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 {
23 pub fn by_kind(self, kind: SyntaxKind) -> Option<TextRange> {
25 TokenTextRange::Token(it) => Some(it),
26 TokenTextRange::Delimiter(it) => match kind {
27 T!['{'] | T!['('] | T!['['] => Some(TextRange::at(it.start(), 1.into())),
28 T!['}'] | T![')'] | T![']'] => {
29 Some(TextRange::at(it.end() - TextSize::of('}'), 1.into()))
37 /// Maps `tt::TokenId` to the relative range of the original token.
38 #[derive(Debug, PartialEq, Eq, Clone, Default)]
40 /// Maps `tt::TokenId` to the *relative* source range.
41 entries: Vec<(tt::TokenId, TokenTextRange)>,
44 /// Convert the syntax tree (what user has written) to a `TokenTree` (what macro
46 pub fn ast_to_token_tree(ast: &impl ast::AstNode) -> Option<(tt::Subtree, TokenMap)> {
47 syntax_node_to_token_tree(ast.syntax())
50 /// Convert the syntax node to a `TokenTree` (what macro
52 pub fn syntax_node_to_token_tree(node: &SyntaxNode) -> Option<(tt::Subtree, TokenMap)> {
53 let global_offset = node.text_range().start();
54 let mut c = Convertor::new(node, global_offset);
55 let subtree = c.go()?;
56 c.id_alloc.map.entries.shrink_to_fit();
57 Some((subtree, c.id_alloc.map))
60 // The following items are what `rustc` macro can be parsed into :
61 // link: https://github.com/rust-lang/rust/blob/9ebf47851a357faa4cd97f4b1dc7835f6376e639/src/libsyntax/ext/expand.rs#L141
62 // * Expr(P<ast::Expr>) -> token_tree_to_expr
63 // * Pat(P<ast::Pat>) -> token_tree_to_pat
64 // * Ty(P<ast::Ty>) -> token_tree_to_ty
65 // * Stmts(SmallVec<[ast::Stmt; 1]>) -> token_tree_to_stmts
66 // * Items(SmallVec<[P<ast::Item>; 1]>) -> token_tree_to_items
68 // * TraitItems(SmallVec<[ast::TraitItem; 1]>)
69 // * AssocItems(SmallVec<[ast::AssocItem; 1]>)
70 // * ForeignItems(SmallVec<[ast::ForeignItem; 1]>
72 pub fn token_tree_to_syntax_node(
74 fragment_kind: FragmentKind,
75 ) -> Result<(Parse<SyntaxNode>, TokenMap), ExpandError> {
76 let buffer = match tt {
77 tt::Subtree { delimiter: None, token_trees } => {
78 TokenBuffer::from_tokens(token_trees.as_slice())
80 _ => TokenBuffer::from_subtree(tt),
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(it) => {
120 let open = TextRange::at(it.start(), 1.into());
121 let close = TextRange::at(it.end() - TextSize::of('}'), 1.into());
122 open == relative_range || close == relative_range
128 pub fn range_by_token(&self, token_id: tt::TokenId) -> Option<TokenTextRange> {
129 let &(_, range) = self.entries.iter().find(|(tid, _)| *tid == token_id)?;
133 fn insert(&mut self, token_id: tt::TokenId, relative_range: TextRange) {
134 self.entries.push((token_id, TokenTextRange::Token(relative_range)));
139 token_id: tt::TokenId,
140 open_relative_range: TextRange,
141 close_relative_range: TextRange,
143 let res = self.entries.len();
144 let cover = open_relative_range.cover(close_relative_range);
146 self.entries.push((token_id, TokenTextRange::Delimiter(cover)));
150 fn update_close_delim(&mut self, idx: usize, close_relative_range: TextRange) {
151 let (_, token_text_range) = &mut self.entries[idx];
152 if let TokenTextRange::Delimiter(dim) = token_text_range {
153 let cover = dim.cover(close_relative_range);
154 *token_text_range = TokenTextRange::Delimiter(cover);
158 fn remove_delim(&mut self, idx: usize) {
159 // FIXME: This could be accidentally quadratic
160 self.entries.remove(idx);
164 /// Returns the textual content of a doc comment block as a quoted string
165 /// That is, strips leading `///` (or `/**`, etc)
166 /// and strips the ending `*/`
167 /// And then quote the string, which is needed to convert to `tt::Literal`
168 fn doc_comment_text(comment: &ast::Comment) -> SmolStr {
169 let prefix_len = comment.prefix().len();
170 let mut text = &comment.text()[prefix_len..];
172 // Remove ending "*/"
173 if comment.kind().shape == ast::CommentShape::Block {
174 text = &text[0..text.len() - 2];
178 // Note that `tt::Literal` expect an escaped string
