmolStr {
428 impl RawConvertor<'_> {}
430 impl<'a> TokenConvertor for RawConvertor<'a> {
431 type Token = (&'a RawToken, &'a str);
433 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
434 convert_doc_comment(&doc_comment(token.1))
437 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
438 let token = self.inner.next()?;
439 let range = TextRange::at(self.offset, token.len);
440 self.offset += token.len;
442 Some(((token, &self.text[range]), range))
445 fn peek(&self) -> Option<Self::Token> {
446 let token = self.inner.as_slice().get(0);
449 let range = TextRange::at(self.offset, it.len);
450 (it, &self.text[range])
454 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
460 id_alloc: TokenIdAlloc,
461 current: Option<SyntaxToken>,
463 punct_offset: Option<(SyntaxToken, TextSize)>,
467 fn new(node: &SyntaxNode, global_offset: TextSize) -> Convertor {
469 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
470 current: node.first_token(),
471 range: node.text_range(),
479 Ordiniary(SyntaxToken),
480 Punch(SyntaxToken, TextSize),
484 fn token(&self) -> &SyntaxToken {
486 SynToken::Ordiniary(it) => it,
487 SynToken::Punch(it, _) => it,
492 impl SrcToken for SynToken {
493 fn kind(&self) -> SyntaxKind {
496 fn to_char(&self) -> Option<char> {
498 SynToken::Ordiniary(_) => None,
499 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
502 fn to_text(&self) -> SmolStr {
503 self.token().text().clone()
507 impl TokenConvertor for Convertor {
508 type Token = SynToken;
509 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
510 convert_doc_comment(token.token())
513 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
514 if let Some((punct, offset)) = self.punct_offset.clone() {
515 if usize::from(offset) + 1 < punct.text().len() {
516 let offset = offset + TextSize::of('.');
517 let range = punct.text_range();
518 self.punct_offset = Some((punct.clone(), offset));
519 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
520 return Some((SynToken::Punch(punct, offset), range));
524 let curr = self.current.clone()?;
525 if !&self.range.contains_range(curr.text_range()) {
528 self.current = curr.next_token();
530 let token = if curr.kind().is_punct() {
531 let range = curr.text_range();
532 let range = TextRange::at(range.start(), TextSize::of('.'));
533 self.punct_offset = Some((curr.clone(), 0.into()));
534 (SynToken::Punch(curr, 0.into()), range)
536 self.punct_offset = None;
537 let range = curr.text_range();
538 (SynToken::Ordiniary(curr), range)
544 fn peek(&self) -> Option<Self::Token> {
545 if let Some((punct, mut offset)) = self.punct_offset.clone() {
546 offset = offset + TextSize::of('.');
547 if usize::from(offset) < punct.text().len() {
548 return Some(SynToken::Punch(punct, offset));
552 let curr = self.current.clone()?;
553 if !self.range.contains_range(curr.text_range()) {
557 let token = if curr.kind().is_punct() {
558 SynToken::Punch(curr, 0.into())
560 SynToken::Ordiniary(curr)
565 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
570 struct TtTreeSink<'a> {
573 open_delims: FxHashMap<tt::TokenId, TextSize>,
575 inner: SyntaxTreeBuilder,
579 // Use for detect ill-form tree which is not single root
580 roots: smallvec::SmallVec<[usize; 1]>,
583 impl<'a> TtTreeSink<'a> {
584 fn new(cursor: Cursor<'a>) -> Self {
588 open_delims: FxHashMap::default(),
590 inner: SyntaxTreeBuilder::default(),
591 roots: smallvec::SmallVec::new(),
592 token_map: TokenMap::default(),
596 fn finish(self) -> (Parse<SyntaxNode>, TokenMap) {
597 (self.inner.finish(), self.token_map)
601 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> &'static str {
602 let texts = match d {
603 Some(tt::DelimiterKind::Parenthesis) => "()",
604 Some(tt::DelimiterKind::Brace) => "{}",
605 Some(tt::DelimiterKind::Bracket) => "[]",
609 let idx = closing as usize;
610 &texts[idx..texts.len() - (1 - idx)]
613 impl<'a> TreeSink for TtTreeSink<'a> {
614 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
615 if kind == L_DOLLAR || kind == R_DOLLAR {
616 self.cursor = self.cursor.bump_subtree();
619 if kind == LIFETIME_IDENT {
623 let mut last = self.cursor;
624 for _ in 0..n_tokens {
625 let tmp_str: SmolStr;
626 if self.cursor.eof() {
630 let text: &str = match self.cursor.token_tree() {
631 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
632 // Mark the range if needed
633 let (text, id) = match leaf {
634 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
635 tt::Leaf::Punct(punct) => {
636 assert!(punct.char.is_ascii());
