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> {
74 let tokens = match tt {
75 tt::Subtree { delimiter: None, token_trees } => token_trees.as_slice(),
77 tmp = [tt.clone().into()];
81 let buffer = TokenBuffer::new(&tokens);
82 let mut token_source = SubtreeTokenSource::new(&buffer);
83 let mut tree_sink = TtTreeSink::new(buffer.begin());
84 parser::parse_fragment(&mut token_source, &mut tree_sink, fragment_kind);
85 if tree_sink.roots.len() != 1 {
86 return Err(ExpandError::ConversionError);
88 //FIXME: would be cool to report errors
89 let (parse, range_map) = tree_sink.finish();
90 Ok((parse, range_map))
93 /// Convert a string to a `TokenTree`
94 pub fn parse_to_token_tree(text: &str) -> Option<(tt::Subtree, TokenMap)> {
95 let (tokens, errors) = tokenize(text);
96 if !errors.is_empty() {
100 let mut conv = RawConvertor {
102 offset: TextSize::default(),
103 inner: tokens.iter(),
104 id_alloc: TokenIdAlloc {
105 map: Default::default(),
106 global_offset: TextSize::default(),
111 let subtree = conv.go()?;
112 Some((subtree, conv.id_alloc.map))
116 pub fn token_by_range(&self, relative_range: TextRange) -> Option<tt::TokenId> {
117 let &(token_id, _) = self.entries.iter().find(|(_, range)| match range {
118 TokenTextRange::Token(it) => *it == relative_range,
119 TokenTextRange::Delimiter(open, close) => {
120 *open == relative_range || *close == relative_range
126 pub fn range_by_token(&self, token_id: tt::TokenId) -> Option<TokenTextRange> {
127 let &(_, range) = self.entries.iter().find(|(tid, _)| *tid == token_id)?;
131 fn insert(&mut self, token_id: tt::TokenId, relative_range: TextRange) {
132 self.entries.push((token_id, TokenTextRange::Token(relative_range)));
137 token_id: tt::TokenId,
138 open_relative_range: TextRange,
139 close_relative_range: TextRange,
141 let res = self.entries.len();
143 .push((token_id, TokenTextRange::Delimiter(open_relative_range, close_relative_range)));
147 fn update_close_delim(&mut self, idx: usize, close_relative_range: TextRange) {
148 let (_, token_text_range) = &mut self.entries[idx];
149 if let TokenTextRange::Delimiter(dim, _) = token_text_range {
150 *token_text_range = TokenTextRange::Delimiter(*dim, close_relative_range);
154 fn remove_delim(&mut self, idx: usize) {
155 // FIXME: This could be accidently quadratic
156 self.entries.remove(idx);
160 /// Returns the textual content of a doc comment block as a quoted string
161 /// That is, strips leading `///` (or `/**`, etc)
162 /// and strips the ending `*/`
163 /// And then quote the string, which is needed to convert to `tt::Literal`
164 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
165 let prefix_len = comment.prefix().len();
166 let mut text = &comment.text()[prefix_len..];
168 // Remove ending "*/"
169 if comment.kind().shape == ast::CommentShape::Block {
170 text = &text[0..text.len() - 2];
174 // Note that `tt::Literal` expect an escaped string
175 let text = format!("{:?}", text.escape_default().to_string());
179 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
180 let comment = ast::Comment::cast(token.clone())?;
181 let doc = comment.kind().doc?;
183 // Make `doc="\" Comments\""
184 let mut meta_tkns = Vec::new();
185 meta_tkns.push(mk_ident("doc"));
186 meta_tkns.push(mk_punct('='));
187 meta_tkns.push(mk_doc_literal(&comment));
190 let mut token_trees = Vec::new();
191 token_trees.push(mk_punct('#'));
192 if let ast::CommentPlacement::Inner = doc {
193 token_trees.push(mk_punct('!'));
195 token_trees.push(tt::TokenTree::from(tt::Subtree {
196 delimiter: Some(tt::Delimiter {
197 kind: tt::DelimiterKind::Bracket,
198 id: tt::TokenId::unspecified(),
200 token_trees: meta_tkns,
203 return Some(token_trees);
206 fn mk_ident(s: &str) -> tt::TokenTree {
207 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
209 id: tt::TokenId::unspecified(),
213 fn mk_punct(c: char) -> tt::TokenTree {
214 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
216 spacing: tt::Spacing::Alone,
217 id: tt::TokenId::unspecified(),
221 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
222 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
224 tt::TokenTree::from(tt::Leaf::from(lit))
228 struct TokenIdAlloc {
230 global_offset: TextSize,
235 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
236 let relative_range = absolute_range - self.global_offset;
237 let token_id = tt::TokenId(self.next_id);