179 let text = format!("{:?}", text.escape_default().to_string());
183 fn convert_doc_comment(token: &syntax::SyntaxToken) -> Option<Vec<tt::TokenTree>> {
184 let comment = ast::Comment::cast(token.clone())?;
185 let doc = comment.kind().doc?;
187 // Make `doc="\" Comments\""
188 let mut meta_tkns = Vec::new();
189 meta_tkns.push(mk_ident("doc"));
190 meta_tkns.push(mk_punct('='));
191 meta_tkns.push(mk_doc_literal(&comment));
194 let mut token_trees = Vec::new();
195 token_trees.push(mk_punct('#'));
196 if let ast::CommentPlacement::Inner = doc {
197 token_trees.push(mk_punct('!'));
199 token_trees.push(tt::TokenTree::from(tt::Subtree {
200 delimiter: Some(tt::Delimiter {
201 kind: tt::DelimiterKind::Bracket,
202 id: tt::TokenId::unspecified(),
204 token_trees: meta_tkns,
207 return Some(token_trees);
210 fn mk_ident(s: &str) -> tt::TokenTree {
211 tt::TokenTree::from(tt::Leaf::from(tt::Ident {
213 id: tt::TokenId::unspecified(),
217 fn mk_punct(c: char) -> tt::TokenTree {
218 tt::TokenTree::from(tt::Leaf::from(tt::Punct {
220 spacing: tt::Spacing::Alone,
221 id: tt::TokenId::unspecified(),
225 fn mk_doc_literal(comment: &ast::Comment) -> tt::TokenTree {
226 let lit = tt::Literal { text: doc_comment_text(comment), id: tt::TokenId::unspecified() };
228 tt::TokenTree::from(tt::Leaf::from(lit))
232 struct TokenIdAlloc {
234 global_offset: TextSize,
239 fn alloc(&mut self, absolute_range: TextRange) -> tt::TokenId {
240 let relative_range = absolute_range - self.global_offset;
241 let token_id = tt::TokenId(self.next_id);
243 self.map.insert(token_id, relative_range);
247 fn open_delim(&mut self, open_abs_range: TextRange) -> (tt::TokenId, usize) {
248 let token_id = tt::TokenId(self.next_id);
250 let idx = self.map.insert_delim(
252 open_abs_range - self.global_offset,
253 open_abs_range - self.global_offset,
258 fn close_delim(&mut self, idx: usize, close_abs_range: Option<TextRange>) {
259 match close_abs_range {
261 self.map.remove_delim(idx);
264 self.map.update_close_delim(idx, close - self.global_offset);
270 /// A Raw Token (straightly from lexer) convertor
271 struct RawConvertor<'a> {
274 id_alloc: TokenIdAlloc,
275 inner: std::slice::Iter<'a, RawToken>,
278 trait SrcToken: std::fmt::Debug {
279 fn kind(&self) -> SyntaxKind;
281 fn to_char(&self) -> Option<char>;
283 fn to_text(&self) -> SmolStr;
286 trait TokenConvertor {
287 type Token: SrcToken;
289 fn go(&mut self) -> Option<tt::Subtree> {
290 let mut subtree = tt::Subtree::default();
291 subtree.delimiter = None;
292 while self.peek().is_some() {
293 self.collect_leaf(&mut subtree.token_trees);
295 if subtree.token_trees.is_empty() {
298 if subtree.token_trees.len() == 1 {
299 if let tt::TokenTree::Subtree(first) = &subtree.token_trees[0] {
300 return Some(first.clone());
306 fn collect_leaf(&mut self, result: &mut Vec<tt::TokenTree>) {
307 let (token, range) = match self.bump() {
312 let k: SyntaxKind = token.kind();
314 if let Some(tokens) = self.convert_doc_comment(&token) {
315 result.extend(tokens);
320 result.push(if k.is_punct() {
321 assert_eq!(range.len(), TextSize::of('.'));
322 let delim = match k {
323 T!['('] => Some((tt::DelimiterKind::Parenthesis, T![')'])),
324 T!['{'] => Some((tt::DelimiterKind::Brace, T!['}'])),
325 T!['['] => Some((tt::DelimiterKind::Bracket, T![']'])),
329 if let Some((kind, closed)) = delim {
330 let mut subtree = tt::Subtree::default();
331 let (id, idx) = self.id_alloc().open_delim(range);
332 subtree.delimiter = Some(tt::Delimiter { kind, id });
334 while self.peek().map(|it| it.kind() != closed).unwrap_or(false) {
335 self.collect_leaf(&mut subtree.token_trees);
337 let last_range = match self.bump() {
339 // For error resilience, we insert an char punct for the opening delim here
340 self.id_alloc().close_delim(idx, None);
341 let leaf: tt::Leaf = tt::Punct {
342 id: self.id_alloc().alloc(range),
343 char: token.to_char().unwrap(),
344 spacing: tt::Spacing::Alone,
347 result.push(leaf.into());