637 let char = &(punct.char as u8);
638 tmp_str = SmolStr::new_inline(
639 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
643 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
645 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
646 self.token_map.insert(id, range);
647 self.cursor = self.cursor.bump();
650 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
651 self.cursor = self.cursor.subtree().unwrap();
652 if let Some(id) = subtree.delimiter.map(|it| it.id) {
653 self.open_delims.insert(id, self.text_pos);
655 delim_to_str(subtree.delimiter_kind(), false)
658 if let Some(parent) = self.cursor.end() {
659 self.cursor = self.cursor.bump();
660 if let Some(id) = parent.delimiter.map(|it| it.id) {
661 if let Some(open_delim) = self.open_delims.get(&id) {
662 let open_range = TextRange::at(*open_delim, TextSize::of('('));
663 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
664 self.token_map.insert_delim(id, open_range, close_range);
667 delim_to_str(parent.delimiter_kind(), true)
674 self.text_pos += TextSize::of(text);
677 let text = SmolStr::new(self.buf.as_str());
679 self.inner.token(kind, text);
681 // Add whitespace between adjoint puncts
682 let next = last.bump();
684 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
685 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
686 ) = (last.token_tree(), next.token_tree())
688 // Note: We always assume the semi-colon would be the last token in
689 // other parts of RA such that we don't add whitespace here.
690 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
691 self.inner.token(WHITESPACE, " ".into());
692 self.text_pos += TextSize::of(' ');
697 fn start_node(&mut self, kind: SyntaxKind) {
698 self.inner.start_node(kind);
700 match self.roots.last_mut() {
701 None | Some(0) => self.roots.push(1),
702 Some(ref mut n) => **n += 1,
706 fn finish_node(&mut self) {
707 self.inner.finish_node();
708 *self.roots.last_mut().unwrap() -= 1;
711 fn error(&mut self, error: ParseError) {
712 self.inner.error(error, self.text_pos)
719 use crate::tests::parse_macro;
720 use parser::TokenSource;
722 algo::{insert_children, InsertPosition},
727 fn convert_tt_token_source() {
728 let expansion = parse_macro(
730 macro_rules! literals {
742 .expand_tt("literals!(foo);");
743 let tts = &[expansion.into()];
744 let buffer = tt::buffer::TokenBuffer::from_tokens(tts);
745 let mut tt_src = SubtreeTokenSource::new(&buffer);
746 let mut tokens = vec![];
747 while tt_src.current().kind != EOF {
748 tokens.push((tt_src.current().kind, tt_src.text()));
753 // [let] [a] [=] ['c'] [;]
754 assert_eq!(tokens[2 + 3].1, "'c'");
755 assert_eq!(tokens[2 + 3].0, CHAR);
756 // [let] [c] [=] [1000] [;]
757 assert_eq!(tokens[2 + 5 + 3].1, "1000");
758 assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
759 // [let] [f] [=] [12E+99_f64] [;]
760 assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
761 assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
763 // [let] [s] [=] ["rust1"] [;]
764 assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
765 assert_eq!(tokens[2 + 15 + 3].0, STRING);
769 fn stmts_token_trees_to_expr_is_err() {
770 let expansion = parse_macro(
782 .expand_tt("stmts!();");
783 assert!(token_tree_to_syntax_node(&expansion, FragmentKind::Expr).is_err());
787 fn test_token_tree_last_child_is_white_space() {
788 let source_file = ast::SourceFile::parse("f!({} );").ok().unwrap();
789 let macro_call = source_file.syntax().descendants().find_map(ast::MacroCall::cast).unwrap();
790 let token_tree = macro_call.token_tree().unwrap();
792 // Token Tree now is :
802 token_tree.syntax().descendants_with_tokens().find(|it| it.kind() == T!['}']).unwrap();
803 let space = token_tree
805 .descendants_with_tokens()
806 .find(|it| it.kind() == SyntaxKind::WHITESPACE)
809 // reorder th white space, such that the white is inside the inner token-tree.
810 let token_tree = insert_children(
811 &rbrace.parent().unwrap(),
812 InsertPosition::Last,
813 std::iter::once(space),
816 // Token Tree now is :
821 let token_tree = ast::TokenTree::cast(token_tree).unwrap();
822 let tt = ast_to_token_tree(&token_tree).unwrap().0;
824 assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
828 fn test_token_tree_multi_char_punct() {
829 let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
830 let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
831 let tt = ast_to_token_tree(&struct_def).unwrap().0;
832 token_tree_to_syntax_node(&tt, FragmentKind::Item).unwrap();