239 self.map.insert(token_id, relative_range);
243 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
244 let token_id = tt::TokenId(self.next_id);
246 let idx = self.map.insert_delim(
248 open_abs_range - self.global_offset,
249 open_abs_range - self.global_offset,
254 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
255 match close_abs_range {
257 self.map.remove_delim(idx);
260 self.map.update_close_delim(idx, close - self.global_offset);
266 /// A Raw Token (straightly from lexer) convertor
267 struct RawConvertor<'a> {
270 id_alloc: TokenIdAlloc,
271 inner: std::slice::Iter<'a, RawToken>,
274 trait SrcToken: std::fmt::Debug {
275 fn kind(&self) -> SyntaxKind;
277 fn to_char(&self) -> Option<char>;
279 fn to_text(&self) -> SmolStr;
282 trait TokenConvertor {
283 type Token: SrcToken;
285 fn go(&mut self) -> Option<tt::Subtree> {
286 let mut subtree = tt::Subtree::default();
287 subtree.delimiter = None;
288 while self.peek().is_some() {
289 self.collect_leaf(&mut subtree.token_trees);
291 if subtree.token_trees.is_empty() {
294 if subtree.token_trees.len() == 1 {
295 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
296 return Some(first.clone());
302 fn collect_leaf(&mut self, result: &mut Vec<tt::TokenTree>) {
303 let (token, range) = match self.bump() {
308 let k: SyntaxKind = token.kind();
310 if let Some(tokens) = self.convert_doc_comment(&token) {
311 result.extend(tokens);
316 result.push(if k.is_punct() {
317 assert_eq!(range.len(), TextSize::of('.'));
318 let delim = match k {
319 T!['('] => Some((tt::DelimiterKind::Parenthesis, T![')'])),
320 T!['{'] => Some((tt::DelimiterKind::Brace, T!['}'])),
321 T!['['] => Some((tt::DelimiterKind::Bracket, T![']'])),
325 if let Some((kind, closed)) = delim {
326 let mut subtree = tt::Subtree::default();
327 let (id, idx) = self.id_alloc().open_delim(range);
328 subtree.delimiter = Some(tt::Delimiter { kind, id });
330 while self.peek().map(|it| it.kind() != closed).unwrap_or(false) {
331 self.collect_leaf(&mut subtree.token_trees);
333 let last_range = match self.bump() {
335 // For error resilience, we insert an char punct for the opening delim here
336 self.id_alloc().close_delim(idx, None);
337 let leaf: tt::Leaf = tt::Punct {
338 id: self.id_alloc().alloc(range),
339 char: token.to_char().unwrap(),
340 spacing: tt::Spacing::Alone,
343 result.push(leaf.into());
344 result.extend(subtree.token_trees);
349 self.id_alloc().close_delim(idx, Some(last_range));
352 let spacing = match self.peek() {
354 if next.kind().is_trivia()
355 || next.kind() == T!['[']
356 || next.kind() == T!['{']
357 || next.kind() == T!['('] =>
361 Some(next) if next.kind().is_punct() => tt::Spacing::Joint,
362 _ => tt::Spacing::Alone,
364 let char = match token.to_char() {
367 panic!("Token from lexer must be single char: token = {:#?}", token);
370 tt::Leaf::from(tt::Punct { char, spacing, id: self.id_alloc().alloc(range) }).into()
373 macro_rules! make_leaf {
375 tt::$i { id: self.id_alloc().alloc(range), text: token.to_text() }.into()
378 let leaf: tt::Leaf = match k {
379 T![true] | T![false] => make_leaf!(Ident),
380 IDENT => make_leaf!(Ident),
381 k if k.is_keyword() => make_leaf!(Ident),
382 k if k.is_literal() => make_leaf!(Literal),
384 let char_unit = TextSize::of('\'');
385 let r = TextRange::at(range.start(), char_unit);
386 let apostrophe = tt::Leaf::from(tt::Punct {
388 spacing: tt::Spacing::Joint,
389 id: self.id_alloc().alloc(r),
391 result.push(apostrophe.into());
393 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
394 let ident = tt::Leaf::from(tt::Ident {
395 text: SmolStr::new(&token.to_text()[1..]),
396 id: self.id_alloc().alloc(r),
398 result.push(ident.into());
408 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
410 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
412 fn peek(&self) -> Option<Self::Token>;
414 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
417 impl<'a> SrcToken for (RawToken, &'a str) {
418 fn kind(&self) -> SyntaxKind {
422 fn to_char(&self) -> Option<char> {
423 self.1.chars().next()
426 fn to_text(&self) -> SmolStr {
431 impl RawConvertor<'_> {}
433 impl<'a> TokenConvertor for RawConvertor<'a> {
434 type Token = (RawToken, &'a str);
436 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
437 convert_doc_comment(&doc_comment(token.1))
440 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