348 result.extend(subtree.token_trees);
353 self.id_alloc().close_delim(idx, Some(last_range));
356 let spacing = match self.peek() {
358 if next.kind().is_trivia()
359 || next.kind() == T!['[']
360 || next.kind() == T!['{']
361 || next.kind() == T!['('] =>
365 Some(next) if next.kind().is_punct() => tt::Spacing::Joint,
366 _ => tt::Spacing::Alone,
368 let char = match token.to_char() {
371 panic!("Token from lexer must be single char: token = {:#?}", token);
374 tt::Leaf::from(tt::Punct { char, spacing, id: self.id_alloc().alloc(range) }).into()
377 macro_rules! make_leaf {
379 tt::$i { id: self.id_alloc().alloc(range), text: token.to_text() }.into()
382 let leaf: tt::Leaf = match k {
383 T![true] | T![false] => make_leaf!(Ident),
384 IDENT => make_leaf!(Ident),
385 k if k.is_keyword() => make_leaf!(Ident),
386 k if k.is_literal() => make_leaf!(Literal),
388 let char_unit = TextSize::of('\'');
389 let r = TextRange::at(range.start(), char_unit);
390 let apostrophe = tt::Leaf::from(tt::Punct {
392 spacing: tt::Spacing::Joint,
393 id: self.id_alloc().alloc(r),
395 result.push(apostrophe.into());
397 let r = TextRange::at(range.start() + char_unit, range.len() - char_unit);
398 let ident = tt::Leaf::from(tt::Ident {
399 text: SmolStr::new(&token.to_text()[1..]),
400 id: self.id_alloc().alloc(r),
402 result.push(ident.into());
412 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>>;
414 fn bump(&mut self) -> Option<(Self::Token, TextRange)>;
416 fn peek(&self) -> Option<Self::Token>;
418 fn id_alloc(&mut self) -> &mut TokenIdAlloc;
421 impl<'a> SrcToken for (&'a RawToken, &'a str) {
422 fn kind(&self) -> SyntaxKind {
426 fn to_char(&self) -> Option<char> {
427 self.1.chars().next()
430 fn to_text(&self) -> SmolStr {
435 impl RawConvertor<'_> {}
437 impl<'a> TokenConvertor for RawConvertor<'a> {
438 type Token = (&'a RawToken, &'a str);
440 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
441 convert_doc_comment(&doc_comment(token.1))
444 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
445 let token = self.inner.next()?;
446 let range = TextRange::at(self.offset, token.len);
447 self.offset += token.len;
449 Some(((token, &self.text[range]), range))
452 fn peek(&self) -> Option<Self::Token> {
453 let token = self.inner.as_slice().get(0);
456 let range = TextRange::at(self.offset, it.len);
457 (it, &self.text[range])
461 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
467 id_alloc: TokenIdAlloc,
468 current: Option<SyntaxToken>,
470 punct_offset: Option<(SyntaxToken, TextSize)>,
474 fn new(node: &SyntaxNode, global_offset: TextSize) -> Convertor {
476 id_alloc: { TokenIdAlloc { map: TokenMap::default(), global_offset, next_id: 0 } },
477 current: node.first_token(),
478 range: node.text_range(),
486 Ordinary(SyntaxToken),
487 Punch(SyntaxToken, TextSize),
491 fn token(&self) -> &SyntaxToken {
493 SynToken::Ordinary(it) => it,
494 SynToken::Punch(it, _) => it,
499 impl SrcToken for SynToken {
500 fn kind(&self) -> SyntaxKind {
503 fn to_char(&self) -> Option<char> {
505 SynToken::Ordinary(_) => None,
506 SynToken::Punch(it, i) => it.text().chars().nth((*i).into()),
509 fn to_text(&self) -> SmolStr {
510 self.token().text().clone()
514 impl TokenConvertor for Convertor {
515 type Token = SynToken;
516 fn convert_doc_comment(&self, token: &Self::Token) -> Option<Vec<tt::TokenTree>> {
517 convert_doc_comment(token.token())
520 fn bump(&mut self) -> Option<(Self::Token, TextRange)> {
521 if let Some((punct, offset)) = self.punct_offset.clone() {
522 if usize::from(offset) + 1 < punct.text().len() {
523 let offset = offset + TextSize::of('.');
524 let range = punct.text_range();
525 self.punct_offset = Some((punct.clone(), offset));
526 let range = TextRange::at(range.start() + offset, TextSize::of('.'));
527 return Some((SynToken::Punch(punct, offset), range));
531 let curr = self.current.clone()?;
532 if !&self.range.contains_range(curr.text_range()) {
535 self.current = curr.next_token();