441 let token = self.inner.next()?;
442 let range = TextRange::at(self.offset, token.len);
443 self.offset += token.len;
445 Some(((*token, &self.text[range]), range))
448 fn peek(&self) -> Option<Self::Token> {
449 let token = self.inner.as_slice().get(0).cloned();
452 let range = TextRange::at(self.offset, it.len);
453 (it, &self.text[range])
457 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
463 id_alloc: TokenIdAlloc,
464 current: Option<SyntaxToken>,
466 punct_offset: Option<(SyntaxToken, TextSize)>,
470 fn new(node: &SyntaxNode, global_offset: TextSize) -> Convertor {
472 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
473 current: node.first_token(),
474 range: node.text_range(),
482 Ordiniary(SyntaxToken),
483 Punch(SyntaxToken, TextSize),
487 fn token(&self) -> &SyntaxToken {
489 SynToken::Ordiniary(it) => it,
490 SynToken::Punch(it, _) => it,
495 impl SrcToken for SynToken {
496 fn kind(&self) -> SyntaxKind {
499 fn to_char(&self) -> Option<char> {
501 SynToken::Ordiniary(_) => None,
502 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
505 fn to_text(&self) -> SmolStr {
506 self.token().text().clone()
510 impl TokenConvertor for Convertor {
511 type Token = SynToken;
512 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
513 convert_doc_comment(token.token())
516 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
517 if let Some((punct, offset)) = self.punct_offset.clone() {
518 if usize::from(offset) + 1 < punct.text().len() {
519 let offset = offset + TextSize::of('.');
520 let range = punct.text_range();
521 self.punct_offset = Some((punct.clone(), offset));
522 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
523 return Some((SynToken::Punch(punct, offset), range));
527 let curr = self.current.clone()?;
528 if !&self.range.contains_range(curr.text_range()) {
531 self.current = curr.next_token();
533 let token = if curr.kind().is_punct() {
534 let range = curr.text_range();
535 let range = TextRange::at(range.start(), TextSize::of('.'));
536 self.punct_offset = Some((curr.clone(), 0.into()));
537 (SynToken::Punch(curr, 0.into()), range)
539 self.punct_offset = None;
540 let range = curr.text_range();
541 (SynToken::Ordiniary(curr), range)
547 fn peek(&self) -> Option<Self::Token> {
548 if let Some((punct, mut offset)) = self.punct_offset.clone() {
549 offset = offset + TextSize::of('.');
550 if usize::from(offset) < punct.text().len() {
551 return Some(SynToken::Punch(punct, offset));
555 let curr = self.current.clone()?;
556 if !self.range.contains_range(curr.text_range()) {
560 let token = if curr.kind().is_punct() {
561 SynToken::Punch(curr, 0.into())
563 SynToken::Ordiniary(curr)
568 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
573 struct TtTreeSink<'a> {
576 open_delims: FxHashMap<tt::TokenId, TextSize>,
578 inner: SyntaxTreeBuilder,
582 // Use for detect ill-form tree which is not single root
583 roots: smallvec::SmallVec<[usize; 1]>,
586 impl<'a> TtTreeSink<'a> {
587 fn new(cursor: Cursor<'a>) -> Self {
591 open_delims: FxHashMap::default(),
593 inner: SyntaxTreeBuilder::default(),
594 roots: smallvec::SmallVec::new(),
595 token_map: TokenMap::default(),
599 fn finish(self) -> (Parse<SyntaxNode>, TokenMap) {
600 (self.inner.finish(), self.token_map)
604 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> SmolStr {
605 let texts = match d {
606 Some(tt::DelimiterKind::Parenthesis) => "()",
607 Some(tt::DelimiterKind::Brace) => "{}",
608 Some(tt::DelimiterKind::Bracket) => "[]",
609 None => return "".into(),
612 let idx = closing as usize;
613 let text = &texts[idx..texts.len() - (1 - idx)];
617 impl<'a> TreeSink for TtTreeSink<'a> {
618 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
619 if kind == L_DOLLAR || kind == R_DOLLAR {
620 self.cursor = self.cursor.bump_subtree();
623 if kind == LIFETIME_IDENT {
627 let mut last = self.cursor;
628 for _ in 0..n_tokens {
629 if self.cursor.eof() {
633 let text: SmolStr = match self.cursor.token_tree() {
634 Some(tt::TokenTree::Leaf(leaf)) => {
635 // Mark the range if needed
636 let (text, id) = match leaf {
637 tt::Leaf::Ident(ident) => (ident.text.clone(), ident.id),
638 tt::Leaf::Punct(punct) => {
639 assert!(punct.char.is_ascii());
640 let char = &(punct.char as u8);
641 let text = std::str::from_utf8(std::slice::from_ref(char)).unwrap();
642 (SmolStr::new_inline(text), punct.id)
644 tt::Leaf::Literal(lit) => (lit.text.clone(), lit.id),
646 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
647 self.token_map.insert(id, range);
648 self.cursor = self.cursor.bump();
651 Some(tt::TokenTree::Subtree(subtree)) => {
652 self.cursor = self.cursor.subtree().unwrap();
653 if let Some(id) = subtree.delimiter.map(|it| it.id) {
654 self.open_delims.insert(id, self.text_pos);
656 delim_to_str(subtree.delimiter_kind(), false)
659 if let Some(parent) = self.cursor.end() {
660 self.cursor = self.cursor.bump();
661 if let Some(id) = parent.delimiter.map(|it| it.id) {
662 if let Some(open_delim) = self.open_delims.get(&id) {
663 let open_range = TextRange::at(*open_delim, TextSize::of('('));
664 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
665 self.token_map.insert_delim(id, open_range, close_range);
668 delim_to_str(parent.delimiter_kind(), true)
675 self.text_pos += TextSize::of(text.as_str());
678 let text = SmolStr::new(self.buf.as_str());
680 self.inner.token(kind, text);
682 // Add whitespace between adjoint puncts
683 let next = last.bump();
685 Some(tt::TokenTree::Leaf(tt::Leaf::Punct(curr))),
686 Some(tt::TokenTree::Leaf(tt::Leaf::Punct(_))),
687 ) = (last.token_tree(), next.token_tree())
689 // Note: We always assume the semi-colon would be the last token in
690 // other parts of RA such that we don't add whitespace here.
691 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
692 self.inner.token(WHITESPACE, " ".into());
693 self.text_pos += TextSize::of(' ');
698 fn start_node(&mut self, kind: SyntaxKind) {
699 self.inner.start_node(kind);
701 match self.roots.last_mut() {
702 None | Some(0) => self.roots.push(1),
703 Some(ref mut n) => **n += 1,
707 fn finish_node(&mut self) {
708 self.inner.finish_node();
709 *self.roots.last_mut().unwrap() -= 1;
712 fn error(&mut self, error: ParseError) {
713 self.inner.error(error, self.text_pos)
720 use crate::tests::parse_macro;
721 use parser::TokenSource;
723 algo::{insert_children, InsertPosition},
728 fn convert_tt_token_source() {
729 let expansion = parse_macro(
731 macro_rules! literals {
743 .expand_tt("literals!(foo);");
744 let tts = &[expansion.into()];
745 let buffer = tt::buffer::TokenBuffer::new(tts);
746 let mut tt_src = SubtreeTokenSource::new(&buffer);
747 let mut tokens = vec![];
748 while tt_src.current().kind != EOF {
749 tokens.push((tt_src.current().kind, tt_src.text()));
754 // [let] [a] [=] ['c'] [;]
755 assert_eq!(tokens[2 + 3].1, "'c'");
756 assert_eq!(tokens[2 + 3].0, CHAR);
757 // [let] [c] [=] [1000] [;]
758 assert_eq!(tokens[2 + 5 + 3].1, "1000");
759 assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
760 // [let] [f] [=] [12E+99_f64] [;]
761 assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
762 assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
764 // [let] [s] [=] ["rust1"] [;]
765 assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
766 assert_eq!(tokens[2 + 15 + 3].0, STRING);
770 fn stmts_token_trees_to_expr_is_err() {
771 let expansion = parse_macro(
783 .expand_tt("stmts!();");
784 assert!(token_tree_to_syntax_node(&expansion, FragmentKind::Expr).is_err());
788 fn test_token_tree_last_child_is_white_space() {
789 let source_file = ast::SourceFile::parse("f!({} );").ok().unwrap();
790 let macro_call = source_file.syntax().descendants().find_map(ast::MacroCall::cast).unwrap();
791 let token_tree = macro_call.token_tree().unwrap();
793 // Token Tree now is :
803 token_tree.syntax().descendants_with_tokens().find(|it| it.kind() == T!['}']).unwrap();
804 let space = token_tree
806 .descendants_with_tokens()
807 .find(|it| it.kind() == SyntaxKind::WHITESPACE)
810 // reorder th white space, such that the white is inside the inner token-tree.
811 let token_tree = insert_children(
812 &rbrace.parent().unwrap(),
813 InsertPosition::Last,
814 std::iter::once(space),
817 // Token Tree now is :
822 let token_tree = ast::TokenTree::cast(token_tree).unwrap();
823 let tt = ast_to_token_tree(&token_tree).unwrap().0;
825 assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
829 fn test_token_tree_multi_char_punct() {
830 let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
831 let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
832 let tt = ast_to_token_tree(&struct_def).unwrap().0;
833 token_tree_to_syntax_node(&tt, FragmentKind::Item).unwrap();