537 let token = if curr.kind().is_punct() {
538 let range = curr.text_range();
539 let range = TextRange::at(range.start(), TextSize::of('.'));
540 self.punct_offset = Some((curr.clone(), 0.into()));
541 (SynToken::Punch(curr, 0.into()), range)
543 self.punct_offset = None;
544 let range = curr.text_range();
545 (SynToken::Ordinary(curr), range)
551 fn peek(&self) -> Option<Self::Token> {
552 if let Some((punct, mut offset)) = self.punct_offset.clone() {
553 offset = offset + TextSize::of('.');
554 if usize::from(offset) < punct.text().len() {
555 return Some(SynToken::Punch(punct, offset));
559 let curr = self.current.clone()?;
560 if !self.range.contains_range(curr.text_range()) {
564 let token = if curr.kind().is_punct() {
565 SynToken::Punch(curr, 0.into())
567 SynToken::Ordinary(curr)
572 fn id_alloc(&mut self) -> &mut TokenIdAlloc {
577 struct TtTreeSink<'a> {
580 open_delims: FxHashMap<tt::TokenId, TextSize>,
582 inner: SyntaxTreeBuilder,
586 // Use for detect ill-form tree which is not single root
587 roots: smallvec::SmallVec<[usize; 1]>,
590 impl<'a> TtTreeSink<'a> {
591 fn new(cursor: Cursor<'a>) -> Self {
595 open_delims: FxHashMap::default(),
597 inner: SyntaxTreeBuilder::default(),
598 roots: smallvec::SmallVec::new(),
599 token_map: TokenMap::default(),
603 fn finish(mut self) -> (Parse<SyntaxNode>, TokenMap) {
604 self.token_map.entries.shrink_to_fit();
605 (self.inner.finish(), self.token_map)
609 fn delim_to_str(d: Option<tt::DelimiterKind>, closing: bool) -> &'static str {
610 let texts = match d {
611 Some(tt::DelimiterKind::Parenthesis) => "()",
612 Some(tt::DelimiterKind::Brace) => "{}",
613 Some(tt::DelimiterKind::Bracket) => "[]",
617 let idx = closing as usize;
618 &texts[idx..texts.len() - (1 - idx)]
621 impl<'a> TreeSink for TtTreeSink<'a> {
622 fn token(&mut self, kind: SyntaxKind, mut n_tokens: u8) {
623 if kind == L_DOLLAR || kind == R_DOLLAR {
624 self.cursor = self.cursor.bump_subtree();
627 if kind == LIFETIME_IDENT {
631 let mut last = self.cursor;
632 for _ in 0..n_tokens {
633 let tmp_str: SmolStr;
634 if self.cursor.eof() {
638 let text: &str = match self.cursor.token_tree() {
639 Some(tt::buffer::TokenTreeRef::Leaf(leaf, _)) => {
640 // Mark the range if needed
641 let (text, id) = match leaf {
642 tt::Leaf::Ident(ident) => (&ident.text, ident.id),
643 tt::Leaf::Punct(punct) => {
644 assert!(punct.char.is_ascii());
645 let char = &(punct.char as u8);
646 tmp_str = SmolStr::new_inline(
647 std::str::from_utf8(std::slice::from_ref(char)).unwrap(),
651 tt::Leaf::Literal(lit) => (&lit.text, lit.id),
653 let range = TextRange::at(self.text_pos, TextSize::of(text.as_str()));
654 self.token_map.insert(id, range);
655 self.cursor = self.cursor.bump();
658 Some(tt::buffer::TokenTreeRef::Subtree(subtree, _)) => {
659 self.cursor = self.cursor.subtree().unwrap();
660 if let Some(id) = subtree.delimiter.map(|it| it.id) {
661 self.open_delims.insert(id, self.text_pos);
663 delim_to_str(subtree.delimiter_kind(), false)
666 if let Some(parent) = self.cursor.end() {
667 self.cursor = self.cursor.bump();
668 if let Some(id) = parent.delimiter.map(|it| it.id) {
669 if let Some(open_delim) = self.open_delims.get(&id) {
670 let open_range = TextRange::at(*open_delim, TextSize::of('('));
671 let close_range = TextRange::at(self.text_pos, TextSize::of('('));
672 self.token_map.insert_delim(id, open_range, close_range);
675 delim_to_str(parent.delimiter_kind(), true)
682 self.text_pos += TextSize::of(text);
685 let text = SmolStr::new(self.buf.as_str());
687 self.inner.token(kind, text);
689 // Add whitespace between adjoint puncts
690 let next = last.bump();
692 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(curr), _)),
693 Some(tt::buffer::TokenTreeRef::Leaf(tt::Leaf::Punct(_), _)),
694 ) = (last.token_tree(), next.token_tree())
696 // Note: We always assume the semi-colon would be the last token in
697 // other parts of RA such that we don't add whitespace here.
698 if curr.spacing == tt::Spacing::Alone && curr.char != ';' {
699 self.inner.token(WHITESPACE, " ".into());
700 self.text_pos += TextSize::of(' ');
705 fn start_node(&mut self, kind: SyntaxKind) {
706 self.inner.start_node(kind);
708 match self.roots.last_mut() {
709 None | Some(0) => self.roots.push(1),
710 Some(ref mut n) => **n += 1,
714 fn finish_node(&mut self) {
715 self.inner.finish_node();
716 *self.roots.last_mut().unwrap() -= 1;
719 fn error(&mut self, error: ParseError) {
720 self.inner.error(error, self.text_pos)
727 use crate::tests::parse_macro;
728 use parser::TokenSource;
730 algo::{insert_children, InsertPosition},
735 fn convert_tt_token_source() {
736 let expansion = parse_macro(
738 macro_rules! literals {
750 .expand_tt("literals!(foo);");
751 let tts = &[expansion.into()];
752 let buffer = tt::buffer::TokenBuffer::from_tokens(tts);
753 let mut tt_src = SubtreeTokenSource::new(&buffer);
754 let mut tokens = vec![];
755 while tt_src.current().kind != EOF {
756 tokens.push((tt_src.current().kind, tt_src.text()));
761 // [let] [a] [=] ['c'] [;]
762 assert_eq!(tokens[2 + 3].1, "'c'");
763 assert_eq!(tokens[2 + 3].0, CHAR);
764 // [let] [c] [=] [1000] [;]
765 assert_eq!(tokens[2 + 5 + 3].1, "1000");
766 assert_eq!(tokens[2 + 5 + 3].0, INT_NUMBER);
767 // [let] [f] [=] [12E+99_f64] [;]
768 assert_eq!(tokens[2 + 10 + 3].1, "12E+99_f64");
769 assert_eq!(tokens[2 + 10 + 3].0, FLOAT_NUMBER);
771 // [let] [s] [=] ["rust1"] [;]
772 assert_eq!(tokens[2 + 15 + 3].1, "\"rust1\"");
773 assert_eq!(tokens[2 + 15 + 3].0, STRING);
777 fn stmts_token_trees_to_expr_is_err() {
778 let expansion = parse_macro(
790 .expand_tt("stmts!();");
791 assert!(token_tree_to_syntax_node(&expansion, FragmentKind::Expr).is_err());
795 fn test_token_tree_last_child_is_white_space() {
796 let source_file = ast::SourceFile::parse("f!({} );").ok().unwrap();
797 let macro_call = source_file.syntax().descendants().find_map(ast::MacroCall::cast).unwrap();
798 let token_tree = macro_call.token_tree().unwrap();
800 // Token Tree now is :
810 token_tree.syntax().descendants_with_tokens().find(|it| it.kind() == T!['}']).unwrap();
811 let space = token_tree
813 .descendants_with_tokens()
814 .find(|it| it.kind() == SyntaxKind::WHITESPACE)
817 // reorder th white space, such that the white is inside the inner token-tree.
818 let token_tree = insert_children(
819 &rbrace.parent().unwrap(),
820 InsertPosition::Last,
821 std::iter::once(space),
824 // Token Tree now is :
829 let token_tree = ast::TokenTree::cast(token_tree).unwrap();
830 let tt = ast_to_token_tree(&token_tree).unwrap().0;
832 assert_eq!(tt.delimiter_kind(), Some(tt::DelimiterKind::Brace));
836 fn test_token_tree_multi_char_punct() {
837 let source_file = ast::SourceFile::parse("struct Foo { a: x::Y }").ok().unwrap();
838 let struct_def = source_file.syntax().descendants().find_map(ast::Struct::cast).unwrap();
839 let tt = ast_to_token_tree(&struct_def).unwrap().0;
840 token_tree_to_syntax_node(&tt, FragmentKind::Item).unwrap();